The impact of taxes on foreign activities of multinational
companies
Name: Bing Sep
Student number: 10010165
Thesis supervisor: dr. ir. S.P. van Triest Date: 26 June 2017
Word count: 13383, 0
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by Student Bing Sep who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Multinational corporations have activities in foreign countries where they produce and sell, just as in their home country. However, tax levels may be different between the home country and foreign countries. Do these difference influence the location choices of foreign activities? This paper investigates whether foreign activities are influenced by tax induced incentives or whether non-tax country characteristics have more influence on foreign activities. The research is approached by making use of a simple linear regression where the foreign effective tax rate (ETR) is the proxy to evaluate the impact of taxes on the foreign activities of multinational corporation’s. Subsequently country attractiveness and entry barriers capture the effect of country characteristics on foreign activities of multinationals. In order to conduct the research publicly available data on U.S. multinational corporations is used for the time period 2006 – 2016. The empirical results show a positive significant relation between the effective tax rate and foreign activities. Indicating that U.S. multinational corporations choose to increase their activities in a country even though they have to pay more taxes. This result in combination with the positive significant influence of country characteristics on foreign activities implies that: U.S. multinational corporations are influenced by non-tax considerations, including country characteristics and thus, are not induced by tax avoidance practices when making foreign activity choices.
Keywords: Foreign activities; Effective tax rates; Country characteristics; Tax avoidance; U.S. multinational corporations
Acknowledgment: I would like to express my graduated to my thesis supervisor dr. ir. S.P. van Triest for steering me in the right direction by providing useful comments, remarks and engagement through the learning process of this master thesis.
Contents
1 Introduction ... 5
2 Literature review ... 8
2.1 Tax planning ... 8
2.2 Taxes and Foreign activities ... 9
2.3 Effective Tax Rate (ETR) ... 11
2.4 Empirical findings: what are the most relevant results from the literature regarding the impact of tax levels on foreign activities? ... 12
2.5 Hypothesis development ... 13
3 Empirical Testing ... 15
3.1 Methodology ... 15
3.2 Data ... 15
3.2.1 Foreign activities and ETR ... 15
3.2.2 Country attractiveness ... 17
3.2.3 Entry Barriers ... 18
3.3 The model ... 19
3.4 Variable description ... 20
3.4.1 Dependent and main independent variables ... 20
3.4.2 Country attractiveness and entry barrier independent variables ... 21
3.4.3 Controlling variables ... 23
3.5 Descriptive statistics ... 24
4 Results... 28
4.1 Correlation Matrix ... 28
4.2 Regression results ... 29
5 Conclusions and recommendations ... 32
References ... 35
1 Introduction
Foreign governments try to attract firms to their shores by offering various inducements, like lower tax rates, tax subsidies and preferential treatments. Whereas the domestic government builds all kind of enforcement walls to prevent their firms from moving to these lower tax jurisdictions. Next to these strict regulatory rules, the home country imposes several tax incentives (e.g., the U.S. production activities deduction) to ensure that their domestic firms maintain competitive (Hines, 1996). In the midst of this all a multinational corporation (MNC) is navigating its way through these different kind of regulations, considering all the tax benefits and costs, in order to maximize their after-tax returns. With this theory in mind Waegenaere, Sansing, and Wielhouwer (2006) showed in their theoretical work that firms indeed optimize their operations in response to the strategic interplay of rules and enforcement by the countries in which they operate.
These optimization strategies in regard to taxes are called effective tax planning1 procedures. The most common of these strategies is locating operations in low-tax jurisdictions after which income is shifted from high-tax locations to these low-tax jurisdictions, so the differences in tax rules will be exploited and at the same time MNCs take advantages of different tax subsidy agreements (Hanlon, 2008).
Policy makers and governmental agencies have expressed their concerns about these tax avoidance practices. Recently, Senator Carl Levine stated, “Too many corporations are using tax trickery to send their profits overseas and by doing so they avoid paying their fair share in the U.S” (Browning, 2008). This statement clearly reflects the current perception on effective tax planning and the way in which it is discussed. However, this does not reflect the latest thinking and empirical evidence on the economic behavior of MNCs which are subject to international taxation. Collins, Kemsly and Lang (1998) for instance reported evidence that U.S. MNCs whose average foreign tax rates exceed the domestic ones shift income into the United States, on the contrary they do not find evidence on U.S. MNCs shifting income out of the United States. Similarly Dyreng and Lindsey (2009) find that profitable manufacturing firms with operations in tax havens do not report lower federal tax liabilities on foreign income opposed to firms without operations in tax havens.
In stark contrast to these findings, Clausing (2009) finds that transfer pricing costs the U.S. government on an annual basis around 35 percent of its corporate income tax revenues which was
1 In this paper the terms “effective tax planning” and “tax avoidance” will be alternating throughout the paper,
whereas both include any tax planning method that taxpayers use to legally reduce their income tax payments. Tax evasion (fraud) is not considered as “tax avoidance” in this paper. The latter part of section 2.1 will explain this in
approximately $21 billion in 2004. In line with this research Taylor, Richardson and Lanis (2015) documented that firms engage in transfer pricing activities, their outcome is based on a hand collected sample of Australian firms over 2009. The discrepancy in the literature is partly attributed to the fact that there is some form of measurement error in the proxies which compute the incentives to shift income, according to Klassen and Laplante (2012).
Although the outcomes of these papers are very different, they are similar in their research approach, which is investigating transfer pricing from the perspective that MNCs use their foreign activities to shift income regardless of their other potentials. However a more generic question would be to ask whether foreign activities are influenced by tax planning expenditures? Like noted in the international business literature location characteristics are also of great concern to foreign activities of MNCs (e.g., Davidson, 1980; Davidson and McFetridge, 1985; Lipsey, 2004). So a more specific question would be; are foreign activities of multinational corporations influenced by tax planning
expenditures or do non-tax country characteristics have more influence on foreign activities?
In order to answer the above stated research question, a right proxy to capture the effect of effective tax planning must be determined. In line with prior research the effective tax rate (ETR) is considered as an effective measure to determine the effectiveness of tax planning (e.g., Mills, Erickson and Maydew, 1998; Phillips, 2003; Rego, 2003). In consistency with such research, the ETR is defined as the ratio of income taxes currently payable to pre-tax accounting income. Which means that if two firms have the same pre-tax accounting income but pay different amounts of income taxes, the firm which pays less tax will have a lower ETR and thus will be noted as being more effective in tax planning (Rego, 2003).
The ETR will be computed with data from the publicly available data source COMPUSTAT in combination with COMPUSTATs segment data for the time period 2006 -2016. By using this data source the research is restricted to U.S. MNCs only. However, this allows us to hold many variables constant, including financial accounting rules and parent country tax rules (Dyreng et al., 2009). The country characteristic data will be computes using various publicly available data sources, which will be described in more detail in section 3.4.2.
The empirical results of the regression analysis indicates that there is a positive significant relation between foreign activities and effective tax planning. Meaning that an increase in effective tax planning leads to an increase in foreign activities. Implicating that U.S. MNCs choose to increase their activities in a country even though they have to pay more taxes. This suggests that non-tax considerations are of a larger importance to U.S. MNCs in choosing where to position their foreign activities. In line with this outcome, the results on the country characteristic variables
also show a positive statistical significance on foreign activities. Indicating that country characteristics have a positive effect on foreign activities of MNCs. To be more specific, a multinational corporation attaches more interest to countries which are more attractive to business and have a potentially better financial long-term prospect while at the same time having lower entry barriers. Which leads to conclude that foreign activities of U.S. MNCs are influenced by non-tax considerations, including country characteristics and thus, are not induced by non-tax avoidance practices.
This paper extends the accounting literature by research a more generic question whether foreign activities are influenced by tax considerations. In order to do this non-tax considerations like country characteristics are taken into account in the form of country attractiveness and entry barriers in the decision making process of looking at foreign activities. Furthermore, the research is conducted in a new time era where most previous research is based on data around the U.S. tax reform act of 1986 (e.g., Lipsey and Weiss, 1984; Grubert and Mutti, 1991; Dyreng, 2008).
The reminder of this paper is designed as follows. Section 2 discusses prior literature and develops the hypothesis after which, section 3 presents the research design by describing the data and variables. Section 4 discusses the results of the empirical test after which section 5 concludes.
2 Literature review
Multinational corporations (MNCs) operate in multiple political, cultural, and economic environments, as well as in different tax jurisdictions. Because of these different fields they must act upon, a single subject like taxation can’t be seen from one perspective. For instance in the tax accounting literature several researchers have examined the topic of tax avoidance on a firm-level basis alone from all kind of angles, for example, the relation of tax avoidance and financial reporting behavior has been examined (Frank et al., 2009, Lisowsky, 2010; Robinson et al., 2010; Wilson, 2009), or the effects that managers have (Dyreng et al., 2010), managerial incentives and governance (Armstrong et al., 2010; Desai and Dharmapala, 2006); auditor expertise (McGuire et al., 2012), or the influence the founding family has (Chen et al. 2010), investment in tax planning (Mills et al., 1998), foreign operations (Rego, 2003), the use of tax havens (Dyreng and Lindsay, 2009), and the new economy business model (Omer et al., 2012).
When we look more closely at these different angles they can be traced back to some core principles in the accounting literature. For instance, accountants are able to use their understanding of tax return data and the income tax data of financial reports to derive measures of firms’ tax avoidance activities and to point out important questions about the determinants and consequences of tax avoidance (Hanlon and Heitzman., 2010). Stating it differently accountants have several comparative advantages over their colleagues in the economic and finance fields to examine the same questions from a different perspective. Because accountant utilize specific knowledge of financial accounting rules and have an understanding of the institutional details of tax and financial reporting, which enables them to identify and examine research questions (Slemrod, 2005). This is why Maydew (2001) pointed out that there is a need for tax researchers from an accounting point of view to bring more depth into the specific field of tax avoidance by incorporated more and different theory.
In the reminder of this paragraph detailed background information will be provided on the topics which will be most discussed during this paper, after which the most prevalent empirical findings from the existing literature will be discussed and finally the paragraph is concluded with the hypothesis development.
2.1 Tax planning
Before the mid-1980s tax research in the academic field of accounting could be divided into two streams: (1) Legal research, which evaluated the effect of taxes on exogenous transactions, these kind of papers were mostly published in tax journals (e.g., Ladd, 1975; Spicer and Becker, 1978).
And (2) Policy studies, which evaluated the efficiency effects of taxes, this kind of research was usually published in public economics journals (e.g., Modigliani and Miller, 1963; Kakwani, 1977). Even though most tax papers at the time were not published in accounting journals, it was an accounting researcher Wolfson together with finance researcher Scholes (1989)2, who designed a framework from which the role of taxes in an organization can be explained. This Scholes-Wolfson paradigm is based on the knowledge found in the microeconomics and taxation fields. However, the paradigm became most central in the field of empirical tax accounting research (Schakelford and Shelvin, 2001).
The core of the paradigm is based around three themes - All parties, all taxes, and all costs – which state that in order to achieve organizational goals all actors involved need to be taken into account (i.e., all parties: all parties involved in a tax incentive transactions need to be considered, All taxes: not only explicit taxes but also implicit taxes need to be taken into consideration, All costs: all the costs need to be considered in the planning process). This implies that tax minimization is not necessarily the goal of tax planning. But instead tax planning is merely there to ensure tax efficiency, which takes all considerations into account and makes sure that they work together in the most tax-efficient way (Scholes et al., 2001).
An implication of effective strategy is given by Levenson (1999) who states that, tax planning strategies can help firms to reduce their effective tax rates - when executed properly – from 35-40 percent to as low as 10 percent. And when a firm reduces its effective tax rates this will result in higher earnings per share and ultimately places companies in a more favorable position opposed to their competitors when they are analyzed. Knowing that managers are partly rewarded on firm performance (i.e., After-tax shareholder value), they will surely not risk prosecution on behalf of their shareholders and thus not try to evade home country taxes. They rather just search for legal means to avoid taxes and thus adopt efficient tax planning activities (McGuire, 2012).
2.2 Taxes and Foreign activities
Hines (1996) starts his working paper with considering why it is even worthy to investigate the impact of taxation on foreign direct investment. He states: “if all other considerations are held constant, international investors would prefer to avoid taxes than to pay them, so there must exist some situations in which tax differences significantly influence investment.” (Hines, 1996, p. 8).
As described at the start of this review, international investments are influenced by a wide range of considerations of which tax is just one. This gives rise to various findings on the impact of taxes on international investments. For instance Vernon (1997) argues that tax differences are too small to have an effect on a firm’s investment location decision, with the underlying presumption that there is too much uncertainty facing firms in making a predictable forecast of economic and political events. On the contrary there are academics who argue that firms invest in low-tax jurisdictions because of tax reasons. As evidence they point out that tax havens host a disproportionate fraction of the world’s foreign direct investment (Desai 2006; Hines, 2006,2007; Hong and Smart, 2007; Dharmapala, 2008).
Before we continue with the analysis I will provide a short summary on how MNCs can create opportunities to avoid taxes via foreign activities. One particular strategy in avoiding taxes via foreign activities involves a mechanism called transfer pricing. Transfer pricing is the price set on goods and services for internal trading among affiliates and the MNC. The prices that are set will affect the income of the MNC across the different jurisdictions it has affiliates. There is legislation on intrafirm trading, which states that firms need to use prices at ‘arm’s length’. This means that firms need to set their prices as if they were engaging in a market transaction with an unrelated party. However this is quite arbitrary, for example, affiliates can – to a significant degree – choose where to locate research and development activity in order to ensure that royalty payments from other affiliates flow towards lower-tax jurisdictions (Dharmapala, 2008). In consistence with this strategy Desai et al. (2006) find that U.S. MNCs with greater focus on research and development are more likely to establish affiliates in a tax haven. These findings are in line with some recent news headlines. On the 30th of august 2016 Apple was fined for 14.5 billion dollar by the European commission because it was accused of shifting its income through Ireland with who they has a special tax deal ("Apple ordered to pay €13bn after EU rules Ireland broke state aid laws", 2016). Like the findings of Desai, Apple is a U.S. MNC with great focus on research and development whereas Ireland is seen as a tax haven by the IMF and the tax research organization (Miedema, 2014).
Another tax avoidance mechanism is based on strategically deploying debt among the MNC affiliates. It is beneficial for a MNC to finances the activities of a high-tax affiliate using debt issued by a low-tax affiliate. This practice is known as ‘earnings stripping’ and it directs interest payments to low-tax countries, while generating interest deduction in high-tax countries. Desai et al. (2004) document that US MNCs tend to locate debt in higher-tax countries, whereas Mintz and Weichenrieder (2005) find that the wholly owned subsidiaries of German MNCs tend to use more debt in countries with higher tax rates. This evidence only supports the theory of earnings stripping
partly because it does not document whether the higher debts are provided from the low-tax jurisdictions.
The above stated strategies are extremely simplified, however, they are sufficient enough to indicate that there are incentives for MNCs to use foreign activities as a way of avoiding taxes, and thus taking foreign activities into account when making an efficient tax plan. If we take a closer look at the empirical research, we find that the first series of academic papers who looked at these incentives can be found in the field of economics, which means that these papers are based on the country level effect. For instance Hines and Rice (1994) find that a large elasticity of US FDI holding and of US affiliates’ profits with respect to foreign countries’ corporate tax rates. Or Grubert and Mutti (1991) documented that the US appears both to import more from and export more to low-tax countries where MNC investment is greater. As mentioned these papers findings are based on aggregated data by country.
From an accounting perspective we are more interested in the behavior of individual firms, and thus tend to look more at tax avoidance from a firm-level basis. Jacob (1996) was the first paper to adopt a firm-level approach in investigating the incentives facing MNCs, he documented that tax-motivated income shifting through transfer pricing is prevalent. These findings are in line with those of Grubert and Mutti (1991). More recent accounting studies have confirmed that MNCs have a preferable advantage over purely domestic U.S. firms because they are able to shift funds, profits and services among variable taxed jurisdictions, resulting in a preferential tax outcome (Rego, 2003; Hanlon, Mills and Slemrod 2007; Tayor, Richardson & lanis, 2015).
2.3 Effective Tax Rate (ETR)
The effective tax rate (ETR) has proven to be an effective tool in measuring tax planning (e.g., Mills, Erickson and Maydew, 1998; Stickney and McGee, 1982 Phillips, 2003; Rego 2003; Hanlon et al., 2010). ETR is defined by these researchers as the ratio of income taxes currently payable divided by pre-tax income. Several reasons can be pointed out why ETR has been so successful in capturing the tax planning effect.
(1) ETR is able to capture book-tax differences as documented by Mills (1998), who noticed that firms with greater book-tax differences had larger Internal Revenue Service (IRS) audit adjustments, which is consistent with greater tax-avoidance activities. Where a Book-tax differences is defined as the difference between a firm’s accounting income and taxable income. When noticing that taxable income is incorporated in the numerator of the ETR formula and that financial accounting income is noted in the denominator, we can conclude that the ETR formula
will capture these kind of variations. Mostly this will result in a lower ETR, because tax-motivated transactions such as tax credits, deferral of income recognition for tax purposes and sales to foreign corporations, will reduce a firms’ taxable income and thus the numerator in the ETR formula (Hanlon, 2005).
(2) Multinational corporations use their foreign activities to avoid taxation, by shifting their income from a high-tax jurisdiction to a low-tax jurisdiction. By doing so the MNCs worldwide taxable income will be reduced (numerator) whereas the worldwide pre-tax accounting income has not changed and thus the denominator remains constant. So the ETR captures the effect of MNCs shifting their income to their foreign operations, because their taxable income will be reduced while their financial accounting income remained the same, which will result in a lower ETR. This is why ETR is a reasonable (indirect) measure of effective tax planning (Rego, 2003).
From the above stated reasoning, we can conclude that a lower ETR indicates less taxes and thus this is an indication of tax avoidance behavior.
2.4 Empirical findings: what are the most relevant results from the literature regarding the impact of tax levels on foreign activities?
Multinational corporations are profoundly different then their domestic counterparts as they operate in multiple cultural, political and economic environments, as well as in different tax jurisdictions. As argued by Rego (2003) MNCs have more extensive foreign operations which enables them to exploit their foreign activities to avoid income taxes. Whereas these opportunities are not available for their domestic counterparts or at least in a lesser sense. Governments are becoming more concerned on the impact of tax avoidance behavior. The U.S. Department of Treasury (1999) underlines this concern by publishing a document containing 188 pages about tax avoidance. As the document shows, MNCs avoid taxes by locating operations in low-tax jurisdictions or by shifting income from high-tax locations to low-tax locations and by exploiting difference in the tax rules of different countries. Unfortunately this document as well as the public discussion surrounding tax avoidance, does not reflect the latest thinking and evidence on the economic behavior of MNCs which are subject to international taxation (Hines, 1996).
As noted by Hines the latest thinking on tax avoidance isn’t as black and white as presented by the U.S. government which marks tax avoidance as a big problem and only sees it as a negative potential. To point this out, Collins, Kemsley and Lang (1998) examined U.S. multinational manufacturing companies from 1984 to 1992 to measure the cross-sectional relation between firm-level foreign profit margins and average foreign tax rates. They documented that U.S. MNCs facing
average foreign tax rates in excess of the U.S. shifted income of approximately 25-30 million of income per year into the United Stated. On the contrary they did not find any opposing evidence on the shifting of income out of the U.S.
Similar to these finding Dyreng and Lindsey (2009) investigate how foreign operations located in tax havens and other jurisdictions affect the income tax rates of MNCs based in the U.S. The research is conducted by means of a newly designed regression framework which identifies the variation in tax expense measure associated with income earned in different locations. With this regression framework they find that, U.S. MNCs with operation in some tax haven countries have higher federal tax rates on foreign income than other MNCs. Suggesting that tax haven operation may increase U.S. tax collection at the cost of foreign country taxes.
On the flip side there is also academic literature which states the contrary and is more in line with the concerns expressed by policy makers. Clausing (2009) investigates tax policy consequences of international tax avoidance, focusing on U.S. MNCs over the period 1982-2004. The research is conducted by examining the estimated relation between U.S. affiliate profit rates and foreign country tax rates. The research estimates that income shifting costs the U.S. government approximately 35 percent of corporate income tax revenues. In line with this research Taylor, Richardson and Lanis (2015) study the major determinants of transfer pricing aggressiveness, based on a hand-collected sample of 183 publicly-listed Australian firm for 2009. They document that firms engage in transfer pricing activities and mostly through the joint effect of intangible assets in combination with their foreign affiliates.
Knowing that the literature provides mixed evidence on the extant of income shifting by U.S. MNCs, Klassen and Laplante (2012) address the conflicting results in the existing literature and point out that, this mixed evidence is the result of measurement errors in the proxies for the incentive to shift income. More specific they find that previous research has made use of annual proxies in the setting of tax avoidance, where either multiperiod proxies or instrumental variables seem more appropriate.
2.5 Hypothesis development
As documented in the previous sub-paragraph, the empirical findings in the existing literature about the impact of tax levels on foreign activities reach no uniform consensus. However these papers show a similarity in the way in which they approach the impact of tax levels on foreign activities. They do so by approaching tax levels on foreign activities from a transfer pricing perspective and investigate whether MNCs use foreign activities to shift income regardless of their
other potentials. By these other potentials is meant the incorporation of non-tax incentives to move foreign operations into other tax jurisdictions. For example, the international business literature points out that country characteristics, like country attractiveness and entry barriers are of great concern to MNCs regarding their foreign activities (e.g., Davidson, 1980; Davidson and McFetridge, 1985; Lipsey, 2004).
So a more generic question on the impact of tax levels on foreign activities would be: are foreign activities of multinational corporations influenced by tax planning expenditures or do non-tax country characteristics have more influence on foreign activities? This could imply that a MNC decides to invest in a specific country not because of its lower taxes but just because of a versatile climate for business operations. Meaning that the assumption to engaging in tax avoidance practices is not the primary reason to use foreign activities.
However, as noted in the prior sub-paragraph the evidence on the tax avoidance practices from multinationals via their affiliates is mixed. Still the majority of research provides evidence – even though the amounts are different – on the existence of this tax avoidance practices (e.g., Grubert et al., 1991; Rego, 2003; Clausing, 2009; Klassen et al, 2012; Taylor et al., 2012). Also as noted by Klassen et al. (2012) the literature which do not find any evidence or even contradict the use of tax avoidance practices of MNCs via their affiliates is attributed to measurement error. This in combination with the logic of Rego (2003) who noted that multinational corporations have the ability to exploit differences in tax jurisdictions. Leads to conclude that it seems rational for a multinational corporation to let the foreign effective tax rate be of influence on foreign activities. Stated differently:
Hypothesis 1: The volume of a multinational corporation foreign activities in a particular country is negatively
3 Empirical Testing
3.1 MethodologyIn an ideal situation one would want to know all the segment and transfer pricing details of foreign affiliates of a MNC in order to study the affect taxes have on foreign activities. Meaning that this information would provide all the necessities to see whether a MNC chooses to set up an affiliate in a particular jurisdiction because of the influences of taxes. However, due to the inability to acquire such private information, previous studies made use of ETR as an indicator of effective tax planning, hence this approach will also be adopted in this paper. Furthermore, to get a more comprehensive view on the possible influencers of foreign activities country characteristics and entry barriers are taken into account. The remaining part of this paragraph deals with the data and how it is retrieved, the model and the variables after which the paragraph is concluded with the descriptive statistics.
3.2 Data
3.2.1 Foreign activities and ETR
This sub-paragraph is concluded with table 1 which summarizes the data selection procedure. The data which is used in this paper is obtained via the publicly available database COMPUSTAT. More specifically two datasets were obtained via this medium in order to create the variables ETR and foreign activities. (1) Compustat Monthly Updates – Fundamentals Annual North America for the time period of 2006-01 to 2016-12 and (2) Compustat Monthly Updates – Segments North America (Non-Historical) ranging from 2006-2016. Combined these datasets provided a total of 85,354 firm-year observations.
The dataset on the Annual fundamentals contained all North American firms where the data set on geographical segments contained data on their affiliates per segment. In order to make sure that both datasets contained the same MNCs there is a need to filter out firms’ which only appear in one particular set. This adoption reduced the initial amount of observations by 14,272 firm year observations. To compute the dependent variable there is a need for total foreign assets and the total assets of the MNC. In order to make this possible the “blanks” needed to be eliminated which are the missing values. This adoption reduced the amount of firm year observations with 49,393. Furthermore observations which had missing data to compute the foreign activities or ETR were deleted (i.e., any missing data on: current income tax expense, pre-tax income, total assets and foreign assets, is eliminated) this amounted in the removal of 4,335
firm year observations. Observations with negative assets or stockholder’s equity were similarly eliminated from the sample amounting in the elimination of 1,869 firm year observations. Likewise the data on insurance companies, banks and utilities were deleted because COMPUSTAT does not provide the domestic and foreign pre-tax income of these branches (4,678 firm-years), which results in an inability to compute the ETR.
To be consistent with previous ETR research, all firms with negative tax expenses or negative pre-tax income are deleted from the sample, which amounts to 4,265 firm-years. This is because loss firms ought to have different incentives concerning financials and tax reporting plus the fact that ETRs with negative components does not have an economic contribution (e.g., Stickney and McGee, 1982; Zimmerman, 1983; Shevlin and Porter, 1992; Manzon and Smith, 1994; Gupta and Newberry, 1997; Rego, 2003).
As will be discussed in the next section of this paragraph country level data will be incorporated in the regression analysis. In order to enable this there is a need for segment information which can be traced back to a particular country or continent. So segments with names such as “Other”, “Rest of the World”, “Headquarters”, “Developing Countries”, etc. were removed from the sample (2,436 firm year observations).
TABLE 1
Summary on data selection
Amount of total firms from the combined datasets, 2007-2016 85,354 Less:
Firm’s that do not appear in both data sets
Firm-years with blanks on both foreign assets or total assets Observations with missing foreign activity or ETR data Observations with negative assets
Financial service institutions were removed Firm-years with negative tax or pre tax expense Observations with untraceable segments Firm-years with ETRs greater than 1
Firm-years in top or bottom 1% of return on assets
(14,272) (49,393) (4,335) (1,869) (4,678) (4,265) (2,436) (634) (411) Number of firm-years available for analysis* 3,061 Notes:
In order to avoid any outliers, all the extreme values are eliminated. First, any ETR with a value excessing one is reduced to one (634 firm-years). Finally, any observation which was in the top or bottom one percent of the assets distribution was deleted (411 firm-years). Which resulted in a total of 3,061 firm-year observations and 585 MNCs.
3.2.2 Country attractiveness
One of the key contributions of this paper is to look at efficient tax planning incentives on foreign operations while we incorporate the influence that country characteristics have on these expenditure decisions. The adoption of country characteristics into the regression analysis is similar to the work of Leung and Verriest (2017), who investigated the link between the characteristics of a geographic segment and reasons for withholding or disclosing information of their segment environment. Their approach to incorporate location characteristics is based around country attractiveness and entry barriers. To determine country attractiveness three variables are created.
(1) The first variable to indicate country attractiveness is constituted via the Forbes “best countries for Business” ranking which is widely read and considered by managers as a principle indicator of business attractiveness. Because of the worldwide attention the ranking receives it seems accurate to treat it as a well seen reflection that managers have of the business climate in a certain country. The ranking is available for all the countries and continents that are used in this paper and for the given time period, it is based on eleven different factors: innovation, technology, freedom (personal, trade and monetary), property rights, taxes, corruption, red tape, stock market performance and investor protection. Each category is weighted equally and it only constitutes of countries which data is available across at least eight of these eleven categories (Leung and Verriest, 2017).
(2) The financial potentials of a country are considered in examining its location attractiveness, specifically in the form of the country susceptibility to currency and solvency risk (i.e., its credit rating). As documented by Jorion (1991) and Choi and Prasad (1995) Currency fluctuations are of great concern to MNCs operating in or penetrating foreign countries. The data to incorporate the financial potential of a country is based on a countries long-term credit rating as provided by Fitch ratings. Which is a measure for the risk MNCs face related to a country’s currency fluctuations and its financial instability (Leung and Verriest, 2017).
(3) The final measure to document a country attractiveness is considered by its future economic prospect. More specifically, the size of a countries middle class is used to measure the future economic performance of a country. As documented by Easterly (2001) there is a strong relation between a countries economic growth and the proportion of income that the middle class
holds. This leads academia to believe that individuals who are part of the middle class hold the key to a countries prosperity and can amplify the potential for economic growth which will eventually result in the reduction of poverty. The most stringent evidence on this phenomena is provided by China. China’s middles class has grown form 5 million in 2000 to 225 million people in 2015 according to the economist of the 9 of July 2016. Whereas its economy has shown the highest and most consistent growth pattern in the last decade compared to other BRIC countries.
Ravallion (2009) noted that a higher middle class comes along with a greater reduction in poverty for developing countries. More importantly the middle class is thought of as influential in political decisions, because politicians see the importance of the middle class and thus try to please them by considering them while making important legislation in return for support of the middle class. Kharas (2010) and Kharas and Gertz (2010) approach the topic from an reverse perspective and discuss the potential negative consequences of a small middle class which can cause countries to get trapped in a stagnating growth cycle and thus disables them to transform to a higher income country. Summing up there is a considerable amount of literature that points out the importance of the middle class of a country in order to indicate future economic prospects (Leung and Verriest, 2017). For this reason data is gathered on the size of a countries middle class, the data will be obtained via the World Bank.
3.2.3 Entry Barriers
A side from country attractiveness, entry barriers also are a point of interest to MNCs, especially in combination with investment location decisions. Next to a countries attractiveness entry barriers can oppose a MNC of moving into a specific country, due to problems in the startup face (i.e., getting food on the ground). So by only pointing out the attractiveness of a country we would not get a clear picture of a countries characteristics. We would neglect the possible difficulties that lie ahead, due to entry barriers. In order to capture the effect of entry barriers the paper of Leung and Verriest (2017) serves as a guideline. Here the authors capture the effects of entry barriers by means of four indicators.
(1) The amount of trade freedom in a particular country is measured. Countries which are more open to trade and thus have more freedom of trade are considered to attract more foreign capital and they are associated with lower entry barriers (Barro, 1997; Kapuria-Foreman, 2007). The data to capture freedom of trade is provided by the Economic Freedom of the world database which provides an aggregated index named “freedom to trade internationally”. This index combines measures of regulatory trade barriers on imports, import tariffs, and controls of the
movement of capital and people for almost all countries in the world. This data is obtained via the Fraser Institute, which is a public think tank. (Leung and Verriest, 2017)
(2) The amount of business days it takes on average for a certain country to start a new firm is widely recognized throughout the literature as measure to indicate entry barriers (e.g., Klapper, Levaen and Rajan, 2006). It is argued that in countries where it takes a long time to establish a new business that they are considered to have higher barriers of entry. This is also documented by Klapper, Amit and Guillen (2010), they find that there is a negative relation between the amount of days it takes to start a business and the number of new entrants and business density in a country. The country data is obtained via the paper of Djankov, LaPorta, Lopez-de-Silanes and Shleifer (2002) who calculated for 85 countries the amount of time it takes to start-up a business. (Leung and Verriest, 2017)
(3) Protectionism is considered to be a measure to indicate entry barriers. As noted by Bloningen and Feenstra (1997) higher levels of protectionism do make it harder for new entrants to start a business. So the third variable is protectionism. The data is obtained via the IMD Business School which surveys business executives from countries all over the world. These executives are asked to provide their opinion over the amount of protectionism which disables the degree of business in their country (Leung and Verriest, 2017).
(4) Finally a variable is constructed that captures how burdensome it is to do trade with a certain country. It seems straightforward that the more difficult it is to trade with a country, the higher its barriers of entries are. In order to incorporate this affect into the regression data from the World Bank is obtained in the form of their so called “Doing Business Report”. Here they measure the amount of documents which are required to import a shipment of goods (Leung and Verriest, 2017).
3.3 The model
The following regression model will be used to test the hypothesis, based on the large sample of firm year observations described in the previous sub-paragraph.
FORACTIVi,j = β0 + β1 FORETRi + β2 USETRi + β3 ATTRACTIVEj + β4 FINPOTj + β5
FUTURECOj + β6 FREETRADEj + β7 STARTUPj + β8 PROTECTj + β9 BURDENj + β10 SIZEi,j
+ β11 YEARi,j + β12 INDUSTRYi,j + εi,j
3.4 Variable description
3.4.1 Dependent and main independent variables
FORACTIV: This is the dependent variable which is continuous and measures the extent of foreign activities. It is constructed as foreign assets divided by total assets. Even though the COMPUSTAT segment data contains the data on foreign operating profit and foreign sales, foreign assets is deemed as the most reliable indicator of a firm’s foreign activities. Because foreign operating profit and foreign sales are seen as more prone to manipulation via income shifting and tax-avoiding activities and thus they are not seen as reliable indicators of the location of a firm’s operating activities (Rego, 2003). It needs to be noticed that FORACT is configured in such a way that is captures the effect of each segment of a MNCs affiliate individually instead of the sum all al segments. Which means that the country characteristics affected every segment individually and thus not need to be averaged over the total sum of segments of a MNC. In this way the country characteristics fairly project the influence on a specific segment. Table 2 provides an overview of all the variables
FORETR (-): This is the main independent variable and captures the effect of foreign effective tax planning. Which is constructed out of the ratio of foreign income taxes currently payable to foreign pre-tax accounting income. This is in consistency with prior ETR research, which ensures that the impact of both temporary and permanent book-tax differences are captured. Since the ETR compares the current tax liability generated by taxable income to pre-tax income based on the generally accepted accounting principles. A ETR is able to measure the ability of a corporation to reduce its current tax liability in contrast to its pre-tax accounting income (Rego, 2003), and thus ETR is an effective measure to capture tax-avoidance activities.
Based on prior research which points out different ways in how taxes can be avoided by means of foreign affiliates (e.g., Desai, 2006; Hines, 2006, 2007; Hong and Smart, 2007; Dharamapala, 2008). In combination with the hypothesis leads to assume that the foreign effective tax rate will have an influence on foreign activities in such a way that it will induce firms to use it in order to engage in tax avoiding practices. Meaning that the volume of MNC activities in a country is negatively related to the MNCs ETR in that country (i.e., FORETR will have a negative effect on the dependent variable foreign activities).
USETR (+): Is constructed in consistency with FORETR, which means that it consists of US income taxes currently payable divided by the U.S. pre-tax accounting income. It is expected that this particular independent variable will have a positive impact on foreign activities. An increase in USETR means relative more U.S. taxes payable. This will provide a larger incentive for U.S. multinationals to engage in tax planning behavior via their foreign affiliates (e.g., Rego, 2003; Hanlon et al., 2010).
3.4.2 Country attractiveness and entry barrier independent variables
ATTRACTIVE3 (+): This variable indicates how well a country is rated on business condition such as economic growth, personal freedom, corruption and taxation, unemployment and regulation. Where the index ranges from 1 for the best business rate in this case Denmark to 139 which is the worst Chad. For the ease of interpreting the regression output, the index is reverse-coded which implies that a higher value on business attractiveness indicates a higher country attractiveness for doing business. Which would positively influence foreign activities.
FINPOT4 (+): Here the variable indicates the size of a countries middle class which is conceived as being an important indicator of future economic prospects. In order to construct this variable data provided by the Word Bank is accessed to be specifically data on the magnitude of the middle class in each country. These percentages of total income per quintile of the population are summed for the three middle quintiles. A higher value of this variable indicates that a location has a larger middle class and thus is more attractive country to conduct business, which is associated with a positive influence on foreign activities.
FUTURECO5 (+): This variable sums up the projected long-term credit worthiness of a country by means of looking at its credit rating. The credit rating ranges from AAA to D this is translated into a numerical variable which ranges from 1 (AAA) to 20 (D). Like ATTRACTIVE this index is also reverse-coded which results in higher values indicating better credit quality, which implicates a better long-run economic forecast. Therefore, it is considered as being more attractive to conduct business and thus is associated with a positive effect on foreign activities.
3 Source: https://www.forbes.com/best-countries-for-business/list/3/#tab:overall (accessed on 15 June, 2017) 4 Source: http://wdi.worldbank.org/table/1.3 (accessed on 14 June, 2017)
FREETRADE6 (+): Here the freedom of international trade is constructed as a variable. The data is derived from the Economic Freedom of the World database. Where scores range from 1 which indicates that a country is restrictive to 10 which is the most open form of international trade. These figures are based on the following dimensions, information about tariffs, regulatory trade barriers, black market exchange and the controls of the movement of capital and people. A higher score here indicates higher freedom of trade which is associated with lower entry barriers and thus should have a positive effect on foreign activities.
STARTUP7 (+): This variable indicates the median number of days it takes to start up a new business in a country. The data is provided in working days, and a higher value means that it takes more days on average to start up a new business. So for the ease of interpreting the regression output, this variable is reverse-coded in such a way that a higher value means less days on average to start up a new business. Which is seen as a lower barrier of entry and thus will have a positive effect on foreign activities.
PROTECT8 (+): Through this variable the extent of an executives’ perception on the degree of protectionism which impairs the conduct of business in a certain country is captured. With a score ranging between 1 to 100, where 1 indicates a high level of protectionism and 100 low level of protectionism. A larger score for PROTECT is associated with a lower level of protectionism and thus lower entry barriers, which is positively associated with foreign activities.
BURDEN9 (+): The final variable captures how burdensome it is to do trade with a specific country. This variable is based on the amount of required documents to import a shipment of goods. For the ease of interpreting the regression output the variable is reverse coded, meaning that a high value indicates less required document which is associated with lower entry barriers and thus will have a positive influence on foreign activities.
6 Source:
https://www.fraserinstitute.org/economic-freedom/dataset?year=2014&page=dataset&min-year=2&max-year=0&filter=0&most-free=1&quartile2=1&quartile3=1&least-free=1 (accessed on 15 June, 2017)
7 Source: Djankov, S., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2002). The regulation of entry. The
quarterly Journal of economics, 117(1), 1-37.
8 Source: http://www.imd.org/uupload/imd.website/wcc/scoreboard.pdf (accessed on 15 June, 2017) 9 Source: http://www.doingbusiness.org/Custom-Query (accessed on 14 June, 2017)
Noting that the country characteristics data is country-level based and the COMPUSTAT segment data does not report all the segments on an individual country level. The following method is applied to correct for this difference. If a segment consists of a single separate country, the value of the country characteristic data is simply assigned to this segment data. However if a firm labels a segment as a subcontinent or continent without specifying the specific counties, the average score across all countries of a sub-continent or continent is assigned. Table 8 in the appendix provides further details on the distribution of the affiliate observations across the countries.
3.4.3 Controlling variables
SIZE: This variable controls for firm size making the ratio of foreign assets to total assets (FORACT) comparable across firms. Because firms can have the same amount of percentages in a foreign company, however foreign companies have different amount of assets. Thus holding 95% of the foreign assets of a company with 5 million total assets compared to 5 billion total assets can have vastly different economic consequences. For this reason the controlling variable SIZE is incorporated in the regression. The variable is constructed out of the natural log of total net assets. Because the natural logarithm of any number between zero and one is a negative corrections have been made by adding one to the reported net sales and pre-tax income. This is in line with previous ETR research (e.g., Swenson, 1999; Rego, 2003).
YEAR: Constitutes of a dummy variable indicating the year of observation. This controls for changing factors over the years such as changes in accounting regulations. For example, the varying impact of depreciation deduction on different industries could result in differences across industries. It needs to be noted that this inclusion of a dummy variable assumes that the particular year observation influences all firm similarly.
INDUSTRY: This final controlling variable is also constructed in the form of a dummy, which indicates the type of industry in which the MNC operates in. It is based on COMPUSTATS SIC industry codes and designed as a 1-digit code (e.g., INDUSTRY1 = SIC 2000-3000, which contemplates: food, apparel, paper and chemical product firms.). The notes of Table 2 provides a more comprehensive overview.
TABLE 2
Variable definitions
Variable Definition Source
FORACT* = foreign assets/total assets IAS/AT
FORETR* = foreign current income tax expense/foreign PTI** TXFO/PIFO
USETR* = U.S. current income tax expense/U.S. PTI TXFED/PIDOM
ATTRACTIVE*** = Forbes best countries for Business ranking Forbes
FINPOT = the size of a countries middle class World Bank
FUTURECO*** = long-term credit rating Fitch Ratings
FREETRADE = the amount of trade freedom Fraser Institute
STARTUP*** = median number of days to start up a new firm Djankov et al., (2012) PROTECT = executives perception on degree of protectionism IMD Business School BURDEN*** = # of document required to import a shipment World Bank
SIZE* = natural log (total net sales) (SALE)
YEAR* = dummy variable for year of observation FYEAR
INDUSTRY* = dummy variable for 1-digit SIC codes SIC
Note:
* This variables use COMPUSTAT data ** PTI stand for Pre Tax Income
*** For the regression this variable is reverse coded.
3.5 Descriptive statistics
This sub-paragraph deals with the descriptive statistics of the data. Since much of the data is retrieved from multiple sources the descriptive statistics will subsequently be divided in three parts. Whereas each part will be accompanied with a complementary table which summarizes the statistics.
(1) There is a wide variation in the amount of total assets the MNCs hold in the data sample with an average of 2,894.6 million dollars and a median of 465.4 million dollars. The amount of foreign assets on average constitutes out of 836.4 million dollars with a median of 48.7 million
dollars. The pretax income also varies substantially among the MNCs in the sample, with an average of 248. Million dollar and a mean of 34.8 million dollar.
On average the MNCs in the sample have 37.43 percent of their assets located abroad with a standard variation of 30.38 percent which is a large variation. Stating it differently U.S. multinational corporations incorporated in the dataset have 37.43 percent of their activities located in a foreign jurisdiction, on average.
The mean of the U.S ETR is 28.99 percent with a standard deviation of 20.40 percent. Where the foreign ETR has a similar distribution only with a mean of 22.35 percent but with a dispersion of 19.38 percent. This lower foreign mean of ETR in comparison to the U.S. ETR means that, on average, the income tax rates of the foreign segments where U.S. multinational corporations pay taxes are less than the U.S. statutory tax rate.
TABLE 3
Descriptive statistics on the data to compute the ETRs and foreign activities. (In millions of dollars, except for the ratios). N= 3,061 firm-year observations
Variable Mean Std. dev Q1 Median Q3
FORACT 0.3743 0.3038 0.0992 0.2625 0.6336 USETR 0.2899 0.2040 0.1488 0.2955 0.3724 FORETR 0.2235 0.1938 0.0453 0.2115 0.3294 Total assets 2,894.6 13,562.8 104 465.4 1,826.0 Foreign assets 836.4 3,879.2 3.4 48.7 380.4 PTI total 248.3 865.4 11 34.8 142.6 PTI foreign 88.5 385.2 2.1 5.6 26.7 Net sales 2,885.6 10,145.0 158.1 412.3 1,758.9
(2) If we look at the country characteristics there is substantial variation among the characteristics of segment locations. On average, the affiliates are located in countries which have a rank of the 45th place on the Forbes best countries for doing business ranking10. With a wide variation of 36 places across the segments. The size of a countries middle class is on average 49.71
percent with a small dispersion of 4.34 percent among the segments. A countries credit rating is on average six which is an equivalent of A, and has the implication that the economic situation can affect finance.
The entry barrier variables also exhibit considerable variation across segments locations. On average the affiliates of MNCs are located in segments which score relatively high on the freedom of international trade, with an average segment score of 7.60 out of ten on the freedom of trade measure. It takes on average around 39 days to start up a new business with a large variation of 27 days among segments. The executives’ perception on the degree of entry barriers on the segments location is relatively low with an average of 76 out of 100 which is the lowest form of entry barrier a country can have. Importing a shipment of goods is associated with filling out six documents on average with a variation of two. Table 4 summarizes the country characteristic descriptive statistics.
TABLE 4
Descriptive statistics on the country characteristics and entry barriers
Variable Mean Std. dev. Q1 Median Q3
ATTRACTIVE 44.89 36.09 10.00 33.00 80.00 FINPOT 49.71 4.34 47.70 51.80 52.52 FUTURECO 5.82 4.176 2.00 5.00 8.00 FREETRADE 7.60 0.91 6.85 7.83 8.27 STARTUP 39.14 26.59 18.00 41.00 51.00 PROTECT 75.90 12.35 68.71 74.81 86.71 BURDEN 5.52 2.23 4.00 5.00 7.00
(3) Table 5 provides an overview of the frequencies and descriptive statistics of the control variables. The most frequently reported industries are that of the Standard Industrial Classification (SIC) code 3000-4000 (29.8 percent of the sample), which stand for heavy manufacturing firms. While the professional service branch accounts for 24.3 percent of the total sample (SIC 6000-8000) and the food plus apparel industry accounts for 18.1 percent of the total sample (SIC 2000-3000). Unfortunately 4.5 percent of the sample observations did not provided their SIC code and thus are coded as missing. The firm year observations are fairly evenly distributed throughout 2006-2016, with only a significant amount of lower observations for 2006 (0.9 percent of the observations) which is the starting date of the sample period. The variable SIZE which controls
for the comparability of FORACT among the MNCs is the natural logarithm of net total sales and has a mean of 7.15 with a dispersion of 1.74.
TABLE 5
Descriptive statistics on the controlling variables
Variable Frequency Percentages of Affiliates Industry INDUS1 (SIC 0-2000) 180 5.9 INDUS2 (SIC 2000-3000) 555 18.1 INDUS3 (SIC 3000-4000) 912 29.8 INDUS4 (SIC 4000-5000) 158 5.2 INDUS6 (SIC 5000-6000) 373 12.2 INDUS7 (SIC 6000-8000) 744 24.3 OTHER (SIC > 8000) 139 4.5 3061 100 Year 2006 27 0.9 2007 264 8.6 2008 283 9.2 2009 306 10.0 2010 324 10.6 2011 325 10.6 2012 332 10.8 2013 326 10.7 2014 317 10.4 2015 301 9.8 2016 256 8.4 3061 100
Mean Std. dev. Q1 Median Q3
SIZE 7.15 1.74 5.88 7.14 8.32
Note:
SIC codes 0-2000 include, agriculture, mining, fishing, oil & gas, electrical work and water services. SIC codes 2000-3000 include food, apparel, paper, and chemical products. SIC codes 2000-3000-4000 include metal, machinery, equipment, and other heavy manufacturing firms. SIC codes 4000-5000 include railroads, trucking, water transportation, air transportation, pipe lines and communication. SIC codes 5000-6000 include wholesale and retail firms. SIC codes 6000-8000 include real estate, hotels and professional services (e.g., advertising).
4 Results
4.1 Correlation Matrix
The Pairwise correlation matrix of the empirical model is presented in Table 6. This matrix provides the strength of the relationship between the dependent variable FORACT and all the independent variables including the control variable SIZE. The underlined numbers indicate the p-values of the correlation and the bold p-values are significant at a level of at least 10 percent. TABLE 6
Correlation matrix for the empirical model constructed in section 3.3. N= 3,061
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) FORACT (1) 1.000 - USETR (2) -0.041 1.000 0.022 - FORETR (3) 0.248 -0.021 1.000 0.068 0.247 - SIZE (4) -0.099 0.095 0.003 1.000 0.000 0.006 0.857 - ATTRACTIVE (5) 0.168 0.008 -0.10 -0.015 1.000 0.054 0.667 0.001 0.415 - FINPOT (6) -0.089 0.015 0.036 0.025 -0.602 1.000 0.391 0.392 0.045 0.175 0.000 - FUTURECO (7) 0.101 -0.112 -0.04 -0.029 0.714 -0.540 1.000 0.098 0.182 0.028 0.109 0.011 0.025 - FREETRADE (8) 0.011 -0.005 0.04 -0.004 -0.783 0.426 -0.549 1.000 0.555 0.791 0.029 0.823 0.000 0.000 0.000 - STARTUP(9) 0.346 -0.033 -0.062 -0.060 0.671 -0.381 0.465 -0.44 1.000 0.000 0.070 0.021 0.001 0.000 0.000 0.000 0.000 - PROTECT (10) 0.118 0.005 0.031 0.010 -0.659 0.546 -0.855 0.534 -0.526 1.000 0.078 0.782 0.091 0.564 0.007 0.000 0.058 0.000 0.000 - BURDEN (11) 0.005 -0.003 -0.038 0.028 0.713 -0.319 0.649 -0.624 0.378 -0.644 1.000 0.777 0.874 0.034 0.128 0.000 0.000 0.000 0.000 0.000 0.000 -
As can be seen in Table 6, more than half of the p-values are significant. The main independent variable FORETR has a moderate positive significant correlation in respect to the dependent variable FORACT (0.248). The positive relation contradicts the hypothesis, however the interpretation will be handled in the next section. Furthermore, most of the country
characteristics variables show low significant signs of correlation, except for STRATUP. The control variable SIZE is except for STARTUP not significantly correlated with the other variables.
4.2 Regression results
The regression results are based on a simple linear regression model with a two-tailed confidence interval of 95 percent. In order to correct the regression analysis for the same affiliate reoccurring over the years the assumption is relaxed that the observations are independent. This is done by adjusting the standard error for all the independent variables on the basis of the 585 companies which appear in the dataset. Which means that the regression is based on heteroscedasticity-robust standard error cluster by firm. Furthermore, all the variables are winsorized at the 1st and 99th percentiles to correct for any extreme outliers in the dataset.
Table 7 provides the main empirical findings of this study. If we look at the analysis it shows that the model designed in section 3.3 is significant (p<.001), meaning that the model is predictive for the dependent variable foreign activities. To be more specific the adjusted R-squared of the analysis output indicates that the independent variables combined have a predictive influence on foreign activities of 5.5 percent. In other words, foreign activities are for 94.5 percent explained by other factors than the independent ones tested here. Which is seen as a low but acceptable level of predictability noticing that the model tries to capture an affect that is largely based on private information (Hanlon et al., 2008).
TABLE 7
Regression analysis with FORACT as dependent variable.
Variable Coefficient (Robust) Std. Err. t-statistic p-value
USETR -0.0401 0.0331 -1.21 0.225 FORETR 0.0011 0.0006 1.77 0.078* ATTRACTIVE 0.0417 0.0207 2.01 0.045** FINPOT -0.0015 0.0027 -0.56 0.579 FUTURECO 0.0052 0.0031 1.68 0.093* FREETRADE 0.0080 0.0165 0.49 0.627 STARTUP 0.0020 0.0004 4.50 0.000*** PROTECT 0.0024 0.0013 1.72 0.086* BURDEN 0.0055 0.0063 0.87 0.387 SIZE -0.0216 0.0053 4.04 0.000*** Constant 0.2408 0.2113 1.14 0.255
Observations Adj. R-squared F-statistic p(F) 3061 0.0555 4.32 0.000*** Notes:
* Indicates significance at a level of 10% ** Indicates significance at a level of 5% *** Indicates significance at a level of 1%
The output is based on a simple linear regression analysis with a confidence interval of 95%. For the ease of interpreting the variables ATTRACTIVE, FUTURECO, STATRUP and BURDEN are reverse coded for the regression analysis (higher values indicate that a segment is located in a more attractive country or has lower entry barriers). So a positive coefficient on the country characteristic variables indicates that a segments in a more attractive location positively influence FORACT. All variables are winsorized at the 1st and 99th percentile. The
standard error is adjusted for 585 clusters in company which represents the 585 MNCs which are incorporated in the dataset. Furthermore, the estimated coefficients for the industry dummy and year dummy are not tabulated here, but are positioned in the appendix under Table 9.
The results of main independent variable FORETR show a positive significant
coefficient which contradicts the hypothesis (significance at a level of 5 percent). It indicates that an increase in the foreign effective tax rates leads to more foreign activities. As known a higher FORETR means relatively more tax. This indicates that U.S. multinationals choose to increase their foreign activities even though they have to pay relatively more foreign tax. This suggests that non-tax considerations are more important for multinationals in choosing where to locate their foreign activities and thus that tax avoidance practice do not influence foreign activities of U.S. multinational corporations. In contrast there is no significant evidence on the influence of effective U.S. tax planning expenditures on foreign activities. Which means that no valid conclusion can be attached to the outcome of this independent variable.
In line with these findings the results on the country characteristic variables indicate a significant influence on foreign activities. First the country attractiveness variables will be discussed after which the entry barrier will be dealt with. As explained in section 4.4, the variables ATTRACTIVE and FUTURECO are reverse coded, meaning that a positive coefficient indicates that foreign activities are influenced by countries which score better on these variables. Turning to ATTRACTIVE the output suggests a positive coefficient which is significant at a 5 percent level. This indicates that foreign activities are more likely to be influenced by countries which are more attractive for doing business in accordance with the Forbes business ranking. As for FUTURECO we also find a significant positive coefficient, indicating that foreign activities are influenced by
countries which have a higher credit worthiness and thus are more attractive to conduct business. Finally there is no statistical significant outcome for the variable FINPOT, meaning that no valid conclusion can be attached to the impact of a countries middle class on foreign activities.
The most significant independent variable of the entry barriers is STARTUP, which has a significance of one percent and a positive coefficient. Noting that this variable is reverse coded this implies that, a MNCs foreign activities are influenced by countries which have lower levels of protectionism in the form that it takes a fewer amount of days to startup a new business. Looking at the variable PROTECT we see a significant positive coefficient at a level of 10 percent. Indicating that foreign activities of MNCs are positively influenced by lower levels of protectionism and thus lower entry barriers, where the data is based on executive’s perception of the degree of protectionism. The regression analysis does not provide any statistical significant evidence on the amount of documents which are required to import a shipment of goods or the amount of trade freedom and the influence of these variables on foreign activities.
Overall the empirical model results contradict the hypothesis which presumed that foreign activities are negatively influence by foreign effective tax considerations. This hypothesis was based on the majority of tax avoidance literature which stated that foreign activities have an influence on tax avoidance practices of multinational corporations (e.g., Grubert et al., 1991; Rego, 2003; Clausing, 2009; Klassen et al, 2012; Taylor et al., 2012). So the assumption was when foreign activities positively influence tax avoidance, it would be evident that foreign effective tax rates negatively influence foreign activities. However the results indicate the contrary, foreign effective tax rates positively influence foreign activities. In combination with the results on the country characteristics this lead to conclude that the foreign activities of U.S. multinational corporations are influenced by non-tax considerations like country characteristics and thus, are not induced by tax avoidance practices.
