On Tax Authority Control and Multinational Profit Shifting Behavior

Transcription

1 On Tax Authority Control and Multinational Profit Shifting Behavior Matthias Dischinger University of Munich Nadine Riedel Said Business School, University of Oxford Preliminary Version Abstract This paper presents evidence that profit shifting of multinational entities (MNEs) critically depends on the level of shifting control by tax authorities in high tax countries - precisely on the authorities access to tax accounts of foreign multinational affiliates as this is presumed to facilitate the detection of shifting behavior. Among tax practitioners, it is well known that this access depends on the headquarter being located in the high tax country. If tax authorities demand information on foreign accounts, a headquarter cannot plausibly argue that this information is refused to be provided by its foreign affiliates while this is known to be common practice of dependent subsidiaries. Using a large panel of European MNEs, we show that profit shifting activities are smaller by more than 70% if the multinational headquarter resides in the high tax country. Sensitivity checks indicate that the result is robust against alternative explanations for this pattern. Thus, the paper suggests that profit shifting activities could be substantially reduced with enhanced information exchange between national tax authorities. JEL classification: H25, F23, H26, C33 Keywords: corporate taxation, multinational enterprise, profit shifting Department of Economics, University of Munich, Akademiestr. 1/II, Munich, Germany, phone: , matthias.dischinger@lrz.uni-muenchen.de. Centre for Business Taxation, University of Oxford, Park End Street, Oxford, OX1 1 HP, nadine.riedel@sbs.ox.ac.uk Financial support by the German Research Foundation (DFG) is gratefully acknowledged. 1

2 1 Introduction This paper presents empirical evidence that profit shifting activities of multinational enterprises (MNEs) critically depend on the control intensity of national tax authorities. Although the economic literature has brought forward conclusive empirical evidence for shifting behavior of multinational firms (for a recent survey, see Devereux, 2006), the interaction between profit shifting activities and tax authority control has not yet been clarified in an empirical framework. This is surprising as profit shifting volumes are identified to be of relevant size both in Europe and the US (e.g. Clausing, 2003 and Huizinga and Laeven, 2008), and policy makers as well as researchers intensively think about ways to avoid profit shifting activities and consequent tax competition behavior. The proposals thereby range from the introduction of alternative taxation schemes for multinational firms like formula apportionment to the international coordination of corporate tax rates (e.g. European Commission, 2001). An alternative route to these proposals would be to provide tax authorities with enough (or at least more) information to effectively control and consequently deter profit shifting activities. This was hardly pursued in the past, probably mostly due to a lack of evidence on the interaction between shifting control and shifting activities caused by the absence of appropriate data about resources spent on the control of profit shifting activities. Our paper tries to make a first step in this direction. We suggest an indirect approach to identify the relationship between shifting activities and the tax authority s screening intensity. Precisely, we exploit the fact that the success of profit shifting investigations by national tax authorities critically depends on the amount of information that the tax offices can gather about a multinational firm. Among others, this especially refers to tax statements not only being available for the investigated domestic affiliate but also for foreign locations belonging to the same multinational group. As profit shifting activities translate in an increased profitability at tax haven affiliates and a simultaneous reduction in profitability at high tax locations, the activities are easier tractable if tax information on the whole multinational group was available. However, since there is hardly any international coordination of tax authority work, national tax offices may only retrieve the information on foreign multinational affiliates directly from the investigated firm. Whether this information is made available, thereby critically depends on the ownership structure of the multinational entity. Precisely, if the investigated affiliate is the multinational headquarter and the tax office demands information on foreign multinational tax accounts, then the firm can hardly argue that its foreign subsidiaries refused to provide it with the demanded tax accounts 2

3 since the foreign subsidiaries are bound to the parent instructions via the ownership chain. In contrast, if the investigated affiliate is a mere subsidiary, it may ask at the parent firm for the foreign tax statement(s) but the parent is not bound to follow the demand and can deny the provision which is well known practice in international tax authority investigations. Combined with the common perception that only tax authorities in high tax countries have an incentive to effectively control profit shifting behavior (see e.g. Peralta et al., 2006), the considerations give rise to the hypothesis that MNEs whose parents reside in a high tax country should observe substantially lower profit shifting activities than MNEs whose parents are located at a low tax country since tax authorities can gather and screen more information for the former group of multinationals. We test this hypothesis using data on European multinational subsidiaries and their parent firms. The data is drawn from the AMADEUS data base provided by Bureau van Dijk and contains detailed accounting and ownership information on multinational subsidiaries in EU 25 for the years 1995 to To test the theoretical prediction, we in a first step define two subsamples of subsidiaries, namely subsidiaries which observe a larger and lower corporate tax rate than their parent firm respectively. As rationalized above, tax authorities of the high tax country (i.e. those which can plausibly be assumed to effectively engage in profit shifting control) will have a more information at their disposal if the headquarter is located within their borders and henceforth, we expect lower shifting activities for the latter sub group. Indeed, our empirical analysis indicates that profit shifting activities are significantly smaller by around 70% if the headquarter tax rate exceeds the subsidiary tax. If we extend our bilateral analysis and account for the structure of the whole multinational group, we find that MNEs whose parent firm resides in a country with a larger corporate tax rate than virtually all of its subsidiaries, shifting activities between the parent and the subsidiaries become indistinguishable from zero. In contrast, if the parent firm resides in a country with a smaller corporate tax rate than virtually all of its subsidiaries, shifting activities between the parent and the subsidiaries are measured to be even larger. Moreover, we run some robustness checks, mainly to determine whether the observed data pattern may be generated by alternative explanations and find our story to prevail. Our analysis contributes to the literature on profit shifting activities of multinational firms. From a theoretical perspective, several papers claim that profit shifting behavior deteriorates the efficiency of the international corporate tax scheme since countries compete for the shifty international tax base and henceforth (from a world welfare perspective) set inefficiently corporate tax rates (e.g. Haufler and Schjelderup, 2000; 3

4 Eggert and Haufler, 2006). We are aware of only a few number of theoretical papers that explicitly model international corporate tax competition and tax authority screening behavior. Among the few exceptions are Peralta et al. (2005) and Raimondos-Moeller and Scharf (2000). Furthermore, the empirical literature has brought forward conclusive direct and indirect evidence for profit shifting behavior (for direct evidence, see e.g. Clausing, 2003; for recent indirect evidence, see e.g. Huizinga and Laeven, 2008). However, to the best of our knowledge none of the empirical work is concerned with the interaction between profit shifting activities and the level of national tax authorities shifting control. Thus, the major contribution of our paper is to provide evidence that additional tax payer information is crucial in preventing profit shifting activities and information on foreign multinational tax accounts is able lower observed shifting activities by a quantitatively and statistically significant 70%. Consequently, from a policy perspective, our paper makes a strong point in favor of approaches to enhance the international cooperation and information exchange between national tax authorities with the purpose to prevent international profit shifting. If multinational tax accounts were made available to tax authorities in high tax countries in cases where the headquarter resides in a low tax economy, the results suggest that profit shifting activities might be substantially deterred. Enhanced international cooperation with respect to profit shifting control may thereby tie in with preceding suggestions by the United Nations Organization (UN, 2001) to found a supra-national tax authority or with the implementation of information exchange with respect to other tax instruments, e.g. the savings tax, in the context of the EU Directive on the taxation of savings income in 2003 (European Council, 2003). Apart from that, our paper may help to shed some light on the motivation for recent multinational headquarter relocations to low tax economies, like for example the IT companies SanDisk and Cryptologic which moved their headquarters to Ireland in 2006 and 2007, or the food company Kraft which moved its European headquarters from Austria and Great Britain to Switzerland. Since our paper suggests that the ownership structure directly influences the tax authorities level of profit shifting control, MNEs have a clear cut incentive to optimize their profit shifting opportunities by locating the headquarter firm at a low tax economy. That may add to other tax explanations which were claimed to determine the multinational headquarter choice, like e.g. the double taxation rules and withholding taxes (see Huizinga and Voget, 2006). 4

5 2 Short Theoretical Motivation To illustrate the basic idea of our analysis, we consider a model with two countries a and b. Both countries host one affiliate of a representative multinational entity (MNE). For simplicity reasons, we consider capital K i to be the only production factor, with i {a, b}, and the production in both affiliates to follow a standard production technology F (K i )withf > 0andF < 0. Capital costs are denoted by r and considered to be fixed from the countries perspective. Affiliate profit is taxed at the corporate tax rate t i with i {a, b} whereas we assume that capital costs are fully deductible from the corporate tax base (this assumption will not be decisive for our results). Additionally, for illustrative reasons, we consider country a to be the high tax country and hence it holds t a >t b. The affiliates pre tax profit is henceforth defined π a = F (K a ) rk a s (1) π b = F (K b ) rk b + s (2) The MNE furthermore has the possibility to shift profits between the two affiliates. We denote the amount of profit shifted from country a to country b by s. According to empirical evidence, MNEs engage in profit shifting activities by distorting the transfer prices for intra firm trade (see e.g. Clausing, 2003, Svensson, 2001), the group s debt equity structure (e.g. Buettner and Wamser, 2006) or the location of highly profitable (intangible) assets across subsidiaries (Dischinger and Riedel, 2008). Intuitively, only tax authorities in high tax countries have an incentive to prohibit profit shifting activities out of their borders while tax authorities in low tax economies lack these incentives since their countries gain corporate tax revenues through the multinational tax avoidance transactions. Thus, following previous papers, e.g. Perwaulty et al., we assume that only tax authorities in high tax countries effectively control profit shifting behavior. Additionally, we have to define the tax authority s screening heuristic which determines the detection probability of profit shifting activities and henceforth the costs of transfer pricing. Since arm s length prices for goods traded within a multinational corporation are usually hard to determine, tax authorities in the European Union have followed a so called net margin based transaction method to control for profit shifting activities of multinational groups. This approach implies the comparison of a multinational affiliate s profitability, measured as declared pre tax profit over sales or input factors, to similar firms productivity, e.g. to an industry average. Since the structure and productivity may, however, exhibit a considerable heterogeneity between firms, a 5

6 more direct way to detect profit shifting behavior between multinational affiliates is to compare the productivity of affiliates belonging to the same multinational group. Formally, both ideas are captured in the cost function C = C ( φ a φ a,α(φ b φ a ) ) with φ i = π i /F (K i ) being affiliate i s productivity, with i {a, b}, andφ a being the average (national) firms productivity in affiliate a s industry which is assumed to be used as a reference value by the tax authorities in country a to control for multinational profit shifting behavior. 1 Therefore, the profit shifting costs C are determined by the difference in the productivity of affiliate a and the average productivity of its pool of reference firms (φ a φ a ) as well as by the difference in the productivity of affiliates a and b of the multinational firm (φ b φ a ). The latter effect however only applies if the multinational headquarter is located in the high tax country a since it is only then that the tax authority is in the position to retrieve information on the tax accounts of the low tax subsidiary. This holds since international coordination among tax authorities is small within the OECD and henceforth, national tax offices are known to derive the required information directly from the tax payer under investigation. If the national tax authorities in their screening effort ask for foreign tax accounts of the multinational entity, it is required to demand the information from foreign affiliates belonging to the same multinational group. However, foreign affiliates may deny the provision of their tax statements if the Weisungsbefugnis In contrast, if it is the subsidiary which is located in the high tax country, it is... Formally, this is captured by the binary variable α which is assumed to take on the value 1 if the headquarter is located in the high tax country and the value 0 otherwise. Moreover, we presume the cost function to exhibit the following characteristics C (φ a φ a ) > 0, C (α(φ b φ a )) > 0,C (φ a φ a ) > 0, C (φ b φ a ) > 0(3) Consequently, the detection risk is assumed to rise convexly in the productivity differences between affiliate a and its national reference firms and affiliate b (if α =1)which is in line with previous studies e.g. Haufler and Schjelderup (2000). 1 The firm is thereby assumed to know the average productivity level of its reference group φ a and to take it as given. Since firms, in reality, are unlikely to observe declared pre tax profits and productivities of other firms in the industry, it might be more realistic to assume that the MNE has to form expectations over the parameter φ a. Nevertheless, introducing a stochastic component to the model just complicates the analysis, without adding new insight. We take a short cut of the deterministic model presented above. The results for the stochastic model are available from the authors upon request. 6

7 The MNE s overall after tax profit is thus defined π =(1 t a )π a +(1 t b )π b C (4) The MNE maximizes equation (4) with respect to s, K a and K b. The first order condition with respect to s reads t a t b = C φ a 1 F (K a ) + α C (φ b φ a ) 1 F (K a )+F (K b ) (5) The first order conditions with respect to K a and K b are displayed in the appendix. Equation (5) equates the marginal benefit from shifting one unit of profit from high tax country a (left hand side) to low tax country b to the marginal costs of profit shifting (right hand side). The first term on the right hand side captures that increases in profit shifting from country a to country b lower the reported pre tax profitability of affiliate a and henceforth increase the screening intensity of the tax authority and henceforth the detection probability and the MNE s profit shifting costs. The second term on the right hand side captures the effect that if the headquarter is located in the high tax country increased profit shifting from country a to country b enlarges the productivity differences φ b φ a and therefore enhances the probability that the tax authorities detect the income shifting activities. 2 This directly leads to the proposition Proposition 1. The volume of profit shifting between two multinational affiliates is larger if the corporate rate at the headquarter location is lower than the corporate tax rate at the subsidiary location (α =1). In the next section, we briefly explain how and under which assumptions the theoretical model translates in an empirical identification strategy. The estimation methodology is then described in section 5. 3 Empirical Identification Our empirical identification strategy relies on three assumptions. First, in line with previous papers in the economic literature, we assume that only tax authorities in 2 Note, that the model predicts the marginal profit shifting costs of a unit s to be smaller, the larger the capital investment at affiliates a and b (see the second and third term on the right hand side of equation (5). The intuition for this is very simple, as one unit of additional profit shifting from country a to country b affects the profitability of affiliates a and b less, the larger are the levels of capital investment and output. 7

8 high tax countries have an appropriate incentive to control for multinational profit shifting behaviour since only their governments suffer revenue losses by the evasion activities. In contrast, tax authorities of low tax economies gain tax revenues through shifting behaviour and henceforth lack appropriate control incentives (see e.g. Peralta et al., 2006). Second, tax authorities base their investigation for profit shifting behaviour on a so called net margin based transaction method which compares the profitability of a multinational affiliate to other comparable firms. Most OECD countries are known to follow this approach as it has proved to be a good heuristic to identify tax evaders (whose reported pre tax profit is likely to be below the one of comparable non shifting firms). One additional advantage of the net margin based transaction method is that it can be implemented when international profit shifting guidelines may not applicable due to a lack of information, as is e.g. often the case with respect to the OECD transfer pricing guideline due to missing arm s length prices for multinational intra firm trade. 3 Third, tax authorities are more likely to detect profit shifting behaviour if they can compare the profitability of the affiliate under investigation, not only to the profitability of comparable unrelated firms, but also to the reported profitability of foreign affiliates within the same multinational group. In many respects this intra firm comparison is a powerful control mechanism as profit shifting activities directly translate in both, downward biased profitability measures in high tax countries and upward biased profitability measures in low tax countries. Moreover, these profitability differences within tax evading firms can often only hardly be blamed to idiosyncratic differences between the affiliates. There are two potential sources from which tax authorities in high tax economies can retrieve tax accounting information on foreign multinational affiliates. On the one hand side, they may address other countries tax authorities for assistance. However, the international cooperation between tax authorities is small at best. On the other hand side, the tax authority may demand the tax accounting information directly from the affiliate under investigation. The tax payer is bound to provide the required information if he is in the position to retrieve it. That in turn is the case if other group affiliates are bound to follow the tax payers instructions which in turn is the case if he holds a considerable ownership share so that he may effectively control the entities. In contrast, if the affiliate does not hold ownership shares in foreign entities, these are not bound to follow its instructions and henceforth the MNE must not provide foreign affiliates tax statements. Thus, we arrive at the testable hypothesis that profit shifting between two entities 3 For a theoretical discussion of tax evasion if tax authority follow this net margin method, see e.g. Reinganum and Wilde (1986). 8

9 is significantly larger if the headquarter is located at the low tax economy while the dependent subsidiary is located at the high tax country. We investigate this question based on subsidiary data for the EU 25 in the years 1995 to To test the hypothesis, we thereby split the data in two subgroups: affiliates which observe a larger tax rate than their parent firms and affiliates which observe a lower tax rate than their parent firms. Our theory predicts that the latter group should engage in significantly lower profit shifting activities. 4 Data Set We use the commercial database AMADEUS which is compiled by Bureau van Dijk. The version of the database available to us contains detailed information on firm structure and accounting of 1.6 million national and multinational corporations in 38 European countries from 1993 to 2006, but is unbalanced in structure. We focus on the EU 25 and on the time period of as these countries and years are sufficiently represented by the database. Our criteria of being a multinational enterprise is the existence of a foreign corporate immediate shareholder (parent) with totally at least 90% ownership. Therefore, the observational units of our analysis are multinational subsidiaries within the EU 25. The AMADEUS data base thereby allows to link the subsidiary information to its parent firm which not necessarily must reside within the EU. In line with previous studies, we restrict our analysis to firms that observe positive pre tax profits. Our panel data sample thus consists of 24,948 observations from 6,450 MNE subsidiaries for the years 1995 to Hence, we observe each affiliate for 3.9 years on average. With all above restrictions, our sample contains firms from all EU 25 countries despite Cyprus, the Baltic states, Malta and Slovenia. The country statistics are presented in Table 1 in the Appendix. The AMADEUS data has the drawback that the information on the ownership structure is available for the last reported date only which is the year 2004 in most cases. Thus, by doing a panel study, there exists some scope for misclassifications of parent subsidiary connections that changed during the sample period. However, in line with previous studies, we are not too concerned about this issue since the described misclassifications introduce additional noise to our estimations that will bias our results towards zero (see e.g. Budd et al., 2005). We merge data on statutory corporate tax rates at the parent and subsidiary location, as well as basic country characteristics like 9

10 GDP per capita and the population size. 4 Table 2 in the Appendix displays basic sample statistics. The subsidiaries in our sample observe fixed asset investments of 52 million US dollars, on average, and employ 188 workers. The average pre-tax profit earned is measured with 4 million US dollars. The average statutory tax rate that is faced by the subsidiaries in our sample is 34.7% and thus slightly exceeds the statutory corporate tax rate at the parent level which is determined to be 35.4%. The average tax differential between the subsidiary and the parent is 0 whereas the variable, however, exhibits a considerable standard deviation varying from to Econometric Approach Our main purpose is to test the hypothesis of our theoretical model which predicts that profit shifting activities are significantly reduced if the headquarter resides in a high tax country with a larger corporate tax rate than the subsidiary. In a first step, we test for that hypothesis by defining two subgroups of subsidiaries, namely those which observe a parent firm with a larger statutory corporate tax rate and those which observe a parent firm with a lower statutory corporate tax rate. For these subgroups, we estimate a standard profit shifting equation of the following form the following form log π it = β 0 + β 1 TAXDIFF it + β 2 X it + ρ t + φ i + ɛ it (6) whereas π it represents a the pre tax profit of subsidiary i at time t. Since the pre tax profit variable exhibits a rather skewed distribution, we employ the logarithm of pre tax profits as a dependent variable. To avoid results that are affected by scale economies, we normalize the pre tax profit measure by the subsidiary s sales. X it comprises time varying affiliate and country control characteristics like the logarithm of the subsidiary s fixed assets stock (per sales), the logarithm of the subsidiary s number of employees (per sales), GDP per capita and the population size. Furthermore, year dummies ρ t are included to capture shocks over time common to all subsidiaries. ɛ it describes the error term. As our micro data is in a panel structure, we are able to add fixed effects of the subsidiaries to control for non-observable, time constant firm specific characteristics φ i. While the use of a fixed-effects model is suggestive while dealing with micro data, it is 4 The statutory tax rate data for the EU 25 is taken from the European Commission (2006), while the rates for affiliates outside the EU are based on data of the tax consultancy firm KPMG (2006). Country data for GDP per capita and population are obtained from the OECD. 10

11 also preferred to a random-effects model suggested by a Hausman-Test. Equation (6) is estimated using a fixed effects Logit model. Note, moreover, that a Chow test suggests separate estimation of equation (6) for the subgroups of affiliates with high tax and low tax parents respectively. The explanatory variable of central interest is TAXDIFF it which is defined as the difference between the statutory corporate tax rate of a subsidiary and the statutory tax rate of the parent firm. As suggested by our theoretical model, the tax rate difference between these two affiliates determines the marginal gains from profit shifting activities and consequently the shifting incentive. Since MNEs observe an incentive to shift profits from high tax to low tax countries, we expect β 1 < 0. In a robustness check, we determine whether the expected pattern strengthens if we investigate two subsamples of subsidiaries which belong to one of two subgroups of extreme multinational firms in the sense that the multinational parent observes a larger and lower corporate tax rate than virtually all of its subsidiaries respectively. Precisely, we identify MNEs in which 90% of the subsidiaries observe a lower corporate tax rate than the parent firm and MNEs in which 90% of the subsidiaries observe a larger corporate tax rate than the parent. For the former group, we expect profit shifting activities between the headquarter and the affiliates to be ceased substantially as the tax authorities at the headquarter location have access to information on the tax accounts of all the multinational subsidiaries and may thus easily identify conspicuous profitability patterns that may point to profit shifting activities. In the latter case of MNEs, instead, virtually none of the high tax countries tax authorities at the subsidiary locations has the chance to get hold of necessary foreign tax account information and thus profit shifting control is weakened. 6 Empirical Results This Section presents our empirical results. Throughout all regressions, the observational units of our panel analysis for the years are the multinational subsidiaries as explained in Section 4. Additionally in all upcoming estimations, year dummy variables are included and heteroscedasticity robust standard errors adjusted for firm clusters are calculated and displayed in the tables in parentheses. First, we dwell on our baseline estimation in Section 6.1 and second, present a set of robustness checks in Section

12 6.1 Baseline Estimation In Table 3, we present the basic results of our regression analysis. In line with previous studies, we give indirect evidence for profit shifting behavior by regressing subsidiary profitability on the tax rate differential between the subsidiary and the headquarter firm. We thereby split the sample into subgroups of subsidiaries that belong to parent firms with a larger and smaller corporate tax rate than the subsidiary respectively, as explained in section 5. According to our theory, we would expect that profit shifting activities are substantially reduced if the headquarter firm is located in the country with the higher corporate tax rate. This is impressively confirmed by the regression results. In specifications (1) and (3), fixed effect estimations for the samples of low tax parent (HTP) and high tax parent (LTP) firms are presented. In line with the expectations, the coefficient for the tax rate differential is estimated with 1.53 for the former group and with 0.47 for the latter group. Thus, both coefficient estimates are negative and statistically significant suggesting that the tax rate differential between the subsidiary and the parent firm negatively affects the subsidiary s profitability which is in line with profit shifting behavior. However, the coefficient estimates substantially differ in size, with the coefficient for the high tax parent group being smaller by 70%. Note that the difference between the coefficient estimates is also statistically significant at the 5% level. This result is confirmed in specifications (2) and (4) where we additionally control for host country characteristics like GDP per capita and the population size. 6.2 Robustness Checks At first, we run all specifications with the additional inclusion of 110 one digit NACE code industry year dummies. This add on does not change any of our qualitative and quantitative results. In another robustness check, we investigate whether the profit shifting pattern of the baseline analysis survives and even strengthens if we restrict our focus to extreme cases of MNEs where the parent company either owns almost only firms in country with a lower corporate tax rate or subsidiaries in countries with a higher corporate tax rate. Since this analysis requires information on all subsidiaries of a respective parent firm, we have to restrict our sample to subsidiaries with parent firms in EU 25 since comprehensive ownership relations are available with AMADEUS only for countries within the EU. The results are presented in Table 4 and confirm the pattern of our 12

13 baseline estimation. Specifications (1) and (2) show the results for subsidiaries that belong to corporate groups whose parents observes at least 90% of its affiliates in a country with a lower statutory corporate tax rate and a higher statutory corporate tax rate respectively. Again, while we find a strong and statistically significant negative influence of the tax rate differential on subsidiary profitability in the subgroup of affiliates that belong to low tax parent firms, the coefficient estimate in the subsample of subsidiaries that belong to high tax parents is basically zero and does not gain statistical significance. Again, the coefficient estimates turn out to be statistically different from each other at the 5% level. In specifications (3) and (4) we add a full set of industry-year dummies to the analysis and in specifications (5) and (6), we additionally control for a set of time varying country characteristics. The result turns out to be robust against these modifications and thus suggests that MNEs profit shifting activities between the headquarter and its subsidiaries critically depend upon whether the headquarter is located in a high or low tax country. 6.3 Discussion One may have two major caveats against this analysis which mainly concern the question whether the observed pattern is indeed caused by the interaction of tax authority control activities and profit shifting behavior. A first caveat may be that multinational firms headquartered in high tax countries and multinational firms headquartered in low tax countries systematically differ from each other. That would suggest that the difference in the coefficient estimate for the two subgroups may be driven by an unobserved heterogeneity. However, we consider such an explanation to be unlikely for two reasons. First, we consider that unobserved host and parent country characteristics are not able to drive the results since headquarters in a certain country, e.g. Sweden, may very well belong to a high tax as well as to a low tax parent sample, depending on whether their subsidiary is located in a country with a higher or lower corporate tax rate. Second, we moreover also do not consider that unobserved subsidiary or parent characteristics drive the results as one and the same firm may change their status as belonging to the high tax parent sample and to the low tax parent sample over time in the course of adjustments in the statutory corporate tax rate. A second caveat may be that some some other explanation may drive the determined connection between the country of headquarter residence and the observed pattern of 13

14 profit shifting behavior. Precisely, one may for example think about the fact that profit shifting volumes between the headquarter and one of the foreign affiliates is strongly determined by the number of connections that a multinational headquarter located in a high tax country observes towards foreign (tax haven) affiliates. Given that tax authorities usually control according to the so called transaction based net margin method that compares profitabilities across firms, a high tax headquarter may be reluctant to shift a substantial amount of profits through various channels since this may strongly reduce its reported pre tax profits and henceforth may raise the suspicion of its local tax authority. If this was the explanation for the observed pattern, we would expect it to vanish if we restricted the analysis to MNEs which comprise only two affiliates. We find that this is not the case and thus, we are confident with respect to the explanatory power of our story. 7 Conclusions This paper provides empirical evidence that the scope of profit shifting between multinational headquarters and their subsidiaries critically depends on whether the parent company is located in a country with a high tax or low tax rate. Precisely, we find that profit shifting activities are around 70% smaller if the parent tax rate exceeds the subsidiary tax rate. We interpret this result to reflect that only tax authorities in high tax countries effectively screen for profit shifting behaviour (since they are the ones to loose corporate tax base by shifting activities) and that the profit shifting control is more efficient if the headquarter of the multinational firm is located at the high tax country. This is due to the fact that if tax authorities ask for foreign tax accounts, multinational headquarters could hardly argue that their foreign affiliates refuse the provide the respective information (since the subsidiaries are bound to follow their owners inquiries) while this is common practice for dependent subsidiaries. We check whether the results may be driven by alternative explanations and find our explanation approach to prevail. Thus, our analysis suggests that a substantial fraction of the well documented inefficiencies caused by multinational profit shifting could be abolished if tax authorities in high tax countries were provided with tax account information on the whole multinational group irrespective of the headquarter location. Precisely, our paper thus suggests that an enhanced international coordination of profit shifting control by national tax authorities, may imply steep welfare gains. 14

15 8 References Buettner, T., 2003, Tax Base Effects and Fiscal Externalities of Local Capital Taxation: Evidence from a Panel of German Jurisdictions, Journal of Urban Economics 54, pp Budd, J.W., J. Konings and M.J. Slaughter, 2005, Wages and International Rent Sharing in Multinational Firms, The Review of Economics and Statistics 87, pp Buettner, T., and G. Wamser, 2007, Intercompany Loans and Profit Shifting - Evidence from Company-Level Data, CESifo Working Paper Series, No. 1959, CESifo Munich Clausing, K.A., 2003, Tax Motivated Transfer Pricing and US Intrafirm Trade Prices, Journal of Public Economics 87, pp Devereux, M.P., 2006, The Impact of Taxation on the Location of Capital, Firms and Profit: A Survey of Empirical Evidence, mimeo, University of Warwick. Dischinger, M., 2007, Profit Shifting by Multinationals: Indirect Evidence from European Micro Data, mimeo, Department of Economics, University of Munich European Commission, 2001, Towards an Internal Market Without Tax Obstacles: A Strategy for Providing Companies with a Consolidated Corporate Tax Base for Their EU-Wide Activities, COM(2001), 582 final (October 23). European Council, 2004, Directive on the Taxation of Savings Income, Council Decision 2004/587/EC. Haufler, A. and G. Schjelderup, 2000, Corporate Tax Systems and Cross Country Profit Shifting, Oxford Economic Papers 52, pp Haufler, A. (2006), Die Besteuerung Multinationaler Unternehmen, Munich Discussion Paper in Economics No , University of Munich. Hines, J.R., Jr., 1996, Altered States: Taxes and the Location of Foreign Direct Investment in America, American Economic Review 86(5), pp Hines, J.R., Jr., 1999, Lessons from Behavioral Responses to International Taxation, National Tax Journal, pp Hines, J.R., and E.M. Rice, 1994, Fiscal Paradise: Foreign Tax Havens and American Business, Quarterly Journal of Economics, 109(1),

16 Huizinga, H., and L. Laeven, 2008, International Profit Shifting Within European Multinationals, Journal of Public Economics, forthcoming. Huizinga, H. and J. Voget (2006), International Taxation and the Direction and Volume of Cross- Border M&As, CEPR Discussion Paper No KPMG, 2006, KPMG s Corporate Tax Rate Survey Mintz, J., 1999, Globalization of the Corporation Income Tax: The Role of Allocation, Finanzarchiv 56, pp Nielsen, S.B., P. Raimondos-Mller and G. Schjelderup, 2002, Tax Spillovers under Separate Accounting and Formula Apportionment, Working Paper Overesch, M., 2006, Transfer Pricing of Intrafirm Sales as a Profit Shifting Channel - Evidence from German Firm Data, ZEW Discussion Paper, No , ZEW Centre for European Economic Research Peralta, S., Wauthy, X. and T. van Ypersele (2006), Should Countries Control International Profit Shifting?, Journal of International Economics 68, Riedel, N. and M. Runkel, 2007, Company Tax Reform with a Water s Edge, Journal of Public Economics, forthcoming Weichenrieder, A.J., 2007, Profit Shifting in the EU: Evidence from Germany, CESifo Working Paper Series, No. 2043, CESifo Munich Wilson, John D., 1999, Theories of Tax Competition, National Tax Journal(52), pp

17 9 Appendix Maximizing equation (4) derives the following first order conditions with respect to capital investment in countries a and b F (K a ) = F (K b ) = r (1 t a ) + C sf (K a ) φ a F (K a ) α r (1 t b ) + α C (φ b φ a ) sf (K b ) F (K b ) C (φ b φ a ) sf (K a ) F (K a ) (7) (8) The first order conditions in equations (7) and (8) equate marginal benefits and marginal costs of capital investment. The marginal costs of capital investment K a in country a comprise the interest payment r for a marginal investment unit (first term on the right hand side) which is, however, reduced by profit shifting considerations in this model, i.e. profit shifting considerations increase capital investment in the high tax country a. The rationale behind this result grounds in two countervailing effects: first, increases in capital investment in country a tend to raise the profit earned by the affiliate in country a and henceforth affiliate a s productivity. Second, increases in capital investment in country a tend to lower the productivity of affiliate a since they increase the output F (K a ) which appears in the denominator of the productivity calculation. The first effect is shown to prevail and henceforth, capital investment at the high tax location is boosted by profit shifting considerations. Two equally opposing effects can be found with respect to profit shifting induced investment in the low tax country b whereas, here, the second effect is shown to prevail and henceforth, profit shifting considerations are found to deter capital investment at b. Nevertheless, note that these investment effects do not play any role for our empirical estimation strategy. 17

18 Table 1: Country Statistics Country Subsidiaries Austria 89 Belgium 466 Czech Republic 221 Denmark 443 Finland 325 France 824 Germany 321 Great Britain 1,008 Greece 59 Hungary 99 Ireland 135 Italy 493 Luxembourg 33 Netherlands 590 Poland 396 Portugal 105 Slovakia 44 Spain 676 Sweden 473 Sum 6,450 18

19 Table 2: Descriptive Statistics Variable Mean Std. Dev. Min Max Subsidiary Characteristics Pre-tax Profits 4, , , 000, 000 Fixed Assets 52, , , 400, 000 Employees , 784 Financial Leverage Ratio Tax Rate Measures Statutory Tax Rate Subsidiary Statutory Tax Rate Parent Tax Difference Dummy for Parent in LTC Dummy for Parent in HTC Other Country Characteristics GDP GDP per Capita Globalization Index (KOF) Corruption Index Notes: In thousand US dollars, current prices. Subsidiaries owned with 90% of the ownership shares. Median: 24, Min.: 3, Max.: 752. Calculated as subsidiary statutory tax rate minus parent statutory tax rate. 19

20 Table 3: Baseline Estimation, OLS Fixed Effects, Panel Dependent Variable: Log Profit Before Tax (per Sales) Sample LTP LTP HTP HTP Variable (1) (2) (3) (4) Tax Differential ( 4.18) ( 3.34) ( 2.10) ( 1.72) Log Fixed Assets (per Sales) (7.54) (7.63) (6.50) (7.99) Log Employees (per Sales) (14.93) (13.58) (17.15) (15.15) Leverage Ratio ( 15.16) ( 18.24) GDP ( 3.08) ( 2.75) GDP per Capita ( 1.30) (3.57) GDP Growth Rate (0.87) (1.15) Observations 25, , , , 659 Number of Firms 6, 450 5, 959 9, 077 8, 248 Adjusted R squared Notes: Heteroscedasticity robust standard errors adjusted for firm clusters calculated. Corresponding t-values in parentheses.,, indicates significance at the 10%, 5%, 1% level. The observational units are multinational subsidiaries, i.e. they exhibit a foreign parent with at least 90% of the ownership shares. Regressions (1) (3) comprise the subsample of firms where the subsidiary observes a larger statutory corporate tax rate than the parent, regressions (4) (6) comprise the subsample of firms where the subsidiary observes a smaller corporate tax rate than the parent. The endogeneous variable if subsidiary pre tax profit divided by operating profits. Log Fixed Assets is the logarithm of the subsidiary s fixed asset investment divided by operating profit, Log Employees stands for the logarithm of the subsidiary s labour costs, Leverage Ratio is the subsidiary s debt to equity ratio, Tax Differential represents the difference in the statutory corporate tax rates between the subsidiary and the parent firm. All regressions include year dummy variables. Moreover, HTP indicates that parent tax rate > subsidiary tax rate and LTP indicates that parent tax rate < subsidiary tax rate. 20

21 Table 5: Fixed Effects Regression, Panel Dependent Variable: Log Profit before Taxation (per Sales) Sample HTPS LTPS HTPS LTPS HTPS LTPS Variable (1) (2) (3) (4) (5) (6) Tax Difference (0.2905) (0.6447) (0.2969) (0.6854) (0.3144) (0.7325) Log Num. of Employees (per Sales) (0.0428) (0.0606) (0.0430) (0.0617) (0.0452) (0.0659) Log Fixed Assets (per Sales) (0.0238) (0.0369) (0.0241) (0.0369) (0.0273) (0.0409) Log GDP per Capita (0.5869) (0.8620) Log Population Year Dummies Industry Year Dummies (1.5330) (2.7170) Number of Observations 14, 546 7, , 331 7, , 590 6, 893 Number of Firms 3, 897 2, 381 3, 832 2, 325 3, 404 2, 082 Adjusted R Robust standard errors adjusted for firm clusters in parentheses.,, indicates significance at the 10%, 5%, 1% level. HTPS: High Tax Parent Sample (solely subsidiaries of MNEs for which more than 90% of the subsidiaries are located in a low tax country relative to the tax rate of the parent) LTPS: Low Tax Parent Sample (solely subsidiaries of MNEs for which less than 10% of the subsidiaries are located in a low tax country relative to the tax rate of the parent)