1
Family firms and the extent of
earnings management
Name: Daphne Brandsma Student number: 10640584 Thesis supervisor: Dr. A. Sikalidis Date: June 25, 2018
Word count: 11761
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
2
Statement of Originality
This document is written by student Daphne Brandsma 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 family firms and earnings management. Family firms have relatively less type I agency costs, known as the alignment effect, and relatively more type II agency costs, known as the entrenchment effect. As a consequence of the alignment effect, family firms have less information asymmetry, more effective direct monitoring, a longer time horizon, a better reputation and a lower need for corporate
governance. However, the entrenchment effect leads to the expropriation of non-controlling shareholders. To determine the extent of earnings management, a distinction is made between real earnings manipulation and accrual based earnings management. The findings show a negative relationship between family firms and real activities manipulation.
However, unlike prior research, the results show a positive relationship between family firms and accrual based earnings management. This study contributes to the literature by
including the impact of auditor's independence on the extent of earnings management. The amount of non-audit fees provided by the auditor firm is used as a proxy for the auditor's independence. However, the findings show no evidence that the auditor's independence has an influence on the difference of earnings management between family and non-family firms.
4
Table of content
1. Introduction ... 6 2. Literature review ... 8 2.1 Family ownership ... 8 2.2 Agency theory... 92.2.1 Agency theory applied to firms ... 10
2.3 Agency theory applied to family firms ... 10
2.3.1 Type I agency problems ... 11
2.3.2 Type II Agency costs ... 11
2.4 Consequences of agency costs ... 11
2.5 Alignment effect ... 11
2.5.1 Information asymmetry and direct monitoring ... 12
2.5.3 Corporate governance ... 14
2.6 Entrenchment effect ... 14
2.6.1 Expropriation of non-controlling shareholders ... 14
2.6.2 Greater information asymmetry ... 15
2.6.3 Inferior corporate governance ... 15
2.8 Earnings management ... 15
2.8.1 Accrual based earnings management and real earnings manipulation ... 16
2.10 Hypothesis ... 17
3. Data and method ... 19
3.1 Sample selection ... 19
3.2 Dependent variable - earnings management ... 19
3.2.1 Discretionary accruals ... 20
3.2.2 Real activities manipulation ... 20
3.3 Independent variable - family firms and non-audit services... 23
3.4 Control variables ... 24 3.5 Regression model ... 24 4. Results ... 26 4.1 Descriptive statistics ... 26 4.2 Correlation matrix ... 28 4.3 Multivariate analysis ... 31
4.3.1 Effect of family ownership on earnings management ... 31
5
5 Conclusion ... 36 References ... 38 Appendix A ... 42
6
1. Introduction
Family firms are an important factor in the world economy (Wang, 2006). Standard & Poor's 500 firms from 1992 through 1999 show that family firms are an important class of investors (Anderson & Reeb, 2003). Anderson and Reeb found that family firms constitute of more than 35 percent of the S&P 500 industrials and that families own, on average, nearly 18 percent of their firms' outstanding equity (2003). Based on these numbers, it can be concluded that family ownership is an important ownership structure.
In addition to having an important position in the world economy, family ownership has some specific characteristics. For example, Vural (2018) states that family firms are special because of the information advantages, the long-term presence of the families in the firm, their close ties with the management and their high involvement in firm decisions. Also, Anderson and Reeb (2003) argue that, as a consequence of these characteristics, founding families are different because they are in a good position to control the firm.
This study discusses the defining characteristics of family ownership based on the agency theory. The agency theory consists of a contract in which one person (the principal) engages another person (the agent) to work on behalf of the principal. Consequently, this leads to the delegation of the decision making authority from the principal to the agent (Jensen and Meckling, 1973). However, according to Chrisman et al. (2004), this contract causes two problems: adverse selection and moral hazard. Adverse selection is a consequence of conflicting interests, whereas moral hazard is caused by information asymmetry. According to Gilson and Gordon (2003), there are two types of agency problems: Type I agency problems, which are caused by a distinction between shareholders and management and type II agency problems, which are caused by a distinction between different
shareholders. Family firms face relatively less type I agency problems because they have less severe hidden-information and hidden-action agency problems. This is known as the
alignment effect and leads to lower information asymmetry, more effective direct monitoring, a longer time horizon, a better reputation and a lower need for corporate governance within family firms. On the other hand, family firms face more type II agency problems because of the separation of controlling and non-controlling shareholders. This is
7
called the entrenchment effect and leads to the expropriation of non-controlling shareholders.
Both effects have an impact on the earnings quality of family firms. Because of the previously mentioned importance and specific characteristics of family ownership, it is important to know how these characteristics influence the earnings. In this way, the stakeholders are able to get a true and fair view of the firm's financial statements and therefore they can make well-founded decisions about their (potential) investments. To determine this, this study looks at the amount of earnings management within family firms and makes a distinction between accrual based earnings management and real activities manipulation (REM). This leads to my research question:
'To what extent does the amount of earnings management differ between family and non-family firms?'
I investigate this based on the database from Anderson, Duru and Reeb (2009) and
Anderson, Reeb and Zha (2012). This database consists of 16,200 firm-year observations for the top-2000 largest firms from 2001 through 2010. The database works with a dummy variable which is 1 when the family owns (or votes) a 5% or larger stake and 0 otherwise. I matched this database with the data of North America (US and Canada) from Compustat and obtained a sample of 986 firms of which 342 are family owned and 642 have no family ownership.
This study contributes to the literature because it also looks at the degree of auditor's independence. Previous research has been done on the relationship between family firms and earnings management but they have never taken the degree of auditor's independence into account.
Consistent with prior research (Jiraporand & Dadalt, 2009 and Achleitner et al., 2014), I find evidence that there is a negative relationship between family ownership and real earnings manipulation. However, I find no supportive evidence that there is a negative relationship between family ownership and accrual based earnings management. I also find no evidence that the degree of auditor's independence has an impact on this.
The remainder of this paper is structured as follows: In paragraph two, I will discuss prior research and develop my hypotheses. Thereafter, in paragraph three, I will give my research
8
design and in paragraph four my results. Finally, I will end in paragraph five with my conclusion.
2. Literature review 2.1 Family ownership
Family ownership is an important ownership structure. Among the Standard and Poor's (S&P) 500 companies, which are most unlikely to be family owned, 35% of the companies have an active involvement of founding family members. These families own, on average, nearly 18 percent of their firms' outstanding equity (Anderson & Reeb, 2003).
Vural (2018) states that families are an important group of shareholders to understand because a lot of listed firms around the world are family owned. The information
advantages, the long-term presence of the families in the firm, the close ties of the family with the management and their high involvement in firm decisions are, according to Vural, the characteristics that make family firms relevant when studying financial disclosure practices. He states that the specific characteristics of family firms indicate that family owners might have different preferences for accounting information than other shareholders.
This is in accordance with Anderson and Reeb (2003), who argue that founding families are in a good position to control the firm. They reason that families are able to take this position because of the number of shares they own, the poorly diversified portfolios and the
opportunity to control senior management positions. They also give the multiple generations of families involved in the firms as reason because this makes families more focused on the long-term.
According to Astrachan and Shanker (2003), there is no accurate information about family firms available because of their private nature. They find it even harder to give a concise, universal and measurable definition of a family firm. In their opinion, there are a lot of different criteria used to make a distinction between family and non-family firms. Examples of these criteria are strategic control, the percentage of ownership, the intention of the firms to remain in the family and the involvement of multiple generations. Astrachan and Shanker (2003) give in their research three different family firms definitions from a broad,
9
inclusive definition to a narrow and more exclusive one. Their level of inclusiveness depends on the degree of family involvement in the firm.
Their broad definition is the most inclusive one and has only one requirement. Which is that some kind of family participation in the firm exists and that the family has control over the firms' strategic decisions. Their middle definition narrows the field by the requirements that the family owner has the intention to pass the firm to another family member and that the founder plays a role in running the firm. Finally, their last definition is the narrowest and requires that multiple generations have a significant impact on the firm (Astrachan & Shanker, 2003).
2.2 Agency theory
A typical characteristic of publicly traded companies is the separation of ownership from management. This separation could lead to agency problems between managers and outside shareholders (Jensen & Meckling, 1976). According to Eisenhardt (1989), the purpose of the agency theory is to resolve the problems that occur in an agency relationship. Jensen and Meckling (1976) describe an agency relationship as a contract in which one person (the principal) engages another person (the agent) to work on behalf of the principal.
Consequently, this contract leads to the delegation of the decision making authority from the principal to the agent. Eisenhardt (1989) states that this relationship leads to agency problems when the interests of the principal and agent are conflicting or when it is difficult or expensive for the principal to observe the actions of the agent. Jensen and Meckling (1976) are of the opinion that if both the principal and agent try to maximize their utility, there is a possibility that the agent does not always behave in the best interest of the principal. This leads to an increase of the residual loss, which Jensen and Meckling define as ''the dollar equivalent of the reduction in welfare experienced by the principal due to this divergence''.
The difficulty and expensiveness for the principal to observe the actions of the agent are caused by information asymmetry between them. Information asymmetry means that the agent has more information available about the day to day operations than shareholders. According to Chrisman et al. (2004), this asymmetric information divides the agency
10
when the principal unintentionally makes a contract with an agent who is less able, ethical or committed or whose interests are less congruent than the principal expected when he hired the agent. Moral hazard happens after signing the contract and means that the agent performs or does not perform actions which are in the interest of the agent but are disadvantageous to the principal.
In summary, agency problems are the result of a relationship between a principal and agent and are caused by different utility functions and information asymmetry. Agency costs arise when the agents' actions are not in line with the interests of the principal and when the principal has to make expenses to control the behaviour of the agent (Chrisman et al., 2004).
2.2.1 Agency theory applied to firms
The agency theory can be applied to firms because of the division between the owners (the principals) and managers (the agents). There are different utility functions because the actions of the management diverge from the actions that maximize shareholders return. Also, the managers have more information available about the firm than the shareholders which leads to information asymmetry. As a consequence, there will be a residual loss, that is the extent to which the returns to the owners fall below what they would be if the owners exercised direct control of the company (Donaldson & Davis, 1991).
2.3 Agency theory applied to family firms
Gilson and Gordon (2003) argue that the agency problems in a family firm can be divided into two types. The type I problem is caused by the distinction between the shareholders and management and the type II problem is caused by the distinction between different shareholders.
Family firms face relatively less severe agency problems caused by the separation of
ownership and management (type I agency problems) than non-family firms. The separation of ownership from management may lead to managers not acting in the best interest of the shareholders. However, family firms face more severe agency problems between controlling and non-controlling shareholders (type II agency problems). Controlling shareholders may seek private benefits at the expense of the non-controlling shareholders (Gilson & Gordon, 2003). The differences between the two agency problems will be discussed in the next paragraphs.
11 2.3.1 Type I agency problems
Family firms face less severe type I agency problems than non-family firms. Ali et al. (2007) give three reasons for this. Firstly, they argue that families are able to directly overview the management because of the undiversified and concentrated equity position they hold in their firms. Secondly, families are able to provide superior monitoring of managers because of their knowledge about the firms' activities. The last reason they give is the longer
investment horizons families have compared to that of other companies. Anderson and Reeb (2003) show that this longer investment horizon is caused by the long-term presence of founding families in the firms. A longer time horizon helps family firms to reduce short-sighted investment decisions of the managers. In summary, family firms face, compared to non-family firms, less severe type I agency problems due to the separation of ownership and management.
2.3.2 Type II Agency costs
On the other hand, family firms tend to have more type II agency problems than non-family firms. Ali et al. (2003) argue that this is caused by the concentrated equity holding of families in their firms, the domination of the families on the board of director's membership and their voting rights exceeding their cash flow rights. This control gives them power to seek private benefits at the expense of other non-family shareholders. This is in accordance with Dal Magro et al. (2017), who argue that family firms have more type II agency costs due to the separation between controlling and non-controlling shareholders. They reason that family owners can worsen the type II agency conflict because family owners might try to expropriate capital from minority shareholders.
2.4 Consequences of agency costs
The relatively lower type I agency costs and relatively higher type II agency affects the earnings quality and the extent of earnings management. This is explained by the alignment and entrenchment effect.
2.5 Alignment effect
Wang (2006) argues that the lower type I agency costs within family firms is known as the alignment effect. This effect is based on the argument that the interest of founding families and other shareholders are better aligned because of the long-term presence of the family
12
members and their amount of stock owned. Therefore, the alignment effect predicts that founding families are less likely to expropriate wealth from other shareholders by managing earnings. The alignment effect leads to a lower information asymmetry, more effective direct monitoring, longer time horizon, better reputation and a lower need for corporate governance. These consequences are discussed in the next paragraphs.
2.5.1 Information asymmetry and direct monitoring
As discussed before, the distinction between management and ownership leads to
information asymmetry. Howorth, Westhead and Wright (2004) state that family firms have a closer relationship and higher level of trust between management and shareholders, which makes the information more effectively shared and communicated with the family.
Therefore, they argue that family firms face relatively less information asymmetry than non-family firms. According to Anderson and Reeb (2003), the lower information asymmetry leads to better direct monitoring. They reason that the knowledge of families about their firms' activities enables them to provide greater and more effective monitoring of the management. According to Wang (2006), families also have stronger incentives to do so, because the wealth of the founding families is closely tied to the value of the firm.
According to Ali et al. (2003), the more effective monitoring leads to a better ground on which the rewards of the management can be based. Non-family firms are more likely to base the rewards of the management on observable earnings-based performance measures. Families, on the other hand, reward their management based on information about
managers' activities obtained through direct and more effective monitoring. This makes the compensation of managers less likely to be dependent on earnings, which gives them less
incentives to manipulate earnings.
2.5.2 Time horizon and reputation
Family firms are more focused on the long term because of their incentives to ensure transgenerational sustainability and the long-term presence of founding families in the firm (Achleitner et al., 2014). The longer time horizon makes it less likely that family firms engage in activities of earnings management that use up resources in the long run. According to Graham et al. (2005), an example of these activities are REM activities. REM activities affect the long term because managers are willing to sacrifice future cash flows for higher current
13
period income, so they have a negative impact on the firm's future operating performances. Archleitner et al. (2014) found that the use of REM activities is lower for family firms. This is the opposite of managers with objectives in the short-run because they may report earnings that maximize their personal wealth instead of focussing on the future.
The theory about the longer time horizon is consistent with Ali et al. (2003), who argue that founding families are more likely to decline short-term advantages from earnings
manipulation because of their incentives to pass the firm to future generations and to protect the reputation of the family. They conclude that family firms are less likely to engage in opportunistic behaviour because it could harm the family's long-term firm performance, wealth and reputation. Anderson and Reeb (2003) emphasize the importance of the reputation of family firms. They argue that the sustained presence of the families suggests that external parties, like providers of capitals or suppliers, are more likely to deal with the same governing bodies for a longer period. As a consequence, the family firm's reputation is more likely to create longer-lasting economic consequences compared to non-family firms, in which managers and directors change on a more continuous basis.
Moreover, Fan and Wong (2002) state that families are, as a consequence of the ownership concentration within their firms, willing to make a credible commitment that it is building a reputation for not expropriating minority shareholders. This commitment is credible because minority shareholders are aware that if a family owner unexpectedly extracts private
benefits by expropriating minority shareholders, they will discount the stock price accordingly and the family owners share value will reduce. In equilibrium, a family
shareholder holds a large ownership stake and their stock price of the company is higher.
According to Diéguez-Soto et al. (2017), the focus of the families on their reputation can be explained by three main theoretical frameworks. The first one, called the familiness view, suggests that family firms are more concerned about their reputation because of the more trustworthiness long-term relationships within the firm and the relationships with clients and other stakeholders. The second one is an organizational identity theory and proposes that the higher the influence is on the identity between the family and the firm, the more the family cares about their reputation. The last framework combines previous views and
14
emphasizes the pursuit of the desirable reputation so that the family goals are also achieved in the future.
Concluded, as a consequence of the longer time horizon and the emphasis on the reputation, less earnings manipulation in family firms is expected.
2.5.3 Corporate governance
Klein et al. (2005) state that the lower type I agency costs result in a lower need for a strong corporate governance. The intention of corporate governance is to reduce the misalignment of interests between shareholders and managers. However, because the interest of
shareholders and managers are better aligned in family firms, there is a lower need for outside directors. Outside directors with less financial interest and less knowledge about the firm may even be more inclined to focus on short-run goals and decrease the company’s efficiency.
2.6 Entrenchment effect
Wang (2002) argues that the higher type II agency costs are known as the entrenchment effect. He states that the entrenchment effect is based on the idea that concentrated ownership gives incentives for controlling shareholders to expropriate wealth from smaller shareholders. Furthermore, the higher type II agency costs have a negative impact on the information asymmetry between the different shareholders and the corporate governance. These consequences will be discussed in the next paragraphs.
2.6.1 Expropriation of non-controlling shareholders
Wang (2002) states that the higher type II Agency costs lead to incentives for the family members to expropriate wealth from smaller, non-controlling shareholders. They do this by managing the earnings. For example, Gugler and Yurtoglu (2003) found that larger
shareholders reduce the dividend-payout ratio to get financial resources which they can invest in positive net present value long-term projects. Also, Harris et al. (1994) showed that managers have incentives to report lower earnings because higher reported earnings create pressures for shareholders to demand more dividend. Moreover, DeAngelo and DeAngelo (2000) found that in 1994 the management team of Time Mirror Company radically cut dividends to non-family shareholders but maintained a special dividend for the Chandler family. Prior literature shows that larger shareholders, such as family members, ensure that
15
management serves their interests. So families may pursue actions that maximize their own utility but lead to suboptimal policies resulting in poor firm performance compared to non-family firms (Anderson and Reeb, 2003).
2.6.2 Greater information asymmetry
It has been shown that the alignment effect leads to less information asymmetry between shareholders and management. However, the higher type II agency costs lead to more information asymmetry between the founding families and non-family shareholders. For example, Fan and Wong (2002) found that the earnings in firms with less concentrated ownership in East Asian countries are more informative. They show that concentrated ownership reduces information outflow to outside investors. As a consequence, family members are able to hide the earnings manipulation from non-family shareholders. This gives them, next to the incentives extract private benefits, the possibility to do so (Fan and Wong, 2002)
2.6.3 Inferior corporate governance
It has been concluded that the alignment effect of family firms causes a lower need for corporate governance. However, Fan and Wong (2002) found that the agency conflicts between family and non-family shareholders lead to an inferior corporate governance. Because family members often hold important positions both on the board and on the management team, the monitoring by the board may be ineffective. This ineffective
monitoring enables families to control and influence the firm and makes it easier for them to engage in earnings manipulation (Ali et al., 2003).
2.8 Earnings management
According to the literature review, family firms could have a negative or positive impact on the extent of earnings management. Schipper (1989, p.92) defines earnings management as ''the purpose intervention in the external financial reporting process, with the intent of obtaining some private gain''. According to Healy & Wahlen (1999), earnings management occurs when the management uses judgment in financial reporting to modify the financial statements to either misguide stakeholders about the underlying economic performance of the firm or to influence contractual outcomes that are dependent on the reported
16
2.8.1 Accrual based earnings management and real earnings manipulation
According to Prencipe et al. (2008), there are two ways managers can manipulate earnings; accrual based earnings management and real earnings manipulation. Accrual based earnings management is the manipulation of accruals, which are the items of the income statement (revenues and expenses) that have not been translated into cash flows. In this case,
managers manipulate earnings by taking advantage of the flexibility allowed by the applied accounting standards and/or regulation in recording transactions or in measuring the value of liabilities and assets. Managers can use these flexibility to reach specific earnings targets.
Real activities manipulation means that the management carries out special transactions or cut expenditures (Prencipe et al., 2008). More specifically, Roychowdhury (2006, p. 337) defines real activities manipulation 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'.' He gives three methods to do this. The first one is sales manipulation, which are the attempts of managers to temporarily increase sales during the year by offering more lenient credit terms or price discounts. The additional sales lead to higher total earnings in the current period. However, the price discounts cause lower margins in the future and this can lead to abnormally high production costs relative to sales. The second method is the reduction of discretionary expenditures, such as advertising and R&D, to meet earnings targets. This leads to lower expenditure cash outflows and results therefore in a higher abnormal CFO in the current period. However, the reduction of discretionary expenditures can decrease the cash flows in the future when there are no new products developed. The last method is overproduction, which means that managers produce more goods than necessary to meet expected demand and to manage earnings upwards. With higher production levels, fixed overhead costs are spread over a larger number of units and fixed costs per unit decrease. This lowers the COGS and leads therefore to better operating margins. However, the overproduction causes higher production and holding costs. These costs will occur in the future, leading to higher earnings in the current period, but lower earnings in the future (Roychowdhury, 2006).
17 2.9 Auditor independence
Next to investigating the extent of earnings management in family firms, this study
investigates how the independence of the auditor impacts this. Auditor independence is a fundamental principle of the accounting profession. Though, there is no uniform definition of auditor independence. Warren and Elzola (p.44, 2009) describe auditor independence as ''the delivery of objective decisions that align with professional obligations to the public'' . Auditors should be independent in mind and in appearance. Quick and Warming-Rasmussen (2009, p. 142) describe independence in mind as ''the state of mind that permits the
provision of an opinion without being affected by influences that compromise professional judgment, allowing an individual to act with integrity, and exercise objectivity and
professional skepticism''. Independence in appearance is described as ''the avoidance of facts and circumstances that are so significant that a reasonable and informed third party would reasonably conclude that a firm’s integrity, objectivity or professional skepticism had been compromised'' (Quick & Warming-Rasmussen, 2009, pp.142). So independence in mind is the factual independence and independence in appearance is how the society perceives this.DeAngelo (1981) states that the audit quality can be affected by impaired auditor independence. She describes auditor quality as the probability that an auditor discovers a breach in the client's accounting system and also reports this breach. The
probability that the auditor will report the breach, is dependent on the independence from a given client. So, when the auditor is less independent, the financial statements become less reliable.
2.10 Hypothesis
The separation of ownership and management within family firms leads to lower type I agency costs -the alignment effect- and higher type II costs -the entrenchment effect. The alignment effect is based on the idea that family firms are willing to report earnings in good faith and are therefore less likely to manage earnings. On the other hand, the entrenchment effect stimulates families to expropriate wealth from non-family shareholders by
manipulating the earnings. So, based on the alignment and entrenchment effect, there are different expectations about the relationship between family firms and earnings
management. The alignment effect predicts a negative and the entrenchment effect a positive relationship (Wang, 2006). Most prior research shows a negative relationship
18
between family firms and earnings management. For example, Jiraporn and DaDalt (2009) found that family firms engage in relatively less earnings management than non-family firms. They give the reason that family firms have a more concentrated ownership, more long-term investors and invest more in their reputation. This is consistent with Achleitner et al. (2014), who found a negative association between family firms and the two types of earnings management. They argue that family firms are more focused on the long-term value of their firm, causing them to be more likely to manage reported earnings downward. Also, they state that family firms are less likely to engage in real earnings manipulation because of their concern about the transgenerational sustainability of the firm.
This leads to the following hypotheses:
H1: There is a negative relationship between family firm and accrual based earnings management.
H2: There is a negative relationship between family firm and real earnings manipulation.
Prior research shows a negative relationship between auditor independence and earnings management. For example, Frankel et al. (2002) found that auditors who are less
independent are less likely to report a small earnings surprise. They also report a lower extent of discretionary accruals and lower extent of both increasing and income-decreasing discretionary accruals. According to Frankel et al. (2002), when an auditor is less independent, he will have a stronger economic bond with the client. This increases the auditor's incentive to obey to client pressure, including the pressure to allow earnings management.
The predicted negative relationship between family firms and earnings management could have an influence on the independency of the auditor. For example, if an auditor knows that family firms engage in less earnings management and has therefore higher earnings quality, he might act less independent because he expects that investors will look less carefully at the financial statements. On the other hand, when family firms engage in more earnings management, the auditor might want to act more independent to compensate for the negative impact of the earnings management. In this way he tries to make the financial statements more reliable.
19
This study expects less earnings management in family firms, so my next hypothesis is:
H3: The negative relationship between family ownership and earnings management is less pronounced for firms where the auditors are less independent.
3. Data and method 3.1 Sample selection
The research question will be answered through database research. The data used to determine family business is from the database of Anderson, Duru and Reeb (2009) and Anderson, Reeb, and Zha (2012). This database consists of 16,200 firm-year observations for the top-2000 largest firms from 2001 through 2010. The database works with a dummy variable which is 1 when the family owns (or votes) a 5% or larger stake and 0 otherwise. The data to determine earnings management is from Wharton Research Data Services (WRDS) and is gathered from the database Compustat in North America (US and Canada) from 2001 through 2010. Because the database from Anderson, Duru and Reeb (2009) and Anderson, Reeb, and Zha (2012) is from 2001 through 2010, I also make my sample from 2001 through 2010.
After matching the family firm data with the Compustat data, I obtain an initial sample of 2,212 firm-year observations from 2001 to 2010. Following Achleitner et al. (2014), I then remove all the financial firms, such as banks and insurances because these firms have unique operating characteristics and are governed by specific regulations. After that, I require the availability of firm characteristics for the earnings management test. So, there has to be data available for the control and independent variables I am going to use. With these restrictions I generate a sample of 986 observations: 342 observations for family firms and 642 for non-family firms.
3.2 Dependent variable - earnings management
As described in the literature review, earnings management is ''the purpose intervention in the external financial reporting process, with the intent of obtaining some private gain'' (Schipper, 1989, p.92). There are two ways in which managers can engage in earnings management: by accrual based earnings management and real activities earnings management. For this reason, I use two proxies for earnings management.
20 3.2.1 Discretionary accruals
The proxy for earnings management are discretionary accruals. Accrual based earnings management is the manipulation of accruals, which are the income statement items (revenues and expenses) that have not been translated into cash flows (Prencipe et al., 2008). Dechow et al. (1995) report that the modified Jones model is the best model to calculate the extent of discretionary accruals. They state this because the Jones model relaxes the assumption that non-discretionary accruals are constant. Also, the model tries to control for the influence of changes in a firm's economic circumstances on non-discretionary accruals. Furthermore, Jones designed a modification to eliminate the presumed tendency of the Jones Model to measure discretionary accruals with an error when discretion is exercised over revenues (Dechow et al., 1995).
I use an extract version of the model with an adjustment for the change in receivables, so I use the following regression:
TAit = α0 + α1 (1 /ASSETSit-1) + α2 (ΔREV - ΔREC)it + α3PPEit + εit,
where:
TAit = Total accruals for a firm i at year t;
ΔREV = Change in net revenues in year t from year t-1;
ΔREC= Change in net receivables in year t from year t-1;
PPEit = Gross property, plant, and equipment;
ASSETSit-1 = lagged total assets.
3.2.2 Real activities manipulation
Roychowdhury defines real activities manipulation as ''the 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'' (2006, p. 337). As described in the literature review, real activities manipulation can be done through sales manipulation, overproduction and reduction of discretionary expenditures (Roychowdhury, 2006).
21
3.2.2.1 Sales manipulation
The first method of real earnings manipulation is sales manipulation. As concluded in prior research (e.g. Roychowdhury, 2006; Cohen et al., 2008), when managers try to manipulate sales, it is expected that this leads to lower current-period operating cash flows. So I look at the level of operation cash flows to determine the extent of sales manipulation. I use Roychowdhury's (2006) model:
CFOt / At-1 = α0 + α1 (1/At-1) + β1 (St / At-1) + β2 (ΔSt / At-1) + εt
where:
CFOt = Cash flow from operations in year t;
A = Total assets at the end of period t;
S = Sales during period t; and
ΔS = St - St-1.
For every firm-year, abnormal cash flows from operation (AB_CFO) are the actual minus 'normal' cash flows from operations (i.e., εt) from the corresponding industry-year model
and the firm-year's sales and lagged assets (Roychowdhury's, 2006). When firms offer more lenient credit terms and price discounts, this leads to a lower cash flow in period t. So a lower value of abnormal cash flows from operation (AB_CFO) indicates more sales
manipulation. Therefore, I expect the value for family firms to be positive and higher than the value of non-family firms.
3.2.2.2 Abnormal production costs
The second method of real earnings manipulation is abnormal production costs. Prior research (for example Roychowdhury, 2006) define production costs as the total of cost of goods sold (COGS) and the change in inventory during the year. Also, they define expenses as a linear function of contemporaneous sales. Following Roychowdhury (2006), I estimate the following model for normal COGS:
COGSt / At-1 = α0 + α1 (1/At-1) + β (St / At-1) + εt,
22
In the same way, I use the next equation for calculating the normal inventory growth:
ΔINVt / At-1 = α0 + α1 (1/At-1) + β1 (ΔSt / At-1) + β2 (ΔSt-1 / At-1) + εt,
Where ΔINVt = the change in inventory in year t.
Following Roychowdhury (2006), I express the production costs as PRODt = COGSt + ΔINV.
Using the last two equations, I estimate normal production costs using the following equation:
PRODt / At-1 = α0 + α1 (1/At-1) + β1 (St / At-1) + β2 (ΔSt / At-1) + β3 (ΔSt-1 / At-1) + εt
Abnormal production cost (AB_PROD) is the residual from the model. Because
overproduction leads to higher production costs, a higher value of AB_PROD indicates more real earnings management. Therefore, I expect the value for family firms to be negative and lower than the value of non-family firms.
3.2.2.3 Abnormal discretionary expenses
The last method of real earnings manipulation is abnormal discretionary expenses. Following Roychowdhury (2006), I estimate the normal level of discretionary expenses using the next equation:
DISEXPt / At-1 = α0 + α1 (1 / At-1) + β(St-1 / At-1) + εt,
Where DISEXPt are the discretionary expenses in year t, defined as the sum of R&D (research
and development), advertising and SG&A (selling, general and administrative) expenses.
For every firm-year, abnormal discretionary expenditure (AB_EXP) is the residual from the model. If managers want to boost earnings by reducing discretionary expenses, this leads to lower abnormal discretionary expenses. So a lower value of AB_EXP indicates more real earnings management. Therefore, I expect a positive value for family firms which is higher than the value of non-family firms.
3.2.2.3 Combined measure of real activities manipulation
Following Cohen (2008) et al., I make a proxy for the combined measure of real activities manipulation by accumulating the three individual real activities manipulation proxies
23
AB_CFO, AB_PROD, and AB_EXP. Considering the direction of each real activities manipulation proxy, the combined measure, COMBINED_RAM, is calculated as
AB_CFO - AB_PROD + AB_EXP.
3.3 Independent variable - family firms and non-audit services
The independent variable makes a distinction between family and non-family firms. To determine this, I use the database from Anderson, Duru and Reeb (2009) and Anderson, Reeb, and Zhao (2012). This database consists of 16,200 firm-year observations for the top-2000 largest firms from 2001 through 2010. The data includes each firm's GVKEY, the data-year (from 2001 to 2010), the company's name, an indicator variable that equals 1 when the family owns (or votes) a 5% or larger stake, and an indicator variable that equals 1 when the firm has a dual-class share structure.
Like Anderson, Duru and Reeb (2009) and Anderson, Reeb and Zhao (2012), I use a dummy variable; FAMFIRM, which equals 1 when the family owns 5% or more of the outstanding shares.
To determine how independent an auditor is, I look at the provided amount of non audit services fees (NAF). Prior research shows that there is a relationship between non audit fees and auditor independency. For example, Carmichael and Swieringa (1986) found that an increase in the amount of non audit fees leads to an impairment of the auditor's
independency. They give three reasons for this. The most important reason is that when auditors provide non audit services, they might be auditing their own work. The second reason is that non audit services might cause auditors to develop mutual interest with the management. Finally, auditors might suffer fiduciary conflicts of interest when their audit client is a shareholder of the firm but their non audit service client is the firm's management.
I determine the amount of non audit fees using the AuditAnalytics database on WRDS. I use the variable NAF which is the total amount of audit related fees, FISDI fees, benefit plan related fees, tax related fees and other/ misc. fees. For my regressions, I use the natural logarithm of NAF to avoid getting extreme values.
24 3.4 Control variables
Following Achleitner et al. (2014), I introduce four control variables. I do this to control for factors that, next to the independent variables, could influence the extent of earnings management. I include the variables LOSS, SIZE, GROWTH and LEV, where
LOSS = dummy variable, equal to one if the net income is less than zero;
SIZE = natural logarithm of the market value of the firm at the end of the fiscal year;
GROWTH = Change in sales for the firm from t-1 to t;
LEV = leverage for the firm at time t, calculated as total liabilities scaled by total assets.
I include LOSS as a control variable for mitigating the risk that managers shift earnings when their bonus is dependent on the profit. In case of a loss, managers are more inclined to shift earnings to the next year because they know that they won't get a bonus this year. I expect a positive relationship between loss and earnings management. The control variable LEV is used to control for firms that are relatively high leveraged. According to Achleitner et al. (2014), firms with more debt reduce the level of free cash flow which leads to a reduction of rent extraction opportunities for managers. Therefore, I expect a negative relationship between leverage and earnings management. The next control variable is SIZE. I expect SIZE to have a negative relationship with earnings management because bigger firms are more often scrutinized by financial analysts and usually have a stronger corporate governance structure (Carter et al., 2003). The last control variable I include is GROWTH. I expect a negative relationship between growth and earnings management, because bigger firms usually have more to lose and care more about their reputation (Rangan, 1998).
3.5 Regression model
To determine whether there exists a relationship between family business and earnings management, I base my regression model on Kim et al. (2012) and use the following models:
For discretionary accruals:
(1) ABS_DA = α0 + α1FAMILYFIRMt + α2LOSSt + α3SIZEt + α4GROWTHt + α5LEVt + εt
25
ABS_DA = value of discretionary accruals (signed discretionary accruals), where discretionary accruals are computed through the modified Jones model with an adjustment for the change in receivables.
For real earnings manipulation:
(2) RAM_PROXY = α0 + α1FAMILYFIRMt + α2LOSSt + α3SIZEt + α4GROWTHt + α5LEVt + εt
where:
RAM_PROXY = AB_CFO, AB_PROC, AB_EXP or COMBINED_RAM, where
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 COGS 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.
To determine whether the level of independency of the auditor has an impact on the relationship between family firms and earnings management, I use the following models based on Kim et al. (2012).
For discretionary accruals:
(3) ABS_DA = α0 + α1FAMILYFIRMt + α2NAS + α3NAS*FAMILYFIRM + α4LOSSt + α5SIZEt +
α6GROWTHt + α7LEVt + εt
For real earnings manipulation:
(4) RAM_PROXY = α0 + α1FAMILYFIRMt + α2NAS + α3NAS*FAMILYFIRM + α4LOSSt + α5SIZEt +
α6GROWTHt + α7LEVt + εt
26 4. Results
4.1 Descriptive statistics
Before I perform the regressions, I winsorize the variables. Winsorizing means that the outliers are removed without reducing the number of observations. Outliers are extreme values which can affect the average behaviour of certain variables within a firm.
Winsorization sets the extreme small and large observations equal to the values of less extreme observations. I do this by setting the observations in the first percentile equal to the values of the second percentile and by setting the observations in the largest percentile equal to the value of the 99th percentile.
Table 1 provides information on the number of observations, the mean, the median, the standard deviation and the minimum and maximum value of the variables.
Table 1 Descriptive statistics
Variable N Mean Median SD Min Max
FAMFIRM 986 0.346856 0 0.476211 0 1 ABS_DA 986 0.0002787 0.0076295 0.0843155 -0.330947 0.214388 AB_CFO 986 0.0446976 0.0648414 0.1396798 -0.3916587 0.4530844 AB_PROD 986 -0.0010474 -0.0010861 0.1248566 -0.3425328 0.3262998 AB_EXP 986 -0.0010277 -0.0096909 0.161476 -0.3376172 0.4986554 COMBINED_RAM 986 0.0451094 0.0297287 0.2927127 -0.6515464 0.9210484 NAF 986 11.36995 12.35455 3.579737 0 15.23403 FAMFIRM*NAF 986 3.688905 0 5.616162 0 14.18569 LOSS 986 0.4310345 0 0.4954722 0 1 SIZE 986 6.511315 6.495378 1.405009 2.206823 10.3267 GROWTH 986 0.1154356 0.0838467 0.3085346 -0.6689458 1.657718 LEV 986 0.4642353 0.4143831 0.3075125 0.0415951 1.632847
This table provides the descriptive statistics for the variables that are used in the regressions.
It shows the number of observations, the mean, the median, the standard deviation and the minimum and maximum value of the variables.
See appendix A for variable descriptions.
The first variable, FAMFIRM, is a dummy variable which equals 1 when the family owns (or votes) a 5% or larger stake. I use FAMFIRM to indicate whether a firm is a family firm or not. The mean is 0,346856 which means that 34.69% (324 out of 986 firms) of my sample is indicated as family firms and 65.31% (644 out of 986 firms) is indicated non-family firms.
27
Although my sample is not equally divided, it gives me an opportunity to make a comparison between family and non-family firms.
On average, the most standard deviations are small, indicating that the values of the observations are concentrated around the mean.
Concerning the control variables, table 1 shows that the mean of LOSS is 0.4310, which means that 43.10% of the firms have a net income less than zero. The mean of SIZE, measured as the natural logarithm of the market value of the firm at the end of the fiscal year, is 6.511315. Also, the mean of GROWTH is 0.1154356 which means the firms have an average change in sales of 11.54% from t-1 to t. Finally, the mean of LEV is 0.4642353.
After I calculated the mean, the median, the standard deviation and the minimum and maximum values of the variables, I performed a two-tailed T-test to compare the means of family and non-family firms. The null hypothesis states that there is no significant difference in mean of variables between family and non-family firms. Table 2 shows the p-value of the alternative hypothesis, which is that the difference in the mean is not equal to 0.
Table 2 Descriptive statistics: two-tailed T-test of the means
Familyfirms (N=342) Non-Familyfirms (N=644)
Variable Mean Mean
Difference in mean P-value ABS_DA 0.0074738 -0.0035423 0.0110161 0.0508 AB_CFO 0.0550418 0.0392043 0.0158375 0.0902 AB_PROD -0.0097162 0.0035562 -0.0132724 0.1122 AB_EXP 0.0006743 -0.0019316 0.0026059 0.8095 COMBINED_RAM 0.0644851 0.0348198 0.0296653 0.1299 NAF 10.64607 11.75438 -1.10831 0.0000 FAMFIRMNAF 10.63526 0 10.63526 0.0000 LOSS 0.4532164 0.4192547 0.0339617 0.3059 SIZE 6.395901 6.572606 -0.176705 0.0601 GROWTH 0.1117449 0.1173955 -0.0056506 0.7845 LEV 0.4925091 0.4492203 0.0432888 0.0353
Table 2 shows the two-tailed T-test to compare the means of family and non-family firms. See appendix A for variable descriptions.
28
Table 2 shows that only the the means of ABS_DA and AB_CFO are different at a 10% significance level.
When looking at the difference in means for discretionary accruals, I have contradictory results with prior research. Unlike Jiraporn and Dadalt (2009) and Achleitner et al. (2014), my sample shows that family firms have on average higher discretionary accruals.
Because offering price discounts and more lenient credit terms lead to lower cash flows, I would expect a higher value of abnormal cash flows from operation (AB_CFO) for family firms. This is the case, indicating that family firms engage on average in less real activities manipulation by sales manipulation.
Also, the difference in means for non-audit fees is different between family and non-family firms. So family firms pay on average less non-audit fees at a significance level of 1%.
Finally, from the control variables, only the means of SIZE and LEV are significantly different at a respectively 10% and 5% significance level. The mean of SIZE for non-family firms is 6.572606 and for family firms 6.395901, indicating that non-family firms are on average larger than family firms. Also, the mean of LEV is significantly higher for family firms, showing that family firms have a higher level of leverage on average.
Table 2 indicates that the hypothesis that family firms perform less earnings management than non-family firms can be rejected. However, I cannot make this conclusion yet because I didn't take the outcome of the control variable into account. Therefore, I perform a
multivariate analysis in paragraph 4.3.
4.2 Correlation matrix
Table 3 presents a correlation matrix based on the Pearson correlation. The Pearson Correlation measures how strong the linear relationship is between the independent,
dependent and control variables. The outcomes can vary between -1 and +1 and an outcome of zero means that there is no correlation between the two variables. A correlation is
assumed to be high if it is below -0.7 or above 0.7 and this affects the reliability of the regression model.
29
Table 3 Correlation matrix
FAMFIRM ABS_DA AB_CFO AB_PROD AB_EXP COMBINED_RAM NAF FAMFIRM*NAF LOSS SIZE GROWTH LEV
FAMFIRM 1 ABS_DA 0.062* 1 AB_CFO 0.060** -0.056* 1 AB_PROD -0.059** -0.010 -0.275*** 1 AB_EXP 0.012 -0.083*** -0.038 -0.543*** 1 COMBINED_RAM 0.056** -0.065** 0.548*** -0.835*** 0.741*** 1 NAF -0.147*** -0.031 0.163*** -0.065** 0.102*** 0.158*** 1 FAMFIRM*NAF 0.902*** 0.045 0.084*** -0.074** '0.040 '0.090*** 0.161*** 1 LOSS 0.029 -0.274*** -0.583*** 0.181*** 0.066** -0.311*** -0.110*** -0.036 1 SIZE -0.059** 0.072** 0.494*** -0.174*** 0.136*** 0.380*** 0.320*** 0.151*** -0.453*** 1 GROWTH -0.012 -0.009 0.125*** -0.002 0.170*** 0.144*** 0.023 0.038 -0.148*** 0.249*** 1 LEV 0.058** -0.173*** -0.036 0.009 0.086*** 0.026 0.088*** 0.019 0.130*** -0.045 -0.106*** 1
This table reports the Pearson correlation between the variables. See appendix A for variable descriptions.
30
Table 3 shows that there is a significant positive correlation between family firms and abnormal accruals and abnormal cash flows from operation. Also, there is a significant negative correlation between family firms and abnormal production. However, there is no signification correlation between the family firms and abnormal discretionary expenses.
Furthermore, it can be seen that there are a lot of significant correlations between the different control variables and other variables. The importance of this will be discussed in paragraph 4.3, where I estimate the relationship between family firms and the extent of earnings management.
There are not many correlations above 0.7 or below -0.7. Between COMBINED_RAM and AB_PROD and COMBINED_RAM and AB_EXP there is a correlation of respectively -0.835 and 0.741. However, this makes sense because AB_PROD and AB_EXP are components of
COMBINED_RAM so these values will not affect the reliability of the regression model. There is also a correlation above 0.7 between FAMFIRM and FAMFIRM*NAF. This will also not affect the reliability of the model because FAMFIRM*NAF is a variable with an interaction with FAMFIRM.
The different independent variables in my regression models could lead to multicollinearity. This means that there is a correlation between different independent variables. I perform the Variance Inflation Factor test (VIF) to determine whether the independent variables are really independent or that they are correlated. The VIF factor indicates whether an
independent variable is explained by another variable. If the outcomes of the VIF test are below 10 there is no multicollinearity between the independent variables. The results of the test are shown in table 4.
31
Table 4 Variance Inflation Factor (VIF) test
Variable VIF 1/VIF
FAMFIRM 10.20 0.098079 NAF 2.08 0.480275 FAMFIRM*NAF 10.21 0.097944 LOSS 1.28 0.781672 SIZE 1.39 0.719579 GROWTH 1.08 0.925093 LEV 1.05 0.951336 Mean VIF 3.90
This table shows the Variance Inflation Factor for the different variables. See appendix A for variable descriptions.
Table 4 shows that the independent variables, on average, don't lead to multicollinearity. However, the outcomes for FAMFIRM and FAMFIRM*NAF are both very high and could lead to multicollinearity. Though, this makes sense because FAMFIRM*NAF is a variable with an interaction with FAMFIRM. Therefore the VIF's of respectively 10.20 and 10.21 present no problem.
4.3 Multivariate analysis
I perform two analyses. First, I look at the impact of family firms on earnings management. Then I perform an additional analysis to decide whether the amount of non audit fees
influences the extent of earnings management within family firms. Furthermore, I look at the difference between accrual based earnings management and real activities manipulation.
4.3.1 Effect of family ownership on earnings management
In this paragraph, I discuss the results of the regressions with regard to abnormal
discretionary earnings management (ABS_DA) and real activities manipulation (AB_CFO, AB_PROD, AB_EXP and COMBINED_RAM). The results are shown in table 5.
32
Table 5 Multivariate analysis
Dependent variables
ABS_DA AB_CFO AP_PROD AB_EXP COMBINED_RAM
(Constant) 0.056*** -0.109*** 0.067*** -0.186*** -0.359*** (0.02) (0.02) (0.02) (0.03) (0.05) Independent Variables FAMFIRM 0.014*** 0.026*** -0.016* 0.005 0.046* (0.01) (0.01) (0.01) (0.01) (0.02) Control variables LOSS -0.049*** -0.130*** 0.033*** 0.051*** -0.107*** (0.01) (0.01) (0.01) (0.011) (0.02) SIZE -0.003 0.030*** -0.012*** 0.020*** 0.061*** (0.00) (0.00) (0.00) (0.00) (0.01) GROWTH -0.016* -0.006 0.020 0.084*** 0.050** (0.01) (0.01) (0.013) (0.02) (0.03) LEV -0.043*** 0.011 -0.002 0.047*** 0.06** (0.01) (0.01) (0.02) (0.02) (0.03) N 986 986 986 986 986 R2 0.1082 0.4153 0.0506 0.0689 0.1809 Adjusted R2 0.1037 0.4123 0.0458 0.0642 0.1767
This table reports the results of the first regression with the following model:
Dependent variable = α0 + α1FAMILYFIRMt + α2LOSSt + α3SIZEt + α4GROWTHt + α5LEVt + εt ,
where the dependent variable = ABS_DA, AB_CFO, AB_PROD, AB_EXP or COMBINED_RAM. See appendix A for variable descriptions.
*, **, *** Significant at 10 percent, 5 percent and 1 percent levels, respectively
Table 5 shows that FAMFIRM has a significant relationship with ABS_DA, AB_CFO, AB_PROD and COMBINED_RAM. There is no significant relationship between FAMFIRM and AB_EXP.
Table 5 shows that the adjusted R-squared for ABS_DA is 0.1082. This means that 10.82% of the variation in ABS_DA is explained by the model. Based on the alignment and
entrenchment effect, there are different expectations about the relationship between family firms and accrual based earnings management. The alignment effect predicts a negative and the entrenchment effect a positive relationship. Table 5 shows that there is a significant positive relationship between family firms and accrual based earnings management, which
33
means that family firms engage in more accrual based earnings management. Therefore, I assume that the entrenchment effect in the firms of my sample is more pervasive than the alignment effect. The outcomes are contradictory with prior research, as Jiraporn and DaDalt (2009) and Achleitner et al. (2014), who found a negative relationship between family firms and accrual based earnings management. In my model, the control variables LOSS and LEV are significantly related to abnormal discretionary accruals.
According to table 5, the adjusted R-squared for AB_CFO is 0.4153, so 41.53% of the variation in AB_CFO is explained by the model. As concluded in prior research (e.g.
Roychowdhury, 2006; Cohen et al., 2008), when managers try to manipulate sales, this leads to lower current-period operating cash flows. My hypothesis predicts less earnings
management for family firms so I expect a positive sign for AB_CFO. The value for family firms is +0.026, so I conclude that real activities management through abnormal cash flows is less likely for family firms, which is in line with my expectations. Moreover, there is a
significant relationship between the control variables LOSS and SIZE and abnormal cash flow earnings management.
For AB_PROD, the adjusted R-squared is 0.0459, which means that 4.59 of the variation in AB_PROD is explained by the model. Because overproduction causes higher production costs, a higher value of AB_PROD indicates more real earnings manipulation. The value in table 5 for family firms is low, -0.018, so there seems to be a negative relationship between family firms and real earnings manipulation based on abnormal production costs. All of the control variables are significant.
When looking at AB_EXP, the value for family firms is positive and insignificant. This means that there is no supporting evidence that family firms engage in less earnings management than non-family firms with regard to abnormal discretionary expenses.
Finally, the value for COMBINED_RAM is positive and significant, with an adjusted R-squared of 0.1767. This means that there is a negative relationship between family firms and total real activities manipulation. This is in accordance with prior research, such as Jiraporn and DaDalt (2009) and Achleitner et al. (2014), who show that there is a negative relationship between family firms and real activities based earnings management. Also for
34
In addition, when looking at the control variables, almost all variables are significant. However, they are not always as how I expected them to be. For example, for LOSS, I expected a positive relationship with earnings management because managers, in case of a loss, might try to shift the earnings to increase the chance that they get a bonus next year. For both accrual based and real activities based earnings management there is a negative relationship, which is contradictory to my expectations. The variable SIZE is also not in line with my expectations. I expected a negative relationship with earnings management, because bigger firms are more often scrutinized by financial analysts and usually have a stronger corporate governance structure. The variable for ABS_DA is negative, but not significant, and COMBINED_REM is positive, so both are not in line with my expectations. For GROWTH, I expected a negative relationship between family ownership and earnings
management. However, this is only the case for ABS_DA. For the last variable, LEV, I
expected a negative relationship because firms with more debt reduce the level of free cash flows which reduces the rent extraction opportunities for managers. This is the case for ABS_DA, but COMBINED_REM has a positive relationship and is not in line with my expectations.
In summary, my model shows a positive relationship between family firms and accrual based earnings management, which contradicts my hypothesis and prior research. On the other hand, there is a negative relationship between family firms and real activities manipulation through abnormal cash flows, production costs and total RAM activities, which is in line with my hypothesis. There is no significant relationship between family firms and real activities manipulation through abnormal discretionary expenses. Based on these outcomes, I reject hypothesis 1 ''There is a negative relationship between family business and accrual based earnings management'', and accept hypothesis 2 ''There is a negative relationship between family business and real earnings manipulation''.
4.3.2 Additional analysis with non audit fees
To test hypothesis 3 ''The negative relationship between family ownership and earnings management is less pronounced for firms where non-audit services provided by audit firms are higher'', I perform an additional analysis and add the variables NAF and FAMFIRM*NAF to the regression. The results are shown in table 6.
35
Table 6 Multivariate analysis
Dependent variables
ABS_DA AB_CFO AP_PROD AB_EXP
COMBINED_ RAM (Constant) 0.060*** -0.121*** 0.069** -0.202*** -0.381*** (0.02) (0.02) (0.03) (0.03) (0.06) Independent Variables FAMFIRM 0.017 0.023* 0.002 - 0.079 0.019 (0.02) (0.02) (0.02) (0.03) (0.05) NAF -0.0006 0.0020 0.0003 0.0027 0.0035 (0.00) (0.00) (0.00) (0.00) (0.00) NAF*FAMFIRM -0.0006 -0.0020 -0.0020 0.0013 0.0024 (0.00) (0.00) (0.00) (0.00) (0.00) Control variables LOSS -0.049*** -0.127*** 0.031*** 0.053*** -0.101*** (0.01) (0.01) (0.01) (0.01) (0.02) SIZE -0.002 0.029*** -0.011*** 0.018*** 0.059*** (0.00) (0.00) (0.00) (0.00) (0.01) GROWTH -0.016* -0.003 0.016 0.089*** 0.059** (0.01) (0.01) (0.01) (0.02) (0.03) LEV -0.042*** 0.014 -0.009 0.046*** 0.065** (0.01) (0.01) (0.01) (0.02) (0.03) N 986 986 986 986 986 R2 0.1097 0.4130 0.0481 0.0769 0.1829 Adjusted R2 0.1034 0.4088 0.0413 0.0703 0.1770
This table reports the results of the second regression with the following model:
dependent variable = α0 + α1FAMILYFIRMt + α2NAS + α3NAS*FAMILYFIRM + α4LOSSt + α5SIZEt + α6GROWTHt + α7LEVt + εt
where the dependent variable = ABS_DA, AB_CFO, AB_PROD, AB_EXP or COMBINED_RAM. See appendix A for variable descriptions.
*, **, *** Significant at 10 percent, 5 percent and 1 percent levels, respectively
When comparing this table to table 5, it can be seen that the outcomes are almost the same. There are no significant differences between the control variables. Also, the outcomes for NAF and NAF*FAMFIRM are very low and insignificant, indicating that the auditors for non-family firmsf are not more or less independent than for non-non-family firms. Only the value of AB_PROD is different, this was before adding NAF and NAF*FAMFIRM -0.012 (significant at 10% level) and after -0.016 (not significant), which could indicate that degree of
36
independence of the auditor influences the amount of real earnings manipulation on abnormal production costs.
Because adding the variables NAF and NAF*FIRM only has an impact on AB_PROD at the 10% significance level and the other variables are the same, I reject hypothesis 3 ''The negative relationship between family ownership and earnings management is less pronounced for firms where auditors are less independent''.
5 Conclusion
Family firms are an important factor in the economy. Anderson and Reeb found that family firms constitute of over 35 percent of the S&P 500 industrials and that families own, on average, nearly 18 percent of their firms' outstanding equity (2003). Next to their important position, family firms have some specific characteristics. These characteristics lead to
relatively less type I agency costs, which Wang (2006) describes as the alignment effect, and relatively more type II agency costs, which he describes as the entrenchment effect. As a consequence of the alignment effect, family firms have less information asymmetry, more effective direct monitoring, longer time horizon, better reputation and a lower need for corporate governance. However, the entrenchment effect leads to the expropriation of non-controlling shareholders. Based on the alignment and entrenchment effect, there are different expectations about the relationship between family firms and earnings
management which leads to my research question: 'To what extent does the amount of earnings management differ between family and non-family firms? ' Prior research (for example Jiraporand & Dadalt, 2009 and Achleitner et al., 2014), found a negative relationship between family firms and the extent of earnings management. This study however shows a negative relationship between family firms and real activities
manipulation, but a positive relationship with accrual based earnings management. This can be explained that by the fact that the entrenchment effect for the firms in my sample is more pervasive. This study contributes to the literature because it also looks at the independency of the firms' auditors. The predicted negative relationship between family firms and earnings management could have an influence on the auditor's independence. For example, if an auditor knows that family firms engage in less earnings management and has therefore higher earnings quality, he might act less independent because he expects that
37
investors will look less carefully at the financial statements. Despite this assumption, I found no evidence that the degree of independency of auditors has an influence on the difference of earnings management between family and non-family firms.
This study has some limitations. For example, I considered a firm to be a family firm when the family owns (or votes) a 5% or larger stake. However, prior research shows that there are more characteristics of family firms that have to be taken into account to determine whether a firm is a family firm or not. For example Achleitner et al. (2014), also looks at whether the families are on the management and/or supervisory board of the firm and the number of voting rights they hold. I recommend further research which includes a broader variable for family firms. Another limitation of this study is the sample size. My sample consists of 986 firms, but the outcomes could have been different when the sample size was bigger. Therefore I recommend further research with a larger sample size.
