KNOCKING ON TAX HAVEN’S DOOR: MULTINATIONAL FIRMS AND TRANSFER PRICING Ronald B. Davies, Julien Martin, Mathieu Parenti, and Farid Toubal*

Abstract—This paper analyzes the transfer pricing of multinational firms. intrafirm, is generally not available. Moreover, it is impos- Intrafirm prices may systematically deviate from arm’s-length prices for two motives: pricing to market and tax avoidance. Using French firm-level sible to observe the counterfactual arm’s-length prices of an data on arm’s-length and intrafirm export prices, we find that the sensi- intrafirm transaction (see Diewert et al., 2006, for details). tivity of intrafirm prices to foreign taxes is reinforced once we control for Since the arm’s-length price is not observed, tax authorities pricing-to-market determinants. Most important, we find no evidence of tax avoidance if we disregard tax haven destinations. Tax avoidance through have to determine the market price, which raises obvious transfer pricing is economically sizable. The bulk of this loss is driven by definitional and methodological issues. the exports of 450 firms to ten tax havens. In this paper, we overcome both difficulties. We observe the export prices under each mode (arm’s length or intrafirm) I. Introduction at the level of firms, countries, and products. Moreover, our econometric methodology allows us to compare the intrafirm wealth of empirical evidence finds that within a multi- price used between a multinational and its affiliate with the Anational enterprise (MNE), reported profits vary sys- corresponding arm’s-length price charged by a firm shipping tematically with local corporate tax rates.1 This may be due to an unrelated party. We show that the bulk of tax avoidance to several types of efforts within the firm, including transfer comes from a few large firms through exports in a handful pricing. From the perspective of the tax authorities, inter- of industries to a relatively limited number of tax havens, nal transactions between related parties should be valued where the baseline estimates find that intrafirm prices are on at their market price: this is the arm’s-length principle (see average 11% lower than arm’s length prices. This suggests OECD, 2012, for details). That said, as described in OECD that by targeting enforcement efforts, tax authorities may be (2010), there are numerous ways to determine the arm’s- able to mitigate transfer pricing and raise tax revenues while length price, including the use of comparable prices and keeping enforcement costs low. The granular dimension of cost-plus methods. Thus, the flexibility in these rules allows tax avoidance should facilitate the implementation of global firms to choose transfer pricing methodologies that support enforcement such as the one proposed by Zucman (2014). the use of internal prices, shifting profits from high- to low- In order to frame our empirical analysis, we discuss tax countries.2 This is in addition to the potential for outright the theoretical determinants of arm’s-length and intrafirm tax evasion via transfer pricing. prices using a highly stylized model. This simple model Direct empirical evidence of tax-induced transfer pricing, captures both tax-induced transfer pricing and pricing-to- however, is scarce. Identifying such a strategy faces two market strategies. The latter has been receiving increasing major difficulties. While multinationals’ exports are directly attention in the field of international trade. The model shows observable, detailed information on the prices of products that were transfer pricing costless, an MNE would systemat- and their modes of transaction, whether it is arm’s length or ically find it optimal to deviate from the arm’s-length price exporting to a country with a different corporate tax. How- Received for publication September 30, 2015. Revision accepted for ever, the presence of a fixed component in the concealment publication December 2, 2016. Editor: Amit K. Khandelwal. cost generates a band of inaction. The deviation is profitable * Davies: University College Dublin; Martin: Université du Québec à Montréal, and CEPR; Parenti: Université Libre de Bruxelles, and CEPR; when the tax differential between the home and the host Toubal: Ecole Normale Supérieure de Cachan, CREST, and CEPII. countries is sufficiently important and when the firm exports We thank the editor and three anonymous referees for their very con- sufficiently large volumes. The wedge between the intrafirm structive suggestions. We also thank Andrew Bernard, Kristian Behrens, Robert Cline, Paola Conconi, Anca Cristea, Meredith Crowley, Lorraine price and the arm’s-length price is a decreasing function of Eden, Peter Egger, Lionel Fontagné, Thomas Gresik, James Hines, Patrick the host tax. We also show that arm’s-length and intrafirm Legros, Lindsay Oldenski, Bernard Sinclair-Desgagné, Nicholas Sly, and prices are likely to have a different sensitivity to transport many seminar audiences for helpful comments. A supplemental appendix is available online at http://www.mitpress costs, tariffs, and GDP per capita—variables governing pric- journals.org/doi/suppl/10.1162/REST_a_00673. ing to market. These results suggest that one should be 1 See Fuest, Huber, and Mintz (2003) for a survey on the impact of taxation mindful of this difference in sensitivity in the empirical on real MNE activity. analysis. If one of these variables is significantly correlated 2 The OECD reaches the conclusion that “tax administrations see trans- fer pricing as one of the most significant tax risks they have to manage” with the level of corporate tax rates (something that is true (OECD, 2012, p. 15). Gresik (2001) provides a survey with an emphasis on in our data), the estimated coefficients would be biased if the transfer pricing literature. Diewert, Alterman, and Eden (2006) examine we were not allowing coefficients to differ across pricing the different rationales for manipulating internal prices. A meta-analysis by Heckemeyer and Overesch (2013) shows that transfer pricing and licens- modes. ing are two important means of shifting profits abroad. Recent theoretical On the empirical side, we rely on a unique data set that has contributions on tax-induced transfer pricing include Behrens, Peralt, and finely grained information on the intrafirm and arm’s-length Picard (2014), Bernard, Jensen, and Schott (2006), and Keuschnigg and Devereux (2013), who provide a recent model of non-tax-induced transfer quantities and prices of exported products at the firm level pricing, one motivated instead by manipulating managerial incentives. for almost all exporting firms in France in 1999. The richness

The Review of Economics and Statistics, March 2018, 100(1): 120–134 © 2018 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology doi:10.1162/REST_a_00673

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 121

of this data set allows us to provide a clear-cut identification directed to countries with very low tax rates. Interestingly, of transfer pricing in the cross-section. low taxes are not the entire story. Transfer pricing strate- The case of France is particularly suited to analyzing the gies of multinational firms systematically involve tax havens. transfer pricing strategies of MNEs as it exempts foreign These countries not only have low corporate tax rates but income from taxation. Compared with the United States and also provide an overall tax environment that facilitates profit other countries where foreign tax credits, income baskets, shifting. According to the classification of Hines and Rice and deferral complicate a firm’s tax planning problem, the (1994), in addition to a low tax rate, a tax haven must have relatively streamlined French system provides a cleaner map- a legal system allowing banking secrecy and a good com- ping between tax differences and firm incentives.3 Roughly munication infrastructure, and it must seek to promote itself speaking, when a U.S. firm earns profits overseas, it adds up as a center for financial offshoring. Thus, as the OECD’s its worldwide income into a single income basket and cal- (2013) Action Plan on Base Erosion and Profit Shifting strives culates the U.S. tax owed on this amount when it repatriates to clarify, a tax haven is not simply a low-tax country but these foreign earnings. The U.S. tax authorities then grant a one that facilitates tax avoidance for firms by “artificially credit against this liability that is equal to the taxes already segregating taxable income from the activities that gener- paid to foreign governments on the firm’s overseas income. ate it” (p. 13). Extending our analysis, we find that profit If the firm has paid more taxes overseas than what is owed to shifting through transfer pricing is primarily done by large the United States, it is in an excess credit position and owes multinational firms.4 no taxes to the United States. If not, it is in an excess limit A simple exercise suggests that the tax losses driven by case and must pay the remainder to the U.S. tax authorities. the profit shifting of multinational firms to the ten tax havens Thus, even when there are two U.S. firms with an affiliate in our sample amount to about 1% of total corporate tax in a tax haven, the incentive to shift profits to the tax haven revenues in France. We further show that 450 MNEs account depends on where else the firm pays taxes. Complicating the for over 90% of intrafirm exports to these ten tax havens. issue, the U.S. tax liability is not triggered until profits are Together, these imply that a large share of transfer pricing repatriated or used inactively (i.e., it is no longer Subpart may be curbed by focusing enforcement on a small number F income), and excess credits in a year can be carried into of firms. the future or applied retroactively, introducing a dynamic Although there is an extensive literature on the impact aspect to the firm’s profit-shifting problem. In an exemption of international tax differences on the location of profits system such as France’s, none of these considerations arise, and firms, the results of which are suggestive of transfer meaning that a much simpler comparison of the French and pricing, there is little evidence of the impact of taxes on destination tax rates describes the profit-shifting motives. transfer prices themselves.5 Bartelsman and Beetsma (2003) We exploit the rich structure of the data set to identify the use aggregated data on value-added across manufacturing impact of foreign taxes on the transfer pricing behavior of sectors in the OECD. They estimate a value added func- multinational firms. We propose a difference-in-difference- tion depending on corporate tax rates and other factors, like strategy that allows us to compare the intrafirm prices finding results suggestive of profit shifting via transfer pric- charged by a particular firm for a particular product across ing. Clausing (2003) uses price indices for U.S. exports and markets with the arm’s-length prices of exports for the same imports that include separate indices for intra- and inter- product exported in the same markets. Indeed in our analysis, firm prices, finding a strong and significant impact of taxes the arm’s-length price of a product is firm market specific. In consistent with transfer pricing. Using an approximation of addition, we include several pricing-to-market variables that intrafirm trade from firm-level balance sheet data, Overesch the trade literature identifies as important (income, distance, (2006) finds that the value of German MNEs’ intrafirm trade and tariffs) and allow the sensitivity of free on-board prices varies with the difference between the German tax rate and to these variables to differ across modes. Controlling for a that of the foreign parent or affiliate’s location. Using firm- set of triadic fixed effects at the firm, product, and export product level data, Swenson (2001) finds that prices react mode levels, our strategy captures any difference between to taxes and tariffs. Using U.S. transaction data, Neiman intrafirm prices and their arm’s-length counterparts, which is (2010) shows that intrafirm prices are less sticky and exhibit systematically related to the corporate tax in the destination more pass-through than arm’s-length prices. He also docu- country. ments that the specific pass-through of exchange rates into In line with the main theoretical prediction, our estimates intrafirm prices is not affected by the tax level in the origin suggest that export prices drop with the destination cor- country. This suggests that the dynamic of intrafirm prices is porate tax rate only for intrafirm transactions. This result not primarily driven by the desire to shift taxable income to is robust once we control for pricing to market. We then low-tax countries. It is worth noting that the analysis treats show that above a certain threshold, differences in the cor- porate tax rates have no effect. Transfer pricing is essentially 4 Consistent with these findings, Gumpert, Hines, and Schnitzer (2011) show that the largest German multinationals are the most likely to use tax havens. 3 Bernard, Jensen, and Schott (2006) provide a discussion of the complex- 5 For a recent discussion of the former, see Huizinga and Laeven (2008). ity of the U.S. tax system. Dharmapala and Riedel (2013) discuss the impact of taxes on firm financing.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 122 THE REVIEW OF ECONOMICS AND STATISTICS

tax havens and other countries with a lower corporate tax compare within-firm prices for a product destination, requir- rate than the U.S. symmetrically.6 ing the MNE to export the product both intra- and interfirm Two recent papers, Vicard (2015) and Cristea and Nguyen to the destination. Rather than working with price gaps, we (2016), exploit the panel dimension of firm-level data on work with prices and identify the effect of taxes using an French and Danish firms, respectively. Both papers tend interaction term.8 To make inter- and intrafirm prices com- to provide evidence of transfer pricing. However, they do parable, we control for firm-product-mode fixed effects and not observe intrafirm and arm’s-length prices but instead allow these prices to differ for pricing to market motives. assume intrafirm prices for transactions with countries where Our methodology allows us to get a narrow benchmark a related party is located. As Ramondo, Rappoport, and Ruhl market price while retaining more intrafirm transaction. (2016) and Atalay, Hortacsu, and Syverson (2014) men- Moreover, we avoid the complication highlighted by Cristea tioned most firms with an affiliate in a country do not trade and Nguyen (2016) in which MNEs selling both intrafirm with this affiliate. Furthermore, a firm that exports a product and at-arm’s-length price at arm’s length differently than to its affiliates might also well export another product to a non-MNEs do (potentially for tax reasons). third firm in the same country. In our sample of firm-country The rest of the paper is organized as follows. In section pairs where we observe positive intrafirm trade, the share of II, we discuss the theoretical determinants of transfer pric- intrafirm trade in a firm’s total trade is below 40% for one- ing in order to guide the empirical analysis. In section III, fourth of the observations. Our empirical tests rely on precise we carefully present the data and the estimation sample con- firm-level data on intrafirm and arm’s-length prices. struction. In section IV, we present the baseline estimation, Finally, our paper is related to the work of Bernard et al. extend the results, and provide a quantification exercise esti- (2006), who examine how internal prices depend on taxes mating the revenue loss for France due to transfer pricing. and tariffs using U.S. firm-level data. Similar to the papers We conclude in section V. already mentioned and to our own results, their estimates are consistent with transfer pricing. That said, our results con- II. Model tribute to the debate in a significant way by examining the granularity of transfer pricing. We do so along three dimen- We present a simple model that illustrates how taxes and sions. First, and most important, we consider destination tax other factors, such as trade costs or GDP per capita, influ- haven status as well as tax rates. In addition to providing low- ence the pricing strategy of a multinational firm (MNE) and tax rates, tax havens offer MNEs a tax avoidance–friendly compare it to the pricing strategy of an exporter. Throughout environment. Since we find that internal and arm’s-length this section, we denote with an asterisk the variables that are prices deviate most in exports to tax havens, this is critical. the outcome of firms’ profit maximization. Second, we examine whether all MNEs use transfer pricing, The MNE produces a good in country H (Home) at zero finding that the intensity of profit shifting is systematically cost and ships it to its affiliate located in F (Foreign) at H greater for larger firms. Third, we show that such transfer a free-on-board price pMNE. Both countries levy taxes on a H pricing is concentrated in differentiated products where it is territorial basis, where the home tax is T and the foreign tax F difficult to compare products across firms.7 Together, these is T , consistent with the tax-exemption rule for the income indicate that rather than being common practice among all earned abroad by French MNEs. MNEs, transfer pricing is primarily found only among the For simplicity, we assume the domestic and foreign sales largest firms and then mostly between the parent and its tax are separable. Exports incur trade costs that are specific to the haven affiliates. This concentration of transfer pricing has destination market F. Specifically there is an ad valorem cost τF H been heretofore unrecognized in the literature. v pMNE, such as a tariff or insurance. Finally, we assume that Beyond this, there are other key differences in terms of the production of the final good is costless. The foreign con- F both data and methodology. First, by using data on firms sumer price is denoted by pMNE, and the individual demand F = α − βF F based in France rather than in the United States, we avoid the for that good is d pMNE : 1 pMNE . complications in taxation introduced by the U.S. foreign tax There is a unit mass of consumers in F.9 Each of them credit system. Second, the comparable market price in our has a willingness to pay for the first unit of the good study is built from the prices of firms that are not selling inter- equal to 1/βF . The parameter α is a demand shifter that nally (including multinationals operating at arm’s length and exporters). Bernard et al. (2006) examine the gap between 8 Our results are not directly comparable to those of Bernard et al. (2006) because of this. They regress the price gap on the logarithm of the tax rate. the intrafirm price of a product and a benchmark arm’s- We regress the logarithm of prices on an intrafirm dummy, the logarithm length price that is charged by the same MNE. They therefore of 1 minus the tax rate, and their interaction. The impact of a 1 percentage point decrease in tax rate in the median country leads to an increase in the price wedge by 0.6% in our sample. This effect varies with the initial tax 6 Our results are robust to controlling for the growth or the volatility of rates. By contrast, Bernard et al. (2006) estimate that a 1 percentage point bilateral exchange rates. Results are available on request. decrease in the tax rate is associated with an increase in the price wedge 7 Bernard et al. (2006) show that the price wedge between intrafirm and ranging from 1.6% to 4.2%. arm’s-length transactions depends on the size of exporters and the differen- 9 This rules out differences in market size across countries. Since these tiation of products. However, they do not study whether the sensitivity of differences could matter for pricing-to-market strategies, they are controlled the price wedge to corporate taxes depends on these characteristics. for in the empirical analysis.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 123

is firm specific (e.g., the appeal for that firm’s product). It H∗ = 1 optimal price is p βF +τF . As expected, the free-on- α¯ 2 (1 v ) is described by a distribution with support [0, ]. On the board price set by the firm depends on the characteristics βF contrary, is common to all firms in destination F.For of the foreign market, such as GDP per capita, tariffs, and, βF instance, is lower in countries with a higher GDP per more generally, trade costs. The arm’s-length price of a given capita. product is therefore country specific. To determine the transfer price set by a multinational, we use the popular concealment cost approach to modeling 10 H B. Transfer Pricing transfer pricing. In this, the transfer price pMNE can differ ∗ from the price pH that would be set by an exporter with Under a territorial tax system, the after-tax profits of a the same marginal cost (equal to 0) selling to an unaffiliated multinational are given by party (i.e., at arm’s length). This incurs a cost       πF = − H H F − Φ  H − H∗   MNE : 1 T pMNEd pMNE pMNE p Φ  H − H∗        pMNE p + − F F − + τF H F  1 T pMNE 1 v pMNE d pMNE . γF   pH − pH∗ 2 + f if pH = pH∗ (1) := 2 MNE MNE , H = H∗ 0ifpMNE p H Profits are maximized by choosing the transfer price pMNE and the price of the final good pF . The maximization of γF MNE where is a parameter that reflects the stringency of the equation (1) with respect to the transfer price leads to the fiscal regime enforcement in Foreign. This function is a first-order condition, tax-deductible cost that occurs in the home country.11 This concealment cost is generally interpreted in transfer pricing   γF   d pF = pH∗ − pH ,(2) models as the cost of hiring accountants to “cook the books” MNE θF [T F ] MNE or the fines that the firm would pay if it were caught. We     allow these costs to entail both a fixed component denoted θF [ F ] = θ F H τF = 1−T F + τF − where T : T ; T , v : 1−T H 1 v 1. by f and a quadratic variable cost that is a function of the A firm finds it optimal to shift a share of its domestic price gap.12 Note that this might vary across countries; for profits to Foreign only if θF [T F ] > 0. When it does, it sets H∗ H∗ θF [ F ] example, if a tax haven makes this concealment relatively its price pMNE < p . Note that T > 0 whenever the easy (low γF ), then this could reduce the total and marginal corporate tax abroad is lower than at home (T F < T H ). This costs of concealment for a given price wedge. For simplic- is a standard result in the transfer pricing literature. We want τF ity and without loss of generality, we assume that the fixed to stress that multiplicative trade costs v (e.g., tariffs) are component does not depend on the characteristics of F. also a motive for transfer pricing and that the two effects We use this stylized model to study the variations of the interact with each other. pH∗ θF [ F ] price wedge ∗ within a market across firms and within Importantly, however, T > 0 is a necessary but pH MNE not a sufficient condition. Engaging in profit shifting is a firm across markets with a different corporate tax T F and costly: depending on the magnitude of the corporate tax gap concealment cost parameter γF . In order to guarantee that between its home and destination countries, a firm may not firms never find it optimal to set a negative intrafirm price, find it profitable to do so. Furthermore, even when the con- we assume throughout that γF is large enough.13 ditions are favorable (low T F ,lowγF ), all firms need not engage in profit shifting, as we shall see. Considering the A. Pricing to Market F second first-order condition of an MNE with respect to pMNE We start by solving the optimal price pH∗ set by an yields the following equation:     arm’s-length exporter. The profits in Foreign are given by   F∗ H   ∗ 1 β 1 − T ∗ πF := pH · d (1 + τF )pH . It is readily verified that the d pF = α + − 1 + τF pH . AL v MNE 2 2 1 − T F v MNE 10 This approach initiated by Kant (1988) is predominant to model transfer Plugging equation (2) into the above equation and rear- pricing (e.g., Haufler & Schjelderup, 2000 and Huizinga & Laeven, 2008). 11 This assumption is made for simplicity and does not change any of the ranging the expression leads to the optimal price wedge: insights from the model. α·θF F 12 Ernst and Young (2012) suggest that prices are one of the issues that H∗ F + τF · (T ) p β θ + 2 (1 v ) γF receives “the greatest scrutiny” while volumes are not mentioned (p. 63). = 1 + · · . (3) Of course, if a firm ever pays a penalty and adjusts its price, it will have to H θ + F F α·θF (T F ) pMNE 2 1 1 − β (1 + τ ) · pay taxes accordingly, which are a function of the quantities sold. Never- v γF theless, our result requires only that a large price gap with small quantities be more likely to catch attention than a small price gap with large quanti- Since θF [T F ] is decreasing in T F , taking the first partial    δ Φ H − H∗ F F ties. Formally, if (.) is a function of pMNE p d pMNE , our result derivative of equation (3) with respect to T shows that the holds for δ < 1/2. price wedge is larger when T F is lower. Similarly, holding 13 γF α¯ · βF + τF × F F Specifically,  this condition reads as > (1 v ) γ F T constant, a lower is associated with lower intrafirm 1−T 1 + τF − 1 . 1−T H v prices. Tax havens provide firms with an additional motive

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 124 THE REVIEW OF ECONOMICS AND STATISTICS

for profit shifting via transfer pricing. In the absence of a any effect on arm’s-length prices. Second, for a given corpo- motive for transfer pricing—when θF [T F ]=0 or when the rate tax rate in the destination country, a lower concealment concealment cost γF is infinitely large—intrafirm prices are cost should be associated with lower intrafirm prices. In H∗ = H∗ equal to arm’s-length prices pMNE p . other words, we expect profit shifting via transfer pricing toward tax havens. Third, the price wedge is a function of destination-specific income and trade costs, so there should C. Firm Heterogeneity and the Inaction Band be differences in the impact of these market-specific char- For equation (2) to hold, however, it must be the case that acteristics (transport costs, tariffs, and GDP per capita) on firms actually choose to engage in profit-shifting activities. arm’s-length prices and transfer prices. Fourth, the price Now, since shifting profits through transfer pricing involves wedge should be larger for larger firms. a fixed cost, only a subset of firms finds it profitable to do ∂πF∗ [α] ∂πF∗[α] MNE AL α III. Data and Identification Strategy so. Formally we have ∂α > ∂α as soon as > 0, so there is a unique cutoff such that πF∗ [αˆ F ]=πF∗[αˆ F ]. MNE AL A. Data Description This cutoff is destination specific and higher when the for- eign corporate tax rate is higher (see the online appendix for To investigate the factors driving transfer pricing, we use details). It delimits an inaction band: there is no profit shift- detailed cross-sectional information on intrafirm and arm’s- ing toward destinations with a high cutoff (formally αˆ F > α¯), length export prices for a set of French firms in 1999. France such as destinations with a small corporate tax differential. issued its first tax rule with respect to transfer pricing docu- To sum up: mentation requirements in 1996.14 What is more, 1999 is the last year before which advanced pricing agreements (APAs) Proposition 1. The deviation of intrafirm prices from started to be granted by French tax authorities. Such an arm’s-length prices is larger in destinations with a lower agreement allows a multinational firm to negotiate an under- corporate tax rate. When a fixed cost must be incurred to standing with one or more tax authorities that approves a engage in profit shifting, there is an inaction band: multina- transfer pricing methodology for a given period, resolving tional firms choose arm’s-length pricing when the corporate the uncertainty about its acceptability and reducing audit tax differential (or tariff) with the destination country is small. risk.15 These agreements do not, however, concern the firms in our sample as the French tax authorities did not grant Importantly, even in this simple model, the marginal any APAs in 1999. In other words, the year 1999 is ideal to impact of T F and γF on the price wedge varies across des- identify transfer pricing practices. tinations. Expression (3) implies also that a higher GDP per In order to construct our estimation sample, using a unique capita (i.e., a lower βF ) has a positive impact on the price firm identifier, we combined three data sets that have detailed wedge. In other words, the sensitivity of arm’s-length and information on the firm-level export values and quantities of intrafirm prices with respect to trade costs and GDP per eight-digit product categories by destination, and data on capita generally differs. Failure to account for this can poten- MNE status. These data sets also provide information on tially result in misleading estimates of the impact of taxes whether a transaction is intrafirm or arm’s length. We merge on the difference between intra- and interfirm prices. these data sets with information on country-level character- Now, considering a destination such that α¯ > αˆ F , all firms istics such as the level of the corporate tax rate, distance, with a type above the threshold set their price below the tariffs, and per capita income. arm’s-length price according to equation (3). Thus, condi- tional on profit shifting, firms facing a higher demand curve Firm-level data. Our first data set comes from French abroad (high α) exhibit larger deviations from the arm’s- Customs, which reports the yearly free-on-board values and length price. Since equation (2) implies a positive correlation quantities of exports by firm, eight-digit CN product cat- relation between the volume exported to a destination by a egory, and destination. The data set is quasi-exhaustive as multinational and the price wedge, we can conclude that a declaration threshold of 1,000 euros for annual exports firms with a higher α also export higher volumes. applies to any given destination. Extra-European shipments under 1,000 euros are subject to a simplified declaration Proposition 2. In a given destination, larger intrafirm trade procedure and do not appear in our data. Within the single flows come with larger deviations of intrafirm prices with European market, firms are not required to submit the regular respect to arm’s-length prices. 14 Article 57 of the French tax code contains the main French legal pro- visions on transfer pricing. It states that in assessing the income tax owed Thus, the deviation from the arm’s-length price is increas- by French taxable entities that are controlled by or control entities outside ing in firm size, presuming that, consistent with the data, France, any profits indirectly transferred to the latter, whether by an increase larger firms export more. This exercise yields a set of pre- or decrease in purchase or sales prices or by any other means, shall be added back to the taxable income. dictions for us to take to the data. First, a lower destination 15 See Borkowski (2000) for a description of the APAs’ program in France tax rate should lower the intrafirm price but should not have in 1999.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 125

customs form. The reporting threshold is based on each When the INSEE survey indicates that an HS4 category has firm’s cumulated yearly export value (all destinations within a share of intrafirm exports exceeding 99%, we classify all the EU). The declaration threshold for European countries’ corresponding CN8 exports by MNEs as intrafirm transac- export flows is higher, at around 150,000 euros in 1999. We tions. If the share is less than 1%, we classify the CN8 codes use the value and the quantity of firms’ exports of a given as arm’s length. When the share is positive but below 99% product (CN8) to a given destination in order to construct and above 1%, we drop the observations for this firm for this the destination- and firm-specific free-on-board unit values, destination-HS4 dyad, as we cannot assess whether the cor- our proxy for the price a firm charges for that product in a responding CN8 transactions are intrafirm or arm’s length. given market. We drop 3% of observations accounting for roughly 11.5% This data set, however, does not provide information on of the total value of French exports because of this.18 There- the export mode, that is, whether a transaction is intrafirm fore, the identification of transfer pricing does not rely on or arm’s length. We obtain this information from a confi- the comparison between intrafirm and arm’s-length prices dential INSEE firm-level survey on the foreign activities of charged by the same firm within a destination. French multinational firms. For budget reasons, the survey Furthermore, in some destinations, we observe only the was taken only in 1999 for all French firms with trade worth intrafirm prices of a product. Our strategy relies on the com- more than 1 million euros.16 In the survey, a firm is part of or parison of arm’s-length and intrafirm prices; we thus have is itself a group that controls at least 50% of the voting rights to exclude these product-destination pairs. Out of 208,904 of a firm outside France. Hence, all the firms have at least product-destination pairs, we drop 6,302 pairs for which we one related party abroad and can be considered MNEs. A observe only intrafirm trade prices. They account for 1.2% French intragroup transaction is thus defined as trade with a of French exports and 4.8% of intrafirm exports.19 directly or indirectly affiliated related party controlled by the This then leaves us with information at the firm, CN8 group. Intrafirm trade does not include trade with firms that product, country, and exporting-mode level. Once merged share a licensing agreement or other nonownership arm’s- with country characteristics, there are 729,737 observations length arrangement. The INSEE survey provides a detailed in our unbalanced baseline sample.20 Our cross-section is geographical breakdown of French MNEs’ export values and composed of 64,633 firms, 5,457 products, and 45 countries. quantities at the product level (HS4) and of their exporting About 9.2% of the total number of observations are intrafirm modes through outside suppliers or related parties. The data- prices. base does not have information on the foreign affiliates. We It is worth emphasizing that most of the prices set by then merge the Customs data with the INSEE survey. This is MNEs are not intrafirm prices. In this sample, only 15.7% a straightforward process because firms are identified by the of the prices set by MNEs are intrafirm prices, accounting for same ID number in the two data sets, and there is a direct 56.5% of their exports. Another interesting fact is that in our mapping between French CN8 product categories and HS4 data, one-third of multinational firms do not report intrafirm product categories.17 trade in countries where they have affiliates (or headquar- We use information from LIFI, a French firm–level data ters) according to LIFI. A last fact pertains to the likelihood set on financial linkages between firms. This is used to deter- that we observe both arm’s-length and intrafirm trade for a mine whether a firm in the French Customs data is an MNE multinational firm exporting to a given country. To study this and, if so, its nationality and the country locations of its point, we restrict the sample of firm-destination pairs to those related parties. LIFI is also interesting because it provides that feature intrafirm exports. Since firms make part of their the countries where the parents of affiliates located in France exports to these countries intrafirm, we can be certain that are located themselves. As this identifies some firms in the the firm has a related party in the destination country. Among French Customs data as MNEs for which we do not have these pairs, there are firms selling all their exports intrafirm, the INSEE data, we drop these 606 firms because we cannot while other firms may export to the country through both know whether their transactions are intrafirm or arm’s length modes. In the data, the median share of intrafirm trade is (2.8% of the value of French exports). We also eliminate the 98%. This means that conditional on exporting intrafirm to observations of state-owned firms as these firms might have a different price-setting mechanism (2.5% of French exports). 18 We use alternative definitions of the thresholds. As shown in tables A.10 The data allow us to tag trade flows as intrafirm or arm’s to A.12 of the online appendix, they do not affect the main results of the length. In our data, 1.6% of the firms do both intrafirm paper. 19 The probability that we observe only intrafirm transactions is negatively and arm’s-length export within an HS4-destination category. associated with the tax rate of the destination country. The selection effect tends to underestimate the effect of taxes on intrafirm prices. We also drop all 16 The 1 million threshold reduces the number of respondents, but the the product-destination pairs where there is no intrafirm transaction (these survey still covers 96% of French international trade (imports and exports). pairs account for about half of French exports). While 8,239 firms were surveyed, only 53% of them responded. However, 20 We lose information on about 10% of the French export value when they cover 84% of French total exports (Bovar, Lebon, & Plateau, 2003, merging the data set with the tax rate data. We also lose an additional 2.3% table 1). of the export value when merging the data with information on tariffs. The 17 We drop 6,417 observations (2% of exports) for which the value of tariff data are not available for 20 countries in the original sample of 65 exports reported in the Customs data is 1% higher or 1% lower than the countries. Reproducing the estimations without the tariff variable leads to value of exports reported in the survey. qualitatively similar results.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 126 THE REVIEW OF ECONOMICS AND STATISTICS

a country, the median firms export almost entirely intrafirm. firms export to these countries, and these exports account for This figure, however, hides large variations. In particular, we roughly 11% of the total number of observations.24 As Hines find that for one-fourth of product-country pairs, the share and Rice discussed in (1994), this classification begins with a of intrafirm trade is at most 40%. To put it differently, even list provided by the U.S. Internal Revenue Service (see Glau- if a firm exports intrafirm to a country, in 25% of cases, tier & Bassinger, 1987), which they modify by requiring four the share of intrafirm exports to the country is below 40%. attributes of a tax haven: a low tax rate, legislation enabling These facts confirm the usefulness of having information business and banking secrecy, a good communication infras- on intrafirm and arm’s-length transactions. They also show tructure, and self-promotion as an offshore financial center. the caveats of databases that report information only on the It is important to recognize that their definition requires all presence of related parties in the destination country. of these and relies on both a low tax rate and features that would aid a firm in undertaking transfer pricing (which can Tax and tax haven data. In our model, equilibrium is be interpreted as reflecting a lower cost of transfer pricing in where the marginal savings from reducing tariff and tax pay- our model). As such, some countries appearing on the IRS ments equal the marginal cost of transfer pricing. Thus, the list (and in our data) are excluded from tax haven status. most appropriate tax measure is the effective marginal tax Specifically, Austria (which, although it satisfies the latter rate (EMTR) on income as this represents the tax savings three, has a relatively high tax rate), Korea, Argentina, and from shifting 1 euro of income. If taxes are flat, the EMTR Ecuador (three low-tax countries that do not claim to be equals the effective average tax rate (EATR). However, if offshore financial centers and are not generally regarded as taxes are progressive (as is typically the case), the EATR tax havens) are not classified as such. By contrast, Malta is will understate the tax savings from transfer pricing. In our added since, although Loretz (2008) lists it as having a fairly baseline results, we use the EMTR from Loretz (2008). In high tax rate, given the low taxes paid by U.S. firms, Hines robustness checks, we use the EATR and the top statutory and Rice (1994) include it. Other lists use similar notions to corporate tax rate instead (both from Loretz, 2008) and find identify tax havens, although the specifics differ. For exam- qualitatively identical results. In the baseline estimation, we ple, the OECD’s (2000) list requires a comparable lack of used the EMTR reported in 1998 or 1997 (whichever was transparency and practices that promote secrecy but tightens closer) when the data for 1999 were missing.21 An important the tax rate criterion to “no or only nominal taxes.” Thus, aspect of these tax rate measures is that they are constructed whereas the approach of Hines and Rice (1994) identifies from statutory tax policies, but unlike the headline tax rate, Ireland, Luxembourg, and Switzerland as tax havens, the they account for factors such as the tax offsets from capital OECD’s does not.25 Other lists, such as those compiled by expenditures for a hypothetical firm (see Loretz, 2008, for the Financial Stability Board (2000) or Tax Justice Network details). As such, EMTR and EATR are exogenous to firm (2007) do include these three countries. decisions, something that would not be the case if we used firm accounting data to construct firm-specific taxes. Pricing-to-market data. Firms adjust their prices to As seen in table A1 in the table appendix, the effec- the characteristics of the destination market. The empiri- tive average and marginal tax rates vary considerably across cal literature has identified two main regularities in firms’ countries. In our estimation sample, the EMTR ranges from pricing-to-market behavior: firms charge higher prices when 0% in the Bahamas to about 46% in the Russian Federation. the destination is farther away and when the destination is Of great concern in policy circles is the use of investment in wealthier (Bastos & Silva, 2010; Manova & Zhang, 2012; tax havens for aggressive tax planning. This is particularly Martin, 2012). Berman, Martin, and Mayer (2012) have true for countries such as France and Germany that exempt shown that small and large firms may react differently to foreign income from taxation.22 We therefore use additional trade costs depending on their size and productivity. In information on tax havens. Our definition follows the one our model, we show that these factors may affect intrafirm in Hines and Rice (1994), which has been used recently prices differently from arm’s-length prices. Furthermore, as by Dharmapala and Hines (2009). This gives us ten tax these market characteristics are correlated to the level of havens in our estimation sample: the Bahamas, Bermuda, the the corporate tax rate, it is crucial to control for them. We Cayman Islands, Cyprus, Hong Kong, Ireland, Luxembourg, therefore use data on per capita GDP (in logarithms) from Malta, Singapore, and Switzerland.23 Approximately 41% of the Internal Financial Statistics of the IMF to control for the level of country-specific income. As measures of trade 21 Notice that the 38 of 49 countries in our sample share a bilateral tax costs, we use the logarithm of the bilateral distance variable treaty with France. Controlling for the existence of treaties (and interacting (the population-weighted average distance between coun- with the intrafirm dummy) does not change the results, and indeed, we find tries’ main cities in kilometers), which is taken from the little impact of tax treaties on transfer pricing. 22 See Gumpert et al. (2011), who consider this issue for a large sample CEPII database (Mayer & Zignago, 2006), and information of German MNEs. 23 The correlation between tax haven status and EMTR is −0.54 and is 24 Details on the composition of exports composition to tax havens are significantly different from 0 at the 1% level of significance. The correlation available in table A.4 of the online appendix. between the tax haven dummy variable and the explanatory variable log(1− 25 In addition, neither Hong Kong nor Singapore is considered by the EMTR) is 0.51. OECD (2000).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 127

from TRAINS on tariffs faced by French exporters, which is for the levels of the price differential (Bernard et al., 2006). developed by the WITS (UNCTAD).26 In our data, distance More importantly, the use of triadic fixed effects along with and per capita GDP are both significantly and negatively the interaction term allows us to compare the export prices correlated with both the effective tax rates and tax haven for different modes (arm’s length and intrafirm) for a given status, suggesting that their omission could bias our results. product across countries. Given the set of controls that we In the online appendix, we extend the baseline results to discuss below, the estimated interaction coefficients give an other factors that might also be relevant (see table A.3). For indication of the price differential due to transfer pricing. example, luxury-goods exporters are likely to take into con- Our identification strategy implies that the deviation from sideration the amount of inequality in a country in setting the arm’s-length price in itself is not evidence of transfer prices. Manova and Zhang (2012) find, moreover, that coun- pricing; rather it is the systematic and significant relation- try size matters, as does remoteness. Controlling for these ship between this deviation and the tax differential across additional factors does not alter our main results. countries that is indicative of transfer pricing. As we will show, we use destination fixed effects in some specifications B. Identification Strategy to control for destination-specific heterogeneity. Given pfpmc, the export price charged by firm f for product p in country Our identification strategy is based on the comparison of c under the export mode m, the empirical strategy involves intrafirm prices charged by MNEs with arm’s-length prices estimating the following model: for a given product across countries. It is important to note that this arm’s-length comparison includes pure exporters, pfpmc = α1Intrafpmc + α2Tax c + α3Tax c × Intrafpmc MNEs without affiliates in a given destination, and those + α4TaxHavenc + α5TaxHavenc × Intrafpmc with affiliates in this destination that nevertheless sell the + γ X + γ X × Intra + μ + .(4) product only to nonaffiliates. Before detailing our strategy, 1 c 2 c fpmc fpm fpmc we clarify the following points. First, we do not identify Tax captures the tax level in the destination country. Our pri- transfer pricing by comparing intrafirm and arm’s-length mary measure is defined as Tax = log(1−τ ), with τ being prices charged by a given firm in a single destination for a c c c the EMTR in country c.27 Our second measure, TaxHaven , particular product. A significant price wedge between inter- c is a dummy variable that takes the value 1 if the country is on nal and arm’s-length prices within firm-HS4-destination may the tax haven list of Hines and Rice (1994). These are both serve to alert tax auditors to tax-induced transfer pricing. also interacted with Intra , a dummy variable that takes As Cristea and Nguyen (2016) pointed out, MNEs tend to fpmc the value of 1 if the export mode is intrafirm and 0 other- manipulate their arm’s-length prices as well as their inter- wise. Since we expect the price wedge to be increasing in the nal prices. Thus, comparing the arm’s-length and intrafirm amount of profits retained by the firm—Tax = log(1 − τ ) prices charged by the same MNE would not allow us to c c is larger or the country is a tax haven—we anticipate both of cleanly identify the extent of transfer pricing. Second, as these interactions to be negative (i.e., a larger absolute value detailed below, our strategy requires information on firms difference between the intrafirm and arm’s-length prices). that are exporting. While we can identify exporters located in The term μ is a comprehensive set of firm-product- France, we cannot identify the firms exporting their product fpm mode fixed effects. X is a vector of country-specific vari- to France. Any comparison between intrafirm and arm- c ables that includes the logarithm of distance, tariffs, and the length import prices might therefore be driven by foreign logarithm of GDP per capita. We interact these variables with exporters’ attributes such as their productivity that are not the intrafirm transaction dummy, as the pricing behavior of related to transfer pricing strategies. For that reason, our firms is also affected by bilateral trade costs and income analysis focuses on export prices. in the destination market and might also vary across export In our empirical estimation, we make use of an interaction modes. Since prices might also be influenced by the mar- term between the tax variables and an indicator of the export- ket structure and the intensity of competition in the foreign ing mode that is equal to 1 if the transaction is intrafirm and market and since these characteristics are unobservable, we 0 if it is arm’s length. This interaction term allows us to introduce a set of country fixed effects in some specifications. compare the arm’s-length and intrafirm prices of MNEs and Finally, is the disturbance term. The standard errors are exporters for a given product. Moreover, our empirical model fpmc allowed to be adjusted for clustering at the country level to includes a set of triadic fixed effects at the firm, product, and account for heteroskedasticity and nonindependence across export mode levels. The use of triadic fixed effects accounts repeated observations within countries.28 Table 1 gives the for a broad set of attributes of the transactions at the firm, product, and exporting mode levels that might also account 27 Alternatively, we use the EATR measure and find comparable results (see tables A.7 and A.8 in the online appendix). 26 We use the logarithm of 1 plus the tariff rate in our specifications as 28 Clustering at the product or product-mode level leads to smaller standard in Hummels and Skiba (2004). The estimates on the tariff coefficient are errors than when we cluster at the country level. The results are shown in sensitive to the log transformation of the variable. We show in table A.5 the online appendix’s table A.6. As Cameron and Miller (2015) showed, in the online appendix that the transformation does not affect the point the use of bigger and more aggregate clusters as the country level in our estimates of other variables. case is the most conservative cluster dimension and should avoid bias.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 128 THE REVIEW OF ECONOMICS AND STATISTICS

Table 1.—Summary Statistics in order to extract rents, this incentive is not present when Variable Nature Mean SD selling to itself. As such, if internal trade includes a larger share of intermediate goods that are not directly sold to over- Intrafpmc 0/1 0.090 0.286 (1 − τc) (EMTR) (log) −0.353 0.095 seas customers, this result makes sense. Turning to the trade − τ − (1 c) (EATR) (log) 0.383 0.091 cost variables, we find a positive effect of distance on export Tax Havenc 0/1 0.106 0.308 prices in line with the literature. A 10% increase in distance 1 + Tariffc (log) 0.011 0.041 Distancec (log) 6.986 0.864 raises export prices by 0.8%. As an example, given the dis- Per Capita GDPc (log) 9.978 0.567 tances between France and the countries in our sample, the × Intrafpmc export prices are on average 0.8% higher in the Netherlands (1 − τc) (EMTR) −0.031 0.104 (1 − τc) (EATR) −0.034 0.112 as compared to Belgium. The effect of distance on internal TaxHavenc 0.007 0.082 export prices is lower, which is also consistent with smaller Tariffc 0.001 0.015 markups for internal trade. Concerning the tariff variable, Distancec 0.641 2.065 Per Capita GDPc 0.889 2.833 the effect on arm’s-length prices is not significant. In other This table reports the summary statistics of the variables in our estimation sample. There are 729,737 words, there is no evidence of dumping by French arm’s- observations. length exporters. This might be due to the low cross-country variation in the tariff variable, as 90% of observations con- cern a transaction toward countries that are members of the summary statistics of the main explanatory variables. The European Union. However, intrafirm prices are significantly distribution of intrafirm and arm’s-length prices is reported lower than arm’s-length prices in high-tariff countries, sug- in figure A.2 in the online appendix. gesting that firms choose to undervalue their exports to pay lower tariffs. IV. Results In column 3, we replace our EMTR variable with a dummy A. Baseline Results variable equal to 1 if the destination is a tax haven. As tax havens not only have low taxes but often provide other According to the theoretical predictions, the average inter- mechanisms that facilitate profit shifting (such as the limited nal price should be lower than the arm’s-length price in a exchange of information between tax authorities), one might country with a lower marginal effective tax rate. The esti- expect that internal prices differ markedly in these nations. mates are reported in table 2. Overall, the specifications The results are striking. The coefficient of the interaction explain about 87% of the variation of the log level of export between the tax haven dummy variable and the intrafirm prices as suggested by the adjusted R2. The estimates using export mode is highly significant. We do not find a significant the EATR variable are reported in tables A.7 and A.8 in effect of the tax haven dummy variable on the arm’s-length the online appendix. In all specifications, standard errors are price. This result suggests that arm’s-length export prices are clustered in the country dimension. the same regardless of whether the destination is a tax haven. In column 1, we do not find a statistically significant The interaction between the tax haven and the export mode effect of the effective marginal tax rate on the level of arm’s- dummy variables is significantly negative, indicating that the length prices. We do, however, find a negative and significant average internal export price for a tax haven is about 11% interaction coefficient between the corporate tax and the lower than the comparable arm’s-length price. This suggests intrafirm dummy: internal export prices are relatively lower that tax havens are playing a major role in the transfer pricing than arm’s-length prices in destinations with a lower corpo- strategies of firms.29 rate tax rate. A 10% decrease in the effective marginal tax This finding remains robust in column 4 when we reintro- rate leads to a reduction of intrafirm prices by 1.2% (2.1% duce the effective marginal tax rate and its interaction term using the EATR variable). with the export mode, where we find the intrafirm export In column 2, we control for other country characteristics prices to be about 9% lower than arm’s-length prices in tax that might influence a firm’s pricing behavior. Nevertheless, haven destinations even when tax rates do not differ. Notice we continue to find comparable results: that taxes influence that the coefficient of the interaction term that involves the internal prices but not those between unrelated parties. If EMTR is smaller and loses statistical significance once we anything, the estimated impact is slightly larger, suggesting control for tax havens. As tax havens tend to have low taxes, a slight bias when they are excluded. In line with the pre- diction of our model, we find a positive impact of per capita 29 Notice that the coefficient on the tax haven dummy is high (0.11) but GDP on the level of prices. A 10% increase in per capita not significant. While the coefficient of the tax haven dummy does not GDP raises the export prices by 0.6%. However, we find a change the interpretation of our results, it suggests higher arm’s-length prices in tax havens. A careful examination of the data shows that the mag- negative and statistically significant interaction coefficient nitude of this particular coefficient is entirely driven by the high prices between the per capita GDP and the intrafirm mode vari- charged by arm’s-length exporters in Switzerland. Including a dummy for ables. This suggests a slightly lower impact of per capita Switzerland pushes the coefficient on the tax haven dummy to 0—but the interaction with the intrafirm dummy variable remains stable. Our main GDP on internal prices. Although a firm has an incentive to results are robust, and the coefficient on tax haven is 0.2 when Switzerland is mark up prices over costs when selling to unrelated parties excluded.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 129

Table 2.—Baseline Regression Effective Marginal Tax Rate Dependent Variables: Export Price (1) (2) (3) (4) (5) (6)

log(1 − τc) 0.09 0.13 −0.00 −0.03 (0.146)(0.144)(0.100)(0.098) ∗ ∗∗ ∗ ∗ −×Intrafpmc −0.18 −0.22 −0.12 −0.08 −0.10 (0.091)(0.106)(0.067)(0.053)(0.054) TaxHavenc 0.11 0.11 (0.073)(0.074) ∗∗ ∗∗ ∗∗∗ −×Intrafpmc −0.11 −0.09 −0.08 (0.040)(0.036)(0.025) ∗∗ ∗ ∗ Per Capita GDPc 0.06 0.04 0.04 0.04 (0.025)(0.022)(0.022)(0.023) ∗∗∗ ∗ −×Intrafpmc −0.03 −0.01 −0.01 −0.02 −0.00 (0.011)(0.010)(0.010)(0.010)(0.008) ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Distancec 0.08 0.09 0.09 0.11 (0.025)(0.022)(0.022)(0.022) ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ −×Intrafpmc −0.05 −0.05 −0.05 −0.07 −0.06 (0.013)(0.010)(0.010)(0.011)(0.009) Tariffc 0.40 0.32 0.32 0.14 (0.324)(0.268)(0.268)(0.274) ∗ ∗ ∗ ∗∗∗ −×Intrafpmc −0.40 −0.33 −0.34 −0.21 −0.36 (0.213)(0.170)(0.174)(0.160)(0.123) Sample Full Full Full Full w/o Tax H. Full Country FE No No No No No Yes Firm-product-mode FE Yes Yes Yes Yes Yes Yes Observations 729,737 729,737 729,737 729,737 652,548 729,737 Adjusted R2 0.863 0.864 0.865 0.865 0.869 0.853 This table investigates the impact of the effective tax rate, GDP per capita, distance, tariffs, and of the tax haven dummy on intra-firm and arm’s length export prices. In column 6, we further include country fixed effects. In column 5, we drop tax haven destinations. Robust standard errors clustered by destination in parentheses. Significance levels: ∗p < 0.1, ∗∗p < 0.05, and ∗∗∗p < 0.01.

this suggests that the results in column 2 were biased due to B. Transfer Pricing in Low-Tax Jurisdictions a failure to control for the tax haven status. Further, it high- lights the important difference between having low taxes and Until now, we have investigated the average effect of tax having low taxes together with other policies that make tax rates on the export price differential. We find evidence of planning easier. Indeed, the OECD (2013) makes precisely transfer pricing, but only in tax haven countries character- such a distinction.30 ized by very low tax rates. Our results therefore suggest a In column 5, we investigate the importance of tax havens threshold effect of the tax rates on the price differential. We further by excluding them from the analysis. Compared to examine this effect further by running a regression using, column 2, the coefficient of the interaction term that involves instead of the EMTR, a set of dummy variables indicating the effective tax rate is about three times lower and becomes the decile in which a country’s EMTR falls. We choose the insignificant, again suggesting that the impact in column 2 ninth decile as our benchmark. This decile is composed of comes from tax havens. five countries that have roughly the same effective marginal Finally, column 6 includes a set of destination-specific tax rate as France and where, in theory, internal and arm’s- dummy variables. Introducing country fixed effects does length prices should be the same. The first decile includes not allow us to estimate the direct effect of the country- countries with the lowest effective marginal corporate tax specific variables (including the tax rate and tax haven rates: the Bahamas, Hong Kong, Ireland, Slovenia, and South status). This, however, comes with the benefit of con- Africa. The tenth decile includes Argentina, Germany, Japan, trolling for other destination characteristics. As can be and Poland, the countries with the highest effective marginal seen, the tax rate interaction remains insignificant, and corporate tax rates in our sample. the tax haven interaction is virtually unchanged in magni- The estimated coefficients of these interaction terms are tude and becomes even more significant. Thus, even after shown in figure 1. We report the median tax rate in each including destination fixed effects, we find evidence of tax- decile and show that the results are not driven by important induced transfer pricing, which is most evident in tax haven jumps in EMTRs between deciles. The pattern described in countries. figure 1 is consistent with the presence of an inaction band driven by the fixed component of the concealment costs as 30 The firms might also operate in tax havens and nontax haven countries. shown by the theory. Each dot corresponds to the interaction We also drop the firms that export to tax havens internally. We do not find coefficient between the effective average tax rate and the a significant tax effect. This suggests that there is no substitution effect. intrafirm export mode dummy variable. We also display the The transfer pricing strategy is not used by firms that do not export to tax havens, while firms that export to tax havens have lower prices in countries confidence intervals at the 5% level. The estimated effects with lower tax rates. are quite heterogeneous. The point estimate of the interaction

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 130 THE REVIEW OF ECONOMICS AND STATISTICS

Figure 1.—Corporate Tax Rate on Transfer Pricing Across Tax Deciles the manipulation of internal prices for tax reasons is pri- marily a phenomenon for large firms in tax havens. Further, .05 we find that the relationship between pricing to market and internal prices is more prevalent in large firms. In columns 3 and 4, we analyze another source of het- 0 erogeneity by operating a distinction across the nationality of ownership of an MNE. In column 3, we include MNEs −.05 that are French residents or are majority owned by a French group (as well as all non-MNEs). In column 4, we include MNEs that are majority owned by a foreign group and all −.1 exporters. Comparing the two, we find that the coefficients are estimated more precisely in the sample of French firms.

Price gap intrafirm − arm’s−length −.15 In particular, we find significance for tax havens only for the French firms. These results therefore again suggest that tax havens play a major role in the transfer pricing strategies of −.2 French firms. 10 15 20 25 30 35 40 Median tax rate in the tax decile In column 4, although the sign of the tax rate and tax haven

This graph displays the price wedge between intrafirm and arm’s-length prices by decile of the destination variables match those in the French-only sample, they are not country corporate tax rate. The price wedge is measured by the coefficients on the interaction between statistically significant. This suggests that similar forces are tax deciles and an intrafirm dummy in a regression of the logarithm of export prices on firm-product- exporting mode fixed effects, tax decile of the destination country, GDP per capita, distance, and tariff, and at play for this sample as well, although there may be greater their interaction with a dummy equal to 1 if the exports are between related parties. Each dot reports the coefficient associated with the interaction between the intrafirm dummy and a dummy for the tax decile of noise due to the variety of parent countries in this sample as the destination country. The first decile is the decile of countries with the lowest corporate tax rates. The tenth decile is the decile with the highest corporate tax rates. The coefficient for the decile 9 is normalized compared to that in column 3. In particular, if MNEs from to 0 (countries with the same tax level as France). The x-axis reports the median tax rate within each tax other countries face worldwide taxation, this may illustrate decile. The gray area corresponds to the confidence interval at 5%. the cleaner tax effects to be found by using data on FDI from a tax-exempting country. In France, as in most other countries, the tax authorities’ effect is, however, negative and significant only for countries expectation is that firms set the price of their internal trans- in the first two deciles. This indicates that our results are actions according to the arm’s-length principle.33 The main heavily driven by the nine lowest-tax countries, five of which force at play in the above model is that deviations from this are also classified as tax havens (see table 5). As explained price come at a cost. This cost includes penalties incurred in section 2, our simple model can easily accommodate such if a firm is caught. When the appropriate arm’s-length price inaction band when the concealment cost function includes is easily determined, as is the case for homogeneous goods a fixed cost. that are traded in organized markets, transfer pricing should therefore be minimal. By contrast, differentiated products C. Additional Results that are by definition specific to the relationship lack such comparable arm’s-length transactions (Blonigen, Oldensk, A relevant concern that has been raised in the literature & Sly, 2014). Thus, MNEs that export differentiated prod- studying the effects of transfer pricing is the differing abil- ucts might more easily reduce taxes via transfer pricing. In ities of firms to engage in transfer pricing (Bernard et al. columns 5 and 6, we use the Rauch (1999) classification 2006). In our model, larger firms are expected to have larger and document the effect of taxes and tax havens on both price differentials. In columns 1 and 2 of table 3, we split the the homogeneous and the differentiated goods category.34 sample according to the size of the MNEs measured by their There is no difference between intra- and interfirm transac- total exports.31 In the first column, we drop MNEs below tions for goods that are exchanged on organized markets (i.e. the 75th percentile of the distribution of multinational firms’ where appropriate prices are easily verified by tax authori- size and thus keep large MNEs and all pure exporters.32 In ties). Differentiated products, however, do show evidence of column 2, we drop observations of MNEs above the 25th such pricing practices. Notice that the estimated coefficients percentile, keeping only small MNEs and all pure exporters. reported in column 6 are in line with the ones found in the Looking at the corporate tax and the tax haven interactions baseline estimation (column 6 of table 2). with the intrafirm trade dummy, we find significance for tax havens only for the large firms, which is consistent with the 33 The French tax legislation is based on the comparable uncontrolled price inaction band emphasized in the model. This indicates that method (CUP). This means applying prices that independent companies would set in identical transactions (BOI, 1999). 34 In the online appendix’s table A.9, we go beyond the binary product 31 Note that as all the estimations in table 3 include destination fixed classification that Rauch (1999) proposed by using alternative measures of effects, only the interaction terms can be estimated. differentiation that Khandelwal (2010) or Nunn (2007) proposed. The esti- 32 Note that since all exporters appear in both columns 1 and 2, the com- mated coefficients are less precise but suggest that transfer pricing strategies bined number of observations across these two regressions is greater than are more pronounced in industries that exhibit more vertical differentiation in the baseline specification. or require relationship-specific inputs.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 131

Table 3.—Additional Results: Effective Marginal Tax Rate (1) (2) (3) (4) (5) (6) (7) (8)

Intrafpmc× ∗ ∗ log(1 − τc) −0.08 −0.13 −0.19 −0.04 −0.06 −0.12 −0.13 −0.09 (0.086)(0.149)(0.128)(0.089)(0.140)(0.069)(0.069)(0.067) ∗∗∗ ∗ ∗∗ ∗∗∗ ∗∗ TaxHavenc −0.11 −0.00 −0.07 −0.11 −0.04 −0.07 −0.07 −0.08 (0.022)(0.069)(0.039)(0.067)(0.087)(0.027)(0.024)(0.036) Per Capita GDPc −0.00 −0.01 −0.02 0.01 −0.04 −0.01 −0.01 −0.01 (0.010)(0.031)(0.023)(0.015)(0.050)(0.011)(0.011)(0.010) ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Distancec −0.05 −0.07 −0.09 −0.04 −0.01 −0.05 −0.05 −0.06 (0.012)(0.023)(0.018)(0.015)(0.055)(0.012)(0.009)(0.011) ∗∗ ∗ ∗ ∗∗∗ ∗∗ Tariffc −0.41 −0.24 −0.44 −0.28 0.17 −0.55 −0.34 −0.25 (0.182)(0.276)(0.248)(0.163)(0.147)(0.184)(0.145)(0.161) Big Small French Foreign Homogeneous Differentiated Without Sample Firm Firms Firms Owned Goods Goods Wholesalers All Firm-product-mode FE Yes Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes No Country Sector FE No No No No No No No Yes Observations 710,924 682,887 567,652 455,508 11,592 599,792 276,451 729,737 Adjusted R2 0.928 0.931 0.934 0.933 0.969 0.915 0.907 0.929 This table investigates the impact of the effective tax rate, GDP per capita, distance, tariffs, and the tax haven dummy on intrafirm and arm’s-length export prices. Column 1 focuses on MNEs whose export sales are above the P75. Column 2 focuses on MNEs whose export sales are below the P25. Column 3 excludes foreign-owned affiliates. Column 4 excludes French MNEs. Column 5 contains only products classified as homogeneous in the Rauch nomenclature. Column 6 contains only products classified as differentiated in the Rauch nomenclature. Column 7 excludes MNEs whose main activity abroad is wholesale. Robust standard errors clustered by destination reported in parentheses. Significance levels: ∗p < 0.1, ∗∗p < 0.05, and ∗∗∗p < 0.01.

In column 7, we show that our baseline results are robust and destination requires us to augment the former specifi- to the exclusion of the observations of firms active in the cations by the size of the destination market. We therefore wholesale sector, as the pricing behavior of such firms may include the destination market GDP as larger destination differ from that of others. As can be seen, the results are markets import more volumes. The results are reported in robust to their exclusion. table A2 in the table appendix. In columns 1 and 2, we restrict In the last column of table 3, we deal with the self- the sample to the intrafirm trade flows only. In columns selection of firms into the multinational status. As Helpman, 3 and 4, we report the results based on estimations using Melitz, and Yeaple (2004) emphasized, the selection of firms the full sample. In either sample, we do not find any quan- into multinationals depends on a set of firm characteris- tity differential between arm’s-length and intrafirm exports tics and on barriers to entry and other destination market in tax havens. These findings suggest that low intrafirm characteristics such as the level of competition. These char- prices in tax havens are not directly related to the exported acteristics may well be influenced by the corporate tax rate quantities. or the tax haven status of the destination market. Our speci- In the online appendix, we provide a set of additional fication allows for the possibility that market characteristics robustness checks. Our results are robust to additional vary across industries and across countries. We include coun- determinants of pricing to market (table A.3), alternative try × HS2-industry fixed effects in column 8 and show that definitions of the variables of interest (tax rates, tariffs, our main results hold. and product differentiation; tables A.5, A.7/A.8, and A.9), One may worry that the specific behavior of transfer different thresholds’ definitions of the export mode (tables pricing to tax havens is due to the very specific nature A.10–A.12), and alternative treatments of the intrafirm trade of transactions between firms and their affiliates in these variable (table A.13). jurisdictions. In the online appendix, we show that the nature of intrafirm transactions to tax havens is very sim- ilar to intrafirm exports toward other destinations (see table D. Back-of-the-Envelope Calculation A.3). Specifically, about 70% of intrafirm exports are resold To quantify the losses for tax authorities due to transfer directly by the affiliates, while about 30% of intrafirm pricing, we use the estimates of the baseline estimation, col- exports are used by the affiliates in their production pro- umn 4. In our quantification exercise, we compute the loss cess.35 We also examine whether the volumes exported by of exports due to lower pricing in tax havens. In 1999, the firms are systematically different between tax havens and French effective corporate tax rate was 31.8%, and brought other countries. We use the same empirical strategy as the in about 36 billion euros of corporate tax receipts.36 one that we implemented for the analysis of export prices. In column 3 of table 2, we find that intrafirm prices are The estimation of quantities at the level of firm-product 10.4% (exp(−0.11) − 1) lower than the market price in tax

35 Note that we have information only on the use of the products by the 36 http://www.performance-publique.budget.gouv.fr/farandole/archives affiliates for 45% of intrafirm transactions, accounting for 77% of intrafirm /1999/lftab99.htm. Taxes are charged on the taxable income, which consists exports. of operational and financial profits minus charges.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 132 THE REVIEW OF ECONOMICS AND STATISTICS

Table 4.—Underreporting to Tax Havens Country Share of French Exports Share of Exports Intrafirm Value Not Reported (million euros) Switzerland 0.0407 0.58 597.0 [241, 932]a [390, 797]b Ireland 0.0083 0.62 131.0 [53, 204]a [85, 74]b Singapore 0.0072 0.58 107.0 [43, 166]a [70, 142]b Hong Kong 0.0071 0.54 97.5 [39, 152]a [64, 130]b Luxembourg 0.0056 0.37 51.9 [21, 81]a [34, 69]b Malta 0.0019 0.88 42.8 [17, 67]a [28, 57]b Cyprus 0.0007 0.53 10.0 [4, 16]a [7, 13]b Bermuda 0.0003 0.85 6.0 [2, 9]a [4, 8]b Bahamas 0.0002 0.51 2.8 [1, 4]a [2, 4]b Cayman Islands 0.0001 0.55 0.7 [0.3, 1]a [0.5, 1]b This table presents the weight of tax havens in French exports, their share of intrafirm trade, and the estimated values of intrafirm exports not reported due to transfer pricing (built from the coefficient estimated in column 3, table 2). a Confidence intervals based on standard errors of coefficients clustered in the country dimension. bConfidence intervals based on standard errors of coefficients clustered in the product dimension.

havens.37 Table 4 reports the share of exports and the share destinations come from 25 firms. This suggests that a small of these exports that were intrafirm for the ten tax havens. number of firms are avoiding a large tax payment.39 Three of these countries are important export destinations. Our results suggest that the lion’s share of transfer pric- Furthermore, the shares of intrafirm exports to Switzerland ing practiced in France is concentrated in the exports to at or Ireland are very high (around 60%). One can see strong most ten countries by about 7% of multinationals. Targeting heterogeneity in the importance of the different tax havens. exports by these firms to tax havens would make enforcement Conduit tax havens like Switzerland, Ireland, Singapore, and of the arm’s-length price principle more efficient. Hong Kong account for the lion’s share of profit shifting. By contrast, profit shifting through the transfer pricing of V. Conclusion physical products is very modest for financial tax havens like the Bahamas and the Cayman Islands. Recognizing this Despite the clear incentive that firms have to shift prof- distinction between types of tax havens can improve the its through transfer pricing and the widespread concern over effectiveness of tax enforcement. its implications, direct and systematic evidence of this prac- Using the intrafirm trade values in the data, the final col- tice remains scarce. This is due to a general lack of data umn gives the value of underpriced intrafirm exports, the sum on the prices used within a multinational and the prices for of which amounts to more than 1 billion euros.38 Without comparable arm’s-length transactions. Thus, the question of this underreporting, French tax authorities would have col- transfer pricing practices in terms of their monetary value lected 333 million euros more. This figure can be compared and the number of firms and countries involved remains to the 36 billion euros collected in 1999 (which includes largely unanswered. both services and manufacturing), meaning that total tax rev- We have built a unique data set that overcomes this enues that year would have been roughly 1% greater were it problem. These data contain prices at the firm-product- not for transfer pricing by manufacturing firms in these ten destination level for both intrafirm and arm’s-length exports. tax havens. Interestingly, only 2,495 firms make intrafirm Having such detailed data is important for three key reasons. exports to these countries, with a scant 450 firms accounting First, it allows us to control for other determinants of prices for 90% of intrafirm exports to these ten countries. What is across firms, such as the relative productivity of multina- more, almost 50% of intrafirm exports to these tax haven tionals as compared to exporters. Second, it allows us to control for the destination country’s characteristics, such as income and trade costs, which are potentially correlated with tax variables yet affect intrafirm and arm’s-length prices in 37 We consider all intrafirm exports, not only those used in our estimates. Note that using the information on the location of foreign affiliate would different ways. Third, the richness of the data allows us to overestimate the losses for the tax authorities due to transfer pricing as some consider not only the effect of foreign corporate taxes on multinationals do not export internally to these locations. 38 These are estimated figures. We report confidence intervals for these figures computed from the standard errors of the coefficients obtained with 39 We present the sectoral composition of exports to tax havens in the clustering in either the country or the product dimension. online appendix (table A.5).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 KNOCKING ON TAX HAVEN’S DOOR 133

pricing behavior, but also the role of tax havens and how Financial Stability Board, Report of the Working Group on Offshore this behavior varies with firm and product characteristics. Financial Centres (Basel: Financial Stability Board, 2000). Fuest, C., B. Huber, and J. Mintz, “Capital Mobility and Tax Competition: We find no evidence of tax avoidance if we disregard tax A Survey,” CESifo working paper series 956 (2003). haven destinations. Furthermore, transfer pricing is primar- Glautier, M. W., and F. W. Bassinger, A Reference Guide to International ily found within large MNEs. These results are crucial for Taxation: Profiting from your International Operations (Lexington, MA: Lexington Books, 1987). two reasons. First, they support the OECD’s (2013) asser- Gresik, T. A., “The Taxing Task of Taxing Transnationals,” Journal of tion that there is a difference between low-tax countries and Economic Literature 39 (2001), 800–838. tax havens that provide a tax environment that is particularly Gumpert, A., J. R. Hines, and M. Schnitzer, “The Use of Tax Havens in Exemption Regimes,” NBER working papers 17644 (2011). amenable to tax avoidance. Second, it shows that although Haufler, A., and G. Schjelderup, “Corporate Tax Systems and Cross Coun- transfer pricing may result in significant revenue losses, such try Profit Shifting,” Oxford Economic Papers 52 (2000), 306–325. losses are primarily due to a small number of firms. Given Heckemeyer, J. H., and M. Overesch, “Multinationals’ Profit Response to Tax Differentials: Effect Size and Shifting Channels,” ZEW that our estimates are for 1999 alone, the cumulative tax discussion papers 13-045 (2013). losses from such transfer pricing should be quite large. This Helpman, E., M. J. Melitz, and S. R. Yeaple, “Export versus FDI with implies that by appropriately targeting enforcement, a sig- Heterogeneous Firms,” American Economic Review 94 (2004), 300– 316. nificant increase in revenues may be achieved at a small Hines, J. R., and E. M. Rice, “Fiscal Paradise: Foreign Tax Havens cost. Moreover, since our data are only for manufacturing, and American Business,” Quarterly Journal of Economics 109 and not services, this tax loss is likely to be the tip of the (1994), 149–182. Huizinga, H., and L. Laeven, “International Profit Shifting within iceberg. Multinationals: A Multi-Country Perspective,” Journal of Public Economics 92 (2008), 1164–1182. REFERENCES Hummels, D., and A. Skiba, “Shipping the Good Apples Out? An Empirical Confirmation of the Alchian-Allen Conjecture,” Journal of Political Atalay, E., A. Hortacsu, and C. Syverson, “Vertical Integration and Input Economy 112 (2004), 1384–1402. Flows,” American Economic Review 104 (2014), 1120–1148. Kant, C., “Endogenous Transfer Pricing and the Effects of Uncertain Bartelsman, E. J., and R. M. W. J. Beetsma, “Why Pay More? Corporate Tax Regulation,” Journal of International Economics 24 (1988), 147– Avoidance through Transfer Pricing in OECD Countries,” Journal 157. of Public Economics 87 (2003), 2225–2252. Keuschnigg, C., and M. P. Devereux, “The Arm’s Length Principle Bastos, P., and J. Silva, “The Quality of a Firm’s Exports: Where You and Distortions to Multinational Firm Organization,” Journal of Export to Matters,” Journal of International Economics 82 (2010), International Economics 89 (2013), 432–440. 99–111. Khandelwal, A., “The Long and Short (of) Quality Ladders,” Review of Behrens, K., S. Peralt, and P. M. Picard, “Transfer Pricing Rules, OECD Economic Studies 77 (2010), 1450–1476. Guidelines, and Market Distortions,” Journal of Public Economic Loretz, S., “Corporate Taxation in the OECD in a Wider Context,” Oxford Theory 16 (2014), 650–680. Review of Economic Policy 24 (2008), 639–660. Berman, N., P. Martin, and T. Mayer, “How Do Different Exporters React Manova, K., and Z. Zhang, “Export Prices across Firms and Destinations,” to Exchange Rate Changes?” Quarterly Journal of Economics 127 Quarterly Journal of Economics 127 (2012), 379–436. (2012), 437–492. Martin, J., “Markups, Quality, and Transport Costs,” European Economic Bernard, A. B., J. B. Jensen, and P. K. Schott, “Transfer Pricing by U.S.- Review 56 (2012), 777–791. Based Multinational Firms,” NBER working papers 12493 (2006). Mayer, T., and S. Zignago, “Notes on CEPII’s Distances Measures,” MPRA Blonigen, B. A., L. Oldenski, and N. Sly, “Separating the Opposing Effects paper 26469, University Library of Munich, Germany (2006). of Bilateral Tax Treaties,” American Economic Journal: Economic Neiman, B., “Stickiness, Synchronization, and Pass-Through in Intrafirm Policy 6 (2014), 1–18. Trade Prices,” Journal of Monetary Economics 57 (2010), 295–308. BOI “Bulletin Officiel des Impots/Official Tax Bulletin, General Tax Nunn, N., “Relationship-Specificity, Incomplete Contracts, and the Pattern Directorate,” technical report (1999). of Trade,” Quarterly Journal of Economics 122 (2007), 569–600. Borkowski, S. C., “Transfer Pricing Advance Pricing Agreements: Current OECD, Towards Global Tax Cooperation (Paris: OECD, 2000). Status by Country,” International Tax Journal 26 (2000), 1–16. ——— OECD Transfer Pricing Guidelines for Multinational Enterprises Bovar, O., S. Lebon, and C. Plateau, “Les filiales commerciales des groupes and Tax Administrations (Paris: OECD, 2010). industriels: un rôle majeur,” INSEE working paper 936-dec2003 ——— Dealing Effectively with the Challenges of Transfer Pricing (Paris: (2003). OECD, 2012). Cameron, A. C., and D. L. Miller, “A Practitioner’s Guide to Cluster- ——— Action Plan on Base Erosion and Profit Shifting (Paris: OECD, Robust Inference,” Journal of Human Resources 50 (2015), 317– 2013). 372. Overesch, M., “Transfer Pricing of Intrafirm Sales as a Profit Shifting Chan- Clausing, K. A., “Tax-Motivated Transfer Pricing and US Intrafirm Trade nel: Evidence from German Firm Data, ZEW discussion papers Prices,” Journal of Public Economics 87 (2003), 2207–2223. 06-84 (2006). Cristea, A. D., and D. X. Nguyen, “Transfer Pricing by Multinational Ramondo, N., V. Rappoport, and K. J. Ruhl, “Intrafirm Trade and Verti- Firms: New Evidence from Foreign Firm Ownerships,” American cal Fragmentation in U.S. Multinational Corporations,” Journal of Economic Jounal: Economic Policy 8 (2016), 170–202. International Economics 98 (2016), 51–59. Dharmapala, D., and J. R. Hines, “Which Countries Become Tax Havens?” Rauch, J. E., “Networks versus Markets in International Trade,” Journal of Journal of Public Economics 93 (2013), 1058–1068. International Economics 48 (1999), 7–35. Dharmapala, D., and N. Riedel, “Earnings Shocks and Tax-Motivated Swenson, D. L., ‘Tax Reforms and Evidence of Transfer Pricing,” National Income-Shifting: Evidence from European Multinationals,” Journal Tax Journal 54 (2001), 7–26. of Public Economics 97 (2013), 95–107. Tax Justice Network, “Identifying Tax Havens and Offshore Finance Diewert, E., W. Alterman, and L. Eden, “Transfer Prices and Import and Centres,” Tax Justice Network briefing paper (2007). Export Price Indexes: Theory and Practice,” in W. Erwin Diewert, Vicard, V., “Profit Shifting through Transfer Pricing: Evidence from French Bert M. Balk, Dennis Fixler, Kevin J. Fox, and Alice O. Naka- Firm Level Trade Data,” Banque de France working paper 555 mura, eds., Price and Productivity Measurement (Bloomington, IN: (2015). Trafford Press, 2006). Zucman, G., “Taxing across Borders: Tracking Personal Wealth and Ernst and Young, Transfer Pricing Global Reference Guide (Ernst and Corporate Profits,” Journal of Economic Perspectives 28 (2014), Young, 2012). 121–148.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021 134 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE APPENDIX

Table A1.—Countries, EATR, EMTR, and Tax Havens in 1999 Country EATR EMTR Country EATR EMTR Country EATR EMTR Bahamasa 0.00 0.00 Ecuador 0.22 0.23 China 0.30 0.31 Bermudaa 0.00 0.00 Bulgaria 0.24 0.27 Uruguay 0.30 0.30 Cayman Isl.a 0.00 0.00 Norway 0.25 0.26 Spain 0.31 0.32 Irelanda 0.08 0.09 Great Britain 0.25 0.26 New Zealand 0.31 0.31 Hong Konga 0.09 0.12 Cyprusa 0.25 0.25 Brazil 0.31 0.32 Slovenia 0.12 0.18 Denmark 0.25 0.28 Canada 0.31 0.32 South Africa 0.15 0.21 India 0.26 0.29 Australia 0.31 0.33 Estonia 0.17 0.20 Trinidad and Tobago 0.26 0.29 Colombia 0.32 0.33 Chile 0.17 0.16 Luxembourga 0.28 0.30 Italy 0.33 0.35 Turkey 0.17 0.23 Portugal 0.28 0.31 Maltaa 0.33 0.32 Switzerlanda 0.18 0.21 Indonesia 0.29 0.29 United States 0.33 0.34 Singaporea 0.19 0.15 Greece 0.29 0.32 Poland 0.34 0.34 Korea 0.20 0.24 Netherlands 0.29 0.31 Argentina 0.35 0.35 Guatemala 0.21 0.23 Austria 0.29 0.31 Japan 0.41 0.41 Sweden 0.21 0.23 Peru 0.29 0.29 Germany 0.42 0.43 Mexico 0.22 0.26 Belgium 0.29 0.33 Finland 0.22 0.24 Czech Republic 0.30 0.31 a Tax havens as defined by Hines and Rice (1994).

Table A2.—Quantities and Tax Havens Dependent Variables: Log of Quantity (1) (2) (3) (4) ∗ TaxHavenc −0.92 −0.10 −0.56 −0.07 (0.492)(0.274)(0.354)(0.361) −×Intrafpmc −0.26 −0.03 (0.208)(0.206) log(1 − τc) −0.19 −0.28 (0.603)(0.398) −×Intrafpmc 0.09 (0.449) Per Capita GDPc −0.21 −0.16 (0.145)(0.121) −×Intrafpmc −0.05 (0.117) ∗∗∗ ∗∗∗ Distancec −0.46 −0.38 (0.091)(0.115) −×Intrafpmc −0.07 (0.079) Tariffc −1.49 −0.83 (1.208)(1.100) −×Intrafpmc −0.66 (1.137) ∗∗∗ ∗∗∗ GDPc 0.54 0.31 (0.072)(0.068) ∗∗∗ −×Intrafpmc 0.23 (0.054) Sample Intrafirm Transactions Full Sample Firm-product-mode FE Yes Yes Yes Yes Observations 65,655 65,655 729,737 729,737 Adjusted R2 0.680 0.694 0.657 0.669 This table investigates the impact of the effective tax rate, GDP per capita, distance, tariffs, GDPs, and the tax haven dummy on intrafirm and arm’s-length export quantities. Robust standard errors clustered by destination in parentheses. Significance levels: ∗p < 0.1, ∗∗p < 0.05, and ∗∗∗p < 0.01.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00673 by guest on 03 October 2021