Amsterdam Business School
The effect of corporate ownership structures on tax avoidance
Name: Job Eshuis
Student number: 10184252
Thesis supervisor: dr. Alexandros Sikalidis Date: 20-06-2016
Word count: 17563
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
1
Statement of Originality
This document is written by student Job Eshuis who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources
other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 Abstract
Previous studies state that tax avoidance is beneficial to shareholders. However, this point of view ignores the potential non-tax costs arising from tax avoidance. Examples of these non-tax costs can be litigation costs or reputational damage. The decision on whether to avoid taxes or not comes with potential agency conflicts between controlling and minority shareholders. This research examines three types of ownership structures: family, government and foreign
ownership. I examine tax avoidance by measuring the effective tax rate. In a setting of German listed firms, I find that family ownership has a positive effect on the effective tax rate and therefore a negative effect on tax avoidance. On the contrary, government ownership has a negative effect on the effective tax rate, while for foreign ownership no clear relation can be determined. Using a matched sample of private firms, these results also hold for private
companies. However, the expected negative effect of foreign ownership on the effective tax rate is significantly present with private firms.
These results suggest that some controlling shareholders, families, are willing to forgo the benefits from tax avoidance in order to protect themselves against potential non-tax costs. On the other hand, for government ownership and foreign ownership, the controlling shareholders emphasize more on the benefits of tax avoidance. This study contributes to the understanding of tax avoidance, ownership structures and the relationship between the two.
Key words: Tax avoidance, ownership structures, family ownership, government ownership,
3 Table of Contents
1 Introduction ... 4
1.1 Background ... 4
1.2 Research question ... 6
1.3 Motivation and Contribution ... 6
2 Literature review ... 8
2.1 Concepts and definitions ... 8
2.1.1 Tax avoidance ... 8
2.1.2 Corporate ownership structures ... 9
2.2 Theory ... 14 2.2.1 Agency theory ... 14 2.2.2 Corporate Governance ... 15 2.3 Hypotheses ... 17 3 Research Methodology ... 19 3.1 Sample Selection ... 19
3.2 Ownership and industry distribution ... 20
3.3 Matched sample ... 24
3.4 Research design ... 27
3.4.1 Tax avoidance measurement ... 27
3.4.2 Model and variables ... 29
4 Empirical results ... 35
5 Sensitivity analysis ... 41
5.1 Elimination of negative pre-tax income ... 41
5.2 Private ownership structures ... 45
6 Conclusion ... 50
7 References ... 52
4
1 Introduction
In this paragraph, the research topic is introduced. At first, a general description of the topic and the background is provided. Secondly, the research question is stated. This research question is based on the information given in the background section. At last, the motivation for this topic and research question is explained, together with the contribution to existing literature.
1.1 Background
There is widespread interest about the determinants, magnitude and consequences of tax avoidance in the recent accounting research field (Desai, Dyck and Zingales, 2007; Hanlon and Heitzman, 2010; Jacob, Rohlfing-Bastian and Sandner, 2014). Research has been done about the determinants of tax aggressiveness and the question why tax sheltering isn’t used even more (Shackelford and Shelvin, 2001; Weisbach 2002). However, a lot of subjects have still not been researched sufficiently yet, or are not researched at all. In review papers, tax avoidance is getting a prominent place in the future research questions. Research also calls for a more elaborate investigation in the subject, since that is lacking (Shackelford and Shelvin, 2001; Hanlon and Heitzman, 2010). The main reason for this lack of research is the absence of a clear universally accepted definition. For example, the concept tax avoidance is used in a lot of different ways. The term means therefore different things to different people (Hanlon and Heitzman, 2010). The interesting challenge for this research is therefore to set a clear definition, making sure that the results can be interpreted correctly.
Tax avoidance is a topic that is not only of interest in accounting research. The public opinion has become more important over the years, especially with the disclosures of certain tax avoiding companies, which are well known to the public. Examples of these companies are Starbucks, Google and Amazon. Because companies are interested in the public opinion, and therefore the image they have, they aren’t avoiding these taxes in obvious ways. However, a lot of companies still avoid taxes. It is interesting to understand what types of firms avoid taxes, and how they avoid them. As stated by Chen, Chen, Cheng and Shevlin (2010), the government takes more than one-third share of the pre-tax profits of a regular firm. Therefore, this tax cost is huge. Of course, this differs per region, but overall, tax is a huge burden for companies. Given this significance, it is likely that shareholders and owners want to use tax aggressiveness or avoidance to lower these costs. However, this leaves out the assumption that tax aggressiveness or avoidance also comes with non-tax costs. Chen et al. (2010) therefore researched the amount of these costs, and the effect on the amount of avoidance firms use. Non-tax costs are for example litigation costs, damage to corporate image and brand costs (Scholes, Wolfson, Erickson,
5 Maydew and Shevlin, 2005).
The question arises, why the concept of tax avoidance is attracting more and more attention of accounting researchers. Above, I described the dynamics of the public opinion. Another way to look at this increase in interest is by looking at the consequences. The consequences of tax avoidance can be direct or indirect. Direct consequences can be for example a deduction of an expense, which normally is deductible, and therefore increases the cash -flow and the wealth of the investor. Indirect consequences are for example that this increase in deduction lowers the interest tax shield, that should say the marginal benefit of this shield, and therefore could change firm’s decisions on their capital structure (Hanlon and Heitzman, 2010; Graham and Tucker, 2006).
Another way to look at the consequences is to examine the company specific consequences. These consequences can be focused on shareholders, creditors, the government and all other stakeholders. Consequences for shareholders and managers can be thought of by the principal-agent theory, wherein the manager’s interest should be aligned with the shareholders interest. Avoiding taxes is at interest of shareholders, because it increases their profit and/or firm value (Desai and Dharmapala, 2009). Summarizing, there are a lot of potential consequences of tax avoidance, and therefore the determinants should be researched well, to create a clear picture.
As stated above, there are a lot of determinants of tax avoidance. Examples could be board structure, the scale of the international operations, auditor characteristics and manager characteristics. In this research, I focus on the relation between corporate ownership structures and tax avoidance. The focus of the research is therefore the effect of different types of corporate ownership structures, for example family companies, government ownership and foreign ownership. Shackelford and Shevlin (2001) called for research on this determinant, because in their view it was on the one hand a very important determinant which was on the other hand understudied. Ownership structures can influence tax avoidance due to their alignment of ownership and control. This alignment, if even present, influences the way managers and owners avoid taxes. Therefore, it is interesting to investigate to what level these structures might have influence on tax avoidance. This study seeks to examine this relation, to give insight in whether this relation is present and in what forms.
6 1.2 Research question
The research question of this paper is:
‘What is the effect of corporate ownership structures on tax avoidance?’
In this research, I therefore examine different types of corporate ownership structures. The research looks for a relation between what ownership structure a company has and how this affects the magnitude of tax avoidance.
1.3 Motivation and Contribution
This research contributes to existing knowledge from both a scientific and societal point of view. The scientific contributions are however more relevant. This research looks to expand the existing literature. As stated in the background, tax avoidance is becoming a more and more relevant topic in accounting research. The determinants of tax avoidance have not been studied sufficiently yet. Although there is knowledge about what determinants there are, the effects of those determinants is not yet covered broadly in literature. In this research, I focus on the corporate ownership structures.
Although there have been a few studies on this determinant, these studies are very time and country specific. This study provides a new view, investigating a new setting, both time and country wise. The new setting is conducting the research for companies in Germany, a Western country. Previous research into corporate ownership structures, especially in their relation with tax avoidance, have been primarily conducted in Eastern countries like China, Malaysia etcetera (Chen et al. 2010; Mahenthiran and Kasipillai; Chan, Mo and Zhou, 2013; Annuar, Salihu and Obid, 2014). As these authors also point out, those specific settings may lead to the untransferability of their results into other settings. Therefore, no clear picture can be created into this relationship in the Western world.
The second reason to choose Germany as country is the transferability of the results into other Western countries, which is one of the main objectives of the research. Germany has incorporated the German civil law system, as defined by La Porta, Lopez-De-Silanes, Shleifer and Vishny (2000). The German civil law system is used by a lot of central European countries, and is comparable with the French civil law system; together the most used models throughout Europe. Civil law systems are most commonly used, and distinguish themselves from common law systems used in the United States and the United Kingdom. Because of this aspect of Civil law systems, the results should be transferable to other Civil law systems (La Porta et al. 2000). Adding up to this, La Porta, Lopez-De-Silanes and Shleifer (1999) state the ownership structures to be determined in this research are all present in Germany.
7 Besides the calls in research and the transferability, the third major benefit of taking Germany as my laboratory is the data availability, or to be more specific the structures of German corporate ownership are all present in order to properly investigate the hypotheses. For example, Germany has a huge presence of family and state owned firms, in comparison with some other European countries. Thomsen and Pedersen (1996) point out in their research on ownership structures in Europe that Germany has a wide variety of ownership structures which are used in this research. For example, the level of family ownership and government ownership is relatively high in comparison with other European countries. La Porta et al. (1999) provide evidence of the existence of these ownership structures, and therefore show that Germany is the natural laboratory to provide data in order to produce generalizable results on the effect of corporate ownership structures on tax avoidance (Faccio and Lang, 2002).
Getting insight in the effects of those structures may add knowledge to some necessary questions about tax avoidance, for example: ‘Why do some firms avoid more taxes than others?’ and ‘How do shareholders influence the amount of tax avoidance’. Moreover, Hanlon and Heitzman (2010) explicitly state in their review paper that more serious research should be conducted on the effect of ownership structures on tax avoidance. They also call for more research on the measurement proxies of tax avoidance. Although this research uses some existing proxies, it might come up with new proxies or at least new insights about the usage of these proxies. Therefore, this study also contributes to the growing literature about the measurement of tax avoidance. Finally, there is a lack of a clear and general definition of tax avoidance, as stated in the background paragraph. This research is not aiming to provide this general definition, because one single research would never be able to do that, but it contributes to a better understanding of the topic, and therefore adds to build-up process to this general definition.
As stated, there is also a small contribution from a societal point of view. As this research aims to provide a clear insight in the concept of tax avoidance and one of the determinants of this concept, it helps the general understanding of tax avoidance. Because tax avoidance has a central place in the public debate, it adds to the understanding and therefore contributes to this public discussion. It also gives insight to both the government and auditors into how corporate ownership structures affect tax avoidance, and therefore what kinds of companies are likely to use it. As mentioned, this is not as important as the scientific contribution, but should be taken into account.
8
2 Literature review
This chapter looks to explain the relevant literature and the theories used to come up with hypotheses. Additionally, some definitions are set up to clarify some concepts used for this research. In conclusion, this paragraph gives the theoretical framework for the conducted research.
2.1 Concepts and definitions
This paragraph provides the general concepts discussed in this research. A clear and understandable definition should provide the framework through which this research is conducted. As stated in the introduction, generally accepted definitions are not available for all concepts. Therefore, existing literature is provided to be as thorough as possible, without losing the ability to empirically substantiate this research.
2.1.1 Tax avoidance
This subparagraph introduces the concept of tax avoidance and provides a definition of what tax avoidance is. This definition is based on findings and thoughts of prior research. One of the ways of looking at tax avoidance is by using the framework of Dyreng, Hanlon and Maydew (2008). These authors conducted a research on a new measure of tax avoidance. The authors emphasize that prior research has come up with a lot of different ways to measure tax avoidance. Examples of these measures are examining book-tax differences for the corporate sector, conducted by the U.S. Department of Treasury, and book-tax differences in publicly traded firms, conducted by Desai, Dyck and Zingales (2007). They recognize the common factor of these measures is that they are based on annual tax avoidance. Dyreng et al. (2008) suggest that focusing on annual measures might include bias. They argue that measuring on an annual base does not give insight in whether the avoidance happens year to year or whether it is based on transitory phenomena which can be based on specific circumstances. Therefore, they provide a framework in which tax avoidance is measures in the long run. This long run measure is able to show whether firms are able to avoid taxes over longer periods of time.
Hanlon and Heitzman (2010) are using an even broader definition, because the definition of Dyreng et al. (2008) is purely empirically based. Hanlon and Heitzman (2010) add a conceptual thought to tax avoidance, stating that avoidance can either be legal or illegal. They won’t distinct between this legal question, because in their opinion a conceptual thought on tax avoidance should consider both parts. In their review paper, Hanlon and Heitzman present 12 different measures which are being used in modern financial accounting research literature. This research uses and combines some of the different measures. As the authors mention, not all measures are useful for every research. Therefore, the different measures are analyzed and the right measure
9 ismeasures are used to come up with a thorough measurement of tax avoidance. The operationalization of the measurement of tax avoidance, which is based on the theory discussed in this section, is mentioned and clarified in the methodology paragraph. This operationalization
comes with formulas to calculate the amount of tax avoidance.
A more recent study which is used in order to build and operationalize tax avoidance is the study of Badertscher, Katz and Rego (2013). The authors of this study reckon that all different kind of measures, for example cash effective tax rate or discretionary permanent book-tax difference, contain errors. In order to control for the robustness of the results, they argue that one should rely on all the different measures in order to come up with the right results.
Graham, Hanlon, Shevlin and Shroff (2013) discuss the different incentives of tax avoidance in their research. Interesting notions are the potential non-tax costs, such as litigation costs and reputational damage, but also the cost of a potential price discount. This discount arises when minority shareholders are concerned that the majority shareholder uses the company for rent-seeking. The authors state that companies outweigh the benefits from avoiding taxes, which are of course the lower tax payments leading to a higher net income, and the costs which are mentioned above. This is the reason that not all firms have the same incentives when it comes to
tax avoidance.
In this paragraph I elaborated on different views in prior literature on the measurement of tax avoidance. The theory build-up is used in the methodology and analysis part, where I operationalize the variable tax avoidance and come up with the formulas that are used for the measurement of the variable.
2.1.2 Corporate ownership structures
This subparagraph looks to point out the different corporate ownership structures examined. These ownership structures are derived from prior literature. The three general ownership structures which are determined are family ownership, based on research by Chen et al. (2010), government ownership, based on research by Chan et al. (2013), Bradshaw, Liao and Ma (2016), and foreign ownership, based on research by Huizinga and Nicodème (2006). In the following part of this subparagraph, I elaborate on the characteristics of the different ownership structures. These characteristics, evaluated with theory build up in this entire paragraph are used to come up with the hypotheses used to answer the research question.
2.1.2.1 Family ownership
As specified, the first ownership structure determined in this study is the family ownership. This includes firms which are owned or run by family members. Chen et al. (2010) define family firms as firms in which members of the founding family still have positions in top management, are
10 settled on the board or are so called block holders of the existing company. They argue that the fact that a firm is family owned leads to different agency conflicts. The authors build from theory that states that tax avoidance or aggressiveness derives from the conflict between costs taxes represent and the potential non-tax costs which can accompany tax avoidance. This agency conflict is discussed in depth in the following theory section. Family ownership is considered a ‘big player’ in the discussion around the effect of corporate ownership structures on tax avoidance, because between 30-40% of S&P 1500 firms are ought to be family owned. This percentage depends on the definition of family ownership used (Chen et al. 2010). To investigate the effect of family ownership, the incentives and risks of managers in this kind of firms to participate in certain kinds of tax avoidance should be researched. This subparagraph builds upon degree of risks and incentives managers of family firms are subject to. Combined with the agency theory and following costs, the effect of family ownership on tax avoidance is stated. Following a general perspective, family firms are ought to be participating in less tax avoidance than their non-family counterparts.
Fama and Jensen (1983) argue that undiversified shareholders, in this case the family, can emphasize different investment decision rules than there diversified counterparts. In other words, Fama and Jensen (1983) show an example on the general implication that family firms, especially their managers, have incentives to focus on investments that benefit themselves instead of the firm. However, this relation between family ownership and firm value has resulted in contradictory results in prior research. Anderson and Reeb (2003) provide for example analysis that family firms actually perform better and are therefore an effective ownership structure. The interesting part is that these authors expect the effect of these family ownership structures on firm value to be negative. This shows the importance of good usage of the relevant theory.
Similar with the mentioned effect on firm value, the effect of family ownership on tax avoidance is subject to some contradictory findings and expectations. Whereas Chen et al. (2010) expect and find a negative relationship of family ownership on tax avoidance, Annuar et al. (2014) expect a positive relation. The difference both authors make is the analysis of the costs and benefits. Chen et al. (2010) state that family ownership is subject to an unique agency conflict between the minority and dominant shareholders. This implies that the different preferences of these shareholders give insight in the amount of tax avoidance family firms engage in. The tradeoff they find is that family members are willing to forgo the benefits of tax avoidance in order to prevent the extra costs of a price discount. This price discount is brought up by Desai and Dharmapala (2006), who state that a price discount can arise because minority shareholders might presume that tax avoidance activities are a way for family members to mask so called
rent-11 seeking. This means that investors are usually protecting themselves against this rent extraction, if they assume family firms are conducting family entrenchment. Besides, both the articles also refer to the potential litigation costs of tax avoidance and the damaging effect to the reputation of the family. To summarize, Desai and Dharmapala (2006), together with Chen et al. (2010), find that family firms are expected to consider both the benefits and the costs of tax avoidance, in order to decide whether to engage in it. The notions brought up in this subparagraph are used to build a hypothesis at the end of this chapter.
2.1.2.2 Government ownership
This subparagraph describes the second ownership category, the firms that are owned by the government. According to prior research, for example Chan et al. (2013), government ownership is present when the government is the majority shareholder of a company, and therefore dictates board structures and certain policies. This government ownership researched in prior research, by Chan et al. (2013) and Mahenthiran and Kasipillai (2012), is predominantly present in Asian companies, where despite of recent privatization still a lot of companies are owned by the government. This study looks into German listed companies, where this form of ownership is a lot less present, but still plays a decent role in the economic landscape. Interesting are the contradictory results which some of the prior research come up with. Chan et al. (2013) find that firms with governmental ownership pursue less tax aggressive activities, and therefore the tax avoidance is lower. However, Mahenthiran and Kasipillai (2012) find that governmental owned firms in Malaysia have a lower effective tax rate, assuming therefore that government ownership structures enhance tax avoidance. This research looks to provide the insight in a new setting, Germany, in order to create a better understanding of the effects of government ownership structures on tax avoidance.
Bradshaw et al. (2016) approach government ownership from a different perspective. They state that the government is a minority shareholder in all firms, because of the tax claim that government has on firms. Therefore, the government is entitled to a large percentage of the profits. Trying to establish a relationship between government ownership and tax avoidance should take into account these potential conflicts of interests between managers and ownership, in this case the government. As is clear from prior literature (e.g. Djankov and Murrell, 2002; Sun and Tong, 2003), governmental ownership usually leads to less incentives for financial performance, and therefore to lower financial results. This reasoning is used to provide the assumption that managers in governmental ownership structures are less likely to use tax avoidance, because there are less incentives to provide higher net income and it wouldn’t be incentivized by the owners because tax costs for a firm are income for the government.
12 Following this reasoning, avoiding taxes would costs the controlling shareholder, the government, income, and is therefore inefficient. These authors argue that government ownership is expected to lead to less tax avoidance, because of the inefficiency.
Chan et al. (2013) approach the ownership effects by focusing on the long-term horizons of executives of government ownership. They base their research on the findings by Khurana and Moser (2009), who found that firms with a high level of institutional ownership, which in this case is the governmental ownership, engage less in tax avoidance practices because the owners think about the long term consequences evolving from tax avoidance. Examples of this long term consequences can be either litigation costs, if illegal avoidance is used, or shifting of profits when legal avoidance is being used. Chan et al. (2013) conclude that government ownership usually should have a negative effect on tax avoidance, meaning that governmental owned firms are engaging in less tax avoidance practices, because of the long term horizon. In addition, managers of governmentally controlled firms are likely to contribute to this negative effect for two reasons. The first reason is that managers that are appointed by the owner, the government, are likely to have political objectives and therefore are actually willing to report high taxes, even at cost of overall firm value (Chan et al. 2013). This creates an interesting agency conflict in which the controlling shareholder has incentives to pursue high taxes at cost of firm value, while minority shareholders obviously pursue firm value and profit. The exact dimensions of the agency theory isare clarified in the theory section. The second reason is that violation of taxes can result in serious litigation and penalties, and moreover can jeopardize the future of the careers of managers in the political scene (Cao and Liu, 2007). Summarizing, Chan et al. (2013) argue that firms with government ownership are less likely to conduct tax avoidance than their non-governmental counterparts.
Concluding, it is clear that there are different perceptions on the effect of government ownership on tax avoidance. The theory brought up in this paragraph is used to build the hypothesis at the end of this chapter.
2.1.2.3 Foreign ownership
This subparagraph describes the third ownership structure, foreign ownership. Annuar et al. (2014) state there has not been much research into the foreign ownership structure. However, they include a study by Christensen and Murphy (2004), which concludes that the presence of foreign investors can be linked with tax aggressive practices, and therefore more tax avoidance. Demirguc-Kunt and Huizinga (2001) researched the effect of foreign ownership in banks on tax avoidance. These authors found that foreign owned banks pay less taxes than non-foreign owned banks, using data from 80 countries around the world, therefore finding that foreign ownership
13 leads to more tax avoidance. However, this finding is limited to the banking sector, and is based on data before 2000. A lot has changed since, and therefore Annuar et al. (2014) call for new research on this matter.
Studies in more recent time period are rare. However, there are some studies which try to examine a comparable relationship. For example, the research by Egger, Eggert and Winner (2010) looks for the relationship between foreign plant ownership and the amount of taxes saved. In their research, they find significant evidence for profit shifting, meaning that multinationals pay relatively huge amounts of taxes in low-tax countries and relatively low amount of taxes in high-tax countries. This shows that foreign plants pay lower taxes than their domestic counterparts, and therefore provides evidence for tax avoidance.
Another analysis which shows some resemblance to this ownership structure is the research by Salihu et al. (2015), which presents a relation between the interests of foreign investors and tax avoidance. This study especially focusses on the investments in developing countries. Again, as presented in the introduction, this research doesn’t focus on European, developed countries, but the results can be used to come up with the hypotheses. The results show a significant positive relationship between the interests of these investors in the magnitude of tax avoidance. This shows again that, in developed countries, there is an assumed positive relation between presence of foreign ownership and tax avoidance.
Huizinga and Nicodème (2006) investigate the effects of foreign ownership on income taxation on a macro level, meaning that they investigate the overall tax burden instead of the difference between domestic and foreign ownership on individual firms. They find that there is a positive relationship between tax burden and the presence of foreign ownership in Europe. Besides they argue that countries with high presence of foreign ownership impose high tax rates. The most comparable study for the relation between foreign ownership and tax avoidance is the research by Langli and Saudagaran (2004). This study provides a Norwegian framework to conclude that foreign-controlled Norwegian firms have systematic lower effective tax rates than their domestic counterparts. Their study is unique because it also provides data on small and medium-sized firms, instead of only multinationals. By assessing these firm sizes, they hope to improve generalizability. However, their data and results are focusing on the end of the 90’s, with huge differences in international tax rates compared to now. Therefore, the incentives to participate in tax avoidance compared to the non-tax costs might be different, leading to different results.
14 on foreign ownership in its relation to income taxation. The insights are used to provide expectations for this research, leading into hypotheses at the end of this chapter.
2.2 Theory
In this section, I focus on the theory provided in different prior research. I combine these theories and evaluate the progression of the usage of these theories throughout the last years, approximately 10-15 years. That way, I can show what theories are still relevant, or are countered by new theories and frameworks. The total picture helps in developing the necessary hypotheses to test the research question.
2.2.1 Agency theory
In this subparagraph I discuss what is more and more seen as one of the most important theories to explain the effect of corporate ownership structures on tax avoidance. In the early and mid-2000’s, the theoretical framework by which tax avoidance could be explained was based around the agency theory. This new interest in the agency theory is built upon the framework of Fama and Jensen (1983). Badertscher et al. (2013) build upon this agency framework in their research. The link in their research is made between the segregation of ownership and control. The segregation of ownership and control influences the extent to which managers are risk-averse. Badertscher et al. (2013) state that the more risk-averse managers are, the less likely they are willing to invest in risky projects. Tax avoidance is seen as a risky investment, because it can impose significant costs on a firm and their managers. Therefore, the link between tax avoidance and risk-averse managers is been made. The connection between tax avoidance and ownership structures is therefore made too. Ownership structures affect how risk-averse managers and owners of companies are. This association between risk-adversity and tax avoidance is built upon in a lot of tax related accounting research, for example Hanlon and Heitzman (2010) and Desai and Dharmapala (2008). The prediction of Badertscher et al. (2013) are concentrated on this agency theory. As described before, the different ownership structures have different types of risks for managers and therefore the manager haswill have different levels of risk-adversity. Following these assumptions, predictions can be made about the effect of certain ownership structures on tax avoidance.
This general application of the agency theory is extended by some other viewpoints and implications of the theory. For example, Chen and Chu (2005) focus on a standard principal-agent model, whereby the focus on the efficiency loss caused by the separation of control and management. Crocker and Slemrod (2005) also use this same model to examine the effect of the compensation contract of the one responsible for the tax deductions. They shift their focus to the penalties based on the principal or the agent.
15 Another important vision on this principal-agent theory is the perspective presented by Desai et al. (2007), who focus their research on the situation in which a manager is self-interested. In this case, a manager would structure transactions in a way by which there will both be tax avoidance and the ability to use resources of the company for private use. They argue that a strict monitoring by tax authorities might be a solution for this particular case, and therefore something which should be added in the theory. In this subparagraph, I extend on the theories provided by Desai et al. (2007).
As is clear, a lot of different perceptions of the agency theory are being used to understand and explain managers behavior and therefore the effect of certain ownership structures on tax avoidance. Additionally, in the following subparagraph about corporate governance, this issue is evaluated throughout the research.
2.2.2 Corporate Governance
In this subparagraph I focus on corporate governance, the theories behind it and the expected influences on tax avoidance. This theoretical part is based on several papers, in which this
phenomenon has been researched. Examples of these papers are Desai et al. (2007), Minnick and Noga (2010) and Lanis and Richardson (2011). Theories about corporate governance could for example be the structure of the board, the presence of certain committees etc. In prior research, by the before mentioned researchers, corporate governance is one of the key factors in
understanding and explaining corporate ownership structures. Chan et al. (2013) and Mahenthiran and Kasipillai (2012) also use corporate governance to evaluate the different
ownership structures. Besides, they emphasize that corporate governance plays a huge role in the segregation between ownership and control. Again, this segregation between ownership and control influences managers and their perception on the costs and benefits of tax avoidance. Understanding the concepts of corporate governance is therefore necessary to understand the relation and the effect of the ownership structures on tax avoidance. The different proxies of corporate governance could help to understand the different corporate ownership structures, and could therefore control for other factors that would influence tax avoidance, besides the
corporate ownership structures. In order to maintain a realizable timeframe for this research, not all the different corporate governance proxies are measured and discussed in depth.
Corporate governance mechanisms has been studied a lot and their relation with tax avoidance has been established in lots of different and sometimes contradictive ways. A recent study by Armstrong et al. (2015) finds that there is no clear relation between several corporate governance mechanisms around the mean and median of tax avoidance. However, they find a relation for low and high levels on tax avoidance. The conclusion they make is that the
16 mechanisms have way more impact on the extreme occasions of tax avoidance. Important to note from this study is therefore that mechanisms do affect tax avoidance activities, and in order to provide a clear results on the effect of ownership structures, should be taken into account when assessing control variables.
Minnick and Noga (2010) state that tax planning, for example avoidance, is a long term investment which is influenced by corporate governance mechanisms. For example board structure may have a huge influence in the way managers are incentivized to avoid taxes, in order to show better performance. Evaluating these mechanisms is therefore vital in order to understand the concept of tax avoidance. Companies with different structures of corporate governance may enhance certain types of tax avoidance more, either by incentivizing or by a lack of controls. This again shows the importance of governance mechanisms in the research on tax avoidance. The authors show that governance is related to long-term tax planning.
Mahenthiran and Kasipillai (2012) argue that corporate governance mechanisms are a way of minority shareholders to protect themselves against the influences and desires of insiders, in many case the major shareholders and therefore the owners. Good mechanisms should prevent the owners to focus only on their interests and expropriate minority shareholders. Given this relation, they argue that corporate governance mechanisms should be taken into account in order to provide insights on tax avoidance. They find that firms audited by a big 4 are having higher effective tax rates. The other governance mechanism they define are not significant. However, following their reasoning, these factors should be included in order to control for the results of tax avoidance.
The last paper to discuss in this section is the research by Lanis and Richardson (2011). These authors focus on another corporate governance mechanism, being the board composition. Important for this study is whether they find a relation between this board composition and tax aggressiveness, because that would mean that it should be included as control variable. They focus their research on the paper by Desai and Dharmapala (2006) and find that board composition does have an effect on tax aggressiveness, based on research in the Australian setting. The argument is that a higher number of outside directors reduces the amount and likelihood of tax aggressiveness, and therefore tax avoidance.
Summarizing on the views provided in this section, it is clear that corporate governance mechanisms can influence the way a company is approaching tax aggressiveness and tax avoidance. It is therefore necessary to include this into the thought process around tax avoidance and the effect ownership structures might have.
17 2.3 Hypotheses
In this paragraph I discuss the hypotheses which are used to answer the research question. The hypotheses are derived from the above mentioned theory and conceptions and definitions. When necessary, some additional explanation isare described in order to make sure the hypotheses are built upon the mentioned theory.
As follows from the research question, this research seeks to examine the effect of corporate ownership structures on tax avoidance. The main theory to explain certain effects is the agency theory, which seeks to explain the way managers behave. Following this theory, the hypotheses isare built up around the risk-adversity of managers, in combination with the segregation of ownership and control.
The first type of ownership is family ownership. As stated before, and based on prior literature, family ownership is most likely to not have a huge segregation in control and ownership. This influences the risk-adversity of managers and therefore the participation in tax avoidance. Besides, expected is that the potential benefits of tax avoidance will not outweigh the potential costs. Therefore, I hypothesize that family ownership is likely to have a negative effect on tax avoidance.
H1: Firms with family ownership are less likely to participate in tax avoidance practices than non-family ownership firms.
The second type of ownership is the government ownership. In the literature section is pointed out that, although conducted in different countries and landscapes, different thoughts and results are found. In this research, focus is on the German setting. Although this form of ownership is not as present as in Eastern countries, for example China, it is an important ownership structure and therefore important for this research. The prediction for this form of ownership arises from the cost and benefit analysis. Chan et al. (2013) argue that government ownership has a negative effect on tax avoidance due to political objectives and potential litigation costs as discussed in paragraph 2.1.2.2. This research, based on the agency framework and therefore the separation of ownership and control, has the same expectation for the German setting. I hypothesize that government ownership is likely to have a negative effect on tax avoidance.
H2: Government ownership is likely to have a negative effect on tax avoidance.
The third type of ownership is foreign ownership. Prior literature has pointed out that there is a lack of research in this field. The research that has been done by Demirguc-Kunt and Huizinga (2001) is concentrated solely on the banking systems and relies on data before 2000. Because a lot of things has changed since for example the financial crisis, new research should be done.
18 Christensen and Murphy (2004) conclude that the presence of foreign ownership leads to more tax avoidance. In comparison with the agency theory, which states that tax avoidance should take place when managers are less risk-averse. Foreign ownership usually enhances the segregation between ownership and control, resulting in managers usually being less risk averse. As discussed in paragraph 2.1.2.3, there are some other studies who discuss similar questions, and the overall tendency is that foreign ownership enhances tax avoidance and is used as a method to create a lower effective tax rate. Therefore, the prediction is that foreign ownership has a positive effect on tax avoidance.
H3: Foreign ownership is likely to have a positive effect on tax avoidance.
Of course, the distribution of ownership in these three categories is not complete. One can also argue that the could be differences between listed firms, which shares are able to be publicly traded, and non-listed firms, which shares are not. Both types of firms therefore have different incentives as it comes to taxes. Publicly listed firms likely prefer profits to be as high as possible, because this enhances their share price and therefore the value of their company, or market capitalization. Non-listed firms have incentives to focus more on long-term profits and might be more likely to be affected with reputational damage and litigation costs. Overall, the expectation is that non-listed firms will participate in less tax avoidance, leading to an higher effective tax rate.
19 3 Research Methodology
In this paragraph I describe the research methodology. The research method consists of a part about sample selection and a part about the empirical framework. The sample selection part concludes motivations on the sample and initial thoughts. The empirical framework is initially based on empirical frameworks used in prior literature. Throughout the research, there might be additions and or different formulas. The empirical framework seeks to operationalize the variables, making clear how research is conducted.
3.1 Sample Selection
The sample I used is a sample of German firms, over the time period of 2010-2014. I chose German firms because, to my best knowledge, no research has been conducted on this interaction, and it gives insight in the relation between corporate ownership and tax avoidance, in the Western market. Important is the note of the Western market. Prior research states that no such relationship has been researched. Comparable research in other countries is presumably not transferable to the Western market, therefore making it hard to predict a relationship in this market. This research seeks to provide insights in the interaction in this Western market, hopefully leading to transferable results. An example of another study is the study by Chea et al. (2013) on government ownership. This research is conducted in China, where a lot of companies are held by the government. The assumption by these authors is that their results are probably not transferable to other markets.
The selected timeframe, 2010-2014, is selected for a couple of reasons. At first, the call by Hanlon and Heitzman (2010) for more recent studies into the determinants of tax avoidance. Prior studies primarily use earlier timeframes. Secondly, the sample starts in 2010 in order to avoid bias from the financial crisis in 2007 and 2008. The year 2015 is excluded because of the lack of data for some firms for this year. All in all, 2010-2014 seems a proper timeframe with sufficient data availability.
Data is gathered through archival database research, primarily based on the Amadeus database by Bureau van Dijk, accessed via WRDS. This database offers financial statement data about listed and non-listed firms in Europe, and therefore information about German companies. In addition, a separate database owned by Bureau van Dijk also has some information on major shareholders, direct ownership and ultimate ownership.
Because of the restrained period of time for this thesis, I decided to focus on listed firms in Germany. In addition to the time frame, the data on listed firms was also more complete. Initially, the sample consisted of 720 listed firms with 4882 firm-year observations. Table 1, Panel A, shows the sample selection process. The first step was taking out 652 observations due to lack
20 of taxation data, which was necessary for calculating the ETR. Lack of data on operating profit (38), intangible assets (16) and long-term debt (3) excluded another 57 observations. Based on prior research, in order to control for skewness and outliers, effective tax rates lower than 0 or higher than 1 should be taken out of the sample (Chen et al. 2010; Chan et al. 2013; Bradshaw et al, 2016). Eliminating these observations, 1145 firm-years were taken out of the sample. Missing data on market capitalization led to another 313 drops of data, while finally double-year data had to be removed (959). Due to these data eliminations, the final sample consisted of 1756 firm-years of 537 publicly listed companies in the period 2010-2014. Financial data on these firm-year observations was gathered through the Bureau van Dijk database.
3.2 Ownership and industry distribution
Panel B of table 1 shows the distribution into ownership structures. As is clear from the table, around 40% of the sample can be classified as one of the ownership structures, with family firms and foreign firms dominantly available. The sample for government ownership is not very large, which is taken into account while discussing the results. The classification for certain ownership structures is clarified based on prior literature.
Family ownership is primarily based on the paper by Chen et al. (2010), which studies the relationship between family ownership and tax avoidance in a Chinese setting. Firms are classified as family firms of one of the following criteria is met: The founding family holds more than 50% of the outstanding shares, founding family members or successors still hold more than 5% of the shares outstanding or founding family members or successors still hold positions on the board. This classification leads to 112 family firms with 416 firm-year observations.
The classification for government firms is primarily based on the papers by Chan et al. (2013) and Bradshaw et al. (2016). In these papers, which look at so-called State-Owned Enterprises, firms are classified as governmentally owned when the government either holds more than 50% of the shares or is the major block holder of the firm. Major block holder of the firms assumes that a government party has the largest part of the shares in possession. Because of the selected sample of German listed firms, in order to keep the sample workable within the timeframe, this type of ownership is not very present. Actually, the subsample consists of 11 firms with 42 firm-year observations. Despite this small sample, I decided to still incorporate the regression, to show the direction of the relationship. Future research should point out whether a sample of non-listed firms contains more government ownership and could therefore shed a better light on the relation between this type of ownership and tax avoidance.
The last ownership structure is foreign ownership. Foreign ownership is a structure that has not seen much research, especially not in relation with tax avoidance. As specified in chapter
21 2.1.2, several researchers call for a more in-depth investigation of the relationship between
foreign ownership and tax avoidance. As shown in panel B of table 1, German listed firms provide a sufficient sample to investigate the relationship. Based on Annuar et al. (2014),
Huizinga and Nicodème (2006) and Egger et al (2010), firms are classified as foreign owned firms when foreigners hold more than 50% of the shares or are the major block holders of the firm. This subsample consists of 80 firms including 244 firm-year observations.
Data on ownership was initially gathered from the Bureau van Dijk database, which contains some information on ultimate ownership. The data gathered contained two problems. At first, the data was far from complete. Ultimate ownership data was available for 180 out of 537 firms. Besides, this data didn’t separate the years. This results into problems when major portions of a company’s shares were traded to different owners throughout 2010-2014. In order to prevent this problem, I hand collected ownership data via firm websites. Almost 90 percent of the firms provided this information via the tabs investor relations and shareholder’s structure. When English websites weren’t available, I searched for the ‘aktionärsstruktur’, which is simply the German word for shareholder’s structure. For the 10 percent which did not provide information on their website, I searched the internet and websites of the exchanges where the firms were listed. This allowed me to finalize my data on ownership. Data on founders and boards of companies were also gathered from firm websites, which allowed to classify for example the family firms.
Panel C presents the distribution of the firms into several industries. Data on the industries was also gathered from the Bureau van Dijk database, by the NACE Rev. 2 primary codes. Classification of these codes is based on the Eurostat report from the European Commission (2008). Chen et al. (2010) state that there is cross-industry variation in tax aggressiveness and therefore avoidance, I control for the different types of industries using dummy variables for each of the industries.
22 Table 1: Full sample descriptive statistics
Panel A: Sample selection
Firm-year observations of German listed firms 4882
(652) (38) (16) (3) (1057) (88) (313) Less: Observations with missing data on taxation
Less: Observations with missing operating profit Less: Observations with missing intangibles Less: Observations with missing long-term debt Less: Observations with ETR<0
Less: Observations with ETR>1
Less: Observations with missing market capitalization
Subtotal after removing observations with missing data
Less: Double-year data entries
Final sample for ETR test
2715
(959)
1756 Panel B: Ownership Distribution
# of firm years Percentage # of firms
Family Ownership Government Ownership Foreign Ownership
Other ownership structures
416 42 244 1054 23,7% 2,4% 13,9% 60,0% 112 11 80 334 Totals 1756 100% 537
23
Notes: This table shows respectively the sample selection in Panel A, the distribution of ownership structures in Panel B and the distribution of industries in Panel C.
Panel C: Industry distribution of firm-years
# of firm years Percentage
Agriculture, forestry and fishing 5 0,28%
Mining and quarrying 7 0,40%
Manufacturing 455 25,9%
Electricity 33 1,88%
Water supply 1 0,05%
Construction 19 1,08%
Wholesale and retail trade 103 5,87%
Transportation and storage 37 2,11%
Accommodation and food service 5 0,28%
Information and communication 167 9,51%
Financial and insurance activities 110 6,26%
Real estate activities 89 5,07%
Professional, scientific and technical activities 617 35,14%
Administrative and support activities 52 2,97%
Public administration and defense 0 0%
Education 5 0,28%
Human health and social work activities 17 0,97%
Arts, entertainment and recreation 24 1,37%
Other service activities 10 0,57%
Activities of households as employers 0 0%
Activities of extraterritorial organizations 0 0%
24 3.3 Matched sample
In this subparagraph I describe the matched sample which is created and used to measure the fourth hypothesis, which tests whether non-listed firms avoid less taxes than listed firms. This matched sample was created to contain around 100-150 firms, in order to be sufficient enough to be used for regression analysis. The easiest way to create a matched sample is to look at firms which have are about the same of size. Because market capitalization, which is used to determine size, is not easily determined for non-listed firms, I chose to focus on the total assets.
After establishing the mean of the total assets of the listed firms, it became clear that no private companies where of the same size as the largest listed companies. Therefore, I tried to winsorize the sample in order to reduce the amount of outliers. I again took the mean of the winsorized sample and allowed the total assets to be 5% higher or 5% lower than this mean. This criterion brought me a total sample of 231 private firms with 1645 firm-year observations.
Table 2 provides the selection criteria for the full sample, including the matched sample. The initial sample consisted of 6527 firm-year observations of which 4882 were observations of listed firms and 1645 were observations of the matched sample of non-listed firms. The major cut in data was deleting the double-year entries (2462). In comparison to the starting sample of listed firms, the same firm-year observations were removed because of missing data. The extra removal consisted of data of 2009, which was included to compute the sales turnover, which is explained in Appendix A. As Panel A shows, after removing all the data, the sample for the matched analysis consisted of 705 unique firms with 2499 firm-year observations. This sample is used to test hypothesis 4.
Panel B of table 2 provides the distribution of firm-year observations to the different industries. Again, this is done by the framework provided by the European Commission (2008). These industry classifications are used to control for cross-industry variation of tax
25 Table 2: Matched sample descriptive statistics
Panel A: Sample Selection (including normal sample)
Firm-year observations of German firms 6527
Less: Double-year entries 2462
Subtotal after removing double entries 4065
Less: Observations with missing sales (300)
Less: Observations of 2009 (627)
Less: Observations with missing data on taxation (37)
Less: Observations with missing operating profit (1)
Less: Observations with missing intangibles (1)
Less: Observations with missing long-term debt (2)
Less: Observations with ETR<0 (527)
Less: Observations with ETR>1 (71)
26 Panel B: Industry distribution of firm-years
# of firm years Percentage
Agriculture, forestry and fishing 5 0,20%
Mining and quarrying 16 0,64%
Manufacturing 634 25,4%
Electricity 91 3,64%
Water supply 17 0,68%
Construction 45 1,80%
Wholesale and retail trade 167 6,68%
Transportation and storage 57 2,28%
Accommodation and food service 5 0,20%
Information and communication 182 7,28%
Financial and insurance activities 134 5,36%
Real estate activities 133 5,32%
Professional, scientific and technical activities 851 34,05%
Administrative and support activities 76 3,04%
Public administration and defense 0 0%
Education 5 0,20%
Human health and social work activities 28 1,12%
Arts, entertainment and recreation 33 1,32%
Other service activities 20 0,80%
Activities of households as employers 0 0%
Activities of extraterritorial organizations 0 0%
Total 2499 100%
27 3.4 Research design
This paragraph elaborates on the research design by describing the tax avoidance measure, showing the research model and providing descriptions of the control variables. The tables provide descriptive and univariate analysis, before the actual regressions are conducted in paragraph 4. Statistical analysis is conducted by using STATA.
3.4.1 Tax avoidance measurement
The tax avoidance measurement that is used is based on prior literature and is the well-known effective tax rate measure. Throughout literature this measure is seen as one of the most reliable measurements of actual tax rates and therefore is a right estimate of tax avoidance (Chen et al, 2010; Mahenthiran and Kasipillai, 2012; Bradshaw et al, 2016). Prior research sometimes uses different measures, such as cash taxes paid or book-tax differences. These measures are not appropriate in a German setting because of limited data availability. For example, the Compustat database (US firms) or CSMAR database (Chinese firms) provide additional information on cash taxes paid and book-tax differences, and therefore more ways of measurement can be used. Whether the results would be different if these ways of measurement were included can be determined by future research.
Table 3 shows the means and medians for the effective tax rate (ETR), separated for each ownership structure. For example, Panel A shows the difference in mean and median for family and non-family firms. The other panels show the results for the other structures. As table 3 shows, the differences in both mean and median are significant for all of the ownership structures except for foreign firms. The results for family firms are consistent with prior literature (Chen et al. 2010), the results for government firms are consistent with some of prior literature (Mahenthiran and Kasipillai, 2012) but also inconsistent with others (Chan et al. 2013; Bradshaw et al, 2016). The results for the private vs. listed firms are significant and so far prove hypothesis 4. This therefore allows proceeding with multivariate analysis.
Summarizing, univariate analysis proves that both family firms and private firms have significant higher ETR’s than there non-family or listed counterparts. With government firms, these firms have a significant lower ETR than their non-government counterparts. The difference between foreign and non-foreign firms is not significant.
28 Table 3: Descriptive statistics
Notes: This table shows the univariate analysis for the relation between tax avoidance (ETR) and the ownership structures. Detailed variable descriptions can be found in Appendix A. Panel A is based on family ownership, Panel B is based on government ownership, Panel C is based on foreign ownership and Panel D is based on private firms. Differences in mean and median are tested with the t-test and Wilcoxon rank test respectively.
***, **, * are signs of two-tailed significance, *** when p < 0,01, ** when p < 0,05 and * when p < 0,1.
Panel A: Family firms vs. non-family firms
Family firms Non-family firms
N Mean Median N Mean Median t-test Wilcoxon
ETR 416 0,297 0,289 1340 0,242 0,245 5,51*** 6,74***
Panel B: Government vs. non-government firms
Government firms Non-government firms
N Mean Median N Mean Median t-test Wilcoxon
ETR 42 0,149 0,112 1714 0,258 0,264 -3,94*** -4,12***
Panel C: Foreign vs. domestic firms
Foreign firms Domestic firms
N Mean Median N Mean Median t-test Wilcoxon
ETR 244 0,253 0,254 1512 0,256 0,264 -0,231 -0,775
Panel D: Non-listed vs listed firms
Non-listed firms Listed firms
N Mean Median N Mean Median t-test Wilcoxon
29 3.4.2 Model and variables
To test our hypotheses, comparable models are used. The first three hypotheses only differ in the ownership variable, which is therefore also the only difference in the model. Hypothesis one seeks to test whether family ownership influences tax avoidance. The expectation is that family firms avoid fewer taxes. The following model is used to test for this:
𝐸𝑇𝑅 = 𝛽0+ 𝛽1𝐹𝐴𝑖,𝑡+ 𝛽2𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛽3𝑅𝑂𝐴𝑖,𝑡+ 𝛽4𝐿𝐸𝑉𝑖,𝑡+ 𝛽5𝑀𝐵𝑖,𝑡+ 𝛽6𝐶𝐴𝑃𝐼𝑁𝑇𝑖,𝑡+ 𝛽7𝐼𝑁𝑇𝐴𝑁𝐺𝑖,𝑡 + 𝛽8𝐵𝐼𝐺4𝑖,𝑡+ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀𝑖,𝑡
After all the models are shown, a concise explanation of the variables is given. For a detailed description, please look at Appendix A.
The second hypothesis seeks to explain a relationship between government ownership and tax avoidance. Here, the expectation is that government owned firms avoid less taxes. The follow model is used:
𝐸𝑇𝑅 = 𝛽0+ 𝛽1𝐺𝑂𝑖,𝑡+ 𝛽2𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛽3𝑅𝑂𝐴𝑖,𝑡+ 𝛽4𝐿𝐸𝑉𝑖,𝑡+ 𝛽5𝑀𝐵𝑖,𝑡+ 𝛽6𝐶𝐴𝑃𝐼𝑁𝑇𝑖,𝑡+ 𝛽7𝐼𝑁𝑇𝐴𝑁𝐺𝑖,𝑡 + 𝛽8𝐵𝐼𝐺4𝑖,𝑡+ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀𝑖,𝑡
The third hypothesis examines the relationship between foreign ownership and tax avoidance. The expectation is different from the other structures, in particular that foreign ownership is expected to have a negative effect on tax avoidance, decreasing the ETR. The used model is as follows:
𝐸𝑇𝑅 = 𝛽0+ 𝛽1𝐹𝑂𝑖,𝑡+ 𝛽2𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛽3𝑅𝑂𝐴𝑖,𝑡+ 𝛽4𝐿𝐸𝑉𝑖,𝑡+ 𝛽5𝑀𝐵𝑖,𝑡+ 𝛽6𝐶𝐴𝑃𝐼𝑁𝑇𝑖,𝑡+ 𝛽7𝐼𝑁𝑇𝐴𝑁𝐺𝑖,𝑡 + 𝛽8𝐵𝐼𝐺4𝑖,𝑡+ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀𝑖,𝑡
The fourth and final hypothesis is somewhat different, as it seeks to examine the difference between private and listed firms, and their effect on tax avoidance. The expectation is that private ownership has a positive effect on tax avoidance, leading to a higher ETR.
𝐸𝑇𝑅 = 𝛽0+ 𝛽1𝑃𝑟𝑖𝑣𝑎𝑡𝑒𝑖,𝑡+ 𝛽2𝑃𝑟𝑆𝐼𝑍𝐸𝑖,𝑡+ 𝛽3𝑅𝑂𝐴𝑖,𝑡+ 𝛽4𝐿𝐸𝑉𝑖,𝑡+ 𝛽5𝑆𝑇𝑖,𝑡+ 𝛽6𝐶𝐴𝑃𝐼𝑁𝑇𝑖,𝑡+ 𝛽7𝐼𝑁𝑇𝐴𝑁𝐺𝑖,𝑡 + 𝛽8𝐵𝐼𝐺4𝑖,𝑡+ 𝑌𝑒𝑎𝑟 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀𝑖,𝑡
The first three models consist of the same control variables, only making a difference in the ownership variable. The control variables are all taken out of prior literature on tax avoidance (e.g. Annuar et al., 2014; Badertscher et al., 2013; Bradshaw et al., 2016; Chan et al., 2013; Chen et al., 2010; Dyreng et al., 2008; Grubert et al., 1993; Hope et al., 2013; Huizinga and Nicodème, 2006; Langli and Saudagaran, 2004; Mahenthiran and Kasipillai, 2012). These controls are in
30 place in order to make sure that the results are not driven by standard differences between certain types of ownership, for example between family firms and their non-family counterparts. The control variable SIZE is included because larger firms might enjoy benefits due to economics of scale (Bradshaw et al., 2016) or might have different incentives to invest in tax beneficial assets (Chen et al., 2010). The variable SIZE is calculated as the natural logarithm of the market capitalization. ROA is included to control for the profitability of firms. The higher the return on assets, the higher the incentives are to avoid taxes. Almost all prior literature includes this variable. ROA is measured as operating profit divided by total assets. The variable leverage (LEV) is included to control for the tax incentives rising from debt financing. LEV is measured by dividing total long-term debt by total assets. The MB variable is included to control for a firm’s growth. Growth of firms can come with different incentives for tax avoidance, which should therefore be controlled. Market-to-book ratio is calculated by dividing the market capitalization, which is the market value of equity, by the shareholders funds, which is the book value of equity. The CAP_INT variable focusses on capital intensity. The variable is calculated by dividing fixed assets by total assets. Together with the INTANG variable, which is calculated by dividing intangibles by total assets, controls for differences in the asset mix which therefore provide different tax incentives. There is also a dummy variable BIG4 in place, which equals 1 if a firm is audited by a big 4 company and zero otherwise.
The fourth hypothesis, examining the difference between private and listed firms, incorporates two different variables, due to a simple reason. The SIZE variable controls the same, but is measured differently, due to the unavailability of a proper estimate of the market capitalization. SIZE is therefore measured as the natural logarithm of the total assets. The variable ST is used instead of MB, for the same reasoning. The variable is called sales turnover and is calculated by dividing the growth in sales by the sales of the previous year. This is again a measure to control for growth. Understandably, the ownership variable also differs.
Finally, we control for industry type and different years by creating dummy variables. For the ease of reading, we excluded these variables in the descriptive statistics, the correlation matrix and the results. It is however important to note that these variables are taken into account. The different industries, shown in table 1C, are based on the NACE Rev 2. provided by the European Commission. The years are simply the different years of the sample period, 2010-2014.
Table 4 shows the descriptive statistics on the control variables. Family firms have a significantly bigger SIZE and have a higher ROA. The findings are consistent with prior literature (Chen et al., 2010). Government firms show a significant higher capital intensity (CAP_INT) and a smaller market-to-book ratio. These findings are similar to previous research on government
31 ownership (Chan et al., 2013; Bradshaw et al., 2016). Descriptive statistics on foreign firms shows only a significantly lower leverage than domestic firms. All the other differences are insignificant. Similar relation have not yet been studied, and can therefore not be confirmed with prior
literature. Finally, the descriptive statistics on the private vs listed firms show a somewhat biased picture. This is the case because the matched sample was created by picking companies around the mean total assets. This matched sample can be seen as ‘an average firm’, making mean differences significant. However, it is for example interesting that private firms are significantly more audited by big 4 companies than listed firms. The elaborate descriptive statistics can be found in table 4, panel A to D.
Table 5 shows the correlation matrix for the control variables. In panel A, the correlation matrix for the three basic types of ownerships is shown. This correlation matrix shows
approximately the same results as table 4. Most of the correlations presented are small, and therefore multicollinearity is not an issue.
