Taxation of Foreign Profits with Heterogeneous Multinational Firms

Abstract

This paper analyses optimal taxation of foreign profits using a model with heterogeneous multinational firms that serve a foreign market through exports or foreign direct investment (FDI). If a firm switches from exporting to FDI, domestic activity and tax payments may decrease, stay constant or even rise because of intra-firm trade. It turns out that, in contrast to recent claims, in all three cases, the optimal tax system implies full taxation after deduction of foreign tax payments. If the country accounts for the effects of its policy on the foreign price level, the case for taxing foreign income becomes even stronger. However, the globally optimal tax system may require exemption of foreign income from tax.

1. Introduction

In 2008, the worldwide income from outward foreign direct investment reached an all-time high of US$1,283 billion, of which the United States alone had US$350 billion and the United Kingdom around US$130 billion.¹ This income is generally taxed at source, that is, in the country where the investment has been made. However, when transferred back to the firm's headquarters, it can additionally be taxed by the country where the headquarters of the multinational firm reside. In this case, the OECD recommends choosing among two standard systems of taxing repatriated business income: the tax credit system where foreign income is taxed at the domestic corporate tax rate and foreign taxes are credited against the domestic tax liability, and the exemption system where foreign income is exempt from domestic taxation. Given the scarcity of public funds, one would expect that governments in residence countries around the world grasp this opportunity and exercise their right to tax. However, the opposite can be observed: Several countries including the UK² and the United States³ have recently switched from the tax credit system to exemption or are considering such a move.

These reform initiatives receive intellectual support from the academic realm which may seem surprising because, for a long time, scholars used to favour the tax credit system for efficiency reasons. However, as proponents of the exemption system argue, empirical findings have altered the view on international capital flows and, thus, the foundation for optimal taxation reasoning. According to these authors, the new view on multinational investment implies the optimality of the tax exemption system. For instance, Desai (2009) states that ’modern welfare norms that capture the nature of multinational firm activity recommend a move toward not taxing the foreign activities of American firms, rather than taxing them more heavily’. In this paper, I examine whether the case for switching to an exemption system is theoretically well-founded.

The superiority of the credit system builds on the classical work by Peggy Musgrave (née Richman, 1963) who describes a world in which a multinational firm allocates its investment projects across locations. At the margin, it chooses between investing the last dollar at home or abroad. In such a situation, full taxation of foreign income after deduction of foreign taxes paid is the optimal tax policy from a national point of view, whereas crediting foreign taxes against domestic taxes leads to global optimality. Musgrave's work (Richman, 1963; Musgrave, 1969) and the following contributions like Hamada (1966) and Feldstein and Hartman (1979) were highly influential in shaping international taxation agreements like, for example, the OECD convention on double taxation treaties.⁴

This view has recently been challenged. The main point of criticism focuses on the multinational's investment behaviour. In the Musgrave model, one dollar of investment abroad crowds out one dollar of investment at home. Proponents of the new view on international taxation argue that this has been proven wrong by empirical evidence.⁵ Instead, a dollar invested abroad can be shown to effectively increase domestic investment within the firm or, at least, to leave it unaffected.⁶ Then, the proponents argue, there is no rationale anymore for taxing foreign income for efficiency reasons. Optimality implies exemption of foreign profits.

In the following, I will restate the proponents’ arguments in a formal model which captures the important features of the ‘nature of multinational firm activity’: imperfect competition, firm-specific advantages and heterogeneous consumer tastes. Firms may choose between foreign direct investment, exporting or not serving the foreign market at all. Firm heterogeneity allows endogenously determining these decisions (as well as those on quantity and prices) as a function of factor productivity, like in Helpman et al. (2004). If the firm chooses investment abroad, part of the production remains at the domestic headquarters and is supplied to the foreign affiliate via intra-firm trade, like in Grossman and Rossi-Hansberg (2008). As a consequence, foreign investment may actually be associated with increased domestic activity and tax payments. In this case, according to the authors favouring the exemption system, a tax on foreign profits unnecessarily reduces the multinational's firm activity without benefiting (or even by harming) the domestic economy. The model presented in this paper allows asking whether this view is correct and how the choice of the optimal tax rate on foreign profits looks like in such a setting.

As the main result, the model shows that, even if foreign investment increases domestic activity and tax payments, a tax on foreign income is optimal for efficiency purposes. It turns out that the standard result proves to be robust in this setting. The nationally optimal tax system implies full taxation after deduction of foreign tax payments. The reason is that firms themselves take into account that domestic profits increase in response to foreign investment. Without a full tax on foreign income, social and private interests diverge. Moreover, I find that, if the home country is able to manipulate the foreign price level, the incentive to levy a tax on foreign profits may even increase.⁷ From a global point of view (i.e. accounting for the welfare of the foreign country's household), the tax rate on foreign profits is inefficiently high. In contrast to the standard literature, the globally optimal tax system may imply exemption of foreign income.

To set the contribution of this paper in a wider context,⁸ it is useful to consider how the literature on optimal foreign profit taxation evolved after the seminal achievement by Musgrave (Richman, 1963). Essentially, the literature has dealt with a number of extensions concerning the assumption of a fixed capital stock (Horst, 1980; Keen and Piekkola, 1997), the implementation of double taxation agreements in a strategic multi-country setting (e.g., Janeba, 1995; Mintz and Tulkens, 1996; Davies, 2003), the role of deferral (Dharmapala et al., 2011) and headquarters mobility (Voget, 2011). Moreover, the implications of alternative forms of investment like R&D spending (Grubert and Mutti, 1995) and mergers and acquisitions (Desai and Hines, 2003, 2004; Becker and Fuest, 2010) have been considered. For the purpose of this paper, extensions regarding the multinational firm's investment behaviour and the introduction of a world capital market are most important. If a country is small relative to the world capital market, capital is virtually infinitely available at a fixed interest rate. Then, investment abroad need not be associated with reduced investment at home, as both investment levels are effectively determined by the world market interest rate. In this setting, there is no need to tax foreign income for efficiency reasons (see Grubert and Mutti, 1995; Mintz and Tulkens, 1996; Devereux, 2004, for a discussion).⁹ However, these studies (implicitly) assume perfect competition and, thus, miss some features of multinational firm activity that have been stressed by recent empirical studies.

Given the literature with its broad range of assumption sets and modelling choices, the question arises which model to choose to answer the research question brought up by the recent debate on switching to the exemption system. I have two answers to this question. First, the model should capture all the stylised facts which the exemption proponents have indicated to be crucial for their argument and which the recent empirical studies have proven to be robust. These are firm heterogeneity, imperfect competition and the simultaneous existence of exporting and FDI. The model should yield that domestic activity may react positively to foreign investment within the firm (Desai et al., 2005) and negatively in aggregate (Feldstein, 1995). Second, the proponents of the exemption system themselves recommend a model framework in the tradition of Melitz (2003) for deriving optimal tax rates on foreign income: ‘This new evidence [...] suggests that further explorations of the application of these models to the question of taxing foreign profits would be highly profitable’ (Desai, 2009, pp. 11–12). It turns out, though, that a Melitz (2003) framework with FDI, as in Helpman et al. (2004), cannot replicate the stylised fact that domestic activity increases in response to foreign investment. Therefore, I adopt model features from Grossman and Rossi-Hansberg (2008) where the multinational firm is allowed to allocate different production ‘tasks’ across locations within the firm. This gives rise to intra-firm trade and links the two activity levels at home and abroad with each other.

The remainder of the paper is organised as follows. In the next section, I present the model and Section 3 outlines the results for nationally and globally optimal tax policy. Section 4 concludes.

2. The Model

As indicated above, the model used in this paper has two building blocks: the first adopted from Melitz (2003) and Helpman et al. (2004) and the second from from Grossman and Rossi-Hansberg (2008). The integration of these two model types becomes necessary because of the tax focus in this paper which will be clarified later on.

Consider a world with two countries labelled home (h) and foreign (f). In each of these two countries, there are a representative household and many heterogeneous firms.

a. Households

The representative consumers in the home and the foreign country derive utility U_h and U_f, respectively, from a numéraire good y and a variety of differentiated goods X. The differentiated goods are either produced in the home country, then denoted as x_h with index i, or in the foreign country, denoted as x_f with index j. To keep things simple, I assume that the household preferences in both countries are equal. Therefore, location indices for representative consumers are omitted until misunderstandings may arise.

The utility function is given by

• ((1)))

where γ and σ are preference parameter, N_h and N_f are the numbers of home and foreign produced varieties, respectively, and σ > 1. I further assume that which ensures that each first derivative of (1) with respect to x_h(i) and x_f(j) describe a utility maximum (see Chor, 2009, for a similar modelling strategy).

The budget constraint is given by

• (2)

where I denotes the household's after-tax income which is the sum of wage income, after-tax profits of firms and lump-sum transfers from the government. Prices are given by p_h, p_f and unity for the numéraire good y. I assume that all home firms belong to the home country's representative household and all foreign firms to the foreign household.

Substituting the budget constraint into the utility function yields:

• (3)

The utility-maximising quantities of x_h(i) and x_f(j) can be written as

• (4)

where is the price index.¹⁰ It can be shown that the representative household's utility can be expressed as

• (5)

where W is the household's wage income, Π is dividend income from firms belonging to the household and T is a lump-sum transfer from the government financed by source-based business taxes.

b. Firms

The numéraire good y is produced by firms in both countries with constant returns to scale technology under perfect competition. Labour productivity in this sector is identical in both countries. The numéraire good y can be freely traded across borders which effectively equalises wages in both countries.¹¹

Upon entering the market, each home firm draws a productivity level, given by 1/a(i), where a(i) is the amount of labour the firm i needs to produce a unit of x(i) (cf. Helpman et al., 2004). Firms differ in a(i). For simplicity, assume that a is uniformly distributed over the interval [a⁻, a⁺] and that a increases in the index i. Similarly, each foreign firm draws a productivity level a(j). For simplicity, I assume that the distributions of a(i) and a(j) are independent of each other and that ∂a(i)/∂i = 1 and ∂a(j)/∂j = 1.

In principle, all firms in the differentiated goods sector are allowed to serve both markets. As both markets are perfectly separated and the paper's focus is on optimal repatriation taxation by the home country government, I will focus on the foreign market in what follows.

Home country firms either export their goods to the foreign market (export regime, denoted by subscript E) or invest in own production facilities in the foreign country (FDI regime, denoted by subscript I). In the export regime, the goods are produced in the home country and sold in the foreign country. Production requires a fixed cost of F_E units of labour input. Variable costs are the wage rate w (the index of which has been omitted because wage rates are identical in both locations) grossed up by the transport cost τ > 1. After-tax profits are then given by

• (6)

where t_h is the corporate tax rate in the home country.

The individual firm chooses x(i) given the choices of all other firms in the economy. It also assumes that its choice has no impact on the price index P_f. Then, profit-maximising production choices under the export regime yield a price of

• (7)

If the firm decides to invest in foreign production facilities, production is shifted to the foreign country. However, a fraction θ_I  ∈  [0,1] of production remains at the headquarters (think of management or technology services; Grossman and Rossi-Hansberg (2008), use the term ‘tasks’ that can be allocated across locations within the firm). For accounting and tax purposes, the foreign affiliate has to purchase the part of the goods produced in the home country from the home country headquarters at a price of δ. Shifting production abroad may lower the variable production cost, because there is a transport cost related to exporting. However, foreign direct investment involves a higher fixed cost of production, F_I > F_E. After-tax profits are given by

• (8)

where t_e is the effective tax rate on foreign income equal to t_e = t_f + t_r(1−t_f) and t_r is the statutory tax rate on repatriated foreign profits. The profit maximising price under the FDI regime is given by

• (9)

If the transfer price exactly reflects the cost of the headquarters input, δ = τw, the price p_I does not depend on tax rate differentials between t_h and t_e.¹² The reason is that, at δ = τw, there is no taxable profit at the headquarters location. As variable costs are deductible at the foreign affiliate and marginal profits are zero, the effective tax on foreign profits, t_e, does not play a role either. However, if δ > τw, part of the foreign profit is shifted to the home country headquarters via intra-firm trade. If the effective tax on headquarters income is higher than on affiliate income, t_h > t_e, this drives up the variable cost and, thus, the profit-maximising price. Put differently, an increase in the repatriation tax t_r reduces the price: . In the following, I will allow for cases in which δ≠τw and consider δ = τw as a special case.¹³

Finally, foreign firms have an after-tax profit of

• (10)

and charge profit-maximising prices of

• (11)

3. Optimal Tax Policy

a. Equilibrium and Welfare

The focus of this paper is on the welfare and efficiency properties of repatriation taxation. I therefore take the tax rates t_h and t_f as given.¹⁴ Consider the following three decision stages. In the first stage, the home country sets the tax rate t_r on repatriated foreign income. In the second stage, all firms choose whether or not to produce and the home country firms choose between the export and the FDI regime. In the third stage, all producing firms and the representative households choose their quantities.

The decisions in the third stage are implicitly determined by equations (7, 9 and 11). The second stage decisions are characterised by three equations which define different margins. At the first two margins, the marginal home and foreign firms make zero profits. Firms with a labour productivity below and , respectively, do not have an incentive to enter the market, where and are defined by π_h(N_h) = 0 and π_f(N_f) = 0 and N_h and N_f denote the indices of the marginal home country and foreign firms, respectively. At the third margin, the marginal firm is indifferent between exporting and FDI. The threshold level is defined by π_E(n^c) = π_I(n^c) where n^c is the index of the marginal firm and

• (12)

• (13)

which are restatements of equations (6 and 8) with optimally chosen prices equations (7 and 9).

Before I analyse the welfare implications of home country tax policy in the next subsection, it is worth briefly discussing some specific features of the model outlined above. First, only the more productive firms, that is, firms with a labour productivity above choose the FDI regime if which is demonstrated in the Appendix. I only consider cases in which this condition holds. The price level on the foreign market is then given by . Second, FDI increases output as long as p_I(i) < p_E(i). Third, FDI may cause domestic activity to rise. Here, domestic activity – measured in labour input – does not decline if . Note that, without intra-firm trade (θ_I = 0), domestic activity would always decline in response to foreign investment which is why the model part adopted from Grossman and Rossi-Hansberg (2008) is needed.

All these features are backed by empirical evidence. As indicated in the introduction, some authors presume that this changes the conditions for optimal tax policy substantially. This is to be examined in the next subsection. Before doing so, welfare has to be defined.

In line with the literature, I assume that the home country government maximises the representative consumer's utility U_h which is the sum of wage income, firm profits, tax revenue and consumer surplus in the differentiated goods sector. This implies that the government has two kinds of incentive to levy a tax, that is, to redistribute funds from the private to the public sector. First, it may want to change the firms’ or households’ decisions for allocative efficiency reasons. Second, it may want to extract rents from foreigners. As the focus of this paper is on the optimal choice of the repatriation tax levied by the home country which only concerns home country firms, it is possible to abstract from the second incentive to levy taxes by assuming that tax rates t_h and t_f are given (which is also in line with the literature, see Richman, 1963, and the subsequent contributions cited above).¹⁵

Because of the quasi-linearity of the utility function, the household's wage income and consumer surplus in the home country are not affected by variations in t_r.¹⁶ Firm profits can be expressed as

• (14)

and tax revenue as

• (15)

b. The Optimal Repatriation Tax

The tax rate on foreign profits, t_r, is optimally chosen if with

• (16)

It can be shown that .

For purpose of better understanding, assume for a moment that the home country government neglects its impact on the foreign price level, that is, it assumes that . Then, optimal tax policy requires t_r = t_h. Thus, the optimal tax system implies full taxation of foreign profits after deducting foreign tax payments. This is the standard result first derived by Musgrave in 1963. The novel result here is that even though the foreign investment generates income and tax revenue in the home country, the home country government has an incentive to fully tax foreign income after deducting foreign tax payments. Note that this is even true if the home country tax revenue of the marginal firm is larger under the FDI regime than under the export regime.¹⁷

Accounting for policy effects on the price level, , introduces a strategic aspect into the analysis which has, in the context of the model, two dimensions. First, because an increase in home country firms’ quantities reduces the market shares of foreign firms, tax policy might be used to extract rents from foreign monopolists, like in Brander and Spencer (1985) and Eaton and Grossman (1986). Second, accounting for the price level widens the perspective to the aggregate level. As mentioned above, empirical analysis finds that, whereas foreign investment is associated with increased domestic investment within the firm (Desai et al., 2005), on the aggregate level, an increase in foreign investment crowds out domestic investment nearly dollar for dollar (Feldstein, 1995). Policymakers might have the incentive to account for this effect on other domestic firms when deciding on tax policy strategies.

The effect of a small increase in t_r on the price level in the foreign country is given by

• (17)

The first term on the right hand side captures the direct effect of t_r on prices p_I(i). The second term represents the price level effect of a tax induced change at the export-FDI-margin. The third and fourth terms are the changes at the market entry margin of home and foreign firms, respectively.

The Appendix demonstrates that is ambiguous. This is because two countervailing effects are at work. Given that an increase in t_r reduces FDI and p_E(n^c) > p_I(n^c), this has, ceteris paribus, a positive effect on the price level. However, an increase in t_r also reduces the prices of all intra-marginal firms under the FDI regime (as long as δ > τw). This has, ceteris paribus, a negative effect on the price level. It can be shown that for small δ approaching the headquarters variable cost τw, the effect of a tax increase is positive: .

With , it follows that, for t_r = t_h, an increase in t_r still increases welfare if and vice versa.

Proposition 1.  (i) If the home country government does not take into account the effects of its tax policy on the price level, the optimal tax on foreign profits is t_r = t_h (full taxation after deduction) and (ii) If an increase in t_r increases the price level P_f, , the optimal tax rate on foreign profits t_r exceeds the tax rate under full taxation after deduction, that is, t_r > t_h.

What is the intuition behind these findings? Consider the first part (i) of the above proposition. If, at t_r = 0, tax revenue in the home country are lower if the firm chooses FDI (as in the Musgrave model), the government has an incentive to increase t_r to force the marginal firm back into the export regime. In contrast, if tax revenue is larger than under the export regime (as suggested by recent empirical evidence), the government has an incentive to subsidise FDI which can be achieved by increasing t_r (recall that the foreign part of the tax base is then negative). Finally, if tax revenues from the marginal firm are equal under both regimes, the foreign part of the tax base is zero. Any tax rate is optimal and increasing t_r does not harm the firm or the economy. Thus, in all these cases the government has an incentive to increase t_r until t_r = t_h. Then, private and social interests are aligned.

Now, consider part (ii) of the above proposition which introduces the strategic effect of price level manipulation. An increase in t_r reduces the number of firms under the FDI regime. The marginal firm is just indifferent between FDI and exporting. However, all firms profit from an increase in the price level. As a consequence, the sum of firm profits and tax revenue increases. Seen from a different perspective, the tax internalises an external effect which the firm investing in FDI does not take into account. By having lower variable cost, the price level is reduced which then translates into lower profits for all other firms. Forcing the firm back into exporting thus corrects for this (from a national perspective) inefficiently high level of FDI.

The results in Proposition 1 are diametrically opposed to the views expressed by the exemption proponents. From my point of view, there are two potential sources of misunderstanding, that is, reasons which have led to the mistaken presumption that exemption is the optimal answer to a situation in which domestic activity by heterogeneous firms increases in response to foreign investment. The first concerns the investment behaviour by firms. It seems that some of the commentators treat the domestic income increase as an external effect of foreign investment which the firm does not account for in its investment decision. However, as long as there are no grave principal-agent issues within the firm, the firm will account for it and invest until the sum of the returns in both locations equals the cost of production. The second source of misunderstanding might be identified in the differentiation of marginal from intra-marginal firms. Seemingly, some commentators have firms in mind which generate positive income abroad and at home. However, such a firm is not at the margin, that is, it will not react to small changes in the tax environment. The model shows that marginal firms have either positive income at home or abroad, but not both.

c. Global Optimality

By assumption, the nationally optimal tax policy by the home country does not take into account the foreign household's utility U_f. Again, it is worthwhile to recall that, in the Musgrave model, the nationally optimal tax policy (full taxation after deduction) implies inefficiently high tax rates from a global point of view. The reason is that the home government considers foreign taxes as costs although, from a global viewpoint, taxes are just funds redistributed from the private to the public sector. What are the efficiency features of nationally optimal tax policy in our model?

Global welfare is simply the sum of home country and foreign welfare. The latter is given by

• 

where is consumer surplus. Foreign profits Π_f are given by

• (18)

and foreign tax revenue by

• (19)

The question arises how t_r affects the foreign household's utility. Global welfare W_g can be expressed as

• (20)

Proposition 2.  If , a reduction of t_r starting from t_r = t_h unambiguously increases global welfare. The optimal tax system implies t_e < t_h.

Proof: See Appendix. ▪

The intuition behind this proposition is the following. An increase in t_r if optimally chosen by the home country government does not affect home country welfare. From the viewpoint of the foreign country, it increases the price level, reduces the number of home country firms in the foreign country, and increases entry of foreign firms into the market. It can be shown that, first, an increasing price level reduces consumer surplus more than it increases firm profits and tax revenue and, second, that the loss in welfare because of a reduction of home country firms’ FDI dominates the welfare gain because of increased market entry.

The above proposition has three important implications. First, the tax credit system is not a globally optimal tax policy choice. Second, it cannot be concluded that the exemption system is globally optimal. At t_e = t_h, (credit system), it is still desirable to reduce the tax to increase consumer surplus. Third, because the tax credit system replicates the allocation, prices and quantities in the absence of taxes, the above proposition implies that a globally optimal tax system with t_e < t_h attains a higher global welfare level than in the absence of taxation. The reason is that the market is characterised by imperfect competition. There is too little FDI in the absence of taxation because firms do not account for the resulting decrease in consumer prices.

4. Conclusion

One of the recently promoted arguments in favour of the exemption system goes as follows: if foreign investment does not affect or even fosters domestic activity, taxing foreign income is no longer desirable. In this paper, I build a model with heterogeneous firms and intra-firm trade and derive the choice of the optimal tax rate on foreign profits. It turns out that, even though foreign investment may be associated with increased domestic investment and higher domestic tax revenues, the optimal tax system implies full taxation of foreign income after deducting foreign tax payments. Thus, exemption is not an optimal choice from the national point of view. The standard view on foreign profit taxation prevails. However, the globally optimal tax rate is lower than the one chosen under a tax credit system (which has been considered globally optimal in the standard model). This may imply, under certain circumstances, that exemption is optimal from a global point of view.

Having stated that some of the classical Musgrave results remain robust, it should be noted that the analysis in this paper is not just a trivial replication of the standard theory. The most important difference between the classical framework and the one presented here concerns the question how domestic and foreign activity are related. In the Musgrave framework, savings are distributed across locations. A dollar invested abroad cannot be invested at home – investment projects in both locations are therefore substitutes. In contrast, in the above-presented framework, an increase in activity abroad (FDI) does not necessarily reduce activity at home. One of the main insights of this paper is that the desirability of taxing foreign income does not depend on the effects of foreign activity on domestic activity (e.g. measured by tax revenue). Moreover, capital productivity in the Musgrave framework entirely depends on the location. A dollar of additional investment reduces the marginal productivity of capital in a given location and vice versa. In the framework presented here, productivity is firm-specific. Furthermore, the Musgrave model assumes homogeneous firms, whereas the framework considered here explicitly allows for firm heterogenity. This allows determining endogenously which firms produce, which ones export and which ones invest in foreign production facilities. Finally, this paper explicitly allows for imperfect competition. This difference to the classical framework is important because a tax on foreign profits is sometimes interpreted in the context of strategic trade policy because, under imperfect competition, national firms may extract rents from foreign markets and thus hurt their competitors.¹⁸ As this paper demonstrates, a tax on foreign profits is nevertheless optimal. Imperfect competition is also the reason that the global optimality of the tax credit system does not hold anymore. As it is shown above, proponents of the exemption system should argue more with the notion of global optimality than with the national interest.

Of course, the model results should be seen in the light of the many restrictions because of the specificity of the model assumption. As the literature cited in the introduction shows a change in assumption is likely to translate into a change in the recommendation for optimal tax policy. In fact, policymakers are aware of this conditionality; for instance, a US Treasury report states that ‘[n]one of the proposed standards [of international taxation] fits all cases and tax policy cannot feasibly be calibrated to have different rules for different cases’ (United States Department of the Treasury, Office of Tax Policy, 2007).¹⁹ It should therefore be recalled that the purpose of this paper is to analyse specific arguments based on the empirical finding that foreign investment need not be associated with reduced domestic activity. The paper's aim is to check whether tax exemption is an optimal tax policy response in this economic environment. It turns out that in such a situation, exempting foreign income from tax is not an optimal strategy although there may be other good arguments in favour of doing so.

Footnotes

• 1

  See http://www.unctad.org (World Investment Report), http://www.bea.gov and http://www.statistics.gov.uk (number for UK from 2007: GBP 89,855 (exchange rate from 31/12/2007).

• 2

  See e.g. HM Treasury and HM Revenue and Customs (2007).

• 3

  See e.g. United States Department of the Treasury, Office of Tax Policy (2007).

• 4

  Other standard references are Bond and Samuelson (1989) and Bucovetsky and Wilson (1991).

• 5

  See e.g. Desai and Hines (2003,2004), Hines (2008), Desai (2009). For an overview of the debate, see Becker and Fuest (2011).

• 6

  See e.g. Egger and Pfaffermayr (2003), Simpson (2008), Desai et al. (2009).

• 7

  Accounting for tax effects on the foreign price level may capture the case in which foreign and domestic investment are complementary within the firm (as in Desai et al., 2005) but are substitutes in aggregate (as demonstrated by Feldstein, 1995) because foreign investment by one firm crowds out activity by other firms.

• 8

  Seen from another perspective, this paper contributes to a literature which confronts recent findings in international trade economics and multinational firm analysis with public finance issues, see e.g. Davies and Eckel (2010), Haufler and Stähler (2010), Krautheim and Schmidt-Eisenlohr (2011).

• 9

  This is true, as long as the domestic tax rate is assumed to be given and no revenue requirement is binding. If the domestic tax rate is endogenised, the government faces a Ramsey style problem of optimisation where optimal tax rates on domestic and foreign income reflect the locational elasticities.

• 10

  Note that, because of the quasi-linearity of the utility function, the utility-maximising quantities do not depend on the income level. Of course, this is different in a world without a numéraire good, which will be discussed in Section 3.

• 11

  I assume that the labour force is large enough such that there is always a positive quantity of y produced in each country.

• 12

  Of course, the same is true if θ_I = 0, that is, all production takes place at the foreign affiliate and no intra-firm trade occurs.

• 13

  Throughout the paper, the transfer price δ will be treated as an exogenous variable. This assumption may be questioned because firms often have some discretion in manipulating transfer prices for tax saving purposes. Moreover, in bilateral tax agreements, national governments often agree on some system of transfer price rules that effectively split the tax base according to some notion of fairness. It should be noted, though, that the results derived here do not depend on the actual level of transfer prices, as is shown later on.

• 14

  This assumption is in line with the standard literature. However, as demonstrated in Devereux (2004), it is of crucial importance. If tax revenue from foreign income can be used to reduce domestic taxes (or increase public goods provision), levying taxes on repatriated dividends may yield efficiency gains. In this setting, the government faces a Ramsey style problem of optimal taxation where optimal tax rates reflect locational elasticities. It should be noted, though, that apart from extreme situations exemption of foreign income is not a likely outcome if domestic taxes are endogenised.

• 15

  Another option would be to assume that the households have a preference for a publicly provided good. However, such an analysis would also require to endogenise the tax rates t_h and t_f.

• 16

  Actually, Desai (2009, p. 11) claims that a tax on foreign profits reduces competition in the home country market–‘Overall welfare is reduced at home given the lowered competition amongst firms’. This may be true if foreign investment lowers variable cost associated with home country production, for example, because of higher R&D activity. These effects would require another model, though, and are therefore neglected in the following.

• 17

  Higher domestic tax revenue (at t_r = 0) implies that the foreign part of the firm's tax base is negative. Then, an increase in t_r subsidises investment and increases the domestic gains in tax revenue. At t_r = t_h, the gains in tax revenue equal the cost of the subsidy.

• 18

  In the context of monopolistic competition, negative external effects of increasing quantities on other firms’ profit levels occur because the price level is affected.

• 19

  Devereux (2008, p. 711) takes this finding as an argument in favour of exempting foreign income from tax: ‘If governments cannot design tax systems which vary according to firm or sector characteristics, but are forced to implement general tax structures, then [...] the underlying presumption should be in favour of a system which does not tax foreign income of domestic corporations.’

• I thank May Elsayyad, Clemens Fuest, Andreas Haufler, Frank Stähler and participants at seminars in Munich for valuable comments.

Appendix

Appendix A: Productive firms choose FDI

Appendix A derives the conditions under which the more productive firms choose the FDI regime and the less productive firms prefer exporting. A firm is indifferent between exporting and FDI at π_E(n^c)−π_I(n^c) = 0. With and , an increase in n^c has the following impact

• 

It follows that the above expression is positive, that is, the more productive firms choose FDI, if

• 

What is the effect of a small increase in t_r on the export-FDI-margin, that is, on n^c? Differentiating π_E(n^c)−π_I(n^c)=0 with respect to n^c and t_r, gives where is derived above. is given by

• 

It follows that is negative if which is the case if

• 

This equation can be expressed as

• 

It follows that as long as the foreign part of the firm's tax base is positive.

Appendix B: Changes in the price level

In this appendix, I derive the effect of a small increase in t_r on the foreign price level P_f. The export-FDI-margin is characterised by π_E(n^c) = π_I(n^c), the entry margin for home country firms by π_E(N_h) = 0, the entry margin for foreign firms by π_f(N_f) = 0 and the price level is given by

• 

These four equations define the endogenous variables n^c,N_h,N_f and P_f. The total differential is given by

• 

This system of four equations allows determining dn^c, dN_h, dN_f and dP_f for an exogenous change of dt_r. After solving for dn^c, dN_h, dN_f and replacing them, the effect of a small increase in t_r on the price level can be expressed as

• 

Appendix C: Global optimality

Here, I derive the first-order condition of t_r for global welfare. Adding home country profits in equation (14), home country tax revenue in equation (15), foreign profits in equation (18), foreign tax revenue in equation (19), and foreign consumer surplus given by gives

• 

The first-order condition with respect to t_r reads

• 

which can be simplified, using , to

• 

Provided that , the consumer surplus decreases more than firm profits are increased (first term). The first term in square brackets which captures the effect of the marginal firm's regime switch on global tax revenue is unambiguously positive. The second term in square brackets depicts the effect of the intra-marginal firms’ price changes on global tax revenue which is positive, too. Thus, the sign of the whole term depends on the sign of t_h−t_e.