Faculty of Economics and Business

(1)

Faculty of Economics and Business

It’s All Politics: A Quantitative Study on the Effects

of Political Induced Risk on Earnings Management.

Evidence from the United States

Master’s thesis June 2019 Word count: 14,226

ABSTRACT Concerns about the effects of political induced risk have nowadays become more prominent

in the western world. This thesis adds to the finance and accounting literature, by examining the effects of political induced risk on earnings management practices. I draw on agency theory and impression management literatures, to argue that executives possess increased opportunity and motivation to engage in real activities manipulation during times of greater political induced risk. Although I expect executives to increasingly engage in (overall) earnings management practices during times of greater political induced risk, I take into account a potential swap from accrual-based earnings manipulation practices, to real activities manipulation practices, due to increasingly cautious executive behaviour. I therefore state the relationship between political induced risk and accrual-based earnings management in a non-directional way, to investigate which effect is dominant. In addition, I hypothesize a negative moderating effect of leverage, on the relationship between political induced risk and both real activities manipulation and accrual-based manipulation. This because of increased monitoring activities performed by debtholders, which diminishes an executives’ opportunity to engage in earnings management practices. I employ a sample of 1,348 U.S. listed firms (representing 10,485 firm-year observations), and find no significant relationship between political induced risk and real activities manipulation. Furthermore, I am unable to identify a significant relationship between political induced risk and accrual-based manipulation. My results also indicate that leverage is not significantly moderating the relationship between political induced risk and both earnings management techniques. Although my results do not indicate significant relationships among political induced risk and earnings management, it still allows academics and practitioners to better understand executive behaviour, during times wherein the degree of political induced risk is fluctuating. JEL classification: D80, G18, G32, M40, M41, M48 Key words: Political risk, Earnings Management, Leverage

Student number: S 3522407

Name: M.L. (Mark) Neimeijer

Track: MSc Accountancy & Controlling

(2)

2

1. INTRODUCTION

While political news is dominating financial markets, concerns about the effects of political induced risk have become more prominent in the western world (Pástor & Veronesi, 2013). Recently, the world for instance witnessed the longest United States federal government shutdown in history. President Donald J. Trump agreed to reopen the federal government on the 25th of January 2019, 35 days after the shutdown had been initiated (Rocha, Wagner & Wills, 2019). This shutdown, which was caused by a conflict between Trump and U.S. congress, is expected to have notable consequences for the U.S. economy. The Congressional Budget Office (2019) lowered the U.S. GDP forecasts by 0.2% and indicated that 3 billion in GDP to be lost forever.

Besides the influence of governmental policies on the economy, individual firms also directly experience that these policies affect their businesses (Boddewyn, 1988). In the recent past, some firms have revealed that governmental policies influences their decision-making processes. Apple, for example, pledged to invest 350 billion dollars into the U.S. economy, after President Trump signed a new tax code in 2018, which led to a reduction in taxation on overseas earnings (Lima, 2018). Moreover, Ford Motor Company announced in 2017 that it had cancelled its plans to build a manufacturing plant in Mexico, after Trump had directly threatened its competitor (General Motors / GM) with the introduction of a ‘big border tax’ if GM would import cars from Mexico to the U.S. (Thielman, 2017). Instead, Ford Motor Company decided to invest an additional 700 million dollars into an existing manufacturing plant in Michigan. Examples like Apple and Ford Motor Company provide anecdotal evidence in support of the idea that governmental policies and actions are very important to firms and their behaviour. While these examples relate to policies whose effects are quite straightforward and easily interpretable for executives, the effects of many other (future) governmental policies are often not that predictable, resulting in political induced risk. Political induced risk, or political uncertainty, originates when executives and/or firms are not fully able to anticipate the future path of governmental actions or policies (Pástor & Veronesi, 2013; Riem, 2016). The executive or the firm is therefore unable to perfectly predict the consequences of these actions and policies.

A recent stream in the finance and accounting literature examines to what degree risk, induced by governmental policies, is influencing the macroeconomy, individual firms, and the behaviour of executives. Studies provide evidence that political induced risk is associated with changes in the macroeconomy (Snowberg, Wolfers & Zitzewitz, 2007), and aggregate economic activity (Baker, Bloom & Davis, 2016; Fernández-Villaverde, Guerrón-Quintana, Kuester & Rubio-Ramirez, 2015). Furthermore, studies indicate that political induced risk affects stock price volatility, stock return volatility (Baker et al., 2016; Hassan, Hollander, van Lent & Tahoun, 2017; Pástor & Veronesi, 2013; Voth, 2002), stock price correlation, equity risk premia (Pástor & Veronesi, 2013), and stock prices (Pástor & Veronesi, 2012). On top of that, studies also provide evidence that political induced risk influences the behaviour and decision-making process of executives. These studies indicate that political induced risk is influencing executive decision-making, in regards to

(4)

4

investments (Baker et al., 2016; Gulen & Ion, 2015; Handley & Limão, 2015; Hassan et al., 2017; Jens, 2017; Julio & Yook, 2012), employment (Baker et al., 2016; Hassan et al., 2017), foreign direct investments (FDIs) (Julio & Yook, 2016), and mergers and acquisitions (M&As) (Bonaime et al., 2018; Nguyen & Phan, 2017).

This thesis aims to extend the finance and accounting literature examining the influence of political induced risk on executive behaviour. It investigates how political induced risk influences the degree in which executives use their discretion, to engage in earnings management practices, while also examining the moderating effect of leverage on the relationship between political induced risk and earnings management. Earnings management (EM) occurs when an executive uses his or her discretion in order to influence the perception of stakeholders or try to influence contractual outcomes (Healy & Wahlen, 1999). Executives employ two major techniques when engaging in earnings management practices, namely the manipulation of accounting accruals (AEM), and/or the manipulation of real activities (REM). Executives engaging in AEM use their discretion and judgement over accounting accruals, within the possibilities provided by a country’s Generally Accepted Accounting Principles (GAAP) (Dechow, Sloan & Sweeney, 1995; Kothari, Mizik & Roychowdhury, 2016). REM entails executives taking suboptimal actions, by amending real operational decisions, like production and investments, in order to make changes to reported earnings (Zang, 2012).

I draw on agency theory and impression management literatures to hypothesize potential relations among political induced risk, EM and leverage. Agency theory asserts that due to the divergence in preferences and goals between executives, shareholders, and debtholders, executives may be motivated to engage in opportunistic behaviour, like EM (Jiraporn, Miller, Yoon & Kim, 2008). I argue it is likely that, during times of greater political induced risk, executives possess greater opportunities and increased motivation to engage in EM. Increased opportunity arises, because greater political induced risk makes it more difficult for executives to assess the consequences of governmental policies for the firm, but this difficulty is expected to be even greater for investors. Executives may possess increased motivation to engage in EM, to calm investors and/or to increase the value of their (out-of-the-money) equity options. In line with this reasoning, I hypothesize a positive relationship between political induced risk and REM. In contrast, I hypothesize a non-directional relationship between political induced risk and AEM, because prior research shows not only that executives become more careful and cautious during times of increased political induced risk, but also that they perceive REM to be harder to detect by external parties (Cohen, Dey & Lys, 2008). I therefore take into account that this cautious behaviour will damage the potential positive effect of political induced risk on AEM, and further strengthen the positive effect on REM. In addition, I hypothesize a negative moderating effect of leverage, on the relationship between political induced risk and both AEM and REM, because of increased monitoring activities by debtholders, which diminishes the opportunity for executives to behave opportunistically and engage in EM practices.

(5)

5

I employ a sample of 10,485 firm-year observations representing 1,348 unique firms, which are all listed in the United States between the years 2002-2016, to test my hypotheses. My firm-level measure of political induced risk is based on the study of Hassan et al. (2017). Consistent with prior studies, I use the ‘Modified Jones Model’ to estimate AEM (Dechow et al., 1995), while using the ‘Roychowdhury Model’ to estimate the degree of REM (Roychowdhury, 2006). The results of this thesis indicate that there is no significant relationship between the degree of political induced risk and the use of REM by executives. This thesis neither is able to report significant evidence concerning the relationship between political induced risk and the usage of AEM tactics. Moreover, the results of this thesis indicate that leverage is not significantly moderating the relationship between political induced risk and both AEM and REM. Lastly, the results show that several control variables are significantly related to the extent in which executives make use of AEM and REM tactics, while additional analysis also indicates that political induced risk is negatively related to the extent in which executives engage in income-decreasing AEM practices.

This thesis makes several contributions to the literature. First, it contributes to the literature concerning earnings management. In particular, it explores how and why political induced risk may affect the degree in which executives use their discretion, to engage in earnings management practices. Although the results do not indicate a significant relationship between political induced risk and either form of earnings management, it still allows academics and practitioners to better understand how and why executives engage in earnings management practices. Second, this thesis contributes to the emerging field in the accounting and finance literature, where academics investigate the effects of political induced risk on firms and executive behaviour, by studying the relationship between political induced risk and earnings management. This thesis differs from most prior studies (Baker et al., 2016) which focused on political induced risk, by employing a firm-specific proxy of political induced risk, instead of a macro-economic proxy. Lastly, this thesis contributes to the literature concerning the importance of the effects of capital structure. It is the first to explore the moderating effect of leverage on the relationship between political induced risk and earnings management. Prior studies (Jelinek, 2007; Franz, HassabElnaby & Lobo, 2014) already explored the effects of leverage on earnings management, yielding mixed results.

In the following section I present the theoretical background surrounding the main topics, and develop four hypotheses. Section three addresses the research design, while in section four I present my results. Section five discusses these results and presents conclusions, limitations and suggestions for future research.

2. THEORETICAL BACKGROUND & HYPOTHESIS

DEVELOPMENT

2.1 Agency Theory

Agency theory provides the starting point of my theoretical arguments to explain how political induced risk is associated with earnings management (EM) practices. Jensen and Meckling (1976)

(6)

6

portray ‘the firm’ as a “black box” (p. 306), which is operated in a certain manner that maximizes profits and present firm value. They illustrate the firm as a form of legal fiction, which can be seen as a nexus for contractual relationships between several parties (Jensen & Meckling, 1976). These parties, whom are all aiming to a maximization of their own utility, have divergent and sometimes conflicting objectives, motivations and risk-appetites.

The rise of publicly held firms, and the resulting increase in the dispersion of ownership in firms, provides the starting point to address issues that are a consequence of the separation of ownership from control. In a typical agency setting, the two main contracting parties are the shareholders (principal/owner) and the executives (agent/controlling party). The executives of the firm are acting as the agents for the shareholders, and are supposed to take actions and make decisions that result in a maximization of shareholder value. Agency theory emphasizes that the operations of firms are subject to uncertainty, which potentially leads to information asymmetry between the shareholders and the executives (Walker, 2013). The information advantage is in favour of the executive, because he or she has better access to detailed information, and has more knowledge of current and future operations, resources, risks, and earnings, compared to the shareholders (Armstrong, Guay & Weber, 2010; Ndofor, Wesley & Priem, 2015; Richardson, 2000).

The key problem that agency theory addresses is that the interests, preferences and goals of executives and shareholders may be conflicting. Specifically, the key premise of agency theory is that executives may have motives to engage in opportunistic behaviour, which might not be in the best interest of the shareholders (Jiraporn et al., 2008). The latter is called ‘the moral hazard problem’. Information asymmetry contributes to the moral hazard problem, because it makes the relationship between managerial actions and firms’ results more opaque, resulting in reduced ability of shareholders to adequately monitor the actions of the executives (Holmström, 1979; Jensen, 1994; Ndofor et al., 2015; Richardson, 2000; Walker, 2013). The accounting literature suggests that executives may employ a variety of opportunistic actions, which include engaging in a high-risk/reward strategy, shirking, reporting in a fraudulous manner (Ndofor et al., 2015), and - which is the focus of this thesis - earnings management. As eluded to in the next subsection, earnings management refers to the deliberate attempt, made by the executives, to manage the firms’ earnings (Chen, Luo, Tang & Tong, 2015).

Agency theory does address some mechanisms through which shareholders can mitigate the problem of moral hazard. These mechanisms include: (1) setting up an incentive compensation scheme, which aligns the interests of the executive with those of the firm, (2) incur (additional) monitoring costs (e.g. through introducing a board of outside directors) (Chen et al., 2015; Ndofor et al., 2015), and (3) pressure the executives to increase the quality of (financial) reporting (Armstrong et al., 2010; Fu, Kraft & Zhang, 2012).

(7)

7

2.2 Earnings Management

Although Earnings Management (EM) has received a considerable amount of attention in the accounting literature, to date academics are still unable to reach a consensus concerning a generally accepted definition of this phenomenon. While Schipper (1989) defines EM as “a purposeful

intervention in the external financial reporting process, with the intent of obtaining some private gain” (p. 92), Healy and Wahlen (1999) define EM as follows:

“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 reported accounting numbers” (p. 368)

The latter definition is the starting point of this thesis, as it incorporates the fact that managerial discretion can both interfere with the financial reporting process through accounting techniques, and by means of real-life operational decision-making (Roychowdhury, 2006).

2.2.1 Earnings management techniques

The two major EM techniques which executives can employ are (1) accrual-based earnings manipulation (AEM) and (2) real activities manipulation (REM) (Cohen et al., 2008; Cohen & Zarowin, 2010; Healy & Wahlen, 1999; Kothari et al., 2016; Schipper, 1989). Up to and including the early 2000s, research concerning EM was mainly focused on AEM practices (Cohen et al., 2008; El Diri, 2018; Roychowdhury, 2006; Walker, 2013). This technique entails the manipulation of accruals, by means of managerial discretion and judgement (Dechow et al., 1995; Kothari et al., 2016), mostly within the possibilities provided by a country’s Generally Accepted Accounting Principles (GAAP) (Chen et al., 2015; El Diri, 2018; Walker, 2013). This, for example, includes manipulation of the depreciation period and salvage value of assets, manipulating the estimation of post-employment benefits or the value of inventories (Healy & Wahlen, 1999; Walker, 2013). AEM is therefore a tactic employed by executives in order to misrepresent firm performance, but does (generally) not have an impact on real operations (Kothari et al., 2016). A consequence of engaging in AEM is the so called ‘accrual reversal’, which indicates that the effect of the manipulation is reversed in the subsequent years (Baber, Kang & Li, 2011), and therefore thus not have an impact on the current or future (free) cash flows of the firm (Roychowdhury, 2006), and (overall) firm value (Walker, 2013).

Since 2005, researchers began shifting their attention to REM (El Diri, 2018; Walker, 2013). REM “is a purposeful action to alter reported earnings in a particular direction, which is achieved

by changing the timing or structuring of an operation, investment, or financing transaction, and which has suboptimal business consequences” (Zang, 2012, p. 676). The two studies which

emphasized the importance of REM were those of Graham, Harvey & Rajgopal (2005) and Roychowdhury (2006). The former concluded, from surveying around 400 executives, that most managers of both public and private firms in the U.S. preferred to manipulate real actions, rather than using accounting choices, in order to manage earnings. The latter found evidence that

(8)

8

executives manipulate real activities, amongst others by giving customers discounts in order to prevent disclosing a loss in their financial statements. A likely reason for the shift in attention from AEM to REM is the introduction of the Sarbanes-Oxley Act (SOx) in July 2002 (Kothari et al., 2016). As a result of the introduction of SOx, the role of the board of directors in supervising executive directors has been strengthened, while at the same time regulatory scrutiny by the SEC increased (Commerford, Hatfield & Houston, 2018), both of which make it harder for executives to resort to AEM practices. Although conclusive evidence concerning the ability of the market to identify the manipulation of real activities is lacking (Kothari et al., 2016), it is likely that executives believe that the manipulation of real activities is harder to detect, compared to AEM (Cohen et al., 2008; Franz et al., 2014), even though its potentially costlier for the firm (Kothari et al, 2016; Roychowdhury, 2006). Executives have a variety of ways through which they can employ REM practices. This thesis focuses on three ways which are most commonly studied by academics, and include: (1) the manipulation of sales numbers trough lenient credit terms or price discounts, (2) manipulating discretionary expenditures, and (3) manipulating production (Cohen et al., 2008; Cohen & Zarowin, 2010; Roychowdhury, 2006).

The REM tactics, in contrast to the AEM tactics, do in fact influence the current and future (free) cash flows of the firm, thereby potentially harming the firm and its value (Roychowdhury, 2006). For example, diminishing R&D expenditures leads to higher cash flows on the short term, but may have a negative effect on future cash flows, because the firm does not innovate its products and processes as well as before. Moreover, overproduction leads to lower fixed costs per unit on the short term, but can result in excessive inventory holding costs in the following periods.

2.2.2 Earnings management motives

Aside from having the opportunity to engage in EM, executives also require certain motives to actually do so. This subsection builds further on agency theory and uses impression management as a starting point. This thesis perceives EM to be opportunistic by nature, which executives employ to pursue their own interests (Jelinek, 2007; Jiraporn et al., 2008)1. Academics have well studied the opportunistic motives for executives to engage in EM in the past, including: contracting motives, capital market motives and external motives (Healy & Wahlen, 1999; El Diri, 2018; Walker, 2013). I only address the motives that are expected to be connected to political induced risk2.

1_{Some researchers have studied the beneficial nature of EM (Healy & Palepu, 1993; Watts & Zimmerman, 1990), and}

others even found evidence of this beneficial nature (Subramanyam, 1996; Jiraporn et al., 2008).

2_{Other motives for EM which are not included in this thesis are: (1) stock-financed acquisitions (Botsari & Meeks,}

2008; Efendi, Srivastava & Swanson, 2007; Erickson & Wang 1999), (2) SEOs (Cohen & Zarowin, 2010; Teoh, Welch & Wong, 1998b), (3) IPOs (Morsfield & Tan, 2006; Teoh, Welch & Wong, 1998a ) (4) employment horizon (Dechow & Sloan, 1991), (5) competitive environment (El Diri, 2018), (6) taxes (Kumar & Visvanathan, 2003), (7) management buy-outs (Perry & Williams, 1994; Wu, 1997), (8) firm reputation (Raman & Shahrur, 2008), and (9) debt covenant violation (Watts & Zimmerman, 1990).

(9)

9

EM is often seen as a tactic of impression management, to deliberately influence the perception of stakeholders (board, shareholders, analysts), and their reaction to certain events (Chen et al., 2015). It provides executives with the ability to ‘manage the perception’ that stakeholders have of them, through manipulating firm performance (Healy & Wahlen, 1999). In times when a firm is confronted with a more challenging external environment, its shareholders might have increased concerns. Shareholders may demonstrate these concerns by marketing their shares, leading to enhanced pressure on stock performance (Chen et al., 2015). This might motivate executives to engage in EM practices, in order to enhance stock performance and restore the confidence of shareholders in the firm and themselves. Executives may also experience increased uncertainty about their position in more challenging times, which might induce them to manipulate earnings in order to strengthen their position (DeAngelo, 1986). Another likely reason for executives to engage in EM is to balance their yearly and quarterly earnings (Walker, 2013). De Jong, Mertens, van der Poel & van Dijk (2014) provide evidence that both executives and analysts (as a proxy for investors) recognize that smoothing a firms’ earnings path has positive consequences for the firm, by reducing the perceived riskiness3.

The remuneration package of an executive often consists of a mix of compensation and other benefits. Some of the pay within the package is fixed (e.g. a base salary), while other is more variable of nature (e.g. bonuses), or more dependent on other contingencies (e.g. stock (options)). Especially the last two forms of compensation are expected to be motives for executives to engage in EM practices. For example, executives may be motivated to (1) boost performance to reach a (bonus) contract-threshold, and defer earnings after a contract-cap is reached (Cheng & Warfield, 2005; Feng, Gramlich & Gupta, 2009; Healy, 1985), or (2) manage earnings to influence the value of stocks or stock-options which they possess (in the near future) (Bergstresser & Philippon, 2006; Cheng & Warfield, 2005; Coles, Hertzel & Kalpathy, 2006; Kuang, 2008; Zhang, Bartol, Smith, Pfarrer & Khanin, 2008).

In the following section, I will present arguments to explain how political induced risk can affect managers’ tendencies to engage in earnings management practices.

2.3 Political Induced Risk

Governments continuously make decisions regarding their policies and future directions. The general aim of policy formulation is to improve the welfare of its citizens, and it is motivated by both financial and non-financial reasons. Political decisions, such as monetary, regulatory or fiscal changes, have both benefits and costs associated with them, and do have an influence on economic activity. Political induced risk, or political uncertainty, arises when executives (or more generally: individuals) and firms cannot fully anticipate the future path of governmental actions and policies, and therefore are unable to perfectly predict the consequences of these actions and policies (Pástor

3_{Although analysts (as a proxy for investors) prefer a smooth earnings path, they dislike it when executives}

intentionally smooth their earnings (De Jong et al., 2014). This research also indicates that both executives and analysts are willing to sacrifice economic value by smoothening earnings, but that executives are willing to go much further.

(10)

10

& Veronesi, 2013; Riem, 2016). This thesis makes use of a firm-level measure, developed by Hassan et al. (2017), which serves as a proxy for political induced risk. Researchers have already shown that political induced risk is associated with changes in the macroeconomy. For example, Snowberg et al. (2007) provide evidence that the probability of a certain party (Democrats versus Republicans) winning the U.S. elections is associated with changes in the financial markets, while increased political induced risk also affects aggregate economic activity (Baker et al., 2016; Fernández-Villaverde et al., 2015).

In the last decade, the influence of political induced risk on firms increasingly receives attention of academics. Studies in this field provide evidence that political induced risk leads to increased stock return volatility and stock price volatility (Baker et al., 2016; Hassan et al., 2017; Pástor & Veronesi, 2013; Voth, 2002), while it also results in an increase in stock price correlation and equity risk premia (Pástor & Veronesi, 2013). Political signals (e.g. policy changes) are more powerful in times when political induced risk is high, thus increasing its effects. Policy changes, on general, affect all firms, and are non-diversifiable for investors and shareholders, thus increasing stock correlation and equity risk premia (Pástor & Veronesi, 2013)4. This evidence is an indication that shareholders and investors have increased concerns during times of greater political induced risk, which does affect their behaviour. These results are also in line with the research of Pástor & Veronesi (2012), which noticed that stock prices decline in times when political induced risk is higher, due to an increase in the discount rate5.

Concerns about future governmental policy also influences the behaviour of executives and firms in a variety of ways. Studies provide evidence that firms cut their investments (Baker et al., 2016; Gulen & Ion, 2015; Handley & Limão, 2015; Hassan et al., 2017; Jens, 2017; Julio & Yook, 2012) and take on less employees (Baker et al., 2016; Hassan et al., 2017) in times of greater political induced risk. This evidence corroborates with the standard models in literature, which predict an increase in any type of uncertainty results in executives preferring to ‘wait and see’, instead of making bolt investment and employment decisions, under challenging circumstances (Bloom, Bond & van Reenen, 2007). More simply said, executives become more careful and cautious in times when political induced risk is high, which affects their investment and employment decisions. Under such challenging circumstances, firms are also less likely to engage in FDI’s (Julio & Yook, 2016) and M&A’s (Bonaime, Gulen & Ion, 2018; Nguyen & Phan, 2017), which are also signs that are in line with the evidence indicating that executives become

4_{Hassan et al. (2017) find contrasting evidence to this viewpoint, indicating that political induced risk is a driver of}

idiosyncratic risk instead of systematic risk, and thus is diversifiable for shareholders and investors.

5_{Not only does political induced risk influence a firm’s stock, it also affects the price of equity options (Kelly, Pástor}

(11)

11

increasingly cautious, and prefer to take less drastic decisions in times of greater political induced risk6.

2.3.1 Political induced risk and earnings management

Executives whom are willing to engage in EM practices require the simultaneous presence of two circumstances, these are: (1) the opportunity to engage in EM, and (2) the motivation to engage in EM. The former is key here, given that even the most motivated executive is not capable of successfully engaging in EM, when the opportunity to do so is not present (Ndofor et al., 2015). Agency theory explains that information asymmetry plays a key role in providing executives with the opportunity to engage in EM, by means of decreasing the ability of shareholders to adequately monitor and identify their opportunistic behaviour (Richardson, 2000). I expect that increased levels of political induced risk provides executives with more opportunity to engage in EM. Increased levels of political induced risk make it more difficult for executives to assess the consequences of governmental policy for the firm, but this difficulty is expected to be even greater for investors, thus enlarging the asymmetry of information between investors, shareholders and executives. In support of this view is prior research showing that the discretion in managerial decision-making increases under increased levels of uncertainty, thereby making executives’ decisions more complex and less comprehensible for shareholders (Demsetz & Lehn, 1985). This view is also supported by the studies of Pástor & Veronesi (2013), and Hughes, Liu & Liu (2007), with the former providing evidence that risk premia increase during times when political induced risk is high, while the latter discovered that this increase is the result of increased levels of information asymmetry. Moreover, political induced risk may also distract investors, by moving their focus from monitoring activities, to focusing more on assessing the potential effects of regular politics7_.

The separation of ownership and control within listed firms contributes to the goal divergence between executives and shareholders, leading to the problem of moral hazard. It exists because the interests of executives are not perfectly aligned to those of the shareholders. Shareholders, especially within U.S. firms, use compensation schemes to counter or discourage this opportunistic behaviour (Jansen, Merchant & Van der Stede, 2009; Zhang et al., 2008), and tie executive pay to measures of performance, stocks or stock-options. In contrast to this view, research is showing that incentive and equity compensation are not always able to perfectly align the interests of executives with those of the shareholders, and may even be a motivator to engage in EM themselves (Healy & Wahlen, 1999). The latter view is supported by research of Zhang et al. (2008) showing that

6_{Executives do not only react passively, but also actively try to mitigate political induced risk. There is evidence that}

firms which are exposed to higher amounts of political induced risk tend to donate more funds to politicians, compared to firms which are exposed to lower levels of political induced risk (Akey & Lewellen, 2017; Hassan et al., 2017).

7_{It is also possible that shareholders may become more protective during times when poltical induced risk is high,}

thereby increasing monitoring activities and/or rely more on mechanisms of corporate governance (Boubakri, Guedhami & Mishra, 2010).

(12)

12

executives are more likely to engage in EM when their stock-options are ‘out-of-the-money’8 (stock-options that have a grant price above the market price). Given that during times when political induced risk is greater, stock prices become more volatile and tend to go down (Pástor & Veronesi, 2012; Pástor & Veronesi, 2013), I expect that increased political induced risk does motivate executives to engage in EM practices. Other evidence in support of this viewpoint, by drawing on impression management, is showing that executives are aware that shareholders and investors become more concerned during times when political induced risk is greater, due to disruptive or sudden changes in policy (Chen et al., 2015). Because executives are aware of the potential negative consequences of these concerns (e.g. selling shares), they might attempt to mitigate these concerns and simultaneously try to restore the confidence of the investors, through employing EM practices (Chen et al., 2015).

In sum, I expect that executives are more likely to engage in EM practices when political induced risk is greater, because under these circumstances, executives possess more opportunity and have increased motivation to engage in opportunistic behaviour. When looking at the techniques through which executives employ EM practices, I expect executives to employ more REM during times when political induced risk is greater. This is inferred from studies which indicate that an executive becomes more reserved or cautious in making decisions, regarding investments and employment. This is a clear indication of the potential positive effects of political induced risk on the use of REM, for example through diminishing R&D spending. Therefore, I posit the first hypothesis, regarding firm-level political induced risk and REM, in a positive direction.

H1: “The degree of firm-level political induced risk is positively associated with the use of

real activities manipulation practices.”

In contrast with the previous hypothesis, I state the relationship between firm-level political induced risk and AEM in a non-directional way. On the one hand, executives are still expected to have increased opportunity to engage in opportunistic behaviour in times when political induced risk is high, due to (1) lower scrutiny by investors (through attention shifting), and (2) increased information asymmetry, while also possessing increased motivation to do so. However, I take into account that executives behave more cautious in times when political induced risk is higher. This line of reasoning is supported by research (Baker et al., 2016; Bonaime et al., 2018; Hassan et al., 2017; Julio & Yook, 2016) showing that executives tend to make decisions more carefully and more cautious. In addition to the latter, it is expected that executives see REM as the more ‘safe

8_{Zhang et al. (2008) draw on both agency theory and prospect theory to examine the influence of various stock-based}

incentives on EM. Prospect theory emphasizes that executives are ‘loss-averse’ and that they won’t pursue gains, if these are connected to (large) potential losses. Conversely, when already at a loss, prospect theory expects that executives become more proactive, and are even capable to sacrifice the interests of the shareholders, in order to make something of the situation.

(13)

13

way’ to manipulate earnings, compared to AEM. This view is supported by the studies of Cohen et al. (2008) and Franz et al. (2014), which indicate that, although conclusive evidence is still lacking, executives have the perception that REM is more difficult for shareholders to detect, compared to AEM. In line with this reasoning, I expect that, even though executives possess increased opportunity and motivation to engage in opportunistic behaviour, they might use this increased opportunity and motivation to engage more in REM, rather than in AEM. I therefore state hypothesis 2 in a non-directional way, in order to examine which of the two effects is dominant, and leads to either a positive or a negative relationship between firm-level political induced risk and AEM.

H2: “The degree of firm-level political induced risk is associated with the use of

accrual-based earnings manipulation practices.”

2.4 The Monitoring Role of Debtholders

According to the literature concerning corporate governance, there are a variety of mechanisms through which shareholders are able to mitigate the agency problem. Firms which have adequately designed their governance systems are expected to be better able to effectively monitor their executives, and are therefore better able to ensure that the behaviour of these executives is in line with the best interests of the shareholders (Chen et al., 2015). These mechanisms can arise from both the design of the internal organization, and from the external environment. Internal mechanisms include installing an independent board of directors, or setting up an incentive compensation scheme (Chen et al., 2015), while external mechanisms may entail the monitoring by debtholders, or financial analysts. In this thesis, I solely focus on one external governance mechanism, namely monitoring by debtholders.

Debt financing became a more important issue for firms, since Modigliani and Miller (1963) argued that firm value could be enhanced as debt increased. When firms substitute equity with debt, agency costs of debt rise. Although the interests of debtholders and shareholders are generally aligned, they also can diverge due to different views or different stakes in performance (Asbaugh-Skaife, Collins & Lafond, 2006). In this case, agency costs of debt may arise when executives act on behalf of themselves or the shareholders, and take actions which do not lead to a maximization of debtholder utility, like asset substitution, claim dilution and debt overhang (Christensen, Nikolaev & Wittenberg-Moerman, 2016). Given that debtholders are aware of the existence of these agency costs, they are incentivized to take actions which oppose these (Armstrong et al., 2010). In doing so, debtholders engage in monitoring activities to assess the actions of the

(14)

14

executive, and to protect themselves from opportunistic behaviour of the executive (Asbaugh-Skaife et al., 2006) 9.

For debtholders of highly leveraged firms it may be more efficient to incur additional monitoring costs, in order to assess whether or not the quality of the borrower is still sufficient to eventually repay its debt (Rodríguez-Pérez & van Hemmen, 2010). This controlling effect of debtholders is consistent with the ‘control hypothesis’, developed by Jensen (1986). This control hypothesis argues that when firms are highly leveraged, it constrains the opportunity of executives to engage in opportunistic behaviour. This because executives of highly leveraged firms are subject to increased scrutiny of lenders (Chen et al., 2015), and because the debt repayment restricts the amount of resources, which executives might use ‘opportunistically’ (Jelinek, 2007). Other studies found evidence that is consistent with this line of reasoning, indicating that leverage limits the degree in which executives behave opportunistically (Denis & Denis, 1993; Gupta & Rosenthal, 1991; Kaplan & Stein, 1993). Moreover, debtholders are often financial institutions, which possess the resources and the ability to effectively monitor the executive and discourage them from engaging in opportunistic behaviour.

In line with the previous reasoning, and given the opportunistic nature of EM, I expect that leverage moderates the relationship between firm-level political induced risk and EM, in such a way that increased leverage will result in lower levels of EM. When looking at both techniques of EM, I expect that the opportunity for executives to engage in AEM will diminish when leverage is higher, due to increased monitoring activities by debtholders. This view is supported by prior research, which indicate that increased levels of leverage are negatively associated with the use of AEM practices (Jelinek, 2007; Wasimullah, Toor & Abbas, 2010). I therefore expect a negative moderating effect of leverage, and thus state the third hypothesis as follows:

H3: “The degree of leverage is moderating the relationship between firm-level political

induced risk and accrual-based earnings manipulation practices in such a way, that increased leverage will result in a lower level of accrual-based earnings manipulation

practices.”

Although I expect an overall decrease in EM practices due to the moderating effect of leverage, I do take into account the possibility that executives will further shift their attention from using AEM practices to REM practices. This is because executives believe that REM is less likely to draw the attention from stakeholders, and so they might have the perception that they will succeed

9_{Debt covenants are also mechanisms which are widely employed by debtholders, in order to prevent executives from}

behaving opportunistically. There is some evidence, in line with the debt covenants hypothesis, that debt covenant violation is a motive for executives to engage in EM practices (Defond & Jiambalvo, 1994; Dichev & Skinner, 2002; Franz et al., 2014; Lazzem & Jilani, 2018; Sweeney, 1994), but these debt covenants are outside the scope of this thesis.

(15)

15

in pulling off REM, even though monitoring activities amplify. Although this swap from AEM to REM is a possibility, I do not expect it to completely offset the negative influence of leverage. I expect that this negative moderating effect of leverage is greater, because debtholders, like banks or financial institutions, are presumed to have greater access to relevant firm information, and possess the ability to adequately process this information (Armstrong et al., 2010). Moreover, large debtholders often demand additional proprietary (private) information from the borrowing firm, including forecasts and budgets, in order to assess whether or not the firm is able to repay its debt. I therefore expect that leverage sufficiently diminishes information asymmetry, and thus limits the executives’ opportunity to behave opportunistically. In support of the latter is the research of Zamri, Rahman & Isa (2013) showing that leverage limits REM practices, by restricting the executive’s ability to engage in real earnings manipulation. I therefore state the fourth hypothesis as follows:

H4: “The degree of leverage is moderating the relationship between firm-level political

induced risk and real activities manipulation practices in such a way, that increased leverage will result in a lower level of real activities manipulation practices.”

3. RESEARCH DESIGN

3.1 Sample and Data Selection

This thesis adopts a firm-level measure of political induced risk, based on Hassan et al. (2017). Their dataset provides the starting point of my research. As result of unavailability of data concerning several variables, the final sample consists of 1,348 firms, which represents approximately 14% of the total sample of Hassan et al. (2017). My sample consists of firms which are listed in the United States between the years 2002 and 2016. I exclude firms which are active in highly regulated industries, financial institutions and banks (i.e. firms with SIC-codes between 4400-5000 and 6000-6500), which is in line with the studies of Roychowdhury (2006) and Zang (2012). Financial and firm data, which are used as an input for measuring AEM, REM and several firm-specific control variables, are derived from the COMPUSTAT database. The MSCI database is employed to obtain board-specific control variables, while the EXECUCOMP database is employed to obtain data on CEO-specifics10.

3.2 Measuring Earnings Management

3.2.1 Accrual-based earnings manipulation (AEM)

Academics developed several models in order to estimate the degree in which executives engage in AEM (DeAngelo, 1986; Dechow et al., 1995; Healy, 1985; Jones, 1991). These models are mostly used to obtain a quantifiable measure of abnormal discretionary accruals (DACC), which is used as a proxy for measuring AEM (El Diri, 2018). The most prominently used models

(16)

16

for estimating DACC are the ‘Jones Model’ and the ‘Modified Jones Model’ (El Diri, 2018; Islam, Ali & Ahmad, 2011). The latter was introduced by Dechow et al. (1995), as a reaction to the former model of Jones (1991). The model of Dechow et al. (1995) adjusts the firms’ changes in revenue for the changes in receivables, thereby reducing the measurement error in the case when managerial discretion is employed over credit sales. I follow several academics (Cohen et al., 2008; Cohen & Zarowin, 2010; Jiraporn et al., 2008; Zang, 2012), and employ the ‘Modified Jones Model’ in order to estimate the DACC. The DACC, which serve as a proxy for AEM11, are estimated by year and by industry, in order to take time and industry variation into account. Moreover, I require a minimum of 15 industry-year observations in order to estimate the DACC, which is in line with the study of Zang (2012)12.

_𝑖𝑡

)

𝐴

_{𝑖,𝑡−1}

+ 𝛼

3

𝑃𝑃𝐸

_𝑖𝑡

𝐴

_{𝑖,𝑡−1}

+ 𝜀

𝑖𝑡 (2) In formula 2, A represents the total lagged assets (the assets in year t-1), ∆Sales represents the change in revenues. ∆R represents the change in receivables. The latter represents the previous mentioned modification that Dechow et al. (1995) made, relative to the original ‘Jones Model’, in order to correct the change in revenue with the change in receivables. Moreover, PPE represents the gross value of the property, plant and equipment. Lastly, the residuals of the regression (

ε

it)

11_{AEM is estimated by including as many observations as possible (i.e. U.S. listed firms which are not part of the}

sample of Hassan et al. (2017) were also included).

12_{I follow prior studies (Cohen et al., 2008; Cohen & Zarowin, 2010) and identify industries by means of a two-digit}

Standard Industrial Classification (SIC) code (i.e. companies are grouped into industries based on the first two digits of the SIC code).

13_{Formula 1 is called ‘the balance sheet approach’. The total accruals (TACC) can also be calculated by using the}

cash-flow-statement-approach (Hribar & Collins, 2002): Earnings before extraordinary items and discontinued operations (EBXI) -/- operating cashflows (CFO). Testing AEM by using the cash-flow-statement-approach, instead of ‘the balance sheet approach’, did not lead to any different results (not reported) concerning Prisk and Prisk*LEV.

(17)

17

represent the firm-year DACC, and these are used as the proxy for AEM. I use the absolute values of AEM, because this thesis focuses only on the degree of earnings management, rather than the direction in which the earnings are managed (upward versus downward).

3.2.2 Real activities manipulation (REM)

Early studies investigating REM had a tendency to focus only on single economic decisions, made by executives. These studies often focused solely on the manipulation of R&D expenses (Roychowdhury, 2006; Walker, 2013). Later studies developed models that included more proxies for measuring the degree in which executives engage in REM (Gunny, 2010; Roychowdhury, 2006.). This thesis follows prior research (Cohen et al., 2008; Cohen & Zarowin, 2010; Zang, 2012), and develops proxies for measuring REM in line with the model of Roychowdhury (2006). The ‘Roychowdhury Model’ is one of the most commonly used models in order to measure REM (El Diri, 2018), and prior researchers (Cohen et al., 2008; Cohen & Zarowin, 2010; Gunny, 2005; Zang, 2012) provide evidence that the measures in this model do indeed capture the manipulation of real activities, by executives. These academics thus show that the measures in the ‘Roychowdhury Model’ possess construct validity.

This thesis focuses on three ways in which executives may employ REM, including (1) the manipulation of sales, (2) diminishing discretionary expenses, and (3) overproduction (Roychowdhury, 2006). Firstly, an executive may attempt to increase sales, by offering customers more lenient credit terms or excessive price discounts. This results in increased earnings and turnover in the current period (given that the margin is positive), but may harm the firms’ cash flow, due to the lower profit-margin (Cohen & Zarowin, 2010). Secondly, discretionary expenses, like advertising expenses, R&D expenses, and SG&A expenses, do often not generate income immediately. These are however often expensed in the same period, as they did occur. Executives therefore may attempt to increase earnings, by reducing the discretionary expenses, while at the same time risking sacrificing future cash flows (Roychowdhury, 2006). Lastly, an executive may be inclined to overproduce, in order to lower the costs of goods sold (COGS). Overproduction lowers the fixed costs per product (assuming that the marginal costs do not offset this effect) and thus increases the reported margins in the current period. Overproduction affects the firms’ cash flow (in comparison to its sales), through increased inventory holding costs and increased production costs (Roychowdhury, 2006).

I employ three proxies, in order to measure the three ways through which executives engage in REM. These proxies being: (1) the abnormal level of cash flow from operations (abn_CFO), (2) the abnormal level of discretionary expenses (abn_DisEx), and (3) the abnormal level of production costs (abn_Prod). All three proxies, which serve as the measure for REM14_{, are estimated by year}

and by industry, in order to take time and industry variation into account. Moreover, I require a

14_{REM is estimated by including as many observations as possible (i.e. U.S. listed firms which are not part of the}

(18)

18

minimum of 15 industry-year observations in order to estimate the three proxies, which is in line with the studies of Roychowdhury (2006) and Zang (2012)15.

_{𝑖,𝑡−1}

+ 𝜀

𝑖𝑡 (4) In this formula, Salest-1 represents the lagged revenue of the firm (the revenue in year t-1). Roychowdhury (2006) does not express DisEx as the linear function of the sales in the current year, because if done so, the regression would exhibit extraordinarily low levels of residuals, even though they might not diminish the actual discretionary expenses. The residuals from the regression (

ε

it)

represent the abn_DisEx, and this is used as a proxy for diminishing discretionary expenses (1 of

16_{The minimal requirement for DisEx to be calculated is the availability of SG&A expenses. If only the data regarding}

(19)

19

3 ways through which executives engage in REM). In order to reflect this proxy’s upward earnings management, it is multiplied by -1 (Cohen & Zarowin, 2010; El Diri, 2018; Zang, 2012).

The abn_Prod represents the difference between actual production costs and normal production costs. In order to derive to this proxy, it is first necessary to calculate the production costs (Prod) of a firm (i) during a certain year (t), this is done by executing the following formula:

𝑃𝑟𝑜𝑑

_𝑖𝑡

= 𝐶𝑂𝐺𝑆

_𝑖𝑡

− ∆𝐼𝑁𝑉

ε

it) represent the abn_Prod, and this is used as a proxy for

overproduction (1 of 3 ways through which executives engage in REM).

I use the absolute values of abn_CFO, abn_DisEx, and abn_Prod, because this thesis focuses only on the degree of earnings management, rather than the direction in which the earnings are managed (upward versus downward). I rely on prior studies (Cohen et al., 2008; Cohen & Zarowin, 2010; Franz et al., 2014) and compute a single measure of REM, because the individual proxies could also affect each other. This single measure of REM represents the sum of standardized abn_CFO,) standardized abn_DisEx, and standardized abn_Prod17_.

3.3 Measuring Firm-Level Political Induced Risk

In the absence of a direct measure, as is mentioned before, I employ a firm-level proxy of political induced risk, which is constructed by Hassan et al. (2017). In their study, the authors used computational linguistic tools to analyse the transcripts of 175,797 quarterly earnings conference calls, which they collected themselves18_{. They constructed a measure of firm-level political induced}

risk, that represents the percentage of the total quarterly earnings conference call, that is focused

17_{Standardized values of these variables are calculated as follows: [Variable – Mean (variable)] / standard deviation}

(variable) (Chi, Lisic & Pevzner, 2011).

18_{Baker et al. (2016) did also apply textual-based analysis, in order to measure political induced risk. These authors}

developed the EPU index, which is widely used by other researchers. Hassan et al. (2017) indicate that the EPU index, and other models (e.g. the one mentioned in Pástor & Veronesi, 2012), do not adequately describe the impact of political induced risk on firms. Therefore, I employ their firm-level measure of political induced risk.

(20)

20

on topics related to political induced risk. The authors employ the following formula to construct their measure, which is called ‘Prisk’.

𝑃𝑟𝑖𝑠𝑘

_𝑖𝑡

=

Σ

_𝑏𝐵𝑖𝑡

_{(1 [𝑏 ∈}

P

N

] ∗ 1[|b − r| < 10] ∗

𝑓

_𝑏,𝑃

𝐵

_𝑝

𝐵

_𝑖𝑡 (7) Firstly, the authors constructed two libraries, one with political texts, and another with non-political texts. Both libraries are filled with ‘Bigrams’, these are two-word combinations, which relate to political or non-political texts. The political ‘Bigrams’ are derived from political sections of newspapers and political science textbooks, while the non-political ‘Bigrams’ are derived from transcripts of non-political speeches, accounting textbooks and non-political sections of newspapers. In a similar vein, the authors identified terms which are associated with ‘risk’ or ‘uncertainty’, including synonyms of these. Hereafter, the authors performed a textual analysis of the transcripts, and developed a list with all bigrams that were present among them. Lastly, the authors counted the frequency of political bigrams, which were in close proximity (within a 10-word radius) to terms like ‘risk’, ‘uncertainty’ or other synonyms, and they divided this frequency by the total quantity of all bigrams, thereby obtaining their measure of firm-level political induced risk. Hassan et al. (2017) provide evidence that their measure of political induced risk does possess construct validity19. I transform the quarterly data of firm-level political induced risk into yearly data, due to availability issues of several control variables. In doing so, I identify the month in which a firm ends its fiscal year, and link that to its corresponding quarter. The logarithm of the firm-level political induced risk data of that quarter, serve as a proxy for the whole year20_.

3.4 Moderating & Control Variables

The moderating variable of this thesis (leverage) is calculated by dividing a firms’ long-term debt by its book-value of equity (LEV) (Jelinek, 2007; Lazzem & Jilani, 2018; Wasimullah et al. 2010)21. I use the book-value of equity, rather than the market-value, because the latter may be affected by share prices, and therefore is less able to adequately reflect the actual level of a firm’s reliance on debt. Whilst now all primary variables are discussed, I do also take into consideration

19_{The authors (Hassan et al., 2017) performed additional analyses, case studies and falsification tests to validate their}

measure. They find that their formula does correctly identify transcripts that have increased focus on topics regarding political induced risk. Moreover, their measure (1) varies intuitively across sectors that are highly influenced by government policy and (2) in times of high aggregate political induced risk (e.g. around federal elections), while their measure is also correlated with (3) stock market volatility, (4) firm behaviour and (5) the EPU index.

20_{In order to address the observations of Prisk with the value zero (0), I use the following formula: log(1+Prisk).} 21_{This thesis follows prior studies (Jazzem & Jilani, 2018) and drops negative LEV ratios, because these could have}

disproportionally impacted the results. Including negative LEV ratios would not have changed the reported conclusions in regards to the hypotheses 3 and 4.

(21)

21

the potential effects of other variables on AEM and REM, including (1) firm-specifics, (2) governance mechanisms, and (3) CEO characteristics.

Firstly, I control for the effect of firm size (SIZE) (Roychowdhury, 2006), it is calculated by the logarithm of total assets (Hooghiemstra, Hermes, Oxelheim & Randøy, 2019). On the one hand, researchers expect that executives of larger firms employ more EM, because of the increased political costs that they are facing (Richardson, 2000; Zmijewski & Hagerman, 1981). In contrast to this, other researchers expect that executives of larger firms are less able to exercise their managerial discretion, because of increased scrutiny by analysts and other stakeholders (Han, Kang, Salter & Yoo, 2010; Sánchez-Ballesta & García-Meca, 2007). Secondly, I control for the effect of future firm growth (GROWTH), measured by dividing the market value of the firm by the book value of its assets (Market-to-Book ratio / MTB) (Jelinek, 2007; Roychowdhury, 2006). Executives of firms with large MTBs are expected to be under great pressure, and are therefore motivated to engage more in AEM and REM practices, to meet earnings thresholds (Roychowdhury, 2006), or inflate earnings figures (Chih, Shen & Kang, 2008). Thirdly, I control for the effect of firm losses (LOSS), by including a dummy variable which indicates 1 if the firm experienced a loss in that year, and zero otherwise (Hooghiemstra et al., 2019). Executives of firms which experiences losses have several motives to engage in EM, like inflating earnings to keep their position within the firm or avoid debt covenant violation (DeAngelo, DeAngelo & Skinner, 1994). Executives may also take a ‘big bath’ in the year that the firm incurs losses, by managing the earnings further downwards, in order to increase the reported earnings in the following years (Jordan & Clark, 2015; Kirschenheiter & Melumad, 2002; Walsh, Craig & Clarke, 1991).

In addition to firm-specific control variables, I also include four control variables related to corporate governance. Firstly, I incorporate board effectiveness (BOARD_SIZE) in my model, which is measured by the logarithm of the number of board members. Larger boards are expected to be more effective, because they possess more experience and expertise. Ghosh, Marra & Moon (2010) and Kang & Kim (2012) conclude that board size does indeed affect the likelihood of AEM and REM practices22. Secondly, I control for board independence (BOARD_INDEP), measured by the proportion of independent outside directors on the board. It is expected that independent outside directors are better able to protect shareholders, by means of more effective monitoring activities, and thereby decreasing the opportunity for executives to engage in AEM and REM practices (Xie, Davidson & DaDalt, 2003; Talbi, Omri, Guesmi & Ftiti, 2015; Visvanathan, 2008). Thirdly, I control for the gender of board members (BOARD_FEM), measured by a dummy variable which indicates 1 if there is at least one female director on the board, and zero otherwise. It is expected that the monitoring activities of the board become tougher when it is comprised of at least one female (Srinidhi, Gul & Tsu, 2011), associating it with lower levels of EM. Lastly, I control for audit quality (AUDIT_BIG4), measured by a dummy variable which indicates 1 if the audit was

22_{In contrast, some researchers (Talbi et al., 2015) argue that larger boards may have more difficulty in communication}

(22)

22

performed by a BIG4 audit firm, and zero otherwise. Some studies point out that big audit firms are more effective in constraining an executives’ EM activities, compared to smaller audit firms (Becker, Defond, Jiambalvo & Subramanyam, 1998), while others indicate that there is indeed a difference in perceived audit quality, but not in actual audit quality (Boone, Khurana & Raman, 2010).

Aside from firm-specifics and governance specifics, I do also take into account the potential effects of four CEO-characteristics. Firstly, I control for CEO tenure (CEO_TENURE), which is measured as the logarithm of the sum of years that the CEO is in his or her position (Hooghiemstra et al., 2019). It is expected that a CEO increasingly engages in AEM and REM practices during the early years of his or her tenure, because of career concerns (Ali & Zhang, 2015; Chou & Chan, 2018). In addition, I also control for CEO turnover (CEO_TURNOVER), which is measured by a dummy variable which indicates 1 if the current CEO is different from last year’s CEO. It is expected that new CEOs engage in the ‘taking a big bath’ -strategy, and manage earnings downward in the year of the CEO change, because they possess increased opportunity in that year to do so (Geertsema, Lont & Lu, 2018; Wells, 2002). Thirdly, I control for the situation wherein the CEO is also chairman of the board (CEO_DUALITY), measured by a dummy variable which indicates 1 if the CEO is chairman of the board, and zero otherwise. It is expected that the monitoring ability of the board impairs when the CEO is also chairman (Chou & Chan, 2018; Cornett, Marcus & Tehranian, 2008; Jensen, 1993), thereby increasing an executives’ opportunity to engage in both AEM and REM. Lastly, I control for CEO incentives (CEO_INCEN), measured by [1- (Base salary / Total compensation)]. When the potential total compensation of CEOs is more closely linked to the value of equity, CEOs are expected to engage more in AEM and REM practices, to maximize the value of their equity (Bergstresser & Philippon, 2006; Cheng & Warfield, 2005; Chou & Chan, 2018). I do also include year and industry dummies23_{, to take}

time-specifics and industry characteristics into account.

3.5 Regression Models

In order to analyse the degree of association, and estimate the significance between the independent, dependent, moderating and control variables, I use OLS regression models, with standard errors clustered by year and by firm to address heteroskedasticity (Petersen, 2009). Linear regression assists in developing explanatory relationships between the key variables. The following models are used to test hypotheses 1 & 2:

Faculty of Economics and Business

It’s All Politics: A Quantitative Study on the Effects

of Political Induced Risk on Earnings Management.

Evidence from the United States

TABLE OF CONTENTS

1. INTRODUCTION

2. THEORETICAL BACKGROUND & HYPOTHESIS

DEVELOPMENT

3. RESEARCH DESIGN

𝑇𝐴𝐶𝐶

= ∆𝐶𝐴

− ∆𝐶𝑎𝑠ℎ

− ∆𝐶𝐿

+ ∆𝐶𝐿𝑇𝐷

− 𝐷𝐸𝑃

𝑇𝐴𝐶𝐶

𝐴

= 𝛼

1

𝐴

+ 𝛼

(∆𝑆𝑎𝑙𝑒𝑠

− ∆𝑅

)

𝐴

+ 𝛼

𝑃𝑃𝐸

𝐴

+ 𝜀

ε

𝐶𝐹𝑂

𝐴

= 𝛼

1

𝐴

+ 𝛼

𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝛼

∆𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝜀

ε

𝐷𝑖𝑠𝐸𝑥

𝐴

= 𝛼

1

𝐴

+ 𝛼

𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝜀

ε

𝑃𝑟𝑜𝑑

= 𝐶𝑂𝐺𝑆

− ∆𝐼𝑁𝑉

𝑃𝑟𝑜𝑑

𝐴

= 𝛼

1

𝐴

+ 𝛼

𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝛼

∆𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝛼

∆𝑆𝑎𝑙𝑒𝑠

𝐴

+ 𝜀

ε

𝑃𝑟𝑖𝑠𝑘

=

Σ

(1 [𝑏 ∈

P

N

] ∗ 1[|b − r| < 10] ∗

𝑓

_{(1 [𝑏 ∈}