Competition as a Double-Edged Sword: How does it affect Earnings Management and can Political Risk explain the direction of the relationship?

(1)

Competition as a Double-Edged Sword: How does it affect Earnings

Management and can Political Risk explain the direction of the relationship?

MSc Thesis Accountancy

Faculty of Business and Economics, Department of Accounting

ABSTRACT

In response to the mixed results in previous research, I examine the effect of competition on the degree of earnings management (EM), including both accrual-based (AEM) and real earnings management (REM). Based on agency theory, competition can either decrease the level of EM by reducing information asymmetry or lead to more EM by stimulating opportunistic behavior. Therefore, I hypothesize that competition affects EM, but do not include a direction. Moreover, I study the moderating effect of political uncertainty by employing a firm-level measure of political risk, where I expect that in context of high (low) political uncertainty, competition leads to an increase (decrease) in EM. I argue this because in times of increased political uncertainty, there is more opportunity and motivation to engage in EM. Using a sample of 9,332 U.S. observations from 2002 to 2016, I find no significant results for both hypotheses. Additional analyses give helpful insights regarding the insignificant results. From these, possible explanations for the outcome are derived, i.e. the noisiness of the proxy for AEM, the fact that managers do not employ all methods of REM and the presence of financial crisis years in my sample. Concerning the moderating effect of political uncertainty, politics might not affect executives as much as expected. Thus, although there are no significant results in this thesis, it still provides useful information that can be used by standard setters and practitioners.

Keywords: market competition, earnings management, accruals earnings management, real earnings manipulation, political risk

Name: F.A.M. (Floor) van Kessel

Student number: 2676931

Supervisor: Prof. dr. R.B.H. Hooghiemstra Co-assessor: Dr. S. Mukherjee

Date: 20-01-2020

(2)

Introduction

Earnings management (EM) has been an intriguing topic for accounting researchers for decades. Generally speaking, earnings management refers to “a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain” (Schipper, 1989: 92). After Enron hiding debts off its balance sheets, Worldcom underreporting on expenses and including fake revenues, and more recently General Electric being under serious investigation for accounting fraud, it is difficult to understand how no one realized what was going on in these firms at the time. To prevent more accounting scandals from happening and its disastrous effects, it is fruitful to be more able to pinpoint circumstances that affect the extent to which firm management engages in earnings management. Previous literature has studied the role of various corporate governance (CG) characteristics in explaining differences in the extent to which managers (can) engage in EM, such as independence of the board and/or audit committee, bonus schemes, CEO tenure and many other factors (Klein, 2002; Bergstresser & Philippon, 2006; Cheng & Warfield, 2005; Ali & Zhang, 2015). Most prior research focuses on internal CG mechanisms, whereas external CG mechanisms are largely ignored (Aguilera, Desender, Bednar, & Lee, 2015). However, recently an increasing number of studies focus on the role of external governance on opportunistic behavior by management. This is due to the realization that internal characteristics do not operate in isolation - they are impacted by external factors that can have a large effect on an organization’s operations (Aguilera et al., 2015). Examples of these external governance mechanisms include legal systems, external audits, and market competition.

This paper investigates how market competition affects earnings management. Previous studies on this topic have found mixed results, which justifies additional research on this relationship in the hope to gain more traction on this topic. For example, Balakrishnan and Cohen (2014) find that competition leads to less EM, where they argue that competition is a disciplining force that retains managers from misreporting. On the other hand, Shleifer (2004) emphasizes the “dark side” of competition and suggests that it can lead to unethical behavior such as earnings management. It seems that competition has two faces – one that solves agency problems and one that creates them. Agency problems are issues created by the separation of ownership and management and their divergent goals (Jensen & Meckling, 1976). Owners want to maximize firm value while management might pursue personal goals which are not necessarily in the best interest of the firm (Eisenhardt, 1989). The existence of information asymmetry characterizing the relationship between owners and managers, make it hard for owners to fully detect managers’ self-serving behaviors. While financial reporting is an important control mechanism to reduce information asymmetry, there are ways – including EM- to distort that information too.

One stream of literature argues that increased competition solves agency problems. Several studies find a negative relationship between competition and EM (e.g. Balakrishnan & Cohen, 2014; Laksmana & Yang, 2014; Marciukaityte & Park, 2009). This is in line with Hart’s (1983) paper, who states that competition reduces managerial slack. He argues that if the costs of competitors go down, every firm’s costs also decrease which makes it harder for managers to slack. In addition, market competition provides more information for conducting peer performance, since there are more firms to compare your results to (Holmstrom, 1982). This leads to less information asymmetry, limiting the opportunity to engage in EM and thereby reducing agency costs.

(4)

Furthermore, false reporting is punished more severely in the stock market in heightened competition. Research shows that if the market discovers earnings management it leads to negative stock market reactions, where this relationship is stronger for firms operating in competitive markets (Marciukaityte & Park, 2009). Given management does not want the stock price to drop, the stock market serves as an incentive to prevent opportunistic behavior in competition. Also, overly risk-taking behavior, as seen in real earnings manipulation, can lead to bad investments that can have a detrimental effect on the firm’s competitiveness (Balakrishnan & Cohen, 2014). Due to predation risk, the risk of losing market share and investment possibilities to the competition, this decrease in competitiveness can lead to the firm being pushed out of the market (Froot, Sharfstein, & Stein, 1993). All in all, taking risks can have larger, worse effects on firms in highly competitive industries, which creates an incentive to not engage in risky behavior such as earnings management practices.

On the other hand, it is argued that heightened competition encourages management to engage in overly risk-taking behavior and other forms of behavior classified as moral hazard, which may include too high levels of earnings management. Milgrom and Roberts (1992) study the impact of competitive pressure on moral hazard problems in the U.S. savings and loan industry. They state that in order to survive in the highly competitive industry, firms resort to risk taking and fraudulent behavior. The reason is that more risk-averse companies are not able to compete with their risk-taking competitors without gambling on the same risky investments (Milgrom & Roberts, 1992). In addition, Shleifer (2004) discusses how competition can lead to unethical behavior. He argues that firms value ethical behavior, but when push comes to shove, the willingness to behave ethically is overpowered by firm survival and profitability. Basically, the idea is that competition potentially threatens firm survival, and subsequently triggers unethical ways of conducting business. Several studies find evidence that is consistent with this view (e.g Markarian & Santalo, 2014; Lin, Officer, & Zhan, 2015; Karuna, Subramanyam, & Tian, 2015; Datta, Iskandar-Datta, & Singh, 2013).

Apart from looking at the direct association between competition and earnings management, this paper seeks to examine if political uncertainty can explain why in one condition, the association between competition and earnings management is positive and in the other negative. Political uncertainty is defined as “uncertainty about the government’s future actions” (Pastor & Veronesi, 2013: 2). A recent stream in the accounting and finance literature focuses on the role that political uncertainty plays in organizations, for example, its influence on risk premia, investment expenditures and the volatility of stocks (Pastor & Veronesi, 2013; Julio & Yook, 2012; Pastor & Veronesi, 2012). Current developments such as the trade war between the U.S. and China prove that business and politics are closely intertwined. In regards to the influence of political uncertainty on EM, I first draw upon agency theory. Due to increased information asymmetry between the principal and agent in times of uncertainty about policies (Nagar, Schoenfeld, & Wellman, 2019), managers are provided with more opportunity to manage earnings, because it makes it more difficult for investors to assess the impact of uncertainty whereas management of the firm can assess this much better. Besides, according to the political cost hypothesis, management has more motivation to engage in EM during politically uncertain times. This means that firms want to avoid government interference and are willing to manage earnings to influence the outcome in uncertain times (Watts & Zimmerman, 1978). I hypothesize that political uncertainty moderates the relationship between competition and EM in such a way that in the context of high political uncertainty, competition leads to EM and vice versa. Given political uncertainty creates more

(5)

opportunity and motivation to manipulate earnings, it creates more incentives for EM whereas low political uncertainty reduces incentives.

I employ a sample of 9,332 observations of 1,571 unique firms from 2002 to 2016. Competition is measured with the Herfindahl-Hirschman Index (HHI), a statistical measure of concentration used as a proxy for competition (Rhoades, 1993). It is calculated as the sum of squared market shares of the firms in the industry, where the lower the index, the higher the level of competition. I use the sales of firms to calculate the market shares per industry. The measurement of earnings management is difficult because the managerial intent behind it is not quantifiable (Dechow & Skinner, 2000). Firms mainly employ two techniques to manipulate earnings, which I both include in this research. The first method is based on accruals, where management changes accounting methods or estimates to present its financial status more favorably, such as changing a depreciation or valuation method (Zang, 2012). I calculate these according to the Modified Jones Model developed by Dechow, Sloan and Sweeney (1995). Another method that mainly affects cash flow is called real earnings management. It can be described as actions managers take that deviate from normal business practices to meet certain earnings goals (Roychowdhury, 2006). Roychowdhury (2006) describes several ways in which firms do this: through sales acceleration, lowering cost of goods sold and altering discretionary expenses. I will use this model to calculate the amount of real earnings manipulation. Lastly, the data on political uncertainty is a firm-level measure of political risk developed by Hassan, Hollander, van Lent and Tahoun (2019). The results of this research display no significant relationship between market competition and the degree of earnings management (both accrual-based earnings management and real earnings manipulation). In addition, political uncertainty is not able to explain the direction of the competition – EM relationship, as I do not document a significant moderating effect. To gain more insights on these results, I conduct several additional tests. I find that the results remain robust to the use of a different measure of competition, separate REM proxies, different industry classification and the signed values of EM. Moreover, in other cases it results in significant results: first, the use of a different formula when calculating AEM results in a significant relation with competition. Second, I replace the Hassan et al. (2019) political uncertainty measure with the EPU index, a macroeconomic measure of political uncertainty, which results in a moderating effect regarding REM. Lastly, I take the financial crisis into account by dividing my sample into three subsamples (before, during and after the crisis) and find that in the years of the financial crisis competition leads to a significant decrease in the use of accrual-based earnings management.

This study makes several contributions to the literature. First, it adds to the literature regarding the competition – earnings management relationship, which is inconclusive to date due to mixed results (e.g. Markarian & Santalo, 2014; Laksmana & Yang, 2014). While this research does not provide significant results, it still provides relevant insights e.g. the theoretical and methodical framework about this subject. Second, I contribute to the literature on real earnings manipulation and what drives it. Previous similar research often only includes accrual-based earnings management (e.g. Markarian & Santalo, 2014; Marciukaityte & Park, 2009), whereas Cohen, Dey, and Lis (2008) show there has been a shift from AEM towards REM since 2002. Consequently, I include a proxy for REM to test for this besides AEM. Third, this thesis adds to the literature regarding political uncertainty in the finance and accounting field. Several studies find a relationship between political uncertainty and firm behavior

(6)

(e.g. Pastor & Veronesi, 2013; Julio & Yook, 2012), but it has not been researched in the context of EM and competition, hence I include it as a moderator. Moreover, I include a more recent, firm-level measure of political risk whereas most other studies use a macroeconomic measure. All in all, this thesis provides useful insights for practitioners and standard setters.

The remainder of this paper is organized as follows. Section 2 provides a theoretical background and develops hypotheses. Next, section 3 discusses the data and methodology on how to execute the research, of which the results are presented in section 4. Lastly, section 5 provides a discussion and conclusion with the limitations of this research and suggestions for further research.

(7)

2. Theoretical framework

2.1. Agency theory

The foundation of the theoretical framework of this thesis is agency theory (Jensen and Meckling, 1976). The key issue agency theory addresses are potential “solutions” for the problems that arise due to the separation of ownership and management of a firm. As a result of this separation and as each individual is presumed to maximize its own utility, a conflict of interest can arise – management might take actions that do not maximize firm value (Eisenhardt, 1989). The manager is able to do so because there is information asymmetry, where one party has an information advantage over the other. In this case, the agent (management) has more knowledge of the firm than the principal (shareholders) because the agent operates the firm on a daily basis and is thus more familiar with the current earnings, risks and opportunities (Ndofor, Wesley & Priem, 2015).

Due to information asymmetry, agents may take part in opportunistic behavior which is also referred to as moral hazard. Moral hazard occurs when the agent can engage in actions that may potentially harm firm value as the principal will suffer the costs if it does not have the desired effect (Holmstrom, 1979). For example, a CEO has a private jet that is very expensive and reduces profits but has a positive impact on the CEO’s status. Here the corporate jet is the agent’s goal which is not aligned with the principal’s goal. In fact, Yermack (2006) finds that the use of a corporate jet leads to significant underperformance of the company’s stock and is the most costly CEO benefit. More common issues involve investing in overly risky projects and managerial shirking. Moreover, earnings management is generally also considered a form of moral hazard. The manipulation of earnings can serve several personal goals of the agent, such as receiving a higher bonus, meeting earnings targets and equity rewards (Healy, 1985; Healy & Wahlen, 1999; Cheng & Warfield, 2005).

2.2. Earnings management

Previous literature shows that earnings management can be viewed from two different perspectives. First, from a contracting perspective, managers can use EM to increase efficiency by protecting the firm from the consequences of unforeseen events when contracts are incomplete (Scott, 2014). Secondly, managers may use EM out of self-interest to avoid reporting losses or to meet earnings forecasts. The focus of this study lies on the latter, the more opportunistic view, therefore the definition used in this thesis is “a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain” (Schipper, 1989: 92).

There are two main earnings management techniques. The first and most discussed method to manipulate earnings is based on accruals. Accruals are used to help investors to evaluate the economic performance for a specific period (Dechow & Skinner, 2000). The U.S. GAAP offers flexibility in financial reporting choices such as valuation and depreciation methods to provide organizations various possibilities to show the true economic position in the financial statements. However, earnings management can also take place within GAAP by choosing the method that makes the financial statements look better than it is in reality, for example switching from conservative accounting procedures to aggressive accounting practices that are allowed by the U.S. GAAP (Dechow & Skinner, 2000). Although it does not impact an organization’s activities, manipulating accruals does have an impact on future financial statements due to accrual reversal (Scott, 2014). This means if a firm boosts

(8)

profits by altering accruals, these accruals will have to be reversed in the future which leads to lower profits.

Secondly, real earnings manipulation can be used to manage earnings by deviating from normal operations, motivated by management who wants to mislead stakeholders (Roychowdhury, 2006). Cohen et al. (2008) show that after the implementation of the Sarbanes-Oxley Act1, there has been a shift from accrual-based earnings management to real earnings manipulation. According to Graham, Harvey and Rajgopal’s (2005) survey, firms shift because they believe that real earnings manipulation is harder to detect, although it is more costly. Roychowdhury’s (2006) paper demonstrates how firms use this method of EM: through sales manipulation, reduction of expenditures and overproduction. Sales manipulation is done by offering special discounts or being more lenient in credit sales to increase revenue. The reduction of discretionary expenditures is for example reducing or postponing investments in R&D, or marketing costs. Lastly, overproduction is used to lower fixed overhead costs which leads to a lower COGS, therefore leading to higher profit margins.

The agency theory illustrates how management can be either motivated or demotivated to engage in earnings management. On the one hand, governance mechanisms such as board independence and/or an audit committee decrease the amount of EM by reducing information asymmetry (Klein, 2002). Thus, it limits the opportunity to manipulate earnings. On the other hand, (bonus) contracts are used to align interests, but can create incentives to manipulate earnings when compensation is tied to a certain benchmark (Healy & Wahlen, 1999; Watts & Zimmerman, 1978). Here the information asymmetry leads to overly risk-taking behavior called moral hazard, including EM. Most prior research has focused on internal CG mechanisms, such as the previous examples given. However, these internal mechanisms do not operate in isolation; they are affected by what happens around them (Aguilera et al., 2015). Thus, it is also important to study the effects of external corporate governance mechanisms on firm behavior, which has been up and coming in the last years. For instance, market competition is a critical external factor that affects all firms and their operations. That is why in this thesis, I will study the effect of market competition on the extent of earnings management. According to agency theory, it can either reduce information asymmetry and lessen EM practices, or entice management to increased moral hazard to beat the competition.

2.3. The relationship between competition and earnings management

2.3.1. Competition as an external governance mechanism

Competition is often praised for its beneficial value – it ensures low prices, better quality and a wider range of products for the customer. Famous economists have argued for the positive effects of competition, including its effect on earnings management. One of the first ones to outline the positive effect of competition on managerial behavior is Hart (1983). He argues that when one firm’s costs decrease, the costs of competing firms in the same industry also decrease, which leads to an increase in supply whereby the product price drops. When only one firm’s costs would decrease, its management can slack, while as when the entire industry has lower costs, they cannot slack as much. In other words, there is less information asymmetry in heightened competition, which suggests that market competition limits the opportunity for earnings management. On a similar note, Holmstrom (1982) highlights the

1_{The Sarbanes-Oxley Act was established in 2002 after large accounting scandals and includes legislation to help protect stakeholders from}

(9)

presence of multiple firms that can be used for peer performance. Moreover, he finds competition itself to have no benefits, there is only value in the information that can be obtained from relative performance measures. This makes it more difficult for firms to show overly positive results without others noticing. This also reduces information asymmetry, which makes it more difficult for management to manipulate earnings. Furthermore, more recent studies have found that competition leads to decreased private benefits for executives, and it leads to higher cash payouts to the shareholders (Guadalupe & Perez-Gonzalez, 2010; Grullon & Michaelly, 2014). All in all, these arguments come down to the fact that competition acts as an external governance mechanism that helps to align the interest between the shareholders and executives, reduces information asymmetry leading to lower agency costs. Competition would thus lead to less earnings management.

Several studies confirm this view and find a negative relationship between competition and earnings management (e.g. Balakrishnan & Cohen, 2014; Laksmana & Yang, 2014; Marciukaityte & Park, 2009). Besides the above-mentioned arguments, they remark that the stock market can act as a disciplining force for management misbehavior. Research shows that false reporting leads to negative stock market reactions, and more specifically, when firms make earnings-decreasing restatements, their stock price drops (Palmrose, Richardson & Scholz, 2004; Marciukaityte & Varma, 2008). This means that if earnings management practices are discovered, there are likely to be negative consequences for the firm’s value and this relationship is stronger for firms that operate in more competitive industries (Marciukaityte & Park, 2009). Thus, firms operating in competitive industries will be punished more severely in the stock market than firms in low competition. Since management does not want that its stock price drops, we can see competition as a force that disciplines managers to not mislead stakeholders by falsely reporting.

Balakrishnan and Cohen (2014) conclude that when a firm in a heightened competitive environment engages in too risky behavior, i.e. investing in projects with a negative net present value and excessive expenditures, the firm becomes less competitive and is thus more likely to be pushed out of the market. Predation risk plays a part here: the risk of losing market share and investment possibilities to the competition (Froot et al., 1993). Firms in competitive industries cannot risk making mistakes, because there is more competition that can take their place in the market and drive them out of business. For accrual-based EM, this risk expresses itself by punishment in the stock market described in the previous paragraph. Furthermore, Shi, Sun and Zhang (2018) find a negative relationship between competition and real earnings management and argue it is due to the high costs, high risks, and limited options. For example, investments in R&D and marketing can be essential for firm survival. Furthermore, offering discounts in order to boosts sales is detrimental to profit margins that are often very low already in competitive industries. To conclude, all arguments in this subsection point at the increased risks firms face in increased competition, which would lead rational managers to not engage in opportunistic behavior such as misreporting financial information.

2.3.2. Competition as a dark force

Another stream of literature argues that competition can lead to unethical business practices. For example, if a bartender does not serve alcohol to a minor, or a taxi driver won’t speed to the airport, they lose customers that will go to their competitors. Such potential loss of clients by obeying the law creates an incentive for firms to engage in unethical behavior (Bennet, Pierce, Snyder, & Toffel, 2013).

(10)

In this vein, Shleifer (2004) highlights the negative effects of market competition, as he calls the “dark side” of competition. He argues that increased competition may stimulate managers to engage in a higher degree of EM. Shleifer (2004) states that many growth firms show their financial statements in a different format than the usual, which displays earnings more favorably. Although it is not considered as earnings management, it may potentially mislead stakeholders. Furthermore, he finds the battle for funds to be a key incentive to manipulate earnings. With a higher valuation, the cost of capital goes down which makes it possible for a firm to grow or survive by making acquisitions for stock or issuing new shares for example. Moreover, when firms are in competition for funds and there are larger financing needs, it motivates managers to use earnings management (Lin et al., 2015). It is assumed that when there are more competitors within an industry, the battle for funds is also larger. All in all, due to competitive pressure to survive in the market and the limited funds available, management is induced to resort to opportunistic behavior such as aggressive accounting methods.

Various recent studies support the ideas echoed by Shleifer (e.g. Markarian & Santalo, 2014; Lin et al., 2015; Karuna et al., 2015; Datta et al., 2013). When competition in a market increases, so does the risk of every firm operating in that market. More specifically, it can reduce market share, firm profitability, and the possibility to achieve excellent results (Shi et al., 2018). This makes it harder to meet all stakeholder needs, ranging from value maximization to not violating debt covenant restrictions. The increased difficulty of operating an organization under competition also enhance managers’ concerns about their careers. These concerns are justifiable, as DeFond and Park (1999) find that intensified competition leads to higher CEO turnover and Jung and Subramanian’s (2017) study shows that CEO talent is more important in competitive markets. Due to these reasons, executives potentially experience higher pressure to prove themselves and safeguards their career opportunities, which possibly creates an incentive to distort (financial) information (Hermalin & Weisbach, 2007). Overall, because of the increased difficulty of managing a firm in high competition, career concerns can play a role in the extent of earnings management of an organization.

Furthermore, a positive relationship between competition and earnings management can be explained by managerial myopia. When a firm is struggling to survive in a competitive industry, its focus is in the first place on short term results. Feinberg (1995) states that competition inherently leads to short-term gains, which leads to myopic behavior. Focus on short term results can incentivize management to use aggressive accounting techniques that boost short-term profits (Karuna et al., 2015). Moreover, Frésard and Valta (2015) find that when competition increases, there is a decline in capital and R&D investments which is a clear example of real earnings manipulation according to Roychowdhury (2006). Besides managers focusing on the short term, it is also proven that investors in the stock market act myopically in the U.S. (Black & Fraser, 2002). This means that good short-term results are rewarded in the stock market, which can be an incentive for firms under competition to act opportunistically and misreport (Markarian & Santalo, 2014).

The last argument important to mention is the desire of firms to hide their true performance from their competitors (Karuna et al., 2015). Theory predicts that in competitive industries, organizations will choose to withhold information, and if that is not possible they might manipulate their earnings to mislead the competition. In a monopoly, there is no need to misreport because there is no competition to fool. Therefore, competition can encourage executives to engage in earnings management.

(11)

2.3.3. Hypothesis development

It can be concluded that previous literature offers contrasting results and views on the influence of competition on earnings management. One side argues for the disciplining effect of market competition where agency problems are reduced, while the other side states competition brings out the worst in executives of an organization where agency problems are increased. Given there is equal evidence and theoretical support for both sides, I formulate the following non-directional hypothesis:

Hypothesis 1: Market competition has an effect on the level of earnings management.

2.4. The explanatory role of political uncertainty

The main novelty of this thesis is that I will study if political uncertainty acts as a conditional variable that can explain why in one circumstance the association between competition and earnings management is positive, while in the other it is negative. In this research, political uncertainty is defined as “uncertainty about the government’s future actions” (Pastor & Veronesi, 2013: 2).

Recently, there is a rise in studies that focus on political uncertainty in regard to the finance and accounting literature. Current events show that firms cannot ignore politics: trade wars, tax cuts, and economic policies are all able to affect business decisions. This is proven by scientific research, which shows that uncertainty about government actions leads to increased risk premia and stock volatility, and in election years corporations tend to reduce investment expenditures (Julio & Yook, 2012; Pastor & Veroseni, 2013; Jens, 2017). Similar to competition, political uncertainty is an external factor that organizations cannot directly influence.

I expect that political uncertainty influences the degree of earnings management. This relationship can be explained by the use of agency theory, and more specifically by information asymmetry. Nagar et al. (2019) and Francis, Hasan and Zhu (2014) find that uncertainty about policies leads to an increase in information asymmetry between an organization and its capital providers. For investors, it is more difficult to assess the impact of the uncertainty than the firm can itself, which is shown by the fact that investors’ reaction to earnings surprises is less strong in times of political uncertainty (Nagar et al., 2019). Information asymmetry is essential for managers to engage in earnings manipulation, otherwise, shareholders would know the financial statement is not in line with the true economic state of the firm. Given in times of higher uncertainty there is higher information asymmetry, I expect that managers will engage in a higher degree of earnings management because there is more opportunity to do so.

Besides more opportunities to use EM practices in times of political uncertainty, there might also be an increase in motivation to engage in it. The political cost hypothesis describes that firms will manage earnings to reduce the costs of a potential regulatory outcome under the uncertainty of changing regulation (Watts & Zimmerman, 1978). In other words, firms are willing to engage in earnings management to avoid the government creating or changing certain policies that will have a negative effect on their organization, i.e. higher taxes or import tariffs. Hence, in times of higher political uncertainty, and thus higher uncertainty on governmental outcomes, managers can experience an increased motivation to manage earnings. In conclusion, I expect that management is more likely to engage in EM when there is high political uncertainty, due to increased opportunity and motivation.

(12)

Given this reasoning, I believe political uncertainty moderates the relationship between competition and earnings management in such a way that in times of high political uncertainty, the relationship between competition and EM is strengthened in a way that competition leads to more EM and vice versa. Higher political uncertainty creates more opportunity and motivation to manipulate earnings, which I believe can entice the “dark side of competition” where management is tempted to behave unethically. Political uncertainty can be the final reason that motivates executives to engage in EM. In addition, in the context of low uncertainty, there is less opportunity and motivation for EM and then political stability can spark the good side of competition. This leads to the following hypothesis:

Hypothesis 2: Political uncertainty moderates the relationship between market competition and the level of earnings management such that in a context of high (low) political uncertainty, the impact of market competition on the level of earnings management is strengthened (weakened).

(13)

3. Research design

3.1. Sample selection

The dataset used in this study comprises 9,332 observations relating to 1,571 unique firms for the years 2002 to 2016 (including). I collect the data from two separate databases. Firstly, information to determine the Herfindahl-Hirschman Index, earning management proxies and control variables come from the COMPUSTAT database2. Secondly, I employ the Hassan et al. (2019) database to gather information about firm-level politically induced risk. The latter database comprises data for U.S.-listed firms for the years 2002 to 2016. This study uses annual data because it is the evident choice regarding the topic of EM: numerous firms experience seasonality where annual numbers offer more significant information and the annual report receives more attention than other reports. Thus, I assume that management focuses on the manipulation of annual numbers. Furthermore, I use the Fama-French 49-industry classification for the calculation of the measure of competition and earnings management, where I use all available data from COMPUSTAT. Additionally, I omit the observations from the “Banking” and “Insurance” industries due to the difficulty of defining (abnormal) accruals for financial institutions (Klein, 2002). In line with prior research, I also exclude observations related to industries that have less than ten observations. To mitigate outliers that can influence inferences from regression analyses, I winsorize all continuous variables at the 1% and 99% levels (Leone, Minutti-Meza & Wasley, 2013). In table 1 below you find the distribution of the sample by industry.

Industry Number of observations Proportion

Almost Nothing3 ₇₅ _0.80%

Apparel 81 0.87%

Automobiles and Trucks 164 1.76%

Business Services 233 2.50% Chemicals 145 1.55% Communication 154 1.65% Computer Hardware 512 5.49% Computer Software 1,798 19.27% Construction Materials 229 2.45% Consumer Goods 251 2.69% Electrical Equipment 216 2.31% Electronic Equipment 799 8.56% Entertainment 167 1.79% Food Products 221 2.37% Healthcare 54 0.58% Machinery 378 4.05%

Measuring and Control Equipment 336 3.60%

Medical Equipment 570 6.11%

Pharmaceutical Products 675 7.23%

Recreation 140 1.50%

Restaurants, Hotels, Motels 461 4.94%

Retail 1,353 14.50%

Trading 71 0.76%

Wholesale 249 2.67%

Total 9,332 100%

Table 1: Distribution of firms by the Fama-French 49-Industry Classification

2_{COMPUSTAT is a part of the Wharton Research Data Services which includes financial information from firms worldwide.}

(14)

3.2. Methodology

3.2.1. Independent variable: market competition

I measure the degree of market competition using the Herfindahl-Hirschman Index (HHI). The HHI is a statistical measure of concentration and it can be applied to various contexts - in this case in the context of market competition. The HHI is a commonly used measure in this type of research as market concentration is an important indicator of market competition, where even the U.S. Federal Reserve uses it to determine the competitive effects of mergers (Rhoades, 1993). Also, various similar studies include it as a measure of market competition (e.g. Markarian & Santalo, 2014; Lin et al., 2015). The formula is as follows:

𝑯𝑯𝑰𝒋,𝒕= ∑(𝑴𝑺𝒊)² 𝒏

𝒊=𝟏

(1)

Where MS represents the market share in year t and industry j of firm i and there are n firms in the market. The market share is calculated based on the amount of sales per firm in the industry. To exemplify, this means that in a monopoly the firm’s market share is 1 and the HHI is 1² = 1 (the maximum amount) whereas when there are two firms in the industry with equal sales, the HHI is 0.5² + 0.5² = 0.5. Thus, the relation between the HHI and competition is negative, meaning a high (low) HHI-score indicates a low (high) degree of competition.

However, there is some criticism on the HHI as a proxy for market competition as well. Firstly, I compute the HHI with available information from public firms. There might be private firms operating in industries that are not included in my dataset, leading to a noisy measure (Markarian & Santalo, 2014). Moreover, some argue that the index is too sensitive and attaches too much weight to the number of firms in the industry (Hart, 1975; Davies, 1980). Nonetheless, the issue of which measurement best represents competition is not yet resolved (Curry & George, 1983). Therefore, although the measure is not free from criticism, I will use the HHI in this thesis given it is still one of the most common proxies of market competition.

3.2.2. Dependent variable: earnings management

3.2.2.1. Accrual based earnings management

The degree of accrual-based earnings management (AEM) is measured through discretionary accruals, which can be done in several ways. Dechow et al. (1995) discuss various methods and develop a modified version of the Jones Model, which differs from the regular model by exercising discretion over revenues by subtracting these by net receivables. Hence, the modified Jones Model is less likely to measure discretionary accruals with error and is the most powerful method according to Dechow et al. (1995). Therefore, I will use this model in my research. In addition, prior research often uses this method to detect AEM (e.g. Markarian & Santalo, 2014; Zang, 2012; Shi et al., 2018). Discretionary accruals are a proxy for abnormal accruals, which are calculated by subtracting non-discretionary accruals from total accruals. I start by calculating total accruals according to the balance sheet approach as follows:

(15)

Where Δ𝐶𝐴_𝑡 represents the change in current assets between year t and t-1, Δ𝐶𝐿_𝑡 stands for the change in current liabilities between year t and t-1 and Δ𝐶𝑎𝑠ℎ𝑡 is the change in cash and cash equivalents between year t and t-1. Δ𝐷𝐶𝐿_𝑡 is added which stands for the change in the current portion of long-term debt between year t and t-1 and finally 𝐷𝑒𝑝_𝑡 is deducted which represents depreciation and amortization expenses from year t.

As mentioned, total accruals (TACC) can be divided into discretionary (DACC) and non-discretionary accruals (NDACC). The non-discretionary part identifies the conditions of a business, i.e. growth and operating cycles. The discretionary accruals reflect the choices made by firm management and are therefore used as a proxy for AEM. The Modified Jones model calculates the parameters necessary to determine the DACC. The parameters are established by an OLS regression as follows:

𝑇𝐴𝐶𝐶_𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1 + 𝛼2 Δ𝑅𝐸𝑉_𝑡− Δ𝑅𝐸𝐶_𝑡 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸_𝑡 𝐴𝑡−1 + 𝜀𝑡 (3) Where 𝐴𝑡−1 stands for total assets at the end of t-1, Δ𝑅𝐸𝑉𝑡− Δ𝑅𝐸𝐶𝑡 represents the change in revenue between year t and t-1 subtracted by the change in net receivables between t and t-1 and Δ𝑃𝑃𝐸_𝑡 means gross property, plant, and equipment in year t. Moreover, 𝛼₁, 𝛼₂, and 𝛼₃ are the estimated parameters and 𝜀𝑡 is the residual. With these estimated parameters and the residual, I can compute the NDACC as follows: 𝑁𝐷𝐴𝐶𝐶_𝑡 𝐴_𝑡−1 = 𝛼1 1 𝐴_𝑡−1 + 𝛼2 Δ𝑅𝐸𝑉_𝑡− Δ𝑅𝐸𝐶_𝑡 𝐴_𝑡−1 + 𝛼3 𝑃𝑃𝐸_𝑡 𝐴_𝑡−1 (4)

Lastly, after subtracting the non-discretionary accruals from the total accruals I end up the abnormal accruals for my regression model. I will use the absolute version of the abnormal accruals (abs_AEM) because I want to capture earnings management practices without a specific directional prediction (Hribar & Nichols, 2007).

3.2.2.2. Real earnings manipulation

In this thesis, real earnings manipulation (REM) is calculated according to the method of Roychowdhury (2006). He argues management can manipulate real earnings in three ways, namely through sales manipulation, overproduction and reduction of discretionary expenditures. Roychowdhury’s (2006) model uses three proxies to capture these, which have been frequently used as a proxy for REM (e.g. Zang, 2012; Cohen et al., 2008). In my final regression, I will use an aggregate measure of the REM proxies.

The first proxy to capture sales manipulation stems from the abnormal level of cash flow from operations. To do so, I first have to calculate the normal level of cash flow per industry per year as follows: 𝐶𝐹𝑂_𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1 1 𝐴𝑡−1 + 𝛽1 𝑆_𝑡 𝐴𝑡−1 + 𝛽2 Δ𝑆_𝑡 𝐴𝑡−1 + 𝜀𝑡 (5)

(16)

Where 𝐶𝐹𝑂_𝑡 is the cash flow from operations, 𝑆_𝑡 represents sales during year t and Δ𝑆_𝑡 stands for the change in sales between year t and t-1. Subsequently, the abnormal degree of CFO (abn_CFO) is the actual CFO subtracted by the normal CFO per year per industry calculated according to formula 5. Abn_CFO is thus the first proxy used to determine the level of REM.

Next, I establish the proxy for overproduction with the following formula: 𝑃𝑅𝑂𝐷𝑡 𝐴_𝑡−1 = 𝛼0+ 𝛼1 1 𝐴_𝑡−1 + 𝛽1 𝑆𝑡 𝐴_𝑡−1 + 𝛽2 Δ𝑆𝑡 𝐴_𝑡−1 + 𝛽3 Δ𝑆𝑡−1 𝐴_𝑡−1 + 𝜀𝑡 (6) Here 𝑃𝑅𝑂𝐷_𝑡 stands for the normal production costs per industry per year, calculated by the cost of goods sold in year t plus the change in inventory between year t and t-1. The abnormal production costs (abn_PROD) are calculated in the same manner as the abnormal cash flows, namely by subtracting the normal costs by the actual costs. This produces abn_PROD, the second proxy of REM.

Lastly, discretionary expenses can be used to manipulate real earnings. Discretionary expenses are calculated as the sum of advertising expenses, R&D expenses, and SGA (selling, general and administrative) expenses. With the following formula, I detect the normal amount of these expenses:

𝐷𝐼𝑆𝐸𝑋𝑃_𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1 1 𝐴𝑡−1 + 𝛽 𝑆𝑡−1 𝐴𝑡−1 + 𝜀𝑡 (7) The normal discretionary expenses are subtracted from the actual expenses, resulting in the final proxy abn_DISEXP. As mentioned, I will use an aggregate measure of REM in my empirical model, which is the sum of abn_CFO, abn_PROD and abn_DISEXP. Here I first multiply abn_CFO and abn_DISEXP by -1 before summing the proxies up (Franz, HassabElnaby & Lobo, 2014). This is due to the different influence the proxies have – the lower the abnormal cash flow and discretionary expenses, the more income increasing real earnings manipulation. Because production costs increase when engaging in income increasing REM, this is not multiplied by -1.

3.2.3. Moderating variable: political uncertainty

In order to measure political uncertainty, I employ the measure of political risk derived from the Hassan et al. (2019) database to assess the political uncertainty of a firm. They develop the level of political risk by looking at the share of their quarterly earnings conference calls that firms devote to political risk and do this for 178,173 conference calls of 7,357 individual firms from the U.S. between 2002 and 2016. By using computational linguistics, Hassan et al. (2019) are able to identify the degree of political risk with the following formula:

𝑃𝑅𝐼𝑆𝐾𝑖𝑡 = ∑ ( 1[𝑏 ∈ ℙ\ℕ] × 1[|𝑏 − 𝑟| < 10] × 𝑓_𝐵𝑏,ℙ ℙ 𝐵𝑖𝑡 𝑏 𝐵_𝑖𝑡 (8)

Hassan et al. (2019) identify two-word combinations, called bigrams, that are frequently used in political texts and apply these to the transcripts of the quarterly earnings calls, and they do the same for

(17)

nonpolitical risk. Next, the authors count how often the political bigrams come into proximity to terms such as “risk” and “uncertainty” and divide this number by the total number of bigrams, which results in the measure of political risk. The measure entails that the more time spent discussing political risks, the higher the amount of political risk. Furthermore, Hassan et al. (2019) prove the validity of PRISK by showing it corresponds with certain actions taken by firms that are exposed to risk, i.e. a decrease in investments and employees and an increase in lobbying and donations to politicians.

As I use annual data in the final regression models for this study, I convert the quarterly data from the Hassan et al. (2019) database to annual data. I choose to generate an average amount of political risk for each year. This is done by summing the political risk from the quarters available each year for a firm and dividing it by the amount of quarters available in that year, given that some years did not contain all four quarters. I will use this yearly average of political risk as a moderating variable in my regression.

3.2.4. Control variables

To avoid invalidate results, I include control variables that can influence the degree of earnings management. This ensures a higher internal validity of this research. First, I control for firm size (SIZE) as is common in prior research (e.g. Markarian & Santalo, 2014; Laksmana & Yang, 2014). Larger firms have more resources to avoid EM practices such as monitoring and control systems, therefore I expect a negative relation between firm size and EM. Firm size is measured by the natural logarithm of total assets, to ensure a normal distribution. Furthermore, I control for growth opportunities by including the market to book (MTB) ratio as a control variable (e.g. Roychowdhury, 2006; Shi et al., 2018). This is calculated by dividing the market value of equity by the book value of equity. Growth firms (high MTB) experience more information asymmetry than value firms, which is the primary reason they are more likely to engage in EM (Madhogarhia, Sutton & Kohers, 2009). Furthermore, the management of growth firms might feel pressure to keep performing well. This means I expect a positive association between the two variables. Next, leverage (LEV) is included which is measured by dividing the long-term debt by total assets (Shi et al., 2018; Laksmana & Yang, 2014; Markarian & Santalo, 2014). The direction of leverage is unclear – on the one hand, it is argued that creditors serve as monitors that help decrease EM while on the other hand debt covenants can create an incentive to engage in EM when a firm is near violation of the covenant (Jensen, 1986; Becker, DeFond, Jiambalvo, & Subramanyam, 1998). The next variable I control for is cash flow volatility (CFVOL), which is correlated with earnings management according to Hribar and Nichols’ (2007) research. To control for performance, I include Return on Assets (ROA) (Laksmana & Yang, 2014). Better performance can explain a higher degree of accruals, so it has a positive effect on EM. Furthermore, I take the audit quality of the firm into account by creating a dummy variable (BIG4AUD) for Big Four 4and non-Big Four auditors as is often done in previous studies (e.g. Laksmana & Yang, 2014; Shi et al., 2018). The dummy takes a value of 1 for a Big Four auditor and zero otherwise, where the Big Four auditor is expected to deliver higher audit quality. Becker et al. (1998) find that increased audit quality constraints EM because large audit firms have more resources and capabilities to detect accounting irregularities. This means I expect the dummy to negatively influence the degree of EM. Next, Franz et al., (2014) argue that AEM and REM can be used as substitutes or complements, therefore it is important to control for one version of EM when regressing the other. Hence, I include ABS_AEM (REM) as a control variable in the regression where

(18)

REM (ABS_AEM) is the dependent variable. Lastly, I include an industry and year dummy to control for time and industry-specific effects.

3.3. Empirical model

In order to gather evidence to test my hypotheses, I use an OLS regression to find the association and significance of this association between my main variables. For the first hypothesis, where I investigate the effect of competition on two different measures of earnings management, I run the following empirical models: 𝐴𝐵𝑆_𝐴𝐸𝑀_𝑡 = 𝛽₀ + 𝛽₁𝐻𝐻𝐼_𝑡+ 𝛽₂𝑅𝐸𝑀 + 𝛽₃𝑆𝐼𝑍𝐸_𝑡+ 𝛽₄ 𝑀𝑇𝐵_𝑡+ 𝛽₅ 𝐿𝐸𝑉_𝑡+ 𝛽₆𝑅𝑂𝐴_𝑡 + 𝛽₇𝐶𝐹𝑉𝑂𝐿_𝑡+ 𝛽₈𝐵𝐼𝐺4𝐴𝑈𝐷_𝑡+ 𝛽₉ 𝑌𝐸𝐴𝑅_𝐷𝑈𝑀𝑀𝑌 + 𝛽10𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌_𝐷𝑈𝑀𝑀𝑌 + 𝜀𝑡 Model 1 𝑅𝐸𝑀_𝑡= 𝛽₀+ 𝛽₁𝐻𝐻𝐼_𝑡+ 𝛽₂𝐴𝐵𝑆_𝐴𝐸𝑀_𝑡+ 𝛽₃𝑆𝐼𝑍𝐸_𝑡+ 𝛽₄ 𝑀𝑇𝐵_𝑡+ 𝛽₅ 𝐿𝐸𝑉_𝑡 + 𝛽₆𝑅𝑂𝐴_𝑡+ 𝛽₇𝐶𝐹𝑉𝑂𝐿_𝑡+ 𝛽₈𝐵𝐼𝐺4𝐴𝑈𝐷_𝑡+ 𝛽₉ 𝑌𝐸𝐴𝑅_𝐷𝑈𝑀𝑀𝑌 + 𝛽10𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌_𝐷𝑈𝑀𝑀𝑌 + 𝜀𝑡 Model 2

In these models, the variable of interest is the coefficient 𝛽1. The direction of this coefficient determines whether competition has a positive or negative effect on earnings management, and the level of significance shows if the hypothesis can be accepted or not. I will accept the hypothesis if the p-value is lower than 0.05.

Moreover, this thesis aims to explain the relationship between competition and earnings management by the use of the conditional variable political risk. To test the second hypothesis where political risk is the moderator, I run the following empirical models:

𝐴𝐵𝑆_𝐴𝐸𝑀𝑡 = 𝛽0+ 𝛽1𝐻𝐻𝐼𝑡+ 𝛽2𝑃𝑅𝐼𝑆𝐾𝑡+ 𝛽3(𝐻𝐻𝐼𝑡 × 𝑃𝑅𝐼𝑆𝐾𝑡) + 𝛽4 𝑅𝐸𝑀𝑡 + 𝛽₅ 𝑆𝐼𝑍𝐸 + 𝛽₆𝑀𝑇𝐵_𝑡+ 𝛽₇𝐿𝐸𝑉_𝑡+ 𝛽₈𝑅𝑂𝐴_𝑡+ 𝛽₉𝐶𝐹𝑉𝑂𝐿_𝑡 + 𝛽₁₀𝐵𝐼𝐺4𝐴𝑈𝐷_𝑡+ 𝛽₁₁ 𝑌𝐸𝐴𝑅_𝐷𝑈𝑀𝑀𝑌 + 𝛽₁₂𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌_𝐷𝑈𝑀𝑀𝑌 + 𝜀𝑡 Model 3 𝑅𝐸𝑀𝑡 = 𝛽0+ 𝛽1𝐻𝐻𝐼𝑡+ 𝛽2𝑃𝑅𝐼𝑆𝐾𝑡+ 𝛽3(𝐻𝐻𝐼𝑡 × 𝑃𝑅𝐼𝑆𝐾𝑡) + 𝛽4 𝐴𝐵𝑆_𝐴𝐸𝑀𝑡 + 𝛽5 𝑆𝐼𝑍𝐸 + 𝛽6𝑀𝑇𝐵𝑡+ 𝛽7𝐿𝐸𝑉𝑡+ 𝛽8𝑅𝑂𝐴𝑡+ 𝛽9𝐶𝐹𝑉𝑂𝐿𝑡 + 𝛽₁₀𝐵𝐼𝐺4𝐴𝑈𝐷_𝑡+ 𝛽₁₁ 𝑌𝐸𝐴𝑅_𝐷𝑈𝑀𝑀𝑌 + 𝛽₁₂𝐼𝑁𝐷𝑈𝑆𝑇𝑅𝑌_𝐷𝑈𝑀𝑀𝑌 + 𝜀_𝑡 Model 4

Here the variable of interest is the coefficient 𝛽₃, which shows the moderating effect of political risk. Since I assume that increased political risk has a positive influence on EM, I expect the coefficient to be positive as well. Furthermore, I will also accept hypothesis 2 if the p-value is lower than 0.05.

(19)

4. Results

4.1. Descriptive statistics

Table 2 provides an overview of the summary statistics of the key variables which are used in the empirical models. In addition, I include the separate REM proxies. The mean of the absolute value of abnormal accruals is 0.063, which is similar to prior research (e.g. Klein, 2002; Laksmana & Yang, 2014). Abn_CFO is positive which implies that most firms in this sample have lower abnormal cash flows than the industry average. Also, abn_PROD has a negative mean, thus the sample-firms produce less than the industry average. Lastly, the mean of abn_DISEXP is close to zero but also negative, which shows that on average abnormal discretionary expenses are higher in this sample than average. In total, the median of the combined measure of REM is -0.154, from which I can conclude that the majority of firms in this sample do not use real earnings manipulation to increase their earnings. They might use it to decrease earnings.

Next, the descriptive statistics of the independent variable HHI are largely consistent with previous literature (e.g. Markarian & Santalo, 2014; Lin et al., 2015). The sample consists of mainly growth firms given the high market to book ratio, where the market value is on average 3.310 as large as its book value. Besides, the firms report a relatively low leverage ratio which implies that the firms are mainly financed by equity. Firms have a negative return on assets on average where the median is positive, which means if a firm has a loss it is larger than the profit. This illustrates that when firms have to report a loss, they often take a “big bath” (Scott, 2014). Lastly, 79.96% of the sample is audited by a Big 4 firm.

Table 3 reports on the correlation between the variables in the empirical model. The coefficients of the pairwise combinations do not trespass the value of 0.7 and thus show there is no collinearity. Due to several higher indicators (>0.4), I perform a more in-depth analysis of collinearity by examining the VIFs (variance inflating factors). These are all between the values of 1.01 and 2.56, from which I can conclude there is no high correlation between the variables.

Variable N Mean σ Quartile 1 Median Quartile 3

AEM5 _9,332 _-0.010 _0.086 _-0.032 _-0.010 _0.053 ABS_AEM 9,332 0.063 0.066 0.020 0.043 0.082 REM 9,332 -0.196 0.473 -0.435 -0.154 0.077 Abn_CFO 9,332 0.117 0.212 0.003 0.100 0.216 Abn_PROD 9,332 -0.105 0.209 -0.212 -0.092 0.013 Abn_DISEXP 9,332 -0.023 0.292 -0.184 -0.051 0.095 HHI 9,332 0.068 0.036 0.047 0.056 0.076 PRISK 9,332 91.275 102.701 28.616 58.809 114.391 SIZE 9,332 6.247 1.921 5.060 6.247 7.665 MTB 9,332 3.310 5.829 1.409 2.442 4.167 LEV 9,332 0.156 0.195 0.000 0.080 0.251 ROA 9,332 -0.014 0.203 -0.035 0.037 0.085 CFVOL 9,332 90.164 241.570 6.097 17.887 56.058 BIG4AUD 9,332 0.796 0.403 1 1 1

Table 2: Summary statistics

5_{The mean and medium of AEM are close to zero. This means the sample represents no systematic upwards or downwards} earnings management and I can use the absolute value of the abnormal accruals (ABS_AEM) (Klein, 2002).

(20)

ABS_AEM REM HHI PRISK SIZE MTB LEV ROA CFVOL BIG4AUD ABS_AEM 1.000 REM -0.038*** 1.000 HHI 0.081*** 0.050*** 1.000 PRISK 0.029*** -0.034*** 0.011 1.000 SIZE -0.236*** 0.085*** -0.089*** -0.034** 1.000 MTB 0.030*** -0.193*** -0.039*** -0.008 0.052*** 1.000 LEV -0.057*** 0.076*** -0.065*** -0.010 0.289*** -0.065*** 1.000 ROA -0.173*** -0.058*** -0.074*** -0.057*** 0.405*** 0.077*** -0.060*** 1.000 CFVOL -0.079*** 0.020* -0.047*** 0.030*** 0.576*** 0.022** 0.104*** 0.117*** 1.000 BIG4AUD -0.173*** -0.052*** -0.061*** -0.043*** -0.472*** 0.058*** 0.114*** 0.220*** 0.163*** 1.000

Notes: N = 9,332. Significance levels are as follows: * p < 0.1, ** p < 0.05, *** p < 0.01

(21)

4.2. Regression analysis

Table 4 presents the output of the regression analyses of the models that test the two hypotheses. The output is established by performing an OLS regression. Here I employ fixed effects for industry and time, and I cluster the standard errors on two dimensions, namely by firm and year.

Model 1 shows that there is no statistically significant association between the degree of competition and the degree of earnings management (β=-0.110 and p=0.171). More specifically, this means that the coefficient does not differ from zero, which means that the first hypothesis based on AEM is not accepted. In addition, only 9.6% of the variance in AEM is explained by the independent variables in this analysis, which means that the explanatory power of this model is low. Furthermore, various control variables are significantly related to the level of absolute accrual based earnings management, where firm size (β=-0.009), return on assets (β=-0.022) and the audit quality (β=-0.008) decrease the level of AEM and market to book ratio (β=0.000) and cash flow volatility (β=0.000) have a slightly positive effect.

Furthermore, the second column of table 4 shows that competition is also not associated with real earnings manipulation (p=0.892). Accordingly, the first hypothesis based on REM can also not be accepted. The adjusted R² of this model is 20.7%, which means that the independent variables including competition explain around 20 percent of the variance in the degree of REM. Table 4 shows that firm size (β=0.044) is positively associated with real earnings management, whereas the market to book ratio (β=-0.012), return on assets (β=-0.312) and audit quality (β=-0.145) are negatively associated to REM. Model 3 and 4 show the results of the moderating effect of political risk on the two techniques of EM. For both models, the interaction term is almost non-directional (for AEM β=0.000 and for REM β=-0.001) as well as not statistically significant (respectively p=0.128 and p=0.737). Hence, hypothesis 2 is not accepted. I notice that the results of models 1 and 3, as well as models 2 and 4, are very similar to each other. The coefficients, explanatory power, and levels of significance are almost the same, where the biggest difference is that in model 4 the coefficient is slightly larger than in model 2 (respectively

β=0.203 and β=0.082). Moreover, the last column of table 4 shows that political risk is significantly

associated with the degree of real earnings manipulation (β=0.000, p=0.044), whereas the coefficient for AEM is not statistically significant (β=0.000, p=0.565).

(22)

Variable Model 1 ABS_AEM Model 2 REM Model 3 ABS_AEM Model 4 REM HHI -0.110 (0.081) 0.082 (0.607) -0.131 (0.086) 0.203 (0.604) PRISK 0.000 (0.000) 0.000** (0.000) HHI * PRISK 0.000 (0.000) -0.001 (0.001) SIZE -0.009*** (0.001) 0.044*** (0.008) -0.009*** (0.001) 0.043*** (0.008) MTB 0.000*** (0.000) -0.012*** (0.002) 0.000*** (0.000) -0.012*** (0.002) LEV 0.004 (0.005) 0.076 (0.055) 0.004 (0.005) 0.077 (0.055) ROA -0.022*** (0.006) -0.312*** (0.062) -0.022*** (0.006) -0.310*** (0.062) CFVOL 0.000*** (0.000) 0.000 (0.000) 0.000*** (0.000) 0.000 (0.000) BIG4AUD -0.008*** (0.003) -0.145*** (0.026) -0.008*** (0.003) 0.144*** (0.026) REM -0.001 (0.003) -0.001 (0.003) ABS_AEM -0.047 (0.130) -0.048 (0.130) Intercept 0.112*** (0.013) -0.236* (0.126) 0.113*** (0.013) -0.258** (0.128)

Industry fixed effects YES YES YES YES

Time fixed effects YES YES YES YES

N 9,332 9,332 9,332 9,332

R² 0.100 0.211 0.100 0.212

Adjusted R² 0.096 0.207 0.096 0.208

F-Value 16.91*** 45.63*** 16.26*** 43.61***

Notes: standard error in parentheses. Significance levels are as follows: * p < 0.1, ** p < 0.05, *** p < 0.01.

(23)

4.3. Additional analyses

In this section, I perform additional analyses to provide useful insights into the results of this study. This is done by e.g. using different proxies for the independent and dependent variables. All continuous variables are winsorized at the 1% level and the standard errors are clustered by firm and year.

4.3.1. Alternative measurement of the HHI

Given the criticism on the Herfindahl-Hirschman Index, I run the regression for an alternative measure of competition. I employ the eight-firm concentration ratio, which is the sum of the proportion of sales of the eight biggest firms in the industry classified by the Fama-French (49) classification (Karuna, 2007). Similar to the HHI, an increase in the concentration ratio is equal to a decrease in competition. Both the HHI and the concentration ratio are based on concentration and industry sales, however the concentration ratio does not attach as much weight to the total number of firms in the industry as the HHI does.

The alternative results in the first four columns of table 5 show little variation in coefficients, significance and explanatory power. None of the relevant coefficients differ much from the original results in table 4 and remain statistically insignificant. Therefore, there is no association and moderating effects between the main variables which means that the results are robust to an alternative measure of competition.

4.3.2. Alternative measurement of AEM

As an alternative to the Modified Jones Model, I investigate the relationship between competition and AEM using the Jones Model (Jones, 1991). These calculations are mainly similar to each other, but slightly differ as is shown in the following formula:

𝑇𝐴𝐶𝐶_𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1 + 𝛼2 Δ𝑅𝐸𝑉_𝑡 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸_𝑡 𝐴𝑡−1 + 𝜀𝑡 (9) The difference lies in the second parameter 𝛼2, where the change in revenue between t and t-1 is not corrected for the amount of receivables because Jones (1991) only assumes that revenues are non-discretionary. Similar to the main analysis, I use the absolute value of the abnormal accruals calculated by the Jones Model.

The last two columns of table 5 show the results of the alternative analysis of models 1 and 3. I find that competition is significantly related to abnormal accruals when I use the alternative measure of AEM (β=-0.163 and p=0.040). This would mean the first hypothesis can be accepted, where it proves the negative influence that competition has on the ethical behavior of management, where they employ a higher level of AEM. Furthermore, the last column of table 5 shows that political uncertainty still does not moderate the relationship between competition and AEM (β=0.000 and p=0.368). The results for model 1 are thus not robust for the alternative measure of AEM whereas model 3 is. A possible explanation lies in the fact that, even though widely used in research, discretionary accruals are a noisy proxy of EM (Jackson, 2018) which I will elaborate on in the discussion section.

(24)

Variable Model 1 C8 Model 2 C8 Model 3 C8 Model 4 C8 Model 1 Jones model Model 3 Jones model HHI -0.130 (0.082) -0.048 (0.591) -0.148* (0.088) 0.072 (0.587) -0.163** (0.079) -0.182** (0.086) PRISK 0.000 (0.000) 0.000** (0.000) 0.000 (0.000) HHI * PRISK 0.000 (0.000) -0.001 (0.001) 0.000 (0.000) INTERCEPT 0.116*** (0.008) -0.296*** (0.097) 0.116*** (0.009) -0.315*** (0.100) 0.117*** (0.009) 0.118*** (0.009)

Firm-level controls YES YES YES YES YES YES

Industry fixed effects YES YES YES YES YES YES

Time fixed effects YES YES YES YES YES YES

N 9,332 9,332 9,332 9,332 9,332 9,332

R² 0.100 0.211 0.101 0.211 0.097 0.097

Adjusted R² 0.096 0.207 0.096 0.207 0.092 0.092

F-Value 16.91*** 45.58*** 16.26*** 43.54*** 16.63*** 15.93***

Notes: standard error in parentheses. Significance levels are as follows: * p < 0.1, ** p < 0.05, *** p < 0.01.

(25)

4.3.3. Alternative measurement of REM

In the main empirical model, I use an aggregate sum of the three components of real earnings manipulation. However, in the use of the aggregated measure, I lose information on how each of the components contributes to the sum. Hence, I will re-perform the second and fourth model with the three separate components of REM (abnormal cash flow from operations, abnormal production costs, and abnormal discretionary expenses) as is done in previous research as well (e.g. Gunny, 2010; Visvanathan, 2008). Also, it is interesting to find out how competition affects the way that management uses real earnings management.

Table 6 displays the results of the alternative regressions. I document one significant relation at the 0.1 level between competition and one of the proxies of REM, namely the abnormal cash flows from operations (β=-0.051 and p=0.090). This indicates that an increase in competition leads to higher abnormal cash flows. In addition, the adjusted R² is 52.7 percent, which means more than half of the variance in the abnormal cash flows is explained by the independent variables in this study. Further, the other proxies of REM, abnormal production costs and abnormal discretionary expenses, show a positive coefficient (respectively β=0.064 and β=0.748) but no significant relation with competition (respectively p=0.775 and p=0.103)

The interaction term remains statistically insignificant in the last three columns of table 6, thus political risk does not moderate the relationship with any of the REM proxies. Subsequently, political risk is significantly related to abnormal discretionary expenses but only at the 0.1 level (p=0.076). To conclude, none of the results in table 6 are significant at the 0.05 level and thus not differ from the results in table 4, which means that my main results are robust to the separate proxies of REM.

4.3.4. Alternative measurement of political risk

Next, I run the regression again with a different measure of political risk. The measure created by Hassan et al. (2019) is very recent and is not widely used yet. Previous research regarding political risk often uses the economic policy uncertainty (EPU) index by Baker, Bloom, and Davis (2016) (e.g. Pastor & Veronesi, 2013; Gulen & Ion, 2015). More specifically, the index reflects the frequency of certain terms (i.e. “economic”, “uncertainty” and “regulation”) in the 10 largest U.S. newspapers, and it corroborates with big events such as elections, wars and terrorist attacks (Baker et al., 2016). The EPU is a monthly index, therefore I create a yearly average to match it with the annual data from other variables.

Concerning accrual-based earnings management, table 7 shows there is almost no difference between the main results in table 4. The interaction term coefficient does not differ and is not statistically significant (β=-0.001 and p=0.260). On the other hand, the results regarding real earnings manipulation are not robust to the use of the EPU instead of the measure of political risk by Hassan et al. (2019). It shows that the EPU successfully moderates the relationship between competition and REM (β=-0.009 and p=0.002) and that the EPU is significantly related to REM (β=0.001 and p=0.036). The coefficient of the interaction term does not show direction, hence I cannot conclude how it moderates the relationship. In conclusion, model 4 is not robust to the analysis using the EPU whereas model 3 is. This might be explained by the nature of the measures, where the EPU involves national political uncertainty and does not consider firm-specific effects, whereas Hassan et al. (2019) do.

Competition as a Double-Edged Sword: How does it affect Earnings Management and can Political Risk explain the direction of the relationship?