Social reporting and company determinants impacting corporate tax avoidance in the Netherlands

(1)

Social reporting and company determinants

impacting corporate tax avoidance in the

Netherlands

Reinier Evenhuis

ABSTRACT

In recent years, a lot of attention has been drawn towards companies avoiding corporate taxes. Society has labelled companies as immoral because of their tax avoidance behaviour. A recent example of this were the public hearings of Amazon, Starbucks and Google. This study examines how corporate social reporting and the company determinants age and government ownership influence a company’s tax avoidance behavior. By using a sample of Dutch companies of the period 2010 to 2014, this study finds that tax avoidance has a significant relation with application of the guidelines of the global reporting initiative and government ownership. No significant relation was found between the degree of corporate tax aggressiveness and company age. A sensitivity analysis was performed in order to verify the results by using two other measures to proxy for tax avoidance. This resulted in different results concerning the application of the global reporting initiative guidelines. The significance was lost, but due to sample restrictions it could not provide the necessary evidence to refute the original result. For company age, the sensitivity analysis does not change the results and remains insignificant. Based on the findings, the study contributes to a better understanding of the relation between corporate social responsibility and tax avoidance, and it identifies company characteristics that indicate tax avoidance practices. The results of this study may also have practical implications. Government and society can gain a better understanding of the conditions under which a higher change of tax avoidance behavior by companies exist.

(2)

PREFACE

About one and a half year ago I completed my bachelor’s thesis for the Bachelor of Accountancy & Controlling. My research topic was the quality of risk reporting within the Dutch pension sector. For the master’s thesis I wanted something completely different. I have always had a strong interest in tax structures and tax minimization structures. I was delighted to hear that for this academic year tax avoidance would be an available research topic. Writing this thesis has gained me a lot of insights into the practices applied by companies in order to avoid taxes and the scale on which this has happened. Furthermore, it gave me the tools to further develop myself in terms of academic writing, thinking and most of all time management.

I would like to thank several people who contributed to this thesis. First of all my supervisor, professor S. Lee for the provided feedback and guidance in writing the thesis. I would also like to thank my colleagues at PwC for the helpful discussions and feedback. Lastly, I would like to thank PwC for providing the opportunity and placement for writing the thesis.

Reinier Evenhuis, June 2016

Name Reinier Evenhuis

Student number S2230984

Date of birth 17-09-1991

E-mail address r.r.evenhuis@student.rug.nl reinierevenhuis@live.nl

University University of Groningen

Faculty Economics and Business

Study Program MSc Accountancy & Controlling (track: Accountancy)

Supervisor (university) dr. S. (Soojin) Lee

Second supervisor (university) dr. V.A. (Vlad Andrei) Porumb

Internship organization PricewaterhouseCoopers (Utrecht)

Supervisor (organization) F. (Francesco) Siefers

(3)

2.5 IMPRESSION MANAGEMENT………...9 2.6 HYPOTHESIS DEVELOPMENT……….10 3. RESEARCH DESIGN……….12 3.1 SAMPLE SELECTION……….12 3.2 DEPENDENT VARIABLES……….13 3.3 INDEPENDENT VARIABLES……….16 3.4 CONTROL VARIABLES………..17 3.5 MODELS………...19 4. RESULTS……….20 4.1 DESCRIPTIVE STATISTICS………...20 4.2 CORRELATION RESULTS……….20 4.3 MULTIVARIATE ANALYSIS……….21 4.4 SENSITIVITY ANALYSIS………...25

5. DISCUSSION AND CONCLUSION………..32

5.1 CONCLUSION………..32

5.2 LIMITATIONS………..33

5.3 FUTURE RESEARCH………..34

(4)

1. INTRODUCTION

In recent years, a lot of attention has been drawn towards companies avoiding corporate taxes by using tax shelters, complex securities or by the use of transfer pricing. Companies are being labelled as immoral because of their tax avoidance practices. One of the most outspoken examples of this judgement from society, were the public hearings of Amazon, Starbucks and Google, with the famous comment: “We’re not accusing you of being illegal, we’re accusing you of being immoral” (BBC, 2012).

Hutchinson (director of global tax for Google) stated back in 2010 that it was their duty to create tax structures that minimizes taxes for their shareholders (Washington post, 2010). That statement indicated the aggressiveness that Google applied concerning minimizing tax expenditures. Since then the times seem to have changed.

Henderson (2012) made the notion that society can be assured that tax transparency will be here to stay. Companies should address the issue as soon as possible. This notion paints a picture of a society that is finished with large corporations exploiting loopholes in tax regulation and subsequently demands transparent reporting on how corporations are managing their taxes.

These national sentiments seem to be shared in the US, where the emphasis on tax avoidance practices used by corporations was also highlighted by president Barack Obama, when he stated there is a small but growing group of large corporations that, in order to avoid paying taxes, flee the country (Reuters, 2014). As a reaction, the Obama administration came with budget proposals in the fiscal year of 2015, designed to decrease the relocation of assets and income overseas (Androff, 2014).

Even more recent, growing dissatisfaction surrounding tax avoidance was flared up by a settlement reached by Google and the British tax authorities (HMRC). A respectively low payment of 130 million pounds was agreed concerning back taxes over the last decade (The guardian, 2016). The French finance minister said that this low payment seemed more like the product of a negotiation, rather than the application of the law. The minister hereby stands behind the widespread allegations made that the deal amounts to an economically favorable one for Google (The Guardian, 2016). The broad media- and political attention to tax avoidance creates awareness in the rest of the world and will probably result in global consequences. The high media coverage that has especially been dominant in the UK may indeed spread further into Europe and consequently to the rest of the world.

(5)

1.1 Understanding tax avoidance and CSR

Over the last decade, the issue of tax avoidance has become a public issue. In a time of economic turndown, this has created societal displeasure. Since there is an existing link between corporations and society, a tension arises when corporations employ tax avoidance strategies in order to minimize their tax expenditures. When trying to understand the concept of tax avoidance, it is commonly seen as reducing explicit taxes (Dyreng, Hanlon & Maydew, 2008; Hanlon & Heitzman, 2010). This can be achieved in a variety of methods. An example is the creation of tax structures and using benefits provided by laws and regulation while remaining in the realms of the law.

The European Commission has a more comprehensive view of tax avoidance, and sees it as certain companies using aggressive tax planning, basically to try to pay as little taxes as possible. This can often be achieved by exploiting legal loopholes in tax systems and mismatches between rules of different nations. The companies shift their profits artificially to low or no tax jurisdictions. However, by doing so, the principle of taxation reflecting the place of the economic activity no longer applies (Europaeu, 2015).

In tax avoiding behavior, corporations make the decision that is in their shareholders best interest, but that penalizes society out of funds that can improve society as a whole. The resulting tension that is created is best described by the concept of corporate social responsibility (CSR). Wood (1991) understands the concept of CSR that corporations and society seem to be connected with one another and hereby cannot be viewed as separate bodies. This connection will result in society expecting that corporations will show correct behavior that is not only in their own best interest, but also in the interest of society. To rephrase, corporations are not solely responsible for themselves and their shareholders, but also bear responsibility towards other stakeholders, for example the environment or society (Keinert, 2008). This is in sharp contrast with tax avoiding behavior.

In order to resolve the issue at hand, the G-20 has called for increased cooperation between countries to create a more international tax regime, by instituting the Base Erosion and Profit Shifting (BEPS) Project. This project provides solutions for closing the legal loopholes that companies are using to artificially shift their profits to low or no tax jurisdictions (OECD, 2016). The extent to which the BEPS project has been successful globally, remains to be seen.

(6)

1.2 Research gap

As previously indicated, tax avoidance is a global matter that is becoming more and more a public debate. Research concerning tax avoidance is mainly based on research from the UK (Henderson, 2012), Australia (Lanis & Richardson, 2012a; Lanis & Richardson, 2012b) and the US (Dyreng et al., 2008; Huseynov & Klamm, 2012; Rego, 2003).

This study will make a contribution by looking at the Netherlands. The Netherlands is renowned for being used by multinationals in reducing their tax obligations. The Netherlands has an extended treaty network, preferential tax regimes and efficient tax administration, which combined makes the Netherlands a possibly favourable country of settlement. Other, more generalized elements of the tax regime of the Netherlands are that there is no taxation on outgoing interest and royalty payments, and there is a special immunity for foreign dividend income (Weyzig & Van Dijk, 2009). This altogether makes the Netherlands a tax regime that makes it possible to aggressively avoid tax and could therefore be considered as a tax haven especially for the larger corporations (Weyzig & Van Dijk, 2009). Furthermore, given the history of the Netherlands as a trading nation, attracting foreign commerce has always been a crucial part of the economy. To attract this foreign commerce, the Netherlands has positioned itself as a fiscally attractive nation for businesses to settle (physically, or only on paper, which is appointed by the term shelf companies).

1.3 Research direction

In order to further investigate the relation between CSR and tax avoidance, this study will look at the impact that corporate social reporting has on the tax avoidance by companies. Furthermore, the study will look at the company determinants: corporate age and government ownership, that can have an impact on tax avoidance practices. Hence, the research questions of this study are: (1) Can tax avoidance be influenced by corporate social reporting in the form of international reporting initiatives? (2) Can government ownership make a difference in the tax avoidance practices of corporations? (3) Does corporate age have an impact on tax avoidance?

Regression results from a sample of 1601 firm-year observations for the fiscal years of 2010-2014 indicate that tax avoidance has a significant relation with application of the guidelines of the global reporting initiative and government ownership. No significant relationship is found between the avoidance of tax and company age. A sensitivity analysis is performed in order to verify the results by using two other measures to proxy for tax avoidance. This results in conflicting results concerning the application of the global reporting initiative guidelines, but ultimately does not provide evidence to

(7)

refute the original results. For company age the results of the sensitivity analysis remain insignificant. Based on the findings, the paper contributes to a better understanding of the relation between corporate social responsibility and tax avoidance. Furthermore, it identifies company characteristics that can be used to indicate a higher probability of tax avoidance practices. The results of this study may also have practical implications. Government and society can gain a better understanding of the conditions under which a higher change of corporate tax avoidance behavior by companies exist. 1.4 Structure

The remainder of this study will be structured as follows. Section 2 will define the concepts and give a review of relevant research and theory. Consequently, hypotheses will be developed. Section 3 will discuss the research design, sample selection, variables and models. Section 4 will present the results and section 5 will present the discussion, conclusion and limitations of the study. The study will conclude with possibilities for future research.

(8)

2. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT

In this section the concepts corporate social responsibility and tax avoidance will be discussed, and the relationship between the concepts will be established through theory. The foundation for this relationship will be based on the use of previous research. Lastly, the hypotheses will be developed. 2.1 Corporate tax avoidance

Tax avoidance is a concept that has recently gained great media attention, but what does it actually imply? Tax avoidance has several different definitions, but a broad explanation of this concept is, as stated in the introduction, “reducing explicit taxes” consistent with the view of Hanlon and Heitzman (2010). A different definition brought forward by Dyreng et al. (2008), sees tax avoidance as “anything that reduces the firm’s cash effective tax rate over a long time period”.

A distinction between two concepts, ‘tax avoidance’ and ‘tax evasion’, has to be made immediately. Even though prior media attention has used the terms interchangeably, there is an important difference between the two. Hanlon and Heitzman (2010) see tax concerning matters on a spectrum, with taxes as a cost on one end, where firms are allowed to engage in tax management or tax planning to reduce the cost, increase their profits and add shareholder value, and reducing taxes on the other end of the spectrum, where firms engage in tax sheltering or tax evasion. This view is consistent with the thoughts of Hanlon and Slemrod (2009), as well as the views of Erle (2008).

From the view of Hanlon and Heitzman (2010) can be derived that tax avoidance is the reduction of tax, in which it is seen as a cost that can be minimized in a legal manner. On the other hand there is tax evasion, in which the only purpose is the minimization of taxes, where there is little regard for legislation. This paper will focus on tax avoidance, which is legal, but remains a grey area since the practice involves exploiting loopholes and small gaps that are found in legislation, in order to reduce or avoid taxes (Killaly, 2009; Braithwaite, 2005).

Tax avoidance can also be denoted as tax planning, which means the maximization of the firm its expected discounted after-tax cash flows (Scholes, Wolfson, Erickson, Maydew & Shevlin, 2002). This will result in a variety of structures with the aim of minimizing the effective tax rate of the corporation.

It seems that the practice of tax avoidance has been the norm for so many decades, that it is now entrenched within the fibres of corporate culture (Rego, 2003; Slemrod, 2001; Braithwaite, 2005). This has implications, since corporations mainly see taxes as costs (Hanlon & Heitzman, 2010). On the other hand, however, taxes are a contribution to society, in the sense that these funds are used for

(9)

the country its infrastructure and social programs ( Huseynov & Klamm, 2012). This creates a tension, which has resulted in companies becoming even more aggressive in avoiding tax to increase profitability and grow their cash flows, while government is trying to scrutinize the companies to protect their tax revenues.

2.2 The concept of corporate social responsibility

In defining CSR, it becomes clear that there is not just one single definition; it has a different meaning for every company, which depends on the overarching perspective of the company (Moir, 2001). The underlying idea of CSR is that there is a connection between companies and society that makes them entangled rather than two independent parties (Wood, 1991). This view concludes on one of the most important aspects of CSR: that there is a relationship between businesses and society, and this relationship creates mutual responsibilities. This has been indicated by Cannon (1992, p. 33), who says that in order for business to contribute to society, it has to be efficient, profitable and socially responsible.

Brummer (1991) gives a distinct explanation of the term responsibility and provides a definition. His conclusion is that responsibility entails that the corporate leaders of corporations are to be held responsible for the decisions they make.

In order to understand the corporate interpretation of social responsibility, a statement was made by Shell: “We all need to assess the impact our business makes on society and ensure that we balance the economic, environmental and social aspects of everything we do” (Moody-Stuart, 1999). This interpretation of CSR is largely consistent with that of The World Business Council for Sustainable Development.1 They see CSR as a dedication by corporations to act ethical and being involved with economic advancements while also trying to raise the overall quality of life for overall society.

The corporate interpretation of CSR shows that the mutual relationship between corporations and society are based mainly on three pillars: economic, environmental and social. The corporations will have to fulfil the needs of society in these areas, and communicate to society how the needs were met. So in reporting about CSR, corporations will have to convey to stakeholders that besides economical, the environmental and social aspects are also included in the business philosophy (Van Marrewijk, 2003).

1

http://www.wbcsd.org/about/organization.aspx

(10)

The way to communicate information to stakeholders is mainly done by the issuance of a report. In the last few decades, reporting about CSR has become a major aspect in terms of the external reporting by companies (James, 2015). A survey that was executed by KPMG (2013) showed that 43% of the biggest corporations compiled out of 41 countries use the collective term of ‘sustainability report’ in their reporting about corporate social responsibility. Corporations that report on sustainability in a formal matter usually do this by means of a detached report that is freestanding of the annual report, and is commonly referred to as a sustainability report.

2.3 The new paradigm of the neoclassical view in CSR

Friedman (1970) saw the maximization of corporate profit as the way for businesses to be social responsible. This view perfectly relates to this study, because it immediately clarifies the fact that social responsibility and profits are seemingly related to one another. Now, more than 45 years later, the issue remains relevant.

The thoughts of Friedman align with the neo-classical view that solely focuses on shareholder interest and therefore looks mostly at maximizing profit. In the view of Friedman, pursuing profits is legitimate as long as corporations stay within “the rules of the game”, so in free competition the use of tactics such as fraud or other kinds of deception are not allowed (Friedman, 1970). Since most listed companies are funded by shareholders (making them the owners), ownership and management will be divided. Management runs the company from day to day, and will therefore gather firsthand knowledge. This knowledge can be used in order to gain personal advantage and impair owner remuneration. The mismatch in the different aims can lead to tension between the managers and the owners of the firm, who cannot monitor all day to day operations and decisions. This tension is materialized by the agency theory in which the shareholders are denoted as the principles and managers identified as the agents (Eisenhardt, 1989; Jensen & Meckling, 1976).

The intention of the agency theory in this principle-agent relation is to define the optimal contract (Eisenhardt, 1989). For both parties this will imply that they will strive for profit maximization, which is consistent with the neo-classical view. The principles will strive for profit maximization in order to achieve the highest yield on their shares. The agents will strive to maximize their compensation and will make choices accordingly (Wallage, 1997). Problems will start to emerge when the optimal outcome for the two parties starts to diverge. In this sense the optimum contract will not be achieved easily, further complicated by the existence of information asymmetry. This implies that the agent will possess more information about the firm than the principle, due to the fact that they run the firm from

(11)

day to day (Eisenhardt, 1989). This asymmetry can be reduced, but that will result in high costs that will further reduce an optimum outcome.

The neo-classical way of thought that is profiled by the agency theory, has been dominant for the better part of the last few decades and has resulted in a lost connection among the corporation and society. To summarize, in this neo-classical way of thought any obligation to improve society will result in a limitation to the organization (Kramer, 2011).

In most organizations, tax is seen as a cost. So consistent with the agency theory, corporations will try to minimize this cost. Christensen and Murphy (2004) state that directors and executives of corporations oftentimes see tax minimization as one of their primary duties. This conflicts with society its point of view, which is argued by Weisbach (2002), who makes a case that activities concerning tax-planning do not bring forth a net social benefit. Tax avoidance by organizations can be classified as costly for society as a whole (Weisbach, 2002). This contradictory is shared by Sikka (2010), who has the concern whether or not the corporation is shrinking taxes in order to maximize shareholder value, and the concern of the community whether corporations are paying their fair share of tax. Summarizing this contradiction leads to a conflict: on the one hand reducing taxes may be in the best interest of shareholders, on the other hand tax avoidance may harm society as a whole (Sikka, 2010).

2.4 The Legitimacy Theory

The legitimacy theory assumes that in order for corporations to exist, they have to comply with the boundaries and norms that are set by society (Brown & Deegan, 1998, p. 22). These expectations by society do not have a permanent nature but tend to adjust and evolve over time, on which corporations need to react and foresee (Brown & Deegan, 1998, p. 22). This dependence on society can be seen as a kind of contract with society, a so-called ‘social contract’ (Hooghiemstra, 2000).

The social contract may restrict corporations to follow a certain course of action to forestall corporate negligence, as perceived by society (Pittroff, 2014, p.401). When society feels that the organization has violated their norms, a gap arises between the corporation and society, which prior literature calls the legitimacy gap (Lindblom 1993, p. 9 ; Deegan & Unerman 2011, p. 329). This failure to comply may result in immense damage reputation-wise (Pittroff, 2014, p.401).

Since tax avoidance practices have become a public eyesore, with as the most recent example the settlement between Google and the British tax authorities concerning a decade of tax avoidance (The Guardian, 2016), it seems like the terms of the social contract have been violated. When the social

(12)

contract has been violated, prior literature by Brown and Deegan (1998, p. 23) and Deegan and Unerman (2011, p. 333) indicates that corporations have several options to reconcile and minimize the reputational damage that has occurred or will occur. One possibility is to report to society about the changes that are being made in the activities of the corporation and what the alterations will ultimately entail, performance-wise. This is to indicate to society that their expectations will be met (Lindblom, 1993; Reverte, 2009; Gray, Kouhy & Lavers, 1995). A different kind of response is to communicate towards society that certain changes are being made, without actually making these changes. A more comprehensive strategy is that management of the corporation will try to alter or mould societal expectations and therefore eliminate the need to make certain alterations to their corporate activities (Dowling & Pfeffer 1975; Lindblom, 1993).

When transferring this way of thought to tax avoidance, the options for corporations become clear. Change your activities and report about it, report changes but do not really change anything or try to mould society its expectations in order to throw society of the scent of tax avoidance.

2.5 Impression management

Corporate social reporting is a communication instrument to enable companies to communicate with society. The aim of this communication is to influence the perception that society has of the company (Hooghiemstra, 2000). A field of study that is in line with these thoughts is impression management, which is defined as: “the conscious or unconscious attempt to control images that are projected in real or imagined social interactions” (Schlenker, 1980). This definition makes clear that impression management is about influencing social interactions.

Although it is originally focused on the individual, in the last couple of decades researchers have started to look at impression management on an organizational level. Organizations use actions that have been deliberately planned in order to influence the perceptions towards themselves held by the audience (Elsbach, Sutton & Principe, 1998, p. 68). The problem that arises with CSR is that the image that is perceived by society may not precisely reflect the actual identity of the company. This CSR image of the organization can be explained as the perception of the audience, with regard to CSR issues ( Tata & Prasad, 2015). This can result in a discrepancy between the perceived and the actual CSR performance. A way to overcome this discrepancy is to use CSR communication (Tata & Prasad, 2015). This means that the organization uses a medium, for example a sustainability report, to establish a better communication about their CSR performance to society.

(13)

When looking at corporate social responsibility, the use of impression management can have a significant impact in two distinguishing ways. Firstly, it can help strengthen the corporations reputation, and secondly it can be used as a way of handling threats of legitimacy that could have a negative impact on the reputation and the stock price of the corporation (Hooghiemstra, 2000).

Since the recent media exposure, like the settlement between the Google and the British tax authorities or the public hearings of Amazon, Starbucks and Google (The guardian, 2016; BBC, 2012), tax avoidance falls in the realm of a threat to legitimacy, and impression management may present the answer. It had also been applied to corporations in the search to explain reactions that corporations make when they are faced with threats of legitimacy (Elsbach & Kramer, 1996; Elsbach & Sutton, 1992).

2.6 Hypotheses development

Prior research (e.g., Hoi, Wu & Zhang, 2013; Huseynov & Klamm, 2012; Watson, 2015) has established a relation between CSR and tax avoidance in US corporations. All found that increased levels of CSR lead to lower tax avoidance. Up until this day, however, the link between social reporting guidelines and tax avoidance has been scarcely made. A way of reporting about CSR is by applying the GRI guidelines.

GRI, which stands for the ‘Global Reporting Initiative’, is an international oriented organization that develops sustainability reporting guidelines for the use of organizations2. GRI is considered to be the most comprehensive guideline concerning sustainability (Toppinen & Korhonen-Kurki, 2013). Furthermore, GRI is regarded as the leading international framework for social and environmental reporting by corporations (Dingwerth & Eichinger, 2010, p. 76).

Since there is a societal displeasure concerning tax avoidance by corporations, the expectation is that companies applying the framework of GRI will feel a higher pressure to reduce tax avoidance practices. In terms of the legitimacy theory they are bound by a social contract when they make the decision to adhere to the GRI guidelines.

Based on the foregoing line of thought, the following hypothesis arises:

Hypothesis 1: Corporations that apply the GRI guidelines avoid less taxes.

As indicated by the legitimacy theory, corporations need to adhere to norms set by society in order to exist (Brown & Deegan, 1998). The expectations molded by these norms shift through the course of

2

(14)

time (Brown & Deegan, 1998). It seems that by looking backwards in time, the neo-classical way of thought was dominant, resulting in the corporate view that taxes were a cost and that minimizing these taxes was a primary responsibility (Christensen & Murphy, 2004). However, the societal norms seem to have shifted concerning minimizing taxes, and now the corporations have violated the societal norms. This violation could result in reputational damages (Pittroff, 2014).

When looking at company age, prior research has suggested that when corporations mature their reputation grows, and can become entrenched (Roberts, 1992). This is in line with the possible effect of legitimacy, that when corporations age, they become better established (Hannan & Freeman, 1984). A violation of the societal norms for older corporations could therefore have a greater impact on their reputation and internal legitimacy, resulting in higher damages.

Based on the above, the following hypothesis arises:

Hypothesis 2: Older corporations avoid less taxes.

Based on the legitimacy theory, society has certain expectations that corporations need to adhere to (Brown & Deegan, 1998). This can be described as a social contract between society and the corporation (Hooghiemstra, 2000). The government can be seen as the spokesperson of society, since it is appointed by society to represent their needs. Since tax avoidance does not bring forth a net social benefit (Weisbach, 2002), it goes against the terms of the social contract.

In this respect, prior research has indicated that a relation exists between government ownership and tax strategy. Chan, Mo and Zhou (2013) found that firms that are government controlled are less tax aggressive than firms without the involvement of the government. This seems to be in accordance with government exerting pressure on corporations to act as better corporate citizens.

Based on the above, the following hypothesis arises:

(15)

3. RESEARCH DESIGN

3.1 Sample selection

The data used to test the hypotheses will be based on an index that is provided by Elsevier.3 This index contains the 500 largest Dutch corporations based on annual revenues and is prepared in cooperation with Bureau van Dijk.4

The procedure of gathering data will be as follows. First, the companies are indicated by the use of the top 500 index by Elsevier. Subsequently, the BvdID5 numbers are manually retrieved from Orbis in order to extract the appropriate financial and other relevant data. In order to get the appropriate BvdID number, the operating revenue in Orbis is converted into millions of euros to match it to the operating revenue as published in Elsevier, to ensure the right legal entity is used. The sample is limited to the companies published by Elsevier for the fiscal year 2014 and extended back in time until the fiscal year of 2010. Of the 500 companies, 15 companies are dropped since the operating revenue cannot be matched with available data in Orbis. This ultimately results in 485 unique companies and 2425 firm-year observations. Since the sample does not contain all the corporations with government ownership, for the fiscal years 2010-2014, 39 additional corporations are identified. The companies that are not readily incorporated in the sample, are added. This results in an addition of 23 companies and 115 firm-year observations.

The financial data, including the necessary data for the independent variables, the control variables and partially for the dependent variable, are gathered from Orbis. The remaining dependent variable information, special items and paid cash tax, are gathered from Compustat Global. The government ownership data in Orbis is not sufficient and therefore extracted from the annual reports, ‘Beheer Staatsdeelnemingen’, from the ministry of finance. The application of GRI guidelines for the companies is manually obtained from the sustainability disclosure database of the global reporting initiative.6

3_{Elsevier, “Top 500 bedrijven in Nederland,” 2015,}

http://www.elsevier.nl/economie/article/2015/11/elseviers-top-500-de-grootste-bedrijven-van-nederland-2716441W/

4_{Bureau van Dijk is a company that specializes in analyzing and comparing companies. They’re information contains} ownership structures, financial data and management details among others. See; http://www.bvdinfo.com/en-gb/home

5

BVDID number stands for the Bureau van Dijk ID number. The ID number consists of the chamber of commerce identification number that is unique for the company.

6

(16)

Two companies are excluded from the sample due to specific reasons. Firstly, Royal Dutch Shell is excluded due to the fact that Shell operates in the oil and gas industry, an industry that is subjected to industry-specific taxes. Due to the global presence of Shell including countries with a higher statutory tax rate than the Netherlands, this results in an above average effective tax rate.7 Due to the size of Shell in comparison to the other companies, this would result in skewing the findings. The second company is ABN AMRO. The reason for excluding ABN AMRO is because this is a special case due to the nationalization in 2008 and the large losses they incurred during the financial crises. These losses combined with limited data result in deceptive tax rates.

In accordance with previous research (e.g, Hoi et al., 2013; Huseynov & Klamm, 2012; Lanis & Richardson, 2012a; Watson, 2015;), excluded from the original sample are (i) firm-year observations with a negative pre-tax income, (ii) firm-year observations with a tax refund, (iii) firm-year observations with missing key Orbis data, taxation and pre-tax income, and (iv) firm-year observations with an ETR that exceeds one. The resulting final sample is consistent of 1601 firm-year observations from 356 unique companies. Table 1 presents the steps taken in order to arrive at the final sample.

Table 1: Sample selection

Firm-year observations

Maximum observations based on index 2500

Potentially valid observations in Orbis 2425

Government ownership observations not readily incorporated 115

2540 Less:

Observations specifically indicated (Shell and ABN AMRO) 10

Observations with negative pre-tax income 63

Observations with a tax refund 127

Observations with missing data (taxation and pre-tax income) 681

Observations with an ETR exceeding one 58

Remaining observations for final sample 1601

3.2 Dependent variables

For the purpose of this study, several measures of tax avoidance will be used. Prior literature has developed several ways to identify tax avoidance, each with their strengths and weaknesses (Hanlon & Heitzman, 2010). The main proxy used for measuring tax avoidance will be the book effective tax rate (B-ETR), following the research of McGuire, Omer and Wang (2012). This will be the starting

7

(17)

point in examining tax avoidance. Other proxies used, are the cash effective tax rate (C-ETR) and the long term cash effective tax rate (LC-ETR), consistent with research done by Dyreng et al. (2008); Hanlon and Heitzman, (2010). Below, each of the proxies will be discussed.

The book effective tax rate, also referred to as the current effective tax rate, can be described as “current tax expense per dollar of pre-tax book income” (Hanlon & Heitzman, 2010). This measure is a prominent proxy in prior tax avoidance research and is used to measure the tax burden of a firm (Dyreng, Hanlon & Maydew, 2010; Robinson, Sikes & Weaver, 2010). The book ETR is measured over a one-year period. An advantage of using book ETR is the fact that it is less volatile from year to year compared to other proxies for tax avoidance, like cash ETR. This is because book ETR looks at changes of net income (so direct changes in tax liability). A major drawback that comes with this is that deferred taxes that are shifted through time are not contained in this measure (Hanlon & Heitzman, 2010).

Because of the shortcomings of book ETR, the so-called cash ETR will also be included in this study. Following the line of thought of Dyreng et al. (2008), the cash ETR entails two alterations made to the book ETR. First of all, instead of using the nominator of tax expense, cash ETR uses the actual cash taxes that are paid. In this way cash ETR also incorporates strategies of tax avoidance that defer taxes to different time periods, but simultaneously do not influence the tax expenses that are recorded in the financial statements (McGuire et al., 2012). Second in the denominator, instead of just using pretax income as a whole, in cash ETR special items are excluded from this figure (Dyreng et al., 2008). The negative aspect of this way of measuring is that it has more volatility from year to year than the book ETR, because it is based on the actual tax payments made each year (Hanlon & Heitzman, 2010). Another point of concern in this matter, as appointed by Dyreng et al. (2008), is that it may contain tax settlements that relate to other years.

In order to overcome the limitation of increased volatility, the long-term cash ETR, which looks at longer time periods, will be used in this study. In order to calculate the long-term cash ETR the cash taxes paid over a n-year period are added and the total is divided by the sum of the pre-tax income over a n-year period, while excluding special items (Dyreng et al., 2008). There will, however, be two major limitations in using this measure. First of all, it is hard to make the distinction between tax favored activities that the firm exploits and have nothing to do with tax avoidance, and behavior that is consistent with tax avoidance strategies (Hanlon & Heitzman, 2010). The reason for this is because the long-term cash ETR looks at the ratio between the cash taxes paid and the pre-tax income, and

(18)

consequently will take into account both the activities, without taking into account the underlying rationalization.

A second point of concern is the fact that corporations may reduce the pre-tax income to minimize tax payments. The long-term cash ETR measure is unable to distract this (Hanlon & Heitzman, 2010). The last point stated may be a major concern in this study, since Hanlon and Heitzman (2010) identified that the tax avoidance strategy of reducing pre-tax income is specifically common in corporations that are under less pressure to manage accounting income upwards, which is mainly consistent with private firms. Since a major part of the sample consists of private firms, this may lead to limitations in the outcomes.

Furthermore, all three measures of the effective tax rate as stated above are multiplied by minus one. This is in accordance with prior research (Landry, Deslandes & Fortin, 2013;Lanis & Richardson, 2012a), in order to obtain an increasing measure for tax avoidance. This can be illustrated by the following example, while using the book ETR:

[

Tax expense

Pre-tax book income

]

2.131.000.000

7.646.000.000 = 0.28

- [

Tax expense

Pre-tax book income

]

2.131.000.000

7.646.000.000 = - 0.28

-0.28 -0.18 +

-0.28 -0.38 -

Because of the multiplication with minus one, the effective tax rate becomes negative. When the measure then increases (+) it comes closer to zero, indicating a higher degree of tax avoidance. When the measure decreases (-) it becomes more negative, indicating a lower degree of tax avoidance. By multiplying the effective tax rate by minus one an increasing measure for tax avoidance is obtained making the results easier to interpret.

(19)

3.3 Independent variables GRI

GRI will be measured by looking at the application of the GRI guidelines by the corporations that are included in the sample. In order to empirically test the third hypothesis, GRI adherence will be checked by looking up the corporations in the GRI database.8 By manually checking each corporation, it can be traced year by year if the corporation was openly adhering to GRI and which version was used as the guideline concerning CSR reporting. The database is set to find Dutch corporations in order to select the appropriate entity. In case a certain corporation is not traceable in the database, the corporation in question is researched in order to establish if it has a foreign parent company. If this is the case then the application of the GRI guidelines of the foreign parent is included. The rationale behind this is that the subsidiaries of multinational corporations gain internal legitimacy when they implement practices that are in line with those of the parent corporation (Davis, Desai & Francis, 2000; Hillman & Wan, 2005). This combined with consolidated reporting will result in corresponding practices in term of reporting.

According to the sustainability disclosure index, there are different versions of application of the GRI guidelines. The sample shows that there is a distribution between the three latest versions, called G3, G3.1 and G4. For the hypothesisin this study, no difference will be made regarding the version that is adhered to, in order to only test the impact that the application of GRI has. In order to run the regression, GRI application is made into a dummy variable. 0 indicates no GRI application and 1 implicates that the company adheres to it.

Government ownership

Government ownership will be measured by looking into whether the Dutch government has ownership in the specific corporations. The data will be obtained from information made available by the ministry of finance in the annual reports of state holdings, ‘Beheer Staatsdeelnemingen’.9

The information is not obtained from Orbis, because ownership structures in Orbis are harder to interpret and are less complete than the ownership as stated in the annual reports. In order to run the regression, government ownership was made into a dummy variable. 0 means no government ownership and 1 means that there is government ownership.

8

https://www.globalreporting.org/information/about-gri/Pages/default.aspx

9

(20)

Corporate age

The age of a corporation can be measured by the number of years it has existed since the date of its foundation (Monteforte & Stagliano, 2015). However, accurately measuring the age of a corporation remains difficult due to mergers, significant changes in business activities or reorganizations (Roberts, 1992). In order to circumvent these possibilities the date of incorporation will be used. The date of incorporation stands for the date the legal entity was established in the Netherlands.

A different reason for using the date of incorporation and not the founding year of the organization, is that using the founding year in the sample would result in difficulties in case the specific company was a subsidiary of a foreign parent company. If the founding year was selected, the age of the foreign parent company would have been selected even though this specific corporation was not even represented in the Netherlands at that time. To overcome this problem and use the most appropriate way to measure corporate age, the date of incorporation is selected.

This proxy for corporate age is in accordance with prior research (e.g., Loderer & Waelchli, 2009; Roberts, 1992). The information will be extracted from Orbis by looking at the date of incorporation and subtracting this from the fiscal year of the observation. For example: a date of incorporation of 1989 combined with an observation of 2014, will result in a company age of 2014-1989= 25. In case of missing data, the date of incorporation is manually retrieved by consulting the website of the ‘Kamer van Koophandel’10

, “the chamber of commerce”. 3.4 Control variables

In accordance with prior research, a variety of control variables will be included. These will include the return on assets (ROA), leverage (LEV), the natural log of intangible assets (INT), and the natural log of assets (ASSETS). The next section will underpin the use of these specific control variables. Return on assets

The return on assets is a variable that measures the profitability of a firm. Prior research (e.g., Chen, Chen, Cheng & Shevlin, 2010; Francis, Hasan, Wu & Yan, 2014; Frank, Lynch & Rego, 2009; McGuire et al., 2012) found a positive relation between the profitability of a corporation and the likeliness of engaging in tax avoidance behavior. This can be rationalized by the fact that more profitable companies have more to gain from tax avoidance. The return on assets (ROA) will be determined by taking the pre-tax income and dividing it by the total assets of the firm.

10

(21)

Leverage

Leverage is a variable that looks at how the assets of a firm are funded. This can be done by equity and debt. A higher leverage ratio means a higher percentage of the corporation its activities will be funded by debt. Because of tax deductibility of interest expenditures, a report of the OECD stated that interest is one of the easiest ways to shift profits (OECD, 2014). This statement is consistent with prior research (e.g., Dyreng et al., 2008; Graham, 2003; McGuire et al., 2012), which found that an increased leverage percentage is positively associated with tax aggressiveness. For this reason the leverage percentage will be used as a control variable in conducting this study. LEV will be determined by dividing the long-term debt by total assets.

Intangible assets

In this study intangible assets will be used as a control variable, because intangible assets create an opportunity for corporations to minimize their taxable income. Measuring intangible assets will be done by using the natural logarithm of the intangible assets. Prior research by Hanlon, Mills and Slemrod (2007) found that firms who have more intangible assets, can deploy more options to avoid taxes. Therefore, consistent with prior research (e.g., Frank et al., 2009; Hanlon et al., 2007; Wilson, 2009), intangible assets will be controlled for. INT will be determined by the natural logarithm of the intangible assets.

Firm size

The size of a company has been found to influence the tax avoidance behavior that corporations exercise (Frank et al., 2009; Zimmerman, 1983). Zimmerman (1983) stated that larger firms have more power as compared to smaller firms, and hereby tend to show higher levels of tax avoidance. A viable explanation for this is provided by Huseynov and Klamm (2012), who stated that larger firms may possess more tax expertise and use this in order to reduce their tax expenses. In this study, company size will be included as a control variable. The size (SIZE) will be determined by the natural logarithm of the total assets.

(22)

Table 2 provides an overview of the used independent and control variables that are applied in the study. All the measures are defined in accordance with previous literature.

Table 2: Constructs of the independent and Control Variables Independent

variables

Control variables

Measurement Global reporting initiative

guideline application

GRI Dummy variable equal to 1

if company applies GRI, 0 otherwise

Corporate age AGE Number of years since

incorporation

Government ownership GOV Dummy variable equal to 1

if government has

ownership in the company, 0 otherwise

Profitability ROA Pre-tax income / Total

assets

Company size SIZE Natural logarithm of total

assets

Intangible assets INT Natural logarithm of

intangible assets

Leverage LEV Long-term debt / Total

assets

3.5 Models

In order to empirically test the hypotheses, the following regression models will be used: Model 1: TA = ß1 GRI + ß2 ROA + ß3 LEV + ß4 INT + ß5 SIZE + ɛ

Model 2: TA = ß1 AGE + ß2 ROA + ß3 LEV + ß4 INT + ß5 SIZE + ɛ

Model 3: TA = ß1 GOV + ß2 ROA + ß3 LEV + ß4 INT + ß5 SIZE + ɛ

The first model describes the relation between the application of GRI guidelines and tax avoidance. The second one describes the association between corporate age and tax avoidance. The third model describes the relation between government ownership and tax avoidance.

In the models, (ß) stands for beta represents the coefficient of the regression. The error term of the

(23)

4. RESULTS

4.1 Descriptive statistics

The descriptive statistics are presented in table 3. The government share is made into a dummy variable so the variation is between 0 and 1. The use of a dummy variable can be explained by elimination of a large portion of the firm year observations because of missing data concerning the (partially) government owned corporations. The measure used to proxy for tax avoidance is multiplied by minus one, as previously explained. This results in a minimum of -1 and a maximum of 0. -1 means an ETR of a hundred percent, while 0 means that no taxes have been paid. The minimum age of a corporation in the sample is 25 years, while the maximum is at 291 years. GRI is also a dummy variable and resulting from the descriptive statistics 23.36 percent of the corporations from the sample apply the GRI guidelines.

Table 3: Descriptive Statistics

Variables N Mean Minimum Median Maximum Standard deviation

TAX AV (B-ETR) 1601 -0.2489 -1 0.2488 0.0000 0.13771 GRI 1601 0.2336 0 0 1 0.42326 AGE 1601 39.0087 0 25 291 40.6261 GOV 1601 0.0312 0 0 1 0.17399 ROA 1601 0.0944 -0.2039 0.0614 2.0697 0.1407 SIZE 1601 13.0111 4.9492 12.7507 17.6873 1.5440 INT 1222 9.5722 0.3001 9.6807 16.9144 2.8520 LEV 1130 0.1695 0.0000 0.1305 1.0547 0.1574

TAX AV (B-ETR)= tax avoidance, measured by multiplying the book effective tax rate by minus one; GRI= global reporting initiative guideline application, measured by dummy variable equal to 1 if company applies GRI, 0 otherwise; AGE= corporate age, measured by number of years since incorporation; GOV= government ownership, measured by dummy variable equal to 1 if government has ownership in the company, 0 otherwise; ROA= profitability, measured as pre-tax income / total assets; SIZE= company size, measured as the natural logarithm of total assets; INT= intangible assets, measured as the natural logarithm of intangible assets; LEV= leverage, measured as long-term debt / total assets

4.2 Correlation results

Looking at existing associations between the variables, an analysis of correlations is performed. Table 4 presents the Pearson correlation matrix. To improve interpretability, the book ETR (B-ETR) is multiplied with minus one (this is further elucidated in the method section). The table shows a positive

(24)

correlation between tax avoidance practices and corporate age, which is also the case for the correlation between tax avoidance practices and GRI application. Both of these coefficients were deemed insignificant (p = 0.226 and p = 0.690). However, government share can be deemed negatively significant on a 0.01 significance level, which is what was expected. Furthermore, positive significant associations were found with ROA and SIZE.

Table 4 indicates that there may be problems concerning multicollinearity, because the correlation between intangible assets and size is rather high (0.670). In order to determine if multicollinearity poses to be a problem, the variance inflator factors (VIF) arecalculated. When establishing the extent of multicollinearity, oftentimes the greatest value amongst the independent variables will be used (Neter, Kutner, Nachtsteim & Wasserman, 1996). A value for multicollinearity that exceeds 10 is seen as problematically affecting the estimate for the least square. For this study the largest VIF equals 2.182, meaning that multicollinearity will not pose a problem.

4.3 Multivariate analysis

A linear regression is used to examine the relation between the book ETR (B-ETR), as a proxy for tax avoidance and the characteristics: GRI application, corporate age and government ownership. Table 5 shows the regression results. For hypothesis 1 the model shows an adjusted R² of 0.059 and the results appear significant at a high degree (F= 12.672, p < 0.001). The results show a negative relationship (β = -0.015) between GRI and book ETR (at a 10 percent level), indicating companies that are applying GRI guidelines will avoid less taxes. This is in line with the expectation that applying GRI guidelines leads to a perceived pressure felt by corporations to uphold the terms of the social contract made with society. Concerning the control variables, it can be noted that ROA and SIZE have a significant positive association with book ETR (p < 0.001 and p < 0.001). These findings support prior research (Frank et al., 2009; Huseynov & Klamm, 2012). INT has a significant negative association with book ETR (p < 0.001), which is in contradiction with previous research (e.g., Wilson, 2009; Frank et al., 2009; Hanlon et al., 2007). LEV showed an positive association with book ETR (β = 0.003) which is in line with previous research (e.g., Graham, 2003; Dyreng et al., 2008; McGuire et al., 2012). However, the association is insignificant (p = 0.917).

For hypothesis 2 the model shows an adjusted R² of 0.057 and the model appears significant on the 1 percent level (F = 12.318, p < 0.001). These results indicate that the amount of explained variation of hypothesis 2 is 0.2 percent lower than for hypothesis 1. The results show a negative relation (β = -0.001) between AGE and book ETR, indicating that companies that have existed for more years avoid less taxes. However the result cannot be statistically substantiated, given the coefficient for AGE is

(25)

not significant (p = 0.967). For the control variables the results show a significant positive relation for ROA and SIZE in association with book ETR (p < 0.001 and p < 0.001). INT has a significant negative association with book ETR (p < 0.001). LEV showed a positive association with book ETR (β = 0.003), but is not significant (p = 0.913).

For hypothesis 3 the model shows an adjusted R² of 0.085 and thus appears to be significant on the 1 percent level (F =18.287, p < 0.001). The results indicate that the amount of explained variation for hypothesis 3 is 2.6 percent higher than for hypothesis 1 and 2.8 percent higher than for hypothesis 2. This result indicates that government ownership explains more in the variation of tax avoidance than corporate age and GRI application. A negative coefficient was expected for GOV in order to test hypothesis 3. This would imply that government ownership in a firm would result in lower tax avoidance. As was predicted, the regression reports a negative relation (β = -0.140) between GOV and book ETR, significant on the 99 percent confidence level (p < 0.001). The findings support research by Chan et al. (2013) who found that government controlled companies show less tax aggressiveness than firms without government involvement. Concerning the control variables the model reports results that correspond to the prior two hypothesis: ROA and SIZE associate with book ETR on the one percent level (p < 0.001 and p < 0.001). INT is negatively associated with book ETR (p < 0.001). LEV has a positive association with book ETR, but is not significant (p = 0.486).

(26)

Table 4: Pearson Correlation – Sig (2-tailed) N=1601 TAX AV

(B-ETR)

AGE GOV GRI ROA SIZE INT LEV

TAX AV (B-ETR) 1 AGE 0.030 1 GOV -0.111*** 0.029 1 GRI 0.010 0.232*** 0.105*** 1 ROA 0.095*** -0.029 0.122*** -0.019 1 SIZE 0.087*** 0.185*** 0.175*** 0.344*** -0.099*** 1 INT -0.032 0.104*** 0.019 0.187*** -0.138*** 0.670*** 1 LEV 0.006 -0.045 0.119*** 0.074** -0.054 0.255*** 0.267*** 1

***Correlation is significant at the 0.01 level ** Correlation is significant at the 0.05 level * Correlation is significant at the 0.10 level

(27)

Table 5: Multiple Regression Results

Hypothesis 1. Hypothesis 2. Hypothesis 3.

TAX AV (B-ETR) TAX AV (B-ETR) TAX AV (B-ETR) GRI -0.015* (-1.289) AGE -0.001 (-0.042) GOV -0.140*** (-5.494) ROA 0.254*** 0.244*** 0.266*** (4.181) (4.048) (4.462) SIZE 0.029*** 0.027*** 0.031*** (6.743) (6.546) (7.669) INT -0.011*** -0.011*** -0.012*** (-4.712) (-4.620) (-5.341) LEV 0.003 0.003 0.022 (0.104) (0.109) (0.697) Constant -0.563*** -0.542*** -0.575*** (-12.049) (-12.261) (-13.271) Observations 1601 1601 1601 Adjusted R-square 0.059 0.057 0.085 F 12.672*** 12.318*** 18.287***

Absolute value of t statistics in parentheses ***Relation is significant at the 0.01 level ** Relation is significant at the 0.05 level * Relation is significant at the 0.10 level

Significance levels are one-tailed for directional hypotheses, two-tailed otherwise

(28)

4.4 Sensitivity analysis

For the main hypotheses the book effective tax rate (B-ETR) is used as a proxy for tax avoidance, which is in accordance with research done by McGuire et al. (2012). However, as explained in the method section, the book ETR has a major drawback, as it looks at direct changes in tax liability hereby not including deferred taxes (Hanlon & Heitzman, 2010). To overcome the drawback posed by not including deferred taxes, the cash effective tax rate (C-ETR) can be used to proxy for tax avoidance. However, this will lead to more volatility from year to year (Hanlon & Heitzman, 2010). To overcome the year to year volatility of the cash ETR a longer time period proxy for tax avoidance can be used, which is called the long-term cash effective tax rate (LC-ETR).

Due to the described circumstances, the cash ETR and the long-term cash ETR will be used to perform a sensitivity check of the main proxy for tax avoidance, the book effective tax rate.

The descriptive statistics in tables 6 and 7 show that the book ETR has the highest mean (-0.2489 ) as compared to the cash ETR (-0.2290) and long-term cash ETR (-0.1958). This indicates that book ETR shows the least amount of tax avoidance for the specific sample. All three measures are used to proxy for tax avoidance and have been multiplied by minus one, as previously explained. The overall spread of the measures is smaller for the cash ETR and long-term cash ETR, which could be due to the smaller sample size (N = 127 and N = 85). The smaller sample sizes for cash ETR and long-term cash ETR can also be the explanation for other smaller variations observed, for example: intangible assets range from 0.3001 to 16.9144 in the book ETR sample and from 7.95 to 16.90 in the cash ETR sample. For the cash ETR and long-term cash ETR descriptive government ownership (GOV) is excluded. The reason is that there are only three firm year observations with government ownership in the sample of the cash ETR and none in the sample of the long-term cash ETR. This low amount of observations would make it impossible to draw a conclusion. Overall, in comparing the descriptive statistics of the three measures used, no major differences are reflected, showing a first indication that book ETR can be used as a reliable proxy for tax avoidance in this specific sample.

(29)

Variables N Mean Minimum Median Maximum Standard

deviation TAX AV (C-ETR) 127 -0.2290 -0.81 -0.23 -0.01 0.13832 GRI 127 0.5039 0 1 1 0.50196 CORP-AGE 127 80.7165 5 75 249 60.08696 GOV - - - - - - ROA 127 0.0754 -0.01 0.0700 0.24 0.04785 SIZE 127 14.7436 11.54 14.7100 17.68 1.53568 INT 127 12.9739 7.95 12.9600 16.90 2.02222 LEV 122 0.1820 0 0.1700 0.55 0.10995

TAX AV (C-ETR)= tax avoidance, measured by multiplying the cash effective tax rate by minus one; GRI= global reporting initiative guideline application, measured by dummy variable equal to 1 if company applies GRI, 0 otherwise; AGE= corporate age, measured by number of years since incorporation; GOV= government ownership, measured by dummy variable equal to 1 if government has ownership in the company, 0 otherwise; ROA= profitability, measured as pre-tax income / total assets; SIZE= company size, measured as the natural logarithm of total assets; INT= intangible assets, measured as the natural logarithm of intangible assets; LEV= leverage, measured as long-term debt / total assets

Variables N Mean Minimum Median Maximum Standard

deviation TAX AV (LC-ETR) 85 -0.1958 -0.53 -0.20 -0.02 0.09739 GRI 85 0.5176 0 1 1 0.50265 CORP-AGE 85 81.9765 12 54 249 59.84364 GOV - - - - ROA 85 0.0720 0.01 0.07 0.22 0.04143 SIZE 85 14.8194 11.54 15.10 17.68 1.60838 INT 85 13.1492 7.95 13.28 16.90 2.09727 LEV 82 0.1816 0 0.16 0.55 0.11778

TAX AV (LC-ETR)= tax avoidance, measured by multiplying the long term cash effective tax rate by minus one; GRI= global reporting initiative guideline application, measured by dummy variable equal to 1 if company applies GRI, 0 otherwise; AGE= corporate age, measured by number of years since incorporation; GOV= government ownership, measured by dummy variable equal to 1 if government has ownership in the company, 0 otherwise; ROA= profitability, measured as pre-tax income / total assets; SIZE= company size, measured as the natural logarithm of total assets; INT= intangible assets, measured as the natural logarithm of intangible assets; LEV= leverage, measured as long-term debt / total assets

(30)

In order to see if existing associations between the variables changed by using a different measure for tax avoidance, an analysis of correlations has been performed. Table 8 and 9 present the Pearson correlation matrixes. The tables show a negative correlation between tax avoidance practices and corporate age. This seems to contradict the previous result where a positive association is found. However, the coefficients are also deemed insignificant (C-ETR, p = 0.721 and LC-ETR, p = 0.130). Tax avoidance and GRI application showed a positive correlation for both measures, consistent with the previous result. However, the coefficients are deemed insignificant (C-ETR, p = 0.280 and LC-ETR, p = 0.976). Furthermore, positive significant associations are found for ROA at the 99 and 95 percent confidence level (C-ETR, p < 0.01 and LC-ETR, p < 0.05) in line with the previous result. For size the results show a deviation from original result. For book ETR a positive significant association is found at the one percent level. The results for the other two measures, however, show no significance (C-ETR, p = 0.697 and LC-ETR, p = 0.213), and for the long-term cash ETR the association becomes negative. The association for INT remains negative and even becomes significant at the one percent level for the long term cash ETR ( p = 0.005). Noticeable about this is the contradiction it poses to previous research that indicates a positive association between tax avoidance and the intangible assets. The association for LEV remains consistent for all the measures of tax avoidance and shows no significance (C-ETR, p > 0.1 and LC-ETR, p > 0.1).

Table 10 shows the regression results while using three distinct measures to proxy for tax avoidance. The model shows an adjusted R² for measure (B-ETR), (C-ETR) and (LC-ETR) of respectively 0.059, 0.175 and 0.139 for hypothesis 1 and 0.057. 0.178 and 0.127 for hypothesis 2. These results indicate that when using the cash effective tax rate as a proxy for tax avoidance, the used variables explain 17.5 percent of the variation in tax avoidance for hypothesis 1. Prior research found that the cash effective tax rate and the long term cash effective tax are better measures for tax avoidance (see method section). The model seems to validate those findings since the variation in tax avoidance is better explained by using the cash ETR (17.5 percent) and the long-term cash ETR (13.9 percent) compared to using the book ETR (5.9 percent). Due to data restrictions, the hypotheses were tested using the book effective tax rate. The regressions will now be re-run using the other measures of tax avoidance in order to establish if the result will significantly differ, keeping in mind the reduced sample sizes. In order to evaluate the multivariate analysis, the main results (table 5) were included in table 10 in order to provide a good overview of the differences resulting from using different measures to proxy for tax avoidance.

For hypothesis 1, which looks at the impact of GRI application on tax avoidance, the results significantly differ between the measures used. For measure one applying the book ETR the results show a negative relationship (β = -0.015) significant at the ten percent level. The significance is lost when measure two and three are applied (p = 0.462 and p = 0.268) and the relation becomes positive

(31)

(β = 0.019 and β = 0.025). A possible explanation for the conflicting results could be due to the relative limited sample sizes for measures two and three (N = 127 and N = 85 ) resulting in skewed observations. For the second hypothesis, which looks at the impact of corporate age on tax avoidance, the results for measure one and three are consistent (β = -0.001 and β = -0.00007) and show no significance. For measure two using the cash ETR the relation becomes positive (β = 0.0002) and remains insignificant. These results could also be due to the increased volatility of the cash ETR as indicated by Hanlon and Heitzman (2010). Hypothesis 3, which looks at the impact of government ownership, could not be tested due to data restriction resulting in three observations relating to measure two and none for measure three. Relating to the control variables ROA, SIZE, INT and LEV, the results seem to correspond between the measures. One difference is in the ROA for hypothesis 2. For measure one and two the ROA is positively related to tax avoidance and highly significant (on the one percent level), however for measure three the relation remains positive but is deemed insignificant (p = 0.149).

Social reporting and company determinants impacting corporate tax avoidance in the Netherlands