Internationalization and tax avoidance practices of publicly listed firms within the European Union: The influence of board independence

(1)

Internationalization and tax avoidance practices of

publicly listed firms within the European Union:

The influence of board independence

Sander Laurent Swart S2169584

Thesis

MSc. International Financial Management University of Groningen

MSc. Business and Economics University of Uppsala Supervisor: Dr. V. Purice Second assessor: Dr. W. Westerman

January 12, 2018

Key words: Tax avoidance, corporate governance, board independence, publicly listed firms JEL classification: F23, H26, H32

(2)

2

Abstract

(3)

3

1. Introduction

On January 26, 2017 the Austrian Minister of Finance announced: “the cross-border fight against tax fraud and tax avoidance is a top priority for me”1_{. Austria will preside the} European Council (EC) during the second half of 2018, and clearly expressed its intentions towards tax avoidance practices of firms within the European Union. This statement from the Austrian Minister of Finance cannot be seen in isolation since other governments, non-governmental organizations (NGOs), the press, and the public increasingly scrutinize the tax avoidance practices of firms. The case of Apple, Starbucks, and Google for instance, is a clear depiction of the current sentiment regarding these practices. The UK parliamentary public accounts committee summoned these firms to clarify their adopted tax strategies; a hearing about tax strategies that was the first ever in the history of the UK2. Besides aforementioned parties, a substantial base of criticism on tax avoidance practices by firms is found within academic research. For example, Bird and Davis-Nozemack (2016) argue that “tax avoidance suffocates the state’s ability to provide essential services today and in the future” (p. 2). Moreover, West (2017) finds arguments to classify tax avoidance practices as unethical in both the utilitarian perspective of Bentham (1789) and Mill (1863), and in deontological ethics as first introduced by Immanuel Kant (2002, orig. 1785).

Even though a large number of parties increasingly scrutinize the tax avoidance practices of firms, there is a limited amount of research that investigates tax avoidance and its determinants. Chen, Chen, Cheng, and Shevlin (2010) state that “despite the important implications of tax planning for shareholders and regulators, our understanding of the determinants of tax reporting aggressiveness is limited” (p. 44). Furthermore, Desai and Dharmapala (2006) who investigate tax avoidance and its determinants – amongst others – call for additional research on the topic. To advance the existing literature and respond to this call, this research empirically investigates two possible determinants of tax avoidance practices.

Firstly, as is explained later, the very nature of tax avoidance practices makes that firms active in more tax jurisdictions have increased opportunities to avoid taxation. Consequently, the level of internationalization of a firm is the first potential determinant included in this research.

1_{Reuters - http://www.reuters.com/article/us-eu-austria-taxavoidance-idUSKBN1AE0F4?il%3D0/ - Accessed}

August 22, 2017.

2_{The Guardian -}

(4)

4 Secondly, an area that receives increasing attention as a possible influence on tax avoidance practices of firms is the area of corporate governance. This stems from the fact that the benefits and costs of tax avoidance practices are not equally distributed amongst shareholders and executive management (McClure, Lanis, Wells, and Govendir, 2017). From this an incentive misalignment arises, and therefore multiple authors argue that an agency theory perspective (Jensen and Meckling, 1976) should be adopted in the research on tax avoidance practices (e.g., Desai and Dharmapala, 2006). One extensively studied corporate governance mechanism that has shown to be of influence on agency problems and its consequences is board independence. That is, the presence of outside (or independent) board members on the board of directors who are appointed with the sole goal of protecting shareholders’ interests (Rosenstein and Wyatt, 1990) increases the monitoring effectiveness of the board of directors, and hereby induces closer monitoring of the decisions made by executive management (Lanis and Richardson, 2011). Therefore, the independence of the board of directors is included in this research as a potential influence on the relation between the level of internationalization of a firm and its engagement in tax avoidance practices.

Summarizing, the research question is: Does a firm’s level of internationalization

influence its engagement in tax avoidance practices, and is this relation influenced by the independence of the board of directors?

The results show that both potential determinants are related to tax avoidance practices of firms. The level of internationalization decreases the engagement in tax avoidance practices, and this relation is found to be weakened by the independence of the board of directors. That is, increased foreign presence decreases the amount of taxation that is avoided, but this relation becomes less strong as board independence increases.

(5)

(6)

6

2. Literature review

Even though the literature on tax avoidance is relatively young (Hanlon and Heitzman, 2010), tax avoidance by firms is discussed in several areas. A substantial part of the literature on tax avoidance is found within the ethical domain (e.g., Prebble and Prebble, 2010; Bird and Davis-Nozemack, 2016). Moreover, it is found in research on, amongst others, corporate social responsibility (e.g., Preuss, 2010; Huseynov and Klamm, 2012; Lanis and Richardson, 2015), ownership structure (e.g., Chen et al., 2010; Richardson, Wang, and Zhang, 2016), and corporate governance (e.g., Desai and Dharmapala, 2006; Chang, Mo, and Zhou, 2013; Armstrong, Blouin, Jagolinzer, and Larcker, 2015). This section continues by defining tax avoidance, describing theories on the benefits and costs of tax avoidance, elaborating on the literature on the relation between multinationality and tax avoidance, and finally discussing the literature on the link between board independence and tax avoidance. From these discussions hypotheses are derived that are empirically tested in later sections.

Tax avoidance definition

(7)

7 Multinationality

After having defined tax avoidance practices, it is important to specify these practices to understand the possible relation between other constructs and these practices. Therefore, subsequent paragraphs provide an overview of the existing literature identifying different manners or practices in which tax is avoided.

Jansky and Prats (2015) state three types of tax avoidance practices. The first type is the distortion of intra-firm transfer pricing. Firms with subsidiaries operating in two or more tax jurisdictions have the possibility to influence the price of intra-firm trade. Hereby the profit shifts to the selling subsidiary if a relatively high price for the intra-firm trade is determined, and the profit shifts to the purchasing subsidiary if a relatively low price is set. In this way firms are able to shift profits to subsidiaries located in low tax jurisdictions. Consequently, the overall tax burden of the firm decreases. As the price of the intra-firm trade is established mainly considering the allocation of profits, this is regarded as a practice intended to avoid taxation. The second type mentioned by the authors is the distortion of the capital structure of the firm. Firms may rely more on debt financing in high tax jurisdictions since the interest payments on debt are tax deductible, while relying more on equity financing in low tax jurisdictions. Amongst others, Collins and Shackelford (1992) and Grubert (1998) find evidence of such practices by multinationals. Lastly, Jansky and Prats (2015) argue that firms shift profits by allocating certain costs and assets (often highly profitable intangible assets) to specific subsidiaries. For example, a highly profitable intangible asset as a patent can be allocated to a subsidiary in a low tax jurisdiction motivated mainly by tax considerations. Arguably, this could be classified as a tax avoidance practice.

(8)

8 From this discussion it can be concluded that there potentially is a relation between the level of internationalization of a firm and its engagement in tax avoidance practices. Even though conflicting evidence exists on the sign or direction of the relation, prior literature seems to support the notion that the number of tax jurisdictions in which a firm is present influences the engagement in tax avoidance practices of that firm. Moreover, even though the sign predictions are debated, rationally, it is more reasonable to expect that firms that have more opportunities to avoid taxation seize these opportunities and lower their overall tax burden accordingly. Formalized, I expect to find a positive relation between the level of internationalization of a firm and its engagement in tax avoidance practices. Therefore, it is hypothesized that the level of internationalization of a firm positively influences the engagement in tax avoidance practices.

(9)

9 Benefits and costs of tax avoidance

Following the discussion on how tax avoidance is defined and which practices are utilized by firms to avoid taxation, the focus shifts to the benefits and costs of tax avoidance for firms. As stated, rationally one would expect that a firm always strives to reduce its tax burden and therefore would maximize its tax avoidance practices. More specifically, by engaging in tax avoidance practices wealth is transferred from the government of a country to the shareholders of a firm, and is therefore shareholder value maximizing (Khurana and Moser, 2012). However, prior research has theorized that tax avoidance is not only associated with benefits for a firm, but also with potential costs. Therefore, successive paragraphs discuss these benefits and costs. Three potential benefits of tax avoidance are identified. Firstly, as described in the previous section, engaging in tax avoidance practices reduces the overall tax burden of a firm. Since a smaller part of the profit before tax is paid to the government via taxation, more profit is retained in the firm or can be distributed to shareholders. Secondly, the lower tax burden is positively received by the shareholders and therefore the cost of equity decreases (Chi, Pincus, and Teoh, 2014; Inger, 2014; McGuire, Omer, and Wilde, 2014). An important detail to consider though, Inger (2014) shows that the investor response is dependent on which tax avoidance practices are utilized by a firm. Thirdly, the lower tax burden can increase the retained earnings of a firm that could be allocated to increase liquidity (Saavedra, 2013).

(10)

10 Board independence

Derived from the discussion on the benefits and costs of tax avoidance practices, it becomes apparent that such practices can be costly to shareholders of a firm, or at least not beneficial for shareholders under specific circumstances. Simultaneously it has to be noted that the potential costs mentioned are for a significant part originated from the provided room for managers to act in their self-interest that is not necessarily in the interest of shareholders. Consequently, Desai and Dharmapala (2006) argue that corporate tax avoidance should be investigated adopting an agency theory perspective (Jensen and Meckling, 1976). That is, the situation originated by tax avoidance practices is one in which an incentive misalignment between shareholders and managers exists, and therefore should be classified as a problem related to separation of ownership and control. As a result, research on tax avoidance has been previously linked to the corporate governance structure of a firm (Desai and Dharmapala, 2006). Whereas these authors incorporated high-powered incentives, I incorporate board independence in this research. Previous research on the topic has shown significant relations between board independence and tax avoidance, and therefore it is of potential influence on the relation between the level of internationalization of a firm and its engagement in tax avoidance practices.

Preliminary to reviewing this research, it is important to define board independence and consider the premise on which the link between board independence and tax avoidance is based. First of all, board independence in this research is defined as the percentage of inside vs. outside

board members. Outside members are not part of management and are considered to be

(11)

11 Then, the remaining question is whether the increased board independence results in more or less engagement in tax avoidance practices by firms. From the discussion of the benefits and costs of tax avoidance earlier, it is really the question what the (potential) benefits and costs of tax avoidance practices are as perceived by the shareholders. If the benefits outweigh the cost, increased board independence would stimulate engagement in tax avoidance practices by executive management. Naturally, the opposite is equally true. If the costs outweigh the benefits, increased board independence would decrease engagement in tax avoidance practices by executive management.

From this reasoning different hypotheses on sign of the relation between board independence and tax avoidance could be derived. As described in the introduction, the scrutiny of tax avoidance practices has increased over the years. This leads to increased chances of reputational damage, fines from tax authorities, and therefore to higher potential costs of tax avoidance practices. Consequently, this research argues that it is expected that firms engage in tax avoidance practices less if management is more closely monitored by independent boards. Firms with more independent boards will seize less opportunities than firms with less independent boards. Therefore, it is hypothesized that board independence weakens the relation between the level of internationalization and tax avoidance practices.

H2: As the independence of the board increases, the relation between the level of internationalization of a firm and its engagement in tax avoidance practices weakens.

In Figure 1 below, both the hypothesis on the link between the level of internationalization and tax avoidance, and the hypothesis on the influence of board independence on the relation between the level of internationalization and tax avoidance are visualized in a conceptual model.

Figure 1: Conceptual model

(12)

12

3. Methodology

Next, this section describes the research design selected to test the hypothesized relations between the level of internationalization, board independence, and tax avoidance practices. First, Sections 3.1 to 3.4 describe the variable construction. Then, Section 3.5 shows the model specifications of the Pooled Ordinary Least Squares (OLS) regressions adopted in this research.

Tax Avoidance

As Hanlon and Heitzman (2010) state in their paper: “A primary issue in the empirical tax avoidance literature is the researcher’s definition and measurement of tax avoidance” (p. 128-129). The authors emphasize that researchers should be aware of the different measurements of tax avoidance available and their limitations. Moreover, processing the data collected in the different measurements requires great caution since this information is made public by firms and therefore is biased. That is, the information required to accurately measure tax avoidance may not be the information that is obtained from firms’ financial statements. However, Gupta and Newberry (1997) argue that obtaining data from financial statements is the only plausible way to gather data within the time and resource constraints. Moreover, Zimmerman (1983) concludes that data gathered from financial statements “yield unbiased estimates of effective tax rates” (p. 137).

3.1.1 Measurement

In their article, Hanlon and Heitzman (2010) discuss twelve measures of tax avoidance and their applicability to different types of research (see: Table 1, p. 140). It is believed that the GAAP effective tax rate (ETR) is the most appropriate measure for this research setting. The GAAP ETR measure is operationalized as the income tax expense divided by the pre-tax book income. Despite the fact that this research recognises the flaws of the GAAP ETR measure3, it is seen as the most appropriate measure of tax avoidance. Even though not all tax avoidance strategies are captured in the GAAP ETR, it provides an indication to what extent firms are involved in tax avoidance practices. Furthermore, prior literature on the topic of tax avoidance commonly employs the GAAP ETR as a measure of tax avoidance (e.g., Hanlon and Slemrod, 2009; Chen et al., 2010). Lastly, other available methods (e.g., Cash ETR, book-tax differences, etc.) are flawed in different ways, as is discussed by Hanlon and Heitzman (2010).

3_{Hanlon and Heitzman (2010) argue that methods focussed on measuring the GAAP effective tax rate of firms}

(13)

13 However, since the imperfections in the measure are acknowledged, robustness analysis will be performed to assess the quality of the GAAP ETR measure. The robustness analysis includes the raw book-tax gap (BTG) as another measure of tax avoidance. The book-tax gap has been included as a measure of tax avoidance by, amongst others, Desai and Dharmapala (2006), Chen et al. (2010), and Richardson et al. (2016). The raw book-tax gap is the difference between the accounting income and the taxable income (Richardson et al., 2016). It is calculated by subtracting the taxable income from the pre-tax income (Manzon and Plesko, 2002) and then dividing this by the total assets (Richardson et al., 2016). Moreover, the taxable income is calculated as the income tax expense divided by the statutory tax rate4_{(Manzon and Plesko,} 2002).

Multinationality

Reeb, Kwok, and Baek (1998) emphasize that in previous research various methods have been used to measure the level of internationalization of a firm. Moreover, Sullivan (1984) and Burgman (1996) argue that there is no consensus on the ‘right’ way of measuring the level of internationalization. Both the foreign sales ratio (foreign sales as a percentage of total sales) and the foreign asset ratio (foreign assets as a percentage of total assets) have been used separately or together (e.g., Rego, 2003) in prior research. However, since the foreign sales ratio might mix international trade and international investment (Lee and Kwok, 1988; Burgman, 1996), the foreign asset ratio (FAR) is included as the proxy of the level of internationalization.

Board independence

Board independence (OBM) is measured as the percentage of outside board members on the board of directors. It is expected that the presence of more outside board members will increase independent monitoring of executive management (e.g., Fama and Jensen, 1983). However, as discussed in Section 2.4, opposing views exist on the result of increased monitoring. That is, increased monitoring could result in lower effective tax rates if one is to follow the reasoning that tax avoidance practices increase shareholder wealth. Alternatively, increased monitoring could result in higher effective tax rates because the potential costs outweigh the benefits.

4_{Statutory tax rates for the individual countries over the years 2008-2016 are collected from KPMG.}

(14)

14 Control variables

Derived from the approach of, amongst others, Rego (2003), Chang et al., (2013), and Richardson et al. (2016), eight control variables are included. Firstly, there is a need to control for leverage (LEV). As Gupta and Newberry (1997), and Chang et al. (2013) argue, firms with higher financial leverage have a lower ETR because interest payments are deductible. Secondly, this research controls for the market-to-book ratio (MTB) of firms since prior research (Chen et al., 2010) has shown a correlation between the market-to-book ratio and tax avoidance. Thirdly, the asset mix (CAP) (capital intensive vs. inventory intensive firms) is included (Gupta and Newberry, 1997). The authors argue that the tax deduction of depreciation leads to a lower ETR for a capital intensive firm. Moreover, firm size (SIZ) is included as control variable. However, opposing views on its influence are found in prior literature (Gupta and Newberry, 1997). The researchers arguing that firm size positively influences the ETR state that larger firms have more resources to engage in tax planning and therefore should have a lower ETR (Siegfried, 1972). The opposing view of Zimmerman (1983) argues that larger firms should have a higher ETR since these firms are more closely watched and scrutinized, and therefore have a higher ETR. Finally, the return on assets (ROA) is incorporated as a fifth control variable. Gupta and Newberry (1997) argue that the profitability influences the book income, and therefore this research controls for profitability.

(15)

15 Table 1: Variable overview and measurement

Variable Abbreviation Description Operationalization Measurement

Dependent ETR Tax avoidance Effective tax rate Income tax expense / Pre-tax book income

BTG Tax avoidance Book-tax gap (Pre-tax income - Taxable income5) / Total assets

Independent FAR Level of internationalization Foreign asset ratio Foreign assets / Total assets

Moderating OBM Inside vs. outside board

members

Percentage of outside board members

Control LEV Firm leverage Leverage Total liabilities / Total assets

MTB Market-to-book Market-to-book ratio Market value of equity / Book value of equity CAP Capital vs. inventory

intensive firms

Capital intensity Plant, property and equipment / Total assets SIZ Firm size Total assets Natural logarithm of total assets

ROA Return on assets Return on assets Net income / Total assets

YDM Year Dummy variable

IDM Industry Dummy variable SIC codes

CDM Country Dummy variable

(16)

16 Regression Analysis

Pooled Ordinary Least Squares (OLS) regressions will be performed to estimate the hypothesized relationships. The reasons to select a Pooled OLS estimator for this research are twofold. First of all, fixed effects models are not able to estimate the influence of a variable if this variable does not vary over time (Brooks, 2014). Since both variables that are of particular interest for this research (FAR and OBM) show only little fluctuation over time, a fixed effects model is limited in estimating the influence of these variables on the tax avoidance practices of firms. Secondly, a Hausman (1978) test shows that the random effects model produces inconsistent results, and therefore this model is rejected. Even though the Pooled OLS estimator does not control for unobserved firm characteristics, this research controls for both industry and country fixed effects, and thereby accounts for a part of these characteristics. Moreover, prior research on tax avoidance has incorporated similar estimation methods (e.g., Richardson and Lanis, 2007). However, tests for potential multicollinearity, heteroskedasticity, and autocorrelation are included in Sections 4.1 and 4.2 to further investigate the appropriateness of the Pooled OLS estimator.

The first regression equation only includes the dependent, a constant, and the independent variable. Consequently, regression equation 1 is depicted as,

𝐸𝑇𝑅𝑖,𝑡 = 𝛽0+ 𝛽1𝐹𝐴𝑅𝑖,𝑡+ 𝜀𝑖,𝑡 (1)

Hereafter, the control variables are included in the regression equation. Regression equation 2 is estimated as,

𝐸𝑇𝑅_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝐹𝐴𝑅_𝑖,𝑡+ 𝛽₂𝐿𝐸𝑉_𝑖,𝑡+ 𝛽₃𝑀𝑇𝐵_𝑖,𝑡 + 𝛽₄𝐶𝐴𝑃_𝑖,𝑡+ 𝛽₅𝑆𝐼𝑍_𝑖,𝑡+ 𝛽₆𝑅𝑂𝐴_𝑖,𝑡 + 𝑌𝑒𝑎𝑟 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜀𝑖,𝑡

(2)

The third regression includes the moderating variable, and is estimated as,

𝐸𝑇𝑅_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝐹𝐴𝑅_𝑖,𝑡+ 𝛽₂𝑂𝐵𝑀_𝑖,𝑡 + 𝛽₃𝐿𝐸𝑉_𝑖,𝑡+ 𝛽₄𝑀𝑇𝐵_𝑖,𝑡+ 𝛽₅𝐶𝐴𝑃_𝑖,𝑡+ 𝛽₆𝑆𝐼𝑍_𝑖,𝑡+

𝛽₇𝑅𝑂𝐴_𝑖,𝑡+ 𝑌𝑒𝑎𝑟 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 + 𝜀_𝑖,𝑡 (3)

Lastly, in the fourth regression the interaction term between the independent variable and the moderating variable is added to the model,

(17)

17

4. Data

The focus in this research is on publicly listed firms within the EU. Since accounting standards influence financial statements of firms - and publicly listed firms within the EU have to comply with the International Financial Reporting Standards, or IFRS (Doupnik and Perera, 2015, p. 96) - all sample firms have to adhere to the same accounting standards. Data on these firms are collected for the period 2008-2016. This period is chosen because of two reasons. Firstly, it encompasses the most recent data available. Secondly, in 2010, Hanlon and Heitzman predicted that “the relevance of tax avoidance research will increase as governments try to close the tax gap, increase compliance, and collect more revenue” (p. 168). The predicted increased efforts of government to reduce tax avoidance practices could become visible in this time period. Consequently, data on publicly listed firms are retrieved from Thompson Reuters Datastream spanning the years 2008-2016. An initial sample of 8,505 firm-year observations on 945 firms is collected from the database. Financial firms (Gupta and Newberry, 1997; Rego, 2003; Chang et al., 2013; Richardson et al., 2016) and utility firms (Gupta and Newberry, 1997; Rego, 2003) are excluded from the sample because both type of firms face regulatory constraints significantly affecting their effective tax rate (Gupta and Newberry, 1997). Moreover, all firm-year observations with missing data on the effective tax rate are deleted (Rego, 2003; Chang et al., 2013). Lastly, cases where the effective tax rate is negative or above 1 are deleted (Chen et al., 2010; Richardson et al., 2016). Table 2 below shows the sample selection procedure.

Table 2: Sample selection

Firm-year observations

Initial sample: 2008-2016 8,505

Less: Financial firms (2,097)

Utility firms (369)

Missing values on effective tax rate (164)

Negative effective tax rate or effective tax rate >1 (757)

Final sample 5,118

(18)

18 Table 3: Division of sample firm-year observations by country, industry, and year

# of firm-year observations % of firm-year observations

(19)

19 As Table 3 shows, firm-year observations from the United Kingdom constitute a relative large part of the final sample. To control for the influence of these firms, country dummies are added to the regression models, as is discussed in Section 3.4. Also, robustness analysis is conducted to test for the influence of the UK firms. From Table 3 there is no evidence that a bias towards a certain industry or year is present in the data.

Table 4 below presents the descriptive statistics of the final sample. To exclude extreme values, all variables are winsorized (0.05, 0.95). Exceptions are the effective tax rate since this variable in truncated to only include values between 0 and 1, and the firm size because the natural logarithm of the total assets is used. The mean of the effective tax rate is 0.265 and has a standard deviation of 0.138. The observed median is 0.258. Second, the book-tax gap is expected to show smaller statistics since it is the difference between the accounting income and the taxable income (Richardson et al., 2016). The collected data for this variable show a mean of -0.002, a standard deviation of 0.029, and a median of 0.000. Third, the foreign asset ratio shows a mean of 0.312, with a standard deviation of 0.280, and a median of 0.267. Lastly, the percentage of outside board members on average is 0.474, with a median of 0.500 and a standard deviation of 0.248. All in all, the descriptive statistics on the dependent, independent, and moderating variable have values that seem reasonable and within the expected boundaries. Table 4: Descriptive statistics

N Mean Std. Dev. Median Minimum Maximum

ETR 5,118 0.265 0.138 0.258 0.000 1.000 BTG 5,042 -0.002 0.029 0.000 -0.068 0.055 FAR 4,106 0.312 0.280 0.267 0.000 0.843 OBM 4,332 0.474 0.248 0.500 0.000 0.889 LEV 5,117 0.582 0.172 0.585 0.261 0.893 MTB 3,645 2.760 2.043 2.120 0.600 8.345 CAP 5,111 0.255 0.195 0.212 0.021 0.686 SIZ 5,117 14.992 1.599 14.926 8.896 19.862 ROA 5,045 0.072 0.054 0.063 -0.016 0.202

i: For variable definitions see variable overview in Table 1.

(20)

20 Multicollinearity

Next, Table 5 on the next page shows the correlation matrix. The effective tax rate and the book-tax gap are negatively correlated, as expected. That is, whereas tax avoidance practices lead to a higher book-tax gap, they result in a lower effective tax rate. Following that reasoning, the two measures should be negatively correlated. It is important to note that the correlation matrix indicates a moderate significant correlation coefficient of -0.476 between the GAAP effective tax rate and the book-tax gap. This means that the variables measure different tax avoidance practices and/or measurement errors are present (Chen et al., 2010). A potential explanation for this is the fact that both the GAAP effective tax rate and the book-tax gap measure tax avoidance, but from a different perspective. For example, deferral strategies are not reflected in the GAAP effective tax rate whereas the book-tax gap does reflect these strategies (Hanlon and Heitzman, 2010). Therefore, the explanation of Chen et al., (2010) that different aspects of tax avoidance are captured by these measures is likely. Moreover, Chen et al., (2010) find correlation coefficients of -0.361 (significant at the 1 per cent level) whereas the observed correlation coefficients in this research (also significant at the 1 per cent level) are higher. Therefore, the correlation between the GAAP effective tax rate and the book-tax gap is not uncommon.

Finally, the largest significant correlation statistic is 0.513 for the correlation between the return on assets and the market-to-book ratio, indication that firms with higher market-to-book ratios have higher return on assets. All in all, the levels of correlation between the control variables are low and therefore multicollinearity is considered not to be an issue.

Heteroskedasticity and autocorrelation

(21)

21 Table 5: Pearson correlation matrix

ETR BTG FAR OBM LEV MTB CAP SIZE ROA

ETR --- BTG -0.476*** --- 0.000 FAR 0.095*** -0.091*** --- 0.000 0.000 OBM -0.045** -0.035* 0.113*** --- 0.023 0.077 0.000 LEV 0.073*** -0.031 -0.047** -0.048** --- 0.000 0.115 0.019 0.017 MTB -0.049** -0.018 -0.033 0.035* 0.136*** --- 0.015 0.372 0.102 0.080 0.000 CAP 0.023 -0.031 -0.002 -0.052** -0.092*** -0.150*** --- 0.243 0.124 0.910 0.010 0.000 0.000 SIZE 0.098*** 0.066*** 0.144*** -0.119*** 0.223*** -0.199*** 0.050** --- 0.000 0.001 0.000 0.000 0.000 0.000 0.013 ROA -0.202*** 0.200*** -0.047** 0.105*** -0.259*** 0.513*** -0.042** -0.276*** --- 0.000 0.000 0.018 0.000 0.000 0.000 0.035 0.000

(22)

22

5. Results

After inspection of the data, Table 6 shows the results of the Pooled OLS regression with the GAAP effective tax rate as dependent variable. To enhance the readability of the research, the ETR variable is transformed by multiplying the variable by -1 (Richardson et al., 2016). Hereby, a positive sign for the independent and control variables in the models represent an increase in tax avoidance. Similarly, a negative sign means tax avoidance practices decrease. The independent, control, and moderating variables are added gradually through the models. Table 6: OLS regression results - ETR

Variable Model 1 Model 2 Model 3 Model 4

FAR -0.034*** -0.038*** -0.044*** 0.001 (-2.87) (-2.76) (-2.97) (0.04) OBM 0.006 0.040 (0.39) (1.59) OBM*FAR -0.092* (-1.69) LEV 0.005 0.003 0.004 (0.20) (0.14) (0.16) MTB -0.007*** -0.008*** -0.008*** (-3.46) (-3.59) (-3.65) CAP -0.048** -0.048** -0.047** (-2.24) (-2.14) (-2.12) SIZ 0.002 0.001 0.001 (0.59) (0.32) (0.23) ROA 0.628*** 0.609*** 0.610*** (7.02) (6.88) (6.91) Constant -0.255*** -0.303*** -0.301*** -0.315*** (-54.90) (-6.18) (-5.38) (-5.49)

Year dummies No Yes Yes Yes

Industry dummies No Yes Yes Yes

Country dummies No Yes Yes Yes

N 4,106 2,881 2,529 2,529

Adjusted-R2 0.004 0.129 0.135 0.137

F-statistic 8.26*** 18.23*** 20.92*** 20.93***

(23)

23 First, models 1 and 2 show a significant negative influence of the level of internationalization on tax avoidance, as measured by the foreign asset ratio. The coefficients of -0.034 and -0.038 are significant at the 1 per cent level. As can be seen from model 2, adding the control variables to the model increases the adjusted R-squared significantly to 0.129. The market-to-book ratio and capital intensity are negatively related to tax avoidance, and significant at the 1 and 5 per cent level, respectively. On the market-to-book ratio prior research shows both a positive and negative influence on tax avoidance (Richardson et al., 2016). In this research the sign is negative, meaning that firms with a higher market-to-book ratio engage less in tax avoidance practices. For the capital intensity the findings are inconsistent with prior research (e.g., Gupta and Newberry, 1997). Lastly, also the prior findings on the return on assets are mixed (Richardson et al., 2016), the results in Table 6 show that the return on assets positively influences tax avoidance.

(24)

24 Second, in model 3 the moderating variable is added to test if the percentage of outside board members in itself influences the tax avoidance practices of a firm. Since the coefficient of 0.006 is insignificant, no direct relation between the percentage of outside board members and tax avoidance practices is observed. That is, the results show that the engagement of executive management in tax avoidance practices is not dependent on the independence of the board considering all firms.

However, in model 4 the cross-term is significant at the 5 per cent level. The coefficient of the cross-term of -0.092 indicates that the influence of the percentage of outside board members is significantly influencing the relation between the level of internationalization and the engagement in tax avoidance practices. That is, comparing model 3 and 4, it seems that the independence of the board of directors becomes relevant if the firm has a foreign presence. The results show that the independence of the board of directors only influences the tax avoidance practices of firms if these firms have a presence outside the home country. Moreover, this influence is negative which means that the negative relation between the level of internationalization and tax avoidance practices of firms is weakened by a more independent board. Differently formulated, if one is to compare two firms with the same level of internationalization, the firm with more independent board members engages more in tax avoidance since the negative influence of the level of internationalization on the engagement in tax avoidance practices is decreased more for this firm than for the firm with less independent board members.

Even though a negative moderating influence was expected and formulated in hypothesis 2, and the results confirm this negative moderating influence, the hypothesis is rejected. Despite the fact that this seems counterintuitive, the implication of confirming the hypothesis would be that the tax avoidance practices of firms increase less with an increase in the percentage of outside board members. Based on models 1 to 4, supporting this premise would be incorrect because the tax avoidance practices decrease less with a higher percentage of outside board members. That is, in firms with a foreign presence the results show that a more independent board of directors stimulates tax avoidance practices instead of rejecting the engagement in these practices.

(25)

25 prior research has found other benefits of tax avoidance that would be in the interest of shareholders, and therefore encourage independent board members to pursue tax avoidance strategies. For example the cost of equity declines because a lower tax burden is a positive signal to investors (Chi et al., 2014; Inger, 2014; McGuire et al., 2014), and the cash and liquidity of a firm increases (Saavedra, 2013).

(26)

26 Additional analysis

After establishing the general relation between the level of internationalization and the engagement in tax avoidance practices of firms, additionally it is interesting to analyse if this relation holds if the foreign asset ratio is transformed into a dummy variable. Hereby, the sample is divided in a group with low or moderate foreign asset ratios, and a group with high foreign asset ratios. Consequently, alongside including a continuous variable for the foreign asset ratio, the foreign asset ratio is transformed into a dummy variable to test for potential differences in the aforementioned relations considering firms with a low foreign presence (FAR<=0.7) and a high foreign presence (FAR>0.7). The descriptive statistics of the respective groups are shown in Appendix A.1. The results (Appendix A.2) for the low foreign presence group show that the relation between the level of internationalization, and the influence of board independence on this relation are similar to the results reported in the previous section. Contrastingly, the result for the high foreign presence group (Appendix A.3) show that neither the foreign asset ratio, nor the percentage of outside board members significantly influences tax avoidance practices of firms.

(27)

27 Robustness analysis

To verify the results, various robustness tests are performed. First, as mentioned in Section 3.1.1, the book-tax gap is introduced as dependent variable. Hereafter, subsamples are created based on the country of the firm-year observations to address the potential bias of the UK, as discussed in Section 4.

5.2.1 Book-tax gap

To verify the results of the OLS regression in Table 6, a different measure of tax avoidance is incorporated in the Pooled OLS regressions. That is, the regression equations described in Section 3.5 remain unchanged except for the dependent variable. The dependent variable GAAP ETR is replaced by the book-tax gap to test if the GAAP effective tax rate is an appropriate measure to investigate tax avoidance practices by firms. The results of the regressions including the book-tax gap as the dependent variable are shown in Table 7.

Table 7: OLS regressions results - BTG

FAR -0.005** -0.007** -0.008*** 0.002 (-2.02) (-2.28) (-2.60) (0.30) OBM 0.001 0.008 (0.22) (1.48) OBM*FAR -0.020* (-1.76) LEV 0.009 0.009* 0.009* (1.60) (1.65) (1.67) MTB -0.003*** -0.002*** -0.002*** (-5.10) (-4.89) (-4.92) CAP -0.008* -0.007 -0.007 (-1.91) (-1.51) (-1.47) SIZ 0.001 0.001 0.001 (0.74) (0.87) (0.79) ROA 0.178*** 0.181*** 0.181*** (9.26) (9.00) (9.03) Constant 0.000 -0.020* -0.024* -0.027** (0.27) (-1.81) (-1.84) (-2.06)

N 4,053 2,866 2,514 2,514

Adjusted-R2 0.002 0.172 0.174 0.176

F-statistic 4.08** 9.61*** 9.49*** 9.37***

(28)

28 First, the results of the regressions confirm the relation found between the level of internationalization and tax avoidance. In models 1 and 2, the coefficient of the foreign asset ratio is negative and significant at the 5 per cent level. These results are similar to the result obtained from the initial regressions. Moreover, models 3 and 4 similarly show no relation between the percentage of outside board members and tax avoidance, and a negative relation between the percentage of outside board members and the relation between the foreign asset ratio and tax avoidance, respectively. The difference between models 1-4 with the GAAP ETR as dependent variable and models 1-4 with the book-tax gap as dependent variable is that firm leverage becomes significant (at the 10 per cent level) in the latter case. Moreover, whereas the capital intensity variable coefficient is negative and significant in the models with GAAP ETR as dependent variable, it is not significant in the models incorporating the book-tax gap as dependent variable. Only in model 2 the coefficient is significant, with a p-value of 0.056. All in all, the results of the initial analysis are similar to the results of the robustness regressions, and therefore the observed effects are robust. Consequently, the robustness check confirms the GAAP effective tax rate as appropriate measure of tax avoidance.

5.2.2 Country sampling

(29)

29

6. Conclusion

This research commenced with the notion that firms engaging in tax avoidance practices are under increased scrutiny of politics, the press, the general public, and academics. Therefore, it is valuable to investigate these practices and their determinants. The results of the regression analyses show that lower levels of tax avoidance are associated with publicly listed firms within the European Union that have a higher level of internationalization. Moreover, the findings show that this relation is weakened by the board independence of the board of directors of the firm. However, from the results it seems that this relation between the level of internationalization and tax avoidance practices of firms, and the influence of board independence on this relation, ceases to exist from a certain level of internationalization. Moreover, the results show that the relations found vary from country to country.

The contribution to the existing literature is threefold. Firstly, this research advances the literature on tax avoidance practices of publicly listed firms by adding an empirical analysis on several potential determinants. Secondly, the call of, amongst others, Hanlon and Heitzman (2010) to research determinants of tax avoidance in a principal-agent setting is answered. And finally, this research contributes to the literature on corporate governance in relation to tax avoidance.

This research is bound to several limitations. Firstly, the selection of a Pooled OLS estimator to investigate the interdependencies between phenomena poses several issues. This estimator does only partly (via industry, and country dummies) control for unobserved firm characteristics. Whereas I am aware of the implications of potential endogeneity problems, the limited variation in both the independent and moderating variable necessitated incorporating the Pooled OLS estimator. However, future research could include other measures of especially the level of internationalization which vary more over time (e.g., foreign sales ratio). In that case fixed effects models could be used that control for unobserved firm characteristics.

(30)

30 Thirdly, this research only incorporates publicly listed firms due to data availability. It could be that results differ for private firms since these firms have to provide only limited insight in their business. Hereby, these firms are less prone to public scrutiny and the decision to engage in tax avoidance practices might therefore be dependent on a different risk-return trade off.

Finally, the data that form the foundation of the analysis in this research are derived from financial statements, and likely to be biased. Publicly listed firms are aware of the increasing attention to tax avoidance practices and are expected to incorporate this consideration when publishing their financial statements. Even though these firms are bound by rules and regulations in reporting their financial position, arguably firms would seek the boundaries of these rules and regulations to optimize their perceived position and behaviour.

(31)

31

7. References

Armstrong, C., Blouin, J., Jagolinzer, A., Larcker, D., 2015. Corporate governance, incentives, and tax avoidance. Journal of Accounting and Economics 60, 1-17.

Bentham, J., 1789. An Introduction to the Principles of Morals and Legislation. Dover, Mineola, New York.

Bird, R., Davis-Nozemack, K., 2016. Tax avoidance as a sustainability problem. Journal of Business Ethics, 1-17.

Brooks, K., 2014. Introductory Econometrics for Finance. Cambridge University Press, Cambridge.

Burgman, T., 1996. An empirical examination of multinational capital structure. Journal of International Business Studies 27, 553-570.

Chang, K., Mo, P., Zhou, A., 2013. Government ownership, corporate governance and tax aggressiveness: evidence from China. Accounting and Finance 53, 1029-1051.

Chen, S., Chen, X., Cheng, Q., Shevlin, T., 2010. Are family firms more tax aggressive than non-family firms? Journal of Financial Economics 95, 41-61.

Chi, S., Pincus, M., Teoh, S., 2014. Mispricing of book-tax differences and the trading behavior of short sellers and insiders. Accounting Review 89, 511-543.

Collins, J., Shackelford, D., 1992. Foreign tax credit limitations and preferred stock issuance. Journal of Accounting Research 30, 103-124.

Desai, M., Dharmapala, D., 2006. Corporate tax avoidance and high-powered incentives. Journal of Financial Economics 79, 145-179.

Desai, M., Dharmapala, D., 2009. Corporate tax avoidance and firm value. Review of Economics and Statistics 91, 537-546.

Doupnik, T., Perera, H., 2015. International Accounting. McGraw-Hill Education, New York. Dyreng, S., Hanlon, M., Maydew, E., 2010. The effects of executives on corporate tax avoidance. Accounting Review 85, 1163-1189.

(32)

32 Fama, E. 1980. Agency problems and the theory of the firm. Journal of Political Economy 88, 288-308.

Fama, E., Jensen, M., 1983. Separation of ownership and control. Journal of Law Economics 26, 301-325.

Grubert, H., 1998. Taxes and the division of foreign operating income among royalties, interest, dividends and retained earnings. Journal of Public Economics 68, 269-290.

Gupta, S., Newberry, K., 1997. Determinants of the variability in corporate effective tax rates: evidence from longitudinal data. Journal of Accounting and Public Policy 16, 1–34.

Hanlon, S., Slemrod, J., 2009. What does tax aggressiveness signal? Evidence from stock price reactions to news about tax shelter involvement. Journal of Public Economics 93, 126-141. Hanlon, S., Heitzman, S., 2010. A review of tax research. Journal of Accounting and Economics 50, 127-178.

Hausman, J., 1978. Specification tests in econometrics. Econometrica 46, 1251-1271.

Huseynov, F., Klamm, B., 2012. Tax avoidance, tax management, and corporate social responsibility. Journal of Corporate Finance 18, 804-827.

Inger, K., 2014. Relative valuation of alternative methods of tax avoidance. Journal of the American Taxation Association 36, 27-55.

Jansky, P., Prats, A., 2015. International profit-shifting out of developing countries and the role of tax havens. Development Policy Review 33, 271-292.

Jensen, M., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.

Kant, I., 2002 (orig. 1785). Groundwork for the Metaphysic of Morals (Wood, A., trans.). Yale University Press, New Haven, Connecticut.

Kim, J., Li, Y., Zhang, L., 2011. Corporate tax avoidance and stock price crash risk: firm-level analysis. Journal of Financial Economics 82, 639-662.

Khurana, I., Moser, W., 2012. Institutional shareholders’ investment horizons and tax avoidance. Journal of the American Taxation Association 35, 111-134.

(33)

33 Lanis, R., Richardson, G., 2015. Is corporate social responsibility performance associated with tax avoidance? Journal of Business Ethics 127, 439-457.

Lee, K., Kwok, C., 1988. Multinational corporations vs. domestic corporations: international environmental factors and determinants of capital structure. Journal of International Business Studies 19, 195-217.

Manzon, G., Plesko, G., 2002. The relation between financial and tax reporting measures of income. Tax Law Review 55, 175-214.

McClure, R., Lanis, R., Wells, P., Govendir, B., 2017. The impact of dividend imputation on corporate tax avoidance and controlling for outside director monitoring. Journal of Corporate Finance, in press.

McGuire, S., Omer, T., Wilde, J., 2014. Investment opportunity sets, operating uncertainty, and capital market pressure: determinants of investments in tax shelter activities? Journal of the American Taxation Association 36, 1-26.

Mill, J., 1863. Utilitarianism. Parker, Son, and Bourn, London.

Mills, L., Erickson, M., Maydew, E., 1998. Investments in tax planning. Journal of the American Taxation Association 20, 1–20.

Newey, W., West, K., 1987. A simple positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703–708.

Prebble, Z., Prebble, J., 2010. The morality of tax avoidance. Creighton Law Review 43, 693– 745.

Preuss, L., 2010. Tax avoidance and corporate social responsibility: you can't do both, or can you? Corporate Governance 10, 365-374.

Reeb, D., Kwok, C., Baek, H., 1998. Systemic risk of the multinational corporation. Journal of International Business Studies 29, 263-279.

Rego, S., 2003. Tax-avoidance activities of U.S. multinational corporations. Contemporary Accounting Research 40, 805-833.

Rego, S., Wilson, R., 2012. Equity risk incentives and corporate tax aggressiveness. Journal of Accounting Research 50, 775-810.

(34)

34 Richardson, G., Wang, B., Zhang, X., 2016. Ownership structure and corporate tax avoidance: evidence from publicly listed private firms in China. Journal of Contemporary Accounting and Economics 12, 141-158.

Rosenstein, S., Wyatt, J., 1990. Outside directors, board independence, and shareholder wealth. Journal of Financial Economics 26, 175-191.

Saavedra, D., 2013. Analysis of unsuccessful tax avoiders. Unpublished working paper. University of California, Los Angeles.

Siegfried, J., 1972. The relationship between economic structure and the effect of political influence: empirical evidence from the federal corporation income tax program. Unpublished dissertation. University of Wisconsin, Madison, Wisconsin.

Sullivan, D., 1994. Measuring the degree of internationalization of a firm. Journal of International Business Studies 25, 325-342.

West, A., 2017. Multinational tax avoidance: virtue ethics and the role of accountants. Journal of Business Ethics, 1-14.

White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct tests for heteroskedasticity. Econometrica 48, 817-838.

Woolridge, J., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, Massachusetts.

(35)

35

8. Appendix A

Appendix A contains the regression results of the additional analysis, and the robustness analysis. Table 8 and 9 show the descriptive statistics of the subsamples created based on the foreign asset ratio, and Table 12 and 13 show the descriptive statistics of the UK and non-UK subsamples. Table 10 and 11 show the regression results of the subsample with a foreign asset ratio below or equal to 0.7, and with the regression results of the subsample with a foreign asset ratio above the 0.7, respectively. Lastly, Table 14 and 15 show the regression results of the UK and non-UK subsamples.

A.1. Descriptive statistics foreign asset ratio sampling

Table 8: Descriptive statistics sample FAR<=0.7

Table 9: Descriptive statistics sample FAR>0.7

ETR 537 0.278 0.157 0.266 0.000 1.000 FAR 537 0.792 0.050 0.799 0.700 0.843 OBM 470 0.536 0.241 0.455 0.000 0.889 LEV 537 0.573 0.169 0.556 0.261 0.893 MTB 402 2.840 2.028 2.185 0.600 8.340 CAP 537 0.285 0.223 0.242 0.021 0.686 SIZ 537 15.277 1.467 15.203 11.419 19.176 ROA 532 0.073 0.053 0.065 -0.016 0.202

(36)

36 A.2. Regression results sample FAR<=0.7

Table 10: OLS regression results sample FAR<=0.7

FAR -0.040*** -0.028*** -0.032** 0.017 (-2.87) (-1.79) (-1.94) (0.50) OBM 0.011 0.041 (0.64) (1.62) OBM*FAR -0.107* (-1.70) LEV -0.026 -0.029 -0.029 (-1.10) (-1.20) (-1.20) MTB -0.006*** -0.007*** -0.007*** (-3.01) (-3.30) (-3.38) CAP -0.014 -0.015 -0.015 (-0.71) (-0.71) (-0.75) SIZ 0.002 0.001 0.000 (0.68) (0.22) (0.14) ROA 0.569*** 0.557*** 0.552*** (6.18) (6.20) (6.16) Constant -0.254*** -0.280*** -0.271*** -0.281*** (-52.69) (-5.40) (-4.52) (-4.59)

N 3,569 2,482 2,174 2,174

Adjusted-R2 0.004 0.130 0.138 0.139

F-statistic 8.22*** 15.56*** 19.73*** 16.66***

(37)

37 A.3. Regression results sample FAR>0.7

Table 11: OLS regression results sample FAR>0.7

FAR 0.008 0.026 0.160 0.516 (0.04) (0.16) (0.91) (1.14) OBM -0.018 0.485 (-0.47) (0.84) OBM*FAR -0.647 (-0.86) LEV 0.228*** 0.248*** 0.252*** (2.82) (2.66) (2.69) MTB -0.020*** -0.021*** -0.021*** (-3.02) (-3.09) (-3.15) CAP -0.136* -0.109 -0.108 (-1.85) (-1.60) (-1.60) SIZ -0.005 -0.005 -0.005 (-0.68) (-0.50) (-0.53) ROA 1.155*** 1.159*** 1.170*** (4.30) (4.09) (4.06) Constant -0.284** -0.471** -0.596*** -0.875** (-2.07) (-2.50) (-2.80) (-2.16)

N 537 399 355 355

Adjusted-R2 -0.002 0.225 0.219 0.218

F-statistic 0.00 3.81*** 3.60*** 3.81***

(38)

38 A.4. Descriptive statistics country sampling

Table 12: Descriptive statistics UK sample

Table 13: Descriptive statistics non-UK sample

ETR 3,335 0.272 0.139 0.264 0.000 1.000 FAR 2,626 0.318 0.272 0.279 0.000 0.843 OBM 2,760 0.435 0.285 0.455 0.000 0.889 LEV 3,334 0.585 0.164 0.589 0.261 0.893 MTB 2,397 2.555 1.829 2.010 0.600 8.345 CAP 3,331 0.247 0.181 0.208 0.021 0.686 SIZ 3,334 15.329 1.534 15.234 8.896 19.862 ROA 3,295 0.067 0.051 0.057 -0.016 0.202

(39)

39 A.5. Regression results UK sample

Table 14: OLS regression results UK sample

FAR -0.060*** -0.098*** -0.103*** -0.100 (-3.52) (-4.57) (-4.67) (-1.38) OBM -0.034 -0.032 (-0.86) (-0.65) OBM*FAR -0.005 (-0.04) LEV 0.026 0.026 0.026 (0.78) (0.77) (0.77) MTB -0.007** -0.007** -0.007** (-2.50) (-2.48) (-2.48) CAP -0.009 -0.018 -0.018 (-0.34) (-0.72) (-0.73) SIZ 0.004 0.004 0.004 (1.00) (0.92) (1.184) ROA 0.416*** 0.439*** 0.439*** (3.82) (3.84) (3.81) Constant -0.236*** -0.357*** -0.344*** -0.345*** (-38.28) (-5.48) (-4.95) (-4.97)***

Country dummies No No No No

N 1,480 1,029 954 954

Adjusted-R2 0.016 0.099 0.114 0.113

F-statistic 12.36*** 5.94*** 6.11*** 5.97***

(40)

40 A.6. Regression results non-UK sample

Table 15: OLS regression results non-UK sample

FAR -0.015 -0.008 -0.009 0.023 (-0.94) (-0.47) (-0.45) (0.70) OBM 0.005 0.030 (0.28) (1.01) OBM*FAR -0.069 (-1.12) LEV -0.039 -0.035 -0.033 (-1.12) (-0.92) (-0.87) MTB -0.010*** -0.010*** -0.010*** (-2.84) (-3.04) (-3.09) CAP -0.075** -0.070* -0.070* (-2.17) (-1.94) (-1.93) SIZ 0.003 0.003 0.002 (0.71) (0.57) (0.42) ROA 0.786*** 0.786*** 0.785*** (5.71) (5.88) (5.89) Constant -0.267*** -0.300*** -0.302*** -0.304*** (-41.66) (-4.34) (-3.43) (-3.42)

N 2,626 1,852 1,575 1,575

Adjusted-R2 0.000 0.172 0.174 0.175

F-statistic 0.89 19.39*** 22.56*** 22.46***