CEO CHARACTERISTICS ON LONG-RUN
CORPORATE TAX AVOIDANCE
Abstract: This paper discusses the relationship between characteristics of the chief executive officer (CEO) and long-run corporate tax avoidance. This is in line with the argument of the upper echelons theory which proposes that CEO characteristics are important determinants of organizational outcomes. This study finds that personal characteristics of CEOs are related to the tax avoidance behavior of the company. The results indicate that on average male CEOs are more likely to engage in tax avoidance. Older CEOs are less associated with tax avoidance than their younger counterpart. Longer-tenured CEOs are less engaged in tax avoidance than their shorter-tenured counterpart, and CEOs who are the founder of the company are more engaged in tax avoidance than CEOs who are not the founder of the firm. Based on the analysis of the nationalities of the CEOs, there is evidence that the culture of the CEOs’ nationality is related to tax avoidance. More specifically, CEOs belonging to countries with a higher uncertainty avoidance score, are less involved with avoiding corporate taxes. There is also some evidence that country governance characteristics from the nationality of the CEO are related to tax avoidance. The above results are gathered through the construction of a new long-run corporate tax avoidance model, regressed over the tenure of the CEO. This fact-based research helps researchers understand the relationship between CEO characteristics and corporate tax avoidance practices. Based upon my fact-based findings, I provide some guidelines for future researchers.
Keywords: CEO characteristics, corporate tax avoidance, effective tax rate, upper echelons theory Author: B.J. Oosterveld
Student Number: S3022706 Supervisor: Dr. N. Hussain
Date: 23-01-2017
CEO characteristics on long-run corporate
tax avoidance
Author: B.J. (Bastiaan) Oosterveld
Student Number: S3022706
Date of Birth: 23rd of August 1993
E-mail address: B.j.oosterveld@student.rug.nl
Additional E-mail address: Bj.oosterveld@gmail.com
Address: Dickenslaan 121
Postal Code: 1102 XR, Amsterdam
University: University of Groningen
Faculty: Economics and Business
Study Program: Msc Accountancy & Controlling (track: Accountancy) Supervisor (university): Dr. N. (Nazim) Hussain, PhD
Second supervisor (university): Dr. W.G. (Wilmar) de Munnik Internship organization: Deloitte (Amsterdam) Supervisor (organization): O. (Otto) ter Haar, RA Msc
Preface
Tax avoidance is one of the most current accounting topics, and my personal interest in this topic makes this even more interesting. Some recent tax avoidance examples include Apple and their tax back-payment of 14.5 billion dollars, IKEA and BHP Billiton which are forced to pay back 1 billion euros and 1 billion dollars accordingly, and Zara and McDonalds which are accused of avoiding taxes of around 500 million dollars. These examples require additional knowledge surrounding tax avoidance and the incentives to engage in tax avoidance. Therefore, I was glad to hear that I was allowed to do research in the topic of tax avoidance. By writing this thesis I have gained a lot of insight in the topic of tax avoidance. This thesis also helped me develop my thought process, academic writing skills and skills surrounding the development of databases.
I would like to thank a few people who contributed to the writing process of this thesis. First of all, I would like to thank my supervisor N. (Nazim) Hussain for the ideas he provided, and the feedback and guidance during the process. Furthermore I would like to thank my Deloitte supervisor O. (Otto) ter Haar, M.H. (Miriam) Riefel, and R. (Robert) Oosterveld for reviewing my paper. I would also like to thank S. (Sjoerd) Verhoogt and J. (Joey) Huisman for help regarding the statistical procedures and my database. I would also like to thank A. (Adriaan) Beeftink for his help regarding the gathering of the nationalities of the CEOs. At last, I would like to thank Deloitte (and its colleagues) for providing me this opportunity to write my thesis in their office in Amsterdam.
23-01-2017
X
B.J. Oosterveld Author
Table of contents
Preface ... 1
1. Introduction ... 3
2. Theory and hypotheses ... 6
2.1 Theory ... 6
2.2 Hypotheses development ... 7
3. Data and methodology ... 14
3.1 Collection of data ... 14 3.2 Methodology ... 14 4. Results ... 18 4.1 Descriptive statistics ... 18 4.2 Main results ... 19 4.3 Robustness checks ... 22
5. Conclusion and discussion ... 26
6. References ... 28
7. Appendixes ... 33
7.1 Descriptive statistics ... 33
7.2 Robustness checks ... 35
1. Introduction
“A growing stream of research examines the determinants of firms’ abilities to avoid corporate taxes. The majority of the research in this area examines the effect of firm characteristics on corporate tax avoidance” (Olsen & Stekelberg, 2016, p. 1). Hanlon & Heitzman (2010) suggest however, that a significant gap in the literature represents the effect of managers on tax avoidance. Regarding the upper echelons theory which proposes that personality characteristics of the firms’ top management at least partially influences organizational choices, one can argue that managers can influence the corporate tax strategy. Dyreng, Hanlon, and Maydew (2010) indicate that the level of tax avoidance firms engage in, is highly determined by individual executives. They were, however, unable to find specific personal characteristics of the executives to explain this notion.
In the wake of recent events, tax avoidance has become a popular topic in the accounting literature and international news. Margrethe Vestager, the European Union (EU) competition commissioner on behalf of the European Commission (EC), is trying to promote fair competition (EC, 2016). The EC ordered Apple to pay a record-breaking amount of $14.5 billion in back-taxes to Ireland. According to Margrethe Vestager, Apple received illegal state aid, which caused unfair competition. Apple was able to pay an effective tax rate (ETR) of only 1% in 2003, down to 0.005% in 2014. Apple is not the only company who has recently been charged for their tax avoidance behaviour. Other examples include: IKEA, BHP Billiton, McDonalds, and Google which are convicted to pay back an approximate of $1 billion, and $0.5 billion in back-taxes accordingly (Financial Times, 2016).
Regarding the case of Apple; Ireland does not desire Apple’s tax billions. Ireland is actually appealing the tax ruling because Ireland’s low tax rate is attracting many international companies. Ireland’s economic growth has been realized by these multinationals. Pursuing the tax billions could discourage new conglomerates to start in Ireland. Two contrasting views exist regarding the tax money. According to Mick Barry, a lawmaker from Ireland, the tax money can create many jobs, and improve the entire economy of Ireland. In line with this argument senator Sam Dastyari (2016) states: “Every dollar that is minimized (on taxes) is a dollar that is not going to a school, that is not going to a hospital, and that is not going to a service that we require in need”. Besides the above arguments, the reason that Apple, and Ireland are both appealing the back-tax payment, is that those multinationals provide thousands of jobs, pump in several millions of dollars through rent, income taxes, and help the local economies through expenditure at local restaurants and shops (New York Times, 2016). Regarding the above, this seems like an insolvable puzzle, and more scientific research is inevitable to provide answers.
Sikka (2010) highlights the debate between the significance, and meaning of tax evasion, and tax avoidance. Tax avoidance usually is considered as rightful, whereas tax evasion are the practices that are against the law. The difference in practice is not always clear. Some companies who followed strategies they described as tax avoidance, were scrutinized, and found to be tax evasion by the court. Even though avoidance may be legal, on ethical and moral grounds, these schemes are considerably unacceptable (Christian-Aid., 2008). The decrease in taxes has adverse effects on the foundation of social stability, public goods, and the decrease of poverty. Tax avoidance can therefore be considered as negative corporate social responsibility (CSR), or as a corporate social incident, rather than just mere compliance with laws and regulations (BBC, 2012). In this research, the definition of tax avoidance follows the definition of Dyreng, Hanlon, and Maydew (2008), who state that tax avoidance includes everything that lowers the companies’ taxes, in relation to the pre-tax accounting income. Many firms seem to avoid taxes lately, and this causes public scrutiny. There is an even higher need for research regarding tax avoidance currently (Kallunki et al. 2016).
The literature regarding long-run tax avoidance is still relatively unaddressed, while many researchers describe the advantages of using a longer-run tax avoidance measure in comparison to the currently used yearly measure (Dyreng et al., 2008; Dyreng et al., 2010; Henry & Sansing, 2014; and Law & Mills, 2016). Dyreng et al. (2008) state however: “Annual effective tax rates appear to vary greatly, but have only modest ability to predict long-run avoidance” (p. 63). Dyreng et al. (2008) argue: “a significant fraction of firms appear to be able to successfully avoid large portions of the corporate income tax over sustained periods of time” (p. 79). Dyreng et al. (2010), and Law & Mills (2016) use the yearly cash ETR and the yearly GAAP ETR. The yearly measure of tax avoidance is not the best way to measure tax avoidance, since specific characteristics of a CEO hold over a certain period. A better way to calculate this tax behaviour is to use the long-run measure of tax avoidance, which is described to be superior to the yearly measure of tax avoidance (Dyreng et al., 2008; Henry & Sansing, 2014; and Law & Mills, 2016). This long-run measure takes away the flexibility during the single firm-years. Henry & Sansing (2014) discuss the weaknesses of the annual ETR. They argue that a selection bias occurs, when choosing the annual ETR, because this measure cannot be used for years with negative income. The use of the annual ETR eliminates firm years with negative income, which overstates the tax aggressiveness for the firm. The problem of negative income can be remediated using the long-run cash ETR measure. Kallunki et al. (2016) also state that varying annual ETRs are biased by the varying income, which is also remediated by using the long-run cash ETR method. As a final advantage, taxes paid to the internal revenues service (IRS) and other national tax authorities, are based on payments and refunds that were settled years ago. These taxes are corresponding better to their time-periods when they are measured over longer periods of time.
Law & Mills (2016) investigate managerial characteristics to explain corporate tax aggressiveness. They confirm that previously studied characteristics of managers do not explain specific heterogeneity on corporate tax avoidance. They find however, that managers with military experience approximately pay 1.5$ million more on taxes per year. Kallunki et al. (2016) find a significant association with personal tax avoidance by the CEO and tax avoidance for the entity the CEO manages. Personal tax aggressive CEOs mostly manage corporations that are more aggressive in their corporate tax behaviour. CEOs who file their personal taxes more conservatively are genuinely managing firms that follow more conservative tax policies. Kallunki et al. (2016) also find that firms replace the current CEO with another CEO who has the same risk propensity, education, age and wealth. These results suggest that the personal behaviour of CEOs is directly linked to the outcomes of the firm they are controlling. This paper advances the existing studies of executive effects on corporate tax strategies. Understanding how CEOs’ personal characteristics influence corporate tax avoidance is important. It is an empirical test to confirm the upper echelons theory (Hambrick & Mason, 1984), and thereby objects the inference that the operating environment of the firm, and the firms’ performance, determine the corporate tax avoidance. Moreover, research regarding the tone at the top can have an implication for tax enforcement policies (Law & Mills, 2016). More practically, firms can use this information to help them in the process of selecting a new CEO, to match their desired risk-propensity to personal characteristics of the CEO.
In this research, CEOs’ personal characteristics will be used to try to explain long-run tax avoidance according to the Dyreng et al. (2008) measure. It can be difficult to consider that CEOs have an actual effect on the tax avoidance policy, since a typical CEO is not an expert in taxes (Dyreng et al., 2010). A CEO can guide activities via ‘tone at the top’ influence, since a CEO has most of the power in the firm. Therefore, it seems reasonable to suggest that a CEO can make and change a firm’s tax behaviour, and more explicitly its tax avoidance scheme. This leads to the following research question: ‘Can long-run corporate tax avoidance of a company be explained by their CEO’s personal characteristics?’
This study finds that CEO personal characteristics do influence the tax avoidance behaviour of the organization. This study therefore confirms the upper echelons theory (Hambrick & Mason, 1984), which states that, outcomes of an organization can be determined by the background of the CEO. Finally, this research contributes by adding additional, to the best of my knowledge, still unexplained variables such as the culture scores of the nationality from the CEO. Whether the CEO is also the founder of the firm, and the worldwide governance indicator (WGI) scores related to the nationality of the CEO.
This study is one of the first studies to find an actual relationship between personal CEO variables and tax avoidance. There is evidence that male CEOs are engaging more in tax avoidance than their female counterpart. Older CEOs are engaging less in tax avoidance than younger CEOs. CEOs who are higher-tenured avoid less corporate taxes than their shorter-higher-tenured counterpart. Furthermore, there is evidence that CEOs who are the founder of the firm engage more in tax avoidance, than their non-founding counterpart. Finally, there seems to be evidence that the nationality of the CEOs matter for their tax avoidance behaviour, with a higher uncertainty avoidance culture score (Hofstede, 1980) from the nationality of the CEO leading to less tax avoidance. There is some evidence regarding better country governance leading to more tax avoidance. Since all personal CEO variables are determining the tax avoidance by the CEOs, there seems to be a confirmation of the upper echelons theory as opposed by Hambrick & Mason (1984). This is interesting for future research because this study shows that CEOs, in fact, influence the bottom lines of their organization, even more than may have been expected.
This research continues with the theory and hypotheses stated in section 2, where the upper echelons theory is used to hypothesize a link between a firm’s leader and organizational outcomes. After the theory and hypotheses, the description of the data and the methodology follow in section 3. In section 4 the results are discussed, starting with the descriptive analyses, after which the main results follow, followed with a couple of robustness tests. The conclusion and discussion are presented in section 5. The references can be found in section 6, and final section 7 includes the appendixes.
2. Theory and hypotheses
In this second section, previous literature regarding managerial characteristics are discussed. This literature is driven from the theory that managers influence organizational outcomes. This theory is known as the upper echelons theory, and is more extensively discussed in the theory below. This theory is followed by the hypotheses section, which states the expected directions of the personal CEO characteristics.
2.1 Theory
There are two views regarding leadership influence through organizations. The first view is that leadership barely influences organization. The opposing view is that leaders do influence company performance.
Hall (1977) considers that large organizations form themselves through events and run themselves somehow, and state that managers barely have influence. Another argument can be given from a sociology perspective, as Freedman et al. (1956) state that the impact a single person has on the progress of groups, is not as considerable as we picture it to be. There seems to be evidence that the impact of people are rarely as conclusive as the great man theory would lead to conclude. This great man theory states that ‘great men’ influence the course of history. There are others that consider the role of management to be generally symbolic (Pfeffer & Salancik, 1978). A leaders’ ability to accomplish certain goals is not just reflected by his/her qualities, but also by the limits from an environmental, internal, and social perspective.
One of the first articles that states that leaders do influence company performance, is an article from the American Sociological Review, written by Lieberson & O’Conner (1972). Lieberson & O’Conner (1972) investigate the effect of leadership and organizational performance, and found that leadership differences were relevant for all of their performance criteria. They investigate large corporations to explain how the influence of leadership is associated with company, yearly, and industry influences. Company and industry variables explain more of the variance in sales and earnings. Leadership influences the profit margins of the company more than these company and industry effects. Lieberson & O’Conner (1972) thus provide evidence that leadership characteristics matter for organizational outcomes.
Hambrick & Mason (1984) are considered to be the founders of the upper echelons theory in their seminal work. In their paper, they put emphasis on observable managerial characteristics as an index of the firm’s organizational outcomes. The upper echelons theory states that: “organizational outcomes – strategic choices and performance levels – are partially predicted by managerial background characteristics” (Hambrick & Mason, 1984, p. 193). They provide a sequential view of the decision making of managers (in accordance with Hambrick & Snow, 1977). A manager is not able to scan every part of the company, a manager has a ‘limited field of vision’. A manager also has a selective perception of these phenomena in his field of vision. Finally, a manager interprets the information through a filter from the managers’ cognitive base and values (Hambrick & Mason, 1984). These values and cognitive base determine how the manager responds to a strategic choice. These values, cognitive bases, and the perceptions of managers are hard to measure. But, these values can be corresponding through age, gender, and a managers socio-economic roots, and is regarded as essential for the development of the upper echelons theory.
In their seminal paper they made propositions for the (observable) characteristics, which are: age, education, tenure, functional background, financial position, and the managers’ socio-economic roots. Hambrick & Mason (1984) gave a summary of existing research per observable managerial
tested these propositions as given by Hambrick & Mason (1984). They find that these propositions were supportive. Managerial characteristics did not only predict the variation of performance through industries, but top performing companies seem to have completely different manager profiles than companies that perform poorly. Norburn & Birley (1984) also found that specific observable managerial characteristics are similar for companies that are performing high. Their research increases the confidence to identify top performing companies, and shows how managers’ commercial success can be increased.
More recently, Wang et al. (2016) investigated if CEOs matter for strategic actions of the firm and firm performance. Through their meta-analytical research they found considerable support for the upper echelons theory, with only a few exceptions. They found that CEO characteristics (e.g. positive self-concept, formal education, career experience, and tenure) significantly determine the strategic actions of the firms. These strategic actions are in turn related to the performance of the firm in the future. Moreover, they found that future firm performance is positively influenced by CEO characteristics (e.g. tenure, age, career experience, and education).
Managerial characteristics are found to affect organizational outcomes. In this research this upper echelons theory is used to provide a link between CEO personal characteristics and the tax avoidance firms engage in. These personal characteristics of CEOs are discussed below in the hypotheses development section.
2.2 Hypotheses development
Recent studies show that gender affects performance in various settings (Atkinson, Baird, and Frye, 2003; Kumar, 2009). Francis et al. (2014) have argued that gender differences have been researched extensively in economics, as well as in psychology literature. These studies have found that women are more risk-averse than men. Women are more likely to comply with regulation and rules (Brinig, 1995), and their investment portfolios are found to have less uncertain assets in them (Sundén & Surette, 1998; Jianakoplos & Bernasek, 1998; and Francis et al., 2014). Barua et al. (2010) find that male CEOs are more often engaging in earnings management than female CEOs, and that female CEOs are more often related to more conservative accounting policies (Francis et al., 2015). Women also tend to make less risky decisions in financing and investing (Huang & Kisgen, 2013).
Huseynov & Klamm (2012) state how board diversity impacts corporate outcomes. Gender diversity in the board of directors influences corporate outcomes in different ways. A board of directors with more diverse genders, influences firm performance in a positive way if the firm has weak governance (Adams & Ferreira, 2009). A more diverse board will impact firm performance in a negative way if this firm has strong governance, which is achieved via over-monitoring. Gender diversity in the board also leads to a higher return on assets (ROA) of companies (Carter et al., 2010). Finally, diversity across genders in the management of a company affects the governance and culture of the corresponding organizations (Dwyer, Richard, and Chadwick, 2003).
A couple of explanations exist for the difference in risk-taking behaviour between genders. Croson & Gneezy (2009) have argued that men are more confident than women. In an uncertain environment, men are experiencing less fear and nervousness than woman, and that women see risk more as a threat, whereas men associate risk more with challenges (Francis et al., 2014).
Female managers only represent a small amount of the leaders nowadays. Francis et al. (2014) have investigated gender differences by the chief financial officer (CFO) on their corporate tax aggressiveness. Tax aggressiveness hereby is the most extreme form of tax evasion activities. They focused on a transition from a male to a female CFO, and a female to a male CFO. They have found
overall support that female CFOs are less associated with tax aggressiveness than their male counterpart. Their study considers the gender from the CFO as an important element for tax aggressiveness. Recent studies suggest that choices from females only seem conservative regarding the overconfidence of male CFOs (Huang & Kisgen, 2013; and Barber & Odean, 2001). Woman may not be more risk-averse in general, but men seem to be overconfident. Francis et al. (2014) took this in account when constructing their robustness checks and found that overall risk aversion by female CFOs is the most essential factor behind the different genders on tax aggressiveness.
Unlike Francis et al. (2014), Dyreng et al. (2010) could not find a specific relation between gender differences and tax avoidance. Dyreng et al. (2010) have studied the effects of 521 executives of which 23 were female. The results may have been inconclusive because of the relatively low sample size of their research, or because they used executives in their analyses, and thereby including CFOs and other executives. There may be a stronger result when using just CEOs. Therefore, this study follows the results and direction of Francis et al. (2014).
Bearing in mind that tax avoidance can be considered as a risky activity, and that female CEOs on average tend to be more risk-averse than male CEOs, it can be expected that male managers are more aggressive in their corporate tax avoidance strategy, leading to the first hypothesis:
H1: Male CEOs are more likely to engage in corporate tax avoidance than female CEOs Law & Mills (2016) state that older managers face different risk preferences, different incentives, have different beliefs and cognitive abilities, compared to younger managers. One of the first studies on birth cohort and organizational outcomes was Bertrand & Schoar (2003). They investigated the effect managers have on firm policies. Fixed effects of managers are mostly contributable to the dividend policy, cost-cutting policy, acquisition decisions, interest coverage, and diversification decisions. They stated that managers from older generations are generally more conservative, since older managers have lower levels of investment, and have different leverages compared to younger managers. They also stated that older managers have a different level of capital expenditures compared to younger managers (Bertrand & Schoar, 2003).
Hong, Kubik, and, Solomon (2000) have found that the age of an analyst is related to different incentives. Younger analysts have a forecast more closely related to the average consensus, which indicates that younger analysts can be more risk-averse than older analysts. Hong et al. (2000) have also found that older analysts are providing forecasts on a timelier matter, and they revise their forecasts less often. Age therefore seems to be related with personal performance and organizational outcomes. The research of Yim (2013) relates the acquisition behaviour of CEOs to their age. Yim (2013) has found that older CEOs are less likely to acquire firms, and are less likely to report such an acquisition to their shareholders. These effects are related to risk aversion because older people are assumed to be more risk averse (Yim, 2013). He has emphasized the relevance of corporate decisions related to CEO personal characteristics, and variation in problems regarding the agency theory on a CEO-level. Halek & Eisenhauer (2001) investigate risk aversion based on demographic factors. Demographic variables (e.g. gender, age, and race) affect the extent of risk aversion. Risk-aversion is a popular topic in psychology literature. Through the means of asset allocation decisions Riley & Chow (1992) found risk-aversion to decrease with age, wealth, and education. Until the age of 65 they argue that a persons’ risk-aversion increases. After the 65th year, this risk-aversion starts to decrease. Because tax avoidance can be regarded as a risky activity, one can argue that older managers are less engaging in tax avoidance than younger managers.
Regarding tax avoidance behaviour, Law & Mills (2016) hypothesize that older managers are engaging less in tax avoidance than their younger counterpart, but are unable to find significant evidence to support this notion. Dyreng et al. (2010) were also unable to find enough evidence to support this notion of risk-averseness. Even though no support has been found, regarding the above literature that older managers are more risk-averse, the same hypothesis is tested with a bigger data set. Based upon the review of related literature the following relationship is hypothesized:
H2: Older CEOs are less likely to engage in corporate tax avoidance than younger CEOs Allgood & Farrel (2003) have found that managers with a longer tenure exhibit different corporate policy styles than managers with a shorter tenure. This is consistent with the job match from Jovanovic (1979). He has stated that the amount of risk a CEO takes increases until his tenure reaches the fifth year, after this fifth year it decreases. The risk-propensity thus seems to be the greatest for the first five years a CEO serves. When a CEO has to be replaced, Kallunki et al. (2016) find that this CEO is most often replaced by a person with a similar age and risk-propensity. This follows the notion of Allgood & Farrel (2003) who suggest that the best match of a new CEO arises when the replacing CEO is an insider and has mostly the same characteristics.
Holmström (1999) has studied the relationship between future career concerns and decision making behaviour of managers. He has found that managers with a longer tenure have reduced career concerns. He considers the reputation implications on risk-taking activities by management and finds that risk-preferences are closely related by the manager’s future career concerns. Because CEOs with a higher tenure may have reduced career concerns, longer-tenured CEOs are expected to avoid more taxes, than shorter-tenured CEOs.
Ekelund et al. (2005) investigate self-employment and risk-aversion. They expect that the measure of risk-aversion will be influenced by the amount of self-employment. They predict that individuals with a longer tenure are less risk-averse, because they successfully are self-employed. Tenure therefore would increase the risk appetite for individuals. Simsek (2007) investigate organizational performance through CEO’s tenure. They find that the tenure of a CEO indirectly influences the companies’ performance by having an influence on the top management teams’ (TMT) risk-taking behaviour. He argues that the likelihood to engage in strategic risk and risk-taking change over the course of a CEO’s tenure. When an individual’s experience with handling strategic risk is greater, this individual identifies less uncertainty with respect to the outcomes of engaging in these risks. These risks will thus seem more reasonable to the individual (Simsek, 2007). Experience has the ability to reduce potential losses or the probability that these losses will occur, with regard to strategic risks.
Simsek (2007) argues that shorter-tenured CEOs may have a lowered consciousness to asses and respect these strategic risks. Shorter-tenured CEOs, since they are less known with strategic risks, may be less risk taking than longer-tenured CEOs. These longer-tenured CEOs have more experience, and a better knowledge of the firm and its environment. Longer-tenured CEOs are more familiar with specific risky situations, and are therefore better able to take risks. Because longer-tenured CEOs are more familiar with taking risks, and tax avoidance can be regarded as a risky activity, longer-tenured CEOs are more likely to engage in tax avoidance. This is stated in the following hypothesis:
H3: CEOs with a longer tenure are more likely to engage in corporate tax avoidance than CEOs with a shorter tenure
Anderson & Reeb (2003) have investigated the relation between firm performance and founding-family ownership. Families are involved in one-third of their sample of the S&P 500. Anderson & Reeb (2003) have found that firms run by the family, are performing in a better way than non-family firms. Firms that are founded by the CEO (family firms), are performing better than firms that are not run by the founding CEO (non-family firms). Their study suggests, that the organizational structure of a family-owned company is effective. Founding CEOs have a positive relation on the company’s profitability, and market performance seems to increase as well.
Chen et al. (2010) have expanded the research by looking at the tax aggressiveness of these firms. They have found family firms are less tax aggressive than non-family firms, which is due to the unique agency conflict these family firms have. They have argued that stakeholders might desire tax avoidance, since tax costs are high and this follows extra profitability. What benefits and costs actually arise from tax avoidance, is a topic that still provides no exact solution as of today. The argument of fewer costs ignores certain non-tax costs associated with being more tax aggressive. Chen et al. (2010) have found that family firms are less associated with tax aggressiveness, contrary to the notion they suppose. Family owners might be more concerned with non-tax costs, such as family reputation, and imposed penalties by the IRS.
Wasserman (2003) explains how founding CEOs and entrepreneurial success are linked. Founding CEOs’ success is achieved, because founding CEOs have a higher level of attachment with their firm. Founding CEOs tend to hold more equity holdings of the firm, which increases their control. Founding CEOs also remain in the firm for longer period, because of their attachment. Founding CEOs paradoxically have more success, because they have more success at reaching milestones. Fahlenbrach (2009) finds that 11% of the large US firms are managed by a CEO who is the founder of the firm. He states that founding CEOs have a different risk appetite, compared to succeeding CEOs. There seem to be systematic differences between founding CEOs and successor CEOs. Founding CEOs are performing more acquisitions, and engage more in mergers. Founding CEOs have a bigger amount of capital expenditures, and founding CEOs are having higher research & development (R&D) expenditure. Performing acquisitions, and engaging in mergers are risky activities, where May (1995) finds evidence that founders are bearing more risk capacities, than non-founders. The OECD (2016) provide evidence that firms are lowering their taxes by moving R&D activities to lower tax countries. Since tax avoidance can be seen as a risky activity, and founding CEOs are taking more risk, and are having a higher level of R&D expenditure, founding CEOs may be more likely to engage in tax avoidance. This leads to the following hypothesis:
H4: Founding CEOs are more likely to engage in tax avoidance than CEOs who are not the founder of the firm
Hofstede (1980) created cultural dimensions by which national values are captured. Hofstede (1980) developed four dimensions based on the scores per country. The four dimensions are increased to six dimensions as of today. These six dimensions are: “power distance, individualism, masculinity, uncertainty avoidance, long-term orientation and indulgence” (Hofstede, 2001, p. 29). Han et al. (2010) investigated how culture influences earnings management. Uncertainty avoidance and the individualism score of a culture, are explaining managers’ earnings management through countries. Individualism is positively related with the extent of earnings management, and uncertainty avoidance negatively influences earnings management. Kanagaretnam, Lim, and Lobo (2011) find these association as well, when looking at the earnings quality, but also found a higher power distance score to be related with smoother earnings. The culture of a country can therefore be linked to irresponsible behaviour of managers. National culture seems to influence the decision making of individuals.
Tsakumis, Curatola, and Porcano (2007) match national culture to firm’s tax compliance. They find evidence that a high score of uncertainty avoidance, a low individualism score, a low masculinity score, and a high power distance score are linked to a countries’ non-compliance. Culture is having an impact on the tax-compliance through the countries. Richardson (2008) has built on this research by performing further empirical analysis. They found that a lower individualism score and a higher uncertainty avoidance score are related with more tax evasion. Because the socio-economic background is related with strategic choices of the manager, as described by the upper echelons theory, this research will not focus on the culture of the country, but rather on the culture of the country the CEO is originally from.
Power distance comprehends the equality or inequality of a country. According to Hofstede (1980) in countries with a high power distance score, inequalities are developing, and more powerful people are entitled to special rights. Tsakumis et al. (2007) find that a higher score of power distance is related with more tax evasion. This tax evasion may be due, because an increased level of power distance is related with a higher grade of corruption. Accordingly, tax evasion can be seen as corrupt activity, therefore leading to higher tax evasion. The direction of their hypothesis is followed, namely:
H5a: When a CEO is from a country with a higher power distance score, the CEO is more likely to engage in corporate tax avoidance
In countries with a high score of individualism, the same rules, standards, and procedures should apply to all inhabitants (Hofstede, 1980). Richardson (2008), and Tsakumis et al. (2007) find evidence that a lower individualism score leads to more tax evasion. In countries with a higher individualism score, individuals should feel less motivated to lower their taxes, because they have better regulatory systems in place. In accordance with previous research, this leads to the following hypothesis:
H5b: When a CEO is from a country with a higher individualism score, the CEO is less likely to engage in corporate tax avoidance
In countries with a high score of masculinity, people strive for success, recognition, wealth and ego boosting. The country hereby strengthens the conventional roles of male (Hofstede, 1980). Tsakumis et al. (2007) find that a lower masculinity score is associated with more tax evasion. In countries with a high masculinity score, there is higher focus on punishment. Countries with a lower masculinity score are more lenient, and focus on adjustment. It appears that these more lenient ruled countries are more associated with tax avoidance. Therefore, a CEO from a country with a lower score of masculinity is expected to be associated with more tax avoidance, hypothesized as:
H5c: When a CEO is from a country with a higher masculinity score, the CEO is less likely to engage in corporate tax avoidance
Hofstede’s fourth dimension is uncertainty avoidance. Uncertainty avoidance is based on the tolerance level for uncertainty. When the uncertainty avoidance score is high in a country, their tolerance will be low regarding ambiguity and uncertainty (Hofstede, 1980). Richardson (2008), and Tsakumis et al. (2007) have found that a higher level of uncertainty avoidance is related with a higher level of tax evasion. They argue that persons from a country with a lower uncertainty avoidance score view institutions more as trustworthy. Tax avoidance will therefore be a less viable option. Countries with a high score of uncertainty avoidance should tolerant risker (and corrupt) activities more. Regarding the above argument the hypothesis is as follows:
H5d: When a CEO is from a country with a higher uncertainty avoidance score, the CEO is more likely to engage in corporate tax avoidance
The nationality of the CEO can be used for other insights than culture. The Worldbank Group establishes the dataset of WGI each year. The WGI consist of six dimensions from over 200 countries, and indicate the country governance characteristics. “The WGI provides six dimensions of governance: (The Worldbank Group, 2016)
- Voice and accountability (VA)
- Political stability and absence of violence (PV) - Government effectiveness (GE)
- Regulatory quality (RQ) - Rule of law (RL)
- Control of corruption (CC)”
Kaufman, Kraay, and Mastruzzi (2010) provide the input for these country governance dimensions. They use over 30 data sources to provide input for these six dimensions for the WGI. Governance characteristics are more focussed on compliance (Ottervik, 2013), these measures can provide better results than the culture scores as mentioned above. The six dimensions are shortly explained below: VA captures the perception of which the countries’ citizens are able to select their own government, freedom of speech, and free media. PV captures the likelihood that the government can be overthrown by violence. GE captures the quality of civil and public services, and the exertion and commitment of the government to their strategy. RQ apprehends the governments’ ability to set good policies and regulations. RL captures the confidence in the rules of the society, and the police’s, courts’, and contracts enforcement’s quality and the likelihood of violence. CC captures the degree of which public power is used for personal gains, which includes corruption (Kaufman et al., 2010).
La Porta, Lopez-de-Silanes, and Shleifer (2008) discover that the origin of the laws from a country are related with their regulations and legal rules. They find that the origins of these laws matter for economic outcomes for firms. The origin of the governance backgrounds of managers, can therefore be related to economic outcomes for the firm. Cobham (2005) investigated the effect of the wealth of countries in their relation to tax evasion and avoidance. He found that rich countries are able to obtain more tax revenues than poorer countries. These poorer countries have a lower amount of governance characteristics. A lower score of country governance may therefore be related to more tax avoidance. Hillier et al. (2011) investigate the relation between R&D, and country-level governance. They find that R&D investment is facilitated by appropriate investor protection. In general it seems that the R&D sensitivity in relation with the cash flow, becomes lower by better corporate governance. The OECD (2016) provide evidence that firms can engage in tax-lowering activities by moving their R&D expenditure to places with a low tax rate. Since better governance is associated with less R&D expenses, and R&D expenses are used to avoid taxes, a higher score of country governance can therefore be related to less tax avoidance. Ottervik (2013) measured tax compliance as a measure of state capacity. He found that tax compliance as a state capacity has validity. This can be assumed because tax compliance is highly correlated with the measures of government effectiveness (GE). Since tax compliance is related with good country governance, tax avoidance may be related with a lower country governance score.
To the best of my knowledge no other author matched country governance characteristics (WGI) to tax avoidance. Regarding the above arguments that better country governance characteristics lead to less tax avoidance, the hypothesis is as follows:
H6: When a CEO is from a country with better country governance characteristics, this CEO is less likely to engage in tax avoidance.
All hypotheses above lead to the following conceptual model as noted in Figure 2.1.
Figure 2.1 Conceptual model
Personal characteristics of CEO
H1: Male (+) H2: Age (-) H3: Tenure (+) H4: Founder (+) Nationality of CEO H5a: CulturePD (+) H5b: CultureInd (-) H5c: CultureMas (-) H5d: CultureUA (+) H6: WGI score (-) Long-run corporate tax avoidance
3. Data and methodology
This study aims to find a relationship between personal CEO characteristics and long-run corporate tax avoidance. To find this relationship the Dyreng et al. (2008) measure is followed to accumulate taxes during a multi-year period in order to get a more meaningful ETR. The corporate taxes in this study are calculated over the tenure of the CEO, in comparison to the year-by-year measure of the preceding studies by Law & Mills (2016) and Dyreng et al. (2010). This study therefore differs from the studies of Law & Mills (2016) and Dyreng et al. (2010) in a significant aspect, because there is a lot of supposed variation in year-by-year tax avoidance measures. The sources of the data are discussed below, followed by the methodology used in this study.
3.1 Collection of data
The collection of the data started with the Morgan Stanley Capital International (MSCI) GovernanceMetrics International (GMI) directorships dataset. This dataset is used instead of the more frequently used Execucomp, because Execucomp has an average amount of CEOs covered per year of 1,628, whereas the GMI directorships dataset has an average coverage of 3,113 CEOs per year. To keep the relevance of this research high, the years 2006 till 2015 are used. Any missing CEO data are hand-collected by the use of Google and the Marquis Who’s Who database. The GMI database includes the variables: age, tenure, founder, and the gender of the CEO.
The dataset of Boardex is used to match the CEOs with their corresponding nationality. Not all nationalities are available via Boardex. For the CEOs whose nationality could not be determined using Boardex, the dataset of Beeftink (2013) is used to match the nationalities. Beeftink (2013) hand-collected nationalities of CEOs and I was granted access to his dataset. Regarding the nationalities, the culture scores of Hofstede are collected from his website. These culture scores from Hofstede (1998) do not change over time, as he argues that the culture scores are representing ‘century old roots’. The country governance scores are taken into account in the last hypothesis. It relates to the WGI scores based on the nationalities of the CEOs. These country governance characteristics are gathered via the Worldbank (Worldbank, 2016). The WGI scores change over time, thus these are calculated per year. On these WGI scores a principal component analysis (PCA) is performed to merge it, which is explained more extensively in the methodology section.
The financial figures such as the ETR, long-run ETR and firm characteristics are retrieved from the dataset of Compustat. The analyst’s forecasts are gathered from the dataset of the Institutional Brokers’ Estimate System (IBES), and the governance characteristics are collected from the dataset of Bloomberg, which could be accessed via Deloitte.
3.2 Methodology
The main analyses are performed using the long-run measure of tax avoidance. This long-run measure of tax avoidance is shown in equation 1, and is regressed on the personal characteristics as hypothesized in the second section.
Where equals the firm, is the time period of the tenure of the CEO, and equals the sum of years. This allows a measure to calculate the ETR over the tenure of the CEO. When this long-run ETR is lower, this means that the firm is capable of achieving lower cash taxes paid in comparison to their pre-tax income. A lower ETR is therefore associated with more engagement in corporate tax avoidance.
− ℎ , =∑ (∑ − ℎ ,
The long-run cash ETR as a measure of tax avoidance is used to examine the prediction that long-run tax avoidance can be explained by CEO characteristics. The following personal CEO characteristics are used: male, age, tenure, founder, culture, and the WGI score of the nationality. The WGI scores show a high correlation between the six dimensions of country governance characteristics, as can be found in Table 3.1. Because the correlation between the six dimensions is high, a PCA is performed to merge these six dimensions into one fitting country governance score. This country governance score is used in the regression, instead of the six different measures of governance scores.
Table 3.1
Principal component analysis (PCA): WGI scores
WGI VA WGI PS WGI GE WGI RQ WGI RL WGI CC WGI VA 1.000 WGI PS .606 1.000 WGI GE .839 .576 1.000 WGI RQ .834 .481 .915 1.000 WGI RL .881 .584 .964 .908 1.000 WGI CC .785 .659 .917 .858 .898 1.000
This table shows the correlations between the WGI scores. Because of the high correlations between the six dimensions of the country governance indicators these scores are merged based on a PCA.
Where: WGI VA relates to the Worldbank voice and accountability score. WGI PS relates to the Worldbank political stability and absence of violence/terrorism score. WGI GE relates to the Worldbank government effectiveness score. WGI RQ relates to the Worldbank regulatory quality score. WGI RL relates to the Worldbank rule of law score and WGI CC relates to the Worldbank control of corruption score.
The dependent and independent variables are discussed above and in the hypothesis development. Table 3.2 shows the description of all variables. Regarding the control variables, a broad range of control variables are used in the analyses. The control variables are grouped as firm characteristics, and fixed effects. This broad range of control variables leaves little for CEO characteristics to capture. This conservative approach is taken to increase the robustness of CEO characteristics on long-run tax avoidance. The control variables are discussed below and can be found in Table 3.2. The variables of the robustness checks are also included in Table 3.2.
Firm characteristics are identified in prior literature that are found to affect tax avoidance. These firm characteristics are used in accordance with Law & Mills (2016), Dyreng et al. (2010), and Chen et al. (2010). The firm characteristics are based on the dataset Compustat. The Compustat pneumonic is given in the calculation in Table 3.2. The expected direction of these variables is shown in the analyses and follows the direction they provided. A higher ROA is therefore associated with less tax avoidance. A higher leverage is associated with more tax avoidance. Losses carried forward (the net operating loss (NOL) indicator equals 1) are associated with more tax avoidance. A positive change in losses carried forward is associated with more tax avoidance. Foreign income is associated with less tax avoidance. A higher amount of property plant, and equipment (PPE) is associated with more tax avoidance. A higher amount of intangible assets are associated with less tax avoidance. A higher amount of equity income is related to more tax avoidance. A bigger firm size is associated with more tax avoidance. A higher market-to-book ratio is associated with more tax avoidance, and a higher expenditure on R&D is associated with more tax avoidance as well.
The results are also controlled for industry and year fixed effects. The year fixed effects control for macro-economic changes in the environment of the firm. Industry fixed effects are used to control for industry driven characteristics in tax avoidance. The results are regressed using the primary standard industrial classification (SIC) code as in Bertrand & Schoar (2003) to control for fixed effects, because
Table 3.2 Description of variables
Variable Description Source
Dependent variables Long-run cash
ETR Long-run cash ETR is the sum of cash taxes paid (TXPD), divided by the sum of pre-tax income (PI) minus special items (SPI) over the tenure of the CEO. Truncated at 0 and 1.
Compustat Cash ETR Ratio of cash taxes paid (TXPD) divided by pre-tax income (PI) minus special items
(SPI). Truncated at 0 and 1. Compustat
GAAP ETR Ratio of total tax expense (TXT) divided by pre-tax income (PI) minus special items
(SPI). Truncated at 0 and 1. Compustat
Independent variables
Male Indicator variable which equals 1 when the CEO is a male and equals 0 otherwise. MSCI GMI
Age Age of the CEO in years. MSCI GMI
Tenure Tenure of the CEO in years. MSCI GMI
Founder Dummy variable which equals 1 when the CEO is the founder of the firm and
equals 0 otherwise. MSCI GMI
CultPD Culture power distance measure, based on the nationality of the CEO. Hofstede CultInd Culture individualism measure, based on the nationality of the CEO. Hofstede CultMas Culture masculinity measure, based on the nationality of the CEO. Hofstede CultUA Culture uncertainty avoidance measure, based on the nationality of the CEO. Hofstede WGI Merged Worldbank worldwide governance indicator rank score on the nationality
of CEOs, based on a principal component analysis (PCA). Includes: voice and accountability, political stability and absence of violence/terrorism, government effectiveness, regulatory quality, rule of law, and control of corruption.
Worldbank WGI Control variables: firm characteristics
ROA Return on assets; calculated as operating income (pre-tax income (PI) minus
special items (XI)), divided by lagged total assets (AT). Compustat Leverage Calculated as long term debt (DLTT), divided by lagged total assets (AT). Compustat NOL An indicator variable for a net operating loss (NOL) which equals 1 when the loss
carried forward (TLCF) in year t-1 is positive and equals 0 otherwise. Compustat ChangeNOL Changes in losses carried forward; calculated as the change in loss carried forward
(TLCF), divided by lagged total assets (AT). Compustat ForeignInc Foreign income; calculated as the foreign pre-tax income (PIFO), divided by lagged
total assets (AT). Compustat
PPE Property, plant and equipment; calculated as net PPE (PPENT), divided by lagged
total assets (AT). Compustat
Intangibles Calculated as intangible assets (INTAN), divided by lagged total assets (AT). Compustat EquityInc Calculated as equity income (ESUB), divided by lagged total assets (AT). Compustat FirmSize Calculated as the natural log of the total assets (AT) at the beginning of year t. Compustat MarkettoBook Market-to-book ratio from the beginning of year t; calculated as market
capitalization (PRCC_F*CSHPRI), divided by lagged total assets (AT). Compustat RDIntensity Calculated as R&D expenses (XRD), divided by lagged total assets (AT). Compustat Robustness check: additional control variables
Big4 Indicator variable which equals 1 when the auditor is one of the ‘Big Four’ auditor
firms (EY, KPMG, Deloitte, or PwC) and equals 0 otherwise. Bloomberg Consensus Mean recommendation of analysts following the firm, where: 1=strong sell, 2=sell,
3=hold, 4=buy and 5=strong buy. IBES
ConDeviation Analyst standard deviation determines how much the standard deviation is on
average over the mean recommendation. IBES
NrAnalysts Amount of analysts following the company. IBES BoardSize Amount of board members during year t. Bloomberg CEODuality Indicator variable which equals 1 when the CEO is also the chairman of the board
and equals 0 otherwise. Bloomberg
Regarding the variables above, the full ordinary least squares regression is as follows and is specified below in equation 2:
Where long-run cash ETR is calculated as in equation 1, relates to the firm, is the time period of the tenure of the CEO, is the yearly period, and is the error term of the regression. The variables are defined above in Table 3.2.
The methodology for the robustness checks is now discussed. In the first robustness check the year-by-year ETR is used for comparison with previous works (Dyreng et al., 2010; Law & Mills, 2016). The year-by-year ETR is calculated via the cash ETR and the GAAP ETR which can be found as dependent variables in Table 3.2. The second robustness check calculates the long-run ETR over a broader range of control variables. These control variables are used, because previous research showed a link between those variables and corporate tax avoidance (Sikka, 2010; Hanlon & Heitzman, 2015; Kallunki et al. 2016). These variables are not included in the main analyses, because of missing data and the reduced sample size correspondently. The broader range of control variables is also listed in Table 3.2 as additional control variables. For the third robustness check the CEO must have worked at the company as a CEO for a minimum of three years, to ensure a better measure for long-run corporate tax avoidance. For the fourth robustness check, and in accordance with previous research (Law & Mills, 2016), the financial institutions and utility firms have been removed to check whether the results hold for firms with a more similar way of generating revenues.
Long − run cash ETR , = + , + , + , +
, + , + , + , + , + , + , + , + , + ℎ , + , + , + , + , + , + , + , + ∑ + ∑ + , (Eq. 2)
4. Results
In this section the results from the ordinary least squares regressions are discussed. The section starts with the descriptive statistics of the sample. Following these descriptive statistics, the main results are discussed, after which the robustness checks follow.
4.1 Descriptive statistics
Table 4.1 presents the descriptive statistics for all used variables. The variables for the main results are included, as well as the descriptive statistics for the robustness checks. As mentioned earlier, the sample is based on the years 2006-2015, and includes a maximum of 31,131 firm-year observations. The long-run cash ETR is the cash ETR calculated over the tenure of the CEO and consists of multiple years (unless the CEO serves one year). The mean of this long-run cash ETR is 20.3%, with a median of 19.5%. The cash ETR in this sample has a mean of 19%, and a median of 15.1%, and the GAAP ETR has a mean of 24.2%, and a median of 27%. The cash ETR is lower than the calculated GAAP ETR, which is consistent with previous research (Dyreng et al., 2008; Dyreng et al., 2010; and Law & Mills, 2016). Most often firms are more capable of realizing a lower taxable income than they are able of realizing a lower pre-tax accounting income. As stated by Dyreng et al. (2010), the volatility of the Cash ETR is higher than the GAAP ETR, which makes results more difficult to document. This volatility is lower for the long-run cash ETR, as can be subtracted from the lower standard deviation (Std.). The observation size (N) changes over the variables, because of the missing observations from using multiple databases. The main results therefore only use the control variables of firm characteristics, because of the lower sample size, while in the second robustness check all control variables are used.
In this sample a typical CEO is a 55-year-old American male with approximately eight years of experience as a CEO and this CEO is not the founder of the firm. Only a small fraction (10.4%) of the CEOs are also the founder of the firm.
To control for the fixed effects, the regression includes the industry code based on the SIC code. The SIC code is a four-digit code. The first digit of the SIC code is used in order to regress the main divisions of the industry code classification. These main SIC code divisions are shown in Figure 4.1 in Appendix 7.1. The most common industry in the sample is the manufacturing industry, which includes 32.71% of the sample. Finance is the second biggest industry after which the services industry follows. In the robustness checks the sample will exclude the financial institutions (6000-6999), and firms in the utilities sector (4900-4999).
The nationalities and the culture scores of the CEOs are shown in Table 4.2 in Appendix 7.1. Nearly all firms in the sample from GMI are incorporated in the United States (US). 92.9% of the known nationalities of CEOs in the sample are therefore from the US. For 45.3% of the CEOs, the nationality could not be determined through the existing datasets.
Table 4.1 Descriptive statistics
Dependent variables N Mean Std. P25 P50 P75
Long-run cash ETR 25,582 0.203 0.195 0.016 0.195 0.307
Cash ETR 27,122 0.190 0.210 0.000 0.151 0.303 GAAP ETR 27,122 0.242 0.204 0.025 0.270 0.355 Independent variables Male 31,131 0.968 0.177 1 1 1 Age 30,778 55.188 7.694 50 55 60 Tenure 30,822 8.145 7.709 3 6 11 Founder 31,131 0.104 0.305 0 0 0 CultPD 16,983 40.203 4.001 40 40 40 CultInd 16,983 89.688 6.967 91 91 91 CultMas 16,983 61.506 4.378 62 62 62 CultUA 16,983 46.277 4.247 46 46 46 WGI 17,030 0 2.225 -0.207 0.110 0.480
Control variables: firm characteristics
ROA 26,693 0.025 0.368 0.001 0.043 0.107 Leverage 26,693 0.233 1.310 0.013 0.151 0.339 NOL 26,693 0.484 0.500 0 0 1 ChangeNOL 26,693 0.043 1.102 0 0 0.003 ForeignInc 26,693 0.013 0.053 0 0 0.014 PPE 26,693 0.238 0.292 0.030 0.125 0.345 Intangibles 26,693 0.185 0.287 0.007 0.073 0.280 EquityInc 26,693 0.001 0.009 0 0 0 FirmSize 26,693 7.310 1.841 5.990 7.234 8.449 MarkettoBook 26,693 1.560 3.878 0.402 0.895 1.733 RDIntensity 26,693 0.046 0.169 0 0 0.029
Robustness check: additional control variables
Big4 20,051 0.852 0.355 1 1 1 Consensus 22,550 2.308 0.505 1.977 2.298 2.653 ConDeviation 22,550 0.744 0.300 0.624 0.799 0.924 NrAnalysts 22,550 9.823 7.336 4.333 7.833 13.667 BoardSize 19,975 9.199 2.490 7 9 11 CEODuality 19,949 0.451 0.498 0 0 1 IndDirectors 18,684 0.788 0.121 0.714 0.833 0.889
This table describes the variables used in the analyses. N equals the amount of observations. Mean relates to the mean value of the variable. Std. is the standard deviation of the mean. P25, P50, and P75 relate to the 1st,
2nd, and 3rd quartile consecutively.
For the description of the variables see Table 3.2.
4.2 Main results
The results are separated in two parts: the first two columns of Table 4.3 relate to the regression excluding the nationalities, whereas the last two columns include the nationalities of the CEO. This is due to the lower sample size caused by the missing nationalities of CEOs. After the regression, the analysis is discussed.
Table 4.3 Main results
Dependent variables:
Pred. Long-run cash ETR Long-run cash ETR
Independent variables (1) (2) (3) (4) Male (-) -.043*** -.048*** -.057*** -.063*** Age (+) .001*** .001*** .001*** .001*** Tenure (-) .002*** .002*** .002*** .001*** Founder (-) -.061*** -.041*** -.055*** -.037*** CultPD (-) .000 .000 CultInd (+) -.000 .000 CultMas (+) .000 .000 CultUA (-) .001*** .002*** WGI (+) -.001 -.002* ROA (+) .036*** .189*** Leverage (-) -.005*** -.105*** NOL (-) -.023*** -.008** ChangeNOL (-) -.001 -.001 ForeignInc (+) .175*** -.082** PPE (-) -.031*** -.014** Intangibles (+) .060*** .063*** EquityInc (+) .502*** .564*** FirmSize (-) .001* -.005*** MarkettoBook (-) .001 -.003*** RDintensity (-) -.131*** -.263***
Constant Yes Yes Yes Yes
Year fixed effects No Yes No Yes
Industry fixed effects No Yes No Yes
Nationalities included No No Yes Yes
Nr. of observations 25,271 24,864 14,930 14,641
Adjusted-R2 .016 .062 .016 .100
Highest VIF 1.365 3.173 2.939 3.047
*, **, *** Represent the statistical significance at the 10%, 5% and 1% level, respectively. Pred. is the predicted direction of the variable. Column (1) reports the regression, using the main personal CEO variables. Column (2) reports the regression using the main personal CEO variables, and the control variables. Column (3) reports the regression of all personal CEO variables, including nationalities. Column (4) reports the regressing using all personal CEO variables, and the control variables.
For the description of the variables see Table 3.2.
Table 4.3 above shows that male CEOs are consistently associated with a significant lower ETR. Even though the male gender takes up 96.8% of the positions of CEOs in this sample, male CEOs are reporting lower ETRs than their female counterpart across all columns of the table. This is in line with hypothesis 1, which suggests that male CEOs are more risk-taking than their female counterpart. This provides enough support to accept hypothesis 1; male CEOs are more engaged in corporate tax avoidance than female CEOs. When a CEO is male, the long-run ETR on average is 5.3% lower through the 4 columns than their female counterpart. Regarding the average ETR of 20.3%, this lower ETR shows that male CEOs are averagely paying 26% less on taxes than female CEOs.
Age is consistently associated with a higher ETR, suggesting that older managers are engaging less in tax avoidance activities. These results are consistent across all columns of Table 4.3. This provides support for hypothesis 2; older CEOs are less engaging in corporate tax avoidance than their younger CEO counterparts. Older CEOs can be considered as more risk-averse than younger CEOs. Each year increase in the age of the CEO is associated with an increase in the ETR of 0.1%. An age difference of 20 years between two CEOs should lead on average to the older CEO paying a 2.0% higher ETR, compared to the younger CEO.
The tenure of a CEO is also significantly associated with a higher ETR. Based on the results, it can be concluded that CEOs working longer for the firm as a CEO, are engaging less in tax avoidance than their lower-tenured CEO counterpart. This is contrary to hypothesis 3, suggesting that higher-tenured CEOs have an increased level of corporate tax avoidance. Each year increase in the tenure of the CEO is associated with an average increase in the ETR of 0.175% through the 4 columns of Table 4.3. On average, a newly appointed CEO of a company is engaging more in tax avoidance activities, than a CEO who already manages a company for a longer time-period. This provides evidence that CEOs take more risk in the beginning of their tenure as stated by Allgood & Farrel (2003). This opposes the thought of Simsek (2007) who stated that more experienced CEOs dare to take more risk, based on their experience. Hypothesis 3 is therefore rejected.
When the CEO is the founder of the firm, this is consistently associated with a significant lower ETR. This suggests that founding CEOs are engaging more in corporate tax avoidance. This is in accordance with hypothesis 4, stating that founding CEOs are less engaged with tax avoidance. When the CEO is the founder of the firm, the long-run ETR on average is 4.85% lower through the 4 columns than their non-founding counterpart. Regarding the mean ETR of 20.3%, founding CEOs are paying an average of 23.9% less on taxes than non-founders.
The culture scores from the CEOs seem to be less associated with tax avoidance; power distance, individualism, and the masculinity score show no significant association with the long-run cash ETR. The culture score of uncertainty avoidance is significantly associated with a higher ETR, which is contrary to hypothesis 5d, which states that CEOs from a culture with a higher score of uncertainty avoidance engage less in corporate tax avoidance. This is contrary to the findings of Richardson (2008), and Tsakumis et al. (2007). They studied the effects of the location of the firm’s culture on tax evasion, whereas in this research the nationality of the CEO influences the regression. A difference of influence in tax avoidance behaviour can lie between the culture of a CEO’s birth location and the national culture based on the location of the firm.
The WGI score shows some significance in the last column. There seems to be some more significance of the country governance indicators of the nationalities of CEOs, in the robustness checks below. Despite these robustness checks, there is deficient significance to confirm hypothesis 6. The direction is also contrary to the supposed notion. There is some evidence that better governance leads to more tax avoidance. The lack of diversity in the sample can be an explanation for the non-significance of the first three cultural dimensions and the WGI score. Future research could provide interesting results, when European firms are investigated on their tax avoidance behaviour of CEOs. This may give more significant results across other dimensions than uncertainty avoidance.
The adjusted-R2 is higher than the adjusted-R2 from the first robustness check with the yearly cash ETR. The variables are therefore explaining the variation in the long-run ETR to a greater extent, in comparison to the yearly cash ETR. Regarding the variance influence factor (VIF), the problem of multicollinearity is not an issue for the main results, or any of the robustness checks.
4.3 Robustness checks
Four measures of robustness checks are identified and discussed below. The robustness checks discussed are: the year-by-year ETR, a robustness check with additional control variables, a robustness check where the minimum tenure of the CEO is 3 to increase the long-term orientated view of the results, and a final robustness check without financial institutions and utility firms.
4.3.1 Robustness check 1: year-by-year ETR
The first robustness check follows previous research (Law & Mills, 2016; Dyreng et al., 2010; and Christensen et al., 2015). The yearly cash ETR and GAAP ETR are used in this robustness check, instead
Table 4.4
Robustness check 1: yearly cash ETR Dependent variables:
Pred. Cash ETR Cash ETR
Independent variables (1) (2) (3) (4) Male (-) -.023*** -.027*** -.016* -.023*** Age (+) .001*** .001*** .000 .001*** Tenure (-) .002*** .002*** .002*** .001*** Founder (-) -.048*** -.029*** -.040*** -.018*** CultPD (-) .000 -.001 CultInd (+) .000 .000 CultMas (+) .000 -.001 CultUA (-) .000 .001** WGI (+) -.001 -.003** ROA (+) .048*** .208*** Leverage (-) -.005*** -.103*** NOL (-) -.025*** -.011*** ChangeNOL (-) -.001 -.001 ForeignInc (+) .215*** -.010 PPE (-) -.027*** -.010 Intangibles (+) .068*** .074*** EquityInc (+) -.064 -.391** FirmSize (-) .003*** -.001 MarkettoBook (-) -.000 -.003*** RDintensity (-) -.082*** -.227***
Constant Yes Yes Yes Yes
Year fixed effects No Yes No Yes
Industry fixed effects No Yes No Yes
Nationalities included No No Yes Yes
Nr. of observations 26,798 26,369 15,756 15,453
Adjusted-R2 .009 .055 .006 .087
Highest VIF 1.370 3,073 2.946 3.063
*, **, *** Represent the statistical significance at the 10%, 5% and 1% level, respectively. Pred. is the predicted direction of the variable. Column (1) reports the regression, using the main personal CEO variables. Column (2) reports the regression using the main personal CEO variables, and the control variables. Column (3) reports the regression of all personal CEO variables, including nationalities. Column (4) reports the regressing using all personal CEO variables, and the control variables.
