The effect of board diversity on tax aggressiveness in Europe
MSc Thesis S.J. Kortstee1
MSc International Financial Management
University of Groningen- Faculty of Economics and Business Supervisor: Dr. M. Hernandez Tinoco2
Co-assessor: Dr. H. Vrolijk January 8, 2016
Abstract
This paper studies the effect of the corporate governance characteristics of board gender diversity and board independence on tax aggressiveness for mainland European firms. A sample of 254 firms over an eight year period (2007-2014) is used. Prior research has so far not analyzed firms from stakeholder-oriented cultures, which are hypothesized to install the board of directors in such a manner so that tax aggressiveness is reduced. Fixed effects tests and a system Generalized Method of Moments (GMM) estimator are applied to deal with potential bias as a result of endogeneity. The findings provide no support for a decrease in tax aggressiveness as a result of increasing the percentage of females or independent directors in the board. The results are supported by several robustness tests. The main implication involves that appointing more female and independent board members in corporate boards is not expected to lead to a decrease in a firm’s tax aggressiveness. Key words: tax aggressiveness, corporate governance, board diversity, board gender diversity, board independence
JEL classification: G34, K34, M14
1 Address: Gedempte Zuiderdiep 148A, 9711HN Groningen, The Netherlands. E-mail: s.j.kortstee@student.rug.nl. Student number: s2028824
2
2 Table of content
1. Introduction ... 3
2. Literature review and hypothesis development ... 6
2.1. Literature review ... 6
2.1.1. Tax aggressiveness ... 6
2.1.2. Stakeholder and legitimacy theory ... 7
2.1.3. Shareholder theory... 8
2.1.4. Agency theory ... 9
2.1.5. Corporate governance ... 9
2.2. Hypothesis development ... 10
2.2.1. Board Gender Diversity ... 10
2.2.2. Board independence ... 12
3. Method... 14
3.1. Sample and data ... 14
3.2. Dependent variable ... 15 3.3. Independent variables ... 17 3.4. Control variables ... 18 3.5. Descriptive statistics ... 19 3.6. Models ... 20 4. Results ... 22
4.1. Results hypotheses testing ... 22
4.2. Endogeneity ... 24
4.3. Robustness ... 29
5. Conclusion ... 29
References ... 32
3 1. Introduction
During the past few years there has been an increased focus on the actions of multinational enterprises (MNEs) involving the avoidance of taxes. While normal citizens transfer a considerable amount of their wages to the relevant tax authorities, it is the large corporations who prove to be able to reduce their tax obligations to a minimum. Microsoft, Google and Facebook, for instance, are accused of not paying their fair share of taxes in multiple countries in which they operate by using transfer pricing schemes to tax profits in countries with beneficial corporate tax rates. It is estimated that this income shifting costs the US government as much as $60 billion annually (Drucker, 2010). Moreover, the Organization for Economic Co-operation and Development (OECD) has estimated that this Base Erosion Profit Shifting (BEPS) leads to non-collected taxations of 4-10% of global corporate income tax revenues, amounting between 100 and 240 billion US$3.
In order to put an end to this loss of tax revenues, the tax authorities of Canada, Australia, the UK and the US set up the Joint International Tax Shelter Information Center in 2004 (Landolf, 2006). More recently, the OECD has collaborated in order to introduce the BEPS action plan. This plan involves Country-by-Country (CbC) reporting for restricting transfer pricing practices (OECD, 2015). Hence, the tax management activities of the major MNEs have been subject to extensive public outcry although being completely legal. Protesters, on the one hand, argue that these corporations should pay their fair share of taxes in the companies they operate in, as it is these societies that enable them to do business in the first place (Barford and Holt, 2013). Firms, on the other hand, start to feel uncomfortable as the OECD plan limits their abilities to design operations in the most efficient manner (Broekhuizen, 2014). These companies argue that looking for the most efficient manner to do business internationally is the raison d’être of the MNE.
However, for the MNEs it may prove difficult to devise a strategy which brings together different objectives of the firm, and this will depend on which party it is the corporation bears responsibility to. Prior literature stresses the importance of including tax management in a company’s Corporate Social Responsibility (CSR) policy (e.g. Landolf, 2006; Huseynov & Klamm, 2012). The payment of a fair share of taxes is valued by the public on the one hand, as it enables governments to fund public facilities such as education and health care. In respect to this view, following an aggressive tax strategy would be seen as immoral towards society. For MNEs, on the other hand, paying the highest tax rates may be disadvantageous vis-à-vis competitors as it can involve considerable costs and thus directly affects the bottom line (Landolf, 2006). The question therefore is when a company is considered to be engaging in tax aggressive behavior and what exactly is considered a fair share of taxes.
Prior research has to a great extent already focused on the tax strategies of firms and their consequences. Hanlon and Slemrod (2009), for instance, find that stock markets react unfavorably
4 towards news announcements in which a firm is accused of engaging in tax sheltering activities. Moreover, stock market reactions may be different for firms in different industries. Slemrod (2004) finds that a firm’s tax aggressive behavior may lead to irrecoverable loss for society, as firms incur costs while trying to look for the most efficient tax management strategies and governments have to make costs in order to develop regulation that refrains these firms from avoiding taxes. Landolf (2006) argues that firms not paying the fair share of taxes limit the government in its main task, which is to meet the needs of citizens as much as possible. The relation between a firm’s corporate governance (CG) characteristics and its tax aggressiveness, however, has been researched considerably less. Lanis and Richardson (2011) studied the relation between Board of Directors (BOD) composition and tax aggressiveness for 16 considered tax aggressive and 16 considered non-tax aggressive Australian firms. They found that a higher percentage of independent board members reduces the probability of tax aggressive behavior. Moreover, Lanis and Richardson (2012) research tax aggressiveness as part of CSR policies, and find that more socially responsible firms show less tax aggressive behavior. Furthermore, Minnick and Noga (2010) found that an increase in BOD independence and size have a negative effect on tax aggressiveness in US S&P 500 firms. Khaoula and Ali (2012) research the role of women in the BOD and the effect on tax aggressiveness in US firms, but find no support for this.
5 appoint a certain number of independent board members (Higgs, 2003). These initiatives have the purpose of reducing the likelihood of opportunistic behavior by management of large corporations. An indirect result of these actions may be lower corporate tax aggressiveness, as board of diversity effectiveness is augmented as a result of an increase in both variables (Lanis and Richardson, 2011; Adams and Ferreira, 2009). This increased effectiveness of the board of directors should enable the board to make decisions that better protect the interests of the shareholder of the firm (Fama and Jensen, 1983).
The study is performed using a panel of 254 European firms included in the STOXX 600 Europe ex UK index and annual firm data over an eight year period (2007-2014). Regarding the characteristics of the data and potential endogeneity issues in CG research (e.g. Wintoki, Linck and Netter, 2012; Minnick and Noga, 2010; Francis et al.,2014), this paper relies on 1) fixed effects testing and 2) the system Generalized Method of Moments (GMM) estimator, which is designed to control for endogeneity issues among others. Endogeneity may lead to biased results due to unobserved heterogeneity, simultaneity or dynamic relations between the variables (Wintoki et al., 2012)
The main contribution of the paper is twofold. Firstly, no significant results were found to support a relation between tax aggressiveness and board independence in Europe. A negative relation between board independence and tax aggressiveness was expected. Contrary, when controlling for endogeneity, a positive effect of board gender diversity on a firm’s tax aggressiveness is found. Again, this is different than hypothesized, as a negative relation between board gender diversity and tax aggressiveness was expected. Secondly, the results suggest that endogeneity does not lead to biased estimates in the relation of CG and tax aggressiveness for European firms. The findings, therefore, suggest differences in the tax aggressiveness of European firms vis-à-vis their US and Australian counterparts.
Implications for management, therefore, may involve that the selection of women in the BOD does not necessarily lead to a lower tax aggressiveness, and critical selection of BOD members may thus be required. Furthermore, increasing the number of independent board members does not lead to a lower tax aggressiveness among European firms.
6 2. Literature review and hypothesis development
2.1. Literature review 2.1.1. Tax aggressiveness
As Hanlon and Heitzman (2010) mention in their paper, the literature on tax aggressiveness is young but very active and discusses the activities of firms in the design of their taxation strategies. Their paper states that the area has not developed a universally accepted definition of tax aggressiveness or tax avoidance so far, which leads to the terms meaning different things to different people. Moreover, several terms indicating activities performed in order to reduce a company’s tax bill are used interchangeably, such as tax avoidance, tax evasion, tax aggressiveness, tax management and tax sheltering. The definition one uses seems to depend on his or her personal values towards the active management of taxes. What is considered tax evasion by one person, may easily be referred to as noncompliance by another (Hanlon & Heitzman, 2010).
Prior literature, as stated above, is characterized by mixed definitions of tax aggressiveness. Shackelford and Shevlin (2001) bring forward the need for further research in order to discover what it actually is that determines tax aggressiveness. Lanis and Richardson (2011), for instance, define TA as “a scheme or arrangement put in place with the sole purpose or dominant purpose of avoiding tax” (p.50) and mention examples such as “the shifting of income or profits to offshore tax havens and the excessive claiming of tax deductions (e.g. interest and R&D expenses) and tax losses that the corporation is not entitled to receive” (p.50). Furthermore, the OECD describes tax planning as “strategies that exploit the gaps and mismatches in tax rules to artificially shift profits to low or no-tax locations where there is little or no economic activity, resulting in little or no overall corporate tax being paid”4.
In the rest of this paper, however, a more broad definition for tax aggressiveness is used. Hanlon & Heitzman (2010) counter the lack of a universal definition for tax aggressiveness by using a broader term to describe this phenomenon. The authors find that a definition for tax aggressiveness should not be bound to activities solely designed for the purpose of avoiding taxes, as different persons may consider different activities to be an example of tax aggressiveness. Hence, the authors refer to tax aggressiveness as the reduction of explicit taxes. By doing so, the authors do not distinguish between activities that are tax favored (such as municipal bond investments), avoidance activities undertaken with the goal to reduce taxes, and targeted tax benefits from lobbying activities. In order to avoid indistinctness as a result of using synonyms of tax aggressiveness interchangeably, the term tax aggressiveness (hereafter: TA) will solely be used to indicate a company’s activities in the pursuit of a lower tax bill throughout the remainder of this paper, which is in line with Hanlon and Heitzman (2010).
4
7
2.1.2. Stakeholder and legitimacy theory
TA may have significant consequences for a firm and should therefore be an important focus point for the firm’s management in designing the firm’s strategy. Huseynov and Klamm (2012) research the relation between CSR and TA, and state that a firm’s tax management should be an integral part of a company’s CSR strategy. TA is closely related to the theories involving the stakeholders and the legitimacy of the firm. Freeman (1984) introduced the idea of the stakeholder theory, stating that the goal of the firm is to create shareholder value but at the same time protect the interests of the firm’s stakeholders. Stakeholders can be defined as entities who may affect, or may be affected by, a firm’s actions. Legitimacy theory suggests that a firm enters into a social contract with society, in which the rights and obligations of society towards the firm are established (Deegan, 2002). As TA involves the reduction of explicit taxes, a firm risks losing its license to operate by paying low rates of taxes. The different constituents of society enable a firm to do business, as the firm is dependent on all sorts of stakeholders such as employees, political groups, customers, communities and suppliers, among others. All of these stakeholders have different levels of power and influence over the firm, and the firm can gain legitimacy by operating in accordance with what is expected of it by these members of society (Roberts, 1992; Deegan, 2002). These stakeholders have the possibility to impede a firm’s operations, and willingness to do so may develop from the fact that the payment of taxes is an important source of income for governments. The government distributes these tax payments to society, and these tax revenues can be used for the funding of public facilities such as education, infrastructure and healthcare (Friese, Link and Mayer, 2008).
Therefore, from a stakeholder and legitimacy perspective, paying a fair share of taxes and maintaining a good relation with the tax authorities would be the right thing to do as the company may benefit from investment in these facilities in the long run (Landolf, 2006). Additionally, failing to do so may in the worst case lead to hostility from society towards the corporation. This could lead to reputational damage among the firm’s stakeholders, or even the cessation of the firm’s operations in the long run (Erle, 2008).
8
2.1.3. Shareholder theory
Opposed to the stakeholder and legitimacy theory is the shareholder theory, which was introduced by Friedman (1970). He argues that the sole purpose of the business is to maximize the value of the firm for its owners, the shareholders. Any activities performed without the goal of value maximization divert potential profit away from the owners into the hands of other stakeholders, such as the government, customers or employees. In the opinion of Friedman (1970) it is not the responsibility of the business to behave in a socially responsible manner unless this leads to shareholder value creation in the future.
Hence, from Friedman’s point of view, pursuing a strategy of reducing a firm’s tax obligations is justified as long as the marginal benefits of such a strategy outweigh the marginal costs of implementing it (Slemrod, 2004). Marginal benefits are the increased cash flows resulting from transferring smaller amounts to the tax services. Marginal costs associated with pursuing a low tax bill are potential tax fines imposed by the tax administration, implementation costs (e.g. time, effort, transaction costs), reputational costs and political costs (Slemrod, 2004; Hanlon & Slemrod, 2009; Chen et al., 2010).
Ball et al. (2000) introduced the idea of ranking countries as either a shareholder or stakeholder oriented country. According to the authors, these classifications may have significant effects on many areas of business in the respective country. Ball et al. (2000) differentiate between countries whose law system is dictated by either common law or code law. In code law countries the government is the main regulative body, whereas in common law countries regulations for businesses is primarily constituted by private bodies. Furthermore, the authors suggest that in code law countries the government makes up rules and regulations together with involved parties such as labor unions, banks and business associations. The interaction with these stakeholders in the setup of guidelines leads to a stakeholder governance model at the firm level, in which agents of the major stakeholder groups of the firm are represented. Contrary, in common law countries guidelines are mainly set up in the private sector and therefore common law countries are characterized by a more shareholder oriented view in the establishment of corporate governance practices within a firm. Ball et al. (2000) conclude that the shareholder oriented countries seem to rely on common law and the stakeholder oriented countries on code law, roughly meaning that Anglo-Saxon countries focus on the shareholder and other countries, such as most countries in Europe, focus on the stakeholder.
9 stakeholders.
For instance, Khaoula and Ali (2012) have researched the effect of higher female presence in the BOD on corporate TA in the US. The authors did not find support for this relation. However, regarding the differences in firm goal orientation between US and European mainland discussed above, results for this relation may be different in the European context.
2.1.4. Agency theory
The ideas of Friedman (1970) led to the development of the agency theory, which discusses the principal-agent relationship that exists between shareholders and managers. Managers are appointed by shareholders to run the company and should therefore act in the best interest of the shareholders (Fama and Jensen, 1983). This is a potential source of conflict of interest, as risk-neutral investors expect managers to act in their interest and, hence, to focus on profit maximization. The question is whether or not tax aggressive policies are value enhancing, as the benefit is only realized in the long run and is quite uncertain (Rego & Wilson, 2009).
As long as there is separation of ownership and control there are opportunities for managers to transfer wealth from the shareholders to themselves. This may influence a company’s TA, as higher reported income may result in higher compensation for management as a result of profit sharing plans and bonuses. Jensen and Meckling (1976) introduced the principle that shareholders and the BOD should try to find the right combination reflecting a desired tradeoff of incentive and agency costs. Thus, by agreeing on a certain compensation for management, shareholders decide to what degree TA is considered desirable.
2.1.5. Corporate governance
CG systems are installed with the main goal of protecting the interests of corporate stakeholders from opportunistic behavior by management. Since the managers of the company control the key decisions of the corporation, these stakeholders have an interest in determining how management will protect their interests. CG, therefore, deals with the manner in which the stakeholders control management. By doing so, it separates ownership from control and minimizes agency costs (John & Senbet, 1998). John and Senbet (1998) further bring forward that the BOD is a major determinant in CG mechanisms, as it is viewed as “a primary means for shareholders to exercise control on top management”(p. 379).
10 most important decisions on their behalf (Grossman and Hart, 1980).
The BOD is able to hire, fire and compensate top level managers within the corporation, making it one of the most influential instances within the corporation (Fama & Jensen, 1983). Most prior research focuses on the ability of the BOD to control management on behalf of the shareholders, but increasing focus is put on the role of the BOD in protecting the interests of the other stakeholders of the organization (John & Senbet, 1998; Rose, 2007).
As the BOD has significant controlling power over a corporation’s top management, it also has considerable power in determining the tax strategy of the respective firm. An example of such a strategy would be the creation of a tax management department, which has the sole purpose of reducing a company’s tax bill. Top management is responsible for an efficient allocation of resources and the pursuit of the highest possible profit by improving company performance. The role of the BOD vis-à-vis top management of the corporation gives it a controlling interest in the decisions of top management.
As indicated by prior research, board effectiveness may affect firm performance (John and Senbet, 1998). A BOD’s effectiveness considers the degree to which the BOD is able to oversee and control management (Fama, 1980). Board effectiveness may, in turn, be affected by board composition. This research focuses on the effects of BOD diversity by taking into account both gender and independence of BOD members. Both diversity indicators are found to affect the effectiveness of BODs (e.g. Kang, Cheng and Gray, 2007; Adams and Ferreira, 2009; Lanis and Richardson, 2011). Since board effectiveness affects firm performance, it may also be important in the determination of a firm’s tax strategy. Next, both diversity indicators and their potential consequences are discussed. 2.2. Hypothesis development
2.2.1. Board Gender Diversity
Adams and Ferreira (2009) suggest that board gender diversity (BGD) affects BOD effectiveness, because the addition of female board members usually leads to better monitoring. However, the authors found a relation only to hold in case of non-representation of women in BODs in weakly governed firms. Contrary, in strong governance firms the addition of female board members led to over-monitoring and decreasing BOD effectiveness. Thus, board gender diversity may affect BOD effectiveness, but is influenced by the governance systems already in place in a firm. This, in turn, may have interesting consequences for the shareholder value of the firm.
11 which members of a group engage in a mode of thinking that is desired by the group, and by doing so ignore their motivation to take into account realistic other scenarios (Janis, 1972). This assumption is further supported by Adams and Ferreira (2009), who state that women directors’ perceptions more closely resemble the role of independent board members as “they do not belong to the old boys club”(p.2).
Existing literature mainly assesses the relation between BGD and effectiveness in a US context. In reaction to this, Campbell and Minguez-Vera (2008) research this relation in a code law environment, specifically Spain. The authors found support for a comparable relationship between female presence and firm performance in Spain, namely that a higher percentage of women in the BOD has a positive effect on firm performance.
Earlier research sheds light on the consequences of appointing women in certain positions for a firm’s tax strategy. Francis et al. (2014) research whether companies with a female CFO behave less aggressively in managing their taxes. The authors found that companies having a female as their CFO engage less in tax aggressive behavior than firms with male CFOs, and contribute this fact to the risk-aversion of the female CFOs. The higher risk-risk-aversion of women compared to men is supported by the findings of Croson and Gneezy (2009), who find that the higher risk-aversion of women can be attributed to emotions, confidence and attitudes towards challenges. Interestingly, the authors find that these differences are less evident between male and female directors and managers because experience and role influence the perception of risk of individuals.
Graham, Hanlon, Shevlin and Shroff (2013) find that the risk of reputational damage as a result of pursuing tax aggressive strategies is an important point of consideration for managers, which can prevent them from to engaging in tax aggressive behavior. Given that women are more risk-averse than men, this would suggest that women cover this risk with more care.
Hasseldine (1999) and Kastlunger et al. (2010) studied if the more compliant nature of women leads to higher tax compliance compared to men. The researchers focus on natural and social factors influencing behavior. Both researches suggest that women show more tax compliance than men, and that men show more behavior of tax evasion. The authors claim that these differences are the result of differences in societal expectations and self-concepts of men and women. Participants in the survey of Kastlunger et al. (2010) showed more tax evasion and strategic use of tax planning in case of high masculine character traits. Examples of attitudes influencing decision making are socially desirable behavior and kindness among women and dominance, competitiveness and aggressiveness among men.
12 BOD. Thus, the tax rate of the company is expected to rise when one more women are making part of the BOD. This leads to the following hypothesis:
H1: A higher percentage of female directors in a firm’s BOD negatively affects a firm’s tax aggressiveness.
2.2.2. Board independence
Next to BGD, board independence (BODI) may be a considerable factor in explaining a firm’s TA. In existing literature multiple terms are used interchangeably for discussing the concept of BODI. Fama (1980) and Fama and Jensen (1983) differentiate outside from inside directors, reflecting the affiliation of the respective director with the firm. Lanis and Richardson (2011) and Minnick and Noga (2010) make the distinction between independent and non-independent directors. Another differentiation commonly used is the executive versus non-executive classification. Board members can be classified as holding either executive and non-executive positions. These positions have a significant difference in responsibility, as it is the executive directors who are responsible for day-to-day operations of the firm, and the non-executive directors who are responsible for assuring that the executive directors act in the best interest of the shareholders.
The different roles come with varying levels of information availability and incentives. The executive directors are part of the management team of the company, and the non-executive directors have the sole task of controlling their counterparts in executive positions. Since the executive directors have more in-depth knowledge on the daily operations of the company, non-executive directors face an information disadvantage. For this reason, Beasley (1996) argues that control systems should be in place, as these systems reduce the likelihood that individual managers are able to make key decisions on their own. In addition, Williamson (1984) finds that the BOD may become a tool of daily management as a result of the executive members taking advantage of their full-time job and the beneficial position in the gathering of information that is accompanied by it.
The non-executive directors should keep in mind that management may tend to pursue their self-interests and consequently ignore shareholder interests, potentially leading to a transfer of stockholder wealth (Fama, 1980). According to Rosenstein and Wyatt (1990), stockholders value the inclusion of independent directors on boards, since the addition of these directors is found to lead positive abnormal stock returns. Carter, Simkins and Simpson (2003) argue that shareholders expect independent board members to perform the controlling function to the best of their capabilities, as the incentive for these board members is the potential for job opportunities once a reliable reputation in such a role is developed. For convenience and clarity purposes, this paper will avoid using the terms of outside board members, non-executives board members and independent board members interchangeably. Hence, solely the term independent board members is used in the rest of this paper.
13 An independent director can be described using the definition introduced by Higgs (2003): a director is “considered independent when the board determines that the director is independent in character and judgement and there are no relationships or circumstances which could affect, or appear to affect, the director’s judgement” (p. 37). The author proceeds by describing relationships that could affect director independence, which involve a past at the firm; close family ties with a firm director, advisor or senior employee; represents a significant shareholder; or has or has had a material business relation with the company the past 3 years, among others.
Fama (1980) and Fama and Jensen (1983) state that it is important to have a balanced distribution of inside and independent directors in the BOD. In case of an unequal distribution between inside and independent directors, BOD effectiveness may easily be mitigated. According to prior literature, decreased BOD effectiveness as a result of an undesired distribution of insiders and independent directors may lead to multiple undesired outcomes (e.g. Lee, Rosenstein, Rangan and Davidson, 1992; Brickley and James, 1987; Weisbach, 1988; Beasley, 1996).
As mentioned before, governments are starting to see that director independence is an important factor in preventing corporate scandals. In the US, the Sarbanes-Oxley Act of 2002 requires the audit committee of a corporation to exist of independent members only. Moreover, the NASDAQ and NYSE adopted new listing rules in which a majority of the board is required to be independent. Other examples of government intervention is the Bouton report in France and the Higgs report in the UK, both recommending that at least half of the board is independent (Higgs, 2003).
Regarding the relation between BODI and a firm’s tax management activities, Williams (2007) researched the role of the board of directors in corporate tax planning. His findings suggest that directors of firms active in common law countries have a duty of care, established by law, to ensure that proper internal control systems are implemented within a corporation and to monitor management. These control systems should address tax controls systems as well. Furthermore, the level to which the firm already carries out its tax management does not relieve the directors from this duty of care towards the shareholders. Thus, it is the BOD that bears the ultimate responsibility for the tax management practices in the firm and will be held accountable for malpractices by shareholders and stakeholders. As discussed above, increasing the level of independent board members is expected to lead to increased BOD effectiveness. This increased effectiveness should in turn lead to better tax management by firms.
14 Lanis and Richardson (2011) state that prior literature has assessed the relationship between CG and TA. These authors, too, mention that so far CG has not been decomposed into its major components such as BODI. For this reason, they researched the influence of including independent board members in the BOD on a firm’s tax strategy in Australian firms. The findings indicate a negative relationship between the percentage of independent board members and a firm’s tax aggressiveness.
From existing literature on CSR and tax management it becomes clear that independent directors have a key role in controlling the strategic choices of executive management of a corporation. Main focus is put on the consequences of these choices for society as a whole. Ibrahim, Howard and Angelidis (2003), for instance, bring forward that the controlling role of independent directors on strategic choices leads to a certain responsibility to ensure greater corporate responsiveness to society’s needs. Additionally, the authors find that independent directors are less likely to focus on a firm’s financial performance and pay more attention to non-financial performance indicators, such as CSR scores, than do executive directors.
Thus, the findings of existing literature suggest that adding independent board members to a firm’s BOD will increase its effectiveness. This should lead to more awareness for the importance of doing business in accordance with expectations raised by society. This, in combination with the more stakeholder orientated view of companies on the European mainland, lead to the expectation that BODI will negatively affect a firms tax aggressiveness. Hence, the hypothesis is as follows:
H2: an increase in board independence negatively affects a firm’s tax aggressiveness.
3. Method
3.1. Sample and data
15 included in the sample were constituents of the index at 21 September 2015.
Using the STOXX Europe 600 ex UK index provides an interesting advantage regarding this research, as mainly the largest corporations in Europe are included. According to Dyreng et al., 2008, it is the larger, multinational operating firms that are better able to implement tax strategies through the creation of extensive tax management departments and the use of transfer pricing for reducing tax rates. Moreover, the sample excludes private firms, which may have different incentives in tax management as discussed by Hanlon and Heitzman (2010).
Data were collected from publicly accessible sources. Financial data were retrieved from the Worldscope database. The scores for the CG aspects, namely percentage of females, percentage of independent board members and board size, are retrieved from the Thomson Reuters ASSET4 database. As firm cash taxes paid were not available for all companies in the sample, the Compustat Global database was used to supplement data on this variable. In case both Worldscope and Compustat Global did not report scores for this variable, data were collected manually from corporation annual reports. However, for numerous firms there was still the problem of non-complete availability of data. Unfortunately though, deleting all firms whose data were incomplete would have considerably reduced the sample size. In order to retain a sufficiently large sample size, firms that had data available on at least half the observations in the 2005-2014 period for all variables were included in the sample. Eventually, the final sample includes 254 firms and 13,363 firm years. Appendix A gives an overview of the companies included in the sample.
In order to deal with comparability issues some assumptions and transformation of the data was necessary. Since this research considers data on multiple European countries, not all company financial statements were reported in the same currency. The Euro (EUR) is used as the main currency throughout this research. For transforming the amounts stated in foreign currency to the corresponding amount of Euros, the annual average exchange rates of the Swiss Franc (CHF), Danish Krone (DKK), Swedish Krona (SEK), Norwegian Krone (NOK), Czech Koruna (CZK) and US Dollar (USD) are used. Data on the exchange rates for applicable years are retrieved from the OECD5.
In order to deal with possible outliers, all variables are winsorized at the 5 percent level at both sides, leading to outliers taking on the values of the 5 percent and 95 percent limits. This should reduce the effect of extreme values in the sample.
3.2. Dependent variable
Companies report taxable income in their tax returns and report income tax expense in their financial statements. As corporate tax returns are not publicly available and financial statements in general are, existing literature has mainly used financial statements for establishing relationships between company policies and TA (Hanlon and Heitzman, 2010). Existing literature has to a great
5
16 extent used the Effective Tax Rate (ETR) as dependent variable in researching TA (e.g. Armstrong et al., 2012; Lanis and Richardson, 2011; Dyreng et al., 2008). ETR is calculated by using a firm’s income tax expense and dividing this by the total pretax income of the firm. Both items are reported in company consolidated financial statements. As such, the formula for calculating ETR is as follows:
Effective Tax Rate=Income Tax ExpensePretax Income (1) Over the years tax management has become an important source of costs for firms, as there are incentives for speeding up and delaying transactions to the tax authorities. For example, the time value of money may make it interesting to defer taxes, leading to accounts receivable and accounts payable with regard to the tax authorities. The possibility of timing certain payments of taxes introduces the necessity of using an alternative method of capturing TA. This alternative measure was first brought forward by Dyreng et al., 2008. In response to this issue, the authors decided to look at actual cash taxes paid, which is reported in supplemental cash flow statement information for many firms. By doing so one is able to account for the effects of accrued and deferred taxes.
However, Dyreng et al. (2008) recognize the problem of possible variability in yearly Cash Effective Tax Rate (CETR), as the value for a certain year is not representative for structural pursuit of a low tax bill. Hence, the authors introduced the idea of calculating long-run corporate tax avoidance through taking ten-year averages of the total sum of cash taxes paid. By doing so, the influence of a one year score is controlled for. Comparing companies on ten year averages would allow one to make assumptions about which firms are more active in striving for lower tax obligations. In line with Dyreng et al., (2008), this research uses a long-run average as a proxy for a firm’s tax aggressiveness. The long-run CETR, however, uses three year averages instead of ten year averages, which is consistent with Huseynov and Klamm (2012). Data for the years 2005-2014 are used for calculating the three year average rates, hereby using the actual year and the values for the two previous years. This leads values on average CETRs for the years 2007-2014, as the years 2005 and 2006 are dropped as a result of calculating average scores.
The three year CETR is calculated by dividing the sum of cash taxes paid over the three years by the sum of pretax income generated over the three years, minus special items. These special items are erased from the equation by Dyreng et al. (2008), as the effects can be quite large. This example is followed in this research. Thus, the CETR is calculated as follows:
Cash Effective Tax Rate= ∑Nt=1cash taxes paid
Pretax Income (2)
17 Moreover, consistent with Dyreng et al. (2010) and Francis et al. (2014), negative scores on three year average cash taxes paid have been set to missing.
Several disadvantages of using long-run CETRs (and other ETR measures) as proxy for a firm’s TA exist. First, the long-run CETR takes into account all transactions that affect a firm’s tax liability. Secondly, CETRs cannot be used for differentiating between activities that are tax-favored, avoidance activities that have the specific purpose of reducing taxes or tax benefits from lobbying activities. Lastly, as CETR is calculated by using reported pretax income, tax aggressiveness in the form of reducing accounting income will not be captured by CETR. (Hanlon and Heitzman, 2010). However, as Huseynov and Klamm (2012) stress, “taxes are the result of a firm’s strategy and decisions”(p. 809), and therefore it is assumed that the three year CETR of a firm is an appropriate proxy for indicating tax aggressive policies.
3.3. Independent variables
Board gender diversity
Consistent with prior literature (e.g. Adams and Ferreira, 2009; Campbell and Mínguez-Vera, 2008), this research focuses on the percentage of female board members for measuring BGD. As hypothesized, a negative relation between the percentage of female directors on the BOD and TA is expected. The percentage of female board members is computed by dividing the total number of female board members by the total number of board members. Thus, BGD is formulated as follows: Xi,t = Percentage of females (total number of females on the BOD divided by the total number of BOD members)
Board independence
Lanis and Richardson (2011) use the percentage of independent board members for measuring BODI. However, since their research focuses on Australian firms, using the percentage of independent board members may cause trouble as in European firms differences in CG systems exist. In Australia the common model of CG involves one board with both insiders and independent directors. This type of board is also referred to as unitary board or one-tier board, and it is the structure applied in the larger part of the world.
18 of independent board members is not as easily measurable in Europe as it is in the case of Australia. In line with Lückerath-Rovers (2013) the international terminology of board of directors, referring to both the executive board and the supervisory board, is adopted. Subsequently, instead of taking the measure of the percentage of independent directors in the BOD as proposed by Lanis and Richardson (2011), this research uses the percentage of independent board members in the BOD as reported by the company. By doing so, it is possible to control for differences that may arise as a result of different CG systems that exist in Europe, as the only independence requirement now comes down to whether or not the director is from outside the company. Thus, this proxy is more appropriate for measuring independence of board members, as it can be used throughout different CG systems. The proxy for BODI therefore is as follows:
Xi,t = Board independence (total number of independent directors on the BOD as reported by the company divided by the total number of BOD members).
3.4. Control variables
This research controls for the effect of certain company characteristics on TA. The control variables include company size, leverage, profitability and board size, each of which is discussed below. Firstly, different support exists for including the size of the firm as control variable. Richardson and Lanis (2007), for instance, find that larger corporations are more likely to be tax aggressive because of the greater economic and political power they have compared to smaller firms. This power is suggested to be used in the reduction of tax obligations. Rego (2003) provides a different approach, that relies on the political costs theory. This research finds that larger firms may face higher political costs, and that firms may want to limit the possibility of incurring these costs. This leads to less activity in TA. Furthermore, prior research has also found no relationship between TA and firm size (Stickney and McGee 1982; Shevlin and Porter 1992; Gupta and Newberry 1997). Thus, although prior research is mixed on the potential effect of firm size on TA, it should be controlled for. Taking into account the mixed support from prior research, no sign predictions are made regarding the expected effect of firm size on TA. The proxy for firm size (FSIZE) selected is the natural log of total assets, which is in line with previous research (e.g. Lanis and Richardson, 2011; Hanlon and Slemrod, 2009; Gupta and Newberry, 1997).
Secondly, leverage is found to potentially impact a firm’s TA, which is found in Graham (2003). Taxes are affected through the tax deductibility of interest, which makes debt financing an interesting option for corporations for raising capital. The effect of leverage is controlled for by adding a leverage variable (LEV), measured by dividing the amount of total debt by total assets. It is expected that there is a positive relationship between a firm’s debt level and tax aggressiveness since a higher debt level allows a firm to subtract more interest payments from its tax bill.
19 to Lanis and Richardson (2012) more profitable firms are expected to engage more in tax aggressive behavior than less profitable companies, as tax aggressive policies would result in less profits being transferred to the tax authorities. Additionally, Chen et al., (2010) state that more profitable firms could have higher incentives for tax planning. Hence, the effect of profitability is controlled for. Taking into account the findings of existing research, a positive effect of profitability on TA is expected. The proxy for profitability is return on assets (ROA), calculated by dividing net income by total assets.
Lastly, in previous research board size is found to affect the effectiveness of the BOD in performing its controlling function over management. Jensen (1993) and Yermack (1996) both found support for an increased ability of the BOD to control management as a result of a smaller board size. The researchers contribute this to a loss of BOD effectiveness, which can be caused by problems in coordination, communication, and decision-making between the BOD members, among others. The loss of board effectiveness that is associated with a larger size of the BOD may leave more room for opportunistic behavior by top management. Therefore, a positive relation between a firm’s board size and its TA is expected. Table 1 provides an overview of the included variables and the expected effect on the dependent variable.
Table 1
Included variables and expected effect on TA Predicted
effect on TA
Measure Abbreviation
Independent variables
Board gender diversity Negative total number of females on the BOD total number of BOD members
BGD
Board independence Negative total number of independent BOD members total number of BOD members
BODI
Control variables
Firm size No
prediction
Ln (Total Assets) FSIZE
Leverage Positive Total debt
Total assets
LEV
Profitability Positive Net income
Total assets
ROA
Board size Positive Total number of directors in the BOD BSIZE
3.5. Descriptive statistics
20 in their boards and that a little more than half (52.3%) of the directors is independent. Furthermore, interesting is the high standard deviation on the score of BODI, indicating that a lot of variance exists in the degree to which firms have appointed independent directors in their BODs. The correlation matrix (Table 3) shows that multicollinearity is not expected to cause problems in the regression, as the highest reported value is close to 0.52 (Brooks, 2008).
Table 2
Descriptive statistics
Mean SD Max Min Observations
CETR -0.276 0.155 -0.657 -0.021 1909 BGD 16.675 12.295 40.000 0.000 1978 BODI 52.326 28.786 94.480 7.580 1590 FSIZE 16.479 1.543 19.673 14.078 2032 LEV 0.254 0.157 0.548 0.012 2032 ROA 0.051 0.052 0.183 -0.022 2032 BSIZE 11.750 3.914 20.000 6.000 1970 Table 3 Correlation matrix 1 2 3 4 5 6 7 1. CETR 1 2. BGD 0.004 1 3. BIND 0.050 0.150 1 4. FSIZE -0.088 0.032 0.048 1 5. LEV -0.026 -0.075 -0.090 0.137 1 6. ROA 0.233 0.045 -0.014 -0.516 -0.302 1 7. BSIZE -0.084 -0.043 -0.384 0.496 0.120 -0.294 1 3.6. Models
Having discussed the variables to include in the analysis, the model that will be used for performing the analysis is discussed next. As there are two different hypotheses to be tested, two different models are required. First, the model to be used in fixed effects (FE) tests is discussed, after which the model of the system Generalized Method of Moments (GMM) is explained.
For analyzing the relation between BGD and TA using FE, the following regression model is used: CETRit= α0 + β1BGDit + β2FSIZEit + β3LEVit + β4ROAit + β5BSIZEit + αi + λt + εit (3) where і represents firm and t represents year. CETRit is the dependent variable, α0 is a constant, β
21 For analyzing the relation between BODI and TA using FE, the following regression model is used: CETRit= α0 + β1BODIit + β2FSIZEit + β3LEVit + β4ROAit + β5BSIZEit + αi + λt + εit (4) where і represents firm and t represents year. CETRit is the dependent variable, α0 is a constant, β
represents the coefficient of the respective variable, αi is firm FE, λt is time FE and εit is the error term. In order to control for potential endogeneity issues, which are discussed in more detail in the next section, the GMM estimator is used. Arellano and Bover (1995) and Blundell and Bond (1998) developed the system GMM as an augmented version of the difference GMM, as the latter one may lead to low power in the instruments. The authors argue that the system GMM is more appropriate than the difference GMM, as this estimator is able to correct for potential correlation between some regressors and the disturbance term. The system GMM estimator can be derived by following some essential steps. First, a dynamic model is written in first-differenced form:
ΔCETRit= α + γpΣp ΔCETRit-p+ β1ΔBGD/BODIit + β2ΔFSIZEit + β3ΔLEVit + β4ΔROAit +
β5ΔBSIZEit + Δεit, p>0. (5)
where Δ captures the use of lagged values of the different variables, γ is the coefficient of the lagged values of the dependent variable CETR, and BGD/BODI indicates that either BGD or BODI is to be selected as the independent variable.
First-differencing allows one to include lagged values of both the dependent and explanatory or control variables, hereby controlling for a firm’s past values on current performance. However, first-differencing may create the problem of correlation between some explanatory or control variables and the idiosyncratic disturbance term. Subsequently, an equation in levels is added to the first-differenced equation, as instrumental variables are required for consistency. As such, the system GMM estimator is calculated as follows:
[∆CETRCETRit
it] = α+ γ [
CETRit-p
∆CETRit-p] + β1[∆BGD/BODIBGD/BODIit
it] + β2[
FSIZEit
∆FSIZEit] + β3[∆LEVLEVit it] +
β4[∆ROAROAit
it] +β5[
BSIZEit
∆BSIZEit] + εit (6)
22
4. Results
4.1. Results hypotheses testing
The basic estimation technique available for panel data is pooled OLS. Pooled OLS uses the same constant and same coefficient for all companies included in the sample. The underlying assumption of pooled OLS is homogeneity among the variables, leaving no room for firm-specific effects and variables that change over time. Hence, an interesting complement to the pooled OLS regression is the use of fixed effects (FE) or random effects (RE) in the model. This is one of the main advantages of using panel data, as it allows one to control for unobserved heterogeneity among the firms and periods. Since the sample involves firms from different countries across Europe, there must some cross-sectional unobservable heterogeneity as a result of i.e. culture or political movements being in power.
Cross-section FE allows one to control for unobserved heterogeneity as a result of factors that affect the separate cross-sections (in this case firms) and do not change over time. An example applicable to this research would be culture. Time FE are also included, as this allows to control for omitted variables constant over cross-sections, and therefore applicable to all of the firms in the sample through the different years. For instance, time FE may be used to control for the business cycle. As the period 2007-2014 is characterized by the consequences of a major global financial crisis, controlling for time FE seems legitimate.
In order to decide which FE is more suitable for the investigation, a Redundant Fixed Effects Test is performed. With the Redundant Fixed Effects Test it can be determined if there is unobserved heterogeneity in the model and what test should be used to control for it. As the test scores for both firm and time FE shows significant results, there is unobserved heterogeneity across firms and over time in the sample. Hence, using pooled OLS is invalid and therefore FE testing on both firms and years is preferred.
An alternative to applying an FE model is using the RE model. An RE model has the advantage of allowing for a common intercept and a firm-specific error term vis-à-vis the FE model. This error term is either varying across firm and fixed over time, or fixed across firms and varying over time. This leads to the inclusion of a richer set of data than in the case of FE, as in the FE models the variables that do not change over either cross-sections or time are deleted from the sample. The main drawback of the RE model, however, is the underlying assumption that there are no omitted variables in the model. Moreover, an assumption is that any omitted variables should not be correlated with the independent variables. As this requirement is hard to meet in practice, an FE model is the preferred one for testing the hypotheses.
23 The RE model therefore is not a consistent estimator, which leads me to conclude that the FE model is preferred.
The results of the FE tests of both hypotheses can be found in Table 4. The tests include cross-section and time fixed effects, and White’s period standard errors are used with the purpose of controlling for heteroscedasticity and serial correlation issues. Regarding the first hypothesis (BGD), the regression results show no significant results for the relation between BGD and TA. Thus, hypothesis 1 is not supported. This is consistent with Khaoula and Ali (2012), who find that TA is not affected by female presence on the board of American firms. The authors contribute this to the low percentage of women in the board and the dominance of masculine preferences in determining a firm’s tax strategy. This suggestion is also brought forward by Elstad and Ladegaard (2012), who find that female influence increases once a higher percentage of the BOD is female. As the average percentage of females in the BOD is comparable in this research, the same may explanation may be applicable. The results of testing hypothesis 2 with FE show similar results. As is the case with BGD, BODI seems not to affect a firm’s TA as the relation between CETR and BODI is insignificant. Therefore, I find no support for hypothesis 2. This is inconsistent with Lanis and Richardson (2011), who find a significant negative effect. The authors contribute this to the increased effectiveness of the BOD as a result of installing more independent directors.On the other hand, Minnick and Noga (2010) find an insignificant effect.
The control variables show different results. In both models, a positive and highly significant relation between TA and both ROA and FSIZE is confirmed. This would imply that TA increases with profitability and the size of the firm. The effects are stronger in case of profitability than in the case of firm size. Chen et al. (2010) found that firms with higher profitability benefit more from lower taxes, and therefore these firms show more tax aggressive behavior. Regarding firm size, the finding is in line with Richardson and Lanis (2007), who state that larger firms may be better able to pursue tax aggressive strategies due to their greater economic and political power.
Leverage shows no significant relation in both models. This is inconsistent with Gupta and Newberry (1997), who confirm a positive relation between leverage and TA due to the tax deductibility of interest payments.
For board size the results are mixed. In case of the BGD model, the result is insignificant, whereas in the BODI model it is significant at the 10 percent level. The economic magnitude of the board size effect, though, is found to be very small. Lanis and Richardson (2011) and Minnick and Noga (2010) find that board size has no significant relation with tax aggressiveness. On the other hand, Jensen (1993) and Yermack (1996) suggest that this may be attributable to more room for opportunistic behavior by top management.
24 reasonable explanatory power, as around 65% of the variance in CETR is explained by the BGD model. In case of the BODI model, this score is approximately 67%.
Table 4
Results cross-section and time fixed effects (FE).
This table provides the results on the cross-section and time FE tests for hypothesis 1, in which board gender diversity (BGD) is the independent variable, and hypothesis 2, in which board independence (BODI) is the independent variable. Cash Effective Tax Rate (CETR) is the dependent variable and equals the three-year average of cash taxes paid/pretax income. BGD is measured as the percentage of female directors in the board and BODI is the percentage of independent directors. FSIZE, LEV, ROA and BSIZE are control variables. Firm size (FSIZE) is the log of total assets. Leverage (LEV) is calculated as debt/assets. Profitability (ROA) equals net income/assets. Board size (BSIZE) is the total number of directors in the board. White’s period standard errors are reported in brackets under the coefficients. ***, ** and * indicate levels of significance at 1, 5 and 10%.
Dependent variable: CETR
Independent variable H1 H2 Intercept -1.813*** -1.705*** [0.311] [0.303] BGD 0.001 0.000 BODI 0.001 0.001 FSIZE 0.089*** 0.081*** [0.019] [0.019] LEV -0.090 -0.042 [0.075] [0.073] ROA 0.988*** 0.957*** [0.156] [0.188] BSIZE 0.003 0.005* [0.002] [0.003] R² 0.646 0.669 Adjusted R² 0.588 0.597 Observations 1851 1483 4.2. Endogeneity
Prior research has discussed the potential problem of endogeneity in researching the relation between CG and firm performance (e.g. Hermalin and Weisbach, 1998; Boone, Field, Karpoff and Raheja, 2007; Coles, Daniel and Naveen, 2007). More specifically, Francis et al. (2014) indicate that studies on gender effects often face the criticism that “the observed effects are not attributable to gender, but instead to some omitted factors, such as situational factors and knowledge disparities” (p. 2). In case of endogeneity issues there may be biased and inconsistent estimators, which means that making reliable inferences is virtually impossible (Wintoki et al., 2012).
25 Simultaneity involves reverse causality, whereas unobserved heterogeneity evolves through unobservable factors that affect both the dependent and explanatory variables. In case of the first issue, OLS and Fixed Effects (FE) tests both lead to biased results. For the latter problem, FE may prove useful for coping with endogeneity as some of the unobserved heterogeneity is dealt with. However, Wintoki et al. (2012) suggest that FE tests deal with this problem in a wrong manner, as it assumes that current observations of the dependent variable are completely independent of past values of the dependent variable. This assumption, however, generally does not hold.
There are grounds to believe that endogeneity may be an issue in this research, as earlier research has already found that CG may affect TA as much as TA affects CG (Wintoki et al., 2012). As discussed earlier, according to Fama and Jensen (1983) the BOD should be formed in such a manner so that it leads to firm performance that is in line with shareholder expectations. This means that the manner in which the BOD is constituted may to a great extent affect firm performance. For instance, if shareholders aim to maximize shareholder value at any cost, the BOD should be constituted in such a manner in which the board members strive for the highest profit. As stated by Hanlon and Heitzman (2010), a tax aggressive policy may lead to a better result below the bottom line. However, past performance on TA may also force shareholders to install more risk-averse board members in case they fear penalties for a firm’s tax aggressive activities. This is an example of in which current BGD is partly determined by past values of TA, and therefore simultaneity may be an issue. Moreover, other research states that CG can reduce corporate risk-taking through higher board effectiveness. This higher effectiveness may be obtained through an increase in the presence of both independent board members and female board members (e.g. Lanis and Richardson, 2011; Abbott et al., 2012; Adams and Ferreira, 2009). A firm’s tax strategy involves a tradeoff of risk and return and it is the shareholders who decide how the company will deal with this tradeoff through installing the BOD in a certain manner. Thus, TA may (through firm performance) have as much an effect on CG as CG may have on TA, and therefore a potential endogeneity problem exists.
In response to this endogeneity problem, Wintoki et al. (2012) discuss the GMM, which uses dynamic panel data in order to take into account the past values of the endogenous variables included. The authors suggest that endogeneity may, next to the problems of unobserved heterogeneity and simultaneity, be the result of a third cause, namely the sort that arises from the influence of past firm characteristics. Subsequently, this effect of some past characteristic of a firm may in turn determine future performance or governance. The GMM can be used to test if endogeneity problems are applicable to a model. Therefore, it is considered a better method for testing CG-performance relations.
26 relation between CG and tax management. Wintoki et al. (2012) and Minnick and Noga (2010) use the system GMM to control for endogeneity issues in CG that may arise from past performance. An example of such a relationship was proposed by Hermalin and Weisbach (1998), who explain that board structure may be partly determined by a CEO’s ability, which is in turn determined by looking at firm past performance. Moreover, Wintoki et al. (2012) find a negative relation between board independence and past firm performance. In the system GMM, controlling for past values is realized by including lags of the dependent variable (CETR) and the independent variables as instruments. Thus, the system GMM uses instruments that are included in the panel itself, hereby eliminating the need for external instruments.
This paper applies the system GMM to include lags of the multiple variables. In line with Minnick and Noga (2010), robust standard errors of the two-step estimator designed by Windmeijer (2005) are used, as this estimator is found to lead to more efficient results in the system GMM. Contrary to Minnick and Noga (2010) who use lags of one and two years, this research uses the second and third lags as instruments. Lags from the second lag and deeper are not correlated with the error term, whereas the first lag is. Wintoki et al. (2012) explain this in more detail. Whereas Minnick and Noga (2010) assume that their independent and control variables are merely predetermined, in this research it is expected that some endogeneity may exist in the relation between CETR and the independent and control variables. Results of the system GMM estimator are reported in Table 5. The results of the system GMM for BGD provide enough support to assume that the model is correctly specified, as the second lag of CETR, the dependent variable, shows no significant effect in explaining the value of the variable. Furthermore, autocorrelation is not a problem as the effect of the second order autocorrelation AR(2) is insignificant, hence the null hypothesis of no autocorrelation is not rejected. The Hansen test indicates that the test may not be ideal since it proves significant. This means that the null hypothesis that the variables are exogenous as a group is rejected, but the model is still consistent.
27 Table 5
Results of running the system GMM estimator on three-year average Cash Effective Tax Rate (CETR).
The system GMM estimator allows to control for endogeneity while researching the corporate governance- tax aggressiveness relation. Column H1 presents results for testing hypothesis 1, in which board gender diversity (BGD) is the independent variable. Column H2 applies to hypothesis 2, in which board independence (BODI) is the independent variable. CETR is the dependent variable and equals the three-year average of cash taxes paid/pretax income. BGD is measured as the percentage of female directors in the board and BODI is the percentage of independent directors. FSIZE, LEV, ROA and BSIZE are control variables. Firm size (FSIZE) is the log of total assets. Leverage (LEV) is calculated as debt/assets. Profitability (ROA) equals net income/assets. Board size (BSIZE) is the total number of directors in the board. Windmeijer corrected robust standard errors are shown in brackets. As the regressors may show cases of endogeneity, lags 2 and 3 are taken as instruments which leads to lags of t-2 and t-3. P-values for additional tests are shown below. AR(1) and AR(2) are tests for first and second order autocorrelation in the first-differenced residuals, under the null of no serial correlation. The null hypothesis of the Hansen test is that the instruments as a group are exogenous. The Difference in Hansen test has as a null hypothesis that the levels of instruments and GMM are exogenous. ***, ** and * indicate levels of significance at 1, 5 and 10%.
Dependent variable: CETR
Independent variable H1 H2 CETRt-1 0.786*** 0.707*** [0.076] [0.07] CETRt-2 -0.088 -0.018 [0.119] [0.137] BGD 0.002* [0.001] BODI 0.001 [0.001] FSIZE -0.073* -0.081** [0.039] [0.034] LEV 0.083 0.055 [0.213] [0.247] ROA 1.653*** 1.388** [0.437] [0.701] BSIZE 0.123 0.001 [0.009] [0.013] Intercept 0.821 1.125** [0.619] [0.554] Observations 1321 1055 AR(1) test 0.000 0.001 AR(2) test 0.588 0.873 Hansen test 0.004 0.110 Difference in Hansen 0.014 0.140
28 resemble those of men than the characters of average women do. This idea was brought forward by Atkinson, Baird and Frye (2003), Kumar (2010), Sheridan and Milgate (2005) and Adams and Funk (2012). Moreover, this different result in comparison with the Wintoki et al. (2012) and Minnick and Noga (2010) papers may be caused by the different geographical region. However, as Europe is found to be more stakeholder orientated (Ball et al., 2000), one would expect an opposite effect of increasing female BOD presence on a firm’s TA.
The results of testing the effect of BODI on TA show that, again, there is enough evidence to assume that the model is correctly specified. The effect of the second lag of the dependent variable is insignificant. The AR(2) estimate indicates that in this test, too, autocorrelation is not an issue as it is insignificant. The effect of BODI is found to be insignificant in explaining a firm’s TA, which is contrary to my hypothesis but in line with the findings in the FE model. Minnick and Noga (2010) arrive at the same conclusion using a system GMM model. Thus, controlling for simultaneity and dynamic relations does not lead to different results compared to the FE model. Therefore, it is safe to state that no support for H2 is found. Thus, consistent with findings of Minnick and Noga (2010), but inconsistent with Lanis and Richardson (2011), the degree of BODI does not explain differences in tax aggressiveness. A possible explanation for this finding is that Lanis and Richardson (2011) do not use panel data. As discussed above, panel data allow researchers to account for unobserved heterogeneity and simultaneity. Minnick and Noga (2010) take into account this issue because endogeneity may affect the results. Thus, this may influence the results of Lanis and Richardson (2011).
Looking at the control variables included in the BGD and BODI models, firm size and profitability are significant predictors of a firm’s TA, although differences in the levels of significance exist. In case of profitability a positive relation exists, but for firm size this is negative. Comparing these results with the FE estimates, the same relation was established regarding profitability and this was according to my expectations. However, the system GMM finds an opposite relation compared to the FE estimator for firm size, namely a negative effect on TA. This would imply that tax aggressiveness decreases once a firm grows larger. The different result in the GMM may be caused since lagged values are taken into account as instruments. Past values have a role in explaining the role of firm size in its relation with TA. The different results of the system GMM and FE test explain why prior research has found mixed results on the firm size- TA relation. The negative effect of firm size on TA was also found by Rego (2003), who uses the political costs theory for explaining why TA decreases once firms grow larger. Furthermore, Zimmerman (1983) suggests that larger firms face more risk of negative media coverage, and therefore engage less in tax planning.
