Effect of blockholders on accruals-based earnings management before, during and after the financial crisis

(1)

Effect of Blockholders on Accruals-based

Earnings Management

Before, During and After the Financial Crisis

by

Mohsan Akbarali

Student number: 11094982

Master’s thesis submitted in support of the degree of Master of Science in Accountancy and Control

Track: Both

University Of Amsterdam Faculty of Economics and Business

Thesis supervisor: Dr. Reka Felleg Word count: 14,521

(2)

Statement of Originality

This document is written by student Mohsan Akbarali 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 empirical thesis aims to investigate whether there is a difference in the relationship of blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. Public listed firms of the United States, during 2005-2014, are used in the dataset. This thesis adopts the Modified-Jones model as the measurement of discretionary accruals. I confirm that there is a negative relationship between blockholders and accruals-based earnings management, what is in line with prior research. The result shows that there is a significant difference in the relationship of blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. Further the relationship of blockholders and accruals-based earnings management does not become less negative after the financial crisis compared to before the financial crisis. Results implies that the last 10 years the relationship of blockholders and accruals-based earnings management has stayed stable, excluding the financial crisis years. During the financial crisis years the relationship between blockholders and accruals-based earnings management is less negative than before and after the financial crisis. This study contributes as being the first study to measure whether the financial crisis changed the effect of blockholders on accruals-based earnings management.

(4)

2.4 Hypothesis development ... 14 3. RESEARCH DESIGN ... 16 3.1 Sample selection ... 16 3.2 Regression model ... 17 3.2.1 Dependent variable ... 17 3.2.2 Independent variables... 18 3.2.3 Control variables ... 19 3.3 Statistical test ... 20 4. RESULTS ... 22 4.1 Univariate analysis... 22 4.2 Regression analysis ... 25 4.3 Robustness analysis ... 27 5. CONCLUSION ... 29 REFERENCES ... 32

(5)

1. Introduction

The global financial crisis of 2008 has led to the questioning of the role of blockholders on protecting the value of the firm (Baryeh, 2014). After the financial crisis researches have focused on financial information quality and corporate controls. According to Salem et al. (2012) quality information is the anchor of capital markets. If there is no quality information, then the market would be inefficient and there would be no liquidity. One of the reasons for the financial crisis was also information quality (Lins, et al., 2011). According to Easley and O’Hara (2004) information asymmetry is created by inefficient markets. One mechanism to reduce information asymmetry is by having investors who are blockholders. More than 90% of the firms worldwide have blockholders that own more than 40% of the shares together (Holderness, 2009). Because of their large shares, they have influence on the decision making process of the management. It could even mean that blockholders play a central role in shaping firms policies and performance (Holderness, 2003).

Beginning this century there were many accounting scandals, like Enron, HealthSouth, Lehman Brothers, Tyco and World.com. These accounting scandals had mainly to do with accruals-based earnings management (Dechow et al. 2012). According to Loomis (1999) earnings management is a big concern for practitioners and controllers. Earnings management hides true firm performance and thereby impacts decisions taken by shareholders. In order to mitigate earnings management, there are several external and internal corporate governance mechanism. One internal mechanism of corporate governance is blockholders which is a part of ownership structure. Generally blockholders are associated with better governance for two reasons. Firstly, there is direct monitoring (intervention) by blockholders on managers at the operation level (Cronqvist and Fahlenbrach, 2009). Secondly, blockholders discipline managers by threatening to sell their shares (exit), which will have negative impact on the stock price what again is related to the bonus of managers (Kahn and Winton, 1998).

Large shareholders can play an important role to mitigate earnings management. Large shareholders, i.e. blockholders, can punish firms by intervention or exit. This means that a firm is put under pressure when having blockholders. On the other hand if a firm is performing well then a blockholder will reward the firm. The reward is done by blockholders by not selling their shares on the capital market. If the capital market anticipate that a blockholder is not selling their shares, this means that a firm is performing well. In turn the capital market will reward the firm by raising the stock price, which again is tied to the compensation of the manager.

This study investigates, this is also the research question, what the effect of blockholders on accruals-based earnings management is before, during and after the financial crisis. In order to

understand better the relationship between blockholders and earnings management, agency theory can provide perspective. The separation between ownership and management, between investor and

managers, leads to conflict of interest and information asymmetry. According to Bukit and Iskandar (2009) managers will not act in best interest of shareholders and will use earnings management for their own benefit. This will lead to agency problem and agency cost. In order to mitigate this problem there is a

(6)

contractual setting in place. A contractual setting is a contract between investors and managers, in order to mitigate conflict of interest and information asymmetry. But only a contractual setting will not be enough to solve this problem (Watts and Zimmerman, 1986). According to Gompers et al. (2003) blockholders can reduce the agency problem and increase the efficiency of the contractual setting.

Prior research regarding effect of blockholders on earnings management shows no general consensus on the effect of blockholders on earnings management. Some studies show there is a positive effect, while other studies indicate a negative effect. For example Yeo et al. (2002) found in their study that blockholders can be an effective internal control mechanism by reducing earnings management. This means that blockholders reduces the level of discretionary accruals. But on the other side Al-Fayoumi et al. (2010) found that blockholders don’t significantly monitor earnings management, be it accruals, real earnings or even restatements. Barclay and Holderness (1991) concluded in their research that the presence of outside blockholders can create pressure on managers to engage in income increasing earnings management. But there is even more controversy regarding the effect of blockholders. Like for example Ding (2007) found that there is both a positive and negative effect of blockholders on earnings management. One of the most prominent suggested reason for these differences is that blockholders aren’t homogeneous. Blockholders are heterogeneous and vary in their abilities to monitor. This means that a bank as an investor have different incentives by being a blockholder then a private investor. Past studies examines the relationship of blockholders and accruals-based earnings management, but this study argues that the financial crisis can exacerbate accruals-based earnings management in the presence of blockholders. I argue that outside capital is less available and the

expected return of existing capital investments decreases (Lins, et al., 2011). This means less investment can be made and the return of investment will decrease. This study suggest, because of previous effect, that blockholders will extract more private benefits that result in higher accruals-based earnings

management. Therefore I expect that accruals-based earnings management will be higher in a crisis-period than in a non-crisis crisis-period. Further I expect that after the financial crisis the earnings management will be less than before the financial crisis. The reason is that Graham et al. (2005) suggest that real earnings management will be used more than accruals-based earnings management in the future. Graham et al. (2005) argues that accruals-based earnings management is easier to detect by

blockholders. Further the financial crisis has taken care of that the annuals reports are subjected to more rules and are more scrutinized by auditors and blockholders (Xu et al., 2013). This all will result in likely less accruals.

Like many previous studies, this study also focusses on the United States for a couple of reasons. The market of the United States is developed, this increase the generalizable effect and gives the study a more reliable image. According to Leuz et al. (2003) in developing countries the determinants could be very different and therefore the results are not generalizable. Also it is possible to do a comparison with previous studies, analyzing what more could be the reason for so divergent conclusions. In order to measure the impact of blockholders on accruals-based earnings management, Modified-jones model

(7)

I confirm first of all that the general relationship between blockholders and accruals-based earnings management is negative. As most prior studies suggested, this study also find a negative relationship. The main question of this study is whether the negative relationship, between blockholders and accruals-based earnings management, is less negative in a crisis period compared to a non-crisis period. I find that there is a significant difference between the relationship of blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. Therefor the first hypothesis is accepted. This is in line with Zhong et al. (2007) and Ely and Song (2000), who suggested that blockholders put pressure on managers when a firm has declining performance. The second

hypothesis is rejected, which suggested that after the financial crisis the relationship between blockholders and accruals-based earnings management will be less negative than before. This is not in line with prior research of Cruijsen et al. (2016) and Xu et al. (2013), who suggested that after the financial crisis there will be less accruals because of the scrutiny done by blockholders on managers and annual reports. Overall it means that there is no difference in the relationship between blockholders and accruals-based earnings management over the last 10 years, excluding the financial crisis years. During the financial crisis years the relationship between blockholders and accruals-based earnings management became less negative.

This study contributes by examining the effect of blockholders on earnings management before, during and after the financial crisis. This is the first study to analyze what the effect of the financial crisis is on accruals-based earnings management moderated by blockholders. Namely, it compares the effect of blockholders on accruals-based earnings management in a crisis period and in a non-crisis period. Further this study contributes by measuring the relationship between blockholders and accruals-based earnings management, from listed firms in the United States, of the last five years. Most recent study about the effect of blockholders on earnings management is of Dou et al. (2016) and even their dataset goes as far as 2009. This study argues that the accruals have decreased after the financial crisis, because of the scrutiny done by regulators and blockholders (Xu et al., 2013). Finally it is interesting to compare this research with past researches and concluded why there have been so many different conclusions regarding the effect of blockholders on earnings management.

The remainder of this thesis is as follows. In section two the relevant literature regarding the concepts of blockholders and earnings management is described. Based on the literature the hypotheses of this study is formulated. In section three the research design, also known as methodology, is explained. Sections three exist of how the sample and data are collected. Further in that section is described how earnings management is measured with the empirical models that are used to test the hypotheses. In section four the result will be set forth and in section 5 this research will be summarized with a conclusion and the limitations.

(8)

2. Literature review and hypothesis development

This chapter is divided in three literatures and one hypothesis section. The first part describes the separation of ownership and management what creates the agency problem and the related costs. The second part shows what one of the consequences of agency problem could be, namely earnings

management. The last part of the literature sections shows how studies differs on the role of blockholders on mitigating or exacerbate earnings management. By combing the literature around blockholders and earnings management the hypotheses are presented in the last section of this chapter.

2.1 Separation of ownership and management

The separation between ownership and management leads to deviant behavior between investors and managers (Berle and Means, 1932). This separation creates conflict of interest and information

asymmetry between shareholders and management, this is called agency problem. Berle and Gardiner (1968) shows in their study that conflict of interest leads to suboptimal management decisions. The consequences of these agency problem are that markets become imperfect and monitoring becomes costly. Thus this separation leads to agency problem and cost.

2.1.1 Agency problem and cost

Large modern firms are directed by managers who don’t necessary have the same goals as the shareholders. Mostly when managers do not own the company they act in self-interest (Jensen & Meckling, 1979). Also Bukit and Iskandar (2009) concluded that managers will not act in the best interest of shareholders and will use earnings management. Managers have incentives to manage earnings through earnings management. For example in order to achieve certain goals and maximize their bonus, managers could take decisions to manipulate figures in the financial report. Such opportunistic behavior of managers is possible because the actions of managers are difficult to observe. This leads to the agency problem between investor and managers. Investors want that managers act in their interest, while managers act in their own interest.

Conflict of interest can have potential agency cost, like management decisions that do not maximize shareholders’ interests. Jensen and Meckling (1976) define agency cost by separating them in three categories. The first category is monitoring expenses, these expenses are incurred when the principals tries to restrict the actions of the agents. These expenses are like voting in general meetings and litigation cost to sue managers. The second one is bonding expenses, these expenses are incurred by the agent when they are obliged to commit to contractual obligations. By committing to these

obligations an agent’s job can be limited or restricted. These expenditure are like disclosure of annuals reports and forego powers by managers. The third is residual loss, also called earnings management, these costs are incurred from divergence of interest between principal and agent. These costs are incurred despite the use of monitoring and bonding expenses.

(9)

2.1.2 Corporate governance

In order to solve agency problem there could be a contractual setting between investors and managers.. This contractual setting is in place to monitor managers and to assure that shareholders and firm value are protected. This means that managers are rewarded if they act in the interest of shareholders and the firm. But only this contractual setting in place may not be enough to resolve the agency problem. The biggest issue here is that managers only disclose information that is best aligned with their interest and shareholders make decisions on basis of this (Watts and Zimmerman, 1986).

According to Gompers, Ishii and Metrick (2003), in addition to a contractual setting, corporate governance can reduce the agency problem and increase the efficiency of the contract setting between investors and managers. They stated that even in developing countries agency problem can be a large source of cost. Gabrielle O’Donovan (1995) defines corporate governance as an internal system encompassing policies, processes and people that serve the needs to shareholders and other

stakeholders by directing and controlling management activities with good business practices, objectivity and integrity.

Corporate governance is designed to pursue interest of shareholders, like maximizing firm value and obtaining reasonable return on capital (Shleifer and Vishny, 1997). Corporate governance can protect shareholders and firm value by external and internal mechanism. External mechanism are market for corporate control, corporate laws, SEC rules and regulations, market takeover force and charters. Internal control mechanism is board independence, board composition, audit committee, compensation

committee, nomination committee and ownership structure. Thus it means that corporate governance is a set of control mechanism were outside investors can protect themselves against inside investors and managers.

2.1.3 Ownership structure

One internal mechanism of corporate governance is ownership structure (Almazan et al., 2005) Ownership structure is for example institutional ownership, ownership concentration and blockholders. Agency theory suggest that monitoring by large investors is an effective corporate governance

mechanism. Almazan, Hartzell & Starks (2005) suggest that large investor can provide more effective monitoring compared to small investors. The reason here for is that small investor are mostly passive and less informed compared to large investor. On contrary Duggal and Millar (1999) suggest that large investor don’t provide an effective monitoring role. According to them they are also passive investor, like small investor. The reason here for is when the firm is performing poorly they don’t monitor (intervention), but sell their shares (exit). This doesn’t lead to better monitoring, because they have sold their shares.

2.2 Earnings management

Earnings management is a part of agency cost and is defined in two categories by Jensen and Meckling (1976). Earnings management can affect the credibility of financial information, which can lead to major

(10)

financial scandals (Enron and World.com) and may even lead to a crash in the capital market. According to McNichols and Wilson (1988) earnings management is one of the best way to measure whether managers are acting in accordance to maximize shareholder value. Earnings management is choice of a manager to intentionally affect earnings, upward or/and downward. Earnings management can have effect on the credibility of financial statements, there could be materially manipulation. This isn’t desirable in a well-functioning capital market. According to McNichols and Wilson (1988) earnings management can occur in two ways. The first one is the structuring of certain revenue and/or expense transactions, also called real earnings management. Second one is accruals-based earnings management. Of these two ways, the method of real earnings management is the most damaging one. The reason here for is that investors are unaware of the extent of such manipulation (Mitra and Rodrigue, 2002). On basis of this, investor can make decision that are not in favorable on the share price.

2.2.1 Reasons for earnings management

Prior studies have researched what the reasons are for earnings management. Healy (1985) provide one of the earliest reason for earnings management, namely compensation scheme. Managers have inside and private information, there for they think in their own interest. They try to maximize their bonus by managing income. Therefore it is very likely that managers will increase current period income. Another reason for managing earnings upward, at expenses of future earnings, is for job security (DeFond and Park, 1997). According to Gaver, Gaver, and Austin, (1995) debt covenant is another reason for earnings management. A long term debt contract has specific covenants to protect debt holders. If a firm violate a debt covenant, this will lead to higher cost and potentially to a lawsuit. Therefor managers will manage their earnings, to not violate this debt covenant. Sweeney (1995) find in his research firms who are on the edge of defaulting, tend to adopt new accounting standards to increase earnings. Also firms are going to avoid reporting losses by earlier managed earnings. Healy and Wahlen (1999) conclude that managers can be motivated to overstate earnings to meet or beat analyst forecast. All this researches concludes that managers tend to use earnings management for many different reason. Even so different that Brown and Higgins (2001) finds that reasons for earnings management differs per country.

2.2.2 Accrual based vs real earnings management

Most studies have focused on accrual based accounting and real management accounting. Accrual is the different between the real earnings and cash flow from operating activities. Accruals exist of

non-discretionary accounting and non-discretionary accounting. Non-non-discretionary accounting aren’t considered manipulation, because managers don’t have discretion over these accruals. On the other hand

discretionary accounting are seen as earnings management. The reason here for is that managers do have discretion and this can become a threat when managers behave opportunistically. Accrual based accounting is like a firm using a provision for credit losses (Ahmed et al.1999), warranty costs, inventory values and the timing of unusual items (Cohen et al., 2011). Dechow et al. (2012) concludes in his

(11)

research that accruals-based earnings management can have a negative effect on revenue. The other type is real earnings management, is the use of direct cash flow, which could affect the firm in long term. The biggest distinction between these two is that accrual based earnings management have no direct cash flow affect, while real earnings management activities have direct cash flow affect. Graham et al (2005) finds that managers prefer real earnings management than accrual based earnings management. The reason here for is that real earnings are much harder to detect than accruals based earnings.

2.3 Blockholders

Blockholders is an internal mechanism of corporate governance. Blockholders are mostly seen to discipline managers by monitoring through voice (intervention in their operation activities) or by exit (threating to sell the shares) (Edmans, 2008). Because of this disciplining effect it is generally agreed that this reduces earnings management. On the other hand blockholders are also seen that put pressure on managers to meet or beat analyst forecast and market expectations, this could potentially lead to an increases in earnings management.

2.3.1 Definition of blockholders

Blockholders are shareholders who have a big proportion of shares compared to other shareholders in a firm. Blockholders can be defined in to two categories, namely outside and inside blockholders. Outside blockholders are defined as those shareholders who own at least 5 percent of a firm’s outstanding common stocks while they serve as neither the firms’ executive officers nor on the board directors (Zhong et al., 2007). Inside blockholders, also called managerial ownership, are shareholders who own at least 5 percent of a firm’s outstanding common stock while they serve as an executive role or a member of board of directors.

According to Edmans (2008) a blockholder gather information about the value of the firm, so he/she can monitor. For example a blockholder knows why a firm expect lower earnings for the coming period. This could be because the firm has made a large investment which will make profit over a couple of years. So if low earnings are because of a desirable long term investment, then a blockholder will not punish the firm. Punishing or monitoring can be done in two ways, by direct monitoring (intervention) or by threatening to sell shares (exit threat). Because the capital market sees that a blockholder is accepting lower earnings, they will anticipate that this has beneficial reason for the future. This decreases information asymmetry, what increases the market efficiency.

2.3.2 Positive view blockholders

Jensen and Meckling (1976), Shleifer and Vishny (1986) and Zhong et al. (2007) stated in their research that outside blockholders have incentives to monitor the action and decisions of managers. The logic behind this is that ordinary shareholders can sell their shares quickly for a stable price if they aren’t satisfied with firm performance. In contrary blockholders can’t sell their stocks for a stable prices. The

(12)

reason here for is that the shares are in a large block and this directly involves decrease of share price when sold. The market reacts to it that the selling of these shares are because of bad firm performance. Defond and Jiambalvo (1991), Dechow et al. (1996) and Yeo et al. (2002) conclude in their research that outside blockholders improves the credibility of financial statements by controlling earnings management practices in a firm. This means that blockholders reduces the level of discretionary accruals. Also Core (2000) shows in his paper that agency costs can be reduced by outside blockholders. The reason here for is that managers are more encourage to act in the interest of shareholders and to limit accounting

manipulation when they are under supervision of blockholders. Consistent with this result, Bedard et al. (2000), Yeo et al. (2002), Bos and Donker (2004), found that outside blockholders can have an effective control mechanism in the financial statements. All with all many researches of above shows that there is a negative relationship between blockholders and earnings management, this means more blockholders is less earnings management.

2.3.3 Negative view blockholder

There is also a view that blockholders could increase earnings management. Shleifer (2004) and Bolton et al. (2006) concludes that the presence of blockholders may increase pressure on managers to meet or beat benchmark. This means that there is pressure on managers, because of blockholders, to meet short-term results that are in line with peers. Because of this pressure managers will upward the earnings, through earnings management, to meet the given norm. McEachern (1975) Shleifer and Vishny (1986), Holderness and Sheehan (1988) and Barclay and Holderness (1991) also concludes in their research that the presence of outside blockholders can create pressure on managers to engage in income increasing earnings management. This is done to meet expectations of capital markets by announcing a favorable financial performance, while it doesn’t reflect economic reality. Also Ely and Song (2000) found that when a firm is underperforming then blockholders could exercises pressure on managers to enhance

performance. By this kind of situation managers are encouraged to do earnings management to keep the revenues stable through the years. In this sense, Zhong et al. (2007) propose that the existence of outside blockholders increases the motivation of managers to manage result upward to hide the decline of

performance. The reason here for is that decline in firm performance will affect the share price in a negative way. What in turn will decrease the share price owned by blockholders.

2.3.4 Controversial view blockholders

Besides the positive and negative effect, there a researchers who found positive as well negative effect in their research. Ding (2007) found in her study that there is positive as well negative relation of

blockholders on earnings management. She concluded this has to with the different types of blockholders. A bank as a blockholders has different incentives than a private investors. This is in line with Xu and Wang (1997) who found that firm performance is better when blockholders are financial institutions than when blockholders are corporations. García-Meca et al. (2009) found a non-linear relationship. This means the

(13)

relationship was positive at low blockholders and negative at high blockholders. The researchers

concluded that the effect was different because at low blockholders there were more chances of collusion. Dempsey et al. (1993) find that outside blockholders don’t lower earnings management, while inside blockholders have the ability to lower earnings management. Warfield et al. (1995) finds that outside blockholders don’t have significant effect on earnings management. While managerial ownership, inside blockholders, have a negative effect on earnings management. The controversial view is even more endorsed by Al-Fayoumi et al. (2010), they found that blockholders don’t significantly monitor earnings management, so the relationship is neither positive nor negative.

Dou et al. (2013) found a so called fixed effect. This means that the blockholder effect on earnings management was caused by the influence that blockholders has on the firms accounting practices. So for example if blockholders has in a firm high influences in their accounting practices, then the earnings management will be low. On the other hand if blockholders has in a firm low influences in their accounting practices, then the earnings management will be high.

2.3.5 Financial crisis and blockholders

There are two major impacts during a financial crisis, first one is that outside capital is less available and the expected return of existing capital investments decreases (Lins, et al., 2011). This means that there are less investment possibilities during a crisis period compared to a non-crisis period. For the existing investments, in a crisis period, there is likely less return on investment. Blockholders provide normally financing (Weinstein and Yafeh, 1998), provide monitoring (Kahn and Winton) and help in product markets (Khanna and Palepu, 2000). Financing is done through buying of outstanding shares, in order to grow their percentage of shares hold in a firm. Monitoring is done through intervention in the operations, in order to maximize the efficiency of managers and employees. Both these services are provided by blockholders in order to maximize their shares and firm value.

If the capital decreases and/or the expected return decreases, then the benefits of blockholders will become less pronounced. During a financial crisis there are high chances that the agency conflict will increase. This means that managers will extract more private benefits and will act in their interest. On the other hand a blockholder will act on his own interest and will try to extract his own benefit by putting pressure on managers and on smaller shareholders. Shleifer and Vishny (1997) concluded in their research that blockholders effort depends on their return on investment and since in a time of crisis the return of investment will be low, blockholders will likely extract more private benefits. Lastly Cimini (2015) found that during the financial crisis years (2007-2010) earnings management increased in the large majority of European countries and in the United States. This increase could have had different reasons, one potential reason could have been ownership structure. When a firm does not have the right ownership structure, then it is way easier to have more manipulation of accruals by managers.

(14)

2.4 Hypothesis development

The literature review shows that there is no general consensus about the effect of blockholders on earnings management. Researches who have found a positive effect, conclude that shareholders put too much pressure on managers to meet or beat benchmark what eventually leads too earnings management. Other research find that there is a negative effect, they conclude that blockholders serve as an internal monitor mechanism to increase financial reporting quality by reducing earnings management.

Dou et al. (2013) mentions two reasons why the effect of blockholders on earnings management has been so divergent in past research. The first reason they mention is that past research have not taken into consideration that blockholders are heterogeneous, rather they treated all blockholders as

homogeneous. To clarify, when blockholders are heterogeneous then they will have different incentives and therefor they will act differently. For example a bank as a blockholder will have different incentives then a blockholder who is a private investor. These different incentives leads to different results on the relationship of blockholders and accruals-based earnings management. The second reason Dou et al. (2013) mentions is that research takes different variables into consideration when measuring the effect of blockholders on earnings management. For example the effect before, during and after the financial crisis of blockholders on earnings management can be very different. As Lins, et al. (2011) concluded during a financial crisis there is less capital available and the return on investment decreases, this eventually can lead to extracting private benefits by blockholders.

Zhong et al. (2007) stated in their research that blockholders have incentives to monitor the action and decisions of managers, in order to maximize their profit. Zhong et al. (2007) proposed that the

existence of outside blockholders increases the motivation of managers to manage results upward to hide the decline of performance. The reason for this is that decline in firm performance will affect the share price in a negative way. Also Ely and Song (2000) found that when a firm is underperforming then

blockholders could put pressure on managers to enhance performance. This suggests that during a crisis, when a firm is underperforming, a blockholder will put pressure on managers to enhance firm

performance. Shleifer and Vishny (1997) concluded in their research that blockholders effort depends on their return on investment and since in a time of crisis the return of investment will be low, blockholders will likely try to extract more private benefits. Further, Cimini (2015) found that during the financial crisis years (2007-2010) earnings management increased in the large majority of European countries and in the United States. Lastly, Zang (2011) concluded in his research that accruals-based earnings management are easier to implement than real earnings management. Thus it means that it takes less time to use accruals-based earnings management than real earnings management. The reason for this is that accruals are changes in the booking keeping, while real earnings management are profound decision taken by the management. For example, delaying investment in order to keep the profit as high as possible. While this does not seem so damaging, it could have severe negative effect. And during a financial crisis managers are looking for a fast solution, thus it is probably that managers would prefer based earnings management. This is indicates that blockholders and managers prefer

(15)

accruals-based earnings management over real earnings management in a crisis period. Therefore it is likely that the effect of blockholders on accruals-based earnings management is different in a crisis period compared to a non-crisis period. This leads to the first hypothesis:

H1: The relationship between blockholders and accruals-based earnings management in

a crisis period is less negative than in a non-crisis period

The financial crisis was also caused by keeping bad debts off the financial statements (Barth and

Landsman, 2010). There were subprime loans that were kept off the balance sheets of a bank to depict a better economic position. The same goes for discretionary accruals, they can be used to depict a better economic position of a firm than it really has. According to Jin et al. (2011) one of the causes of the financial crisis was that a bank or a firm was depicted in a better economic position than reality. The culprit behind this were auditors, banks, blockholders and managers and this gave them a bad reputation

(Cruijsen et al., 2016). In order to win back the trust of shareholders and stakeholders, it is likely that blockholders will have an influence on managers to use less accruals, in order to prevent a crisis what affected the share price of blockholders (Cruijsen et al., 2016). Also Xu et al. (2013) concluded that the annual reports, after the financial crisis, will be more scrutinized by large shareholders (blockholders). This will result in likely less accruals. Further Graham (2005) found that managers prefer real earnings

management, because these are very hard to detect for an auditor. Especially after the financial crisis, managers would prefer real earnings management instead of accruals-based earnings management. In addition the second hypothesis adds to the literature in another way. The most recent research of the effect of blockholders on accruals-based earnings management is of Dou et al. (2016) and even their sample is until 2009. There are not any researches available that measure the effect of blockholders on accruals-based earnings management after the financial crisis. I argue that this adds to the literature because there are a couple of indications that after the financial crisis accruals-based earnings

management has declined because of blockholders. Xu et al. (2013) found that annual reports are more scrutinized by blockholders after the financial crisis. Zang (2011) concluded that accruals-based earnings management are easier to detect than real earnings management. Therefore it is likely that the effect of blockholders on accruals-based earnings management has become less negative. This leads to the second hypothesis:

H2: After the financial crisis the relationship between blockholders and accruals-based

(16)

3. Research design

This chapter starts with describing how the sample is gathered. Afterward the regression model is explained in three sub divisions. The first part explain the dependent variable, how the discretionary accruals are measured with the Modified-jones model (1995). The second part explains the independent variable, namely how blockholders are measured. Also the creation of the dummy variable is explained. In the third part the control variable are explained. To conclude this chapter the statistical test is explained.

3.1 Sample selection

The research sample of this thesis is from public listed companies from the United States. United States is been selected for the following reasons; it has a developed capital market, the data is highly available and the possibility for a comparison with previous studies. The sample period ranges from 2005 till 2014. This period is chosen, because there should be some years before the financial crisis and some years after. So there could be a comparison of what the effect of the financial crisis was on accruals-based earnings management. Therefore the time period before the crisis will be 2005 and 2006. Following on from 2007 until 2010 there was a financial crisis (Cimini, 2015). From 2011 there were signs of economic growth in the United States (Cimini, 2015). So from 2011 till 2014 will be the time period after the financial crisis. However, the proxies of earnings management requires lagged variables for past two years. Therefore the lagged variables data goes two years back, so for the lagged variable the data of 2003 and 2004 is also gathered.

The sample selection process starts by using Orbis, which is available at the University of

Amsterdam. First, all public listed companies in the United States are selected. This gives a result of 4,614 companies. Consistent with Cohen & Zarowin (2010) companies with the SIC codes ranging between 6000-6999 (financial firms) and 4000-4999 (utility companies) are excluded from the sample as these firms may have additional reporting requirements. Also financial firms are excluded because they have different incentives that managers use to manipulate earnings and have specific regulations that affects the discretionary accruals (Becker et al. 1998). After excluding these two segmentation, the sample results in 3,489 listed firms for which the financial data is gathered.

From here on all firms are deleted who have missing data, like missing ticker symbol (Orbis identifier). The ticker symbol is necessary to eventually merge the financial data of Compustat with the sample selection of Orbis. This reduces the sample even more and gives an output of 1,782 listed firms. The financial data for the sample is collected from Compustat which is available at the University of Amsterdam. The financial data is merged with the sample data. Afterwards for all firms-years who have missing financial data are deleted. So missing data about total asset, total debt, net income (loss), net cash flow, PPE (plant, property and equipment), sales, cost of goods sold, total inventories, advertising expenses, R&D expenses, selling general administration expenses, receivables and revenue. The financial data is also gathered for the year 2003 and 2004, as it is necessary to calculate the lagged variables. After deleting all the financial data, the sample consist of 919 listed firms.

(17)

To decide the final sample, both firms that had constantly a blockholder for the last 10 years and firms that had not constantly a blockholder for the last 10 years are included. The reason here for is that by choosing the same companies over the sample period, it reduces the noise in the sample. Also

according to Klein (2002) by choosing the same companies it gives a more biased result. Because having the same companies for each of the 10 years, it reduces the noise that other variables have on accruals. After deleting all those firms, the final sample consists of 372 firms. The total firm year observations are 2275.

3.2 Regression model

The first hypothesis (H1) predicts that the effect of blockholder on accruals-based earnings management will be less negative in non-crisis period compared to a crisis period. The second hyptotheses (H2) predicts that the effect of blockholders on accruals-based earnings management after the financial crisis will be less negative than before. The hypotheses are linked to the regression model. For both hypothesis one regression model will be sufficient. The only difference between the hypothesis is in which time period it is measured. It is expected that accruals will differ in the different time period, because of the financial crisis. To test the effect of blockholders on accruals-based earnings management the following regression is used:

𝐷𝐴 = 𝛽0 + 𝛽1 𝐵𝐿𝑂𝐶𝐾 + 𝛽2 𝐵 + 𝐵3 𝐴 + 𝐵4 𝐵𝐿𝑂𝐶𝐾 ∗ 𝐵 + 𝐵5 𝐵𝐿𝑂𝐶𝐾 ∗ 𝐴 + 𝛽6 (𝐷𝐸𝐵𝑇

_𝑅

)

+ 𝛽7 (𝑆𝐼𝑍𝐸) + 𝛽8 (𝑅𝑂𝐴) + 𝛽9 (𝐿𝑂𝑆𝑆) + 𝐵10 (𝑅𝐸𝑀) + 𝜀

3.2.1 Dependent variable

DA stands for discretionary accruals and is the only dependent variable of the regression model. Firms can use discretionary accruals to increase or decrease their net income. As most past researches used accruals-based earnings management, this study will also use discretionary accruals to calculate earnings management. By doing this, a comparison will be possible with previous studies. Total accruals exist of non-discretionary accruals and discretionary accruals. The non-discretionary accruals are the expected level of accruals for a company based on its size, operating, industry and revenue growth (Jones, 1991). The discretionary accruals are the unexpected component of total accruals. According to Geiger & North (2006) the level of discretionary accruals is a reflection of management’s use of financial reporting discretion to either increase or decrease the income.

The model that will be used in this thesis is Modified-jones Model (1995) as used by Dechow et al. (1995). First of all the original Jones Model (1991) is one of the most used models in calculating accrual-based earnings management. It corrects the weakness in model of Healy and Deangelo. The two models keep the non-discretionary accruals constant over time, but this models doesn’t. This Jones model (1991) tries to control the effect of economic changes that has on non-discretionary accruals. For example this models add total lagged asset to show the size of the company. This model also add REV (changes in revenue), in order to reveal a firms business activities. Further this model adds plant property and

(18)

equipment (PPE) to capture the long-term accruals.

In the Modified-jones model (1995), as used by Dechow et al. (1995), there is a slight change. In the original formula, Jones model (1991), net receivables is added in the revenue part. In the Modified-jones model (1995) account receivable is subtracted from total revenue. The reason is that it is easier for managers to manage earnings by exercising discretion over the recognition of revenue on credit sales than over the recognition of revenue on cash sales (Dechow et al. 1995).

The non-discretionary accruals is calculated as follow:

The given formula stands for:

NDAt = non-discretionary accruals in year t scaled by total assets in t-1 year ΔREVt = revenues in year t less revenues in year t-1 scaled by total assets at t-1

ΔRECt = net receivables in year t less net receivables in year t-1 scaled by total assets at t-1 PPEt = gross property plant and equipment in year t scaled by total assets at t-1

At-1 = total assets in t-1 year a1, a2, a3 = firm-specific parameters

The total accruals are calculated as follow:

The given formula stands for:

TAt = total accruals scales by lagged total assets

a1, a2, a3 = the estimates of a1, a2, a3 that are calculated by OLS regression Ɛt = the measurement error in the year t

Finally the discretionary accruals are calculated by subtracting the non-discretionary accruals from the total accruals.

3.2.2 Independent variables

The proxy for financial crisis is divided in two dummy variables, namely in B and A (Lins, et al., 2011). B stands for before the crisis and A stands for after the crisis. In dummy variable B 0 is for during and after the crisis, 1 is for before the crisis. In dummy variable A 0 is for before and during the crisis, 1 is for after the crisis. Lastly the interaction between blockholders and financial crisis (B and A) is set up, this will create two variable, namely Block_B and Block_A.

The academic research is very divergent related to the association between blockholders and earnings management. Iqbal et al. (2010) found a significantly negative relationship between blockholders and earnings management by using public listed firms from the United Kingdom. They suggested that a

(19)

firm who has at least 2 blockholders, has lower accruals-based earnings management than a firm who has less than 2 blockholders. Similarly, Klein (2002) found that having more than one blockholder decreases the total earnings management, accruals-based and real earnings management, in public listed firms. Both these studies gives a reason why there should be more than one blockholder in a firm to have an effect on accruals-based earnings management. The reason is that one blockholder is not able to significantly infleunce the management in a way that two or more blockholders could do. Both these studies assumed blockholders as an effective internal corporate governance mechanism to monitor management and to reduce discretionary accruals.

On the other hand Dechow et al. (1996) conclude in their research that outside blockholders improves the credibility of financial statements by controlling earnings management practices in a firm. They suggested that having only one blockholder is enough to lower the accruals-based earnings management in a firm. Although they suggested more blockholders could even have a larger impact on accruals-based earnings management. Similarly, Shleifer (2004) suggested that having one or more blockholders can have an effect on the total accruals-based earnings management. This study takes the same assumption as Iqbal et al. (2010) and Klein (2002) that having more than one blockholder lowers the level of discretionary accruals. The reason here for is that having only one blockholder, as Dechow et al. (1996) and Shleifer (2004) suggested, will not have as much as an impact as having a collaboration between blockholders. Although Iqbal et al. (2010) and Klein (2002) have a distinguish between outside and inside blockholders, this study takes blockholders as a whole. The reason here for is that the

database does not provide a specification for blockholders. Therefore the method of Iqbal et al. (2010) and Klein (2002) will be followed, assuming that if there is more than one blockholder in a firm then it will have a negative effect on accruals-based earnings management.

As both studies used a dummy variable for blockholders, this study will also apply the same method. The variable BLOCK stand for blockholders, and should test the effect of blockholders on earnings management. The proxy for blockholders (BLOCK) is a dummy variable, were 0 is a firm having 1 or less blockholder and were 1 is a firm having atleast 2 blockholders (Iqbal et al., 2010; Klein, 2002). The measurement of blockholders is done by choosing only blockholders that have more than 5% of outstanding shares. This is in line with Yang et al. (2008), who used percentages of outstanding shares held by any 5% or greater shareholder to estimate blockholder.

3.2.3 Control variables

In order to be sure that the earnings management effect is only of blockholders, there should be control variables in place. It could be that these variable have effect on blockholders or vice versa. The following control variables will be included to be sure that the impact on earnings management is only from blockholders: debt_r, size, return on asset (ROA), loss and real earnings management (REM).

The first one is violation of debt covenant. So debt will be used to proxy for a firm’s debt covenant violation. According to Press and Weintrop (1990) managers have incentive to manipulate financial figures i.e. earnings management when a firm is close to violate a debt covenant. In order to not violate this debt

(20)

covenant, managers will use earnings management (Watts and Zimmerman, 1986). In a research done by Warfield et al. (1995) and Klein (2002) both studies found a positive relation between debt and

discretionary accounting. Debt_R is measured as the ratio of debt to total asset.

Firm size is included because of the famous political cost hypothesis (Kim et al., 2003). This implies that managers of large firms have more incentives than small firms to keep earnings stable over time. This could both ways, so either push earnings upward or push earnings downward (Watts and Zimmerman, 1978). Kim et al. (2003) concludes that large size firms do more earnings management in order to meet or beat capital market and financial analysis expectations. But on the other hand large firm have higher disclosure duties this reduces information asymmetry. Also large firms have mostly effective monitoring (Meek et al., 2007). So there is an ambiguous relation between firm size and discretionary accruals. There for no expectations can be made and to calculate this the approach of Warfield et al. (1995) will be used. That is measured as the natural logarithm of the total asset of a firm.

Return on asset (ROA) is another important variable which needs to be controlled. As Meek et al. (2007) concluded that firms with higher return on asset, have higher discretionary accruals. It is possible that managers need achieve a certain percentage of return on asset to get their bonus. Return on asset is measured as the net income before extraordinary items divided by lagged total assets.

A dummy variable is included as Loss which is one when the net income if negative and zero otherwise. Loss is included as dummy variable because firm that are suffering from losses might increase earnings (Burgstahler and Dichev 1997). Also Wang (2006) finds that firms who are suffering from losses tend to use more earnings management than firms who aren’t suffering from losses. Because of this a positive relation is expected between firms that are suffering from losses and discretionary accruals. Real earnings management is included as a control variable in order to be sure that the effect of blockholders on earnings management is only the part of accruals. Real earnings management is calculated for example by reducing discretionary expenses or not investing in capital. Not taking the right investment choices, could have negative effect in the long term. Real earnings management is calculated following Cohen and Zarowin (2010) as abnormal cash flow from operations, abnormal production costs and abnormal discretionary expenses, existing of advertising expenses, R&D expenses and SG&A expenses.

3.3 Statistical test

All variables that were not normal distributed are winsorized until they were normal distributed, so the skewness and kurtosis are not extremely high anymore. The variables are winsorized firstly at 1th_{and 99}th percentile in order to remove high outliers. If after this, variables still have high outliers then they are winsorized at 5th_{and 95}th_{percentile. A linear regression is carried out in order to calculate the effect of} blockholders on accruals-based earnings management. And in order to be able to run a test for H1 and H2. A test will be carried out in order to calculate the relationship between blockholders and financial crisis. This will be done with command test, comparing variable Block with Block_A and Block_B. H1 test

(21)

the relationship between blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. It is expected that the relationship is less negative in a crisis period than in a non-crisis period. H1 is rejected when the p-value is higher than 0.05 (5%). This indicates that there is not a great difference in the relationship between blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. If the p-value of H1 if lower than 0.05 (5%), then it indicates that there is a significant difference in the relationship of blockholders and accruals-based earnings management in a crisis period compared to a non-crisis period. H2 tests the relationship between blockholders and accruals-based earnings management before and after the financial crisis. H2 expects that the relationship after the financial crisis is less negative than before the financial crisis. H2 is rejected when the p-value is higher than 0.05 (5%). This is done with command test, testing Block_B with Block_A. If the p-value is lower than 0.05 (5%), then the second hypothesis is accepted. This means that after the financial crisis the relationship between blockholders and accruals-based earnings management is less negative than before the financial crisis.

(22)

4. Results

In this chapter the result will be discussed. First of all the univariate analysis will be discussed, that exists of descriptive statistics and Pearson correlation test. In the next part of this chapter the regression will be discussed. After discussing the regression results, a test is carried out in order to give answer on the first and second hypotheses. Finally a robust regression is carried out in order to analyze how result can differ when another measurement for blockholders is used.

4.1 Univariate analysis

Table 1 presents the dependent variable, independent variable and the control variable in a descriptive statistics. The total number of firm year observations are 2275. As observed in the table, the dependent variable (DA) has a mean of -0.00 with a standard deviation of 0.09. This is in line with prior literature of Cohen et al. (2008). The minimum and maximum value of discretionary accruals is -0.19 and 0.15, respectively. Further the median is 0.01. On average the sample firms have a slightly negative discretionary accruals. This may indicate that the US firms, in this sample, are using discretionary accruals. The independent variable BLOCK is a dummy variable, and it is indicating that there exist at least two blockholder for 66% for all 2,275 firm-year observations. The standard deviation is 0.47, indicating a low spread between the numbers. This is in line with Iqbal et al. (2010) who also found a similar standard deviation of 0.50. The other independent variable is financial crisis, split in 2 categories, namely before the financial crisis (B) and after the financial crisis (A). The mean for B 0.20 is and for A is 0.40, this is in line with the given years. As it was chosen that before the crisis 2 years will be analyzed and during and after the crisis 4 years.

Control variable DEBT_R is the leverage ratio, calculated by dividing total debt by total asset. The mean is 0.13 and standard deviation is 0.14, with a median of 0.10. The minimum is 0.00 and maximum is 0.39. This indicates that the chosen sample firms have a very low leverage ratio. There are even firms that do not have debt, meaning those firms are all equity based. The maximum debt of a firm is even under a ratio of 50%, indicating that the sample firms were mostly equity invested. This leverage ratio is far lower than prior research, like Hsu and Koh (2005) found a leverage ratio of 0.54. Another noticeable result is the high mean and standard deviation of control variable Size. Size is calculated as the natural logarithm of the total asset of a firm. The mean is 6.14, while all other variables means are between -0.00 and 0.66. The standard deviation is 2.24, while all other variables means are between 0.09 and 0.49. This high standard deviation and mean of control variable Size may indicate that the firms sample sizes are very divergent. The median is 6.06, with a lower quartile of 4.52 and a higher quartile of 7.54. The minimum is -5.81 and the maximum is 12.35, indicating a large spread between the sizes of the firms.

The control variable ROA has a mean of 0.01 and a standard deviation of 0.14, with a median of 0.05. Noticeable here is, that there is quite a difference between firms in their return on asset. Because the minimum is -0.29 and the maximum is 0.17. Although the mean is 0.01, this strongly indicates that

(23)

Variable N Mean S.D. Minimum .25 Median .75 Maximum DA 2275 -0.00 0.09 -0.19 -0.05 0.01 0.06 0.15 Block 2275 0.66 0.47 0.00 0.00 1.00 1.00 1.00 B 2275 0.20 0.40 0.00 0.00 0.00 0.00 1.00 A 2275 0.40 0.49 0.00 0.00 0.00 1.00 1.00 Debt_R 2275 0.13 0.14 0.00 0.00 0.10 0.23 0.39 Size 2275 6.14 2.24 -5.81 4.52 6.06 7.54 12.35 ROA 2275 0.01 0.14 -0.29 -0.04 0.05 0.10 0.17 Loss 2275 0.32 0.47 0.00 0.00 0.00 1.00 1.00 REM 2275 0.12 1.74 -3.23 -0.62 0.18 1.26 2.89

there are some strong outliers. Thus there are firms with a good ROA and there are firms with a very bad ROA. The control variables Loss has a mean of 0.32 and a standard deviation of 0.47, with a median of 0.00. The control variable Loss indicates that 32% of the 2,275 firm-years observations had a net income of 0 or lower. The control variable REM indicates a strong difference compared with the dependent variable DA. REM has a mean of 0.12 and a standard deviation of 1.74, with a median of 0.18. The minimum of REM is -3.23 and the maximum is 2.89. This shows there is far more variation in real earnings management than in discretionary accruals. This can be seen by the great variation of standard deviation of real earnings management and discretionary accruals, namely there is a difference of 1.63.

Table 2 gives an overview of Pearson correlation test, correlation coefficient with an asterisk are significant at the level of 0.05 (5%). The coefficient is a statistical measure, measuring the strength of a linear correlation between the variables. The values are between -1 and 1. When the value is between 0 and -1 then it indicates a negative relationship between the variables. On the other hand when the value is between 0 and 1 then it indicates a positive relationship between the variables.

Discretionary accrual is significant related to blockholders, with a significant level more than 10%, namely 0.1380. This indicates that having more discretionary accruals, requires more blockholders. Discretionary accruals has a negative relationship with dummy variable A and B. Indicating that during the crisis there are more discretionary accruals and after the crisis there are less discretionary accruals. Also it is noticeable that discretionary accruals are not significant with real earnings management. This

indicates that discretionary accruals and real earnings management are likely used as substitutes, as Graham et al. (2005) suggested. Discretionary accruals is significant with all other variables, indicating a positive or negative relationship. There is a positive significant relationship with control variable Size and ROA. This means firms that are larger and have a higher return on assets, likely have more discretionary accruals. There is a negative significant relationship with control variable Debt_R and Loss. This means that firms that have higher debt and are having losses, likely have less discretionary accruals.

The independent variable blockholders has a significant relationship with the financial crisis dummies. Dummy variable Block has a significant negative relationship with dummy variable B, indicating that there are less blockholders required. After the financial crisis there is a positive relationship, indicating that there a more blockholders required. This is in line with Xu et al. (2013), they suggested that after the financial crisis there is a need for more blockholders. Further blockholders has a significant negative Table 1, Descriptive statistics

(24)

DA Block B A Debt_R Size ROA Loss REM DA 1.0000 Block 0.1380* 1.0000 B -0.0010 -0.1218* 1.0000 A -0.0021 0.0935* -0.4090* 1.0000 Debt_R -0.0513* 0.2210* -0.0214 0.0358 1.0000 Size 0.2243* 0.4749* -0.0662* 0.0679* 0.3794* 1.0000 ROA 0.6822* 0.3553* 0.0471* 0.0025 0.0307 0.4323* 1.0000 Loss -0.5697* -0.3277* -0.0505* -0.0179 -0.0749* -0.4029* -0.8341* 1.0000 REM 0.0368 0.0084 -0.0607* -0.0042 0.0971* -0.0055 0.0203 -0.0126 1.0000 relationship with control variable Loss. This is indicating firms with losses have proportional less blockholders than firms with no losses. Blockholders are significant positive related to control variable Size. This is indicating that how larger the firm is, how more blockholders are needed in the firm. The same goes for control variable ROA and Debt_R. So how more debt and return on asset in a firm is, the more blockholders are in a firm. The dummy variable B has a significant positive relationship with control variable ROA. This means before the financial crisis there was a lot of emphasis on return on asset. Further the dummy variable B has a significant negative relationship with control variable Size, Loss and REM. This mean before the financial crisis there were already firms with losses and their real earnings management were high. The dummy variable A has a positive significant relationship with control variable Size. This means that after the financial crisis the size of the firms got larger compared to before and during the financial crisis.

As for the control variable Debt_R, it has a significant positive relationship with Size. This is likely indicating that firms that have higher debt have higher total asset. Also a noticeable significant negative relationship is with Loss. This is likely indicating that firms with more debt have less losses. Control variable Debt_R has a positive significant relationship with control variable REM. Indicating firms with higher debt have higher real earnings management. This is logical, because firms with high debt have debt covenant. In order not to break this debt covenant, many firms use real earnings management, because these are difficult to detect (Graham et al., 2005). The control variable Size is positive related to control variable ROA, indicating that larger firms have higher return on their asset. Also that larger firm’s places emphasis on return on asset. The control variable Size has a significant negative relationship with control variable Loss. Control variable ROA has a significant negative relationship with control variable Loss. This is likely indicating that firms who are having a loss, will likely place less emphasis on return on asset. Lastly the control variable Loss has a negative relationship with REM. This is indicating that firms who are having losses are using real earnings management to cover those losses. Control variable ROA has a significant negative relationship with control variable Loss. This is indicating firms who have losses have lower return on asset. This is in line with prior research of Warfield et al. (1995) who suggested that firms who are having losses do not put their emphasis on return on asset.

(25)

4.2 Regression analysis

Following the result the regression model will be explained, table 3 shows the regression results. First of all the adjusted R-squared is 0.4821, this is indicating that almost 50% of the regression analysis is explained by the model. This means that the regression model is well set up in order to explain what influences the discretionary accruals. As observed in table 3, the relationship between discretionary accruals and blockholders has a coefficient of -0.02. The p-value is highly significant, namely 0.000, indicating a significant negative relationship between discretionary accruals and blockholders. The controversial view that exist about blockholders is further enhanced by this study, indicating a negative relationship between blockholders and accruals-based earnings management. I find that this effect is economically significant, because if one blockholder increase in a firm then the total discretionary accruals will decrease with 2% in the firm. The coefficient of dummy variable B is 0.015 and of dummy variable A is 0.003 with a t-value of -2.74 and -0.69, respectively. The p-value of B is 0.006 and of A is 0.046. This indicate that A and B are highly significant. Although the main effects in the regression model are significant, the interaction variable will decide whether both hypothesis are supported. Block_B has a coefficient of -0.005 and t- value of 0.77. Block_A has a coefficient of -0.001 with a t-value of 0.30. Neither interactions variables are significant, Block_B has p-value of 0.439 and Block_A has a p-value of 0.765. In order to know whether the hypothesis are supported, a couple of test are conducted after the explanation of the other variables.

The coefficient of the control variable Debt_R is significantly negative with a p-value of 0.009 and a coefficient of -0.028. This indicate that firms with a high level of debt have low discretionary accruals. This finding is not in line with prior research done by Warfield et al. (1995). One explanation could be that firms with higher debt have higher scrutiny in order not to exceed their debt covenant. By having higher scrutiny it is more difficult for managers to manipulate accruals. There for it is likely that firms with debt covenant have less discretionary accruals. I find that this effect is economically significant, because if debt ratio increases with one then the total discretionary accruals decrease 2.8%. The control variable Size has a coefficient of -0.001 and a p-value of 0.122. This indicate not a significant effect between discretionary accruals and the size of a firm. Therefore it is not necessary that larger firms have higher discretionary accruals. The coefficient of the control variable ROA is significant positive with a p-value of 0.000 and a coefficient of 0.465. This indicate a high positive effect between discretionary accruals and return on asset. This means firms that have higher return on asset, also have higher discretionary accruals. This is in line with prior research of Meek et al. (2007) were they concluded that firms with higher return on asset, have higher discretionary accruals.

The control variable Loss has a coefficient of -0.005 and a p-value of 0.318. This indicate not a significant effect, indicating there is not a relationship of firms who are having losses that they also have higher discretionary accruals. This is not in line with prior research, were Wang (2006) provide evidence that firms who are having losses have more accruals-based earnings management. One explanation could be that, that research was done on firms who are family invested. As Dou et al. (2016) explained,

(26)

Table 5, test H1 (1st_part) Table 4, test H1 (2nd_part) Number of observations 2275 F (10, 2264) 212.71 Prob > F 0.0000 R-squared 0.4844 Adj R-squared 0.4821 Root MSE 0.0635 Source SS df MS Model 8.599 10 0.8599 Residual 9.153 2264 0.0040 Total 17.752 2274 0.0078

DA Coefficient Standard error T-value P-value

Block -0.0219 0.0047 -4.64 0.000 -0.0311 -0.0126 B -0.0157 0.0057 -2.74 0.006 -0.0207 -0.0044 A -0.0037 0.0054 -0.69 0.489 -0.0143 0.0068 Block_B 0.0057 0.0074 0.77 0.439 -0.0089 0.0204 Block_A 0.0019 0.0065 0.30 0.765 -0.0108 0.0147 Debt_R -0.0280 0.0108 -2.60 0.009 -0.0492 -0.0068 Size -0.0011 0.0007 -1.55 0.122 -0.0027 0.0003 ROA 0.4656 0.0184 25.17 0.000 0.4293 0.5019 Loss -0.0052 0.0052 1.00 0.318 -0.0154 0.0050 REM 0.0012 0.0007 1.57 0.118 -0.0003 0.0027 _cons 0.0264 0.0056 4.71 0.000 0.0154 0.0374 95 % Confidence Interval Block = Block_A F(1, 2264) = 5.56 Prob > F = 0.0184 Block = Block_B F(1, 2264) = 6.40 Prob > F = 0.0115

blockholders are heterogeneous and there for each blockholder has different incentives to manage earnings. I find this effect not economically significant, because if a firm is having a loss then the discretionary accruals only decreases with 0.5%. A noticeable effect is if of real earnings management, REM has a coefficient of 0.012 and is not significant with a p-value of 0.118. This indicates that real earning management is not treated as complementary, but rather as a substitute of discretionary accruals. This is in line with prior research of Graham et al. (2005), were they provide evidence that real earnings management will be used as substitutes of discretionary accruals. The constant variable has a coefficient of 0.02 and a p-value of 0.000.

In order to give answer on both of the hypothesis a couple of test are conducted. For the first hypothesis a test is conduced to measure the relationship of Block versus Block_B and Block versus Block_A, so crisis period versus non-crisis period. Below in table 4 and table 5 the result for the first hypothesis are

presented. As it is noticed both p-value are significant, namely 0.0115 and 0.0184. This means that the first hypothesis, namely that the relationship between blockholders and accruals-based earnings management is less negative in a crisis period compared to a non-crisis period, is accepted. So the assumption, following Zhong et al. (2007), Ely and Song (2000) and Cimini (2015), that blockholders put more pressure on managers during a financial crisis to increase earnings is confirmed by this study.

For the second hypothesis a test is conducted to measure the relationship Block_B versus Block_A. So after the financial crisis period versus before the financial crisis period. The second hypotheses result is Table 3, Linear regression

Effect of blockholders on accruals-based earnings management before, during and after the financial crisis