Political uncertainty and firm tax
avoidance
R.J.H. (Rick) Kooistra
S 3789101
University of Groningen Faculty of Economics and Business
Department of Accounting
Master thesis Accountancy (EBM869B20) January 19, 2020
Word count: 12,512
Supervisor:
prof. dr. R.B.H. (Reggy) Hooghiemstra
Co-assessor:
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
2
Table of Contents
1. INTRODUCTION 4
2. PRIOR LITERATURE AND HYPOTHESES DEVOLPMENT 8
3. RESEARCH DESIGN 18
4. EMPIRICAL RESULTS 22
5. DISCUSSION AND CONCLUSION 33
REFERENCES 36
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
3
Political uncertainty and firm tax
avoidance
R.J.H. (Rick) KooistraUniversity of Groningen
Abstract
The debate about corporate tax avoidance behavior of firms has become more heated lately. In this thesis I investigate whether there is a relation between the degree of political uncertainty a firm experience and the level tax avoidance. From the agency theory, this could be explained by the fact that political uncertainty creates an opportunity for managers to engage more in tax avoidance because of the increased information asymmetry. By engaging in more tax avoidance, they can make the company more complex and thus create a shield to hide their opportunistic behavior. In addition, in times of uncertainty, it is harder for a firm to attract capital, so they use tax avoidance to keep cash within the firm. The relationship between political uncertainty and tax avoidance may be strengthened by powerful CEOs, therefore a CEO power measure has been added as moderator variable. I used a sample of U.S. listed firms in the period 2002 and 2015. The political uncertainty measure was measured on firm-level, which is quite unique. Tax avoidance was determined using the cash effective tax rate. The analysis shows that there is a significant association between the degree of experiencing political uncertainty and lower cash effective tax rates, which is an indication that more tax avoidance takes place. Marginal significant evidence was found that powerful CEOs engage in higher levels of tax avoidance, although this does not hold in robustness analyses. There was no evidence was found that CEO power influences the relation between political uncertainty and tax avoidance.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
4
1. INTRODUCTION
In America well-known companies such as IBM, Netflix and Amazon paid no income tax in America in 2018. The Institute on Taxation and Economic Policy (ITEP), an independent think tank that works on federal and state tax issues, reports that the number of Fortune-500 companies that do not pay corporate income tax in America has doubled in 2018 (Berentsen, 2019). Apple CEO Tim Cook had to appear at a hearing of a committee of the Senate of the United States, to explain the tax avoidance behavior of Apple (Drucker & Bowers, 2017). After the revelations of the so-called Panama Papers in 2016 society, firms and politicians wonder whether tax avoidance is still morally justifiable (Goudswaard, 2016). Tax avoidance is a topic that is widely discussed in the media and by politicians. At the same time, tax avoidance attracts considerable scholarly attention.
While there is no generally agreed upon definition of tax avoidance, one frequently used definition refers to tax avoidance as “the reduction of explicit taxes” (Hanlon & Heitzman, 2010 p. 137). This is a broad definition. Some papers only speak about tax avoidance when the more aggressive tax avoidance methods are used (e.g., Slemrod, 2004; Slemrod & Yitzhaki, 2002). However, in this paper I will follow the broader definition of Hanlon & Heitzman (2010).
Research on tax avoidance has been done in different areas and from different disciplines. Prior research suggests different incentives to play a role in explaining why firms engage in tax avoidance (e.g. Desai & Dharmapala, 2006; Graham et al., 2013; Armstrong, Blouin & Jagolinzer, 2015; Lietz, 2013). These incentives include firm characteristics as firm size, profitability, leverage, foreign operations and industries. In addition to firm characteristics corporate governance dimensions such as board characteristics, incentive-based managerial compensation and ownership structures are studied. In general, these studies indicate that internal factors are associated with the extent to which firms engage in tax avoidance. To date, however there is limited research done into the external factors that triggers tax avoidance behavior. In this study I intend to fill that void by focusing on one specific external factor, namely political uncertainty.
Political uncertainty can be defined as the uncertainty that arises in firms because they cannot properly predict the consequences of actions of the political system. Firms respond with their behavior to developments of external events. For example, the trade war between America and China and the Brexit has an impact on companies and stock exchanges (McGregor, 2019; FD, 2019). The Fed chairman, Jerome Powell, confirmed in a speech that uncertainty about political policy leads to uncertainties in companies and in the financial market (Zandbergen, 2019). Various studies show the influence of political uncertainty on companies. Research show that events like the Arab Spring and elections have an impact on the behavior of companies (e.g., Jens, 2017; Acemoglu, Hassan & Tahoun, 2018; Hassan, Hollander, Van Lent & Tahoun, 2019). The stock market also considers political uncertainty by valuating stock prices (Pastor &
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
5
Veronesi, 2013). Duong, Gul, Nguyen, & Nguyen (2017) hypothesized that economic policy uncertainty has a positive relation with tax avoidance and their research shows evidence for this. In their study they used the economic policy uncertainty (EPU) index from the paper of Baker, Bloom & Davis (2016). This index measures the degree of political uncertainty at a national level, based on the frequency use of certain words which are associated with uncertainty from newspaper articles from the 10 leading U.S newspapers. Hassan et al. (2019) have developed another method, which makes it possible to measure political uncertainty at firm level. They look at certain two-word combinations (bigrams) in quarterly conference call meetings about earnings releases. These specific bigrams are divided by the total number of bigrams in the transcript and this will give a number that represents the degree of political uncertainty that the firm experiences.
As with the research of Hassan et al. (2019) I expect that external factors such as political uncertainty lead to certain behavior of firms. In this research I will investigate whether firms that experience more political uncertainty engage more in tax avoidance activities. There are actually several reasons why political uncertainty can lead to more tax avoidance. From the agency theory perspective there can be a difference in interests between the principal and the agent (Eisenhardt, 1989). If the agent does not want to behave optimally and wants to extract managerial rent, he will create a shield to hide his opportunistic behavior by engaging in more tax avoidance (Desai, Dyck & Zingales, 2007). Desai and Dharmapala (2006) suggest namely that tax avoidance and managerial rent extraction are complementary. In times of uncertainty there is more information asymmetry that makes it easier for an agent to engage in tax avoidance. In addition to this argument, it may also be the case that companies find it more difficult to meet their financing needs in times of political uncertainty. D'Mello, Jha and Toscano (2019) argues that in times of political uncertainty, companies find it harder to raise capital. Edwards, Schwab and Shevlin (2015) indicates that companies with financing problems can overcome this by engaging in tax avoidance activities. Duong et al. (2017) shows that there is an association between lower tax rates and more political uncertainty.
The relationship between political uncertainty and tax avoidance may be influenced by the CEO of the firm. The Dyreng et al. (2010) study show that executives play a significant role in the degree of tax avoidance. Armstrong et al. (2015) found that firms with a more independent board engage in less tax avoidance. Previous studies have shown that CEO power influences firm outcomes. ‘CEO power is about the ability of the CEO to overcome resistance and consistently influence key decisions within a firm’ (Haleblian & Finkelstein, 1993 p. 848). Baker et al. (2019) showed evidence that the level of CEO power influences the level of earnings management. In addition, Mamun (2016) showed that powerful CEOs have a positive association on tax avoidance. CEO power is therefore likely to strengthen the relationship between political uncertainty and tax avoidance.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
6
The purpose of this research is to provide new insights into the understanding of external factors for firms to commit tax avoidance. In this case political uncertainty. The research question that will be answered in this thesis is:
What is the influence of experiencing political uncertainty on the level of tax avoidance activities, for US listed firms, and is this influenced by CEO-power?
In this thesis I look at the firm-level to see whether there is a relationship between the political uncertainty that a company experiences and the extent to which they engage in tax avoidance. To my knowledge, this is something that has not been done before. The independent variable in this study is political uncertainty. For measuring political uncertainty, I used the same firm-level data as in the Hassan et al. (2019) paper. My sample consist 5,282 firm-year observations of 1,055 U.S. listed firms between 2002 and 2015. Tax avoidance will be measured according to the cash effective tax rate (ETR). The analyses carried out showed as predicted that there is a positive significant relation between the degree of experiencing political uncertainty and the degree of committing tax avoidance. This association holds after additional analyses have been carried out. For CEO power a marginally significant effect has been found between the powerful CEOs and the level of tax avoidance. However, in additional robustness analyses this relation is not confirmed, which makes it hard to conclude that CEO power moderates the main association.
This research contributes to the literature in several ways. First, because it contributes to the expansion of the literature on determinants of tax avoidance. Previous research mainly looked at internal factors that tax avoidance can explain such as firm characteristics (e.g., Hanlon & Heitzman, 2010; Badertscher, Katz & Rego, 2013) and at the characteristics of the CEO and of the board (Armstrong et al., 2015). In this thesis, I will focus specifically on an external factor that may be an incentive for tax avoidance, namely political uncertainty. Second, limited research has yet been carried out into the consequences of political and policy uncertainty at firm-level. The majority was carried out at the macroeconomic level and looked at investment decisions, stock returns and lobbying activities (e.g., Jens, 2017; Baker et al. 2016). I will look at the degree of political uncertainty which firms experience on a firm-level and whether this leads to other tax avoidance behaviors. Conducting research at firm firm-level provides more insight into the relationship between the two variables than at national level. The aim of this research is therefore to contribute to the existing literature on political uncertainty and tax avoidance and perhaps offering new insights for policy makers.
The remaining of this thesis proceeds as follows. In the second chapter I discuss the theoretical background which results in hypotheses. The third chapter is about the research methodology that will be used in this research. In chapter four the empirical results will be shown. The last chapter contains a
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
7
discussion of the results, a conclusion will be drawn up, the implications of the research are discussed, and recommendations are made for future research.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
8
2. PRIOR LITERATURE AND HYPOTHESES DEVOLPMENT
2.1 Tax avoidanceWhile tax avoidance has been a frequently discussed topic in the accounting, finance, and management literature, today there is no generally agreed upon definition of tax avoidance in the literature (Hanlon & Heitzman, 2010). In their review of the accounting-based literature on tax avoidance, they use the following definition: “tax avoidance is the reduction of explicit taxes” (Hanlon & Heitzman, 2010 p. 137). In this thesis, I follow this broad definition. The definition of Hanlon and Heitzman (2010) follows the concept of Dyreng et al. (2008) that all transactions that have an effect on the company's tax liability must be included. Hanlon and Heitzman (2010) described that tax avoidance activities can be arranged on a spectrum where on one end use of certain deduction rules, while on the other end they identify for example tax evasion and tax aggressiveness. The main notion underlying their view on tax activities is the extent to which these activities reduce the tax burden. In this paper I make no distinction as to how the tax reduction is to be achieved, and the whole spectrum is therefore taken into account.
2.1.1 Theories of tax avoidance
Several theories are relevant to discuss when doing research into tax avoidance. However, I will
only discuss the theories suitable for explaining why managers of firms are more or less involved in tax avoidance activities. Manny research into tax avoidance takes place in agency setting, although this is not undisputed or the only relevant theory for research into tax avoidance (Lietz et al., 2013). Theories about the individual taxpayer are also interesting in this field of research. Below the trade-off theory and the agency theory will be discussed.
Trade-off theory
While trade-off theory in essence focuses on choices individual taxpayers make, the theory has been increasingly used to explain corporate tax decisions (Hanlon & Heitzman, 2010). Trade-off theory started with a paper in which a model for the individual taxpayer and their tax behavior was presented. This model was developed by Allingham and Sandmo (1972). The main idea of Allingham and Sandmo (1972) is that that compliance with tax rules is determined by the height of the tax rate, the probability to be caught, the penalty that has to be paid in case the individual is caught, and her degree of risk aversion. The key notion that tax decisions are essentially a trade-off decision also applies to corporate taxes (Slemrod, 2004). It must therefore be examined whether the benefits of paying less tax, outweigh possible costs of tax avoidance, such as fines and reputation damage. So a off should be made, hence the name trade-off theory (Lietz, 2013). The literature describes various risks that may arise at firms as a result of tax
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
9
avoidance behavior. Most of the named risks are tax risk, reputational risk, political risk, the risk of managerial rent extraction and stock price risk. The tax risk is about uncertainty as to whether the tax position will remain. The IRS can carry out additional research and decide that the company still has to pay more tax. This may also involve penalties and interest costs (Christensen et al., 2015; Slemrod & Yitzhaki, 2002). A company that is widely in the news due to tax avoidance can suffer a loss of reputation as a result (Bankman’s, 2004; Zimmerman, 1983). Companies must undertake activities that comply with the socially prevailing values in society, this is described in the legitimacy theory (Suchman, 1995). There is actually a 'social contract' between society and companies (Laguir, Stagliano & Elbaz, 2015). Because society expects a company to contribute a fair share to the costs of society as a whole, by paying taxes. Tax avoidance can be seen as a breach of the contract and could lead to reputational loss. Christensen et al. (2015) describe the events at Starbucks in the United Kingdom. Starbucks legally avoided the tax but did suffer such damage to their image that they suffered from a customer backlash, stock price decline and were severely affected by action groups. They even had to close stores in the United Kingdom. The political risk is about the event that politics will interfere in the tax avoidance of companies because society observe that companies do not contribute their fair share to the tax revenues (Hanlon & Slemrod, 2009). Desai and Dharmapala (2006) argue that managers can use tax avoidance for their own interests. They make use of tax avoidance to make the company more complex, thus creating more information asymmetry. This leads not only to government rent extraction, but it can also increase the managerial rent extraction (Desai & Dharmapala, 2006). This last risk mentioned arises from a problem which is described in the agency theory, this theory is discussed in the next paragraph. Because tax avoidance involves certain risks as described above, this can also lead to reactions from investors. This is why tax avoidance can also lead to stock price risks, this is the risk that the stock price reacts negatively to tax avoidance by declining or becoming more volatile (Kim et al., 2011).
Agency theory
The agency theory is about the conflict that arises from the separation between ownership and control (Eisenhardt, 1989). This theory refers to 'the principals' which are the owners of the firm and the 'agents' which refers to the management of the company. The agents are actually hired by the principals, which are the shareholders of the company to pursue activities in the shareholders’ interests. Once the agents have been hired, there is a possibility that agents do not act in the best interests of the principals. This may be because there is a divergence of interests between the agents and principals. Agents can be self-interested, but they can also be averse to risk, and as a result decisions may not always be in the best interest of the shareholders (Gerhart et al.,2009). Behavior of a manager who not necessarily acts in the best interest of the shareholder is referred to as moral hazard (Ndofor et al., 2015). Agents can engage in moral hazard behavior because there is an asymmetry of information between the principals and the agents.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
10
Ndofor et al. (2015) suggests that if the information asymmetry is bigger agents have more opportunities to engage in moral hazard behavior. If agents do not behave optimally towards the principals and therefore show moral hazard behavior, a conflict can arise between agents and principals. From the agency theory, tax avoidance can actually be viewed from two perspectives.
The first perspective assumes that the agent behaves in the best interest of the shareholder. Graham and Tucker (2006) present evidence that shareholders see tax avoidance as a value creating activity. Koester et al. (2011) say that managers involved in tax avoidance can be seen as good stewards of the company assets. So, if the agent behaves in the best interest of the shareholders, he will focus on profit maximization, so reducing the tax burden contributes to the profitability of a company as long as the incremental revenues exceed the incremental costs (Slemrod, 2004). Hanlon & Heitzman (2010) also argue that risk neutral shareholders expect the management to reduce the tax burden as long as the benefits outweigh the costs. These benefits can be savings on tax expenses, which consequently imply that more funds are available to shareholders, for instance in the form of dividend pay outs and/or share repurchases. However, tax avoidance may also be risky and may potentially lead to costs such as: legal costs, fines and reputational damage.
Due the separation between ownership and control, agents can possibly not behave optimally, in contrast of the principals, because they have more information at their position and may put their own interests first (Jensen & Meckling, 1976). From this point of view, this can mean that this conflict of interest results in more or less tax avoiding behavior by managers than the shareholders expect. The first studies on tax avoidance in an agency perspective looked at the efficiency loss of tax decisions due the separation (Chen & Chu, 2005), and how this problem could be addressed (Crocker & Slemrod, 2005). Desai & Dharmapala (2006) built upon the agency view on tax avoidance, arguing that managers who are self-interested use tax avoidance to hide their opportunistic behavior. Desai et al. (2007) describe a situation in which the managers structure the firm in such a way that it becomes very complex. This enables them to perform actions that reduce the tax burden and they can extract rents from the firm. So, tax avoidance increases the information asymmetry and as a result, the manager's opportunistic behavior is more difficult to detect by the principals. The managers create a shield to hide their behavior. Therefore, Desai & Dharmapala (2006) see tax avoidance and managerial rent extraction as complements. It is the role of the board of directors to monitor the behavior of the managers. They should try to reduce the agency problems. This is why corporate governance mechanisms are often involved in studies to agency problems.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
11 2.1.2 Tax avoidance determinants
Tax avoidance is a highly researched subject and has been researched from different disciplines (Hanlon & Heitzman, 2010). Many studies have been carried out into the incentives for committing to tax avoidance activities, which may arise within the own organization, but can also arise to external factors. Studies were carried out into whether certain company characteristics are related to tax avoidance. Studies have shown what possible predictors for tax avoidance can be. For example, there are studies that examined the influence of firm size and the extent to which foreign activities affect the influence of tax avoidance (e.g. Wilson, 2009; Rego, 2003; Hanlon, Mills & Slemrod, 2007; Hanlon & Slemrod, 2010). Ownership structure also affects the degree of tax avoidance (Baderetscher et al, 2013; Chen et al, 2010). These studies conclude that companies with a higher degree of ownership concentration engage in less tax avoidance because there is less information asymmetry in these companies. Another finding in this stream in the literature is that firms facing greater financial constraints are associated with tax avoidance activities in an attempt to save cash (e.g., Edwards et al., 2015).
As mentioned earlier, some papers look at tax avoidance from an agency perspective and therefore corporate governance mechanisms are often included in the research. Tax avoidance can be seen as a value-creating activity for shareholders and should therefore be encouraged in the compensation contract between the agents and the principals (Scholes, Wolfson, Erickson, Maydew & Shevlin, 2009). Phillips (2003) finds evidence that after-tax performance measures for business unit managers lead to lower ETRs. He does not find this evidence for CEOs, but Rego and Wilson (2012) do find evidence for this. An increase in the pay-performance sensitivity for both the CEO and the board leads to more tax avoidance behavior, according to Minncik & Noga (2010), Graham & Tucker (2006) and Desai and Dharmapala (2006). Dyreng et al. (2010) have investigated the role of executives on tax avoidance. They have found evidence that executives play a significant role in the degree of tax avoidance. In that study, they looked at executives who are both directly and indirectly responsible for the tax planning process. Not all test results were significant, but they concluded that top executives set a 'tone at the top'. This means that individuals are motivated or demotivated to engage in tax avoidance activities by top executives. The Armstrong et al. (2015) paper shows that the independence of the board of directors also plays a role in the degree of tax avoidance. A more independent board will result in lower levels of tax avoidance.
In this thesis, I will not look for a yet undiscovered determinant of tax avoidance from inside the firm, but I will look at an external factor, in this case political uncertainty, for committing a higher degree of tax avoidance. In the next section I will discuss political uncertainty and why political uncertainty is a possible incentive to engage in tax avoidance activities.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
12 2.2 Political uncertainty
For this study, it is relevant to look at literature that has examined the way firms react to political uncertainty. First, I will briefly discuss previous studies on political and policy uncertainty and firm behavior. In the literature, the terms political uncertainty and policy uncertainty are often used interchangeably, because political uncertainty can lead to uncertainties about the policy. Conversely, policy is determined by politics, and these concepts can therefore be seen as synonyms of each other. After that, I will discuss the possible relation between political uncertainty and tax avoidance I will formulate a hypothesis for this research.
2.2.1 Political uncertainty and firm behavior
Politicians constantly make choices about all kinds of things. For example, they determine the policies conducted at state, national and international levels. Many of the policies and laws that result from political decisions affect companies. The influence of government policy as well as other events such as trade wars and Brexit also potentially affect the outcomes of companies. The impact is unclear and creates political uncertainty and is felt by companies (Zandbergen, 2019). Political uncertainty can be defined as the uncertainty that arises in companies because they cannot properly predict the consequences of actions of the political system. Uncertainties and associated risks that arise from the political system can have an impact on firm decisions and behavior (Jens, 2017).
Various studies have shown that companies respond to political uncertainty with their behavior (e.g. Jens, 2017; Acemoglu, Hassan & Tahoun, 2018; Hassan, Hollander, Van Lent & Tahoun, 2019). Most studies use only two techniques to determine political uncertainty. One of the techniques used is to measure around events that generally cause political uncertainty, such as elections and events that disrupt society, such as Brexit and the Arab Spring (e.g. Achemoglu et al.,2018; Jens, 2017; Çolak, Durnev & Yiming Qian, 2017). The second technique commonly used is the economic policy uncertainty index (EPU), which was developed in the paper by Baker et al. (2016). This index is based on newspaper coverage of policy related economic uncertainty, which is measured at the national level. Hassan et al. (2019) have developed a model that does not measure political uncertainty at the national level but at the firm-level. Specifically, they used the quarterly conference call meetings, measuring the time that is spent on discussing political risk and show that firms indeed react to political uncertainty.
Hassan et al. (2019) have namely found that political uncertainty leads to stock market volatility, a decrease in investment and employment growth. Firms that experienced high political uncertainty also engage in more donating and lobbying activities (Hassan et al., 2019). Jens (2017) shows that there is a link between political uncertainty about elections and a decrease of corporate investments. Pastor & Veronesi’s (2013) model shows that the stock market demands a risk premia for political uncertainty that can cause
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
13
stock prices to fall. This would mean that the cost of equity of the company increases, this effect is stronger in times of economic weakness. Acemoglu et al. (2018) also show that there is a relation between political uncertainty and the share price. They show that during the Arab Spring there was a relation between protests against a certain political group and the market value of companies associated with that group. Firms do not want investors to be worried and will therefore disclose more voluntarily, according to research by Nagar et al. (2019). The study by Çolak et al. (2017) also shows that companies postpone their IPOs around events that increases political uncertainty because the uncertainty increases the cost of capital. In this study I will investigate whether political uncertainty leads to a change in the level of tax avoidance behavior.
2.2.2 Political uncertainty and tax avoidance
There are two possible arguments that can explain the relationship between political uncertainty and tax avoidance. Both arguments are discussed below and will lead to the main hypothesis for this study. The agency theory is about the separation between ownership and control, and the managers should work in the best interest of the shareholders. In the case of risk-neutral shareholders, the agents should aim for maximum profit (Hanlon & Heitzman, 2010). Scholes et al. (2009) argue that tax avoidance can be seen as a value-creating activity. Companies should therefore engage in tax avoidance if the benefits exceed the costs (Slemrod, 2004). According to the trade-off theory (e.g., Allingham and Sandmo, 1972), a trade-off should be made whether the revenues (a lower tax burden) exceed the costs (e.g. prosecution, penalties & reputational damage). However, as the agency theory also emphasizes, there can also be a conflict between the principals and the agents, since the manager can behave opportunistically (Jensen & Meckling, 1976). In times of political uncertainty, the information asymmetry between management and shareholders is increasing (Dai & Zhang, 2019). This is because shareholders are less able to make good predictions about the future of the company, and therefore stock prices are also becoming more volatile (Dai & Zhang, 2019; Kholbadolov, 2012). Anyway, it is difficult to make estimates about the future, but if there is more political uncertainty it is more difficult. Making predictions is easier if you have more information at your disposal. Because the fact that managers have more information at their disposal instead of the shareholders there will be an increase in the information asymmetry in times of political uncertainty. The increased information asymmetry provides managers with the opportunity to engage in higher levels of tax avoidance in order to make the company more complex. A company becomes more complex when the cause-effect attributions are more difficult to make (Ndofor et al., 2015). By means of tax avoidance, a company can become more complex by using tax havens, the creation of complex structures involving tax-indifferent parties or by negotiating deals with tax authorities (Desai et al., 2006). By making the company more complex they thus create a larger shield to hide their opportunistic behavior (Lietz, 2013; Desai et al., 2007). This shield gives them more opportunities for managerial rent extraction.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
14
The second argument concerns the relation between political uncertainty and the financing needs of firms. The Pastor and Veronesi (2012, 2013) paper shows that political uncertainty leads to a demand for a risk premium from shareholders, which in turn leads to a decrease in stock prices and adds to their volatility. This decrease leads to higher financing costs and difficulties in obtaining loans (Bordo et al, 2016). D'Mello et al. (2019) argue that an increase in political uncertainty leads to an increase in the volatility of cash flow. They argue that this increased volatility leads to an impediment in raising debt because debt providers prefer not to see too high volatile cash flows. The cost of borrowing in times of financial constraints is also rising. In addition, there is empirical evidence that companies hold more cash during times of political uncertainty in order to be less dependent on credit. One way to keep more cash is by lowering the tax burden, which can be done through tax avoidance activities. Edwards et al. (2015) also argue that due to the increase in the costs of attracting external funds, companies are more involved in tax avoidance in order to keep cash within the company. They also find evidence that financial constraint firms generate significantly more cash savings through tax avoidance activities. The paper by Edwards et al. (2015) assumes that the implementation costs of tax avoidance activities are independent of the costs of financial constraints. The study by Chen et al. (2016) shows that companies maintain higher levels of cash holdings around elections by means of tax avoidance behavior. Li et al. (2018) have also observed an increase in tax avoidance at times of political uncertainty. The study by Duong et al. (2017) shows that the tax rate of companies is lower in times of political uncertainty, and that companies are also more concerned with more aggressive forms of tax evasion. As a result of the above, it can be said that due to financial constraints that arise as a result of the experience of political uncertainty, managers make the trade-off to engage in more tax avoidance, in order to keep more money in the company.
Bases on above explications I expect management to make the decision (trade-off) to engage in more tax-avoiding activities during times of political uncertainty and predict:
H1: There is a positive association between firm-level political uncertainty and the degree of corporate tax avoidance.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
15 2.3 The moderating role of CEO power
So far, I have argued that political uncertainty has an effect on the level of tax avoidance. However, I also will argue that depending on the degree of power the firm’s CEO has, this association may be stronger in some firms than in other firms. First of all, I will discuss the dimensions from which CEOs can be derive power. I will describe how and why these dimensions have effects on firm outcomes. After that, I will discuss why CEO power can influence the level of tax avoidance and how it can moderate the relationship between the degree of political uncertainty and the level of tax avoidance. Two hypotheses follow from this.
2.3.1 CEO power dimensions
From his position, the CEO can influence the policy of the company (e.g. Baker et al., 2019; Dyreng et al., 2010). An overall definition of CEO power is that ‘CEO power is the ability of the CEO to overcome resistance and consistently influence key decisions within a firm’ (Haleblian & Finkelstein, 1993 p. 848). The literature recognizes that CEOs can derive their power from various sources. Finkelstein's research (1992) is the basis for many studies into the influence of power executives. He describes four dimensions from which executives can derive power: structural power, ownership power, expert power and prestige power.
Structural power is about the formal position of the CEO in the company (Lewellyn & Muller-Kahle, 2012). Finkelstein (1992) describes it as the legitimacy authority over others because of the nature of his position. A CEO has more power if he is also chairman of the board, which is called CEO duality (e.g. Lewellyn & Muller-Kahle, 2012; Combs et al., 2007; Haynes & Hillman, 2010; Baker et al., 2019). If the board of directors is less independent from the CEO, the CEO will have more power. Directors who were first employees of the company or who were appointed at the time that the CEO was already in charge, can be more loyal and sympathetic to the CEO, which is why they are regarded as less independent (Lietz, 2013). Less independent boards are therefore also seen as a source of CEO power.
Ownership power is about a CEO who holds shares in the company. This gives him power because of his voting rights. Owning shares also increases the status of the CEO in the firm and in the board (Lewellyn & Muller-Kahle, 2012). In addition, it is easier to replace low ownership CEOs (Ocasiao, 1994). CEOs who own a lot of shares can also influence the board's agenda and thus the decision-making process (Schmid et al., 2018).
Expert power is about the CEO's ability to handle environmental contingencies properly in order to still achieve the company's goals (Finkelstein, 1992). In the early years of the CEO, he is still very dependent on other executives and has to develop his leadership skills (Combs et al., 2007). A CEO with a lot of experience can use this preponderance to influence the board's choice. This is why CEO tenure is seen as
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
16
an important measure of CEO expert power (e.g. Combs et al., 2007; Lewellyn et al., 2012; Schmid et al., 2018).
Prestige power is about the status of the CEO in the company but also with all other stakeholders (Finkelstein, 1992). It can be about the legitimacy of his decisions, for example a CEO with many corporate board mandates has a higher prestige power (Schmid et al, 2015). Having important contacts, other directorships and having studied at renowned universities are also factors from which prestige power can be derived (Finkelstein, 1992).
2.3.2 CEO power and tax avoidance
Several studies have shown that powerful CEOs can influence the outcomes of firms. Combs et al. (2007) show that CEO power has a moderating effect on the relationship between board composition and firm performance. Haynes and Hillman (2010) find partial support that CEO power moderates the effect between board capital and strategic change. The study by Sheikh (2010) shows that powerful CEOs are less involved in CSR activities. Lewellyn and Muller-Kahle have hypothesized that CEO power is positively related to excessive risk-taking behavior; they also find general support for this. Schmid et al. (2018) investigated whether Americanization and the level of CEO pay is related and if this relation is moderated by CEO power. They conclude that this relationship is strengthened through powerful CEOs. Schmid et al. (2018) argue that this relation exists because CEOs think in their own interest and therefore want to earn more. Baker et al. (2019) investigated whether powerful CEOs and CFOs influence the degree of earnings management, looking at both accruals earnings management and real earnings management. The result of this research is that CEOs are more powerful compared to CFOs, and that CEO power has significant influence on both ways of earnings management. One of the reasons that is mentioned for this associations is that firms want to beat the forecast of analysts, this would give a positive impression on the firm but also on the board. So, this puts the CEO more in the spotlight.
With regard to tax avoidance in particular, the CEOs set the 'tone at the top' and thus determine the extent to which other employees engage in tax avoidance activities (Lietz, 2013). The extent to which the CEO can exert influence is related to corporate governance factors (Minnick & Noga, 2010). They found amongst others that board independence influences tax rates. Lanis and Richardson (2011) and Armstrong et al. (2015) also conclude that more independent directors lead to less tax avoidance. The Mamun (2016) study provides evidence that powerful CEOs have a positive effect on the level of tax avoidance. He also concluded that powerful CEOs generally extract more rent through more tax avoidance activities.
From the agency theory it can as well be argued why CEOs do engage in more tax avoidance and especially powerful CEOs. A CEO should work at the best interest of the shareholders, but because of the separation between ownership and control, there is information asymmetry which allows the CEO to behave
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
17
self-serving (Jensen & Meckling, 1976). By engaging in higher levels of tax avoidance, the firm becomes more complex and this increases the degree of information asymmetry (Desai et al., 2007). This may create a shield that managers can take advantage of to behave more opportunistically and extract more rents (Desai & Dharmapala, 2006). It is likely that CEOs want to make use of this because they often consider their own interests as important.
Literature shows that CEOs often act out of self-interest. For example, they demand more pay, they engage in more earnings management from the idea of improving themselves (Schmid et al., 2018; Baker et al., 2019). The decision to engage in more tax avoidance also increases the possibility that the CEO can behave opportunistically (Desai & Dharmapala, 2006). This decision can therefore also be made for personal gain, and this is something CEOs often take into account when making decisions (Gerhart et al., 2009). Powerful CEOs suffer less resistance, from other board members, when making decisions (Haleblian & Finkelstein, 1993). This makes it easier for them to make the decision to engage in more tax avoidance. It also appears that CEOs can exert influence on other directors and that the directors succumb to pressure from powerful CEOs (Friedman, 2014; Feng et al., 2011). It is therefore likely that it will be easier for powerful CEOs to make the decision to engage in more tax avoidance.
In view of the above, I expect CEO power has a positive effect on the level of tax avoidance, which leads to the following hypothesis.
H2: CEO power is positively related to the level of tax avoidance.
Political uncertainty leads to more information asymmetry between management and shareholders (Dai & Zhang, 2019). This increases the opportunity for managers to behave more opportunistically. Powerful CEOs can engage in more tax avoidance to create a more complex environment. This increases the information asymmetry and their shield to hide managerial rent extracting. Because powerful CEOs have more influence and can therefore more easily engage in tax avoidance activities, I expect that the relationship between political uncertainty and the degree of tax avoidance will be strengthened by powerful CEOs, therefore I come to the hypothesis below.
H3: The positive association between firm-level political uncertainty and the degree of tax avoidance is strengthened by CEO power.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
18
3. RESEARCH DESIGN
3.1 Sample selection and data collection
My sample is compiled from data from various sources, including the data of Hassan et al. (2019), Compustat, and ExecuComp. My sampling procedure starts with the data from Hassan et al. (2019). For the data concerning political uncertainty, I use the data of Hassan et al. (2019). Their sample consists of 209,092 quarterly observations of 9,478 American listed companies over the period 2002 to 20161. The
financial data for the calculation of tax avoidance and the control variables came from the Compustat database. To determine CEO power, I use data from the MSCI database and the ExecuComp database. Companies from highly regulated industries and the financial service sector (SIC codes 4400-5000 and 6000-6500) have been excluded from the sample, which is in line with previous studies (e.g. Hanlon & Shevelin, 2005). In addition, data was lost on the merging of the various data files due missing data, in particular data relating to CEO power. The final sample consists of 1,055 firms with 5,282 firm-year observations.
3.2 Measurement of variables Dependent variable
The dependent variable in this study is tax avoidance. This can be determined via multiple measures, and for these measures there are certain strengths and limitations in use (Hanlon & Heitzman, 2010). In this study, the cash effective tax rate (Cash ETR) is used as the primary measure for tax avoidance. The Cash ETR (Cash_ETR) is determined by the cash taxes paid divided by per-tax income minus the special items. The Cash ETR has been chosen as the primary measure for the following reasons. Of all the tax avoidance measures, the Cash ETR is the most direct measure of tax avoidance (Edwards et al, 2015). If a company experiences political uncertainty and changes his behavior, this will first be measurable via the Cash ETR. In addition, the Cash ETR is an accurate measure because it is less subject to accounting practices that is not related to tax avoidance (Dyreng et al, 2008; Hanlon & Heitzman, 2010). The Cash ETR is truncated between 0 and 1 and negative denominators are deleted. A lower Cash_ETR stands for a higher level of corporate tax avoidance.
Independent variable
The independent variable in this study is political uncertainty. The data on political uncertainty originates from the Hassen et al. (2019) study. Hassan et al. (2019) construct a firm-level measure of
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
19
political uncertainty by means a computational linguistics method. They analyzed 175,797 transcripts of earnings conference calls from 9,478 listed companies in the United States. To determine the political uncertainty, they looked at certain two-word combinations (bigrams) in the transcripts of the call. They first made two libraries with bigrams one related to political subjects and one to non-political topics. Then they counted how often a political bigram occurs and they also searched for the bigrams close to the words of risk and uncertainty and their synonyms. Each bigram is also weighted with a score that indicates how strong the relations is between the bigram and the discussion of political topic. These scores are divided by the total number of bigrams in the transcript. The outcome represents the degree of political uncertainty that a firm experience. By performing additional analyses, Hassan et al. (2019) has proven that their measure is valid. Hassan et al. (2019) measured political uncertainty at quarterly level. In this study I do the analyses at annual level and therefore the political uncertainty data is converted into a proxy at the year level. This is done by linking the political uncertainty quarter data with the corresponding quarter in which the fiscal year ends. The logarithm of the value of the political uncertainty measured of that quarter, is used as political uncertainty proxy (Prisk) for the whole year of a certain firm. Whereby, a higher Prisk variable stands for experiencing a higher degree of political uncertainty.
Moderating variable
The moderator in this thesis is CEO power. Finkelstein (1992) suggests that power is a multidimensional concept and that power should therefore be measured as a composite measure reflecting the various sources from which power can be derived. Finkelstein (1992) describes four dimensions from which CEOs can derive power. The proxies used for the dimensions are indicated hereafter. In this study, structural power is measured by looking at the independence of the board of directors. In the study of Haynes & Hillman (2010) board independence was also used as a proxy for structural power. For this proxy, the number of outside directors is divided by the number of directors. Then 1 minus this percentage is conducted. The percentage of shares held by the CEO is used in this study as a proxy for ownership power, as also has been done in some previous studies (e.g. Combs et al., 2007; Lewellyn et al., 2012; Schmid et al., 2018). Finkelstein (1992) emphasizes that ownership power can also be obtained if the CEO is the founder of the company. Therefore, a dummy variable is added which indicates 1 if the CEO is also the founder of the company. Expert power can be determined by looking at CEO tenure (Schmid et al., 2018). Tenure is determined by the number of years that the executive has been active as CEO of the firm. A higher tenure suggests more power. Prestige power can be determined by looking at the number of boards the CEO sits on, functioning in more boards gives more prestige power (Finkelstein, 1992). The proxy for prestige power is therefore the number of boards the CEO sits on. Again, a higher number signals a higher degree
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
20
of power. To determine the index for CEO power (CEO_Power), all proxies are standardized and summed. A higher value on CEO_Power indicates a more powerful CEO.
Control variables
To prevent biased results, I have identified several control variables, namely: firm size, return on assets, financial leverage, foreign income, property, plant and equipment, intangibles, equity income, market-to-book ratio, investment intensity and R&D intensity.
A firm characteristic that can influence the level of tax avoidance is the firm size, therefore firm size is a control variable. The results of scientific research into the influence of size on tax avoidance are mixed and therefore I do not make a prediction in the direction (Wilson, 2009; Lietz, 2013). The firm size (Size) is determined by the logarithm of the total assets. Companies with higher profitability have a greater incentive to reduce their tax burden. And that is why profitability is often used as a controlle variable in tax avoidance studies (Lietz, 2013). Profitability is determined by the return on assets (ROA). This is obtained by dividing the pre-tax income minus extraordinary items by the total assets. I also control for financial leverage (Lev) because the form of financing of a company may also affect the level of tax avoidance. Debt financing already brings tax benefits, and therefore companies with a high level of debt financing may feel less inclined to reduce the tax burden in other ways as well (Graham & Tucker, 2006). Badertscher et al. (2011) emphasizes that it may also be the case that high leveraged firms have a greater incentive to avoid tax since they have to pay a lot of interest. Financial leverage (Lev) is calculated as the long-term debt divided by the total assets. Multinational companies have foreign activities and the study by Rego (2003) show that companies with more foreign operations also engage in higher levels of tax avoidance. Research by Hanlon, Mills, and Slemrod (2007) shows that multinationals are also more often investigated by the IRS for violations. However, there are also papers that argue that large companies receive more attention and are therefore more controlled by regulators and society. If it turns out that they are involved in tax avoidance activities, they have a greater loss of reputation (Watts and Zimmerman, 1986). As probability of political cost of these companies increases, they will be less concerned with tax avoidance (Hanlon & Slemrod, 2010). In order to check this, I look at the foreign income (Foreign_Inc). This is determined by dividing the foreign pre-tax income by the total assets. Because companies with a lot of property, plant and equipment have more opportunities to avoid tax, there is controlled for this item in tax avoidance research (Dyreng et al., 2010; Lietz, 2013). The control variable (PPE) is determined by dividing the property, plant and equipment by the total assets. Other studies also often control for intangible assets, equity income, market to book ratio investment intensity and R&D intensity (Dyreng et al., 2010; Duong et al., 2017; Lietz, 2013). The intangible assets (Intan) are determined by dividing the intangible assets by the total assets. Equity income (Equity_Inc) is calculated by dividing the equity income by the total assets. The market to
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
21
book ratio (Market_to_Book) is calculated by dividing the market capitalization by the total assets. The investment intensity (Invest) is determined by dividing the capital expenditures by the property, plant and equipment. To determine the R&D intensity (R&D), the R&D expenses are divided by the total assets.
3.3 Empirical models
To determine the association between the variables I use a cluster regression in which the standard errors are clustered by company (i) and by year (t). Model 1 is used to test the first hypothesis. The ! stands for the slope of the variables and " is the error term. For the first hypothesis it is expected that the slope of Prisk is negative and significant. Model 2 shows the relation between CEO power and tax avoidance, which is hypothesis 2 of this study. The formula for model 2 is shown below. I expect the !1 in this model to be negative and significant. Model 3 shows a moderated regression and therefore the interaction term (Prisk * CEO_Power) is added. I expect this interaction term to be positive and significant.
Cash_ETRit = !0 + !1Priskit + !2Sizeit + !3ROAit + !4Levit + !5Foreign_Incit + !6PPEit
+ !7Intanit + !8Equity_Incit + !9Market_to_Bookit + !10Investit+ !11R&Dit + !12Year_Dummyit
+ !13Industry_Dummyit + "it
[ Model 1]
Cash_ETRit = !0 + !1CEO_Powerit + !2Sizeit + !3ROAit + !4Levit + !5Foreign_Incit + !6PPEit
+ !7Intanit + !8Equity_Incit + !9Market_to_Bookit + !10Investit+ !11R&Dit + !12Year_Dummyit
+ !13Industry_Dummyit + "it
[ Model 2]
Cash_ETRit = !0 + !1Priskit + !2CEO_Powerit + !3(Prisk*CEO_Power)it + !4Sizeit + !5ROAit + !6Levit
+ !7Foreign_Incit + !8PPEit + !9Intanit + !10Equity_Incit + !11Market_to_Bookit + !12Investit
+ !13R&Dit + !14Year_Dummyit + !15Industry_Dummyit + "it
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
22
4. EMPIRICAL RESULTS
4.1 Descriptive statistics
Table 1 shows the descriptive statistics for the variables of interests. The table shows that the average Cash_ETR is 0.23 with a standard deviation of 0.14. Prisk reflects the perceived political uncertainty at firm-level, this data has been made available by Hassan et al. (2019). Prisk has an average of 4.19 and a standard deviation of 1.05. The CEO_Power index variable has an average of .084, a minimum of -3,76 and a maximum of 9,45. Table 2 also shows the means of the three most important variables per year. It demonstrates that the Cash_ETR fluctuates over the years without a clear pattern. For the Prisk, you can recognize the financial crisis of 2008. That year has the highest mean value after which the mean decreases a bit each year. For the CEO_Power index, you can see a pattern where the power of the CEO decreases every year. This could partly be explained by strict governance regulations. Appendix B contains another table in which the means of the separate CEO power dimensions per year are shown. This table shows that the CEO tenure and ownership in particular have decreased. In addition to these variables, the descriptive statistics of the control variables are also shown in table 1. Firms in the sample have an average return on assets of 0.09 and an average leverage of 0.19 with a standard deviation of 0.17. Income from foreign operations has a mean of 2.7% of total assets for this sample. The average R&D expenditure is 3% of sales.
Table 3 presents the Pearson correlation matrix, showing the degree of correlation between all variables. To avoid multicollinearity problems, the correlation between the variables should not be too high. The table shows that only the correlation between Market_to_Book and the ROA is above 0.5. Therefore, variance inflation factors have been calculated. Theses lies between 1.003 and 1.844 with a mean VIF of 1.336, which indicates that there are no severe multicollinearity issues.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
23
Table 1 - Descriptive statistics
N Mean St.Dev Min Max Q1 Median Q3
Cash_ETR 5282 .228 .135 0.000 .953 .131 .230 .312 Prisk 5282 4.191 1.05 1.952 6.72 3.475 4.224 4.903 CEO_Power 5282 .084 2.605 -3.764 9.454 -1.617 -.623 .983 ROA 5282 .094 .098 -.938 .398 .049 .088 .139 Lev 5282 .185 .166 0.000 .945 .024 .165 .286 Foreign_Inc 5282 .027 .041 -.210 .200 0.000 .012 .046 PPE 5282 .225 .204 0.000 .893 .079 .160 .306 Equity_Inc 5282 .001 .004 -.011 .024 0.000 0.000 0.000 Size 5282 7.736 1.484 3.967 11.289 6.661 7.616 8.728 Market_to_Book 5282 1.518 1.173 .051 7.539 .759 1.179 1.890 R&D 5282 .030 .049 0.000 .576 0.000 .006 .044 Invest 5282 .103 .070 0.000 .601 .059 .087 .131 Intan 5282 .219 .195 0.000 .776 .043 .179 .354
All continuous variables are winsorized at the 1st and 99th percentile. In appendix A all variables are described
Table 2 – Means by year
N Cash_ETR Prisk CEO_Power
2002 117 .219 4.052 1.673 2003 193 .215 4.091 1.643 2004 232 .194 4.051 1.369 2005 238 .237 4.008 1.353 2006 263 .239 4.108 .958 2007 85 .271 3.996 .729 2008 307 .240 4.407 .271 2009 483 .229 4.304 -.471 2010 497 .226 4.201 -.390 2011 598 .219 4.302 -.308 2012 579 .232 4.328 .778 2013 596 .234 4.168 -.586 2014 559 .235 4.128 -.547 2015 535 .219 4.057 -.55
Due the low number of observations in 2016, this year is excluded.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
24
Table 3 - Pearson correlation matrix
Variables Cash_ETR Prisk CEO_Power ROA Lev Foreign_Inc PPE Equity_Inc Size Market_to_Book R&D Invest Intan Cash_ETR 1.000 Prisk -0.056*** 1.000 CEO_Power -0.025* 0.018 1.000 ROA 0.161*** 0.000 0.051*** 1.000 Lev -0.141*** -0.019 -0.079*** -0.238*** 1.000 Foreign_Inc -0.037*** -0.019 -0.089*** 0.306*** -0.126*** 1.000 PPE 0.032** -0.024* 0.002 0.002 0.141*** -0.089*** 1.000 Equity_Inc -0.043*** 0.017 -0.009 0.041*** 0.052*** 0.029** 0.045*** 1.000 Size -0.035** -0.016 -0.150*** -0.052*** 0.343*** 0.204*** 0.123*** 0.130*** 1.000 Market_to_Book 0.009 -0.000 0.111*** 0.573*** -0.311*** 0.176*** -0.095*** -0.064*** -0.221*** 1.000 R&D -0.150*** 0.038*** 0.032** -0.024* -0.264*** 0.154*** -0.270*** -0.091*** -0.188*** 0.294*** 1.000 Invest 0.027* -0.001 0.109*** 0.164*** -0.197*** 0.049*** -0.003 -0.023* -0.086*** 0.238*** 0.091*** 1.000 Intan 0.038*** -0.006 -0.095*** -0.101*** 0.186*** -0.014 -0.380*** -0.047*** 0.170*** -0.102*** -0.022 0.020 1.000
Significance level: *** p<0.01, ** p<0.05, * p<0.1
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
25
4.2 Hypotheses testing
Table 4 is the main table of this thesis and shows the results of the tests of the hypotheses. To test the hypotheses, I used OLS regression with standard errors clustered by company and by year, to control for heteroskedasticity.
The second column in the table show the results of the test of the first hypothesis. The first hypothesis is that there is a positive association between firm-level political uncertainty and the degree of corporate tax avoidance. The model shows that there is a significant relationship between Prisk and Cash_ETR. The coefficient (! = -0.0053; p = 0.002) is negative, which means, in line with hypothesis 1, that firms experiencing higher political uncertainty are associated with a lower Cash_ETR. This model shows that it determines the variation in the Cash_ETR for approximately 9% (adjusted R-square = 0.1385). The results show that if there is more talking about political uncertainty in conference calls, all else being equal, the Cash_ETR of the firm decreases. Therefore hypothesis 1 can be accepted.
Model 3 shows the effect of CEO_Power on the Cash_ETR. This model shows the results of the test of the second hypothesis of the thesis, namely that powerful CEOs engage in higher levels of tax avoidance. The coefficient of CEO_Power (! = -0.0012) shows that companies with powerful CEOs have lower Cash ETRs, the coefficient is marginally significant (p = 0.075). The adjusted R-square for model 3 is 0.1372.
The fourth column shows the model in which the interaction effect of Prisk*CEO_Power on the Cash_ETR is tested. The third hypothesis of this thesis is that the positive association between firm-level political uncertainty and the degree of tax avoidance is strengthened by CEO_Power. The coefficient of the interaction term is negative (! = -0.0002), but it is not significant. Therefore, hypothesis 3 has to be rejected.
In the first column of the table you can see the influence of the control variables on the Cash_ETR. According to the model, a higher ROA results in a higher Cash_ETR (! = 0.2447). The same applies to the amount of PPE a company has (! = 0.0448). A higher Lev results in a lower Cash_ETR (! = -0.1111), all other control variables also show negative coefficients. For example, Foreign_Inc has a coefficient of (! = -0.1992) and R&D (! = -0.3060). Apart from Size, all control variables show significant coefficients. The adjusted R-squared of the control variables model is 0.1370, this means that the variance in Cash_ETR is explained for about 14% by these variables.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
26
Table 4 – Main regression results
Model 1 Model 2 Model 3 Model 4
VARIABLES Cash_ETR Cash_ETR Cash_ETR Cash_ETR
Prisk -0.0053*** -0.0053*** (0.0019) (0.0019) CEO_Power -0.0012* -0.0001 (0.0008) (0.0027) Prisk*CEO_Power -0.0002 (0.0006) ROA 0.2447*** 0.2465*** 0.2448*** 0.2465*** (0.0343) (0.0340) (0.0339) (0.0335) Lev -0.1111*** -0.1117*** -0.1118*** -0.1125*** (0.0190) (0.0188) (0.0190) (0.0189) Foreign_Inc -0.1992** -0.2060*** -0.2038** -0.2103*** (0.0806) (0.0799) (0.0807) (0.0800) PPE 0.0448* 0.0441* 0.0455* 0.0448* (0.0253) (0.0255) (0.0254) (0.0255) Equity_Inc -1.4621** -1.4433** -1.4650** -1.4471** (0.6870) (0.6835) (0.6871) (0.6835) Size -0.0017 -0.0016 -0.0017 -0.0017 (0.0017) (0.0017) (0.0017) (0.0017) Market_to_Book -0.0113*** -0.0114*** -0.0111*** -0.0111*** (0.0025) (0.0026) (0.0025) (0.0026) R&D -0.3060*** -0.3041*** -0.3058*** -0.3040*** (0.0659) (0.0665) (0.0661) (0.0666) Invest -0.0505* -0.0515* -0.0490* -0.0502* (0.0290) (0.0288) (0.0290) (0.0289) Intan 0.0525*** 0.0511*** 0.0516*** 0.0503*** (0.0172) (0.0172) (0.0173) (0.0172) Intercept 0.3821*** 0.4017*** 0.3818*** 0.4016*** (0.0162) (0.0180) (0.0162) (0.0181)
Year fixed-effects YES YES YES YES
Industry fixed-effects YES YES YES YES
Observations 5,282 5,282 5,282 5,282
R-squared 0.1496 0.1512 0.1500 0.1516
Adjusted R-squared 0.1370 0.1385 0.1372 0.1385
Note: Standard errors in parentheses are clustered by year and firm
Significance level: *** p<0.01, ** p<0.05, * p<0.1, whereby the reported p-values for coefficients regarding hypotheses (Prisk, CEO_Power and Prisk*CEO_Power) are showed one-tailed. The p-values of the other variables are showed two-tailed.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
27 4.3 Robustness checks
In order to verify the robustness of the results, I have carried out a number of robustness analyses.
Alternative regression
For the main analysis, I used the ordinary least squares regression in which the standard errors are clustered by firm and year. As a robustness check I used another form of regression, namely the panel data regression. The panel data regression is a more conservative form of regression. By using this form of regression I checked if the results of the main regression still holds. To see if a random or fixed effects model should be used, a Hausman test has been carried out. The test showed that a fixed effects model is the most suitable, and therefore I used a fixed effects model. The results can be seen in model 2 of table 5. The coefficient (! = -0.0045) is still negative although it has decreased slightly compared to table 3 (! = -0.0053). The coefficient remains significant at the level at which the p < 0.01. Based on the results, it can be concluded that the relationship found between Prisk and Cash_ETR holds when another form of regression is used. For hypothesis 2 (model 3), the coefficient in the main regression is negative and marginally significant. The panel data regression shows that the coefficient becomes positive but is no longer significant (! = 0.0007; p = 0.281). Model 4 shows the test for the third hypothesis. For Prisk*CEO_Power a negative coefficient was predicted, and this was also found in the main regression, but the coefficient was not significant. In the model with panel data regression this coefficient is not significant either, so no statistical evidence is found for the third hypothesis.
Alternative tax avoidance measure
Tax avoidance can be measured in several ways. The most direct form is through the Cash_ETR, however, the GAAP_ETR is also often used in research (Hanlon & Heitzman, 2010). It is expected, partly as a result of the study by Duong et al. (2017), that political uncertainty will lead to a lower GAAP_ETR but to a lesser extent compared to the Cash_ETR. The GAAP_ETR is determined by dividing the total tax expense by the pre-tax income minus special items. As with the Cash_ETR, it is truncated between 0 and 1. Model 2 in table 6 shows that a higher Prisk results in a lower GAAP_ETR (! = -0.0026). The coefficient is as expected smaller compared to the Cash_ETR (! = -0.0053; -0.0026). The p-value for the coefficients of the GAAP_ETR and Cash_ETR are both significant at the p < 0.01 level. The relationship found between firm-level political uncertainty and tax avoidance hold if the GAAP_ETR measure is used. For model 3, a negative coefficient has been found in the main regression which is marginally significant (! = -0.0012; p = 0.075). With GAAP_ETR as tax avoidance measure, a negative coefficient is also found, but this coefficient is not significant (! = -0.0004; p = 0.281). Therefore, hypothesis 2 does not hold when GAAP_ETR is used as tax avoidance measure. The expected moderating effect of CEO_Power on the relationship between Prisk and tax avoidance, measured in this model with the GAAP_ETR, does not occur here either.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
28 In the main analysis, political uncertainty was measured with the data of Hassan et al. (2019). They use a firm-level measure of political uncertainty, which is still quite unique. The economic policy uncertainty index (EPU), by Baker et al. (2016), is used more often in uncertainty studies. The EPU index is determined at the national level instead of the firm-level. A high score on the EPU index means that there is a greater degree of uncertainty. A higher score on the EPU index is expected to result in a lower Cash_ETR. Model 2 of table 7 shows that there is a slightly negative relationship between the EPU and Cash_ETR (! = -0.0001) with a marginally significant p-value (p = 0.070) and the adjusted R-squared of this model is 0.1299. Therefore, for hypothesis two it applies that if political uncertainty is measured differently, at the national level, it also leads to a lower Cash_ETR. However, this relationship is less strong, and the coefficient is less significant. Hypothesis three is about the moderating effect of CEO_Power on the relationship between political uncertainty and tax avoidance. In model 3 of table 7 the EPU is used instead of the Prisk variable. The coefficient is negative but really close to zero and, as with previous regressions, the coefficient is not significant (! = -0.0000; p = 0.287). Again, no support is found for the third hypothesis.
Different samples
In table 8 I show the results of regressions using 3 different samples namely before, during and after the financial crisis. Model 1 shows the influence of Prisk on the Cash_ETR. The coefficient in the before sample (2002-2006) is negative and significant (! = -0.0103; p = 0.003). For the 2007-2008 sample, the coefficient is also negative but not significant (! = -0.0019; p = 0.249). This is possible because the sample size is rather small, because little information is available about the year 2007. In the sample after the financial crisis the coefficient is again negative and significant (! = -0.0040; p = 0.042). It can be seen that the coefficient in the before sample is stronger than in the other samples. Model 2 shows the test for the second hypothesis. In the before-crisis sample a positive coefficient is shown, but this one is not significant (! = 0.0005; p = 0.375). In the during-crisis sample it is negative and also not significant (! = -0.0027; p = 0.185). In the after-crisis sample, the coefficient is negative and significant (! = -0.0013; p = 0.043). Model 3 shows that the coefficient of the variable Prisk*CEO_Power is not significant for all three samples.
Kooistra RJH, Rick
Political uncertainty and firm tax avoidance
29
Table 5 – Main models with panel data regression
Model 1 Model 2 Model 3 Model 4
VARIABLES Cash_ETR Cash_ETR Cash_ETR Cash_ETR
Prisk -0.0045*** -0.0046*** (0.0017) (0.0017) CEO_Power 0.0007 -0.0013 (0.0012) (0.0032) Prisk*CEO_Power 0.0005 (0.0007) ROA 0.0973*** 0.0975*** 0.0975*** 0.0978*** (0.0287) (0.0287) (0.0287) (0.0287) Lev -0.0573*** -0.0572*** -0.0575*** -0.0570*** (0.0216) (0.0216) (0.0216) (0.0216) Foreign_Inc -0.3789*** -0.3820*** -0.3788*** -0.3833*** (0.0799) (0.0799) (0.0799) (0.0799) PPE 0.0789* 0.0799* 0.0784* 0.0800* (0.0434) (0.0434) (0.0435) (0.0434) Equity_Inc -3.0563*** -3.0490*** -3.0501*** -3.0438*** (0.7353) (0.7348) (0.7355) (0.7350) Size 0.0298*** 0.0301*** 0.0296*** 0.0299*** (0.0068) (0.0068) (0.0068) (0.0068) Market_to_Book -0.0173*** -0.0173*** -0.0174*** -0.0175*** (0.0031) (0.0031) (0.0031) (0.0031) R&D 0.2292* 0.2322* 0.2288* 0.2336* (0.1329) (0.1328) (0.1329) (0.1329) Invest -0.0337 -0.0356 -0.0334 -0.0347 (0.0354) (0.0354) (0.0354) (0.0354) Intan 0.0484* 0.0480* 0.0480* 0.0476* (0.0269) (0.0269) (0.0269) (0.0269) Intercept -0.0045 0.0119 -0.0016 0.0136 (0.0559) (0.0562) (0.0561) (0.0564)
Year fixed-effects YES YES YES YES
Industry fixed-effects YES YES YES YES
Observations 5,282 5,282 5,282 5,282
R-squared 0.0412 0.0427 0.0412 0.0429
Number of firms 1,055 1,055 1,055 1,055
Note: Standard errors in parentheses are clustered by year and firm
Significance level: *** p<0.01, ** p<0.05, * p<0.1, whereby the reported p-values for coefficients regarding hypotheses (Prisk, CEO_Power and Prisk*CEO_Power) are showed one-tailed. The p-values of the other variables are showed two-tailed.
