PROFIT SHIFTING IN DEVELOPING COUNTRIES Esther Helling

(1)

PROFIT SHIFTING IN DEVELOPING COUNTRIES

Esther Helling1

Master’s thesis Economics of Taxation Supervisor: prof. dr. I.J.J. Burgers

Date 10-05-2014

Abstract –Several studies focus on tax-induced profit shifting. These studies however do not take into account that companies may have other reasons to shift profits out of developing economies. This study assesses five variables; the level of corruption, political stability, quality of the rule of law, the corporate statutory tax rate and the degree of investment freedom of a country, that may affect the profit shifting behavior of companies located in developing economies using an OLS regression with cross-section and period fixed effects. The results point out that the level of corruption and the quality of a country’s rule of law play a role in profit shifting behavior of multinational companies.

Keywords: Base Erosion and Profit Shifting; Multinational corporations; Developing economies JEL Classification: H20, H25, F23, K34

(2)

LIST OF ABBREVIATIONS

BEPS Base Erosion and Profit shifting

CFC Controlled Foreign Company

CG Corporate Governance

CSR Corporate Social Responsibility CST Corporate Statutory Tax Rate

ETR Effective Tax rate

FDI Foreign Direct investment

IP Intangible Property

MiDi database Micro Database on Direct Investment database of the Deutsche

Bundesbank, which contains detailed information on German privately and publicly traded foreign direct investment affiliates

MNE Multinational Enterprise

OECD Organisation of Economic Cooperation and Development SPE Special Purpose Entity

UN United Nations

US United States

(4)

(5)

1. INTRODUCTION

The sovereign debt crisis has led to increased attention of governments on tax avoidant behavior of companies. Due to differences in national tax law, gaps and loopholes are created which companies can use to reduce their overall tax bill. Google and Apple for example, were able to reduce their effective tax rate on foreign profits to respectively 3 and 1% (Sullivan, 2012). The OECD (2013) is determined to tackle tax avoidance and profit shifting by companies and has set up an action plan to fight base erosion and profit shifting (BEPS). Developing countries are assumed to suffer more from BEPS activities, due to less developed tax administrations and a lack of a strong tradition of voluntary tax compliance (Kaur and Susarla, 2011). The corporate tax revenue as a percentage of GDP in developing countries is around half of that in OECD countries (House of Commons International Development Committee, 2012). Most studies on BEPS focus on developed countries, as for these countries data on profit shifting variables are available. Empirical evidence of BEPS in developing countries is still limited.

(6)

countries and the information sharing between these countries. Recently, the OECD has extended its focus to reduce harmful tax practices of MNEs (OECD, 2013).

Following the definition of the OECD (2013), BEPS is defined as ‘tax planning strategies that exploit gaps and mismatches in tax rules to make profits ‘disappear’ for tax purposes or to shift profits to locations where there is little or no real activity but the taxes are low resulting in little or no overall corporate tax being paid.’ Hence, BEPS activities take place in companies that have the ability to make use of gaps and mismatches in tax rules. The companies are therefore active in more than one country, and are also known as MNEs. The most common used instruments by MNEs to shift profits are debt-shifting, intra-group licensing of intangible property (IP) and transfer mispricing (Huizinga and Laeven, 2008). Intra-group debt for example opens up the possibility to reduce the overall tax bill by letting an affiliate of a group in a low tax country extend loans to affiliates in high tax countries. The interest is deductible as cost against the high corporate tax rate, and is taxed as income according to the low corporate tax rate. The potential withholding tax on the interest payment can be reduced either directly through an bilateral tax treaty between the countries of the affiliates that reduces withholding tax to 0%, or the affiliates can use a conduit company through which the funds are channeled. The conduit company is located in a country that has favorable tax treaties that enable the MNE to avoid the withholding tax. In a similar way, IP, which is highly mobile, can be easily transferred to a low tax country where subsequently the royalty income is taxed, and companies can under- or overstate transfer prices, due to the usual unavailability of direct comparables for intra-group transactions. The company can reduce its accounting profits in a high tax country if it overstates the transfer price for imports, and understates the transfer prices for export (Huizinga and Laeven, 2008).

(7)

on the firm and welfare of society (Weisbach, 2010) cannot be considered ‘good corporate governance’ and neither corporate social responsible.

However, the profit shifting of companies out of developing countries might not be tax-induced. Some developing countries suffer from a corrupt government, a weak protection of property rights or political instability. For fear of expropriation of their property and income, multinationals can try to reduce the profits they earn in the developing countries using profit shifting techniques (Boyrie et al., 2005). Besides, many developing countries offer multinationals attractive tax arrangements (e.g. a tax holiday) to attract investment and often have a lower corporate statutory tax rate than developed countries (Park et al. 2012). The relatively low level of taxes paid by multinationals in developing countries might therefore not be the result of profit shifting, but rather of the tax system of the developing country. Assuming that companies that pay a low level of taxes are tax avoiders might be wrong.

Profit shifting out of developing countries is a widely debated issue, but empirical evidence on this topic is still scarce. Hence, this study tries to fill part of this void by examining what the main drivers are for profit shifting by affiliates of multinational companies located in 39 developing countries in Asia, Latin America and the Caribbean for the period 2004 to 2012. This study uses Bureau van Dijk’s ORBIS database, which contains information on the balance sheets, profit and loss account and other financial information of both listed and unlisted companies worldwide. This database makes it possible to request the unconsolidated accounts, which enables us to investigate the financial information of individual affiliates of a group of companies. By making use of ordinary least squares regressions, evidence on different drivers potentially influencing profit shifting behavior of companies in developing economies is gathered. So, the research question is: ‘What are the driving factors behind profit shifting by multinational companies in developing economies?’

(8)

(9)

2. LITERATURE REVIEW ON PROFIT SHIFTING

First, a general description of the relevant parts of the international tax system on the taxation of business profits for this study is given (section 2.1). This paragraph is followed by three subparagraphs containing an explanation of the main instruments used to shift profits to other countries and the relevant empirical evidence on these instruments. The three instruments are: Transfer mispricing (section 2.1.1), Debt-usage (section 2.1.2) and Intra-firm licensing and taxation of intangible property (section 2.1.3) The next section (2.2) discusses a second strand of literature that focuses more on the result of profit shifting. Subsequently, the empirical literature on the drivers of profit shifting is discussed in the final section (2.3).

2.1 INTERNATIONAL TAXATION OF BUSINESS PROFITS

This paragraph discusses the relevant parts of the international tax system for the shifting of business profits by companies operating across borders. The following example is used to explain how the parts of the international tax system work that are relevant for this research. A parent company located in country A has one subsidiary that is located in country B. The MNE sells consumer products. The subsidiary manufactures the products and delivers them to the parent company, which is responsible for selling the products.

(10)

the OECD model of 2000 (Burgers et al., 2011). The UN Model tax treaties give developing countries a stronger fiscal position than developed countries by assigning them more taxing rights. Both the OECD Model and the UN Model for tax treaties prescribe two ways to prevent double taxation on foreign income: the tax exemption method and the tax credit method. The first method relies on exempting the foreign income from the taxable base in the home country. The second method taxes worldwide income, but reduces the tax bill in the home country with taxes paid over foreign income; a tax credit is given for taxes paid on foreign income. Another set of countries allows foreign taxes to be deducted from worldwide income. The tax is than seen as being a cost of doing business abroad (Huizinga and Laeven, 2008). Countries can determine in their bilateral tax treaties which method to prevent double taxation is applied.

Thus, double taxation is prevented, but still an acceptable distribution of profits over country A and B has to be arranged. The parent company is deemed to pay an ‘arm’s length’ intra-company price (transfer price) as remuneration for the subsidiary’s manufacturing activities. The arm’s length principle is the international transfer pricing standard that both the OECD and the UN Model Tax Treaties prescribe and has been adopted by almost all countries that have adopted transfer pricing rules. The remuneration paid to the subsidiary is in line with the arm’s length principle if “the conditions of those transactions do not differ from the conditions that would have applied between independent enterprises in comparable transactions carried out under comparable circumstances” (OECD, 2010). The members of a MNE are thus treated as separate entities for tax purposes and a ‘comparability analysis’ is made of the intra-company transactions and transactions that non-related parties would agree upon. Factors identified as potentially important for comparability are inter alia characteristics of the property or services transferred, the contractual terms, the economic circumstances, the business strategies pursued, and the functions performed taking into account the assets used and risks assumed by the parties (1.36 OECD Transfer Pricing Guidelines 2010). On the basis of the comparability analysis the parent and subsidiary determine an arm’s length transfer price. The subsidiary gets assigned part of the revenue, and subtracts its own costs to arrive at the taxable base. The parent reports the total revenue, and subtracts the costs it has made, including the arm’s length remuneration for the activities of the subsidiary, to arrive at his taxable base. The shifting of profits between the parent and subsidiary can however erode the taxable base.

(11)

country, and also themselves levy no WHT for outgoing payments. This gives incentives for MNEs to place a ‘special purpose entity’ (SPE)2_{in this country, through which funds are} channeled from country A to country B without facing any withholding taxes. Country B can be any country that has a favorable tax treatment of repatriated profits, often a tax haven. Weyzig (2013), who has written a dissertation about the adverse effects of the Dutch tax system on the tax revenue of developing countries, finds that on average 7% of all foreign investments in the main developing regions is held through Dutch SPEs in 2005, and that developing regions miss at maximum €640 million, and at minimum €110 million tax revenue due to Dutch SPEs. According to Weyzig, the role tax havens, or preferential tax regimes, play in profit shifting out of developing countries is not negligible. Instead of focusing on the magnitude of the profits shifted out of developing countries by multinationals, the aim of this study is to assess what the driving factors are of this profit shifting. Preferential tax regimes enhance the ability of the MNEs to shift profits, and the factors that drive profit shifting might therefore be more pronounced for companies that have the ability to shift profits to preferential tax regimes.

Finally, it might seem that profit shifting would not be attractive if a country uses a tax credit method, as the total tax bill will be determined on the profits obtained in both the country of the parent and the susidiary, and hence shifting income has no effect on the total tax bill. There are however two factors that might blur the distinction between the tax credit and tax exemption system (de Mooij and Ederveen, 2003). The first of these factors is deferral: often the taxes levied on the foreign income are delayed until the foreign earnings are repatriated as a dividend. The second factor is cross-crediting: the excess credit on one foreign subsidiary (A) can be offset against tax payments in the MNE’s home country on income from another foreign subsidiary (B). Excess credit exists when the foreign subsidiary (A) pays more tax on its income in the foreign country, than would have to be paid in the home country of the MNE on this income. If another foreign subsidiary (B) has paid less tax on its income in the foreign country than what would have to be paid in the home country of the MNE, the tax payable on the foreign income of the subsidiary (B) in the MNE’s home country can be reduced by the excess credit of foreign subsidiary (A). Markle (2010) examines whether differences in taxing foreign income, either on the basis of a tax exemption or tax credit method, is related to differences in tax-motivated income shifting behavior of MNEs. He finds that MNEs subject to tax exemption systems shift more income than MNEs subject to tax credit systems, but this is only the case if the MNE cannot

2_{The OECD defines SPEs as: ‘entities with no or few employees, little or no physical presence in the host}

economy, whose assets and liabilities represent investments in or from other countries, and whose core business consists of group financing or holding activities’. Examples are financing subsidiaries, conduits,

(12)

defer repatriation of the shifted income in the tax credit system. To assess the profit shifting behavior of MNEs, the treatment of foreign income by the home country of the parent of a subsidiary located in a developing country does not play an important role.

2.1.1 PROFIT SHIFTING BY MAKING USE OF TRANSFER MISPRICING

Transfer mispricing occurs if the specified price for an intra-company transaction is not in line with the functions performed, the assets used and risks assumed of the respective parties to the transaction. As it is often difficult to find a comparable transaction, companies have some freedom in setting transfer prices. An incentive to misprice transactions is present when there are differences in effective tax rates (ETR) between countries. The focus is on the ETR, as it is the ratio of the actual taxes paid to a measure of pre-tax profit. The CST on the other hand is a ‘headline ratio’, which is applied to a taxable base (OECD, 2013). Although the CST might be the same between countries, the ETR can diverge due to a different definition of the taxable base. MNEs have an incentive to report the most profits in a country with a low ETR. Note that if there would be no differences in tax laws, there is no tax-induced incentive to shift profits.

Assuming that country A and B have the same taxable base, it is in the interest of the MNE to overstate the transfer price for the products of the subsidiary. If the parent company in country A pays a higher price for the products, the profits of the parent company will be reduced. The subsidiary on the other hand receives a higher price for its products and will report higher profits. This transfer mispricing is considered to be one of the instruments MNEs use to erode the tax base and shift profits between countries.

(13)

The mispricing approach has however a few shortcomings which make the results difficult to draw reliable conclusions on (Fuest and Riedel, 2010). First, price differences with a group of products might reflect quality differences. Products imported out of developed economies are likely to be of higher quality, and hence have a higher price. If the products are assumed to be homogenous, the mispricing approach leads to overestimation of profit shifting due to transfer mispricing. This can be corrected by making use of micro-level import data. Another shortcoming is that mispricing studies assume that the lowest and highest quintiles of prices for a product are under- and overpriced, while this does not have to be the case. There are several reasons why prices for a certain product group diverge, and these do not have to be profit shifting reasons. A third shortcoming of mispricing studies is that there is no clear counterfactual, which is the hypothetical situation where no mispricing occurs, making it hard to determine whether certain transactions are mispriced. A final remark on the mispricing approach is that many studies based on this approach only report the underpriced imports and overpriced exports of a country, and not the overpriced imports and underpriced exports (see for example Christian Aid, 2009).

2.1.2 PROFIT SHIFTING BY MAKING USE OF DEBT

(14)

The empirical evidence on the use of debt for profit shifting by MNEs in developing countries is scarce. There is however ample evidence of the use of debt for profit shifting (see for example Desai, Foley and Hines 2004; Buettner et al., 2006; and Buettner and Wamser, 2009). A study by Fuest, Hebous and Riedel (2011) explicitly distinguishes between developed and developing countries using the Micro Database on Direct Investment (MiDi) database of the Deutsche Bundesbank, which contains detailed information on German privately and publicly traded foreign direct investment affiliates. The study assesses whether international profit shifting through debt usage differs for developing and developed countries. They find that affiliates located in high tax countries favor debt financing. However, the effect of the corporate tax rate on the level of intra-company debt financing is twice as strong for affiliates located in developing countries than for the affiliates located in developed countries. Their result confirms that developing countries might be more sensitive to profit shifting through debt usage by MNEs. Developed countries might also have more (effective) rules on limiting interest deductibility on debt, reducing the attractiveness to use debt for profit shifting.

2.1.3 PROFIT SHIFTING BY MAKING USE OF INTRA-FIRM LICENSING AND LOCATION

OF INTANGIBLE PROPERTY

MNEs can make use of the location of intangible property (IP), and the remunerations paid for the use of the IP by the affiliates of the MNE, to shift profits from one country to another. IP is highly mobile, and consists inter alia of patents, trademarks, customer lists and copyrights. IP is often identified as the key value driver of a company, and is therefore a major determinant of firm value (Edmans, 2007). If one of the group’s affiliates owns all the IP, and the other related affiliates of the group make use of the IP, they have to pay a royalty to the owner of the IP. A royalty is a payment for the right to use the IP, and is based on the arm’s length principle.

Generally the arm’s length remuneration for IP is determined as the royalty third parties would pay for use of the IP. The arm’s length remuneration for use of IP, and the value of the IP itself, is often hard to determine, especially for outsiders such as the tax authorities. The valuation difficulties and the high mobility of the IP make it an attractive profit shifting instrument. A MNE has an incentive to locate IP in a country that taxes the income of the IP at a low tax rate. Even if the IP is first located in a high-tax country, it might be attractive for the MNE to transfer the IP to a low tax country. Although most countries charge an ‘exit tax’ when the IP is located out of the country, MNEs are assumed to choose a tax-optimized transfer price for the IP that minimizes the exit tax (Dischinger and Riedel, 2011).

(15)

(CCA), which third party relations may also agree upon (8.3 OECD TP Guidelines, 2010). The costs are shared proportional on the anticipated benefits of the exploitation of the intangible for each of the parties individually. After development, each of the parties to the CCA own part of the IP, and hence no royalty payments have to be made for use of the IP. Often one of the parties contributes existing IP to the CCA, and the other party is deemed to make an arm’s length ‘buy-in’ payment for this IP. If an affiliate of a MNE in a high tax country is owner of the IP and enters into a CCA with an affiliate located in a low tax jurisdiction, part of the profits derived from the IP are allocated to the low tax affiliate. By entering into a CCA early in the development process of the IP, there is still a lot of uncertainty about the value of the IP, and thus about the correct buy-in payment. If the buy-in payment is set at a too low level, profits are shifted from the high to low tax affiliate.

The empirical evidence on the use of IP as profit shifting instrument is scarce, and especially for developing economies due to data unavailability. To the best of my knowledge, there are no studies yet that focus on income shifting out of developing economies by making use of IP. A study by Grubert (2003) focuses on Controlled Foreign Companies (CFCs) of US MNEs. He finds that half of the income shifted from high to low tax countries is due to R&D intangible related income. CFCs that are R&D intensive, or are located in either a high or a low tax country, have a stronger incentive to engage in profit shifting, and undertake more inter-company transactions. The R&D intensive US parents are more likely to invest in either very high tax, or very low tax countries. Grubert’s last finding is that debt usage and R&D based IP income together account for the observed difference in profitability between high and low tax countries. Desai et al. (2006) finds that R&D intensive MNEs are most likely to invest in tax havens, to which the profits can be shifted. Dischinger and Riedel (2011) focus on IP holdings of the affiliates of European MNEs, and find that affiliates of European MNEs located in a country with a relatively lower tax rate than affiliates in other locations have a higher share of IP holdings. These studies underline that IP plays an important role in profit shifting.

2.2 EMPIRICAL EVIDENCE ON PROFIT SHIFTING: STATE OF THE ART

This section focuses on the empirical evidence of the result of profit shifting, instead of focusing on the specific instruments used to shift profits. The information on profit shifting instruments is often unavailable, and this is especially so for developing economies. Hence, focusing on the result of profit shifting is considered to be a more feasible approach for developing countries.

(16)

world, estimates that tax revenue losses due to corporate profit shifting in developing countries amounts to US$50 billion per year. US$ 15 billion of this number is lost revenue due to evasion of income from financial assets held abroad. The other US$ 35 billion is derived from the missing tax revenue on the return on foreign direct investment (FDI). Fuest and Riedel (2010) call into question this estimation, and note that both the assumed average tax rate as the assumed return on FDI in the calculation are too high, potential tax incentives offered to companies are not taken into account, and the return on FDI is assumed to be the same as the taxable base, which does not have to be true. Fuest and Riedel (2010) further emphasize that this study does not give any insight in the drivers of profit shifting out of developing countries, which do not have to be solely tax related.

Another estimate of profit shifting out of developing countries is done by Baker (2005), who estimates that developing countries loose around US$ 500 to 800 billion yearly due to illicit financial flows. He bases this number on 550 interviews he conducted with officials from trading companies in 11 countries. 40% of this number is due to criminal activity, and the other 60% due to legal commercial activities. Baker considers three potential channels through which money is leaving developing countries; transfer mispricing, mispricing of goods traded between unrelated parties, and non-existing transactions. The Global Financial Integrity Fund (GFI, 2010) applies to these illicit financial flows the relevant country corporate tax rates, and estimates that the average tax revenue loss due to profit shifting in developing countries is US$98 billion per year between 2002 and 2006.

(17)

companies, and MNEs that have a connection to a tax haven report even less profits and less taxes. A study by Christian Aid (2013) on profit shifting in India supports this result; MNEs with a link to a tax haven are found to engage in more profit shifting than MNEs without a link to a tax haven.

2.3 EMPIRICAL EVIDENCE ON DRIVERS OF PROFIT SHIFTING

(18)

3. RESEARCH QUESTION AND HYPOTHESES DEVELOPMENT

As the literature on the drivers of profit shifting of multinationals out of developing economies is scarce, this study aims to fill this void by answering the following research question:

‘What are the driving factors behind profit shifting by multinational companies in developing economies?’

The following variables are taken into account as potential variables affecting profit-shifting behavior of companies in a country, resulting in five hypotheses.

Corporate Statutory Tax rate –Companies use the effective average tax rate to decide where to locate their activities, and subsequently the effective marginal tax rate to decide how much to invest in the activities conditional on the location. However, companies determine the level of profits they want to report in a country based on the corporate statutory tax rate (Dischinger and Riedel, 2007), and is therefore considered to be the appropriate tax rate for this research. An increase in a country’s corporate statutory tax rate incentives companies that have the ability to shift profit to lower the profits reported in this country and shift the profits to a country with a lower tax rate, to reduce their absolute tax bill. Hence, we expect that companies that have the ability to shift profits, thus multinational companies, shift more profits out of the developing country after an increase in the corporate statutory tax rates relative to companies that do not have the ability to shift profits, which are defined as domestic companies.

H1: An increase in a country’s corporate statutory tax rate is related to an increase in profits shifted out of the country by multinational companies.

Corruption index –I postulate that a decrease in a country’s level of corruption is associated with a decrease in profit shifting out of the developing economy. Corruption lowers the trust of companies in the government, agencies and the political agenda, leading to an unsecure environment for companies in which they want to report relatively less profits.

H2: A decrease in a country’s level of corruption is related to a decrease in profits shifted out of the country by multinational companies.

(19)

of multinational companies to shift profits out of a country. In comparison to companies that do not have the ability to shift profits, I expect that multinational companies report relatively less profits after an increase in the investment freedom index than domestic companies.

H3: An increase in a country’s degree of investment freedom is related to an increase in profits shifted out of the country by multinational companies.

Rule of Law – The Rule of law index reflects perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence (Kaufmann et al., 2008). An increase in the rule of law index increases the level of protection of a company’s earnings and assets, and therefore, reduces the incentive of companies to shift profits out of the developing economy. Hence, companies that have the ability to shift profits will report relatively more profits after an increase in quality of the rule of law than companies that do not have the ability to shift profits.

H4: An increase in the quality of a country’s rule of law is related to a decrease in profits shifted out of the country by multinational companies.

Political stability - A company that has the ability to shift profits might be less inclined to report profits in a country that has a high degree of political instability, due to uncertainty about the political agenda. Relative to companies that do not have the ability to shift profits, a company that has the ability to shift profits is postulated to report relatively less profits after a decrease in a country’s political stability, and thus shifts more profits out of the developing country.

H5: An increase in the level of a country’s political stability is related to a decrease in profits shifted out of the country by multinational companies.

(20)

4. METHODOLOGY

This study follows partly the methodology proposed by Fuest and Riedel (2010), and which has been applied to India by Christian Aid (2013). Fuest and Riedel (2010) argue that if MNEs shift profit, this will show up in a different level of reported pre-tax profit in comparison with the situation in which MNEs do not shift profits. MNEs shifting profits out of a country, will report a lower level of pre-tax profit per unit of assets, and will pay lower taxes per unit of assets or per unit of profit. It is expected that profit-shifting MNEs hold higher level of intra-firm debt, lower levels of IP, and have stronger distortions of intra-firm prices. MNEs with a tax haven connection, or a preferential tax regime, are assumed to have even stronger incentives for profit shifting, because the tax haven allows the shifted income to be taxed at a very low rate, or not at all (Fuest and Riedel, 2010).

The perfect setup is to compare a group of companies that shift profits to a similar group of companies that does not shift profits. The two groups only differ in profit shifting, and any change in the outcome variables is an effect of profit shifting. However, this setup is not feasible, as it is not possible to find two similar groups that only differ in their profit shifting behavior. Another problem is that we cannot identify which companies shift profits. Hence, a different setup is needed. Instead of focusing on which companies shift profits, the focus of this study is on those companies that have the opportunity to shift profits across borders and those companies that do not have this opportunity. The former group consists of affiliates of multinational companies, and the latter of domestic companies. The affiliates of MNEs are identified as companies that either have a parent or a subsidiary abroad. The control group consists of domestic companies that do not have a foreign parent or subsidiary and thus do not have the opportunity to shift profits3_{to another country. The affiliates of MNEs are divided in} two groups: a group of affiliates of MNEs with a tax haven connection, and a group affiliates of MNEs without a tax haven connection. An affiliate of a MNE has a tax haven connection if at least one of the subsidiaries is located in a tax haven, or if it has a corporate owner located in a tax haven.

If we would compare the absolute values of the profit shifting variable for these three groups, a potential ‘selection bias’ is present (Fuest and Riedel, 2010): the MNEs might diverge on several important characteristics from national companies, which can lead to differences in the outcomes. For example, MNEs have the opportunity to structure their supply chain around

(21)

several countries. If we compare domestic manufacturing companies to MNE manufacturing affiliates, we may find that the profitability ratios of the two groups differ. MNEs may however choose to only locate low risk manufacturing subsidiaries in developing countries. The domestic manufacturing companies will perform all related functions themselves, including inherent risks, and hence, their profitability ratio may diverge of the profitability levels of the MNE manufacturing affiliates. Comparing these two groups without accounting for the bias leads to wrong conclusions. Differences between the profit shifting variables for affiliates of the MNEs and domestic companies may thus be driven by unobserved and observed heterogeneity between the two groups. To control for this, I estimate an ordinary least squares (OLS) regression with cross-section and time fixed effects45_{. Heterogeneity across firms that is} constant over time is controlled for by cross-section fixed effects, and time fixed effects are included to control for heterogeneity that differs per time period, but is constant across cross-sections. Additionally, I control for heteroskedasticity clustered at the cross-section by including robust standard errors. One assumption of OLS is that the error term has constant variance. If this assumption is violated the coefficient estimates are still unbiased, but the standard deviations of the coefficients are wrong, leading to wrong statistical inferences (Brooks, 2008).

The sample is further restricted by excluding observations years in which the profit shifting variable for a company takes a negative value as companies may behave differently when making a loss instead of a profit, and the variables are winsorized at the 0.5% and 99.5th percentile to control for potential outliers. The final restriction on the sample is that only companies with at least a value of US$ 500,000 for total assets in one of the years in the sample are included.

The following formula is estimated:

Where stands for a specific firm, stands for a country, and for time. The dependent variable

( ) is the profit-shifting variable: the level of profits before taxes per 100 units of assets. As

robustness test the taxes paid per 100 units of assets and the taxes paid per 100 units of profits are also taken as profit shifting variable. is a set of country-level variables that may explain

4_{Another strategy is to use ‘propensity score matching’ (see e.g. Egger et al. (2009a) and Egger et al.} (2009b)). This approach is also proposed by Fuest and Riedel (2010).

(22)

profit-shifting of companies. is a dummy variable which takes 1 for MNE’s affiliates, and 0 for domestic companies. The dummy is interacted with the country level variables, which makes our coefficient of interest. and are respectively a set of time and cross-section fixed effects. The value and significance for indicates if the reaction of a company that has the ability to shift profits, an affiliate of a MNE, to a certain change in one of the is different from companies that do not have the ability to shift profits, domestic companies. Most companies in the sample are located in China, and therefore, the same regression estimations are run excluding China from the sample to check whether results are driven by a potential selection bias.

(23)

5. DATA

This study makes use of the ORBIS database to gather information on company’s unconsolidated accounts, both the balance sheet and profit and loss accounts. The country level control variables are derived from the Worldbank’s statistical database. Data on countries’ corporate statutory tax rate are gathered using the Ernst & Young’s Worldwide Corporate Tax Guides, KPMG’s Tax tools and resources, and Deloitte’s International Tax Source. The corruption index, the political stability and the rule of law index are all Worldwide Governance Indicators, and range from -2.5 (low level) to 2.5 (high level) and are derived from Kaufmann et al. (2008). The corruption index reflects the perception of the extent of the influence of private interests on public power, including petty and grand forms of corruption, and of the degree of capture of the state, which indicates the efforts of companies to shape the institutional environment in which they operate to their advantage. The rule of law index reflects perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. Political stability captures the perception of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including domestic violence and terrorism. Finally, the degree of investments freedom measures the ease of in- and outflows of capital in a country and is derived from the Heritage Foundation (2010). The index ranges from 0, no freedom of investments, to 100, high freedom of investments.

The focus is on companies located in developing economies in Asia, Latin America and the Caribbean, for the period 2004 to 2012. The data coverage for companies located in these regions is reasonable (Fuest and Riedel, 2010). I exclude companies located in Newly Industrialized countries, and companies located in countries of which the economy is based on fuel (or natural resources), as these countries often have a different tax structure. From the remaining set of countries, I remove the alleged tax havens. There is no consensus on the criteria used to identify a tax haven or on the lists of tax havens (Weyzig, 2013). For example, the OECD Black List only focuses on countries that do not apply to the transparency and information exchange rules for tax purposes. However, this does not take into account those countries that let companies use their preferential tax regime without demanding substantial economic presence. For that reason, tax haven is broadly defined in this study: a jurisdiction is a tax haven if it is on three or more of the 11 lists mentioned in Murphy’s paper (2009)6. This leaves a sample of 39 countries7.

(24)

The companies in the sample are divided in the following three groups:

1. Domestic Companies: companies that belong to a domestic group with no affiliates abroad (>50% ownership)8_{, or a parent abroad (>50% ownership). There are 3,384} domestic companies.

2. Affiliates of MNE that do not have a tax haven connection: Companies that belong to a group that has at least one foreign subsidiary (>50% ownership), or has a parent company abroad (>50%). The group of affiliates of MNEs that do not have a tax haven connection includes 1,461 affiliates of multinational companies.

3. Affiliates of MNEs with a tax haven connection: Companies that belong to a group that has at least one foreign subsidiary (>50% ownership) located in a tax haven, or a parent company located in a tax haven (>50% ownership). 258 Companies are identified as affiliates of a MNE with a tax haven connection.

Appendix A includes the search strategy in ORBIS. The number of affiliates that are part of a group with a link to a tax haven is relatively small. Hence, I will also combine the two different MNE groups in one group and run the same analysis.

Zoromé 2007 academic paper for the IMF, Senator Carl Levin 2007 for the Stop Tax Haven Abuse Act in the USA, and the Lowtax.Net (accessed 22-1-08) web site promoting secrecy jurisdictions

7_C_{ountries in the sample: Argentina, Bangladesh, Bhutan, Bolivia, Brazil, Cambodia, Chile, China,} Colombia, Dominican Republic, El Salvador, Fiji, Guatemala, Guyana, Haiti, Honduras, India, Indonesia, Jamaica, Kiribati, Lao People's Democratic Republic, Maldives, Mexico, Myanmar/Burma, Nepal, Nicaragua, Pakistan, Papua New Guinea, Paraguay, Peru, Philippines, Solomon Islands, Sri Lanka, Suriname, Thailand, Tuvalu, Uruguay, Venezuela, and Vietnam

8_{The ORBIS database only contains ownership data for the last available year, leading to potential}

(25)

6. RESULTS

In this section (6.1), I first discuss the descriptive statistics of the variables used in this study. In the next section (6.2), the results are presented for the regression model in which the sample of companies is split up in two groups: a group of domestic companies and a group of affiliates of MNEs. In the final section (6.3), I divide the MNE group in two: one group of affiliates of MNEs that have no tax haven link, and one group of affiliates of MNEs that do have a link to a tax haven, and run the same analysis. For reasons of brevity, the profits before taxes per unit of asset are also referred to as the profits per unit of asset, or the level of profitability.

6.1 DESCRIPTIVE STATISTICS

Table 1 contains an overview of the average values for several variables for the three groups individually, respectively, the group with domestic companies, the group with affiliates of MNEs that do not have a link to a tax haven, and the affiliates of MNEs with a tax haven link.

Table 1. Mean of profit shifting variables and instruments for the three groups of companies.

Domestic company MNE No tax haven MNE tax haven Debt per unit of asset 0.612 0.594 0.574

Leverage 2.070 1.866 1.643

Intangible assets per unit of asset 0.037 0.041 0.045 Profit before tax per unit of asset 0.085 0.101 0.098 Taxes paid per unit of profit 0.293 0.305 0.249 Taxes paid per unit of asset 0.018 0.022 0.019

Domestic companies have a higher level of debt (0.612) per unit of asset than the MNE groups (0.594 and 0.574), although the difference is relatively small. Note that leverage in the table is debt per unit of equity, and can be interpreted the same way as debt per unit of asset. For the other profit shifting instrument, intangible assets, we see that on average MNEs, and especially MNEs with a link to a tax haven, hold the highest level of intangible assets per unit of asset (4.1% and 4.5%). The level of intangible assets per unit of asset is the lowest for domestic companies (3.7%). MNEs are generally more intangible intensive firms than domestic companies. The higher level of intangible assets therefore does not indicate that MNEs use the intangible assets for profit shifting.

(26)

shelters, which Lin, Tong and Tucker (2013) find to be inversely related to leverage: a company that increases its degree of tax aggressiveness in the form of the use of tax shelters decreases its leverage. The domestic company is limited in the location of its activities, and would therefore, instead of making use of aggressive tax planning strategies, make more use of debt to benefit from the interest debt tax shield to effectively reduce its tax bill.

Besides the profit shifting instrument, the means of the profit shifting variables are given in table 1. Affiliates of MNEs have a respectively higher level of profit per unit of asset (10.1% for the MNE group without a tax haven link, and 9.8% for the MNE group with tax haven link) than the group of domestic companies (8.5%). This higher level of profitability for affiliates of MNEs can be explained by the productivity of companies – MNEs are found to be more productive than domestic companies, which results in higher profitability levels for MNEs (Maffini and Mokkas, 2011). If we look at the taxes paid per company group, we see that domestic companies and affiliats of MNEs without a tax haven link pay the most per unit of profit (respectively, 29.3% and 30.5%), and the affiliates of MNEs that have a tax haven link pay the least (24.9%). Only looking at the taxes paid per level of profit does not take into account that companies may reduce the level of profit to lower their tax bill. Although the taxes paid per unit of profit might look acceptable, the profit level might have been changed in order to reduce the absolute tax bill. Hence, the tax paid per unit of asset is a better indicator. Affiliates of MNEs without a tax haven link pay the most taxes per unit of asset (2.2%), and domestic companies and affiliates of MNEs with a tax haven link the least (respectively 1.9% and 1.8%). A higher level of profitability is normally accompanied by a higher level of taxes paid. It is therefore not surprising that the group of MNEs without a tax haven link has both the highest profitability level and the highest level of taxes paid per unit of asset, and the domestic group the lowest profitability and lowest level of taxes paid per unit of asset.

Note that these variables do not show whether MNEs engage in profit shifting behavior. They are merely descriptive.

(27)

Table 2. Descriptive statistics of independent variables.

Variables Mean Median Maximum Minimum Std. Dev. Observations Corruption -0.494 -0.534 1.562 -1.729 0.612 351 CST 0.288 0.300 0.450 0.000 0.070 360 Political stability -0.482 -0.500 1.480 -2.812 0.863 351 Rule of Law -0.511 -0.612 1.367 -1.787 0.616 351 Investment Freedom 42.319 40.000 90.000 0.000 19.691 317 LOG(GDP per capita) 3.256 3.194 3.975 2.499 0.371 328 GDP growth rate 3.592 3.720 17.513 -10.231 3.688 334 LOG(Population) 7.060 7.124 9.131 3.984 1.063 351

6.2 RESULTS FOR DOMESTIC COMPANIES AND MNES

This section elaborates on the results of the differences in reactions to changes in the explanatory variables between domestic companies and affiliates of MNEs. First, I present and discuss the results for the regression model that has as independent variable the profits per 100 units of asset (Table 3). Subsequently, I show and interpret the results for the regression models that have as dependent variable either the taxes paid per 100 units of asset, or the taxes paid per 100 units of profit (Table 4). The results for the control variables are included in appendix B. Regression model (1) in Table 3 (located on the following page) – which has as dependent variable the profits per 100 units of asset - has cross-section and period fixed effects. In regression specification (2) there is additionally controlled for heteroskedasticity at the cross-section, and in regression specification (3), the sample is reduced by excluding companies located in China from the sample.

(28)

Table 3. Regression analysis for two groups of companies, dependent variable: profits per 100 units of

asset.

This table contains the results of the model that splits up the companies in domestic companies and affiliates of MNEs. The results show whether the MNE group reacts differently on explanatory variables than the group of domestic companies. The dependent variable is the profits before tax per 100 units of asset. Regression (1) includes cross-section and period fixed effects, and regression (2) additionally controls for heteroskedasticity by clustering standard errors at the cross-section. In regression (3) the sample is limited by excluding companies located in China. Standard errors are in parenthesis. * indicates a significance level of 10%, ** a significance level of 5%, and *** a significance level of 1%.

The coefficients on the investment freedom index have a negative sign for domestic companies in the different regression specifications. Hence, an increase in the level of investment freedom, or openness to trade of a country, is associated with a decrease in the profits reported per 100 units of asset by domestic companies. This coefficient is not significant however. Neither are the coefficients on the interaction terms of MNE status and investment freedom. In this study, investment freedom therefore does not play a role in determining the level of profit shifting by MNEs.

For domestic companies, a significant positive relation between the corruption index (which ranges from -2.5 high corruption, to 2.5 low corruption) and the profitability level is present: a

Regression (1) (2) (3)

Dependent variable Profits per 100 units of asset

Countries All All All, except China

(29)

1% increase in the corruption index is associated with an increase of 10.477 in the profits reported per 100 units of asset by domestic companies. Affiliates of MNEs report (10.477 – 8.767) 1.710 more profits per 100 units of asset after a decrease in the level of corruption. The increase in the profits reported per 100 units of asset for an affiliate of a MNE is 8.767 profits less per 100 units of assets than the profits reported by a domestic company if the corruption index increases by 1%. This is attributed to the ability of the MNE to shift profits, and thus profits are shifted out of the developing country if corruption index increases.

The coefficient on political stability is significant for domestic companies in regression model (1), who report 1.817 more profits per 100 units of asset if the political stability index increases by one. This relation does not hold in regression model (2) and in regression model (3). None of the coefficients on the interaction terms is significant, although they are all negative. The role of political stability in profit shifting by MNEs is therefore considered to be limited.

The interaction term on the rule of law index and MNE status is significantly positive (9.884). The decrease in profits reported per 100 units of asset for domestic companies are -10.381, and for affiliates of MNEs (-10.381 +9.884) -0.497. The negative reaction of domestic companies to an increase in the rule of law index is of greater magnitude than the reaction of affiliates of MNEs. With respect to the increase in the profits reported per 100 units of asset by domestic companies after a 1% increase in a country’s rule of law index, affiliates of MNEs report 9.884 profits more per 100 units of asset. Affiliates of MNEs are therefore considered to shift less profits out of the developing country if the rule of law index increases, as the decrease in the level of profits reported per 100 units of assets is smaller for affiliates of MNEs than for domestic companies.

The results for regression model (3), which excludes companies located in China from the sample to control for a potential selection bias, point out that there are no significant differences between the sample of companies that includes the companies located in China, and the sample of companies that does not include companies located in China.

(30)

variable, there should also be a positive relation between the taxes paid per unit of asset and this explanatory variable.

In table 4, the dependent variable is either the taxes paid per 100 units of asset (regression 1 and 3), or the taxes paid per 100 units of profit (regression 2 and 4). In regression (3) and (4) the companies located in China are excluded from the sample to check for potential selection biases.

Table 4. Regression Analysis with two groups of companies, dependent variable: taxes paid per 100 units

of asset and the taxes paid per 100 units of profit.

Regression (1) (2) (3) (4)

Dependent variable

Taxes paid per 100 units of asset

Taxes paid per 100 units of profit

Taxes paid per 100 units of asset

Taxes paid per 100 units of profit Countries Sample All All All, except China All, except China

(31)

The results for regression model (1) show that domestic companies pay 0.881 more taxes per 100 units of asset if the political stability index increases by one. A significant negative relation exists between the taxes paid per 100 units of asset and the quality of a country’s rule of law for domestic companies (coefficient is -3.658). The relations for political stability and the rule of law index with the taxes paid per unit of asset hold if China is excluded from the sample in regression model (3). The results for regression model (3) further indicate that there is a positive relation between the corruption index and the taxes paid per unit of asset for domestic companies – a domestic companies pays 4.837 more taxes per 100 units of asset if the corruption index increases by 1%. None of the coefficients on the interaction terms is significant for regression model (1) and regression model (3), indicating that MNEs do not pay more or less taxes per unit of asset than domestic companies after a change in one of the explanatory variables.

The results for regression model (2), where the dependent variable is the taxes paid per 100 units of profit, show that the independent variables are not assigned any explanatory power. The coefficient on the tax rate is positive (22.027), but not significant. An increase in the tax rate might therefore be positively related to the taxes paid per unit of profit, as would be expected if a country raises its tax rate, all other things being equal. There is however too much variation in the sample to single this effect out. On overall, the findings indicate that companies report less profits per unit of asset after a change in the tax rate, but do not significantly pay less taxes per unit of asset, or less taxes per unit of profit. Although companies report less profits per unit of asset after an increase in the corporate statutory tax rate, they pay relatively the same amount of taxes as before the raise in the tax rate.

The results for regression (4), in which companies located in China are dropped from the sample, indicate that for the taxes paid per unit of profit, affiliates of MNEs react differently to changes in the corruption index and changes in political stability than domestic companies, although the latter is only significant at the 10% level. A company that has the ability to shift profits pays relatively more taxes per unit of profit than a domestic company if the corruption index or the level of political stability increases, but does not pay more taxes per unit of asset. The standard deviations on the explanatory variables are all quite high; probably the explanatory power of the variables is low.

6.3 RESULTS FOR DOMESTIC COMPANIES, MNES WITH AND WITHOUT A TAX

HAVEN LINK

(32)

changes in the explanatory variables than domestic companies, and that this reaction is more pronounced for affiliates of MNEs that have a link to a tax haven. First, I present and discuss the results of the regression model that has as dependent variable the profits per 100 units of asset (Table 5). Subsequently, I show and interpret the results for the regression model that has as dependent variable either the taxes paid per 100 units of asset, or the taxes paid per 100 units of profit (Table 6).

Regression specification (1) of Table 5 (located on the next page) has cross-section and period fixed effects. Regression (2) additionally controls for heteroskedasticity clustered at the cross-section. In regression (3) companies located in China are excluded from the sample.

The results of Table 5 show that, for domestic companies, there are significant relations between the profits per unit of asset and the tax rate, the corruption index, and the rule of law. The relation between the tax rate and the profit shifting variable is only significantly different for affiliates of MNEs with a link to a tax haven in regression (1). Affiliates of MNEs with a tax haven link report 22.938 profits more per unit of asset after a change in the tax rate than domestic companies. However, after the inclusion of robust standard errors, this relation is no longer significant (see regression (2)). On overall, we see that the coefficient for the tax rate is positive for the interaction terms in regression (1) and (2), indicating that the reaction of affiliates of MNEs to changes in the tax rate might be less negative than the reaction of domestic companies. However, after exclusion of the companies located in China in regression (3), the sign of the coefficient is reversed for both the affiliates of MNEs with and without a tax haven connection, pointing out that affiliates of MNEs might report relatively less profit per 100 units of asset than domestic companies if the tax rate increases 1%. These coefficients are however not significant.

(33)

Table 5. Regression Analysis with three groups of companies, dependent variable: profits before taxes per 100

units of asset.

Regression (1) (2) (3)

Dependent variable Profits before taxes per 100 unit of asset

Companies in sample All All All, except China

CSTCSCST -18.491 -18.491 -18.505 (5.771)*** (7.575)** (13.015) Investment freedom -0.034 -0.034 -0.030 (0.021) (0.029) (0.030) Corruption 10.479 10.479 12.722 (2.824)*** (3.671)*** (3.973)*** Political stability 1.816 1.816 1.955 (0.809)** (1.206) (1.214) Rule of Law -10.369 -10.369 -12.153 (2.442)*** (3.658)*** (3.970)*** CST*MNE No tax haven 6.183 6.183 -6.559

(6.027) (7.258) (20.971) CST*MNE Tax haven 22.938 22.938 -0.310

(12.172)* (15.211) (43.492) Corruption*MNE No tax haven -8.849 -8.849 -8.143

(4.191)** (4.934)* (5.466) Corruption*MNE Tax haven -7.658 -7.658 -13.730 (9.115) (7.617) (6.756)** Investment Freedom*MNE No tax

haven (0.031) 0.041 (0.042) 0.041 (0.045) 0.020 Investment Freedom*MNE Tax haven -0.017 -0.017 -0.027

(0.070) (0.072) (0.073) Political Stability*MNE No tax haven -0.753 -0.753 -1.187

(1.180) (1.560) (1.603) Political Stability* MNE Tax haven -0.671 -0.671 -2.697

(2.459) (2.794) (2.638) Rule of Law*MNE No tax haven 9.597 9.597 8.935

(3.489)*** (4.682)** (5.212)* Rule of Law* MNE Tax haven 11.764 11.764 20.873

(8.256) (10.308) (10.891)*

R-squared 0.712 (0.712) (0.727)

(34)

The coefficient for the interaction term between the corruption index and affiliates of MNEs with no tax haven connection is significantly negative (-8.849). Domestic companies report 10.479 profits per 100 units of asset more if a country’s corruption index increases by 1%, and affiliates of MNEs without a tax haven link report (10.479 - 8.849) 1.630 profits per 100 units of assets more. Hence, in comparison to domestic companies, the profitability levels of affiliates of MNEs without a tax haven link are less sensitive to changes in the level of corruption. I do not find that this effect is significant for affiliates of MNEs with a tax haven link in regression (1) and (2), although the coefficient is negative as well. In regression (3) the interaction term for MNE status with a tax haven link is significant (coefficient on the interaction term is -13.730), pointing out that if a country’s level of corruption decreases, affiliates of MNEs with a tax haven link report relatively less profits per unit of asset than domestic companies.

Affiliates of MNEs without a tax haven link report a higher level of profits with respect to domestic companies after a change in the rule of law index. This result holds for all regressions. The coefficient on the interaction term between the rule of law index and MNEs with a tax haven link is only significant in regression (3), with a value of respectively 20.873. This is twice the value as for the MNE without a tax haven link (8.935). Hence, in regression (3) MNEs with tax haven link react, as expected, stronger to changes in the explanatory variables than MNEs without a tax haven link. However, the standard deviation for this variable is very high, and I find no significant results for MNEs with a tax haven link in regression (1) and (2). My overall conclusion is therefore that MNEs, both with and without a tax haven connection, report relatively more profits per 100 units of assets after an increase in the rule of law index than domestic companies.

(35)

Table 6. Regression analysis with three groups of companies, dependent variable: taxes paid per 100 units of asset, or

taxes paid per 100 units of profit.

Regression (1) (2) (3) (4)

Dependent variable Taxes paid per 100 _{units of asset} Taxes paid per 100 _{units of profit} Taxes paid per 100 _{units of asset} Taxes paid per 100 _{units of profit} Countries in Sample All All All, except China All, except China

CST -2.769 21.959 0.859 261.109 (2.268) (49.872) (4.637) (172.521) Investment freedom -0.012 -0.040 -0.008 -0.082 (0.011) (0.215) (0.013) (0.243) LOG(Corruption) 1.498 -31.333 4.842 2.225 (1.398) (29.152) (1.828)*** (38.909) Political Stability 0.883 -5.028 1.243 -8.324 (0.372)** (6.328) (0.424)*** (7.241) LOG(Rule of Law) -3.662 8.976 -6.198 15.908 (1.270)*** (27.526) (1.847)*** (33.288) CST*MNE No tax haven -1.083 -7.533 -1.731 -213.201 (1.921) (44.169) (8.953) (232.727) CST*MNE tax haven -5.331 -29.841 -1.073 -207.268

(3.508) (51.888) (9.061) (205.699) LOG(Corruption)*MNE No tax haven (1.913) -0.064 (36.206) 55.554 (2.842) 0.380 (52.567)** 104.029 LOG(Corruption)*MNE tax haven (2.550) -0.784 (47.526) -0.128 (3.557) 0.286 (64.423) 36.797 Investment Freedom*MNE No tax haven (0.014) 0.012 (0.281) 0.077 (0.018) 0.018 (0.330) 0.193 Investment Freedom*MNE tax haven (0.021) 0.014 (0.347) 0.347 (0.021) 0.017 (0.357) 0.150 Political Stability*MNE No tax haven (0.538) -0.597 (9.758) 15.454 (0.708) -0.654 (11.909)* 21.593 Political Stability*MNE tax

haven (0.842) -0.386 (17.047) -4.804 (1.136) 0.081 (24.721) 1.942 LOG(Rule of Law) *MNE No

tax haven (1.527) 2.214 (31.360) -49.954 (2.392) 2.358 (42.949) -67.373 LOG(Rule of Law)*MNE tax

haven (2.380) 2.388 (46.030) -17.257 (3.703) 0.850 (69.030) -47.490

R-squared 0.749 0.426 0.764 0.402

(36)

The results indicate that only two of the interaction terms are significant. The coefficients on the interaction terms for the tax rate are all negative, but not significant. There might thus be a negative relation present, but the sample might have too much noise to clear this out. This counts for both MNEs with and without a tax haven connection.

In regression (4) the coefficient is significantly positive for the interaction term for the corruption index and affiliates of MNEs that have no tax haven connection. Hence, affiliates of MNEs pay relatively more taxes per 100 units of profit than domestic companies if there is an increase in the corruption index, or a decrease in the level of corruption. The other significant interaction term is political stability and MNE status in regression (4): Affiliates of MNEs pay relatively more taxes per 100 units of profit after an increase in a country’s political stability than domestic companies. However, differences in the taxes paid per level of profit do not necessarily indicate profit shifting.

(37)

7.DISCUSSION

This study tries to find evidence on variables driving the profit shifting in developing economies by MNEs. I compare companies that do not have the opportunity to shift profits to other countries, to companies that do have this opportunity, and check whether the reaction of multinational companies to changes in variables potentially affecting the degree of profit shifting differs from the reaction of domestic companies. Differences in reaction are attributed to the ability of the multinational to shift profits. This section contains an overview of the hypotheses tested in this study, and the conclusions I draw based on the result section.

CORPORATE STATUTORY TAX RATE

H1: An increase in a country’s corporate statutory tax rate is related to an increase in profits shifted out of the country by multinational companies.

The results of this study show that domestic companies report less profits before taxes per unit of asset after an increase in the corporate statutory tax rate. Affiliates of MNEs, that have the ability to shift profits to other countries, do not react significantly different than domestic companies to changes in the tax rate (see section 6.2). When the affiliates of MNEs are subsequently divided in two groups: one group of affiliates of MNEs with a tax haven link, and one group of affiliates of MNEs without a tax haven link, there is still no consistent evidence for a relation between the tax rate and the ability to shift profits across borders (see section 6.3). The tax rate is therefore not a variable that explains the profit shifting behavior of MNEs in this study.

(38)

CORRUPTION

H2: A decrease in a country’s level of corruption is related to a decrease in profits shifted out of the country by multinational companies.

The results of this study indicate that domestic companies report more profits per unit of asset if there is an increase in the corruption index, or a decrease in the level of corruption (see section 6.2). Affiliates of MNEs also report more profits if there is a decrease in a country’s level of corruption, this reaction is however smaller in magnitude than the reaction of domestic companies. The difference in reaction is negative: after the increase in the corruption index affiliates of MNEs report relatively less profits per unit of asset than domestic companies, compared to the situation before the increase in the corruption index. The results of section 6.3, where the companies are divided in a group of domestic companies, a group of affiliates of MNEs without a tax haven link, and a group of affiliates of MNEs with a link to a tax haven, confirm that both of the groups with affiliates of MNEs report less profits per unit of asset after an increase in the corruption index relative to domestic companies. Whether this effect is stronger for affiliates of MNEs that have a link to a tax haven, than for affiliates of MNEs that do not have this link, cannot be said with certainty, as the standard deviations for the coefficients are quite high. The difference in reaction between domestic companies and affiliates of MNEs is attributed to the ability of the affiliates of MNEs to shift profits across borders. As affiliates of MNEs report relatively less profits per unit of asset than domestic companies after a decrease in a country’s level of corruption, the affiliates of the MNEs are assumed to shift part of their profits outside the developing country. This is in contrast to the hypothesis. Another explanation might be more appropriate. An affiliate of a MNE may receive a fixed reward, based on the functions performed, assets owned and risks assumed by the affiliate, which are not changed in case a country’s level of corruption changes. A small decrease in the level of corruption might not motivate MNEs to allocate more profits, or more functions, assets and risks, to the affiliate in the developing country, as the MNE rather allocates these profits to an affiliate in a country that has a low level of corruption. However, as countries with a low level of corruption are not in the sample, this does not show up in the results.

PROFIT SHIFTING IN DEVELOPING COUNTRIES Esther Helling