T he effect of financial distress on real earnings management and accrual earnings management and the impact of CEO tenure on this relationship

T

he effect of financial distress on real earnings

management and accrual earnings management and the

impact of CEO tenure on this relationship

To what degree does financial distress influence the degree of earnings management and what is the influence of CEO tenure on this relationship?

Josefien Kooistra

ABSTRACT: The purpose of this study is to examine the effect of financial distress on earnings management and the impact of CEO tenure on this relationship. Based on an analysis of 1017 US publicly listed companies, I find a negative relationship between financial distress and accrual earnings management. This suggests that financially distressed companies engage less in accrual earnings management. Moreover, I find a positive relationship between financial distress and real earnings management. This implies that financially distressed companies are managing earnings through real activities. However, I do not find a significant moderating effect of CEO tenure. Nevertheless, I do find that CEOs in the later years of their tenure engage in real earnings activities. The results indicate that companies, despite regulations to avoid earnings management, are still managing earnings through real earnings management.

Keywords: financial distress, earnings management, real earnings management, accrual earnings

management, CEO tenure, covenant violation. Master Thesis MSc Accountancy

Supervisor: Shuo Wang 24-06-2019 11817 words Molenerf 43, Buitenpost s.j.kooistra.2@student.rug.nl 06 20 30 54 45 S2961636

1

INTRODUCTION

ccounting rules and principles enable managers to make judgments and estimations within the lines. This freedom in financial reports allows managers to present a positive view of the financial position of the company. This will influence the decision-making of the users of the financial statements because earnings are the most important element of the financial statements for analysts, investors, creditors, regulators and researchers (Tabassum,Kaleem, and Nazir, 2015). Recent studies argue that this effect is strengthened when companies are in financial distress (Ahmadpour and Shahsavari, 2016; Campa and Camacho, 2015; Adu-Boateng, 2011). Financially distressed companies have stronger incentives to manipulate earnings than healthier firms because they want to conceal the negative signals of financial distress (Campa and Camacho, 2015; Adu-Boateng, 2011). Earnings management can thus ensure that financially distressed companies report a better income than they actually have (Ahmadpour and Shahsavari, 2016).

However, research does not give an unequivocal answer about the relationship between financial distress and accrual and real earnings management. One the one hand, there is a stream of research that argues that financially distressed companies manipulate earnings through accrual earnings management (AEM). AEM refers to the use of legal discretion to manage accounting methods and accounting estimates to influence the reported earnings without alteration in the underlying operations of the company (Walker, 2013). Joosten (2012) argues that financially distressed companies that are focusing on real earnings management have adverse economic consequences. Therefore, she suggests that financially distressed companies should focus on AEM because there is no alteration in the underlying operations of the company and this has no adverse consequences for the company (Joosten, 2012). On the other hand, there is a stream of research that indicates that financially distressed companies should engage in real earnings management (REM). REM is defined as managing earnings through operational decisions (Walker, 2013). In this way, managers are changing the level of cash flows (Walker, 2013). Xu, Dao, and Wu (2018) suggest that financially distressed companies often have insufficient cash and liquidity and, therefore, inflate earnings by changing the level of cash flow, thus engaging in REM. Consequently, companies can accelerate cash flows and improve their performance and liquidity. Moreover, Graham, Harvey, and Rajgopal (2005) argue that through the implementation of the Sarbanes-Oxley Act (SOX) there is more auditor and regulatory scrutiny on AEM and less on REM, and that manager, therefore, prefer REM. However, there is also a stream of research that

2

suggests that AEM and REM are complementary. Franz, HassabElnaby, and Lobo (2014) argue that earnings management objectives cannot only be achieved through REM. They argue that firms have a lack of economic resources to engage in REM and therefore also perform discretionary accrual activities (Franz et al., 2014). The findings of these studies imply that there is not a straightforward answer to the relationship between financial distress and AEM and REM. Therefore, the aim of this study is to examine whether financially distressed companies engage in accrual, real, or total earnings management (TEM).

The pressure that companies face during financial distress can increase the likelihood of Chief Executive Officers (CEOs) to influence firm outcomes (Besancenot and Vranceanu, 2009). However, financial distress is not the only incentive managers have to engage in earnings management. Earnings management may be used to avoid debt covenant violations (Obeidat, 2016; Scott, 2015), to increase CEO compensation (Dechow and Sloan, 1991), to avoid reporting losses (Burgstahler and Dichev, 1997) or because CEOs have to prove themselves to the shareholders. CEOs may have the fear to be replaced, which results in reputation damage and negative career perspectives (Magro, Klann, and Mondini, 2018; Charitou, Lambertides, and Trigeorgis, 2007; Ali and Zhang, 2015). Although CEOs have different incentives to engage in EM, those incentives and the opportunities for EM differ during the tenure of the CEO. For example, Ali and Zhang (2015) argue that CEOs in the early years of their tenure manipulate earnings more than CEOs in the later years of their tenure because they want to enhance the market perceptions of their ability. Moreover, CEOs in the early years of their tenure engage more in earnings management due to the fact that they are more afraid of losing their jobs (Magro et al., 2018). However, other researchers argue that CEOs in the later years of their tenure have more opportunities to manage earnings. For example, Wang, Holmes, Oh, and Zhu (2016) claim that CEOs in the later years of their tenure manage earnings more because they have more knowledge and experience within the firm to improve financial performance. Furthermore, Dechow and Sloan (1991) find that CEOs at the end of their tenure manage earnings in order to increase their financial year’s compensation. The findings of these studies indicate that CEOs in the early years of their tenure have more incentives to engage in earnings management, while CEOs in the later years of their tenure have more opportunities to manage earnings.

Prior research states that the pressure of financial distress will influence the behavior of the CEO towards earnings management (Ghazali, Shafie and Sanusi, 2015; Besancenot and

3

Vranceanu, 2009). However, previous literature has not yet investigated the effect of CEO tenure on the relationship between financial distress and earnings management. Therefore, I also examine what the effect of CEO tenure is on the relationship between financial distress and earnings management.

The research question of this paper is: To what degree does financial distress influence the

degree of earnings management and what is the influence of CEO tenure on this relationship? The

analysis is based on financial accounting data from US publicly listed firms from a period from 2003 to 2016. The findings of this study have implications for investors considering earnings management behavior in financially distressed firms. Investors should be aware of the behavior of management to manipulate earnings when the company is in financial distress. Therefore, a better understanding of earnings management behavior in financially distressed companies will assist them in their investment decision making process.

This research contributes to the existing literature in the following ways. First, as mentioned before, extant literature does not give a straightforward answer about the relationship between financial distress and AEM and REM (Muljono and Suk, 2018). Therefore, this research contributes to the existing literature by examining the impact of financial distress on AEM, REM, and TEM.

Second, this research contributes to the extant literature by examining the effect of CEO tenure on the relation between financial distress and earnings management. As mentioned before, the incentives and opportunities to engage in earnings management differ during the tenure of the CEO and may differ when the company is in financial distress. Nevertheless, prior research has not yet investigated the effect of CEO tenure on the relation between financial distress and earnings management. Therefore, I contribute to the existing literature by examining the impact of CEO tenure on the relationship between financial distress and earnings management.

Third, I have taken a longer time period (2003 to 2016) than prior research. Both the SOX, which was implemented in 2002, and the financial crisis of 2008 are part of this research. Prior research only examined the effects of the SOX or the financial crisis.

The remainder of this paper proceeds as follows. In the next section, I provide an overview of the relevant literature and the development of the hypothesis. In section three and four, I describe the sample selection, research design, and the results. In the final section, I provide a conclusion and discussion of the findings.

4

THEORETICAL BACKGROUND Agency theory as the basis

According to Healy and Wahlen (1999) “earnings management occurs when managers use

judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on earnings accounting numbers”. Earnings

management thus emerges when management influence reported accounting information. Managers often use earnings management for their own private benefit. This behavior can derive due to information asymmetry. Therefore, a useful theoretical basis for this research is agency

theory. This theory addresses problems that arise due to conflicts of interest between the principal

and the agent (Jensen and Meckling, 1976). According to this theory, companies operate under uncertain conditions and this results in information asymmetries between managers and external investors (Walker, 2013). As a result, managers engage in earnings management to benefit from this information asymmetry.

Earnings management can be divided into AEM and REM. AEM refers to the legal discretion to manage earnings through accounting choices, such as depreciation methods, and accounting estimates, for example, write-offs (Walker, 2013; Zang, 2012). The operating income can be divided into a fixed component, free cash flows, and a subjective component, accruals (Walker, 2013). This subjectivity in accruals gives rise to influence future accounting numbers (Allen, Larson and Sloan, 2012).

Moreover, earnings can also be influenced by changing the level of cash flow (i.e. REM). These real cash flow consequences are due to the fact that firms change economic decisions to achieve financial targets (Walker, 2013). Examples of REM activities are cutting research and development (R&D) expenditures, sales of profitable assets, and temporary price discounts. These activities can impose higher long-term costs on shareholders than is the case with AEM because REM has negative consequences for future cash flows and the long-term firm value (Cohen and Zarowin, 2010; Graham et al., 2005). However, managers prefer REM as a consequence of the SOX. Regulators want to discourage earnings management and, therefore, they implemented SOX (Cohen and Zarowin, 2010). As a result, they detect earnings management through AEM and not

5

through REM because AEM is easier to detect (Graham et al., 2005). As a consequence, managers manage earnings through REM because the likelihood of being detected is lower (Franz et al., 2014; Graham et al., 2005).

Financial distress and earnings management

According to Wruck (1990), financial distress is a situation in which firms are not able to discharge financial and other obligations, which increases the likelihood of default or bankruptcy. The pressure of financial distress will influence management’s behavior towards earnings management (Ghazali et al., 2015). Prior research argues that managers of financially distressed firms have stronger motivations to manipulate earnings that firms that are more healthy (Adu-Boateng, 2011; Burgstahler and Dichev, 1997). Especially, subsequent to bankruptcy procedures, managers manipulate earnings upwards and conceal the negative signals of financial distress in their financial reporting (Ahmadpour and Shahsavari, 2016; Campa and Camacho, 2015). Unfavorable outcomes can, for instance, lead to operating difficulty, financial penalties, and loss of support from important stakeholders (Sweeney, 1994; Opler and Timan, 1994; Purnanandam, 2008).

Financial distress situations thus have significant consequences. Moreover, lenders can suffer financial loss as a consequence of financial distress. For this reason, debt contracts include debt covenants to protect lenders. These covenants are agreements between the company and the lender and contain restrictions for specific financial ratios that the company may not violate (Bhaskar, Krishnan, and Yu, 2017). The primary objective of covenants in contracts is to mitigate agency problems (Jensen and Meckling, 1976) because the information asymmetry between managers and lenders will be constrained through the covenants in the contracts (Franz et al., 2014).

According to the debt covenant hypothesis, firms that are close to violating debt covenant manipulate earnings in such a way that future earnings will be shifted to the current period (Chamberlain, Butt and Sarkar, 2014). Extant literature argues that companies have motivations to engage in earnings management to reduce the costs of covenant violation (Franz et al., 2014; DeFond and Jiambalvo, 1994; Sweeney, 1994). Moreover, when the company is in financial distress, lenders can impose costly actions on the firm, such as increasing the interest rate or adding more covenants to the contract (Bhaskar et al., 2017; Franz et al., 2014). Financially distressed

6

firms can prevent these actions and reduce the probability of covenant violation by engaging in earnings management (Franz et al., 2014; Sweeney, 1994). This implies that financially distressed companies have more incentives to manage earnings to prevent costly actions and avoid covenant violation (Franz et al., 2014; Adu-Boateng, 2011). Consequently, debt covenants provide an early warning of financial distress for lenders (Franz et al., 2014). Moreover, Franz et al. (2014) argue that earnings management objectives cannot only be achieved through discretionary accruals or only through real manipulation activities. Companies with a lack of economic resources engage in both REM and AEM to achieve their earnings objectives (Franz et al., 2014). Therefore, Franz et al. (2014) argue that AEM and REM are complements.

Based on the above literature, it can be concluded that financially distressed companies engage in a higher level of earnings management to avoid covenant violation. Therefore, the following hypothesis is presented:

Hypothesis 1a: Firms in financial distress are more likely to engage in total earnings management than firms that are not in financial distress.

Financial distress will thus influence management’s behavior towards earnings management. However, prior research does not give a straightforward answer about whether financially distressed companies engage in AEM or REM (Muljono and Suk, 2018). As mentioned before, Franz et al. (2014) claim that AEM and REM are complements to each other. However, in contrast to Franz et al. (2014), Zang (2012) argue that REM and AEM are substitutes. The findings of Zang (2012) show that REM is positively related to the costs of AEM and REM and AEM are negatively correlated. Therefore, she argues that REM and AEM are substitutes. This is consistent with the findings of Cohen and Zarowin (2010).

Moreover, there is also a stream of research that argues that in the higher stages of financial distress companies engage more in AEM. Joosten (2012) investigated publicly listed firms in Europe and finds, when financial distress is low, companies will engage in REM to achieve earnings objectives because the capabilities of the companies allow them to weaken the adverse economic consequences of REM. However, when companies are in financial distress, they engage in higher levels of AEM, since taking distance from optimal business operations has than positive effects (Joosten, 2012). She argues that manipulating earnings by REM in financially distressed

7

companies can lead to adverse economic consequences. This is consistent with Nagar and Sen (2016), who find that firms in the initial stages of financial distress manage earnings through REM activities and in high stages of financial distress they engage in AEM. They argue that this difference is the result of CEOs of financially distressed firms who have to make a trade-off between the impact on the short-term and on the long-term with regard to liquidity, profitability, and solvency (Nagar and Sen, 2016). Concluding, this stream of research argues that financially distressed companies engage in AEM.

In contrast to the above literature, Xu et al. (2018) argue that financially distressed companies improve their performance and liquidity through REM because these companies often have insufficient cash and liquidity. Financially distressed firms can use different techniques to accelerate cash flows. For example temporarily offering discounts to customers to boost sales and cutting down R&D expenditures to improve their margins (Roychowdhury, 2006). Roychowdhury (2006) concludes that in this way, managers conceal their losses. Moreover, García et al. (2009) claim that managers see REM as more costly, and use it as the last action to lift up. This is consistent with the findings of Campa and Camacho (2015), who find that managers prefer REM over AEM when the company is close to facing bankruptcy, even if it produces costs for the company on the long term (Campa and Camacho, 2015). Through the pressure managers experience in situations of financial distress, managers use real manipulation activities to achieve immediate results to accomplish specific objectives (Campa and Camacho, 2015). They argue that REM is the ‘last chance’ to avoid bankruptcy. Moreover, managers prefer REM because the change of detection is lower in comparison to AEM (Franz et al., 2014; Graham et al., 2005).These results indicate that in the higher levels of financial distress firms manipulate earnings through real activities.

Consequently, research gives not a conclusive answer on the relationship between financial distress and REM and AEM. Based on the above literature, the following hypotheses will be tested:

Hypothesis 1b: Firms in financial distress are more likely to engage in accrual earnings management than firms that are not in financial distress.

Hypothesis 1c: Firms in financial distress are more likely to engage in real earnings management than firms that are not in financial distress.

8

CEO Tenure

“CEO tenure is the number of years that a CEO continuously holds the position in a company” (Baatwah, Salleh and Ahmad, 2015). The upper echelons theory suggests that organizational outcomes are dependent on management characteristics (Hambrick and Mason, 1984). The values, personalities, and experiences of CEOs determine their interpretation and choices (Hambrick, 2007). However, their experiences and thus their incentives differ during their tenure (Ali and Zhang, 2015; Cella, Ellul, and Gupta, 2014). Therefore, decisions, and also the decisions with regard to engaging in earnings management, can be influenced by CEO tenure (Cella et al., 2014). Prior research shows that CEOs in the early years of their tenure have other incentives to report positive outcomes than CEOs in the later years of their tenure and this will impact whether they engage in earnings management. For example, Ali and Zhang (2015) investigated the tenure of CEO in a period from 1992 to 2010 and find that CEOs in the early years of their tenure report higher discretionary expenses and lower abnormal discretionary expenses, such as R&D expenditures (Ali and Zhang, 2015). This indicates that CEOs in the early years or their tenure are manipulating earnings more than CEOs in the later years of their tenure (Ali and Zhang, 2015). They claim that this is due to the fact that CEOs in the early years of their tenure want to enhance the market perceptions of their ability (Ali and Zhang, 2015).

Axelson and Bond (2009) claim that there is adverse selection at the beginning of the tenure of the CEO and they show that managers are seen as “low ability” managers if they report poor results at the beginning of their tenure (Axelson and Bond, 2009; Oyer, 2008). Consequently, their career perspectives are impacted. This is consistent with Magro et al. (2018), who argue that earnings management is greater in the early years of CEO tenure than in the later years, because CEOs in the early years of their tenure still have a whole carrier in front of them. Coherent with the career perspectives is job security. CEOs in the early years of their tenure could engage in earnings management due to the fact that they are more afraid of losing their jobs. Charitou et al. (2007) argue that managers are stimulated to manage their earnings to avoid management turnover. Therefore, CEOs in the early years of their tenure want to secure their job since they have more to lose in comparison to long-tenured CEOs. Concluding, career perspectives and job security are incentives of CEOs in the early years of their tenure to manipulate earnings to influence firm outcomes.

9

Ali and Zhang (2015) give two reasons why CEOs in the later years of their tenure have incentives to not engage in earnings management. First, they argue that CEOs in the later years of their tenure have incentives to prevent overstating the earnings because they want to protect their reputation (Ali and Zhang, 2015). If the market detects an earnings overstatement, this could harm the reputation of these CEOs (Ali and Zhang, 2015). Second, the costs of engaging in earnings management are higher than the potential benefits at the end of the CEOs tenure (Ali and Zhang, 2015). When a company has poor financial performance, the market is likely to blame other factors than the ability of the CEO, and, therefore, the benefits of manipulating the earnings are small (Ali and Zhang, 2015). These two reasons indicate that CEOs in the later years of their tenure are less likely to manipulate earnings.

However, some research indicates that CEOs with a longer tenure have more to lose relative to newly appointed CEOs. They argue that CEOs manipulate earnings at the end of their tenure in order to increase their final year’s compensation (Dechow and Sloan, 1991). Moreover, Butler and Newman (1989) find that CEOs in the final year of their tenure reduce discretionary expenses to increase their short-term earnings in order to receive their bonuses. Furthermore, some research argues that CEOs in the later years of their tenure manipulate earnings through their power, knowledge, and skills. CEOs in the later years of their tenure exercise more power than CEOs in the early years of their tenure (Cremers and Palia, 2011). Results show that CEOs in the later years of their tenure can use this power to increase their compensation because empowered CEOs can exploit their power to construct the board of directors in their favor (Boone, Field, Karpoff, Raheja, 2007). According to Francis, Huang, Rajgopal, and Zang (2008) CEOs in the later years of their tenure accumulate their power to impose their status to manipulate earnings. Moreover, these CEOs gather more knowledge of the company and skills during their tenure. CEOs in the later years of their tenure know the firm better and have more experience which could improve firm performance through effective management (Wang et al., 2016). Gul, Khedmati, Lim, and Navissi (2018) suggest that high-ability managers are in a better position to hide manipulation and conceal fraud. They argue that this is because matured-CEOs have more expertise and knowledge within the firm to override internal controls (Gul et al., 2018). This is consistent with Demerjian, Lev, and Lewis (2013) who argue that higher-ability managers may successfully manipulate earnings because they have more knowledge and experience to perform this complicated reporting strategy. Chou and Chan (2018) find that experienced CEOs are better

10

in decorating the financial statements. This indicates that CEOs in the later years of their tenure may be more skilled in managing earnings. Concluding, research argues that CEOs in the later years of their tenure have more opportunities, such as knowledge and experience, to engage in earnings management.

Based on the above literature, CEOs in the early years of their tenure have more incentives to engage in earnings management when the company is in financial distress. However, prior research does not give a straightforward answer about whether CEOs in the earlier years of their tenure manipulate earnings through AEM or REM when the company is in financial distress. As mentioned before, if a company is in financial distress, it has insufficient cash flow (Xu et al., 2018). As a consequence, CEOs in the early years of their tenure manipulate earnings through REM, because they want to accelerate cash flows and, therefore, they reduce discretionary expenses (Xu et al., 2018). Moreover, there is more auditor and regulatory scrutiny on AEM. As a result, AEM is easier to detect than REM (Graham et al., 2005). Therefore, CEOs in their earlier years prefer REM because there is a lower risk of detection and, therefore, it is less likely that their reputation and job security will be harmed.

However, in the long term, REM has adverse consequences on the long-term value of the firm (Kim and Song, 2013; Chamberlain et al., 2014). Kim and Song (2013) argue that in the term REM is value destroying. REM could hide the real performance of the company in the long-term and this allows managers with opportunities to misappropriate cash (Kim and Song, 2013). Moreover, investors are recognizing the negative consequences of REM on future cash flows and therefore, they estimate that future cash flows are lower for firms that use REM (Kim and Song, 2013). This has consequences for the long-term value of the firm, and, therefore, CEOs in the early years of their tenure engage in AEM because they want to secure their job on the long-term.

Hence, based on the above literature it is more likely that CEOs in the early years of their tenure will influence the relationship between financial distress and earnings management. This is because they have more incentives to engage in earnings management. However, there is not an unequivocal answer about whether CEOs in the earlier years of their tenure engage in AEM or REM. Therefore, I propose the following hypotheses;

11

Hypothesis 2a: The association between financial distress and total earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure.

Hypothesis 2b: The association between financial distress and accrual earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure.

Hypothesis 2c: The association between financial distress and real earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure.

METHODOLOGY Sample selection

The aim of this study is to answer the research question: “To what degree does financial distress influence the degree of earnings management and what is the influence of CEO tenure on this relationship?”. To answer this research question, I use archival research data. Archival research reflects the consequences of the behavior of CEOs when the company is in financial distress. Other alternative methods, such as questionnaires or interview studies, examine the opinions and beliefs of the CEO when the company is in financial distress, but do not describe the consequences of CEO behavior. Moreover, by using archival data, I can use real-world data, which improves the external validity of this research. Furthermore, I can use content analysis to draw valid conclusions about AEM, REM, and TEM.

The sample consists of financial data from US publicly listed firms from a period of 2003 to 2016. I chose this period for several reasons. In 2002 the SOX was implemented. In order to discourage earnings management, supervision by accountants and regulatory became stricter (Graham et al., 2005). As a result, many papers investigated the effect of the implementation of SOX on earnings management behavior. Moreover, there were a lot of economic and financial problems as a consequence of the financial crisis in 2008 (Manzaneque, Priego and Merino, 2016).

12

Consequently, many researchers investigated the effect of the financial crisis on earnings management. However, the time period of these papers were short-time periods. By using a longer time period (2003 to 2016), I can examine the effects of both major impacts on financial reporting.

I used several databases to gather information. I used the automated search features in Compustat to obtain financial data over a sample of 17.626 companies. Subsequently, I obtained information about the covenant violation of companies from the Amir Sufi database. These covenant violations were used to measure financial distress. The databases are matched by using the Gvkey. Afterward, information about CEO tenure was gathered from the ExecuComp database. This database provides information about the executives of the corporations directly obtained from each company’s annual proxy. CEO tenure was calculated by subtracting the date the CEO became CEO from the date the CEO left the company. However, not all information about CEO tenure was complete and consistent. Therefore, I collected extra information about CEO tenure from annual reports and the companies’ websites. This enhances the reliability of this study to a certain degree. Fogarty (2006) argued that the reliability of archival research will improve through hand collecting one variable. However, it should be mentioned that I could not obtain the CEO tenure of 17626 companies due to the time limit. Moreover, Compustat or ExecuComp did not provide complete information about the measurement needed and, therefore, the final sample is smaller. The sample selection resulted in a final sample of 1.017 companies.

Research design

Dependent variable Measuring accrual earnings management

The degree of AEM is measured through discretionary accruals. The Modified Jones model is used to determine the discretionary accruals (Jones, 1991; Dechow, Sloan and Sweeney, 1995; Klein, 2002). The Modified Jones model reduces the measurement error of discretionary accruals when discretion is used over revenues and makes a distinction between discretionary and non-discretionary accruals (Dechow et al., 1995). Prior literature also used the Modified Jones model as a proxy for AEM (Zang, 2012; Franz et al., 2014; Campa and Camacho, 2015). Discretionary accruals are calculated by deducting the non-discretionary accruals from the total accruals. The total accruals are determined as follows:

13

𝑇𝐴𝐶𝐶𝑡= 𝛥𝐶𝐴𝑡− 𝛥𝐶𝑎𝑠ℎ𝑡− 𝛥𝐶𝐿𝑡 + 𝛥𝐷𝐶𝐿𝑡− 𝐷𝐸𝑃𝑡 (1)

In this model, ΔCAt is the change in current assets in year t, ΔCash implies the change in

cash and cash equivalents in year t, ΔCLt is the change in current liabilities in year t, ΔDCLt stands

for the change in short term debt included in the current liabilities in year t and DEPt means the

depreciation and amortization expense in year t.

The Modified Jones model is calculated as follows:

𝑇𝐴𝐶𝐶𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1+ 𝛼2 (ΔREV𝑡− Δ𝑅𝐸𝐶_𝑡 ) 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸𝑡 𝐴𝑡−1+ 𝜀𝑡 (2)

In this formula, At-1 stands for the total assets in year t – 1, ΔREVt is the change in revenues

in year t, ΔRECt implies the change in net receivables in year t and PPEt means the gross property,

plant, and equipment in year t. The parameters α1, α2, and α3 are estimated through the ordinary

least squares regression. The residuals in year t are represented in the formula with εt. This formula

is used to calculate the parameters, whereby the discretionary accruals can be calculated through the following formula:

𝐷𝐴𝐶𝐶_𝑡 = 𝑇𝐴𝐶𝐶_𝑡− 𝑁𝐷𝐴𝐶𝐶_𝑡 (3)

In this model, DACCt implies the discretionary accruals in year t, and NDACCt is the total

non-discretionary accruals in year t. The total non-discretionary accruals can be established through the following model:

𝑁𝐷𝐴𝐶𝐶𝑡 𝐴𝑡−1 = 𝛼̂1 1 𝐴𝑡−1 + 𝛼̂2 (ΔREV𝑡− Δ𝑅𝐸𝐶_𝑡 ) 𝐴𝑡−1 + 𝛼̂3 𝑃𝑃𝐸𝑡 𝐴𝑡−1 (4)

In this model, the parameters α̂₁, α̂₂ and 𝛼̂₃ are estimated through the ordinary least squares regression.

14

Measuring real earnings management

The degree of REM is determined by the methodology of Roychowdhury (2006). This methodology is used because it has been developed for the analysis of distressed companies and therefore it is robust for this research. The construct validity of the proxies presented by Roychowdhury (2006) is supported by Franz et al. (2014) and Campa and Camacho (2015). The total REM is established by the sum of abnormal levels of cash flow from operations, production costs and discretionary expenses.

To calculate the normal level of cash flow, the following formula is used:

𝐶𝐹𝑂𝑡 𝐴𝑡−1 = α + β1 1 𝐴𝑡−1+ β2 𝑆𝑎𝑙𝑒𝑠𝑡 𝐴𝑡−1 + β3 ΔSales𝑡 𝐴𝑡−1 + 𝜀𝑡 (5)

In this model, CFOt stands for the cash flow from operations. This is obtained by the

earnings before interest and taxes plus the depreciation and amortization plus or minus the change in inventories, the change in trade and other receivables and the change in trade and other payables. At-1 stands for the total assets and the Salest implies the total net sales in year t. The change in net

sales in year t is defined by ΔSalest.

The abnormal cash flow (ABN_CFO) is determined as the variance between the actual CFO and the ‘normal’ CFO. The normal CFO is calculated by using the estimated coefficients from the above equation (5).

The manipulation of production costs is established by using the following model:

𝑃𝑅𝑂𝐷𝑡 𝐴𝑡−1 = α + β1 1 𝐴𝑡−1+ β2 𝑆𝑎𝑙𝑒𝑠𝑡 𝐴𝑠𝑡−1 + β3 ΔSales𝑡 𝐴𝑡−1 + β4 ΔSales𝑡 𝐴𝑡−1 + 𝜀𝑡 (6)

PRODt stands for the cost of goods sold plus the change in inventory in year t.

The abnormal production costs (ABN_PROD) is obtained as the difference between the actual costs and the ‘normal’ level of costs, which is calculated using the above equation (6).

15 𝐷𝐼𝑆𝑋𝑡 𝐴𝑡−1 = α + β1 1 𝐴𝑡−1+ β2 𝑆𝑎𝑙𝑒𝑠𝑡 𝐴𝑡−1 + 𝜀𝑡 (7)

In this model, DISXt is the discretionary expenditures, which is the sum of R&D,

advertising and selling, general and administrative expenditures in year t.

The abnormal discretionary expenses (ABN_DISX) are the difference between the actual discretionary expenses and the ‘normal’ discretionary expenses, which is gathered from the above equation (7).

Measuring total earnings management

TEM is measured by combining AEM with the total value of REM (Franz et al., 2014). Earnings management is winsorized at the top and bottom at a five percent level to mitigate any influences from outliers in order to enhance the results.

Independent variable

Financial distress will be defined as a situation where firms are not being able to discharge financial and other obligations (Wruck, 1990). The proxy for financial distress (FD) is the number of covenant violations. This proxy is used because this is an incentive to engage in earnings management based on the study of Franz et al. (2014). The higher the number of covenant violation, the greater the financial distress in the company. The sample of firms with debt covenants is separated into two groups; firms that are not violating covenants and firms that are violating covenants (0= firms that are not violating covenants, 1= firms that are violating

covenants).

Moderating variable

CEO tenure is defined as “the number of years that a CEO continuously holds its positions” (Baatwah et al., 2015). This research divides CEO tenure in the earlier years of their tenure and CEO tenure in the later years of their tenure. Most CEOs take major actions during the first three years in the office (Lewis, Walls and Dowell, 2014). Therefore, and following the research of Ali and Zhang (2015), CEO tenure in the early years of the tenure of the CEO is determined as less

16

than four years and CEO tenure in the later years of their tenure is four years or more (0=

short-term tenure, 1= long-short-term tenure).

Control variables

To mitigate the omitted-variable problem, I include other potential factors that may affect the hypotheses in the regression. These so-called control variables will reduce the effect of confounding variables. This means that when looking at the effect of one variable, all other variables are held constant. This is to increase the internal validity of the research. Based on prior research, the control variables used in this research are firm size, leverage, growth, CEO ownership, CEO age, CEO gender, CEO compensation and beating analysts forecast.

Prior research also used firm size as a control variable (Campa and Camacho, 2015). They argue that smaller firms tend to perform more in earnings management activities than medium and larger firms. In this research firm size (SIZE) is measured by the natural logarithm of the total assets expressed in dollars (Campa and Camacho, 2015). The log ensures a normal distribution.

The degree of earnings management is also likely to be influenced by the extent of leverage. Prior research has shown that companies with high debt leverage have strong incentives to engage in earnings management activities in order to avoid debt covenant violations (Franz et al., 2014). Leverage (LEV) is measured by dividing the total debt by total assets.

Moreover, companies with high growth have a strong incentive to manage earnings (Fengyi, Li-Jung, Chin-Chen and Teng-Shih, 2018). Therefore, the third control variable is growth. Growth (GROWTH) is controlled by dividing the book value of equity by the market value of equity. This proxy implies that firms with higher growth opportunities be inclined to have a lower market to book ratios (Franz et al., 2014). The firm-level characteristics are gathered from the database Compustat.

Furthermore, the following CEO characteristics are included as control variables, CEO ownership, CEO age, and CEO gender. Prior research suggests that CEO ownership have a strong motivation for CEOs to boost earnings because this has an effect on the stock prices and thus increase the value of their stocks (Ali and Zhang, 2015). CEO ownership (CEO OWN) is the percentage of outstanding stocks of the firm owned by the CEO at the beginning of the year (Ali and Zhang, 2015). According to Ali and Zhang (2015), CEO age is negatively associated with earnings management. CEO age (CEO AGE) is the age of the CEO at the beginning of the year.

17

Prior research also used CEO gender as a control variable (Chou and Chan, 2018). CEO Gender (CEO GEND) will be indicated with a dummy variable (0=male, 1=female).

In addition, managerial incentives as CEO compensation and beating earnings targets are also used as control variables. Several researchers examined that CEO compensation is an inducement to engage in earnings management activities because this will influence the stock prices and increase their wealth (Cella et al., 2014). CEO compensation (CEO COMP) is determined as the natural log of the total compensation salary the company is paying to its CEO. The log is used to enhance the distribution of data. The CEO characteristics are collected from the database ExecuComp.

Previous research indicates that beating analysts’ forecast is an important motivation to manage earnings (Franz et al., 2014; Roychowdhury, 2006; Graham et al., 2005). Beating analysts’ forecasts (BEAT) is determined whether firms are beating analysts’ forecasts consensus. This is determined with a dummy variable (0=not beating analysts’ forecasts consensus, 1=beating

analysts’ forecasts consensus). Information about the analysts’ forecasts is gathered from the

database IBES.

All continuous control variables are winsorized at the top and bottom at a one percent level to mitigate any influences from outliers.

Empirical models

To investigate whether financial distress has an impact on earnings management, I address three measures of earnings management, TEM, AEM, and REM. Franz et al. (2014) argue that companies might use AEM and REM as complements of each other. Therefore, this research will control REM (AEM) in the regression when the dependent variable is AEM (REM). Based on the research of Franz et al. (2014), I do not include AEM or REM as a control variable, when TEM is the dependent variable.

To gather evidence to test our hypothesis, I estimate the following models.

T𝐸Mit = α + 𝛽1 × FDit + 𝛽2 × FDit × CEO TENUREit + 𝛽3 × 𝑆𝐼𝑍𝐸it + 𝛽4 ×

LEVit +𝛽5 × GROWTH it + 𝛽6 × CEO OWNit + 𝛽7 × CEO AGEit + 𝛽8 × (8)

CEO GENDit + 𝛽9 × CEO COMPit + 𝛽10 × BEATit + 𝜖it

18

LEVit +𝛽5 × GROWTHit + 𝛽6 × CEO OWNit + 𝛽7 × CEO AGEit + 𝛽8 × (9)

CEO GENDit + 𝛽9 × CEO COMPit + 𝛽10 × BEATit + 𝛽11 × REMit + 𝜖it

R𝐸𝑀it = α + 𝛽1 × FDit + 𝛽2 × FDit × CEO TENUREit + 𝛽3 × 𝑆𝐼𝑍𝐸it + 𝛽4 ×

LEVit +𝛽5 × GROWTHit + 𝛽6 × CEO OWNit + 𝛽7 × CEO AGEit + (10)

𝛽8 × CEO GENDit + 𝛽9 × CEO COMPit + 𝛽10 × BEATit + 𝛽11 × AEMit + 𝜖it

RESULTS Descriptive statistics

Table 1 presents the descriptive statistics for the dependent variables, independent variable, moderating variable, and the control variables. The average AEM is -0.014. This implies that companies in the sample reduce their earnings through AEM. The median of REM is –0.119, which also indicates that companies reduce their earnings through REM. These findings are consistent with the findings of Joosten (2012). Joosten (2012) find also a slight decrease in AEM and a higher decrease in REM. Logically, the average of TEM is also negative. This means that the companies in the sample decrease their earnings through AEM, REM, and TEM.

Furthermore, the independent variable, financial distress, has an average of 2,6%. This indicates that 2.6% of the companies are in financial distress. CEO tenure as an average of 86,4%, which implies that 86,4% of the companies in the sample have a CEO that has been in the office for 4 years or more. The average firm size is 2,8 thousand US dollars of total assets. The largest firm has total assets of 66 thousand US dollars. The maximum leverage 3.081, which indicates that the maximum debt is three times the assets of that company. The growth has a median of 1.0, which indicates that the market value is on average not significantly lower or higher than the book value. CEO ownership has an average of 2.9%, which means that on average 2,9% of the outstanding stocks of the firm are owned by the CEO at the beginning of the year. The age of the CEO varies from 27 to 96, with an average of 57. The average of the variable CEO gender is 0.0334027. This means that 3,34% of the firms have a female CEO. The median of CEO compensation is 680, which indicates that the average compensation of the CEO is 680 US dollars. The average of beating the financial analysts is 0.568, which implies that 56,8% of the companies is beating the financial analysts’ forecasts.

19

Correlation analysis

The correlation analysis determines the reliability scores and the correlations between the variables. The results are presented in tables 2, 3, and 4. The correlation analysis show that TEM is positively correlated with CEO tenure (r = 0.0213, p < 0.01). Moreover, the correlation analysis illustrate that AEM is negatively correlated with financial distress (r = -0.0.0397, p < 0.001) and positively related to CEO tenure (r = 0.0184, p < 0.05).

TABLE 1 Descriptive Statistics

Variable n Mean Median Standard deviation Minimum Maximum

AEM 16825 -0.014 -0.016 0.119 -0.451 0.420 REM 16825 -0.119 -0.058 0.283 -0.920 0.856 TEM 16825 -0.134 -0.079 0.315 -1.034 1.062 FD 16825 0.026 0 0.160 0 1 CEO TENURE 16825 0.864 0.86 0.343 0 1 SIZE 15498 2821.662 1588.60 8925.894 0.014 66955 SIZE (LOG) 15498 3.180 3.152 0.708 -1.678 4.872 LEV 15498 0.201 0.170 0.221 0 3.081 GROWTH 16558 1.001 1.000 0.139 0.556 2.217 CEO OWN 13260 2.919 0.524 6.991 0 100 CEO AGE 15962 57.175 57 8.305 27 96 CEO GEND 16825 0.033 0 0.180 0 1 CEO COMP 16650 638.566 680 369.217 0 1950 CEO COMP (LOG) 16650 2.770 2.796 0.265 1.775 3.681 BEAT 10562 0.568 1 0.495 0 1

20

TABLE 2

Correlation analysis of total earnings management

TEM FD CEO

TENURE

SIZE LEV GROWTH CEO OWN CEO AGE CEO

GEND CEO COMP BEAT TEM 1 FD -0.003 1 CEO TENURE 0.021** -0.002 1 SIZE -0.022** -0.067*** 0.013 1 LEV -0.003 0.031*** -0.042*** 0.209*** 1 GROWTH 0.009 -0.002 -0.013 -0.019* 0.034*** 1 CEO OWN 0.002 0.031*** 0.064*** -0.151*** -0.091*** 0.017 1 CEO AGE -0.012 -0.015 0.086*** 0.111*** 0.003 -0.008 0.193*** 1 CEO GEND -0.017* -0.006 -0.030*** -0.006 -0.016 0.008 -0.037*** -0.066*** 1 CEO COMP -0.057*** -0.038*** 0.163*** 0.544*** 0.146*** -0.022*** -0.108*** 0.145*** -0.021** 1 BEAT -0.014 -0.050*** 0.002 0.068*** -0.062*** -0.034*** -0.019* -0.030** 0.017 0.049*** 1 * p < 0.05, ** p < 0.01, *** p < 0.001

21

TABLE 3

Correlation analysis of accrual earnings management

AEM REM FD CEO

TENURE

SIZE LEV GROWTH CEO OWN CEO AGE CEO

GEND CEO COMP BEAT AEM 1 REM -0.003 1 FD -0.040*** 0.014 1 CEOTENURE 0.018* 0.013 -0.002 1 SIZE 0.037*** -0.039*** -0.067*** 0.013 1 LEV 0.016* -0.007 0.031*** -0.042*** 0.209*** 1 GROWTH 0.008 0.008 -0.002 -0.013 -0.019* 0.034*** 1 CEO OWN -0.008 0.004 0.031*** 0.064*** -0.151*** -0.090*** 0.017 1 CEO AGE 0.025** -0.023** -0.015 0.086*** 0.111*** 0.003 -0.008 0.193*** 1 CEO GEND 0.006 -0.025** -0.006 -0.030*** -0.006 -0.016 0.008 -0.037*** -0.066*** 1 CEO COMP 0.015 -0.068*** -0.038*** 0.163*** 0.544*** 0.146*** -0.022** -0.108*** 0.145*** -0.021** 1 BEAT 0.027** -0.027** -0.050*** 0.002 0.068*** -0.062*** -0.034*** -0.019 -0.030** 0.017 0.049*** 1 * p < 0.05, ** p < 0.01, *** p < 0.001

22

TABLE 4

Correlation analysis of real earnings management

REM AEM FD CEO

TENURE

SIZE LEV GROWTH CEO OWN CEO AGE CEO

GEND CEO COMP BEAT REM 1 AEM -0.003 1 FD 0.014 -0.040*** 1 CEOTENURE 0.013 0.018* -0.002 1 SIZE -0.039*** 0.037*** -0.067*** 0.013 1 LEV -0.007 0.016* 0.031*** -0.042*** 0.209*** 1 GROWTH 0.008 0.008 -0.002 -0.013 -0.019* 0.034*** 1 CEO OWN 0.003 -0.008 0.031*** 0.064*** -0.151*** -0.090*** 0.017 1 CEO AGE -0.023** 0.025** -0.015 0.086*** 0.111*** 0.003 -0.008 0.193*** 1 CEO GEND -0.025** 0.006 -0.006 -0.030*** -0.006 -0.016 0.008 -0.037*** -0.066*** 1 CEO COMP -0.068*** 0.015 -0.038*** 0.163*** 0.544*** 0.146*** -0.022** -0.108*** 0.145*** -0.021** 1 BEAT -0.027** 0.027** -0.050*** 0.002 0.068*** -0.062*** -0.034*** -0.019 -0.030** 0.017 0.049*** 1 * p < 0.05, ** p < 0.01, *** p < 0.001

23

Regression analysis

The regression analysis is used to investigate the relationship between variables. The results of the regression analysis are presented in tables 5, 6, and 7. The results are presented in three different models. Model 1 includes the impact of the control variables on earnings management. Model 2 contains the impact of financial distress on earnings management and model 3 includes the impact of CEO tenure on the relationship between financial distress and earnings management.

Total earnings management

The relationship between financial distress and TEM is described in table 5. Hypothesis 1a posits that firms in financial distress are more likely to engage in total earning management than firms that are not in financial distress. To test this hypothesis, the effect of financial distress and the main effects are included in model 2 (table 5). The results show that there is no significant association between financial distress and TEM (β = 0.017, p > 0.05). Therefore, there is no significant effect of financial distress on TEM. I thus do not find support for hypothesis 1a. However, there is a significant positive relation between CEO tenure and TEM (β = 0.048, p < 0.001). This implies that CEOs in the later years of their tenure will engage in TEM.

Hypothesis 2a posits that the association between financial distress and total earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure. To test this hypothesis, an interaction term of financial distress and CEO tenure (FD x CEO TENURE) was constructed and included in model 3 (table 5). The results show that there is no significant moderating effect of CEO tenure on the association between financial distress and TEM. Therefore, I can conclude that CEO tenure has no significant impact on the relationship between financial distress and TEM. Hence, hypothesis 2a is rejected.

Based on model 1 in table 5, I can conclude that there is a significant negative relationship between CEO gender and TEM (β = -0.055, p < 0.01). This indicates that male CEOs engage more in TEM. Moreover, results show a significant negative relationship between CEO compensation and TEM (β = -0.091, p < 0.001). This indicates that when a CEO receives more compensation, there is less TEM.

24

Accrual earnings management

The relationship between financial distress and AEM is illustrated in table 6 (model 2). H1b posits that firms in financial distress are more likely to engage in accrual earnings management than firms that are not in financial distress. To test this hypothesis, the direct effect of financial distress and other main effects on AEM are examined. The results confirm that there is a significant negative relationship between financial distress and AEM (β = -0.028, p < 0.01). This implies that financially distressed companies engage in less AEM. However, I did not find a significant relationship between CEO tenure and AEM (β = -0.001, p > 0.05).

Hypothesis 2b posits that the association between financial distress and accrual earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure. To test this hypothesis, financial distress and CEO tenure (FD X CEO TENURE) are constructed and incorporated in model 3 (table 6). The results show that there is no significant moderating effect of CEO tenure on the relationship between financial distress and AEM (β = -0.016, p > 0.05). Therefore, hypothesis 2b is rejected.

Moreover, I can deduce that there is a significant positive relationship between size and AEM (β = 0.010, p < 0.001). This implies that bigger firms tend to perform AEM. Furthermore, there is a significant positive relationship between CEO age and AEM (β = 0.001, p < 0.001). This indicates that older CEOs engage more in AEM in comparison with younger CEOs.

Real earnings management

The relationship between financial distress and REM is illustrated in table 7 (model 2). H1c posits that firms in financial distress are more likely to engage in real earnings management than firms that are not in financial distress. To test this hypothesis, the effect of financial distress and other main effects on REM are analyzed. The findings present a marginally positive relationship between financial distress and REM (β = 0.043, p = 0.066). Therefore, I can conclude that there is a marginally significant association between financial distress and REM. This implies that financially distressed companies engage in more REM. I thus find support for hypothesis H1c. This is consistent with the findings of Campa and Camacho (2015). Moreover, I did find a significant positive relationship between CEO tenure and REM (β = 0.048, p < 0.001). This suggests that CEOs in the later years of their tenure engage in REM. Interestingly, the coefficient and the significance level of REM are the same for TEM. TEM is measured by combining REM

25

and AEM, and because the coefficient and significance level are the same, I can infer that TEM is entirely explained by REM. This means that only CEOs in the later years of their tenure manipulate earnings through REM.

Hypothesis 2c states the association between financial distress and real earnings management is stronger when the CEO of the firm is in the early years of its tenure than when the CEO of the firm is in the later years of its tenure. To test this hypothesis, the interaction term of financial distress and CEO tenure (FD x CEO TENURE) is included in model 3 (table 7). The results indicate that there is no significant moderating effect of CEO tenure on the relationship between financial distress and REM (β = -0.123, p > 0.05). Therefore, I can conclude that there is no moderating impact of CEO tenure on the association between financial distress and REM. Hence, I do not find support for hypothesis 2c.

Based on the findings in table 7, I do find a significant negative relationship between CEO gender and REM (β = -0.053, p < 0.01). This implies that whether the CEO is a female there is less REM. Moreover, there is a significant negative relationship presented between CEO compensation and REM (β = -0.089, p < 0.001). This implies that whether the CEO receives more compensation, there is less REM.

26

TABLE 5

Regression analysis of total earnings management

Model 1 t-Statistic Model 2 t-Statistic Model 3 t-Statistic

INTERCEPT 0.067 1.23 0.054 0.99 0.051 0.94 Controls SIZE 0.013 1.91 0.016** 2.39 0.016** 2.41 LEV 0.013 0.63 0.016 0.77 0.016 0.79 GROWTH -0.008 -0.33 -0.008 -0.30 -0.008 -0.30 CEO OWN 0.001 1.14 0.001 0.83 0.000 0.67 CEO AGE 0.000 0.03 -0.000 -0.20 -0.000 -0.14 CEO GEND -0.055** -2.73 -0.052** -2.57 -0.051** -2.56 CEO COMP -0.091*** -5.25 -0.104*** -5.90 -0.104*** -5.92 BEAT -0.004 -0.51 -0.004 -0.50 -0.004 -0.49 Main effects FD 0.017 0.65 0.143 1.61 CEO TENURE 0.048*** 4.20 0.050*** 4.35 Two-way interactions FD X CEO TENURE -0.138 -1.48 R-SQUARED 0.0051 0.0075 0.0078 ADJUSTED R-SQUARED 0.0040 0.0061 0.0063 F-VALUE 4.75*** 5.62*** 5.31*** * p < 0.05, ** p < 0.01, *** p < 0.001 n= 7466

27

TABLE 6

Regression analysis of accrual earnings management

Model 1 t-Statistic Model 2 t-Statistic Model 3 t-Statistic

INTERCEPT -0.086*** -4.38 -0.083*** -4.24 -0.084*** -4.25 Controls SIZE 0.010*** 4.23 0.009*** 3.93 0.009*** 3.94 LEV 0.004 0.51 0.005 0.70 0.005 0.71 GROWTH 0.012 1.25 0.012 1.21 0.012 1.21 CEO OWN -0.000 -0.70 -0.000 -0.54 -0.000 -0.59 CEO AGE 0.001*** 3.36 0.001*** 3.33 0.001*** 3.34 CEO GEND -0.003 -0.44 -0.003 -0.47 -0.003 -0.46 CEO COMP -0.004 -0.64 -0.003 -0.54 -0.003 -0.55 BEAT 0.006* 2.36 0.006* 2.28 0.006* 2.28 REM -0.003 -0.71 -0.003 -0.62 -0.003 -0.63 Main effects FD -0.028** -3.05 -0.014 -0.42 CEO TENURE -0.001 -0.30 -0.001 -0.23 Two-way interactions FD X CEO TENURE -0.016 -0.47 R-SQUARED 0.0065 0.0077 0.0078 ADJUSTED R-SQUARED 0.0053 0.0063 0.0062 F-VALUE 5.40*** 5.28*** 4.86*** * p < 0.05, ** p < 0.01, *** p < 0.001 n= 7466

28 * p < 0.05, ** p < 0.01, *** p < 0.001

n= 7466

TABLE 7

Regression analysis of real earnings management

Model 1 t-Statistic Model 2

t-Statistic Model 3 t-Statistic INTERCEPT 0.152** 3.05 -0.137** 2.75 -0.135** 2.69 Controls SIZE 0.004 0.71 0.008 1.33 0.008 1.35 LEV 0.007 0.41 0.009 0.49 0.009 0.51 GROWTH -0.017 -0.68 -0.016 -0.64 -0.016 -0.64 CEO OWN 0.001 1.47 0.001 1.08 0.001 0.93 CEO AGE -0.001 -1.36 -0.001 -1.60 -0.001 -1.54 CEO GEND -0.053** -2.86 -0.049** -2.69 -0.049** -2.68 CEO COMP -0.089*** -5.64 -0.102*** -6.37 -0.102*** -6.39 BEAT -0.011 -1.58 -0.011 -1.55 -0.010 -1.54 AEM -0.021 -0.71 -0.018 -0.62 -0.019 -0.63 Main effects FD 0.043 1.84 0.156 1.91 CEO TENURE 0.048*** 4.63 0.050*** 4.77 Two-way interactions FD X CEO TENURE -0.123 -1.45 R-SQUARED 0.0076 0.0109 0.0112 ADJUSTED R-SQUARED 0.0064 0.0095 0.0096 F-VALUE 6.36*** 7.50*** 7.05***

29

ROBUSTNESS TESTS

Measurement of AEM

The relationship between financial distress and AEM indicates that financially distressed firms engage in less AEM. To examine whether different proxies for AEM change this significant negative result, I conduct three other proxies for AEM, the Jones model, the Jones model with return on assets, and the Modified Jones model with return on assets.

First, I investigate the relationship between financial distress and AEM using the Jones Model. The Jones model was the first measure of discretionary accruals discovered by Jones (1991). The difference between the Modified Jones model and the Jones model is that in the Jones model the change of receivables are excluded because it only assumes that revenues are non-discretionary. Therefore, the Jones model is organized as follows:

𝑇𝐴𝐶𝐶𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1+ 𝛼2 ΔREV𝑡 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸𝑡 𝐴𝑡−1+ 𝜀𝑡 (11)

I repeated the regression analysis using the Jones model as another estimation of abnormal accruals. As a consequence of the change of AEM, there is also a change in TEM.

The results of this robustness test are presented in the tables B1, B2, and B3 in Appendix B. These results show a significant negative relationship between financial distress and AEM (β = -0.096, p = 0.001) and a marginally significant positive association between financial distress and REM (β = 0.043, p = 0.054). Moreover, the results show that there is no significant association between financial distress and TEM (β = -0.050, p > 0.10). Therefore, I can conclude that these findings are consistent with the findings of this current study.

Second, I examine the relationship between financial distress and AEM using the Jones model with return on assets (ROA). Kothari, Leone, and Wasley (2005) claim that there is a measurement error in discretionary accruals which are calculated through the Jones model because the Jones model does not take the prior performance of the company into account. Therefore, Kothari et al. (2005) include ROA of the previous year (ROAt-1) as a performance measure. This

30 𝑇𝐴𝐶𝐶𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1+ 𝛼2 ΔREV𝑡 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸𝑡 𝐴𝑡−1 + 𝑅𝑂𝐴𝑡−1 + 𝜀𝑡 (12)

Based on the findings of Kothari et al. (2005), I examine whether the use of the Jones model with ROA, have an impact on the significant levels observed in this paper. I repeated the regression analysis using the Jones model with ROA as another proxy for abnormal accruals.

The findings of the robustness are presented in the tables C1, C2, and C3 in Appendix C. Interestingly, these findings show a relationship between AEM and REM. Based on the research of Zang (2006), this implies that financially distressed firms use AEM and REM as substitutes. Moreover, the results indicate that financially distressed companies use that there is no significant relationship between financial distress and TEM (β = -0.060, p > 0.10). However, there is a significant negative relationship between financial distress and AEM (β = -0.107, p = 0.000) and positive association between financial distress and REM (β = 0.048, p = 0.039). Results show that the significance levels increase using the Jones model with ROA. The reason for this is that the performance measure of ROA is included in the model.

Lastly, I investigate the relationship between financial distress and the Modified Jones model with ROA as another proxy for AEM. Kothari et al. (2005) claim that the discretionary accruals calculated through the Modified Jones model contain a measurement error because the model does not include the prior performance of the company. Therefore, they include ROA (ROAt-1) as a measurement of the prior performance of the company (Kothari et al., 2005). The

following equation presents the Modified Jones model with the ROA;

𝑇𝐴𝐶𝐶𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1+ 𝛼2 (ΔREV𝑡−ΔREC𝑡 ) 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸𝑡 𝐴𝑡−1+ 𝑅𝑂𝐴𝑡−1 + 𝜀𝑡 (13)

Based on the findings of Kothari et al. (2005), I examine whether the implementation of the ROA alters the significance levels. I re-run the regression analysis using the Modified Jones model with ROA as a proxy for AEM. The measurement of TEM is also adjusted, because of the change of measurement of AEM. The findings of the robustness tests are illustrated in the tables D1, D2, and D3 in Appendix D. The results illustrate that there is no significant impact between financial distress and TEM (β = 0.020, p > 0.10). However, the results indicate that there is a

31

significant negative relationship between financial distress and AEM (β = -0.022, p = 0.012) and a marginally significant positive relationship between financial distress and REM (β = -0.044, p = 0.062). The results of this robustness test show that the significance levels are the same as the findings observed in this current study. Therefore, the findings are consistent.

Measurement of REM

The marginally significant negative relationship between financial distress and REM indicates that financially distressed firms engage in REM. Therefore, I analyze the impact of financing distress by examining another proxy for REM. Based on the findings of Zang (2012), I repeat the regression analysis using abnormal levels of production costs and discretionary expenses as other proxies for REM. Zang (2012) claims that abnormal cash flow from operations should not be examined when determining REM, therefore she determines REM through the abnormal levels of production costs and the discretionary expenses. Zang (2012) argues that REM activities have a different impact on the cash flow from operations and this leads to a doubtful net effect. For example, price discount and overproduction decrease the cash flow from operations, when reducing discretionary expenses increases the cash flow from operations (Zang, 2012; Roychowdhury, 2006). Therefore, I examine whether abnormal levels of production costs and discretionary expenses as other estimations of REM might change the significant level.

The results of this robustness test are demonstrated in the tables E1, E2, and E3 in Appendix E. These findings show a significant negative relationship between financial distress and AEM (β = -0.020, p = 0.002) and a marginally significant positive relationship between financial distress and REM (β = 0.056, p = 0.063). Results also show no significant relationship between financial distress and TEM (β = 0.017, p > 0.10). Based on the findings of this robustness tests, I can conclude that this is consistent with the observed findings presented in this paper.

Measurement of financial distress

Prior research has investigated the relationship between financial distress and earnings management. A lot of research uses ROA, credit rating, and the Altman Z-Score as other measurements for financial distress. To analyze whether other proxies of financial distress might alter the observed relationship presented in this study, I examine the effect of ROA and credit rating on earnings management. I choose these proxies because they are related to the financial

32

performance of the firm and are used in prior research as a measure of a firm’s likelihood of distress (Franz et al., 2014). Prior research also used Altman Z-Score as a proxy of financial distress, however, there is some criticism about this measurement. García et al. (2009) argue that the Altman Z-Score used the technique LDA. This technique demands that the financial ratios are normally distributed and that the ratios of financially distressed companies should have the same variance-covariance structures as those of healthy firms. They argue that these requirements have an impact on the structure of data and are therefore this measure is not appropriate as a proxy for financial distress (Garciá et al., 2009).

The first proxy for financial distress is the ROA, which implies the abnormal operating performance of the company (Easterwood, Ince, and Raheja, 2012). Franz et al. (2014) find that the return on assets is lower for defaulting firms than firms far form covenant violation. This is not surprising, because firms in financial distress are more likely to violate debt covenants (Franz et al., 2014; Dickev and Skinner, 2002). Therefore, I use the ROA as a proxy for financial distress. The ROA is estimated by the net income in a given period divided by its total assets. I use equations 8, 9, and 10 and replace financial distress with the ROA. The findings of the results are illustrated in in the tables F1, F2, and F3 in Appendix F. The results confirm a significant negative relationship between ROA and TEM (β = -0.175, p = 0.000), ROA and AEM (β = -0.041, p = 0.000), and ROA and REM (β = -0.208, p = 0.000). The results of this robustness indicate that firms with a lower level of ROA manage earnings through AEM, REM, and TEM.

Second, I use credit rating as a proxy for financial distress because it is a measurement of the financial condition of the firm (Franz et al., 2014). Credit rating is an evaluation of the ability of the organization to comply with their financial obligations (Gul and Goodwin, 2018). Franz et al. (2014) argue that firms with a lower level of credit rating imply a higher level of financial distress. Therefore, I expect that firms with a lower level of credit rating engage in earnings management activities. Based on the research of Franz et al. (2014), I operationalize the variable credit rating in such way that the lowest rating (D rating) is allocated with value one and the highest rating (AAA rating) is allocated with value twelve. Next, I use equations 8,9, and 10 and replace financial distress with the variable credit rating. The results imply that there is a significant negative relationship between credit rating and TEM (β = -0.023, p = 0.000), between credit rating and AEM (β = -0.003, p = 0.001), and between credit rating and REM (β = -0.020, p = 0.000). The