1 Amsterdam Business School
Family firms and earnings management
Name: Tarik Bittich
Student number: 11077387 Supervisor: Dr. Mario Schabus Words count: 13218
Date: 26 June 2017
MSc Accountancy & Control, track Accountancy
2
Statement of Originality
This document is written by student Tarik Bittich 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
Earnings management in family firms
ABSTRACT: This study explores the effect of a founding-family ownership structure, on the levels of earnings management in a U.S context. Earnings management is divided into
(ABEM) accruals- based earnings management and (REM) real earnings management. According to the (SEW) socioemotional wealth theory, family firms value the sustainability of the business and they avoid actions that hinder future value. Family firms find it important to gain sustainable success and to retain the family reputation. Second, according to the agency theory there are fewer conflict of interest between agent and principal within family firms. Because of the alignment effect of family ownership there is no need to base
management compensation on reported earnings. Based on the two assumptions earnings management would be less likely in family firms. As result of the institutional and legal differences of prior results, the results don’t hold for U.S listed firms. First of all, based on the SEW theory, I predicted that family firms use less REM than non-family firms.
Inconsistent with the prediction made, the results show that there is no evidence that family firms use less REM than non- family firms. Second, based on the absence of the adverse incentives arising from compensation structures such as bonuses and the conservative accounting style within family firms, I expect that family firms less often engage in ABEM than non-family firms. Consistent with the prediction made, the results show that family firms use less ABEM than non- family firms. Moreover, the results are robust with respect to the inclusion of year fixed effects. Overall, these findings show how ownership structures, institutional characteristics and managerial behavior influence financial reporting.
4
ACKNOWLEDGMENT
At the end of my education in MSc Accountancy & Control at the University of Amsterdam, I conducted an investigation. The purpose of this master thesis was, to investigate if there is a correlation between the choice of types of earnings management ( REM, ABEM) and family ownership in a firm. In order to test my ability as a MSc student, I was expected to be able to deliver a master thesis that complies with the norms of the education Accountancy & Control. The final result is a report that contributes to the academic literature as a practical contribution as well.
I would first like to thank my thesis supervisor, Dr. Mario Schabus, for his guidance during the master thesis process. His remarks and recommendations were useful to steer me into the right direction. The discussion during the period with my fellow students were also very helpful. Also, I would like to express my deepest gratitude towards my family for supporting and motivating me during this tough and educational process.
Tarik Bittich
5 Table of Contents
1 Introduction 6
2.1 Family firms 9
2.1.1 Socioemotional wealth theory 10
2.1.2 Agency theory in family firms 11
2.2 Earnings management 12
2.2.1 Motives earnings management 13
2.2.2 Real earnings management 14
2.2.2 Accrual-based earnings management 15
2.3 Hypotheses development 16
3.1 Sample selection 17
3.2 Family firms 17
3.3 Real earnings management model 18
3.3.1 Accrual-based management model 20
3.4 Empirical model 21 3.5 Control variables 22 4.1 Descriptive statistics 24 4.2 Multivariate analyses 28 4.3 Robustness test 32 5 Conclusion 33 References 35
Appendix A : Descriptive statistics 37
Appendix B : Regression analyses 38
6
1 Introduction
Firms managed or controlled by founding families are referred to as family firms. They own a large block of shares and hold at least one key management position within the family firm. Wang (2006) stressed the importance of family firms for the world economy. In the US, 50% of all listed companies are characterized as family firms (Sirmon & Hitt, 2003). Furthermore, in Europe, 60% of the companies are owned by families, while in Germany 95% of both listed companies and private companies fits into this category. Finally in Australia, 67% of the firms are family firms. Clearly, family firms represent a major part of the world economy. In this paper, I respond to the call of Achleitner et al. (2014) for conduct further research into family firms and earnings management.
This study will focus on listed companies in the U.S (United States), because prior studies contain several shortcomings. Prior research studies, such as those of Wang (2006) and Ali et al. (2007), investigated the effect of family firms on the quality of financial reporting using accrual-based earnings management (ABEM) based on accounting choices. These studies also used agency theory to explain family attitudes toward the quality of financial reporting. However, the contribution of this paper is also to consider a second type of earnings management: real earnings management (REM), based on real economic decisions. Moreover, this research will also be based on another theoretical framework: the concept of socioemotional wealth (SEW). According to the socioemotional wealth theory, family firms attach great importance to sustainable success (Berrone et al., 2012). The socioemotional wealth theory considers that the goal of a family firm is to keep itself intact and leave it to the next generation of the family. Agency theory assumes the opposite: the agent pursues other objectives than the principal and will make suboptimal decisions in favor of his/her reward. Socioemotional wealth theory provides a better framework on which predictions about earnings management between family and non-family firms can be made. The conflict of interest between principal and agents is less present within family firms, because the agent and the principal may belong to the same family, the agent is often close to the principal, and agents are frequently not pulled from the labor market for executives. In other words, it can be assumed that the majority shareholders and agents have the similar interests. Therefore, it is expected that family firms will exhibit less in real earnings management. This is mainly caused because REM, in the long term, will have a negative impact on the operational activities, thereby affecting the sustainability of the company. I expect that family firms will engage less in REM.
7
Next, Ali et al. (2007) and Wang (2006) suggested that because of the alignment effect of family ownership there are fewer conflicts of interest between principal and agent. Chen (2006) argued that under the classic agency theory, family ownership may align with the interest of the family firm and its managers, which reduces the agency problem that is caused by a separation of ownership and control. According to this study, there is no need to base managers’ bonuses upon earnings in family firms. So overall, one would expect, then, that family firms have less ABEM because it is unnecessary for compensation purposes to manage accruals opportunistically. In other words: the bonus plan hypothesis does not hold. This leads to the presumption that family firms are less concerned with ABEM than non-family firms. Regarding both types of earnings management, one can argue that non-family firms are less pressured by the shareholder or investment community to meet or beat earnings benchmarks to maintain stock-prices. Abundant evidence shows that earnings management is mostly applied to meet benchmarks, but one may expect that family firms place a lesser value on these benchmarks.
As mentioned, in this study, I will respond to the call of Achleitner et al. (2014) for further research into family firms and earnings management. The findings of Achleitner et al. (2014) are specific to companies in Germany. The scientists indicated that the results have to be handled carefully because the results are not generalizable to other countries. It is expected that the use of a wider population will lead to research results that can be generalized to foreign listed companies established outside the US, listed in the US. In this study, the sample will only consist of a distribution of US firms listed in the US. Prior studies of Wang (2006) and Ali et al. (2007) investigated only family firm’s effect on the quality of financial
reporting using ABEM.Furthermore, the research of Jiraporn and Dadalt (2009) showed that
family firms use less accrual-based earnings management than non-family firms. However, the focus of this research was limited to Italian firms. I will respond to these limitations by introducing REM. In addition, Achleitner et al. (2014) based their research on a timeframe of 10 years (1998-2008). The research results of this research period may have been affected by undesirable influences because in 2005, Germany and Europe adopted the international financial reporting standards (IFRS), which did not go well. The effect of the change is not limited to the year of transition, but also to the surrounding years. Chapcun et al. (2012) indicated that there were early adopters of IFRS before 2005, because of incentives awarded, and other companies only applied IFRS after 2005. I will reduce the research period so that the findings are updated, and similar effects to previous investigations are mitigated. Furthermore, the use of a more recent time period will also increase the actuality of the
8
literature on earnings management in family firms. The implementation of the Sarbanes-Oxley (SOX) legislation will have an impact on the level of earnings management. The study of Cohen et al. (2008) showed that, after the introduction of the SOX legislation, ABEM gradually decreased and REM gradually increased. The empirical results will show the impact on family firms and non-family firms. Finally, this study will differ from that of Achleitner et al. (2014) because of the institutional differences between Europe and the US. Leuz et al. (2003) claims that listed firms in Europe have less developed stock markets and weak investor protection rights compared to listed firms in the US. These factors seem to influence levels of earnings management in this study.
Overall, this has led to the following research question: Is there a correlation between the choice of types of earnings management ( REM, ABEM) and family ownership in a firm?
In addition to the contribution to the academic literature, my study has a practical contribution as well. The findings of this study may be relevant to the profession of auditors and investors. The research results can contribute to the planning stage of audit services in family firms and non-family firms and to the decision-making process of investors.
The remainder of this paper is structured in the following manner: in section 2, I will discuss literature and develop hypotheses; in section 3 the sample and methodology will be described and section 4 will show the results. Finally, section 5 will show the conclusions of this study.
9
2.1 Family firms
There is no consensus about the definition of a family firm. Different aspects are taken into account when characterizing a family firm. The European Commission (EC, 2009) defined family firms as follows, with the addition of four categories: “A firm, of any size, is a family firm, if: 1) The majority of decision-making rights is in the possession of the natural
person(s) who established the firm, or in the possession of the natural person(s) who has/have acquired the share capital of the firm, or in the possession of their spouses, parents, child or children’s direct heirs. 2) The majority of decision-making rights are indirect or direct. 3) At least one representative of the family or kin is formally involved in the governance of the firm. 4) Listed companies meet the definition of family enterprise if the person who
established or acquired the firm (share capital) or their families or descendants possess 25% of the decision-making rights mandated by their share capital” (p. 4).
According to Chrisman et al. (2012), to be a family firm, a company should not only be in a family’s possession, but also be controlled by that family. Furthermore, another important aspect is the continuity of the family firm in terms of progeny. In this aspect, it is important that the family is aware of the preparations necessary to ensure the continuity of the company.
Deephouse and Jaskiewicz (2013) defined family firms as requiring at least one representative of the family or next of kin to be formally involved in the company’s governance. In addition, the firm’s establisher or acquirer has at least 25% of the decision-making rights on the basis of his/her share. Also, the share capital in the family’s possession is at least in the second generation or after. Clearly, defining a family firm is not a simple task: there is considerable overlap, but also some differences.
In the U.S, 50% of the listed companies are characterized as family firms. The difference between family firms and non-family firms is the ownership structure. In contrast to listed non-family firms, the family in a family-controlled firm holds a significant number of shares and participates in a management position as well. Prior studies have shown that there is a behavioral difference between family and non-family firms. Lins et al. (2013) stated that family firms are associated with risk-adverse behavior. They take fewer investment risks, particularly during a crisis. Furthermore, Lins et al. (2013) argued that family firms are more conservative and long-term oriented. The survival of the company is more important, while non-family firms are more focused on maximizing shareholder value. The main purpose of a family firm is not to sell the shares of the firm but to be continued by the next generation.
10
Lins et al. (2013) concluded that family firms underperform in comparison with non-family firms, in part because they limit executive management members, suggesting a restricted labor pool from which to gain qualified and capable talent, potentially leading to competitive disadvantages relative to non-family firms.
However, there are also studies that contradict these findings. Anderson and Reeb (2003) claimed that family firms are positively related to firm performance. Family wealth is closely linked to firm value. Those family members working in the firm are motivated to increase firm performance and thereby maintain their family’s reputation. Founding families make fewer decisions to boost short-term earnings and focus more on long-term value.
2.1.1 Socioemotional wealth theory
Considerable research has been conducted in current academic literature on earnings management and the motives to engage in earnings management. Academic literature has particularly focused on the agency theory (Jensen & Meckling, 1976). To understand the notion of earnings management from another perspective, Gomez-Meja et al. (2007) developed the socioemotional wealth theory (SEW). In this setting, the main goal of family firms is to protect the utility they also gain from non-economic aspects of the firms. These aspects include improving the family’s ability to control the firm, protecting the wealth of the family members, and ensuring the continuity of the business for future generations. The SEW approach argues that family members consider the firm as a long-term investment to be bequeathed to descendants.
In the agency theory of Jensen and Meckling (1976), managers manipulate profit to maximize their bonus. Moreover, Jensen and Meckling (1976) found evidence that managers reduce research and development expenditures a year before they retire, in order to increase the profit and maximize their bonus. The socioemotional wealth theory approached it from another perspective, by including the relationship between family firms and the two types of earnings management. Prior research from Tong (2008) reports findings that large shareholders, such as family members, have the incentive to dispose of small shareholders. They report conservatively. For instance, family firms want to invest in long-term growth and reduce dividend payouts. They argue that earnings reinvestment will boost future returns. Using the SEW framework, family firms attach more importance to long-term value and are averse to short-term capital market pressure to report inflated earnings.
11
Furthermore, Achtleitner et al. (2014) argued that the SEW theory socioemotional wealth. The main goal is to leave a successful firm for the next generation of the family. To achieve this, the company should be healthy and make decisions that ensure the sustainability of the firm as a primary goal. Therefore, it is expected that family firms will exhibit less real earnings activities than non-family firms, since the REM activities will, in the long-term, have a negative effect on the family firm’s operational activities and harm its sustainability. assumes that family firms are primary controlled from the viewpoint of potential gains and losses in terms of
2.1.2 Agency theory in family firms
Chrisman et al. (2007) claims that family firms are a unique area within the agency theory. The conflicts of interest between principal and agents do not exist or are naught due the family involvement in both ownership and management. Chrisman et al. (2007) argues that agency problems only exist in family firms if a member of the firm’s management is not a family-member. Family members normally act in the interest of the family firm. Monitoring family members in the management board will motivate those managers to pursue family goals and increase firm performance.
This reasoning is also shared by Kang (Kang, 2014): within family firms, the agent is the principal and the principal is the agent. Less monitoring is needed, which has a positive influence on the agency costs. He goes as far as to argue that family firms have more reliable managers in comparison with non-family firms, because the firm is not managed by opportunistic incentives.
Meanwhile, Schulze et al. (2001) contended that it is almost impossible to have only experienced and skilled family members in the management board of listed family firms. There are always non-family members involved in the management of a listed family firm. As result of that, it is plausible that family firms will experience problems with respect to information asymmetry: it is not unlikely that family members will share inside information with other family members/owners, while excluding non-family members of this, relevant, information.
Overall, following prior literature, agency costs within family firms are only lower in comparison to non-family firms when the management board consist of family members. Family firms are long-term oriented, pursue family goals and wish to maintain success within the family. This means that less incentives are needed to manage the earnings and monitor the
12
agents. However, agency problems will increase within family firms when there are non-family owners involved as well.
2.2 Earnings management
In this section, a detailed explanation will be provided of the incentives of earnings management and the two earnings management methodologies that can be applied. Scott (2011) described earnings management as: “a situation where managers are incentivized to manipulate accounting data willfully for their own interests. A manager has discretion over accounting choices to influence how economic events are reflected in measures of earnings’’ (p. 99). With regards to earnings management, the manager has a goal to reach a specific reported objective. Earnings management can be divided into real earnings management, based upon operational decisions, and accrual-based earnings management, based upon accounting choices regarding financial reporting.
According to Dechow et al. (1995), certain criteria must be met to apply earnings management. The first condition is that there must be a principal-agent problem. This occurs when the agent pursues other interests than the principal. The agent has an information and knowledge advantage over the principal, which is also called information asymmetry. As a result, the agent can pursue his/her own interest without the knowledge of the principal. Normally, agency problems are minimized within family firms, since the principal is the agent and the agent is the principal (Kang, 2014). However, this does not always apply to listed companies. Agency problems exist in family firms when the firm has at least one manager who is not the owner of the firm. However, most scientists argue that agency problems do not exist within family firms or is many times lower in family firms than in non-family firms (Lins et al., 2003). The second condition is that there must be a bonus system, which leads to the incentive for the managers to manipulate accounting numbers to maximize his/her bonus. For instance, a manager tries to maximize the revenues when his/her bonus is linked to the revenues. Finally (third condition), the manager needs to have a discretionary decision competence. This means that the manager has the ability to make decisions that influence the results of the firm. The manager has the power over the operational process and can, therefore, influence the operational numbers, for instance by delaying decisions that are of influence to the result of a certain period.
13
2.2.1 Motives earnings management
Bhaumik and Gregoriou (2010) provided several motives to engage in earnings management. Market expectation and stock market price
The interaction between accounting numbers and stock market reaction can push management toward earnings management. Listed companies periodically publish financial numbers, which are scrutinized by analysts to assesses the current performance of the firm and to gain information about expectations with respect to the firm’s future performance. Investors rely on the views of market analysts to put together a portfolio of successful firms (Anderson and Reeb, 2003). Innovation can lead to a strong growth of a firm, for instance by launching a new product. This growth will form the basis of new targets. However, the effect of launching the new product will decrease at a given moment. The problem will be exacerbated when there are no new products to be launched. To lengthen the growth line, the firm seeks possibilities to keep the artificial growth intact to meet or beat expectations of market analysts. If the firm does not achieve these expectations, it can have a negative impact on the share price.
Performance bonus
The purpose of artificially manipulating accounting numbers may relate to the achievement of personal economic benefits. In a situation where the profit does not reach the upper limit of lower limit, the profits will be adjusted. This happens through discretionary accruals to maximize bonuses for the current period and next period. There may also be situations where it goes economically bad or goes well. If all goes well and the manager exceeds all expectations, the manager may choose to shift some of the good results to the next period. In this way, it is easier for the agent to achieve the same targets in the next period. Thus, his/her target is not increased, and the agent receives his/her bonus. This is called “target ratcheting”. This situation can also occur when poor results are achieved. If the performance for the current period is bad, the agent can shift the next period’s poor performance to the current period because the current period is considered lost. According to Scott (2009), this happens during reorganizations and organizational changes. For instance, because of accelerated depreciation, future depreciation costs will be lower and the profits will be higher.
14
Debt covenants
A firm is usually financed with debt capital, and to acquire debt capital, a firm enters into an agreement with a credit lender. The contract often stipulates the minimum requirements that must be met. Often, these are additional requirements about liquidity. If the firm does not meet the contractual agreement, the loan is often immediately claimable. Therefore, managers tend to use earnings management to present a better liquidity position. The investigation of Roychowdhury (2006) showed that if there is a weak liquidity position, managers pass on earnings management to present a stronger liquidity position in order to avoid lender sanctions. For instance, the manager can strengthen the liquidity position to comply with the bank covenants by postponing expenditures of the current book year to the following book year.
2.2.2 Real earnings management
Cohen & Zarowin (2010) defined real earnings management as: “management actions that deviate from normal business practices, undertaken with the primary objective to mislead certain stakeholders into believing that earnings benchmarks have been met in the normal course of operations” (p. 14). Real activities manipulation has an immediate impact on the cashflow and reduces or improves firm value, while accrual-based earnings management does not harm firm value. Nevertheless, managers still prefer to use real activities-based earnings management because real earnings management activities are less likely to be scrutinized by regulators and auditors, and thus potentially have a greater probability of not being detected (Zang, 2006).
According to Roychowdhury (2006), managers use real earnings management to obtain financial targets beforehand. There are three methods by which real earnings management can be achieved. The three methods are listed below (Gunny, 2005).
Sales manipulation
Sales manipulation is used by managers through price discounts (acceleration of sales timing) or more lenient credit terms (Cohen & Zarowin, 2010). The discounts and lenient credit terms will increase sales volumes, but these are likely to disappear when the firm reverts to old prices. When the product price returns to normal, negative consequences may arise, such as lower returns and lower firm value.
15
Discretionary expenditures
Discretionary expenditures include advertising, R&D, and SG&A expenses. Postponing such expenses will improve current period earnings and lower cash outflows. On the other hand, increasing discretionary expenses does not lead to direct profit, but may only be seen in the subsequent years’ results. Roychowdhury (2006) showed, in his study, that more than 80% of the firms use delayed discretionary expenditures when they apply real earnings management.
Overproduction
The third way to manage earnings upward is overproduction. Firms report lower cost of goods sold through increased production. Overproduction leads to lower cost of goods and provides higher operating margins. When the production level increases, fixed overhead costs can be spread out over more product units, which automatically lowers the cost per unit and the firm can report higher operating margins. However, the firm will still incur other
production and holding costs that will lead to higher annual production costs, relative to sales, and lower cash flows from operations given sales levels ( Cohen & Zarowin, 2010).
2.2.2 Accrual-based earnings management
Accrual-based earnings management involves booking the costs and revenues as higher or lower than they actually are. A firm’s manager can choose, at the end of the year, whether to take an accrual. Accruals are reservations incorporated at the end of each financial year. They can be costs without a related cash flow, but which do relate to the financial year. Accruals are the difference between operating cash flow and net income. For instance, if a firm receives a payment in January of the following year, a reservation must be booked in the financial statements of the current year, because the associated costs are related to the current year (Dechow et al., 2000).
Scientists provide different reasons for using accrual-based earnings management. Jones (1991) stated that companies artificially increase the costs in the financial statements to reduce the profit. The purpose is to influence the government to realize an important ban to protect the own market. However, Graham et al. (2005) claimed that companies decrease their costs, reporting fewer profits than they actually earned. Therefore, the financial results will improve. Overall, accrual-based earnings management implies that a manager can manipulate the results by using accounting choices. The accounting choice depends on the
16
manager’s motive. Accrual-based earnings management can be realized by means of shifting the revenues, incorporating reservations (or not), or by not booking the credit notes in the correct year.
Managers can use different accounting policies within the rules to manipulate earnings. Accrual-based earnings management has no immediate impact on the cashflow because accruals only account for shift in time. Accruals do not affect the total profit of the firm, only the annual profit (Roychowdhury, 2006).
2.3 Hypotheses development
As explained, the central aspect of the socioemotional wealth theory is the generational sustainability of the business. Family firms attach importance to sustainable success and retention of family reputation. This leads to the concern that family firms decrease firm value over time when they engage in real earnings manipulation. My expectation is that family firms are less likely to apply REM because applying REM has a negative effect on the firm’s operational performance and contradicts the SEW theory. My hypothesis on the relationship between family firms and REM is as follows:
Hypothesis 1: The amount of REM in family firms is lower than in non-family firms.
Because of the absence of the adverse incentives arising from compensation structures such as bonuses and the conservative accounting style within family firms, I expect that family firms less often engage in accrual-based earnings management than non-family firms do. The agency theory does not apply to family firms because there is no separation of control, unless there is a manager who is not the firm’s owner. Several studies have shown that agency costs do not exist within family firms or are many times lower than in non-family firms. My hypothesis on the relationship between family firms and ABEM is as follows:
17
3.1 Sample selection
This study investigates the differences in earnings management between family firms and non-family firms. I will gather data from listed firms in the U.S. The purpose of the regression analysis is to find evidence as to whether the amount of earnings management statistically differs between listed family firms and listed non-family firms. Because the focus of this study will be on publicly listed firms, all private family firms are removed from the
sample. I will gather hand collected data because in COMPUSTAT there is no option present
where I can distinguish listed firms in family firms and non-family firms. Anderson & Reeb (2003) and Wang (2006) used the S&P index to gather data. I will do it differently. Besides the S&P index the empirical is also based on NASDAQ and the NY( New York) stock exchange. To collect as much data as possible - the bigger n the more reliable the sample - I will collect my data from the COMPUSTAT industrial files for the years 2011 to 2014. I will use the two following databases, namely COMPUSTAT executive compensation and COMPUSTAT fundamentals annual, which will be used respectively for information about the CEO and for the other information and variables. I estimate n will be approximately 700 for each year, without missing values, consist of 350 family firms and 350 non-family firms. Deleted from the sample are family firms and non-family firms of which no data are available to compute earnings management. This means that if data are missing from a firm related to year x (between 2011 and 2014), that firm will not be included in the sample.
Furthermore, I will exclude financial institutions (SICs between 6000 and 6999) and public utilities (SICs between 4900 and 4999). Financial institutions are excluded because they have incomparable accounting information and public utilities will be removed because government regulations affect firm performance (Anderson & Reeb, 2003). Those two industries will be eliminated because they will generate difficulties in the calculation of earnings management. In total, there are 2,666 firm-year observations, of which 1,315 are family firms and the other 1,351 are non-family firms. There are almost no differences between the years and the sample is almost equally distributed.
3.2 Family firms
One of my primary concerns is the identification of family firms. Many theoretical studies have shown that there is no uniform definition of a family firm. I will use publications of Campden FB to collect a sufficient number of family firms. Campden FB is a magazine that provides business-owning families of substantial wealth with the knowledge, insights and
18
vital business intelligence they require to manage their enterprises and their families successfully. Since the establishment in 1987, this magazine focuses entirely on family firms around the world. Campden FB applied the following criteria for the characteristics of a family firm:
At least one representative of the family or kin is formally involved in the governance of the company.
The founder of the family firm or the family member who acquires the business has at least 25% of the decision-making rights on the basis of his / her shares.
The share owned by the family is in the hands of the progeny of the family.
This set of criteria shows an overlap with the theoretical framework discussed in chapter 2.1 (EC, 2009). This can thus be viewed as a reliable approach for the determination of the definition of family firms. Finally, I build a binary variable that takes the value of 1 for family firms and 0 for non-family firms.
3.3 Real earnings management model
To measure the REM-activities I will use the model employed by Roychowdhury (2006) and Zhang (2012). Roychowdhury (2006) and Zhang (2012) defines real earnings management as “management actions that deviate from normal business practices, undertaken with the primary objective to mislead certain stakeholders into believing that earnings benchmarks have been met in the normal course of operations’’. Real earnings management will be examined based on abnormal cashflow from operation (CFO), abnormal production (PROD) and discretionary expenditures (EXP).
CFO
CFOt/At-1 = β0 + β1 (1 / At-1) + β2 (St / At-1) + β3 (ΔSt / At-1) + εt
CFOt = Operational cashflow from year t
At-1 = Total assets in year t – 1
St = Sales in year t
ΔSt = Mutations in net sales in year t relative to t -1
19
PROD
PRODt / At-1 = β0 + β1 (1 / At-1) + β2 (St / At-1) + β3 (ΔSt / At-1) + β4 (ΔSt-1 / At-1) + εt
PRODt
=
The sum of cost of sales in year t and the mutation in inventory in year t relative to
t-1.
At-1 = Total assets in year t – 1
St = = Net sales in year t
ΔSt= = Mutation in net sales in year t relative to t -1
The residual of the model above represents the abnormal level of the production costs.
DISEXP
DISXt / At-1 = β0 + β1 (1 / At-1) + β2 (St-1 / At-1) + εt
DISXt =
The sum of research- and development expenses , selling, general and administrative expenses
At-1 = Total assets in year t–1
St = Net sales in year t
The residual of the model above represents the abnormal level of the discretionary expenses.
The first step to examine levels of real earnings management is to calculate the normal levels of CFO, PROD and DISEXP, and compare this to the actual levels. The residuals in these three models describe the abnormal levels of real earnings management. This is based on three respective formulas described above. The three models are measuring the different REM activities. I follow Cohen et al. (2008) to construct an overall summary measure of REM. In order to capture the effects of real earnings management through all these three variables in a comprehensive measure, I compute a single variable ( REM_AGG) by combining the three individual real earnings management variables. The measure, REM_AGG, is defined as the sum of the three standardized REM metrics, REM_AGG= (R_PROD- R_CFO- R_DISEXP ). However, the three models that measure the REM activities may display opposite outcomes, whereby the effects cancel each other out on aggregate level. I will report results corresponding to the single real earnings management proxy (REM_AGG) as well as the three individual real earnings management proxies.
Negative values on R_CFO and R_DISEXP indicate an income increasing earnings management, meanwhile positive values on R_PROD indicate an income increasing earnings management. Finally, a higher level of REM_AGG indicates also an upwards earnings management.
20
3.3.1 Accrual-based management model
Accrual-based earnings management is known as a management from in which “managers try to influence reported earnings and costs with different accounting choices and policies”. To measure the total level of discretionary accruals, several accrual-based earnings management models are put forward. The Jones model was the first model to take into account the impact of the economic conditions of the firms non-discretionary accruals by taking the revenues in the model (Dechow et al. 1995). However, the Jones model does not take into account that revenues can be manipulated, and that the firms may thus be involved in earnings management through revenues manipulation. In the Jones model, revenues are seen as a non-discretionary variable, which does not serve as a possible earnings management tool. Therefore, the modified Jones model has been developed. Different from the Jones model, the modified Jones model takes the change in cashflow into account instead of the change in revenues. It is easier to manage earnings by exercising discretion over the recognition of revenue on credit sales than to manage earnings by exercising discretion over the recognition of revenue on cash sales. According to Dechow et al. (1995) the modified Jones model is the most powerful and widely used in recognizing discretionary accruals. The difference between the normal accruals and the actual total accruals (TA) of the company are the discretionary accruals (DA). I will use the modified Jones model as in (Dechow, 1995).
TAt/At-1= β0 + β1 (1/At-1) +β2 (ΔREVt-ΔRECt / At-1) + β3 (PPEt / At-1) + Et Total
accrualst =
Income before extraordinary items minus operating cash flows for year
t
At-1 = Total assets at the beginning of year t– 1
ΔREVt = Change in revenues for the year t-1 to year t
ΔRECt = Change in receivables year t-1 to year t
PPEt = Property, plant, and equipment for year t
The residual of the model above represents the abnormal level of the discretionary accruals.
In order to determine the discretionary accruals, the normal accruals have to be computed first. The modified Jones model estimates normal accruals as the change in revenue minus the change in receivables, and the level of property, plant, and equipment. These variables control the changes in accruals that are due to changes in the firm’s normal operating activities. The total of accruals that are unexplained by the firm’s economic condition are the discretionary accruals.
21
Important to mention is that a positive value indicates that discretionary accruals are higher and this means that income increasing earnings management has taken place. A high amount of DACC indicates lower-quality earnings and is a red flag for aggressive accounting to manipulate earnings upwards. A negative value, however, indicates that discretionary accruals are lower.
3.4 Empirical model
Based on the following regression model I have tested hypothesis 1: ‘’The amount of REM in family firms is lower than in non-family firms’’:
Model 1: REM summary measure
REM_AGGit = β0 + β1 F_Firmit + β2 LOSSit + β3 SIZEit + β4 ROAit + β5 LEVit + β6 CEOit
+ εt.
Model 2 : CFO
R_CFOit= β0 + β1 F_Firmit + β2 LOSSit + β3 SIZEit + β4 ROAit + β5 LEVit + β6 CEOit + εt.
Model 3 : DISEXP
R_DISEXPit= β0 + β1 F_Firmit + β2 LOSSit + β3 SIZEit + β4 ROAit + β5 LEVit + β6 CEOit +
εt.
Model 4: PROD
R_PRODit= β0 + β1 F_Firmit + β2 LOSSit + β3 SIZEit + β4 ROAit + β5 LEVit + β6 CEOit +
εt..
Ultimately, REM_AGG will be leading in answering hypothesis 1 and subsequently, I will split it up in three proxies. Based on my hypothesis, I expect that the value of family firm in model 1 will be negative, model 2 and model 3 also will be positive and finally, model 4 will be negative.
Furthermore, based on the following regression model I have tested hypothesis 2: “The amount of ABEM in family firms is lower than in non-family firms.”
22
Model 5 : Discretionary accruals
DACCit = β0 + β1 F_Firmit + β2 LOSSit + β3 SIZEit + β4 ROAit + β5 LEVit + β6 CEOit + εt..
H2 predicts a negative relationship between family firms and DACC, hence I expect a negative value of family firms in model 5.
Variable definition:
REM_AGGit = Aggregate REM variables (CFO, PROD and DISC_EXP)
R_CFOit = Abnormal cashflow from operations for firm I and year t
R_DISEXPit = Abnormal discretionary expenses for firm I and year t
R_PRODit =Abnormal production costs for firm I and year t
DACCit = Discretionary Accruals for firm I and year t
F_FIRMit = Dummy variable which has a value of 1 if it meets the criteria of a family
firm
LOSSit = Dummy variable which has a value of 1 if the net income is negative
SIZEit = Total assets at the end of year t
ROAit = Return on assets; net income divided by total assets at the end of year t
LEVERAGEit = Leverage for firm I at the end of year t; calculated total debt in relation to
total assets
CEOit = Dummy variable for firm I, which has a value of 1 if the CEO in year t is
in his/her first or last year of his/her function
3.5 Control variables
I will introduce several control variables into my analysis in order to control the factors that can influence the level of earnings management. I follow the studies of Wang (2006) and Achleitner et al. (2014) and I will introduce five control variables: LOSS, SIZE, ROA, LEVERAGE, and CEO. These variables refer to control variables which other factors that may affect the level of earnings management are monitored. Many studies have demonstrated the usefulness of these control variables. By controlling for LOSS the effects of managers who adjust earnings downwards become mitigated because there is a ‘’lost period’’ and the profits are shifted to the next period to increase success in the next period. Therefore, I expect that LOSS has a positive relation to earnings management. SIZE is expected to have a negative relation to earnings management, because big listed firms normally have a strong governance structure and they are often scrutinized by financial analysts. As result of that, it
23
is expected that there is less information asymmetry present in big listed firms. The third control variable is return on assets (ROA). ROA is a measure that provides insights into the profitability of a firm. Both Zang (2012) and Roychowdhury (2006) included ROA as control variable. ROA controls for the potential effect of profitability on the choices of discretionary accruals. This variable has a raising effect on earnings management. The expectation is that ROA is positively related to earnings management. The fourth control variable LEVERAGE is used to control firms that are relatively highly leveraged. Expected is that strongly leveraged firms come close to violating debt covenants, and that they will want to renegotiate these debts in order to obtain new favourable conditions that adjust the profits downwards. LEVERAGE is expected to have a negative relation to earnings management. Finally, the fifth control variable is CEO. It is expected that the more powerful the CEO is, the more likely it is that the CEO commits to earnings management. CEO’s who are in their first or final year have been known to use earnings management for adjusting profit respectively downwards and upwards. CEO’s can then argue the poor performance of their first year is caused by the previous CEO, whereas the upward adjustment in their final year will maximize their bonus.
24
4.1 Descriptive statistics
Family firms
Variable N Mean Median St. Dev Q1 Q3
REM_AGG 1315 0.3807 0.2661 0.71468 -0.074 0.7036 R_CFO 1315 0.104 0.1041 0.09837 0.0645 0.1532 R_DISEXP 1315 0.3053 0.2636 0.22436 0.1361 0.4209 R_PROD 1315 0.7974 0.5921 0.68366 0.3237 10.835 DACC 1315 -0.0037 -0.0008 0.0632 -0.0306 0.0244 Control Variables LOSS 1315 0.152 0.000 0.3592 0.000 0.000 SIZE 1315 34.355 33.908 0.70653 29.474 39.027 ROA 1315 0.0485 0.0587 0.10595 0.0252 0.0944 LEVERAGE 1315 0.5219 0.5066 0.22027 0.3798 0.6457 CEO 1315 0.092 0.000 0.2892 0.000 0.000
Table 1 descriptive statistics family firms
Non-family firms
Variabel N Mean Median St. Dev Q1 Q3
REM_AGG 1351 0.4163 0.2743 0.67256 -0.0017 0.7149 R_CFO 1351 0.0867 0.0924 0.10383 0.0419 0.143 R_DISEXP 1351 0.3036 0.2613 0.24806 0.1096 0.4315 R_PROD 1351 0.8084 0.639 0.65528 0.3049 11.013 DACC 1351 0.004 0.0021 0.06688 -0.0292 0.0306 Control Variables LOSS 1351 0.201 0.000 0.4006 0.000 0.000 SIZE 1351 30.309 2.987 0.69467 26.028 34.567 ROA 1351 0.0329 0.0483 0.11625 0.0123 0.0846 LEVERAGE 1351 0.5067 0.4894 0.24673 0.3284 0.6491 CEO 1351 0.067 0.000 0.2495 0.000 0.000
Table 2 descriptive statistics non-family firms
In this section, I will discuss the descriptive statistics, the independent t-test and a correlation analysis. Table 1 shows the descriptive statistics of the family firms and table 2 shows the descriptive statistics of non-family firms. These tables show a research population of almost 50/50. Based on the two tables there are several differences between family firms and non-
25
family firms. To examine the significance of the means of these differences, I will perform a t-test.
First of all, the REM values (REM_AGG) show positive values for both samples. The non-family firm proxy shows a higher positive value, which indicates that non-family firms use more income increase real earnings management. The mean and median for family firms are 0.3807 and 0.2661 relative to 0.4163 and 0.2743 for non-family firms. The REM results are in line with hypothesis 1, because the mean and median of the family firms are lower than those of non-family firms. The differences, however, are minimal.
Furthermore, as mentioned before, I divided the REM summary measure into three proxies to measure each proxy on single level. Family firms have positive abnormal cashflows, positive discretionary expenses and positive abnormal production costs - and non-family firms also have positive abnormal cashflows, positive discretionary expenses and positive abnormal production costs. Also for these three proxies, the differences are minimal
Furthermore, The ABEM (DACC) proxy of family firms shows negative discretionary accruals, while non-family firms show positive discretionary accruals. The mean and median for family firms are -0.0037 and -0.0008 relative to 0.004 and 0.0021. The positive values indicate that earnings are managed upwards in the non-family sample. Assuming that on average managers have incentives to increase income, this result seems consistent with less opportunistic behavior in family firms. This is in line with hypothesis 2.
With regard to the control variables, the table first shows that family firms on average report less LOSSES than non-family firms do. In general, the more LOSSES a company reports, the greater the chance of earnings management. Second, the values of SIZE show that family firms on average report more total assets than non-family firms do. Third, family firms report a higher ROA-ratio, which indicates that family firms are more profitable. Moreover, family firms on average report a higher level of LEVERAGE. Consistent with Tong (2008), this could mean that family firms use high levels of debt to reduce free cashflow, which is very interesting for small outside dispersed shareholders. Finally, the CEO value is also bigger, which means that more CEO changes are taking place within family firms. This may be driven by several factors.
Second, I performed a t-test to measure the significant differences in means. It can be read from the table that of the means of the dependent variables, only R_CFO and DACC are significantly related to each other. This means that family firms engage less significantly in ABEM than non-family firms do and that family firms have significantly higher abnormal cashflows than operations of non-family firms. For the other variables, there are no
26
significant differences between both means.
With regard to the control variables, they are almost all significant, except for LEVERAGE. Inconsistent with Achleitner et al. (2014), family firms are less likely to report a loss than non-family firms. After all, paying a compensation to a creditor is cheaper than paying a compensation to an outside shareholder. Further, the means of the variables ROA and SIZE are significantly related to each other, which indicates that family firms are more bigger in terms of total assets and they are more profitable. Finally, the means of the variable is CEO also significant related to each other, which indicates that significantly more CEO changes are taking place within family firms. It seems that hypothesis 1 will be rejected and hypothesis 2 will be accepted. With respect to the hypotheses, no conclusions can be made, since the analysis above was performed without the input of control variables. Therefore I will add a multivariate analysis in section 4.2.
Independent Samples Test
Significance level. (2-tailed) Mean Difference REM_AGG 0.162 -0.03755 R_CFO .000*** 0.0185 R_DISEXP 0.582 0.00503 R_PROD 0.645 -0.01191 DACC 0.002*** -0.00786 LOSS 0.001*** -0.0472 SIZE .000*** 0.40013 ROA .000*** 0.01598 LEVERAGE 0.144 0.01323 CEO 0.019** 0.0245
***. Correlation is significant at the 0.01 level (2-tailed). **. Correlation is significant at the 0.05 level (2-tailed). *. Correlation is significant at the 0.1 level (2-tailed).
Table 3 Tests for significant differences in means
Finally, I conducted a correlation test to explore the correlations between the independent variable and dependent variable. The correlations give an indication of the linear relationship between the two variables. If high scores on one variable match with high scores on the other variable, there is a positive correlation between the two variables. If high scores on one variable match with low scores on the other variable, then there is a negative correlation
27
between the two variables. The scores are between 1 and -1. I also computed the multicollinearity. The results are not of a concern because none of the VIFs have a value
above 10. I winsorized the sample at the 1st and 99st percentile. From the correlation table
below, the results show that the family firm indicator has a negative significant correlation to discretionary accruals and a positive significant correlation to abnormal cashflows from operation. Meanwhile, there is no significant correlation between family firm and abnormal production costs, abnormal discretionary expenses and the REM summary measure.
Consistent with the study of Achleitner et al. (2014), I found a negative significant correlation between LEVERAGE and discretionary accruals, which indicates that discretionary accruals goes down when LEVERAGE increases. Debt is used to reduce the level of free cashflow. This also reduces rent extraction opportunities by managers. Finally, as expected, it can be observed that the dependent variables have a significant correlation to most of the other control variables. These results reported in table 4 illustrate the importance to control for these factors when estimating the relation between family firms and REM/ABEM.
Correlationsc
DACC R_CFO R_PROD R_DISEXP REM_AGG F_FIRM SIZE ROA LEVERAGE LOSS CEO
DACC 1 R_CFO -.389*** 1 R_PROD .061** 0.034 1 R_DISEXP -.128*** -0.027 .070*** 1 REM_AGG .166*** -.108*** .924*** -.276*** 1 F_FIRM -.059*** .085*** -0.008 0.003 -0.026 1 SIZE -.049** .329*** -.040*** -.312*** 0.019 .278*** 1 ROA .179*** .684*** .066*** -.138*** 0.009 .070*** .368*** 1 LEVERAGE -.046** -.143*** .149*** -.133*** .206*** 0.032* .272*** -.151*** 1 LOSS -.123*** -.496*** -0.032* .165*** -0.015 -.064*** -.355*** -.693*** .113*** 1 CEO -0.025 0.018 -0.006 -0.016 -0.002 .047** .069*** -0.01 0.027 0.024 1
***. Correlation is significant at the 0.01 level (2-tailed). **. Correlation is significant at the 0.05 level (2-tailed). *. Correlation is significant at the 0.1 level (2-tailed).
28
4.2 Multivariate analyses
In the next chapter, the results of the regression analyses will be discussed, and the two hypotheses will be tested. The multivariate analyses are used to examine differences in earnings management between family and non-family firms controlling for firm and manager characteristics. The results of all five regressions analyses are incorporated in table 5.
REM_AGG R_CFO R_DISEXP R_PROD DACC
B Sig B Sig B Sig B Sig B Sig
(Constant) 0.0462 .000*** 0.043 .000*** 0.666 .000*** 1.185 .000*** 0.033 .000*** F_FIRM -0.017 0.53 0.003 0.343 0.046 .000*** 0.04 0.124 -0.005 LOSS -0.073 0.129 -0.005 0.334 0.047 0.003*** -0.029 0.526 -0.004 0.374 SIZE -0.144 .000*** 0.014 .000*** -0.116 .000*** -0.255 .000*** -0.012 .000*** ROA 0.34 0.041** 0.564 .000*** 0.031** 0.576 0.969 .000*** 0.128 .000*** LEVERAGE 0.733 .000*** -0.033 .000*** -0.055 0.005*** 0.664 .000*** 0.008 0.19 CEO -0.008 0.875 0.007 0.209 -0.001 0.967 -0.005 0.916 -0.003 0.518 R 7.36% 48.11% 12.66% 9.67% 5.01% Adjusted R 7.12% 47.97% 12.43% 9.43% 4.76% N 2666 2666 2666 2666 2666
***. Correlation is significant at the 0.01 level (2-tailed). **. Correlation is significant at the 0.05 level (2-tailed). *. Correlation is significant at the 0.1 level (2-tailed).
Table 5 multivariate analyses
Hypotheses 1 REM
In this session I will discuss the results of the four regressions with regard to REM. First of all I will discuss R_CFO, subsequently R_DISEXP, third R_PROD and finally, the summary measure REM_AGG.
The results of the regression analysis, in which I tested the abnormal cashflow as a single measure, contradicts my expectations. From the table it can be read that the adjusted R-squared is 47.97%. The family firm value is positive, which indicates that family firms have higher abnormal cashflows from operations than non-family firms do. This indicates that family firms use less sales manipulation to manage earnings. Roychowdhury (2006) states that sales manipulation through reduces prices, higher price discounts and lenient credit terms decreases cashflow from operations. However, the p-value is not significant, which indicates that there is no evidence that family firms engage less in real earnings management
29
with regard to abnormal cashflows form operations than non-family firms do. Moreover, the control variables SIZE, ROA and LEVERAGE in this model are significantly related to abnormal cashflow from operations. Overall, no statistically significant relation between family firm and abnormal cashflow from operations was found.
Column R_DISEXP displays the results of the discretionary accruals. From the table it can be read that the adjusted R-squared is 12.43%. This is in line with my expectations. The value family firms is positive and significant at a 1% level. This indicates that family firms with higher discretionary expenses manipulate their discretionary expenses less than non- family firms do. This means that discretionary accruals within family firms are not decreased to lower cash outflows and do not manage earnings upwards. All control variables, at least ROA and CEO, are significant. Overall, I found a statistically significant relation between family firm and discretionary expenses.
The final individual proxy, R_PROD, can be found in the penultimate column of the table. The R-squared is 9.43%. The value of family firm is positive, which indicates that family firms have higher abnormal production costs. However, the p-value is not significant and there is no evidence that family firms engage in more real earnings management with regard to abnormal production costs than non-family firms. Also, in this model, most of the control variables are significant – except for CEO and LOSS. Overall, no statistically significant relation was found between family firm and abnormal production costs.
Finally, in this session I will discuss the REM summary measure. The previous session, in which I divided the REM measure into three proxies, indicated a contradictory view of my expectations. Only the proxy R_DISEXP was in accordance with my hypothesis and for R_CFO there was no evidence because the p-value was not significant. Meanwhile, R_PROD contradicted my expectations, but also for this proxy no evidence was found, because the p-value was insignificant. Taken together, it can be seen that the adjusted R-squared with regard to the REM measure is 7.12%, which indicates that 7.12% of the variation in REM is explained by the model. When the individual variables are viewed, it appears that the family firm variable is negative. However the p-value is 0.530. This means that there is no evidence that family firms engage less in real earnings based earnings management than non-family firms. This is not consistent with the study of Achleitner et al. (2014). They found that the variable family firms is negative and significant in the German context. This indicates that family firms in Germany use less real earnings management than non-family firms do, despite of the weaker investor protection in Europe (Leuz et. al 2003). From the results in the U.S, I have not found any evidence. An explanation for this may be
30
that the family firms are so large that there is a similarity in characteristics of non-family firms, which affect in the way earnings management is conducted – they become similar to the investigated non-family firms. Overall, the results are not in line with the socioemotional wealth theory. The results are not entirely strange, because Razzaque et al. (2016), found in their research that family firms even use significantly more real earnings management than non-family firms do. However, this research is conducted in a developing economic context, namely Bangladesh, which bears the characteristics of a relatively weaker investor protection regime including underdeveloped capital markets, with a dominant presence of family ownership. The possibilities of private rent seeking activities of family owners are higher with the existing frailties in Bangladesh's governance environment. This study, however, has been conducted in a strong investor environment. Another reason may be that many family members are involved in the management of listed family firms, which leads to non-significant differences between family firms and non-family firms. Finally, the research of Cohen et al. (2008) showed that, after the introduction of the SOX legislation REM gradually increased in the U.S. This has probably had an impact on the results. After the introduction of the Sarbanes-Oxley (SOX) legislation chief executive officers and chief financial officers could be held personally liable for mismanagement ( Cohen et al., 2008). Since REM is harder to be detect, this is a logical consequence.
In addition, most of the control variables are significant and in accordance with my expectations. The control variable SIZE is negatively significant. This is in line with the expectations. Large firms generally have a strong governance structure, more strict supervision and there is less information asymmetry. The coefficient ROA is also in line with the expectations. The variable is positively significant, which indicates that profitable firms report more real earnings activities based earnings management to meet expectations of financial analysts and investors. The third variable LEVERAGE is also positively significant. This indicates that strong leveraged firms in a renegotiation process adjust the profits to gain the maximum result. Meanwhile, the variables CEO and LOSS have no significant relation to REM. Overall, no statistically significant relation was found between family firm and real earnings management.
31
Hypotheses 2 ABEM
The results of the regression analyses, which hypothesis 2 is tested are included under the column DACC. The R-squared of the model is 4.76%, which indicates that 4.76% of the variation in the model in the discretionary accruals is explained by the model. When the individual variables are viewed, it appears that the family firm variable is also significant. This means that the hypothesis is confirmed and in accordance with the expectations. Family firms use less ABEM than non-family firms do. Higher quality financial statements with regard to ABEM are in line with the studies of Wang (2006), Ali et al., ( 2007) and Jiraporn and Dadalt (2009).
With regard to the control variables: SIZE and ROA are significant. The control variable SIZE is in line with the expectations; as the size of a company increases, ABEM decreases. Furthermore, ROA is also in accordance with the expectations: highly profitable firms report more discretionary accruals. Consistent with Wang (2006), my findings are that firms that report a LOSS have greater discretionary accruals. However, no evidence is found for this assumption. The signs of these coefficients are consistent with prior studies, such as Wang (2006). Furthermore, the value LEVERAGE is positively related to ABEM, which indicates that strong leveraged firms adjust the profit, in order to obtain the best possible result in a renegotiation process. But also this control variable is not confirmed by evidence. Finally, the value CEO is negative, but also this variable is not confirmed by evidence. Remarkable in this study is that the variable CEO is not significantly related to either REM or ABEM. This probably indicates that newly appointed and abandoned chief executive officers are scrutinized in their first or final year. Furthermore, based on the socioemotional wealth theory, board chairmen of family firms in charge of the future performance of the firm will have no incentive in their first of final year to engage in earnings management. This may have had an impact on the variable CEO as a whole, which explain the reason why it is not significantly related to earnings management.
Overall, family firms have lower levels of ABEM compared to non-family firms. This is in line with hypothesis 2.
32
4.3 Robustness test
In this section, I will perform an additional test to examine whether the results are robust. In the first robustness test, I will control the model for year fixed effects, to capture the influence of aggregate time-series trends. The differences in years that could lead to more or less earnings management is taken out from the model. The results of this additional analysis are tabled in table 6. I expect that the results remain the same.
REM_AGG R_CFO R_DISEXP R_PROD DACC
B Sig B Sig B Sig B Sig B Sig
(Constant) 0.461 .000*** 0.041 .000*** 0.671 .000*** 1.187 .000*** 0.035 .000*** F_FIRM -0.017 0.529 0.003 0.341 0.046 0.000*** 0.04 0.124 -0.005 0.046** LOSS -0.072 0.132 -0.005 0.34 0.048 .003*** -0.028 0.542 -0.004 0.374 SIZE -0.145 .000*** 0.014 .000*** -0.116 .000*** -0.256 .000*** -0.012 .000*** ROA 0.342 .040** 0.565 .000*** 0.033 0.55 0.974 .000*** 0.128 .000*** LEVERAGE 0.734 .000*** -0.033 .000*** -0.055 .006*** 0.665 .000*** 0.008 0.184 CEO -0.006 0.898 0.008 0.168 -0.004 0.825 -0.006 0.908 -0.004 0.414 R 7.38% 48.17% 12.68% 9.70% 5.14% Adjusted R 7.03% 47.97% 12.35% 9.36% 4.79% N 2666 2666 2666 2666 2666
***. Correlation is significant at the 0.01 level (2-tailed). **. Correlation is significant at the 0.05 level (2-tailed). *. Correlation is significant at the 0.1 level (2-tailed).
Table 6 Robustness test regression analyses
As expected, the results have remained the same. In addition, the control variables also show
the same significant values. Hence, hypothesis 1 remains insignificant and hypothesis 2 remains significant. From this, it can be concluded that the results are robust with respect to the inclusion of year fixed effects.
