Risky profit shifting

Transcription

1 Risky profit shifting Manthos D. Delis Surrey Business School, University of Surrey, Guildford, GU2 7XH, UK Iftekhar Hasan Gabelli School of Business Fordham University, New York, NY 10023, USA Panagiotis I. Karavitis Department of Economics, University of Cyprus, POBox 20537, CY1678, Nicosia, Cyprus 1

2 Risky profit shifting Abstract We demonstrate the importance of risk factors related to the macroeconomic and fiscal stability in the countries of multinational subsidiaries residence for the potency of profit-shifting activity of multinational enterprises. Using firm-level data for 1,241 parent firms from 24 countries and 12,698 subsidiaries in 43 countries, we first identify more potent profit shifting in periods (or subsidiary s countries) with low macroeconomic risk. Subsequently, we show that even within periods of low macroeconomic risk, profit shifting is mainly potent toward subsidiaries in countries with a stable corporate tax rate across a number of years (low fiscal-risk countries). We contend that especially low fiscal risk is a prerequisite for the identification of significant flows of profit shifting. Keywords: International taxation, profit-shifting, multinational firms, macroeconomic and fiscal risk JEL codes: F23, H25, H32, M41; M48 2

3 1. Introduction Is there a role for macroeconomic and fiscal risk in identifying the magnitude of profit-shifting activity of multinational enterprises (MNEs)? The answer to this question is important for understanding the profit flows from parent firms to subsidiaries for tax-related reasons, identifying the true level of profit shifting, and informing policy. We argue that the macroeconomic and fiscal risk in the subsidiaries countries are decisive elements in shifting profit and this empirical observation has important policy implications for the way resources and policy tools are used to limit international profit shifting and improving social welfare and fairness. Macroeconomic and fiscal risks provide a possible explanation to the identification of relatively low levels of profit shifting when examining average estimates across a number of countries (e.g., Dharmapala, 2014). We begin by providing a formal argument in which the degree of costly profit shifting varies with an exogenous country-specific uncertainty component, related to either macroeconomic or fiscal risk. Assuming that each firm chooses the level of profit shifting to maximize profits, we show that the optimal fraction of profit shifted or the total amount of profit shifted are both non-decreasing in the tax differential between the parent firm s country and the subsidiary s country. We thus hypothesize that in periods (or countries) with low macroeconomic and/or fiscal risk, the profit-shifting activity is quite higher compared to periods (or countries) with high macroeconomic and/or fiscal-risk. We empirically study the validity of our theoretical argument using a panel dataset of parent firms from 24 countries and subsidiaries in 43 countries, with a maximum of approximately 36,200 subsidiary-year observations for the period To examine the role of macroeconomic risk, we exploit the outbreak of the global financial crisis as a seminatural experiment. This period entails a massive increase in macroeconomic risk, which 3

4 represents an exogenous shock to the decision of MNEs to shift profits abroad. Further, countries changing their corporate tax rates more frequently should pose higher fiscal risk for MNEs to shift profits to their subsidiaries residing in such countries. Thus, the decision of governments to change tax rates entails increased fiscal risk that MNEs potentially consider in shifting income. Our main empirical identification method builds on the differences-in-differences (DID) model introduced by Dharmapala and Riedel (2013). This model exploits earnings shocks at the parent level and examines their propagation toward subsidiaries. The main premise is that an exogenous increase in the profits of the parent company will imply partial shifting of profits to subsidiaries in low-tax countries. Exogeneity of the shocks to the parent firm is established through the construction of a variable based on the pre-tax profits of firms operating in the same industry and country with the parent firm (Bertrand et al., 2002). We restrict the empirical analysis to subsidiaries operating in a different industry (and of course country) than the parent companies, and this further restraints the profit shocks to be endogenous. Using this approach and the full sample, we fail to identify significant profitshifting activity, with this finding being in stark contrast with previous studies and earlier samples (e.g., Dharmapala and Riedel, 2013; Huizinga and Laeven, 2008). Subsequently, we split our sample to crisis and post-crisis periods and estimate the DID models separately for each of these periods. In line with expectations, we find potent profit-shifting behavior of MNEs mainly in the post-crisis period. Specifically, we find that if the parent firm experiences an earning shock of 10%, then the low-tax subsidiaries report approximately 0.56% higher profits than the high-tax subsidiaries, according to our most conservative estimates. These results are (i) robust to slight changes in the period of the global financial crisis and (ii) economically more significant when identifying crisis 4

5 periods on a country-specific basis. The results are also robust to the identification of profit shifting through the use of the so-called indirect approach (instead of the DID method) of Hines and Rice (1994), as refined by Huizinga and Laeven (2008). Our results also indicate that the source of profit shifting during periods of low macroeconomic risk is mainly transfer pricing and, to a more limited extent, debt shifting. Specifically, and again referring to our most conservative estimates, we find that a 10% positive earnings shock on parent firms increases the low-tax subsidiaries reported earnings before interest and taxes (i.e., our measure for transfer pricing) by 0.46%, while the equivalent response for debt shifting is only about 0.10%. However, we do find evidence that, unlike transfer pricing, debt shifting is also potent during the crisis period. We next turn to the role of fiscal risk in the profit-shifting behavior of firms. Given that we identify potent profit shifting mainly in periods (or countries) with low macroeconomic risk, we restrict our analysis to non-crisis periods and further split our sample between countries that have changed the level of their corporate tax rates over the last three or four years (high fiscalrisk countries) and those that have not (low fiscal-risk countries). Our findings indicate significant profit shifting only toward subsidiaries in low fiscal-risk countries and insignificant profit shifting toward subsidiaries in high fiscal-risk countries even within the non-crisis period. Importantly, the economic significance of these results almost doubles compared to the specifications considering macroeconomic risk alone. These results are robust to respecifications equivalent to those for the specifications on macroeconomic risk and also show that transfer pricing is the main channel of profit shifting. In brief, our results highlight that our testable hypothesis is validated: profit shifting is quite potent, as long as the macroeconomic and, more importantly, fiscal riskiness in the 5

6 subsidiaries country is low. These findings provide an explanation for the relatively low level of profit shifting that is empirically identified when using global samples without differentiating between types of countries: once we generate a more level playing field in terms of the macroeconomic and fiscal riskiness in subsidiaries countries, the profit-shifting flows are quite substantial. The remainder of this paper is organized as follows. Section 2 sets the context of our study by summarizing existing literature and providing a formal argument on the role of risk in the profit-shifting behavior of firms. Section 3 discusses our empirical strategy and presents the data set. Section 4 presents the empirical findings and Section 5 concludes the paper. 2. Setting the context 2.1. Summary of the related literature It is well documented in the empirical literature that multinational corporations engage in taxmotivated profit (income) shifting towards their subsidiaries located in low-tax jurisdictions. In an influential study, Hines and Rice (1994) suggest that the total reported subsidiary income is divided into the true income that originates in the transformation of inputs of production (labor and capital) to outputs and the shifted income that originates in profit-shifting activities. Phrased differently, income that cannot be attributed to the subsidiary s own resources is attributed to profit shifting. Since the contribution of Hines and Rice, the literature significantly advanced in terms of the procedures used to empirically identify profit shifting and here we outline only the most influential of these studies. 1 Huizinga and Laeven (2008) construct a weighted tax difference using information for all the affiliates of a multinational group (instead of the simple tax 1 For a thorough review of the literature and identification methods, see Dharmapala (2014). 6

7 difference between parent and subsidiary firms used by Hines and Rice). This procedure takes into account the possibility of shifting income from a high-tax subsidiary to a low-tax subsidiary and not only income shifting between the parent and each subsidiary. The results provide strong evidence for the existence of profit shifting. In turn, Dharmapala and Riedel (2013) propose a new approach to empirically identifying profit shifting through exogenous industry shocks on the earnings of the parent firms and their propagation towards their subsidiaries (a DID model). 2 A number of empirical studies examine the potency of profit shifting using the aforementioned methods. Dischinger and Riedel (2011) and Karkinsky and Riedel (2012) examine the tax-motivated shifting of intangible assets towards low tax subsidiaries. They find that the lower a subsidiary's corporate tax rate relative to other affiliates of the multinational group, the higher is its level of intangible asset investment. Markle (2015) investigates the role of territorial vs. worldwide tax systems in the multinational groups decisions to shift taxable income abroad. He finds that multinationals subject to territorial tax regimes shift more income than those subject to worldwide tax regimes, but that the difference in shifting is not statistically different when the worldwide firms can defer repatriation of the shifted income. Klassen and La Plante (2012) use a panel of US multinationals for the period to examine the role of the cost of profit shifting. They show that U.S. MNEs become more active at shifting income as the regulatory costs of shifting decrease. A common characteristic of these and other related studies is that they do not relate the potency of profit shifting to the riskiness of the macroeconomic or fiscal environment. For example, the global financial crisis led to an economic downturn in most developed countries 2 Alternative approaches are widely used in the tax accounting literature. For example, Collins et al. (1998) use consolidated data and assume that the accounting pre-tax rate of return on foreign sales is a function of the return on worldwide sales in the absence of income shifting. If, instead the return on foreign sales is a function of tax incentives after controlling for the worldwide return on sales, then this is attributable to profit-shifting activity. 7

8 and affected most sectors of the economy, sharply increasing the uncertainty in the macroeconomics and fiscal environment faced by MNEs. We posit that within this environment of increased uncertainty, the profit shifting activities of multinational firms might have become less potent. We analyze these issues by providing a simple analytical framework of profit shifting with country risk A formal argument on risk and profit shifting In this section, building on Dharmapala and Riedel (2013), we analytically link profit shifting to the riskiness of the economic and fiscal environment. Consider a representative multinational enterprise that consists of affiliates in countries and. These affiliates earn (exogenous) pretax profits and face corporate tax rates, where,. Without loss of generality, we assume that country is the high-tax country (i.e., ). We also assume that the multinational enterprise's home country has a territorial (or exemption) tax system that does not seek to tax the multinational enterprise's profits earned abroad. 3 The multinational enterprise can shift accounting profits between the two affiliates, for instance by charging a lower transfer price for goods and services bought by affiliate b from affiliate, or by creating financial arrangements in which the high-tax affiliate borrows from the low-tax affiliate b. We also assume that each country defines taxable income as being identical to accounting income (i.e., we assume book-tax alignment). The fraction of affiliate 's pre-tax profit that is shifted to the low-tax affiliate is denoted by x. Consistent with the previous literature (e.g., Haufler and Schjelderup, 2000), we 3 Under worldwide taxation, foreign profits are subject to taxation by the home country (in terms of our example, country taxes profits earned in country ). A pure form of worldwide taxation would eliminate the incentive to shift profits. In reality, however, worldwide systems have features such as the deferral of country 's tax on income earned in until affiliate pays a dividend to the parent in country that result in incentives for profit shifting. 8

9 assume that profit shifting behavior imposes costs C on the multinational enterprise, even if profit-shifting activities are not necessarily illegal. For example, the necessity for a multinational to defend its position either in a court (due to tax evasion) or to the public (due to negative publicity from extremely low tax payments) bears considerable cost. We assume that the cost of profit shifting is likely to depend on the fraction x of the hightax affiliate's profits shifted, as well as on the amount of income shifted (denoted here by ). Thus, we introduce in the model the parameter 0,1, which describes the degree of the uncertainty arising either due to the risk of the macroeconomic environment or due to the country-specific fiscal risk. This uncertainty affects the multinational enterprise s willingness to shift accounting income towards the lower tax jurisdiction affiliate in country. Uncertainty is assumed to be exogenous for the multinational (thus a parameter for our model), 4 with a higher value of implying safer profit shifting for the MNE. The uncertainty component varies between countries and across time for a number of reasons. From a more general viewpoint, uncertainty relates to macroeconomic risk due to potential political turmoil and policy fluctuations that might hurt the firms performance and incentives to shift profits to subsidiaries in these economies. From a more focused fiscal viewpoint, the frequency of the corporate tax rate changes in a subsidiary country potentially increases the uncertainty for profit shifting for the parent firm and thus it has a negative impact on the tendency to shift taxable income. The fiscal developments (mostly increases in corporate tax rates) in many countries after the eruption of the global financial crisis to compensate for higher levels of public debt are a good example of increased fiscal uncertainty. 4 We impose explicitly the uncertainty as an exogenous parameter to the multinational corporate in the sense that it cannot be affected from a single firm s activity. In contrast, we allow the cost function to entail idiosyncratic risk factors for each firm, like financial leverage, etc. 9

10 expressed as: The multinational enterprise's worldwide after-tax profits (denoted by ) can be 1 1 1,, (1) or in terms of total shifted amount: 1 1 1,. (2) Note that for 0 (i.e., extremely high risk) there is no profit shifting, while for 1 (i.e., risk does not affect the willingness for profit shifting) equations (1) and (2) reduce to the respective of Dharmapala and Riedel (2013). The multinational enterprise chooses x (or equivalently y) to maximize. Using Eq. (1), the FOC with respect to x is: 0. (3) Alternatively, using equation (2), the FOC with respect to y is: 0. (4) Clearly, the comparative statics show that the optimal fraction of and the amount of profit shifted ( ) are both non-decreasing in the tax differential between countries and : 0, 0. (5) These results retain the main finding of the literature on the gains from shifting profit from a to b (Hines and Rice, 1994; Huizinga and Laeven, 2008). The new element here is that as profit shifting becomes more risky, i.e. as the parameter falls, the effect of the tax differential also falls. This conclusion is intuitive because the willingness of the multinational corporations to engage in profit shifting activities stems from their objective to reduce their total tax liabilities and increase their after tax profit margin. However, for each monetary unit of pre-tax and pre- 10

11 shifting profit, the higher the uncertainty component, the lower are the expected gains for the pre-tax and after-shifting profits. The optimal fraction x is increasing in if the expression (6) is positive -- i.e. if the derivative of the cost of profit shifting with respect to the amount of profit shifted (evaluated at the optimal choice) is sufficiently small in relation to the tax differential between the affiliates and 0. When these conditions are satisfied, it is optimal for the multinational enterprise to shift a larger fraction of affiliate's profit to the low-tax affiliate in b when increases. We can see from equation (6) that the higher the uncertainty (i.e. the smaller the parameter ), the smaller is the effect of the increase of earnings on the optimal profit shifting fraction. Further, the amount of profit shifted ( ) is non-decreasing in the affiliate's pre-tax profit 0 (7) and decreases as the whole function becomes more and more risky due to uncertainty (i.e., as falls). In turn, the pre-tax and post-shifting profits of b increase in, given that 0. These results entail the core of our empirical analysis, which examines the role of macroeconomic and fiscal uncertainty in the potency of profit-shifting activities. Based on the above formal arguments, we formulate the following testable hypothesis: H 0 : The profit-shifting activity of firms is lower in countries with high macroeconomic and/or fiscal risk. 3. Data and empirical strategy 11

12 We aim to estimate a causal effect of taxation differences between the countries of the parent and subsidiary companies on a subsidiary s profits and, to this end, we resort mainly the DID model of Dharmapala and Riedel (2013). This approach relies on the identification of the propagation of an earnings shock only from the parent firm to each subsidiary and not between subsidiaries. The basic idea is to observe the effect of an exogenous shock on the parent s pre-tax and preshifting profit, constructed in this way to be unaffected by any potential profit-shifting activity on the subsidiary s profits. We consider the subsidiaries in the low-tax countries as the treatment group and the ones in the high-tax countries as the control group. We expect that an increase in the parent pre-tax and pre-shifting profits (i.e., a positive earnings shock) would exert a positive impact on the pre-tax profits of the low-tax subsidiary relative to the high-tax subsidiary. The empirical model takes the form:. (8) The dummy variable takes the value one if the subsidiary faces a lower corporate tax rate than the parent firm and the value zero otherwise. The variable is an exogenous proxy for the subsidiary parent profits, instead of the endogenous parent s own profits. Of course must closely represent the parent s profitability and thus we use comparable firms, which are defined as such when belonging in the same 4-digit NACE industry and same country per year. 5 More formally, following Bertrand et al. (2002), we use the following formula:,, 1,,, (9) 5 We use all the national and multinational firms included in Orbis for which information on profits and total assets is available (this amounts to more than a million observations). To avoid the correlation that arises if we include a firm itself in the calculation of its industry profitability and then use that industry s profitability to predict the firm s own profit, we exclude the firm itself from the set of comparable firms. 12

13 where denotes the total assets of comparable parent firms in year and denotes the comparable parents pre-tax profit over total assets. We keep only the subsidiaryyear combinations in our sample if (i) each set of comparable firms includes at least 10 firms and (ii) the subsidiaries operate in a different 4-digit NACE industry than their parent company. The first requirement increases the accuracy of our measure by providing a sufficient level of information about each industry. The second requirement allows avoiding the reported pre-tax profits of each subsidiary to be directly driven by an industry shock. From (9) we obtain the parents i pre-tax and pre-shifting profit by calculating. (10) If tax-motivated profit shifting occurs, then we expect a positive sign on. This implies that for each given level of corporate-tax difference between the parent and the subsidiary firms, we expect that a parent-firm earnings shock,, will propagate asymmetrically towards the low-tax subsidiaries rather than the high-tax subsidiaries. Our empirical analysis relies mainly on data from Orbis, which provides accounting data for national and multinational firms worldwide, as well as detailed information on their ownership structure and links between parent companies and subsidiaries. A firm is defined as a subsidiary if more than 50% of its shares are owned by another firm. For subsidiaries we use unconsolidated statements. The parent firms in our sample are global ultimate owners of multinational groups with at least one subsidiary in a foreign country. Information from Orbis has the drawback that the ownership structure is only available for the last reported year. In line with previous studies, this is not a key concern because the potential misclassification of parent-subsidiary connections would, if anything, bias our results towards zero (e.g., Budd et al., 2005). 13

14 For parent companies we have to rely on consolidated statements because otherwise our sample would shrink dramatically given the requirements needed to calculate equations (9) and (10). However, we do not expect this to be a serious problem in our analysis. First, it is important to note that consolidated parent profits can be shifted to low-tax subsidiaries and these should be included in the analysis (and not only unconsolidated profits), given that we examine profits shifted only from the parent firm to each subsidiary and not between subsidiaries. If we do not include the profits of the consolidated parent firms we might lose an important part of profit shifting. The only part of these profits that needs to be excluded is the profits of subsidiary i. However, as discussed above, for the calculation of the average industry profitability index in equation (9) we use data for comparable firms and thus the profits of subsidiary i are not included. Further, for the calculation of equation (9) it might even be preferable to use consolidated data as potential profit shifting is then netted out (as the considered firm is a multinational) when determining the earnings shock variable. In turn, for the component of equation (10), we avoid double counting the assets of the subsidiary i in the consolidated statement of the parent firm by subtracting this subsidiary s total assets from each parent firm s total assets. After dropping missing observations for our main variables, we are left with a sample with a maximum of 36,189 observations from 12,698 subsidiaries and 1,241 parent firms over the time period This sample contains subsidiaries from 43 countries and parent firms from 24 countries. In Table A1 of the Appendix we provide summary statistics for the parent firms by country and in Table A2 the equivalent information for the subsidiaries. 14

15 We formally define all variables used in the empirical analysis in Table 1 and also provide the data sources. We measure the subsidiary i s profits at time t using either the log of earnings before interest and taxes (EBIT), or the log of pre-tax earnings (EBT). 6 EBIT does not include financial income and payments and thus it can detect profit shifting activity that comes only from transfer pricing and not from debt shifting. On the other hand, EBT includes financial income and payments and it is thus suitable for the detection of both profit shifting channels. 7 [Inset Table 1 about here] We include in equation (8) the log of total assets as a measure of the subsidiary s size 8 and a vector of time-varying subsidiary and country characteristics, X it. Specifically, we use financial leverage of firms, the subsidiary country s population to control for market size, and GDP per capita to control for the level of economic development. Notably, we use a number of different types of fixed effects in alternative specifications. These include subsidiary fixed effects; year fixed effects; industry-year fixed effects; country-year fixed effects (that control for time invariant characteristics in the subsidiary countries over our panel s years, e.g., effects common to all affiliates settled in Australia in 2009); and country-pair-year fixed effects (that control for time invariant characteristics in the parent-subsidiary countries over our panel s years, e.g., effects common to all U.S. subsidiaries settled in Australia in 2009). As regards the tax measure, we use the statutory tax rate (see Table 1 for data sources). Our choice is theoretically justified given that multinationals shift profits among subsidiaries 6 We use the variables in logs due to their high skewness (e.g., Hines and Rice, 1994). This practice also limits our sample to subsidiaries with positive earnings before interest and taxes. 7 The choice between EBT and EBIT also depends on the type of industries included in our sample. If, for example, the research design requires each subsidiary to belong in a different industry than its parent firm, as we require in most of this paper, this implies a somehow lower impact of transfer pricing. Consequently, it is preferable to use EBT (it tracks both strategic use of debt and transfer pricing) and leave EBIT as a variable tracing the potency of the transfer pricing channel. 8 Alternatively, we could use other control variables for the firm s size like the log of fixed assets or the number of employees. The results are quantitatively and qualitatively very similar. 15

16 already placed abroad and thus take advantage of any tax allowances in any country in which they operate. Having done so, the advantage in being able to transfer a dollar of profit from a high-tax country to a low-tax country must depend on differences in the statutory rate (for a thorough discussion, see Appendix and Deveraux, 2007). Table 2 reports summary statistics for the subsidiaries (the unit of our empirical analysis) and the parent firms. Our sample includes relatively large parent firms, which operate a number of foreign subsidiaries worldwide. Our choice is dictated from the need to have parent firms with subsidiaries in a relatively large number of countries. In turn, this allows us to empirically exploit differences even among subsidiaries of the same parent firm that reside in different countries. The average parent firm in our sample exhibits pre-tax profits equal to 2.82 billion USD and total assets equal to 36.5 billion USD, while the average subsidiary exhibits pre-tax profits equal to million USD and total assets equal to million USD. Note that the rather high mean value for the parent profits relative to the literature is driven by a few very profitable firms (the median for the parent profits equals to 890 million USD). Further, the average statutory corporate tax rate for the subsidiaries in our sample is 28.3%, varying between 0% and 39.54%. The respective tax rate for the parent firms is 33.1%, ranging between 12.5% and 39.54%. Thus, the average parent firm faces an approximately 5% higher corporate tax rate than the average subsidiary. In total, for 54.3% of the subsidiaries in our sample, the corporate tax rate is lower than the one faced by its parent firm. [Inset Table 2 about here] The distinguishing element in our analysis, is to examine whether our results differ based on the riskiness of the period under consideration or the riskiness of the subsidiary country s macroeconomic and fiscal environment. To this end, we first estimate equation (8) for the crisis 16

17 and the post-crisis periods to exploit the outbreak of the global financial crisis as a period where the riskiness of the macroeconomic environment increases exogenously to the firms managerial activities. Table A3 reports summary statistics for the two sub-periods ( and ). Second, to examine the effect of fiscal risk, we separately estimate equation (8) for the subsidiaries located in countries where the corporate tax rate changed during the last three years (i.e., high fiscal-risk countries) and the ones located in countries where the corporate tax rate remained unchanged during the last three years (i.e., low fiscal-risk countries). In Table A4 we report summary statistics for these two sub-groups. 4. Empirical results 4.1. Profit shifting in the full period We begin by estimating equation (8) for the full period and report the results in Table 3. All specifications are estimated using OLS with robust standard errors clustered by subsidiary. The fit of the models is very good, with R-Squared values higher than 80%. In columns 1 to 4 we use EBT as dependent variable to examine the potency of profit shifting, while in columns 5 to 8 we use EBIT to examine profit shifting due to transfer pricing. We report results from different specifications according to the type of fixed effects (see the lower part of the table). This analysis essentially attempts to replicate the equivalent baseline results of Dharmapala and Riedel (2013). Contrary to our expectations, the results do not support the existence of profit shifting. Across all specifications, the coefficient on the interaction ( ) is statistically insignificant at conventional levels. Thus, this initial analysis shows that we cannot identify profit-shifting behavior of firms during , a result in stark contrast to the existing literature. These results are robust to a series of robustness tests. In Table A5 we examine whether our results are driven by the selection of countries. To this end, we separately run the regressions 17

18 for the EU28 and high-income countries. Again, the results do not provide evidence for profit shifting. We also verify that these results are not driven by the DID model by using the so-called indirect approach (e.g., Huizinga and Laeven, 2008). Table A6 reports the results, which show a very similar picture. In what follows, we show that the discrepancy in our findings compared to the existing literature is indeed due to the increased macroeconomic and fiscal risk during our sample period. [Inset Table 3 about here] 4.2. Profit shifting and the macroeconomic environment We exploit the global financial crisis as a period where the riskiness of the macroeconomic environment increases exogenously to the firms managerial activities. To this end, we first separately examine profit shifting in the crisis ( ) and the post-crisis ( ) periods and report the results in Table 4. Columns 1 to 4 report estimations from the crisis period and show that the profit shifting activity during the years of the crisis is weak. In contrast, profit shifting during the period (columns 5 to 8) is quite potent. The positive sign for the coefficient on Low-tax subsidiary * Parent profit implies that for each dollar of profit that the parent company obtains, a higher fraction goes to the low-tax subsidiaries compared to the hightax subsidiaries. For instance, an estimate of implies that, for a 10% increase in the parent s earnings, the low-tax subsidiaries receive 0.56% more profit than the high-tax subsidiaries. [Inset Table 4 about here] We perform extensive sensitivity analyses on these baseline results. First, to investigate only the transfer pricing channel, we use EBIT as the dependent variable instead of EBT. The 18

19 results in column 1 of Table 5 for show that the coefficient on Low-tax subsidiary * Parent profit is statistically insignificant. On the other hand, in column 5 the equivalent specification for yields a positive and statistically significant coefficient at the 5% level. Thus, transfer pricing is a potent channel of profit-shifting activity only in the period after the global financial crisis. [Inset Table 5 about here] Subsequently, we examine the potency of the debt-shifting channel of profit shifting during and after the crisis. Even though Orbis does not provide data on internal-group debt, and thus we cannot directly detect the debt-shifting channel, we can infer the potency of this channel from the impact of the parent s profitability shocks on the parent s leverage (Dharmapala and Riedel, 2013). Formally, we estimate:, (11) where / and is the fraction of the parent s subsidiaries that are located in a lower-tax country than the tax in the parent s country at time t. Unlike previous specifications, equation (11) is estimated using data for parent firms. A positive and significant coefficient on the interaction term Fraction of low-tax subsidiaries * Parent profit suggests that the higher the parent s earnings, the higher is its debt-to-asset ratio if the parent owns a high fraction of subsidiaries located in countries with lower corporate tax rate than its own (i.e., significant debt shifting). Columns 2 and 6 of Table 5 report the results for and , respectively. The coefficient on the interaction term is statistically significant at 5% for the crisis period and positive and statistically significant at the 1% level during Thus, our results provide 19

20 evidence for debt-shifting behavior during both periods. Nevertheless, the economic significance of these results is small, especially for the crisis period (a 10% increase of the parent s earnings leads to a 0.17% increase of parent leverage if the parent owns a 1% higher fraction of low tax subsidiaries). Unfortunately, this result should be interpret with caution as it might not be purely driven by tax-related motivation. An alternative explanation is that during a crisis there is a credit crunch and multinationals might strategically use intra-group debt to satisfy their funding needs. 9 Further, if book income and tax income are not the same, the parent firm may shift tax income instead of book income, resulting to a downward bias in identifying profit shifting. Following previous studies (e.g., Desai and Dharmapala, 2006; 2009), we examine the sensitivity of our findings by restricting our sample only to subsidiaries operating in countries with high degree of book-tax alignment. The literature identifies Denmark, United Kingdom, and the Netherlands to be such countries (Burgstahler et al., 2006). The results for the period and are reported in columns 3 and 7 of Table 5, respectively and show quantitatively negligible differences compared to our baseline results. 10 We also examine the sensitivity of our results to clustering the standard errors at the country-pair-year level, which is a more restrictive type of clustering. We report the results in columns 4 and 8 of Table 5. Clearly, this sensitivity test does not affect our estimations. Re- 9 To strengthen our evidence for debt shifting during the crisis period, we provide as a robustness test in Table A7 a series of specifications for different country groups (EU 28, high-income countries, G7 countries, and countries with a low degree of divergence between cash-flow rights and voting rights). Results provide evidence for a small but statistically potent intra-debt shifting even during the crisis period. 10 We attribute the drop in the statistical significance from the 5% to the 10% level to the large drop in the sample s size. In line with Atwood et al. (2010), we also attempt a similar sample restriction by dropping countries with the low degree of book-tax alignment (Germany, U.S., and India). The results are again qualitatively the same with our baseline results. 20

21 estimating our baseline specifications with standard errors clustered at the country-year level also yields very similar results. 11 In our analysis so far we assume that the crisis period covers Of course, this might not be the case for all countries, some of which have exited the crisis earlier. We carry out an extensive sensitivity analysis on this issue first using the years as the crisis period and the years as the post-crisis period. The results are reported in Table A9, and once again show insignificant profit shifting in (see columns 1 to 4) and significant profit shifting in (columns 5 to 8). Note that according to the most restrictive specification in column 8, the coefficient estimate decreases to 0.42 from 0.56 in the equivalent column 8 of Table 4. Thus, in line with our theoretical conjectures, we find weaker evidence for profit shifting as we move our post-crisis period closer to the crisis period. Importantly, an argument against our baseline findings on relatively weak profit shifting during the crisis is that our results are driven not by the increase in macroeconomic risk but by lower parent profitability during this period. In Table A10 we re-estimate the baseline specifications of Table 4, restricting our sample only to parent firms that did not exhibit lower return on assets during the crisis period. We present only the two most restrictive specifications with country-year and country-pair-year fixed effects for the crisis and post-crisis periods. Evidently, the results are equivalent to those of our baseline specifications. We next consider the sensitivity of these findings to potential differences in the crisis periods among countries and use the country-specific GDP growth rates to determine macroeconomic risk. Specifically, we define as high-risk countries those with negative GDP 11 As an additional sensitivity test, in Table A8 we re-estimate the baseline specifications of Table 4 but this time excluding the subsidiaries owned by a U.S. parent firm. The goal of this tests is to examine whether results are driven solely by U.S. parent and subsidiary firms, as the U.S. is the country that contributes the most observations in our sample. The results are qualitatively similar to our baseline ones. 21

22 growth rates in a given year and as low-risk countries those with positive GDP growth rates. This practice only considers the effect of de facto being in a recession and not the perception that 2008 to 2010 were very risky times globally. Also, based on our previous findings, we use only the period , during which we identify significant profit-shifting behavior of firms. Table 6 reports the estimation results. Columns 1 to 4 show estimations for the high-risk countries and columns 5 to 8 for the low-risk countries. Evidently, the coefficient on Low-tax subsidiary * Parent profit is lower and less statistically significant for the high-risk countries. Thus, the results show that, even within the post-crisis period, country-specific macroeconomic risk in the subsidiary s country is quite relevant in explaining the relatively low level of profitshifting activity. [Inset Table 6 about here] In Table 7, we conduct a series of robustness tests for the results of Table 6. The results in columns 1 and 5 show significant evidence for transfer pricing only in countries with positive growth rates, while the equivalent in columns 2 and 6 provide evidence for debt shifting. Finally, in columns 3 and 7 we restrict our sample to countries with high degree of book-tax alignment and in columns 4 and 8 we cluster the standard errors by country-pair-year. Clearly, the results show profit shifting-behavior only in low macroeconomic-risk countries. [Inset Table 7 about here] 4.3. Profit shifting and fiscal risk Intuitively, the frequency in which the corporate tax rate changes in a country is a factor that has a direct impact on the business environment: it increases the fiscal uncertainty that multinationals 22

23 face and therefore reduces the expected gains from profit shifting. In this section, we empirically examine whether this type of fiscal policy affects the intensity of profit shifting. We mainly use the period , in which we find potent profit shifting, and ask whether subsidiaries located in countries where the corporate tax rate remains unchanged for a reasonably long time receive higher profit shifting compared to subsidiaries located in countries where there are frequent tax changes. We define as low fiscal-risk the countries where the corporate tax rate remained unchanged over the last three years and as high fiscal-risk the countries where the corporate tax rate has changed during the last three years. 12 [Inset Table 8 about here] We report our baseline estimation results in Table 8. Columns 1 to 4 report estimations for the high fiscal-risk countries and show that profit shifting activity is weak, especially as we use the more restrictive fixed effects models. In contrast, profit shifting for the low fiscal-risk countries (columns 5 to 8) is quite potent, irrespective of the type of fixed effects used. The coefficient estimates are also quite stronger compared to the equivalent ones in previous tables. Based on the model with the most demanding fixed effects (column 8), if the parent firm experiences an earning shock of 10%, the low-tax subsidiaries report approximately 0.92% higher profits than the high-tax subsidiaries. This estimate is considerably higher compared to the equivalent 0.56% in column (8) of Table 4. Thus, our results confirm the crucial role of fiscal risk in limiting profit shifting, even inside stable economic periods. This finding is interesting from a policy perspective because it shows that even countries with low tax rates and low macroeconomic risk may not be good 12 For example, for the observations in 2012, the low fiscal-risk countries are those in which the corporate tax rate remained unchanged in the period Note that with this measure of fiscal risk, we include both the countries that increase their tax rates and those that decrease it. We note in our sample that there is a high correlation between within country increases and subsequent decreases and vice versa. Thus, even tax decreases can contribute to fiscal uncertainty because it is highly possible that they were preceded or are followed by increases. 23

24 candidates for considerable volumes of profit shifting, and a pre-condition in attracting these volumes is corporate tax rate stability. On this line, our results extent the list of the tax-havens characteristics discussed by Dharmapala and Hines (2009): besides small size, affluence, and high quality of governance, a potential tax haven should also provide stability in corporate tax rates. These results are robust to a series of sensitivity tests. First, to investigate only the transfer pricing channel, we use EBIT as the dependent variable instead of EBT. The results in column 1 of Table 9 for the high fiscal-risk countries show that the coefficient on Low-tax subsidiary * Parent profit is relatively small (i.e., 0.030) and statistically insignificant. On the other hand, in column 5 the respective coefficient for the low fiscal-risk countries equals 0.081, exhibiting higher economic and statistical significance. 13 Next, we examine the potency of the debt-shifting channel of profit shifting. We re-estimate equation (11) separately for the high fiscal-risk countries (column 2 of Table 9) and low fiscal-risk countries (column 6 of Table 9). In both cases the coefficient of Fraction of low-tax subsidiaries * Parent profit is positive and significant at the 1% level, providing evidence for debt shifting. Thus, in all specifications that consider the parent firm s debt, there exists indirect evidence for debt shifting even during the crisis period. However, as in the case with macroeconomic risk, the coefficient estimates in columns 5 and 6 show that transfer pricing is more potent than debt shifting. [Inset Table 9 about here] In the rest of the columns of Table 9, we cluster the standard errors by country-pair (columns 3 and 7) and by country (columns 4 and 8). Evidently, in both cases the significance of our results is not affected. In turn, in Table A12 we use a four-year period of corporate tax 13 In Table A11 we replicate the full results of Table 8 using EBIT as the dependent variable. Consistent with previous findings, the results show no evidence for transfer pricing for the high fiscal-risk countries and show considerable transfer pricing for the low fiscal-risk countries. 24

25 stability (instead of a three-year period used in the previous specifications). Once again, changes in the results compared to Table 8 are relatively small. As a final sensitivity test, we replicate Table 8 for the period of instead of and report the results in Table A13. Our results are again consistent with those of Table Conclusions With an aim to examine the distribution of profit shifting of MNEs across subsidiaries in countries with different characteristics, we place the role of the macroeconomic and fiscal environment at the heart of this debate. We use data for subsidiaries of MNEs operating in 43 countries, parent firms from 24 countries, and a maximum of approximately 36,200 subsidiaryyear observations for the period Our identification method follows a well-established DID approach, which looks at earnings shocks at the parent level and examines their propagation toward subsidiaries operating in a different industry and country. Our results using the full sample suggest limited profit-shifting activity. In contrast, when differentiating between periods of high and low macroeconomic and fiscal risk in the countries of subsidiaries residence, we identify quite potent profit-shifting activity. Economically, the results show that an increase in the parent profits by 10% leads to 0.56% higher reported earnings before taxes for the subsidiaries during the post-crisis period of Moreover, a 10% increase in the parent firm s profits, results to a 0.81% higher earnings before taxes for the subsidiaries located in low country-specific macroeconomic-risk. Finally, and perhaps most importantly, a 10% increase in the parent firm s profits increases by 0.92% the earnings before taxes of the subsidiaries located in low fiscal-risk countries within periods of a stable macroeconomic environment. 25

26 The implications and directions of future research emerging from our results are then straightforward. First, to identify profit shifting, researchers and policy-makers need to perhaps focus on low-risk countries, especially as regards fiscal risk. In future research, other types of risk, such as institutional risk, could also be shown to play an important role in limiting profitshifting activity. Related to this, any policy action to countervail profit-shifting of MNEs should target low-risk countries and especially transfer-pricing practices. Based on our findings, the extent of other types of tax-avoidance practices due to the riskiness of subsidiaries countries is also a fruitful direction for future research. Finally, profit shifting seems indeed to be underestimated when global samples of firms are used that do not account for the immense differences in subsidiaries countries. More work is needed to further refine our results in terms of profit shifting toward tax heavens or in countries with very low quality of relevant institutions. Given that we have covered considerable ground in this paper, we leave these ideas for future research. 26

27 References Atwood, T.J., Drake, M., Myers, L., Book-tax conformity, earnings persistence and the association between earnings and future cash flows. Journal of Accounting and Economics 50, Bartelsman, E. J., Beetsma, R. M Why pay more? Corporate tax avoidance through transfer pricing in OECD countries. Journal of Public Economics 87, Bertrand, M., Mehta, P., Mullainathan, S., Ferreting out tunneling: an application to Indian business groups. Quarterly Journal of Economics 117, Buetner, T., Overesch, M., Schreiber, U., Wamser, G., The impact of thin-capitalization rules on multinationals financing and investment decision. Journal of Public Economics 96, Budd, J. W., Konings, J., Slaughter, J., Wages and international rent sharing in multinational firms. The Review of Economics and Statistics 87, Burgstahler, D.C., Hail, L., Leuz, C., The importance of reporting incentives: earnings management in European private and public firms. The Accounting Review 81, Clausing, K. A., The impact of transfer pricing on intrafirm trade. In J. R. Hines, Jr., (ed.) International taxation and multinational activity. Chicago: University of Chicago Press, Clausing, K. A., Tax motivated transfer pricing and US intrafirm trade prices. Journal of Public Economics 87, Collins, J., Kemsley, D., Lang, M., Cross-jurisdictional income shifting and earnings valuation. Journal of Accounting Research 36, Desai, M.A., Dharmapala, D., Corporate tax avoidance and high-powered incentives. Journal of Financial Economics 79,