The Effect of Implicit Taxes on Multinational and
Domestic Companies
MSc Thesis by
Johanna Ziesing
Department of Economics and Business
Rijksuniversiteit Groningen
Academic Supervisors: Heijn Vrolijk
Dr. H. Gonenc
Information
Author Johanna `Ziesing
Abstract
This study examines the differences in the effective tax rate for multinational enterprises and domestic companies in the 28 European countries. Previous literature suggests, that the effective tax rate for domestic companies is higher than for multinational companies, due to the fact that multinational companies are able to take advantage of tax avoiding strategies. In contradiction to those studies, my study finds that the effective tax rate is actually lower for domestic companies. I examine this phenomenon and introduce in this context the concept of implicit taxes. Although the effective tax rate will not change when considering implicit taxes, the tax burden of a firm would. I show, that
List of Figures and Tables
Figure 1 Implicit Tax Principle
Figure 2 Comparison ETR1 of MNEs and DCs 31
Figure 3 Comparison ETR2 of MNEs and DCs 31
Table 1 Previous Research 15
Table 2 Conceptual Model 18
Table 3 Selection Process 20
Table 4 Explanatory Variables 25
Table 5 Durbin Watson Statistics 27
Table 6 Correlation Matrix 29
Table 7 Descriptive Statistics 30
Table 8 High and Low Tax Group Differences 32
Table 9 Correlation Analysis 34
Table 10 ETR1 output 36
Table 11 ETR2 output 28
Table 12 Expectations and Findings 39
Appendix
Table 1 VIF Results 47
Table 2 Fixed Effects Model ETR1 48
1. Introduction
The advantages of Multinational Enterprises (MNE) concerning tax avoidance has been a widely discussed topic in the literature.
When looking at nowadays development of MNEs, many companies have been getting a lot of attention because of their tax avoiding behaviour by moving their headquarters or operations abroad. Cases like Google, Starbucks and Pfilzer have been discussed widely throughout the media.
Special attention goes as well to the disclosure of several clients of a Panamanian law firm, which was dealing with clients’ offshore accounts, amongst them wealthy well-known people and high acting politicians and their close acquaintances.
With shining light on the legal but hidden avoidance of taxes, by using offshore companies, the question of whether tax avoidance is profitable must be risen. When looking at the efforts made by MNEs in order to avoid taxes, the answer seems clear.
However, in a recent study published by Dyreng, Hanlon, Maydew and Thornock (2014), the authors conclude, that the tax burden for MNEs is actually higher than for their domestic counterparts. The authors used the effective tax rate (ETR) in order to compare the tax burden of MNEs and domestic companies (DC).
This paper will analyse the effective tax rate of MNEs and DCs in Europe. In order to do so, the concept of implicit taxes will be included into the research of the effective tax rate, as it is called for by Sheckelford and Shevlin (2001). The aim of this study is for one to shed light on the tax advantages MNEs have compared to DCs in Europe. Secondly to prove the existence of implicit taxes in Europe. And finally to interpret whether the ETR is a valid measurement for tax burden. The results of this analysis can give new insights into tax burden measurement as well as the question of tax advantages for MNEs.
My thesis additionally contributes to the existing empirical literature by using most recent data available for a set of countries that have not been
countries yet. Of equal importance I add to the research in the field of tax policy as the tax advantage of multinational companies is a crucial problem in the tax laws of each country, and in how far the effective tax rate is sensitive to certain economic determinants in the country set. Furthermore, I am posing the question whether it is of value to internationalize in order to avoid taxes. In the following I will introduce the basic concepts used in the thesis as well as previous research on the topic. The third part will explain the methodology, with the sample selection and the important variables to be used. The fourth chapter will present the results as well as the discussion of the results. Chapter five will conclude the thesis and give recommendations for future research as well as limitations.
2. Literature review
This chapter will give an overview of previous research on the ETR and the different determinants as well as implicit tax theory. First, I will explain basic concepts and afterwards the important relationships between the basic concepts will be introduced and are used in order to develop my hypotheses. 2.1 Basic Concepts 2.1.1 Multinational Enterprises and Domestic CompaniesWhen examining tax avoidance, the concept of Multinational Enterprises (MNE) and Domestic Companies (DC), is a crucial part of the analysis. As described in the introduction, MNEs have tax advantages due to the possibility of avoiding national corporate tax.
paid on subsidiaries as part of the income, is based on the taxation of
subsidiaries in the subsidiaries’ country of residence. Taxation policies across especially the EU have often agreements amongst the countries, thus the actual amount of subsidiaries, and the amount of subsidiaries which pay foreign tax, might differ tremendously. Dyreng et al. (2014) are making the distinction of MNEs and DCs by classifying the firm samples as multinational if either its pre-tax foreign income is greater than zero or if its absolute value of foreign tax expense is greater than zero. However, the availability of those data is rather limited for the European region, thus I chose to follow a combination of Kim et Al. (1986) and use as a distinction the presence of foreign subsidiaries.
2.1.2 Effective Tax Rate as a measure for tax planning activities
The research in the field of effective tax rate has been growing in the last years.
The most important point is, that the effective tax rate is not a term which is interchangeable with the official tax rate. Rather it is the actual tax rate each company pays.
The ETR has been used widely as a tax burden measurement, as it represents a reliable indicator for unequal or unfair distribution of taxes on average. However, research has proven the ETR to not always be reliable as an indicator for tax burden, as it is sensitive to the firm’s capital structure. (Wilkie, 1988)
as Worldwide total income tax expense over worldwide total pre-tax
accounting income) is the most fitting. According to the authors the three most used ETR methods are, the current ETR, the GAAP ETR and the Cash Flow ETR.
Markle and Shackelford (2013), are using in their study of the effective tax rates the ratio of the total worldwide income tax expense divided by net income before income taxes for a firm in a certain year. Rego (2003) defines
ETRs as the ratio of income taxes currently payable to pre-tax accounting income. The author argues that, MNEs often use international operations in order to avoid income taxes. The ETR reflects this. When taking the definition of the ETR with the denominator defined as pre-tax income, it stays constant, when income shifting takes place. However, if the nominator is defined as income tax currently payable, it will decrease and with this the whole ratio. Thus, MNEs which avoid taxes, will have lower ETRs and with this it is a good measure to use for effective tax planning. Lazar, (2014) uses in his study about the sensitivity of the ETR two different measures in order to prevent biases in his research. One, he calculated the Cash Flow ETR, however using the combined Cash Flow of the company, and two he uses an ETR
calculation, which has the earnings before taxes and interest as a
denominator. He chose those two particular calculations, because those calculations are less sensitive to finance and investment decisions.
2.1.3 Implicit Taxes
a lower return, even though they have the same risk as other bonds. The question that arises from such a scenario is whether the investor deems it worthy to invest into an asset with lower return which is tax exempt. This decision is a simple question of comparison, if the tax to be paid by the investor is higher than the amount that is lost by the lower return asset, the investor would opt for the tax exempted asset. However, if this is not the case, he would rather prefer the taxed asset. However, this is only one case of implicit taxation. Another example would be the decreased output prices which are a result of increased competition due to governmental policies that aim at tax exemption within certain industries. When considering MNEs, many countries that offer a low corporate tax rate would compensate this loss by charging more money for input such as land or property. While, the explicit tax for the firm would be lower than in its home country, the actual expenses due to higher prices in input could offset the tax advantage.
Shackelford and Shevlin (2001) have investigated the necessity of implicit taxes in order to determine a firm’s tax burden. They argue, that implicit taxes are pervasive due to their reoccurring nature in different fields and with that have to be encountered into tax burden calculations. Figure 1 shows a simplified version of implicit taxes, which was explained by Schackelford and Shevlin (2001):
Figure 1 - Implicit Tax Principle (Schackelford and Shevlin, 2001)
As can be seen in the end Investor 1 and Investor 2 receive the same after-tax income. Their theory states, that the reason for the lower pre-after-tax return, is due to the tax exemption of the asset. Thus, especially high taxed investors are rather prone to invest into those assets. Although the actual outcome is the same (both receive their 7$) their tax burden is a different one.
explicit tax rate for Investor 1, whereas there is a 30% implicit tax rate for Investor 2 (the pre-tax rate of return for investor 2 is 30% less from the beginning). This rather simplified example only stands ground, when the actual after tax income is equal. However, that is not always the case. The actual decision of the investor for which asset to invest in, depends on his own marginal tax rate.
Furthermore, the concept that capital is attracted towards corporate activity that is taxed at lower rates has to be supported. Only if this is the case, competition would rise which leads to higher input costs and lower output costs for companies. When this scenario holds true, implicit taxes arise. Competition can in this case increase out of many reasons, however implicit taxes are usually said to be related to governmental policies which aim at compensating for tax exemption. Then, implicit taxes arise due to the positive or negative tax preferences. (Scholes and Wolfson, 1990).
Additionally, by introducing tax exemption for one region or industry, and the increase of competition resulting, gives a competitive disadvantage to the companies already present, which can be seen as a form of implicit tax disadvantage as well. The most important notion is that, implicit taxes are not paid directly to the government, rather are they paid to the beneficiaries of the tax preferences, as the decrease in explicit tax, would benefit them. According to Chyz, Luna and Smith (2014) and Scholes and Wolfson (1990), there is an offsetting effect which should balance an explicit tax advantage (thus, a lower tax rate) and an implicit tax disadvantage (a lower pre-tax return). Which will additionally be the basis of this research concerning implicit taxes.
2.2 Previous Research
In the following previous research considering the relation of
internationalization and tax avoidance is discussed. The relationship is important to mention, as it shows that tax avoidance can be a reason for internationalization.
2.2.1 Relationship between Tax avoidance and Internationalization
globalization and the resulting as well as developing technological possibilities, facilitate tax avoidance and with this increases the number of firms engaging in tax-planning behaviour. (Zucman, 2014, Hines 1999, Clausing, 2009, Grubert 2012)
There is a general perception in the current academic literature, that MNEs have an advantage due to the fact, that they are able to decrease their tax base by tax avoidance strategies. According to current research the growth of intangible assets, which are easier to move across boarders, increases the possibilities for MNEs to avoid taxes, as it facilitates increased income shifting (Gravelle, 2013). They furthermore are able to shift their income from high-tax to low-tax countries, including tax havens for lower costs. Klassen and Laplante, 2012; Dharmapala and Riedel, 2013 and Klassen, Lisowsky and Mescall 2013.
They have several unique tax reduction possibilities, which are not available to DCs in the same magnitude. The decision to internationalize depends amongst others on the tax rate in the chosen country. Once, the decision for a country is made, MNEs are able to shift economic activity, employees and intangibles assets into the low-tax country relatively easier than DCs could (Clausing, 2009).
construct of offshore companies. He points out that most nations do not work sufficiently on preventing this tax avoidance behavior.
2.2.2 Relationship between MNEs and DCs and their ETR and implicit taxes
The relationship between the ETR and implicit taxes concerning MNEs and DCs, only recently gained academic attention. Several authors discussed the difference of ETRs for MNEs and DCs in the last years, mostly coming to the same conclusions: MNEs have a lower ETR than DCs due to extensive tax planning possibilities. (Rego, 2003; Liu and Cao, 2007; Jennings et Al., 2012; Lazar, 2014) Rego (2003) examines the difference of ETRs of U.S. MNEs and DCs concluding, that MNEs have a lower ETR and higher pre-tax income, supporting the notion, that MNEs have more incentives to involve themselves into tax planning activities. However, he as well determines in his findings that a higher level of U.S. pre-tax income leads to lower U.S. and foreign ETRs, whereas a higher foreign pre-tax income leads to the opposite effect. He furthermore concludes that MNEs, which conduct more aggressive tax planning, have lower ETRs.
Klassen and Laplante (2012), support this argument in their findings. The authors find in their study about profit and income shifting, proof for aggressive income shifting by MNEs in order to lower their ETR, thus tax base.
However, Markle and Shackelford (2009), find in their study of MNEs and DCs, that those two entities face similar ETRs. Dyreng et Al. (2014), are analysing in their recent paper the changes in corporate tax rates over the last 25 years. They find in their data set of US MNEs and DCs, that the effective tax rate of both categories is decreasing over the period of time. This indicates new findings, which are in contradiction with earlier results, concluding MNEs do have the possibility of international tax avoidance but the advantage might not be as high as expected. He argues that the results are based on new possibilities of US DCs. Amongst other options US DCs have as well potential alternatives such as profit shifting and substitutions.
Jennings et. Al. (2012), examine in their study the ETR differences but
by the shareholders of the corporations. The ETR used in order to compare the offsetting effect of implicit taxes does not decline on average over the years. Theory states that, if implicit taxes are higher for domestic firms, there is a higher positive correlation between ETR and pretax return for DCs than for MNEs. They argue that implicit taxes offset the tax benefits of firms with a lower ETR to a greater extent in domestic firms.
Chyz et Al. (2014) argue for including implicit taxes, when analyzing ETR differences. Their findings indicate that implicit taxes are lower for U.S. MNEs than DCs. When considering a policy change, which aims at lower tax
burdens in order to increase domestic economic activity, might lead to the opposite effect. They state that more competition indicates a lower pre-tax return, which then leads to higher implicit taxes. Hence, even if the effective tax rates are lower for MNEs, the implicit tax rates are higher, which is incentivizing MNEs to engage in more foreign activity, as the after tax rate of return is higher there, due to lower implicit taxes. Additionally, the decrease in tax base leads to higher competition in the tax-exempted industry, which gives MNEs a relatively higher competitive advantage to their domestic
counterparts. Furthermore, it has to be considered that, most of the explicit tax advantage due to the lower tax burden, is enjoyed by the shareholders in the company rather than lead to a positive policy outcome. With this study they give a counter argument to the study of Dyreng et al. (2014), and concluding that the ETR of MNEs and DCs, when considering implicit and explicit taxes, is not similar neither higher for MNEs in the case of US companies.
Internationalization has thus several implications, one the one hand it furthers the tax avoidance behavior of companies, but it as well due to wrong policy implementation can lead to a competitive disadvantage for DCs.
2.2.3 Relationship ETR and other economic determinants
The most common determines that were researched are firm size, capital intensity, leverage, profitability as well as variations of asset mixes and foreign operations. Especially firm size was of interest for many authors, as it is
surrounded by contradictory theories. Political power theory describes that larger firms are rather making use of their power and can influence provisions and are more aggressive in tax planning. (Becker, 1983) However on the other side political cost theory suggests, that larger companies tend to have more attention on them, which leads them to less aggressive tax planning due to increasing scrutiny by the public. (Zimmerman, 1983)
Concerning the profitability, authors generally agree on, that a higher
profitability leads to a higher ETR (Gupta and Newberry, 1997; Jennings et Al. 2012; Lazar, 2014). Which has the logic behind it that, as more profitable the firm is as higher its pre-tax income is. (Wilkie, 1988) However, recent
research has proven, that higher profitability leads to a lower ETR and vice versa, as more profitable companies have better tax planning opportunities. (Zinn and Spengel 2012)
Concerning the profitability differences between MNEs and DCs Chen,
Cheng, He and Kim (1997), have been analysing the capital structure of firms by debt ratio, in order to examine the influence of the extent of
internationalization on the capital structure of a firm. The study analyses amongst other factors the ROA of MNEs and DCs, coming to the conclusion that MNEs are more profitable than DCs.
Hsu, Chiang and Liao (2013), examine the behaviour of MNEs and DCs on the base of the pecking order theory. They conclude amongst other things, that MNEs have greater profits, when they have a significant degree of internationalization. Following this trait of argument, Debaere, Lee and Lee (2006) come to the same conclusion, however making an important distinction. They argue that, MNEs that move to a more advanced country, are stronger, more efficient and more profitable, that their national counterparts. However, at the same time MNEs that go to less developed countries are similar to their national counterparts in terms of efficiency, capital and profit.
advantage of their tax shield, thus they have lower ETRs. (Stickney and McGee, 1982; Gupta and Newberry, 1997)
In relation to the difference of MNEs and DCs Chyz et Al. (2014), find that leverage has greater effect on ETRs for MNEs than for DCs, which indicates that MNEs have better opportunities to reduce their tax burdens.
Table 1 shows an overview of the most important research that has been done in this field up until now. As it can be seen there have been mixed results concerning the difference of ETRs of MNEs and DCs.
Table 1 Previous Research
Study Number of firm year observations Period Dependent Variable: ETR Independent Variable:
Main Tax Variable
Finding Jennings et. Al (2012) 75,000 Panel data: 1976-1985 and 1988-2005
ETR (total tax expense/PTI1)
PTR2, ROE3 +
+
Rego (2003) 19,737 Pooled, Cross-sectional data: 1990-1997
ETR(worldwide current income tax expense/total PTI) PTI - Chyz et Al. (2014) 40,038 Panel data: 1989 - 2012
ETR (Total tax expense /pretax income before special items) PTR, ROE + + Dyreng et al (2014) 54,005 Panel data: 1988-2012
ETR (current tax expense/PTI) TIME4 () and FORSTAT5 () Decrease in ETR for all Markle and Shakelford (2009)
10,642 Panel data and Cross-country: 1988-2007 ETR (worldwide income tax expense/net income before income taxes) ETRmne (current income tax expense/net income before income taxes)
ETRdom(current income tax expense)
Decrease in ETR
for all
My study 7768 Panel Data:
2011-2015
ETR (Total tax expense/PTI)
ETR (Total tax expense/Cash Flow)
*Notes:
1 Pre-tax income
2 Pre-tax return: Pre-tax income/owner’s equity 3 Return on Equity: Net income/Equity
4 Fiscal year for a given firm-year observation less the number of the first year in the dataset
5 Average of the foreign statutory tax rates for the countries in which the firm has significant subsidiaries
My study will investigate the difference of the effective tax rate for MNEs and DCs in the EU. To measure the difference of the ETR, I will partially follow Chyz et Al. (2014) and Jennings et Al. (2012). I will make use of two different approaches for the calculation of the ETR. My first approach follows Lazar (2014), which entails the use of the Cash ETR. My second approach will make use of the GAAP ETR as proposed by Roemgens and Steinweg (2015). I opt for two different ETR calculations, in order to prevent any dissimilation. My study uses different methods in order to determine the difference of the ETR. First, the ETR of MNEs and DCs will be compared over the time span of 5 years, using both methods. Afterwards, I will compare high and low tax preference groups. Thus, first I divide DC and MNE samples into high and low tax preference groups and compare their differences in implicit and explicit tax rates. This step is done by analyzing the pre-tax return of a company (for implicit taxes) and its tax expense by owner’s equity (for explicit taxes). In the next step I will analyze the correlation of ETRs and the pretax return and after-tax return for the MNE and DC samples. The last step will be a regression analysis of the ETR and certain economic determinants. This will deepen the understanding of the differences in ETR of DCs and MNEs. The use of the independent and dependent variables will be further explained in Chapter 3.
which suggests higher implicit taxes for DCs. Following the literature in Chapter 2.2.2, I expect to find higher implicit taxes for DCs than for MNEs. Additionally, I expect to find a bigger offsetting effect by implicit tax disadvantage to explicit tax advantage for DCs. This goes align with the assumption that, MNEs have more opportunities to prevent certain taxation. I furthermore, expect to find a higher ETR for MNEs than for DCs, in the simple comparison between the two. Due to the fact, that MNEs have less implicit taxes, their tax base might be higher, but their pre-tax return will be as well. Depending on the correlation of the pre-tax return and ETR, it is possible to determine whether there are implicit taxes. Concerning the leverage of a firm, research indicates a positive relationship between Leverage and ETR (Wilkie, 1988; Gupta and Newberry, 1997). However, ETR 2 already deducted the tax expense, thus the second outcome is ambiguous.
According to previous research (Gupta and Newberry 1997; Janssen 2005; Janssen and Buijink 2000; Richardson and Lanis 2007; Lazar, 2014) the ROA is used in order to examine whether the profitability of a company influences its ETR, it is stated that the ROA does increase the ETR. However, according to new research it can as well be used as a proxy for tax-planning activities, which would lead to a decrease of the ETR. (Zinn and Spengel 2012). Hence, there is no sign predictable for the ROA as of yet.
contains of 28 countries, which have different societal, governmental and financial structures. Which makes it impossible to estimate whether one of the two theories is applicable. For example, Lazar (2014) examines the economic determinants for Romania, although the correlation for Size is positive, he argues that, Romania’s limited amount of publicly listed non-financial
companies is rather unlikely to cause political interest on either side. Thus, the predicted sign for Size will stay unknown.
I contribute to the existing empirical literature in four important ways. First, I am using a data set consisting of European MNEs and DCs, up until now mostly US companies have been analysed or one specific country. Second, I am using the most recent data available. Third, I am conducting several analyses, using different measurements for the ETR, usually authors are using either one ETR calculation or only regress the determinants of it. While authors have analysed the difference of the ETR between MNEs and DCs, including correlation and regression analysis, the new data set consisting of European countries and the fact that I am using two calculations for the ETR, is to my knowledge, the first study in this field. Fourth, I am contributing to the tax research by analysing the differences between MNEs and DCs before and after tax profit, depending on their ETR. Which can lead to important insights concerning the worthiness of internationalization.
In summary, the aim of the following empirical study is for one, to examine the difference of the ETR between MNEs and DCs. Secondly, show which
determinants are more relevant for which kind of firm. And thirdly, deepening the understanding of the difference of profitability for MNEs and DCs
depending on their ETR.
Table 2 gives and overview of the effects for MNEs and DCs.
Table 2 Conceptual Model
Variable Effect Explanation ETR1 ETR2
PTR Positive Implicit taxes * *
Negative No implicit taxes
ROA Positive Less tax planning activities ? ?
Negative More Tax planning activities ? ?
LEV Positive Firms make use of tax shield *
SIZE Positive Bigger firms do not have the opportunity to engage in more tax planning activities
? ?
Negative Bigger firms have the opportunity to engage in more tax planning activities
? ?
ROE Positive
Negative * *
2.6 Hypotheses
The ETR literature clearly shows contradictions about the question whether MNEs have a lower ETR and with that a tax advantage. While Dyreng et al (2014) made, considering the previous finding in the literature, a rather astonishing observation, Chyz et Al. (2015) opposed those results with their extension. However, both studies have been taken place with data of US firms. Thus, it is of interest to see whether the findings of those studies hold up to European data as well. Furthermore, the concept of implicit taxes is not a widely researched one yet. In order to conduct the research on the basis of this, the two following hypotheses have been developed from the literature review. In step one, the data will be analysed following the approach of Dyreng et Al. (2014), according to the the literature review, MNEs have a lower ETR than DCs, leading to the following hypothesis:
H1: MNEs have a similar or lower ETR than DCs
In the second step, the analysis will follow Chyz et Al. (2015) including the concept of distinguishing between implicit and explicit taxes, stating the following hypotheses:
H2: MNEs have a similar or higher explicit tax advantage than DCs
This hypothesis stems from previous research, stating that MNEs have better tax planning opportunities leading to smaller taxation. On the other side DCs have a smaller tax base (Rego, 2003), which would suggest a higher implicit tax advantage than MNEs have, leading to Hypothesis 3:
In the following step, a regression analysis, will show how the ETR behaves, considering certain economic determinants. As mentioned in Chapter 2.X, firm leverage and ETR are negatively correlated which would lead to the
conclusion, that firms with higher leverage can take more advantage of their tax shield, as interest expenses are tax deductible, which leads to Hypotheses 4:
H4: MNEs are able to make more use of their tax shields
Afterwards, the question of profitability is included in the analysis. As
Dharmapala and Riedel (2013) argue, the profit shifting increased significantly but the question is whether this actually profits the MNEs in the long run. It could be argued in this context, that due to the fact, that MNEs have more tax planning opportunities, their profit is relatively higher than for DCs and with that their ETR. Thus, hypotheses five is:
H5: MNEs have a higher profit due to a higher ETR.
In the following chapters the hypotheses will be answered with the help of analyses.
3. Methodology and Data
This chapter will cover the sample selection as well as the primary variables and the data analysis. I begin with the collection of the country data and the selection process. The second section describes the different statistical methods used. Which is followed by the description of the necessary main variables. Section 4 will give the overview and present the descriptive statistics and the regression equation.
3.1 Sample
As the previous research covered primarily US companies, my thesis will test whether those studies can be applied to European companies as well,
however do I add the comparison of two different ETR calculation methods. The data set consists of listed MNEs and DCs from EU countries.
Concerning the time period, I will analyze the data from 2011-2015, which is including thus more recent data, than the previous studies.
Table 3 Selection Process
Steps Selection Process
1. Region EU (EU-28)
2. Foreign Subsidiary No Foreign subsidiary mean DC, Foreign Subsidiary means MNE 3. Data availability DV Data used for calculation of ETR1 and ETR2
4. Data availability IV Data regarding the 5 IVs
5. Years 2011-2015
6. Last available year 2015
The first criterion cuts down the number of companies to a total of 65,887,05. When applying the second criteria, which is needed in order to distinguish between an MNE and an DC as discussed in chapter 1.1, the number got cut down to 52,948 for MNEs and 669,515 for DCs.
I only chose countries with the complete data available, in order to prevent biased outcomes due to missing data. The time criterion included first 10 years, however the full data availability was only possible for a shorter time period, thus I cut it down to 5 years. After further cut downs due to the data availability for the calculation of the independent and dependent variables, the final sample included in total 1729 firms (MNEs and DCs).
3.2 Statistical techniques
The model is making use of the ordinary least squared (OLS) method in order to estimate the correlation. This method requires several assumptions which have to be tested for and appropriately dealt with: Homoscedacity,
Endogeneity, Multicollinearity and Normality. In the following I will introduce my main variables, the regression equation and the appropriate tests for the data set at hand.
3.3 Variables
observations (after missing value and outlier reduction), in order to obtain a robust result of the regression analysis.
3.3.1 Dependent Variable
Effective Tax Rate
My dependent variable is calculated according to two different methods. As introduced in chapter 1.2, there are several different approaches for the calculation of the ETR. I chose to use the GAAP ETR (ETR1) as well as the Cash Flow ETR (ETR2).
The distinction is important in order to control for eventual bias. The GAAP ETR is calculated as the ratio of total tax expense (TX) to pre-tax income (PTI):
𝐸𝑇𝑅$ = 𝑇𝑋 𝑃𝑇𝐼
and the Cash Flow ETR is calculated as the ratio of total tax expense (TX) and cash flow (CF):
𝐸𝑇𝑅) = 𝑇𝑋 𝐶𝐹
For the Cash Flow ETR I am using the Cash Flow function in Orbis, which is not limited to only operational Cash Flow, as due to the structure of the data, a comprehensible function of the Operating Cash Flow could not be
constructed. ETRs will be used in separate regressions, in order to check robustness of the sample.
3.2.2 Independent Variables
Pre-tax return
In order to analyse the presence of implicit taxes, the pre-tax return variable is introduced. Following Jennings et Al. (2002), the pre-tax return is used to determine whether the explicit tax advantage is offset by an implicit tax disadvantage. As explained in chapter 2.1.3, the implicit taxes, which occur due to capital being attracted to tax exempted assets, can offset the advantage companies have due to tax exemption. This can lead to a less strong explanatory value, when comparing effective tax rates. The pre-tax return (PTR) is calculated by dividing the pre-tax income (PTI) over the beginning of year owners’ equity (total assets of the year before – total liabilities of the year before):
𝑃𝑇𝑅 = 𝑃𝑇𝐼 𝐸𝑞𝑢𝑖𝑡𝑦
Furthermore, I am using an interaction term in the regression analysis of PTR and a dummy variable for MNE. This term, can indicate whether the implicit tax advantage is higher for MNEs or DCs.
Return on Equity
The Return on Equity (ROE) to ETR correlation will be used in order to examine whether there are implicit taxes present. The bases of implicit tax theory is, that implicit taxes and explicit taxes have a relative relation, which means that the after tax return, thus the ROE is the same whether implicit taxes occur or not. The ROE is calculated as the ratio of net income (NI) to beginning-of-period owners’ equity (OE):
𝑅𝑂𝐸 = 𝑁𝐼 𝑃𝑇𝐼
Return on assets
The Return on Assets is used in order to measure the profitability of companies (Lazar, 2014; Richardson and Lanis, 2007) it is calculated as the ratio of pre-tax income to total assets:
𝑅𝑂𝐴 = 𝑃𝐼 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠
research MNEs are more profitable, thus I expect a positive relationship for the interaction term.
Leverage
The firm leverage is used as a variable in order to examine whether MNEs or DCs can take more advantage of their tax shield, generally leverage affects a firms effective tax rate through interest expenses (Lazar, 2014), which are tax deductible. A firm’s leverage is calculated as the ratio of long-term debt to total assets.
MNE
I introduce the dummy variable MNE in order to examine the difference between MNEs and DCs towards the ETR sensitivity. I set the variable to 1 when the firm has a foreign subsidiary and to 0 if not. Although this approach has the disadvantage that small countries have a higher percentage of MNEs than bigger countries, the fact that 28 countries of different sizes equalized this problem. Common wisdom suggests that MNEs have more possibilities for tax planning activities, which suggests a positive relationship. However, many EU countries have a low tax rate, which could lead to some low tax countries absorbing taxation from higher taxed affiliates, which can lead to a higher tax burden of those firms due to profits shifted to the subsidiary in the low tax country. Thus, there is no expected sign to be determined.
Explicit Tax Rate
The explicit tax rate is used in order to compare the offsetting effect of the implicit tax advantage to the explicit tax rate disadvantage and vice versa. This approach follows Jennings et Al. (2012) and Chyz et Al. (2015). The Explicit Tax Rate is calculated as the ratio of tax expense to beginning of the year owner’s equity:
𝑇𝐴𝑋 = 𝑇𝑋 𝐸𝑞𝑢𝑖𝑡𝑦
3.2.3 Control Variables
The model will include as well a time variable using the 5 years as dummy variables and Firm Size, calculated as the natural log of total assets, this calculation is the common one found in previous literature (Lazar, 2014; Gupta and Newberry, 1997; Richardson and Lanis, 2007).
To control for firm size is particularly important, as the profitability of a company depends on it. (Rego, 2003) The year dummy variable is introduced in order to control for unexpected change of the year due to the introduction of for example new tax regulations, as the EU has not yet introduced a common tax system. Table 4 gives an overview of all the variables used as well as their expected signs.
Table 4 Explanatory Variables
Explanatory variables Expected signs
ETR1 ETR2
PTR Pre-tax income over the beginning of year owners’ equity
+ +
Leverage Ratio of long-term debt to total assets - ?
Return on Assets Net profit over total assets + +
Size Natural logarithm of total assets ? ?
Year Dummy variable 1 for year of data 0 for any other
MNE Dummy variable 1 for MNE and 0 for otherwise ? ?
3.3 Statistical Model
Having introduced the variables, in the following step I will construct the models for the multiple regression analysis. I will run 5 tests, using variations of the regression.
The first test I am running is containing only the dummy variables. Afterwards, I will run a test without the interaction. Third, I will run a test with all variables, except for the size as the control variable. Forth, I run the regression only years as control variable and fifth, I will run the tests with all variables. I will do so with ETR1 as the dependent variable, and afterwards with ETR2:
where, i denotes the firm and t denotes the year; 𝛼= stands for the intercept and 𝑢:;is the error term.
3.4 Modelling Procedure
After collecting the data from Orbis, I proceeded to work with the data in Microsoft Excel. Due to some missing data points as well as outliers I had to omit several firm observations, ending with 8085, which then was processed in Stata 14.
In the case of PTR and TAX, I winsorized the values at the 1% and the 99% level, in order to reduce the impact of outliers. Concerning the ETR, I applied recoding treatment in order to deal with ETRs, that are of unusual magnitude. I censor the data between 0 and 1, thus I set all negative values to 0 and all values above 1 to 1, in order to be able to obtain a meaningful interpretation of the ETR. This is in line with other research in this field (Gupta and Newberry,1997; Chyz et. Al. 2014; Richardson and Lanis, 2007).
3.5 Estimation Method
The data set used is a panel data set. The advantage is, that the size of the sample can be fairly large and the variables can be regressed over time. Although I am controlling for years, there still might be influences which can not be controlled for. However, I am using several tests in order to assess whether my model is robust, the tests are described in the next section. Furthermore, the model uses ratio, interval and nominal data, however with the creation of dummy variables it is possible for me to use a linear regression model.
3.6 Evaluation of Method Assumptions
three ways: the standardised residual1, the studentised residuals2 and the DFBETA3. Using the first test we found 315 outliers, which is about 4% of the data, this falls into the range that is expected by this test. The test for studentised residuals as well as the DFBETA, lead to 353 further reductions. Which entails that 96% percent of the standardised residuals fall into the margin between -2 and +2.
By excluding the observations that were found to be outliers my R^2 of ETR1 increased from 0.044 to 0.144, and for ETR2 from 0.048 to 0.068. After the outliers were deleted, 7689 observations are left.
4. Empirical Results
The following chapter will discuss the four assumptions: Homoscedacity, Endogeneity, Multicollinearity. Afterwards I introduce the regression, followed by the conduction of robustness tests.
4.1 Homoscedasticity
The first assumption, that has to be valid is that the level of variance of regression errors is equal for all observations. If this is the case the data is homoscedastic (Field, 2009)
If this is not the case the data is heteroskedastic, which can cause problems due to the fact that the OLS model would attach more weight to the observations with rather large error variances. This can lead to biased estimates (Pindyck & Rubinfeld, 1991). I conducted the Breush-Pagan, in order to test for heteroskedasticity. For the both models the p-value is 0, which leads to not rejecting 𝐻=. This implies heteroskedasticity. I use the “robust” command on Stata for both, in order to deal with the heteroskedasticity, which is in line with Stock and Watson (2003).
1 Transformation of the data indicating the mean as zero and the standard deviation as one. At least 95% of the standardised residuals should fall between two standard deviations from the mean, hence have values between -2 and +2 (Torres-Reyna, 2007; Williams, 2016)
2 This computation takes the variability into account that the variance of the predicted value used for calculating the residuals is not constant, by dividing the observed residual by an estimate of the standard deviation of the residual at the respective point. Once again, 95% of the values should be within two standard deviations of the mean, which is zero (Torres-Reyna, 2007; Williams, 2016)
4.2 Endogeneity
If the data is endogen, one of the explanatory variables is correlated with the error term. This can lead to an overestimation of the influence of one of the independent variables on the dependent variable. Error terms are expected to be homoscedastic and uncorrelated.
In order to determine whether the data is endogen, I conducted a Durbin-Watson test, as suggested by Keller (2012). The table below shows the outcome of both tests, one time the model using ETR1 as dependent variable, and one time with ETR2.
Table 5 Durbin Watson Statistics
ETR1 ETR2
Value of Durbin-Watson 0.0261 1.4862
Range of D-Statistic 10, 7689 10, 7689
As can be seen, both values are below 2 which indicates positive autocorrelation. As both values are as well below the range of the d-statistic, it can be concluded that both data models are show autocorrelation. This is not surprising, when considering the fact that I am dealing with Panel data. In the next step I conducted the Hausman test in order to decide whether fixed or random effects rather should be used, this aligns with the suggestions by Stock and Watson (2003). The resulting p-value is 0.0172 (0.000 for ETR2), which leads to the rejection of H0, which indicates the use of the fixed effects model, which compared to the OLS model controls for omitted variables that are different in each case but constant over time. Recent research has by Williams (2015) has showed that fixed effects should be only used if omitted variables have time-invariant values with time-invariant effects. However, in both cases in my models variables have already been omitted. Further on, they are not time-invariant, as the effect of a certain year can be different over time. Consequently, I decided to continue with the OLS regression. However, a comparison of both methods can be found in the appendix 3 and 4.
4.3 Multicollinearity
between the sample coefficient and the actual population parameter. Secondly, coefficients will have small t-statistics, which can lead to misinterpretation of the linear relation between the independent and the dependent variable. (Keller, 2012).
In order to test for multicollinearity, I follow Williams (2015) and calculate the variance inflation factor (VIF). The VIF value shows the degree of inflation of the coefficient variance. This is a different approach to the usual approach which focuses on bivariate correlations, which would indicate how high the coefficient variance would be, if it were not to be related to other variables in the model. This however, only detects pairwise variances, that might be small, which does not mean there is no linear dependence between two or three variables. This is however detectable with the VIF. The generally accepted level of inflation is 10 (Wiliams 2015, Field 2010). Furthermore, the tolerance statistic will be used, which is an indicator calculating 1/VIF, the value of this indicator is considered critical if it is below 0.1. The test results show no presence of multicollinearity. The results can be found in appendix 3.
4.4 Normality
The last assumption, that has to be tested for is normality. Normality assumes that the error terms are normally distributed along the corresponding means. If this is not the case, the p-values might be biased. Normality is tested by the Skewness and Kurtosis test in Stata, which is called the Jarque-Beta test. Both p-values take the value 0, which indicates the rejection of H0. Although this indicates prevalent normality, the fact that my sample is fairly big, this can be ignored. (Brooks, 2014)
4.5 Correlation Matrix
Table 6 shows the correlation matrix for all variables. The correlation values, seem to indicate the same results I expect. Furthermore, there are no strong correlations between the independent variable, thus I can proceed with my analysis as planned.
Table 6 Correlation Matrix
ETR1 ETR2 PTR SIZE Leverage ROA MNE Y2011 Y2012 Y2013 Y2014 Y2015
PTR 0.3310 0.1904 1.000 SIZE 0.2100 0.1186 0.2048 1.000 LEV 0.0120 -0.0631 -0.0446 0.3852 1.000 ROA 0.2765 0.1953 0.7514 0.1318 -0.1215 1.000 MNE 0.1710 0.1277 0.1588 0.3853 0.0979 0.1052 1.000 Y2011 0.0073 0.0206 0.0388 -0.0123 -0.0054 0.0311 0.0008 1.000 Y2012 0.0047 -0.0032 -0.0043 -0.0022 -0.0127 0.0029 -0.0031 -0.2515 1.000 Y2013 -0.0202 -0.0095 -0.0163 0.0108 -0.0068 -0.0071 0.0006 -0.2491 -0.2477 1.000 Y2014 0.0104 0.0130 -0.0174 0.0037 0.0107 -0.0132 0.0000 -0.2524 -0.2510 -0.2486 1.000 Y2015 -0.0024 -0.0210 -0.0010 0.0001 0.0142 -0.0138 0.0017 -0.2514 -0.2499 -0.2476 -0.2509 1.000 4.6 Results of Hypothesis
In this chapter I will present the summary statistics as well as the conducted analysis. The comparison was conducted with EXCEL, the correlation and regression was conducted with Stata14.
Table 6 below shows the descriptive statistics of the variables used, including the number of observations, their mean, standard deviation as well as their minimum and maximum values.
Table 7 Descriptive Statistics
(1) (2) (3) (4) (5)
VARIABLES N mean sd min max
For the dependent variable, ETR1 has a mean of .214, and ETR2 has a mean of .153. Given that both ETR measures have the same numerator, and that operating cash flows are normally greater than book income, the mean for ETR1 is greater than ETR2, as expected.
In the next chapters I will represent the outcome of my different analyses. The outcome will answer my hypothesis, for this I will refer to the different output tables.
4.6.1 Comparison of Effective Tax Rates between MNEs and DCs
At first, I compare both ETRs between MNEs and DCs, in order to see
whether I have the same result as Dyreng et Al. (2014), concluding that MNEs have a higher ETR than DCs, although they have more tax-planning
possibilities. As can be seen in Figure 2 and Figure 3, MNEs in the EU region have in fact a higher ETR than their domestic counterparts.
Figure 2 Comparison ETR1 of MNEs and DCs
Figure 3 Comparison ETR2 of MNEs and DCs
However, this observation as conducted by Dyreng et Al. (2014), does not include the concept of implicit taxes. Thus, in the following I will conduct several tests, in order to examine whether the impact of implicit taxes, might have skewed the results above.
4.6.2 Comparison of high and low tax preference groups
In the following I divided the samples of MNEs and DCs individually into high and low tax groups. The purpose of this distinction is the comparison between the implicit and explicit tax offsetting effects. In other words, implicit tax theory suggests, that implicit tax advantages are offset by explicit tax disadvantages, thus, a high PTR comes with a high explicit tax (Jennings et al., 2012). By subdividing MNEs and DCs in high and low tax groups, the existence of implicit taxes can be proven and compared between the two kinds of corporations. The existence of implicit taxes is correlated to high and low explicit tax, due to the fact, that low taxed firms, pay less taxes but on the offside have a lower pre-tax return. Thus, by dividing the sample in high and low tax companies, it can be seen whether low taxed firms have a greater explicit tax advantage or an implicit (thus, lower PTR) tax disadvantage. By comparing these ratios in the end between MNEs and DCs, it can be seen whether the explicit tax advantage of DCs shown by a lower ETR in chapter 4.1 is reflecting the truth or whether this tax advantage is offset by an implicit tax disadvantage. Depending on which offsetting effect by an implicit tax disadvantage is higher, the result of Analysis 1, can be doubted.
I chose to set the limit between high and low tax firms, at an ETR of 25%. The results are shown in Table 8.
Table 8 High and Low Tax Group Differences
ETR1 ETR2
Domestic Multinational Domestic Multinational
High tax Firms
PTR 0.0784 0.16449 0.0491 0.1813
TAX 0.0267 0.05504 0.0275 0.0656
Low Tax Firms
PTR 0.0329 0.08746 0.0414 0.0964 TAX 0.0118 0.02072 0.0127 0.0234 Difference PTR 0.0455 0.07703 0.0077 0.0849 TAX 0.0148 0.03432 0.0148 0.0421 Ratio (PTR/TAX) 3.0617 2.24426 0.5178 2.0146
3.06 time by an implicit tax advantage. Which, indicates as well that the explicit tax advantage for low taxed MNEs is higher than for low taxed DCs. When analyzing the PTR and TAX differences for MNEs similar results are obtained. The explicit tax advantage for low taxed companies is 0.034, which indicates roughly 3 percent. However, the implicit tax advantage of high tax firms, is with a value of 0.077 almost 8 percent. When calculating the
offsetting ratio, it can be concluded that the implicit tax advantage for high tax firms is 2.24 times higher than the explicit tax advantage for low tax MNEs. Thus, two things can be concluded, one, both MNEs and DCs bear implicit taxes and two, the implicit tax advantage is for DCs significantly higher. This, will affect the conclusion whether Analysis 1 can be accepted in the latter part of the thesis.
However, when looking at the results of low and high tax groups categorized by ETR2, a different image arises. The explicit tax advantage for low taxed DCs is similar to the one before with roughly 0.015. However, the implicit tax advantage for high taxed DCs is 0.007. Which indicates an offsetting ratio of only 0.518. This leads to the conclusion that the explicit tax advantage for high tax DCs is only by half offset by an implicit tax disadvantage. When looking at the differences in PTR and TAX for MNEs the values are fairly similar to the one for the ETR1 group, resulting in an offsetting effect of 2.015. Due to the big difference between the two results for DCs, I set the
benchmark for high and low tax companies down to 20 percent in order to assure the result. However, the resulting values are significantly higher, showing an offsetting effect of 3.2. Thus, I drew the conclusion that due to a lower ETR2, the sample is unbalanced and lead to the low offsetting ratio.
4.6.3 Correlation
the correlation, the results will tell whether MNEs or DCs bear more implicit taxes.
Table 9 Correlation Analysis
MNE DC
ETR1/PTR p-value ETR1/ROE p-value ETR1/PTR p-value ETR1/ROE p-value -0.5024 0.3883 -0.57099 0.3801 0.2269 0.7136 0.0614 0.9218
ETR2/PTR p-value ETR2/ROE p-value ETR2/PTR p-value ETR2/ROE p-value
0.1531 0.8059 0.1454 0.6189 0.2657 0.8265 0.1256 0.8406
When looking at the data in table 9 for ETR1, the correlation between ETR and PTR is strongly negative but not statistically significant for MNEs, thus an increase of explicit tax leads to a decrease of pre-tax return. This indicates, that MNEs do not bear any implicit taxes. The correlation between ETR and ROE is strongly negative, which demonstrates that MNEs can retain some of their low tax rate advantage in their after-tax returns. In other words, when the ETR is low the ROE is increasing, which leads to the assumption that there are no implicit taxes, as otherwise the correlation would be either positive or weakly negative. On the other side, the value for DCs indicates a strong positive correlation, which represents the existence of implicit taxes and an offsetting effect of implicit taxes for DCs with a higher tax burden (ETR). At the same time there is a weak correlation of ETR and ROE for DCs, which supports the assumption that there is an offsetting effect of implicit taxes, as firms are supposed to have the same return on equity. As
mentioned in chapter 2.1.3 the after tax return should be the same whether implicit taxes exist or not. Looking at the data for ETR2, the assumptions concerning DCs is strongly supported. However, the values of MNEs for both correlations are contradictory. Although they are not statistically significant, both correlation coefficients indicate a positive (but not strong) correlation. Which indicates that MNEs bear some implicit taxes.
4.6.4 Regression
The Regression analysis will be used in order to discuss the last hypotheses. I will refer to the tables throughout the discussion.
Model 1 is my baseline model. It includes the control variables and shows that there is no significant difference among the years, that I control for. Model 2 includes the different coefficients, but excludes the control variables in order to see the actual impact on the dependent variable and the influence of the independent variables on it. Model 3 includes the first control variable Size. Due to the emphasis in literature on the strong influence of Size on the ETR, I decided to first only control for Size and then in Model 4 only for Year. Before in Model 5 all interactions and variables are included. As can be seen the variable for the year 2015 has been omitted due to multincolinearity, however, due to the big amount of firm observations, this should not be a problem. The output of the regression is shown below in table 10.
Table 10 ETR1 output
ETR 1 (1) (2) (3) (4) (5)
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5
(0.0104) (0.0112) (0.0104) (0.0112) ROA 0.000704* 0.000490 0.000712* 0.000499 (0.000386) (0.000383) (0.000386) (0.000383) MNE 0.0366*** 0.0200*** 0.0366*** 0.0199*** (0.00389) (0.00412) (0.00389) (0.00412) MNE*ROA 0.000201 0.000412 0.000191 0.000401 (0.000447) (0.000444) (0.000447) (0.000444) MNE*PTR 0.0746*** 0.0687*** 0.0752*** 0.0693*** (0.0257) (0.0255) (0.0257) (0.0255) Constant 0.0250*** 0.137*** 0.0442*** 0.137*** 0.0441*** (0.00941) (0.00357) (0.00872) (0.00463) (0.00918) Observations 7,689 7,689 7,689 7,689 7,689 R-squared 0.045 0.129 0.144 0.129 0.144
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
When looking at the pre-tax return, it can be seen that the correlation stays in all cases positive and significant. When considering the interaction term MNE*PTR, it is positive and partially significant. Those two results show, that there are implicit taxes for both DCs and MNEs, and that MNEs bear more implicit taxes than DCs. However, especially when controlled for size and years, the interaction value is low, which does not have a profound base to reject Hypotheses 3, when considering the results in the analysis before. Thus, the results align with my expectations of a positive correlation between PTR and ETR, however the expectation that the interaction term is negative, has not been supported.
The Leverage coefficient is statistically significant when including the control variables and shows a negative sign, which aligns with previous research and my expectations. Thus, Hypotheses 5 can be accepted, however due to the low value with caution. The return on asset coefficient shows positive values, however they do not indicate a strong correlation. Which indicates that
are more profitable are not more engaged into tax planning activities. The interaction as well shows that profitability is not a strong indicator for the level of ETR. Thus, profitability is not a strongly supported reason for the higher ETR for MNEs compared to DCs, rejecting hypotheses 6.
The dummy variable MNE, shows straight through all models a positive significant value, which was to be expected, as the ETR is higher for DCs than for MNEs.
Concerning Size, it can be seen that it does have a positive statistically significant influence on the ETR, which can support political cost theory, however the heterogeneity of the different EU countries, makes a strong support of the theory impossible.
4.6.5 Robustness check As a robustness check for my model, I regressed the same independent variables, with ETR2 as a dependent variable, the results are shown in Table 11. Although, the calculation is different, due to the fact, that it has the same nominator, there should not be extensive differences. The model shows that there are no substantial differences in most of the variables. The values for the PTR are representing weaker correlation, the values for ROA are not significant and the Leverage coefficient has a stronger correlation. However, all values have the same sign, which indicates no major differences. In sum it can be concluded that the regression model is robust.
Table 11 ETR2 Output
(1) (3) (4) (5) (6)
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5
(0.00596) (0.00583) (0.00581) o.Y2015 - - - SIZE 0.0195*** 0.0159*** 0.0159*** (0.00186) (0.00217) (0.00217) PTR 0.0586** 0.0507* 0.0582** 0.0502* (0.0285) (0.0284) (0.0285) (0.0284) Leverage -0.0621*** -0.103*** -0.0621*** -0.103*** (0.0132) (0.0143) (0.0132) (0.0143) ROA 0.000677 0.000506 0.000686 0.000515 (0.000489) (0.000488) (0.000489) (0.000488) MNE 0.0393*** 0.0259*** 0.0393*** 0.0260*** (0.00493) (0.00524) (0.00493) (0.00524) MNE*ROA 0.00137** 0.00154*** 0.00135** 0.00152*** (0.000567) (0.000565) (0.000567) (0.000565) MNE*PTR 0.0149 0.0102 0.0154 0.0107 (0.0326) (0.0325) (0.0326) (0.0325) Constant 0.0508*** 0.132*** 0.0575*** 0.126*** 0.0514*** (0.0116) (0.00453) (0.0111) (0.00586) (0.0117) Observations 7,689 7,689 7,689 7,689 7,689 R-squared 0.015 0.058 0.064 0.059 0.065
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
4.6.6 Discussion of the Results
Table 12 below summarizes the expectations and findings concerning the impact on the ETR and the differences for MNEs and DCs.
Table 12 Expectations and Findings
ETR1 ETR2 ETR1 ETR2
PTR Positive Positive Positive Positive
ROA ? ? Weakly positive Weakly positive
SIZE ? ? Positive Positive
The aim of this analysis, was to obtain insights into the differences of MNEs and DCs in their effective tax rate, in order to do so I chose four different analyses. The first analysis was a comparison of the ETR of MNEs and DCs in the last five years. In contradiction to hypotheses 1, as well as previous literature, which predicted that MNEs have a lower or similar ETR than DCs, MNEs actually have a higher ETR. The analyses conducted in chapter 4.2 and chapter 4.3, are examining this result. Due to the fact that DCs bear implicit taxes, the correlation of their ETR and PTR is positively correlated. Thus although their ETR might be lower, their PTR is as well, which does not indicate a lower tax burden. Additionally, the correlation between the Return on Equity and the ETR is positive, which suggests the same conclusion when considering after-tax income. However, the correlation between the ETR and ROE for MNEs is negative, which indicates that, for one they do not bear implicit taxes and two, that they are able to carry some of their tax benefits into their after-tax returns. The analysis in chapter 4.2 suggests that, explicit tax advantages are higher for MNEs, which supports Hypothesis 2. The outcome of the Analyses in chapter 4.2 and 4.3 concerning the bearing of implicit taxes for MNEs, is contradictory, thus, I turned to Analysis 4 for further investigation. The regression output shows a low non-significant correlation of the interaction term PTR*MNE, which suggests rather the support of
PTR decreases as well. Which supports implicit tax theory, which states that tax exempted assets attracts capital.
Previous research has proven that higher leveraged firms can take more advantage of their tax shields, the regression analysis proves this point. Which suggests that MNEs have a higher tax advantage. However, as the correlation is weak, the conclusion can not be supported strongly.
Concerning Hypotheses five, according to previous research MNEs are more profitable than DCs, however when the regression outcome is considered the higher profit has not a strong explanatory value when analysing the higher ETR of MNEs.
Thus, I can conclude, although the ETR of MNEs is higher, the occurrence of implicit tax disadvantages as well as better tax planning opportunities, can alter the actual tax burden of MNEs.
5. Conclusion
DCs have to stay in the country and are bound to the domestic corporate tax is not valid. The sole fact, that DCs face higher taxes is not enough to explain the difference of taxation between the two entities. When explicit tax is high, DCs actually have an advantage due to the offsetting effect of the implicit tax advantage. Although, MNEs bear some implicit taxes as well, the effect is greater for DCs, which can give them a tax advantage in the end.
According to my analysis, contradictory to most of the previous research, MNEs face a higher ETR than DCs. However, by considering implicit tax theory, the difference can be relativized.
The implicit taxes DCs face is an advantage, when taxed high. However, when taxed low it can become a disadvantage, because of a lower pre-tax of return.
According to my analysis, MNEs bear a lower amount of implicit taxes. This is an advantage when taxed low. Furthermore, they are able to retain some of their tax benefits for their after-tax return.
In combination with other factors analyzed, such as the better tax planning opportunities especially for MNEs with higher leverage, it can be seen that the sole comparison of the ETR, does not represent the actual tax burden.
the ETR depends on a firm’s capital structure, it is not advisable to be used for the comparison of MNEs and DCs.
My results can be concluded in the following, first EU-DCs do have a lower ETR than EU-MNEs. Second, DCs bear more implicit taxes than MNEs, giving them an advantage when highly taxed. And third, implicit taxes, relativize the impact of a high ETR, making it difficult to use as a tax burden measurement when comparing MNEs and DCs.
5.1 Future research and Limitations
The research done, is only a small contribution to the field of implicit tax
research. Academic literature in the field is in need of research concerning the deeper understanding of implicit taxes and the difference and effect of this concept on firms, policy making as well as comparison of tax burdens.
Furthermore, more research has to be done concerning the effect of different economic determinants on the ETR. The fact that my results concerning the profitability have not met the expectations of previous research, shows that there is a need for research of capital structure and profitability in order to obtain a deeper understanding of those determinants relating to the ETR or other tax rates. Furthermore, if the actual tax burden of MNEs is lower, the question whether international tax avoidance strategies are profitable should be investigated in.
My study only represented a limited sample of firms due to missing data and the scope of the thesis. Furthermore, the fit of the model has not been optimal. However, this was to be expected, due to the fact, that only a small amount of variables has been used. A model with more variables should show a better fit. The fact that the R^2 is considerably low for both regression
variety of influences to face. This makes is not the most convenient tax rate measurement when comparing companies.
Thus, a call for future research with different tax rates is in order.
Sources
Becker, G., 1983. A Theory of Competition Among Pressure Groups for Political Influence. Quarterly Journal of Economics 98:3. 371–400.
Brooks, C. 2015. Introductory Econometrics for Finance. Cambridge University
Press.
Calculating the effective tax rates of large Dutch companies and identifying tax avoidance. Report. Centre for Research on Multinational Corporations, Netherlands.
Chen C.J.P., Cheng, C.S.A., He, J., Kim, J., 1997. An Investigation of the Relationship between International Activities and Capital, Structure. Journal of
International Business Studies, 28:3. 563- 577.
Chyz, J.A., Luna, L., Smith, H., 2014. Implicit Taxes of U.S. Domestic and Multinational Firms Over the Past Quarter-Century. Working Paper.
Department of Accounting and Information, University of Tennessee. Clausing, K.A., 2009. Multinational Firm Tax Avoidance and Tax Policy.
National Tax Journal 62:4, 703-725
Click, R.W., Harrison P., 2000. Does multinationality matter? Evidence of value destruction in U.S. multinational corporations. Working Paper
Debaere, P., 2006. Does where you go matter? The impact of outward foreign direct investment on multinationals’ employment at home. Discussion Paper
No. 5737
Dharmapala, D., and Riedel, N., 2013. Earnings shocks and tax-motivated income-shifting: Evidence from European multinationals. Journal of Public
Dyreng S. D., Hanlon M., Maydew E. L., Thornock J. R., 2014. Changes in Corporate Effective Tax Rates Over the Past Twenty-Five Yeats. Working
Paper.
Fatemi A.M., 1988. The Effect of International Diversification on Corporate Financing Policy. Department of Finance. Journal of Business Research. 16:1. 17-30.kin
Field, A., 2009. Discovering statistics using SPSS, Sage publications.
Gravelle, J., 2013. Tax havens: International tax avoidance and evasion. Congressional Research Service, CRS Report for Congress, January 23. Grubert, Harry. Foreign Taxes and the growing share of U.S. multinational company income abroad: profits, not sales are globalized. National Tax
Journal. 2012, 65:2. 247–282
Gupta, S., and K. Newberry., 1997. Determinants of the Variability in
Corporate Effective Tax Rates: Evidence from Longitudinal Data. Journal of
Accounting and Public Policy 16:1. 1–34.
Hines, J.R., 1999. Lessons from behavioral responses to international taxation. NBER Working Paper. Cambridge.
Hsu, C.H.H., Chiang, Y.C. and Liao, T.L., 2013. Testing pecking order behaviors from the viewpoint of multinational and domestic corporations.
Investment Management and Financial Innovations, 10:2, 158-165.
Janssen, B., 2005. Corporate Effective Tax Rates in the Netherlands. The
Economist 153:1. 47–66.
Janssen, B., and W. Buijink., 2000. Determinants of the Variability of
Corporate Effective Tax Rates (ETRs): Evidence for the Netherlands. MARC
Working Paper no. 3/2000-08, University of Maastricht, Maastricht, the
Netherlands
