The impact of board interlocks on earnings management

1

Amsterdam Business School

MSc Thesis

The Impact of Board Interlocks on

Earnings Management

Name: Pasquinel de Roode Student number: 11120576

Thesis supervisor: Dr. A.K. Sikalidis Date: 15th August 2016

Word count: 13,617

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

2

Statement of Originality

This document is written by student Pasquinel de Roode who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3 Abstract

This study examines the relationship between board interlocks and earnings management. Board interlocks occur when members of the board of directors are also active in other boards and vice versa. Board interlocks create social networks, because they act as communication channels, and directors from the connected firms get to know each other. Based on prior research, this can lead to collusion, improper corporate control, cohesion, coordinated action, and unified political-economic power of corporate executives. Therefore, supported by the familiarity threat, it was hypothesized that board interlocked firms are more involved in earnings management than firms without interlocked boards. Since earnings management can be divided in accrual-based earnings management and real activities manipulation, the main hypothesis is also divided in multiple hypotheses. Therefore, it was hypothesized that board interlocked firms exhibit more discretionary accruals, lower levels of abnormal cash flow from operations, higher levels of abnormal production costs and lower levels of abnormal discretionary expenses than firms without interlocked boards. All hypotheses are not supported by the empirical findings, based on the 507 firm-year observations in the United States between 1997 and 2001. The coefficients of the indicator variable Interlocking Directorship are insignificant in both earnings management regressions. Thus, there is no evidence that interlocking directorship is associated with earnings management and that board interlocks lead to more implementation of earnings management.

Key words:

Board interlocks, Interlocking directorship, Affiliated directors, Earnings management, Real activities manipulation, Abnormal discretionary accruals, Abnormal cash flow from operations, Abnormal production costs, Abnormal discretionary expenses, Social network.

Acknowledgement

I would like to thank Dr. Alexandros Sikalidis for his expert guidance, valuable feedback and constant support through the stages of my MSc thesis.

4 Table of Contents

1 Introduction ... 5

2 Literature overview ... 8

2.1 Introduction ... 8

2.2 Composition of the board and independence ... 8

2.3 Board interlocks and prior research ... 10

2.4 Influencing decision making behavior through social networks ... 11

2.5 Familiarity threat ... 12

2.6.1 Earnings Management definition ... 13

2.6.2 Measuring Earnings Management ... 14

2.7 Hypothesis Development ... 16

3 Research design ... 18

3.1 Introduction ... 18

3.2 Data and Sample selection ... 18

3.3 Operationalization of variables ... 20

3.4 Empirical Models ... 23

4 Results ... 25

4.1 Introduction ... 25

4.2 Descriptive Statistics and Univariate Analysis ... 25

4.3 The Relation between Board Interlocks and Accrual-Based Earnings Management 30 4.4 The Relation between Board Interlocks and Real Activities Manipulation ... 32

5 Conclusion and Discussion ... 34

6 References ... 37

5

1.Introduction

Board interlocks occur when the members of a corporate board of directors serve on the boards of multiple corporations and vice versa. For example, the current CEO of firm A serves as a director of firm B and the current CEO of firm B serves as a director of firm A. A multiple director is knows as a person that sits on multiple boards (Scott, 1997). Board interlocks can be a direct or direct interlock. A direct interlock between two firms is present if a director or executive of one firm is also a director of the other, while an indirect interlock occurs when a director of each board sits on the board of a third firm (Salinger, 2005)

Interlocks act as communication channels, because boards are able to share information with each other via multiple directors who have access to inside information for multiple companies (Scott, 1997). Board interlocks also have benefits over cartels, trusts, and other monopolistic/oligopolistic forms of organization, because they are more fluid and less visible which make them less open to public scrutiny. They also benefit the involved firms, because of increased information availability for directors, reduced competition, and increased prestige (Salinger, 2005).

There are theorists that believe that multiple directors are more likely to think in terms of general corporate class interests than simply the narrow interests of individual corporations, because a lot of directors often have interests in firms in different industries (Bowman, 1996; Beder, 2006; Barrow, 1993). Also, most of the time these individuals come from wealthy backgrounds, socialize with the upper classes, and have worked their way up the corporate hierarchy. This will make it more likely that they have internalized values that will cause them to personally support policies that are beneficial to business in general (Bowman, 1996). The importance of these connections on politics and economy was recognized by the Pujo Committee of the US House of Representatives early in the twentieth century. They identified board interlocks as a vehicle for improper corporate control and collusion, particularly by US banks. So the subsequent Clayton Act prohibited interlocks between competitors (Mizruchi & Marquis, 2006).

According to some observers, board interlocks can lead to cohesion, coordinated action, and unified political-economic power of corporate executives (Asimakopoulos, 2009). As an effect, the influence of corporations will increase by exerting power as a group and working together towards common goals (Salinger, 2005). Corporate executives in board

6

interlocks maintain an advantage, and gain more power over workers and consumers, because intra-class competition reduces and cooperation increases (Mizruchi & Schwartz, 1992; Salinger, 2005).

This practice, even though widespread and lawful, creates questions about the independence and quality of the decisions of the board of directors. There are multiple ways how board interlocks could lead to earnings management. Firstly, nepotism can evolve between connected firms, which means that they can make an agreement with each other to take it easy with overseeing their management. In that way, it leads to an open door to earnings management. Secondly, interlocked directors are busy directors, because they are active in multiple boards. So, they would be less effective than directors who are active in only one board. This means that the use of earnings management would be more likely to apply. Board interlocks are social networks where the decision making behavior of directors can be influenced. They tend to follow the majority of the directors to whom they are connected. This kind of behavior is not perfectly rational in the sense of economics and is known as ”herd behavior” (Battiston, 2004). In that way, the following research question is formulated: What is the impact of board interlocks on the decision of the firm whether to use earnings management or not?

This paper contributes to the literature in several ways. Earnings management is still a societal problem, so it is relevant to investigate how to solve that problem or how to limit the use of earnings management. There is also not much prior research or literature about the relationship between board interlocks and earnings management. In sum, the findings of this research will contribute to the accounting and board governance literature by examining if board interlocks stimulate the use of earnings management within the firm. This study expands earnings management research into social network settings, since board interlocks are considered as a social network. This paper also contributes to the corporate governance literature, because it will show if monitoring each other’s board is affected by the social ties within the board interlocks. The findings will be of importance for regulators if board interlocks clearly lead to more earnings management. It will stimulate them to pay more attention to the connectivity of the company’s board to improve the financial reporting quality. So the outcome of the research question would be a big contribution to the existing knowledge. It will extend our understanding of the relation between board interlocks and earnings management.

The rest of the thesis is structured into four chapters and organized as follows. Chapter two presents the literature review consisting the description of the composition of the board, a

7

review of prior board interlock research and the definition of earnings management. It also discusses decision making behavior en presents the hypothesis development. Chapter three presents the research design in this study. It describes the used data extracted from two sources: Compustat and ISS (formerly RiskMetrics). The sample selection procedure is described as well as the criteria to construct the sample. The operationalization of the earnings management variables are divided into two categories (discretionary accruals and real activities manipulation. For each category, the construction and measurement of the variables are explained. At last, the chapter presents the empirical models with the description of all the variables including the control variables. Chapter four comprises three main parts. The first part of the chapter shows the descriptive statistics and univariate analysis of the independent, dependent and control variables. The second part of the chapter shows the multiple regression of accrual-based earnings management, while the third part presents the multiple regression of real activities manipulation. Finally, chapter five presents the conclusion and discussion of the thesis, which focuses on the findings, research limitation, implications and potential opportunities for further studies and improvements.

8

2. Literature overview 2.1 Introduction

This chapter reviews the literature on board interlocks, social networks within firms, and earnings management in order to identify the research question and research gap that will be used to guide this study.

A large body of research that examines corporate governance and board governance in general is present, both theoretical and empirical, but research on the role of board interlocks in governance more specifically is limited. Since the effectiveness of the board is determined by the structure of the board of directors, this chapter will first discuss the composition of the board and its independence. After that, prior research about board interlocks will be presented. Drawing on psychology theories, the effectiveness of the board’s decision making can be influenced by social ties within the firm. Since board interlocks create social networks between the connected firms, prior literature about the relationship between social ties and firm performance, and the familiarity threat will be reviewed. Finally, the chapter will provide various definitions of earnings management and insight into real activities manipulation. Altogether, this will lead to the hypothesis development that will be presented at the end of the chapter.

2.2 Composition of the board and independence

In general, there are three types of directors that are active and can be classified in the board of directors: inside directors, outside directors and affiliated directors. The insiders are known as the employees like CEO and other officers. This means that the insiders are both managers and directors. Outside directors are the independent directors who are not connected with the firm in another way than being a director. Affiliated directors are connected with the firm in some way. That can be through multiple ways like business dealings, such as consumers, suppliers, employees of the connected firms and the audit firm, former employees, lawyers, investment bankers, consultants, executives of advertising agencies and directors who are connected to charitable organizations because of the contributions they make to them. The affiliated directors classification can be applied to interlocked directors, because they sit on each other’s board. This means that director of company A would be employed by company B and a director of company B would be employed by company A. According to rule 34-50298 of the Security Exchange Commission (SEC), interlocked directors are considered to be dependent if they sat on the compensation committee of each other in the past three years (Ronen, J. & Yaari, V., 2008, p. 248).

9

Each type of directors adopts a different role. Outside directors are associated with monitoring, because of their independence. The SEC comments that board of directors that make use of independent director’s judgment to carry out their responsibilities, are mostly effective. Inside directors are considered to be opposite of the outside directors, mainly because they would be an obstacle to efficient monitoring. It’s common that in one directorship, monitoring performance is almost impossible to implement, because the head of the division who is included in a committee can hamper the free communication between other independent committee members (Blair, 1995; Salmon, 1993).

Inside directors are employed by the firm, which means that they have more specific information about the firm than the outside directors. So, they are better informed about the operations of the firm. Outside directors are likely to have little specific knowledge about the firm’s operations, because they arrive at their firm with little direct knowledge and just spend a couple days per year at meetings (Yermack, 2004). Inside directors are thus better able to predict the reactions of a decision on the firm because of its unique business area. This is supported by the findings of Klein (1998) that increase in inside directors on finance and investments committees leads to a higher rate of return on the investments earnings for the firm compared to a decrease in inside directors. The research of Bathala and Rao (1995) also supports the claim that inside directors are better able to predict the reactions of a decision on the firm. They found that independence and growth opportunities were negatively related. But with regard to the value of inside directors, there hides a crucial issue that concerns their loyalty and commitment. Insiders can be brought to the board to be prepared for being the next CEO of the firm (Vancil, 1987). This can affect the behavior of the inside director by associating with outside directors to increase their chance to be selected as the next CEO. Besides that, they are also incentivized to conspire with the management, because their careers are controlled by the CEO. In contrast, outside directors are not affected by such career concerns, especially when they have the ability to join other alternative boards (Beatty & Zajac, 1994). For example, when insiders are in the compensation committee, a situation can occur in which the committee can be under pressure to award management options with the same vesting dates for all managers. This would lead as an effect to incentives to conspire to mange earnings by reducing the strike price and increasing the exercise price. Prior research of Boumosleh and Reeb (2005) confirms the reliability of the example. They found that the difference between the CEO’s pay and the other executives’ pay is lower in firms with a high percentage of inside directors than in firms with a low percentage of inside directors. This indicates that inside directors and the CEO collude to jointly earn more.

10

Affiliated directors are a hybrid, because they have characteristics of insiders and outsiders. These directors know that the managers should be monitored by them, but they are less likely to monitor the managers than the independent directors. The reason for that is that the affiliated directors don’t want to damage their rapport with the management. This can be strengthened by the evidence from prior literature that the board has more affiliated directors and fewer outside directors when the CEO has more legal power over the nomination of directors (Shivdasani, A. & Yermack, D., 1999).

2.3 Board interlocks and prior research

Since the Modern Corporation and Private Property was published in 1932, interlocking directorates or multiple directorships have been studied several times. A board interlock is present when a person serves on the board of more than one company and thereby creating a link or interlock between the companies (Pombo & Gutierrez, 2011). According to Mizruchi (1996), interlocks have been analyzed as monitoring mechanisms, as mechanisms of cooptation, as mechanisms of collusion, and as a reflection of social cohesion. However, ‘’if interlocks are to be worthy studying, it is essential that they be shown to have consequences for the behavior of firms’’ (p. 280).

The first empirical papers about interlocking directorship found that the 250 largest companies in 1965 had the most interlocks (Dooley, 1969). Management-controlled companies were trying to avoid interlocks, while non-financial companies tended to have more interlocks relatively, and there were proportionally more interlocks among competing companies. Dooley (1969) found that interlocks between financial and non-financial companies increased when the non-financial companies became less solvent and the assets of it increased. Another finding was that interlocks seemed more frequent among companies within an industry, which was consistent with the view that board interlocks are mechanisms to decrease competitive uncertainty (Dooley, 1969).

Most recent studies about interlocking directorship and its effect on performance have been done markets that are developed like those in Europe, United States and Australia (Pombo & Gutierrez, 2011). For instance, Ferris and Jagannathan (2001) concluded that multiple directorships are not a pervasive phenomenon, because only 13 percent of the directors sample were active in more than one board. They found that firms with a great number of board seats, larger firms, and more profitable firms were more likely to extend or have directors that sit on multiple boards (Ferris and Jagannathan, 2001).

11

performance. An outside director was defined as busy if they were active on three or more boards. They found that both the busy outside directors and busy boards were negatively and statistically significant related to the market-to-book ratio (Pombo & Gutierrez, 2011).

Fich and White (2005) studied reciprocal CEO interlocks which means that the CEO of one firm sits on the board of a second firm and the CEO of the second firm sits on the first firm’s board. They found that the CEO’s presence on the nominating committee has an influence on the board composition. The main conclusions is that a reciprocal CEO interlock is more likely to be a mechanism to enhance a CEO’s private interests and less likely to be a corporate governance tool to enhance the interests of the firm’s shareholders.

Boards have the important task to monitor management (Pombo & Gutierrez, 2011). Devos et al. (2009) investigated whether interlocked directors and connected boards can be related to weak governance. They found that interlocks with CEO outsiders occurred the most and concluded that the interlocks do not appear to significantly lower firm value.

However, Fich and White (2005) found that firm performance was impaired when both chairman and other directors hold seats on boards of non-listed companies. This means that busy boards do not perform their monitoring task. Besides that, Santos et al. (2009) investigated board interlocks across 320 Brazilian listed firms and found that interlocking is a common practice across firms and that busy boards have a negative influence on firm valuation.

2.4 Influencing decision making behavior through social networks

There are studies that focus on networks within firms (Hwang and Kim, 2010; Fracassi and Tate, 2009). For example, Hwan and Kim (2010) studied the social ties between the CEO and audit committee board members. They found that increasing social ties within the firm stimulates the use of earnings management and leads to higher CEO bonuses. Another example, Fracassi and Tate (2009) studied the social ties between the CEO and the directors of the firm and found that greater social ties lead to fewer voluntary restatements and also more value destroying acquisitions. They concluded with these results that social ties within boards decrease the effectiveness of board monitoring. This research will focus on the social ties between firms through interlocked directors and not within the firms.

The psychology literature also mentions that unethical behaviors can be spread through social ties (Ariel, Ayal and Gino, 2009). The study of Sah (1991) showed that managers can be influenced by the exposure of dishonesty by others. It can lead to managers changing their subjective estimation of manipulations benefits and costs. Other social

12

psychology literature stated that individuals who belong to a group have the tendency to conform to the social norms and values even when they are clearly incorrect (Asch, 1951). Fich and Shivdasani (2007) showed that firms with a board member who also sits on the board of another firm that has been sued for fraud in the past have a bigger chance of facing a financial lawsuit. This is consistent with the study of Bizjak, Lemmon and Whitby (2009) finding the evidence that firms with board interlocks have a higher frequency of stock option backdating.

The observation of actions and preference interactions, and direct communication of information signals through social networks via board interlocks can lead to opportunistic behaviors like earnings management than can spread from one firm to another. It is unlikely that firms and its directors publicize earnings management widely for obvious reasons. However, there is the possibility that the opportunistic behavior from one firm to another diffuse quietly through individuals conversations between directors from multiple boards. But the subjective perceived benefits and costs of such opportunistic behavior determine whether a firm manages its earnings. This means that directors that are linked to a firm that manipulates its earnings, are more likely to estimate a higher perceived benefit than costs of earnings management. As a result, herd behavior can occur (Chiu, Teoh and Tian, 2012).

Furthermore, the preferences of executives and directors can be affected when their boards are linked to firms that apply earnings management because it can change the view of them whether earnings management is a social norm or not. They can view the use of earnings management at other firms as a moral justification to also apply it.

2.5 Familiarity threat

The familiarity threat is ‘’the threat that due to a long or close relationship with a client or employer, a professional accountant will be too sympathetic to their interest or too accepting of their work’’ (IESBA, 2012).

In July 2003, International Federation of Accountants (IFAC) published a report, called Rebuilding Public Confidence in Financial Reporting, which described familiarity as helpful in an audit process to create a bigger understanding and improved ability to identify and evaluate risks. However, IFAC also issued familiarity as one of the significant threats to auditor independence. IFAC’s main concern is that excessive familiarity can lead to auditor’s hesitancy to challenge appropriately and therefore can reduce the level of skepticism that is necessary for an audit (Chi and Huang, 2005).

13

on the behavior in decision making in a negative way. The reasoning for this can also be applied to the influence of board interlocks on the use of earnings management. Board interlocks create social networks, which increase the familiarity between directors from the connected companies. As a consequence, excessive familiarity between the directors could lead to a reduced level of skepticism during the monitoring process.

2.6.1 Earnings Management definition

Earnings management is not a concept with a standard or universal accepted definition. Therefore, multiple definitions of earnings management from different authors will be given to explain the concept as clear as possible. Three definitions will be given and will be shortly discussed.

Schipper (1989: p. 92) was one of the first who gave earnings management a definition. He defined it as: ‘’Earnings management is really disclosure management in the sense of a purposeful intervention in the external financial process, with the intent of obtaining some private gain as opposed to merely facilitating the neutral operation of the process’’. The definition of Schipper considers earnings management as the intentional interference of the manager in the external financial reporting to achieve the objectives of the firm and management. The behavior in which the manager pursues his own objectives first, is called opportunistic behavior. The opposite behavior, pursuing organizational objectives first, occurs by decreasing the contract costs of the firm.

Healy and Wahlen (1999: p. 361) describe earnings management as following: ‘’Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to either mislead some stakeholders about the underlying economic performance of the company or in influence contractual outcomes that depend on reported accounting numbers’’. The manager can form a judgment about the financial reporting in different ways. Firstly, the manager need to make some estimates about the depreciation period, the residual value and the provision for doubtful receivables. Secondly, the manager has to choose some accounting methods. These can influence the earnings on the short-term, but don’t have an effect on the long-term earnings. Thirdly, the manager need to form a policy for the working capital, receivables period, payables period and the amount of inventory. Finally, the manager should form a judgment about the way of financing the firm.

Scott (2003: p. 369) defines earnings management as following: ‘’Earnings management is the choice by a manager of accounting policies so as to achieve some specific objective’’. This definition assumes that earnings management has a good and bad side, but

14

does not indicate which specific objectives are pursued. But later he argued that earnings management can be used by the management to share private information on the future prospects of the company, which is beneficial to investors (Scott, 2011).

Out of all the given definitions, it can be concluded that applying earnings management will lead to a distorted picture of the earnings and a infestation of the reliability. Managers can cause harmful practices if they get stimulated to apply earnings management. The composition of the board and the independence of the directors can have an influence on the decision making of the management whether to use earnings management or not. This is discussed in the previous paragraph.

2.6.2 Measuring earnings management

Earnings management is an unobservable connected process, which means that behavior as if earnings are managed is the only thing that can be used for the empirical analysis (Van Praag, 2001; 49). The operational criteria for inferring earnings management needs to be specified by examining the afforded latitude of the accrual based accounting system. For making the measurement possible, proxies are needed. There are two forms of proxies for discretionary accounting: the discretionary accruals model and the real activities manipulation methods.

Real Activities Manipulation

Real activities manipulation is defined by Roychowdhury (2006, p. 337) 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.” Under certain economic circumstances, some methods such as sales promotions using discounted price and/or a discretionary expenditures reduction are optimal for manipulating real activities. However, when managers engage in these activities more extensively than is normal, because of the given economic circumstances and the objective to meet or beat an earnings benchmark, they are manipulating real activities according to Roychowdhury’s (2006) definition.

Graham, Harvey and Rajgopal’s survey finds that (a) financial executives consider meeting earnings targets such as zero earnings, analyst forecasts and previous period’s earnings as highly important, and (b) they are motivated to manipulate real activities to meet these earnings targets, even when the manipulation can lead to potentially firm value reduction (Roychowdhury, 2006). Real activities manipulation can reduce firm value in a way that current actions to increase earnings can lead to a negative effect on cash flows in the

15

future. An example of firm value reduction due to real activities manipulation is the aggressive price discounts to increase sales volumes and meet short term earnings target. This can lead to the expectations of customers that such discounts will exist in future periods as well. As a consequence, it can imply lower margins on future sales. Besides that, overproduction generates excessive inventories that have to be sold in subsequent periods and leads to greater inventory holding costs for the company (Roychowdhury, 2006).

Most of the prior research studying real activities management focuses on the opportunistic reduction on reported expenses by decreasing expenditures on R&D. Dechow and Sloan (1991) discovered that CEO’s increased short term earnings by reducing R&D spending when they approached the end of their tenure. Similar evidence is provided by Baber et al. (1991) and Bushee (1998), who showed that reducing R&D spending helps hitting the earnings benchmark.

Lots of other evidence revealed that managers have a lot of other activities to choose other than reduction of R&D to adopt for real activities manipulation. Bens, Nagar and Wong (2002) and Bens, Nagar, Skinner and Wong (2003) report that managers repurchase stock because of the dilution of earnings per share. That dilution can be caused by employee stock option exercises and employee stock option grants. Bens, Nagar and Wong (2002) find evidence that these repurchases are partially financed by reducing R&D. In addition, in Graham, Harvey and Rajgopal’s (2005) survey, many survey respondents admitted that they used methods like reducing discretionary expenditures and/or capital investments more often than other manipulation methods. Bartov (1993) shows that firms will report higher profit from sales when there are negative earnings changes. However, Roychowdhury (2006) showed that there are other real activities management instead of just reductions in R&D. He found that managers used real activities manipulation to avoid reporting annual losses. These activities are reduction of discretionary expenditures to improve reported margins, price discounts to temporarily increase sales and overproduction to report lower cost of goods sold.

16 2.7 Hypothesis development

Multiple hypotheses will be developed based on the information in the literature overview. Board interlocks occur when companies are linked by shared board directors. On the one hand, this can have benefits since these directors come from wealthy backgrounds and have worked their way up the corporate hierarchy. This means that they have probably internalized value that will lead them to personally support policies that are beneficial to business in general. So good decision making will be achieved with board interlocks.

On the other hand, prior literature (Shivdasani, A. & Yermack, D., 1999) stated that affiliated directors are less likely to monitor the managers than the independent directors, because they don’t want to damage their rapport of the management. Besides that, there are also studies that state that social ties with the firm stimulates the use of earnings management, higher CEO bonuses and lower effectiveness of board monitoring (Hwang and Kim, 2010; Fracassi and Tate, 2009). Prior literature also found that herd behavior can occur through social networks via board interlocks where directors adopt the norms and values from the board members of the linked company no matter the correctness of these norms and values. Finally, the familiarity threat occurs when directors get to know each other through board interlocks. Therefore, the following hypothesis is developed:

Excessive-Familiarity Effect Hypothesis: Board interlocked firms are more involved in earnings management than firms without interlocked boards.

In this hypothesis, board interlocks can be considered as the independent variable and earnings management can be considered as the dependent variable. However, earnings management is not directly observable, so the literature overview mentioned some methods to measure earnings management. Since discretionary accruals and real activities manipulation will be used to measure earnings management, the hypothesis is divided in two hypotheses:

H1A: Board interlocked firms exhibit more discretionary accruals than firms without interlocked boards.

H2: Board interlocked firms exhibit more real activities manipulation than firms without interlocked boards.

To investigate real activities manipulation, patterns in cash flow from operations (CFO), discretionary expenses and production costs will be examined for firms with board interlocks

17

and without board interlocks. The same model as Roychowdhury’s (2006) model will be used to derive the normal levels of CFO, discretionary expenses and production costs for each firm-year. The deviations from the normal levels of the three variables will be interpreted as abnormal, like abnormal discretionary expenses etc. Therefore, H2 will be further divided into three hypotheses.

Low abnormal cash flow from operations often represent unexpectedly poor performance, which tends to put pressure on firms to apply real activities manipulation like price discounts to temporarily increase sales. Therefore, the following hypothesis is:

H2A: Board interlocked firms exhibit lower levels of abnormal cash flow from operations than firms without interlocked boards.

Production costs are defined as the sum of cost of goods sold (COGS) and the inventory change during the period. Managers can use real activities manipulation to avoid reporting annual losses. These activities can be overproduction to report lower cost of goods sold. Overproduction generates excessive inventories that have to be sold in subsequent periods and eventually leads to greater inventory holding costs for the company. Therefore the following hypothesis is developed:

H2B: Board interlocked firms exhibit higher levels of abnormal production costs than firms without interlocked boards.

Discretionary expenses are taken as the sum of advertising expenses, R&D expenses, and selling, general, an administration expenses (SG&A). Most of the prior research studying real activities management focuses on the opportunistic reduction on reported expenses by decreasing expenditures on R&D. Firms can reduce R&D spending to hit the earnings benchmark and increase short term earnings. Therefore, the final hypothesis is developed:

` H2C: Board interlocked firms exhibit lower levels of abnormal discretionary expenses than firms without interlocked boards.

18

3. Research design

3.1 Introduction

To test whether board interlocks have an impact on the use of earnings management by the firm, a quantitative research will be implemented. To obtain all required data, different databases will be used. The research starts with developing one sample that consists of random firms with interlocked boards and random firms without interlocked boards. The next step is to examine whether firms with interlocked boards are more likely to apply earnings management than firms without interlocked boards. Therefore, different ways of measurement and multiple sources will be used to make the result more reliable. This chapter presents the data and sample selection with the adopted criteria that was needed for the sample selection procedure. It also shows the operationalization of the variables consisting the measurement models for the earnings management proxies. At last, this chapter will present the empirical models used for the regression analysis to determine the relationship between board interlocks and earnings management.

3.2 Data and Sample selection

Boards of directors of listed firms have the duty to monitor and supervise the operations of the firms, and approve management decisions that are considered as important. Directors in the United States generally sit on more than one board and every board has multiple meetings a year. These board interlocks have as a consequence that board members get to know each other and form a social network. In this social network, board members can carry their knowledge and practices from one firm to another regardless of the location and the correctness of these practices. Because of the fact that board interlocks occur a lot in the United stated, this research will focus on the listed firms in North America. So the sample shall consist of publicly traded companies in North America.

The American board interlock network showed a stable cohesiveness that was characterized by the existence of an ‘inner circle’ of directors who are well-connected and the short path of lengths between companies (Chu and Davis, 2012). This stable cohesiveness of board interlocks was the result of a elite social cohesion and the key mechanism for maintaining this elite cohesion. However, Chu and Davis (2012) showed that during the first decade of the 21st century, this stable cohesiveness crumbled. They found that the average path of lengths between companies increased and that the numbers of directors on six or more boards dropped from 17 to zero. They also found that in 1997, minority directors were 1,52

19

times more likely to gain an additional board seat for the next year than non-minority directors, while in the beginning of the 21st _{century there was no difference anymore. So they} showed that the American board interlock network has changed in fundamental ways around the transition from the 20th to 21st century. Based on this information, this research will focus on the period between 1997 and 2001, because more board interlocked firms with the same characteristics are expected in that period.

To detect whether firms deal with board interlocks, information about directors from ISS (formerly RiskMetrics) database will be used. The Directors Legacy Data from RiskMetrics includes a range of variables that relate to individual board directors for example name, age, independence classification, company name they work for, interlocking directorship, other affiliation etcetera. The variable that will be used from that database is the independent variable ‘interlocking directorship’.

Beside that, data will be used from the database Compustat – Capital IQ to get financial statement information needed for the measurement of earnings management. The Compustat sample consists of the same firms from the ISS (formerly RiskMetrics) and includes balance sheet items, income statement items and cash flow items about the selected firms with board interlocks and firms without board interlocks. These items will be used to determine the discretionary line items that infer earnings management.

The initial sample covers all of the firm-year observations from 1997 to 2001: a total of 62,236 firm-years. Financial institutions are already excluded from the initial sample, because their specific asset and liability structure typically produces high financial leverage, which hinders the comparability with those of non-financial firms. Besides that, the estimation of discretionary accruals is not possible for these firms (Mohrmann, 2015). 59,427 observations are deleted, because of incomplete financial statement data and a further 10 observations with missing auditor data. This procedure should also eliminate 1,975 firm-year observations that couldn’t be matched with the ISS (formerly RiskMetrics) database, which means that these firm-year observations had missing interlocking directorship data. Finally, all the firm-year observations with missing data from previous firm-year are deleted, which is necessary for the calculation of numerous variables such as discretionary accruals. The resulting final sample consists of 507 firm-year observations. Table 1 summarizes the sample selection procedure.

20 TABLE 1 Sample Selection

Initial sample 62,236

Firm-year observations with incomplete financial statement data (59,427)

Firm-year observations with missing auditor data (10)

Firm-year observations with missing interlocking directorship data (1,975) Firm-year observations with missing data from previous firm-year (317)

Final sample 507

Note: This table summarizes the sample selection procedure. The initial sample covers all firm-years from COMPUSTAT from 1997-2000. ‘’Firm-year observations with missing interlocking directorship data’’ refers to firms that are not registered in the ISS(formerly RiskMetrics) Directors database.

3.3 Operationalization of variables Discretionary Accruals

Discretionary accruals are known as one of the most frequently used earnings characteristics. Measures of discretionary accruals are often used in numerous prior studies on earnings management and earnings quality (e.g., Jones, 1991; Subramanyam, 1996; DeFond and Subramanyam, 1998; Kothari et al., 2005; Kim et al., 2012) as an indication for earnings management. In general, the purpose of accruals models is trying to untangle the normal portion of accruals that reflect the fundamental performance from discretionary accruals that represents earnings management. Prior empirical research considered predominately discretionary accruals as opportunistic earnings manipulation, which means that high levels of discretionary accruals (positive and negative) are assumed to indicate the low quality of reported earnings (Mohrmann, 2015). According to Perotti and Wagenhofer (2014), discretionary accruals are the measure for earnings quality that is most positively associated with the mispricing of a firm’s share, which is in accordance with the previous view. However, managers can also use earnings management as a signaling channel to communicate their private information, which indicates a positive relationship between the discretionary accruals and the earnings’ quality (Mohrmann, 2015).

Following DeFond and Subramanyam (1998), Kothari et al. (2005) and Kim et al. (2012), the residuals from the annual cross-sectional industry regression model will be used as estimates of firm i’s discretionary accruals, because of its superior specification and less restrictive data requirements (Kim et al. 2012). Following Kim et al. (2012), the modified Jones model will be augmented by including ROA t-1 as a regressor in the estimation model as a control of the effect of performance on measured discretionary accruals. This is to avoid

21

potential misspecification and thus to enhance the reliability of inferences from discretionary accrual estimates (Kim et al. 2012). Specifically the following regression will be estimated:

TAit/ Ait-1 =

α

α

α

+

α

ε

where:

TAit

=

ΔREVit = change in net revenues in year t from year t-1; ΔRECit = change in net receivables;

PPEit= gross property, plant, and equipment;

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

The abnormal accruals are the residuals of Model (1.1). Since earnings understatements and overstatements are both indications of earnings management, the absolute value of the discretionary accruals (ABS_DA) will be used for the main analyses as the first earnings management characteristic (Warfield et al. 1995; Klein 2002; Kim et al. 2012; Mohrmann 2015).

Real Activities Manipulation

Real activities manipulation means that management takes actions that deviate from normal business practices undertaken to meet or beat certain earnings thresholds (Kim et al., 2012). This research relies on the prior study of Kim et al. (2012) to develop the proxies for real activities manipulation. Following Kim et al. (2012), four measures to detect real activities manipulation will be used: (1) abnormal levels of operating cash flows (AB_CFO), (2) abnormal production costs (AB_PROD), (3) abnormal discretionary expenses (AB_EXP), and (4) a combined measure of real activities manipulation (COMBINED_RAM). Kim et al. (2012) measured the abnormal levels of operating cash flows, production costs and discretionary expenses as the residual from the relevant models estimated by year and the two-digit SIC industry code. The combined proxy COMBINED_RAM will be calculated as AB_CFO - AB_PROD + AB_EXP.

22

Roychowdhury’s (2006) model will be used to estimate the normal level of operating cash flows:

CFOt / At-1 =

α

α

ε

,

where:

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

S = net assets;

ΔS = St – St-1.

For every firm-year, the abnormal cash flow from operations (AB_CFO) is the residual from the regression model (1.2).

Abnormal production costs is another measure of real activities manipulation. Production costs is defined as the sum of COGS and change in inventory during the year, and expenses are expressed as a linear function of contemporaneous sales (Roychowdhury 2006; Cohen et al. 2008; Badertscher 2011; Zang 2012; Kim et al. 2012). Production costs will be measured as PRODt = COGSt + ΔINVt. Following these studies, the following model for

normal production costs is estimated:

PRODt/At-1 =

α

α

ε

,

The abnormal production cost (AB_PROD) is the residual of the regression model (1.3)

The final measure that will be used for real activities manipulation is the abnormal discretionary expenses. Following Kim et al. (2012), the normal level of discretionary expenses will be estimated using the following equation:

DISEXPt/At-1 =

α

α

ε

,

where DISEXPt is the discretionary expenses in year t and consists of the sum of R&D,

Advertising and SG&A expenses. The abnormal discretionary expenditure (AB_EXP) is the residual of the regression model (1.4) for every firm-year (Kim et al. 2012).

23 3.4 Empirical Models

To capture the relationship between board interlocks and earnings management in financial reporting, the following models will be estimated:

ABS_DAt (or DAt) =

α

α

+

α

α

α

α

α

α

α

+ε t ; (2.1)

RAM_PROXYt =

α

α

α

α

α

α

α

α

α

+ ε t ; (2.2)

where:

ABS_DA (DA) = absolute value of discretionary accruals (signed discretionary accruals), where discretionary accruals are computed through the cross-sectional modified Jones model adjusted for performance;

RAM_PROXY = AB_CFO, AB_PROD, AB_EXP, or COMBINED_RAM: AB_CFO = the level of abnormal cash flows from operations;

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

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

COMBINED_RAM = AB_CFO - AB_PROD + AB_EXP;

INTERLOCKING_DIRECTORSHIP = an indicator variable that takes a value of 1 if the firm has a interlocked board, and 0 otherwise;

SIZE = natural logarithm of total assets; GROWTH = the change in total assets;

BIG4 = an indicator variable that takes the value of one if the firm is audited by a Big 4 auditor;

LEV = the total liabilities divided by the total assets;

RD_INT = R&D intensity (R&D expense/net sales) for the year;

AD_INT = advertising intensity (advertising expense/net sales) for the year

Equations (2.1) and (2.2) will be estimated with multiple regressions. Firms have many options to manage their reported earnings and are likely to use the technique that is less costly

24

to them. They can use discretionary accruals, real activities manipulation and even a mix of the two tools. The trade-off between the two methods to manage earnings is a function of their relative costs (Zang, 2012; Kim et al., 2012). Following Kim et al. (2012), the substitutive nature of these earnings management methods will be controlled by including ABS_DA, which is a proxy for accrual-based earnings management, as a control variable in the real manipulation (RAM_PROXY) regressions and by including a proxy for real activities manipulation as a control variable in the accrual-based earnings management (ABS_DA or DA) regressions.

Various other control variables will be included that could affect financial reporting behavior like earnings management. The size of the firm and firm-specific growth opportunity can potentially explain the significant variation in earnings management according to Roychowdhury (2006). Therefore and following Kim et al. (2012), proxies for growth opportunities (GROWTH) and firm size (SIZE) will be included.

Besides that, the extent of earnings management might differ for firms when they are audited by large audit firms (Becker et al., 1998; Francis et al., 1999; Kim et al., 2012). Therefore, the indicator variable, BIG4, will be included in the regressions for the firms using a Big 4 audit firm. Further, a leverage variable (LEV) will be included to control for the leverage-related incentives for earnings management (Teoh et al., 1998; Kim and Park, 2005; Kim et al., 2012).

McWilliams and Siegel (2000) found that advertising intensity and R&D intensity in the industry are positively associated with earnings. Following Kim et al. (2012), RD_INT and AD_INT are included to control for a company’s R&D expenditure and advertising expenditure.

25 4. Results

4.1 Introduction

As discussed in chapter three, two empirical models were constructed to test the effect of board interlocks on earnings management and the results are presented here. At first, chapter four presents the results of the descriptive statistics for the used data in this study. After that, this chapter reports the results of the multiple regression analyses undertaken to test the hypotheses that were developed in chapter two. It will report on the results looking at the hypothesized relationships between board interlocks and higher levels of discretionary accruals, lower levels of abnormal cash flow from operations, higher levels of abnormal production costs and lower levels of discretionary expenses.

4.2 Descriptive Statistics and Univariate Analysis

In Table 2, descriptive statistics and Pearson correlations are presented (see Appendix A for variable definitions). All continuous variables are winsorized at the top and bottom 1 percent of their distributions to control for outliers. The descriptive statistics of the used sample are shown in panel A and B of Table 2. Panel A of Table 2 presents the descriptive statistics of the full sample. It shows a mean value of -0.014 for the absolute value of discretionary accruals (ABS_DA). This is a very low mean compared to prior research of Kim et al. (2012). The difference in mean may be explained by the huge difference in sample size, since the full sample of Kim et al. (2012) consisted of 18,160 firm-year observations and this research just consists of 507 firm-year observations. The mean values of abnormal cash flows (AB_CFO), abnormal production costs (AB_PROD), and the combined real activities manipulation proxy (COMBINED_RAM) are -0.001, 0.001 and 0.000, which respectively suggests that, on average, real activities are not manipulated through production costs and cash flows. So, at first sight, firms are not likely to engage in real activities manipulation such as overproduction or sales manipulation. On the other hand, the mean value of abnormal discretionary expenses (AB_EXP) is -0.003, which indicates that the firms in the sample tend to reduce discretionary expenditures to manage earnings.

For the control variables, 96 percent of the full sample firms are audited by Big 4 auditing firms. On average, the sample firms’ advertising and R&D expenditure are approximately 5 percent and 0.2 percent of their net sales. Besides, on average, the firms in the sample have an increase of total assets of 13 percent and a leverage of 54 percent.

26

board interlocks and firms without board interlocks. Panel B shows and compares the descriptive statistics of variables between board interlocked and non-board interlocked firms. This is a preliminary univariate analysis, which indicates an answer to the hypotheses before the multivariate analysis will be implemented. In this way, early differences in the earnings management proxies and control variables between the board interlocked and non-board interlocked firms can be noticed. Both board interlocked and non-board interlocked samples exhibit income-decreasing accruals (e.g., mean ABS_DA = -0.064 for the board interlocked sample and -0.010 for the non-board interlocked sample). The mean and median of the variable ABS_DA differ between the two groups, but are far from statistically significant (p = 0.727), which indicates that there is no evidence that board interlocked firms are more likely than non-board interlocked firms to use income-decreasing discretionary accruals for earnings management. For the real activities management (RAM) proxy, the mean and median values of AB_CFO, AB_PROD, AB_EXP and COMBINED_RAM are higher for the board interlocked firms than for the non-board interlocked firms (e.g., mean value of COMBINED_RAM is 0.229 for the board interlocked firms and -0.019 for the non-board interlocked firms). However, the differences of the AB_CFO, AB_PROD, AB_EXP and COMBINED_RAM mean and median between the two independent samples are statistically insignificant (p = 0.990, p = 0.377, p = 0.524 and p = 0.528). Altogether, there is no evidence that board interlocked firms are more likely to engage in real activities manipulation than non-board interlocked firms, a finding that is not consistent with the excessive-familiarity effect hypothesis.

Furthermore, there is a difference in the control variable BIG4 between the board interlocked and non-board interlocked sample. This means that non-board interlocked firms are more likely to be audited by a Big 4 accounting firm than board interlocked firms (mean BIG4 is 0.87 for the board interlocked firms and 0.97 for the non-board interlocked firms with p = 0.084). Besides, Panel B shows that SIZE is higher for board interlocked firms than non-board interlocked firms with 0.091, but without any significance. Since the BIG4 difference in mean between the two groups is significant and larger firms are more likely to be audited by a Big 4 accounting firm, it makes sense that the SIZE difference in mean between the two groups is insignificant. However, the other control variables don’t show a significant difference between board interlocked and non-board interlocked firms.

In Panel C of Table 2, the Point-Biserial Correlation coefficients are presented for the selected variables. It shows that the INTERLOCKING_DIRECTORSHIP is negatively correlated with ABS_DA, but it is not statistically significant. As well,

27

INTERLOCKING_DIRECTORSHIP is insignificantly correlated with AB_CFO, AB_PROD, AB_EXP, and COMBINED_RAM. This means that there is no evidence that board interlocked firms are more engaged in earnings management than non-board interlocked firms.

TABLE 2

Descriptive Statistics of Selected Variables Panel A: Full Sample

n Mean Median Std. Dev. Min Max Dependent Variables ABS_DA 507 -0.014 -0.217 0.919 -1.220 3.484 Positive_ABS_DA 185 0.884 0.516 0.931 0.003 3.484 Negative_ABS_DA 322 -0.530 -0.514 0.321 -1.220 -0.005 AB_CFO 507 -0.001 -0.075 0.975 -2.521 3.408 AB_PROD 507 0.001 0.095 0.993 -3.123 2.165 AB_EXP 507 -0.003 -0.002 0.977 -1,988 3.551 COMBINED_RAM 507 0.000 -0.384 2.323 -5.346 6.552 Variable of interest INTERLOCKING_DIRECTORSHIP 507 0.070 0.000 0.264 0.000 1.000 Control Variables SIZE 507 7.550 7.488 1.465 4.272 11.956 GROWTH 507 0.126 0.067 0.282 -0.346 1.832 BIG4 507 0.960 1.000 0.195 0.000 1.000 LEV 507 0.537 0.532 0.232 0.088 1.306 RD_INT 507 0.002 0.000 0.010 0.000 0.089 AD_INT 507 0.046 0.036 0.045 0.001 0.210

28

Panel B: Descriptive Statistics by Board Interlocked versus Non-Board Interlocked Firms

Board Interlocked Firms Non-Board Interlocked firms

Difference Tests: p-value

n Mean Median n Mean Median t-test

Dependent Variables 0.727 ABS_DA 38 -0.064 -0.326 469 -0.010 -0.205 Positive_DA 11 1.071 0.636 174 0.872 0.502 0.493 Negative_DA 27 -0.527 -0.616 295 -0.531 -0.503 0.954 AB_CFO 38 -0.003 -0.145 469 0.001 -0.073 0.990 AB_PROD 38 -0.137 0.099 469 0.012 0.095 0.377 AB_EXP 38 0.095 -0.061 469 -0.010 -0.002 0.524 COMBINED_RAM 38 0.229 -0.457 469 -0.019 -0.366 0.528 Variable of interest INTERLOCKING_DIRECTORSHIP 38 1.00 1.00 469 0.00 0.00 Control Variables SIZE 38 7.635 7.577 469 7.544 7.471 0.713 GROWTH 38 0.101 0.066 469 0.128 0.067 0.583 BIG4 38 0.87 1.00 469 0.97 1.00 0.084 LEV 38 0.523 0.519 469 0.538 0.532 0.686 RD_INT 38 0.002 0.000 469 0.002 0.000 0.862 AD_INT 38 0.050 0.035 469 0.046 0.035 0.550

Panel C: Correlations among Interlocking Directorship, Earnings Management Proxies, and Other Selected Variables

1 2 3 4 5 6 1. INTERLOCKING_DIRECTORSHIP 1.000 2. ABS_DA -0.016 1.000 3. AB_CFO -0.001 0.239** 1.000 4. AB_PROD -0.039 -0.024 -0.426** 1.000 5. AB_EXP 0.028 0.057 0.104* -0.797** 1.000 6. COMBINED_RAM 0.028 0.133** 0.639** -0.936** 0.799** 1.000 7. SIZE 0.016 -0.018 0.054 0.039 -0.057 -0.018 8. GROWTH -0.024 0.105* 0.067 -0.067 0.146** 0.108* 9. LEV -0.018 0.029 -0.102* 0.029 -0.087 -0.086 10. RD_INT 0.008 0.255** 0.036 -0.065 0.146** 0.102* 11. AD_INT 0.027 -0.088* 0.116** -0.353** 0.533** 0.425**

29

Panel C: Correlations among Interlocking Directorship, Earnings Management Proxies, and Other Selected Variables (continued)

7 8 9 10 11 7. SIZE 1.000 8. GROWTH -0.060 1.000 9. LEV 0.39** -0.157** 1.000 10. RD_INT -0.020 0.071 -0.120** 1.000 11. AD_INT 0.153** -0.114* -0.015 -0.058 1.000

*, ** Indicate statistical significance at the 0.05 and 0.01 levels, respectively, based on a two-tailed test.

Variables are defined in Appendix A.

Models used in order to calculate the earnings management proxies are explained in Appendix B. In Panels B and C, a firm is defined as a board interlocked firm if the firm has a board with directors serving on the boards of multiple corporationsand vice versa and takes the value of 1; and a non-board interlocked firm otherwise. In Panel B, significances of means and medians are evaluated based on the t-test, respectively (p-values for the t-statistic).

30

4.3 The Relation between Board Interlocks and Accrual-Based Earnings Management

The results of the multivariate regression analyses of discretionary accruals are shown in Table 3. For all multivariate analyses, test statistics and significance levels are reported. To measure accrual-based earnings management, the absolute value of discretionary accruals (ABS_DA), positive and negative discretionary accruals (Positive_ABS_DA and Negative_ABS_DA) are used to report the results. Since hypothesis 1A states that board interlocked firms exhibit more discretionary accruals than non-board interlocked firms, a positive relation between INTERLOCKING_DIRECTORSHIP and ABS_DA is expected. Not consistent with hypothesis, a negative relation between INTERLOCKING_DIRECTORSHIP and ABS_DA is found. To be more specific, the estimated coefficient on INTERLOCKING_DIRECTORSHIP is negative (-0.041), which indicates that board interlocked firms use less accruals for earnings management than non-board interlocked firms. However, the p-value is 0.784, so it is far from significant. In the second and third column, a positive relation is found between INTERLOCKING_DIRECTORSHIP and Positive_ABS_DA (Negative_ABS_DA), indicating that board interlocked firms are more engaged in income-increasing (income-decreasing) earnings management. Unfortunately, the coefficients on INTERLOCKING_DIRECTORSHIP for both samples are not significant (p = 0.504 for Positive_ABS_DA and p = 0.911 for Negative_ABS_DA). Concluding, there is no evidence that board interlocked firms are more engaged in earnings management through discretionary accruals than non-board interlocked firms. Therefore, hypothesis 1A will be rejected.

Furthermore, there is a positive and significant relation between the combined proxy for real activities manipulation (COMBINED_RAM) and ABS_DA (p = 0.001). This indicates that firms managing earnings through accruals are less likely to manipulate real activities, and versa. This means that earnings management through discretionary accruals and real activities manipulation are substitutes of each other, which is consistent with prior research (Graham et al. 2005; Cohen et al. 2008; Badertscher, 2011; Zang, 2012; Kim et al. 2012). Moreover, RD_INT is positively and significantly (p < 0.01) associated with ABS_DA, which indicates that firms with high R&D expenditures also have higher levels of discretionary accruals.

31 TABLE 3

Multiple Regression of Accrual-Based Earnings Management on Interlocking Directorship (n = 507)

ABS_DA Positive_ABS_DA Negative_ABS_DA

Coefficient Coefficient Coefficient (t-stat) (t-stat) (t-stat)

INTERLOCKING_DIRECTORSHIP -0.041 (-0.275) 0.191 (0.669) 0.007 (0.111) COMBINED_RAM 0.066 (3.438)*** 0.017 (0.607) 0.022 (2.104)** SIZE -0.014 (-0.481) 0.005 (0.100) 0.048 (3.533)*** GROWTH 0.221 (1.549) 0.208 (1.032) -0.009 (-0.112) BIG4 0.109 (0.537) 0.125 (0.349) 0.056 (0.613) LEV 0.347 (1.847)* 0.609 (2.171)** -0.088 (-0.916) RD_INT 21.075 (5.448)*** 10.731 (2.489)** -2.581 (-0.469) AD_INT -2.698 (-2.726)** -2.626 (-1.367) -0.550 (-1.253) Adj. R2 _{0.090 0.040 0.027}

*, **, *** Indicate statistical significance at the 0.10, 0.05, and 0.01 levels, respectively, based on two-tailed tests.

Variables are defined in Appendix A.

32

4.4 The Relation between Board Interlocks and Real Activities Manipulation

In Table 4, the results of the multiple regression analyses with the measures of real activities manipulation are reported. The coefficients and significance levels are shows for each proxy. Four proxies were used to measure real activities manipulation: abnormal cash flow from operations (AB_CFO), abnormal production costs (AB__PROD), abnormal discretionary expenses (AB_EXP) and the sum of real activities manipulation proxy (COMBINED_RAM). Since higher levels of AB_CFO, AB_EXP and COMBINED_RAM, and lower levels of AB_PROD indicate more conservative operating decisions (Kim et al. 2012), the opposite is expected from the INTERLOCKING_DIRECTORSHIP coefficient for the regressions of AB_CFO, AB_PROD, AB_EXP and COMBINED_RAM.

The estimated coefficient for INTERLOCKING_DIRECTORSHIP in the AB_CFO regression is negative (-0.015), which is consistent with hypothesis 2A. However, the p-value is 0.923, so it is insignificant by far. Therefore, hypothesis 2A must be rejected. Turning to the control variables, the coefficients of SIZE, LEV, and AD_INT are significant (p = 0.036, p =0.003 and p = 0.005), indicating that larger firms, low leveraged firms, and firms with high advertising intensity are less likely to engage in real activities manipulation through abnormal

cash flows.

Next, the estimated coefficient for INTERLOCKING_DIRECTORSHIP in the AB_PROD regression is negative (-0.109), which is inconsistent with hypothesis 2B. However, the p-value is 0.486, so it is insignificant. Therefore, hypothesis 2B must be rejected. The control variables SIZE, GROWTH, RD_INT and AD_INT are significantly associated with AB_PROD (p = 0.021, p = 0.014, p = 0.071 and p < 0.01). So, the level of abnormal production costs is influenced by those four variables.

Besides, the estimated coefficient for INTERLOCKING_DIRECTORSHIP in the AB_EXP regression is positive (0.061), which is not in the expected direction. But again, the result is not significant (p = 0.645). Thus, hypothesis 2C must be rejected. Just like in the AB_PROD regression, only the variables SIZE, GROWTH, RD_INT and AD_INT influenced the level of AB_EXP.

At last, the estimated coefficient for INTERLOCKING_DIRECTORSHIP in the COMBINED_RAM regression is positive (0.150), which is not the expected direction. However, the coefficient is not significant (p =0.665). The coefficient on ABS_DA is positive and significant (p < 0.01) in the AB_CFO and COMBINED_RAM regression, which is consistent with Cohen et al. (2008) and Kim et al. (2012).