International Trade Agreements, Environmental Policy, and Emergence of Multinational Firms

Transcription

1 International Trade Agreements, Environmental Policy, and Emergence of Multinational Firms Essi Eerola FPPE, University of Helsinki February 2002 Abstract During recent decades there has been substantial momentum for trade liberalization. At the same time, multinational production has become increasingly important in international competition. In this paper we show that these phenomena may be linked through environmental policy. When the national governments cannot use direct trade policy measures, environmental policy is distorted towards enhancing competitiveness of the domestic producers. In this environment, signing agreements that prevent the use of trade policy may lead to Þrms investing in production capacity abroad. Keywords: Environmental policy, foreign direct investment, multinational production, trade agreements JEL ClassiÞcation: H23, H77, F18, F23 1 Introduction During the last few decades multinational production has become increasingly important in international competition. An overwhelming majority of foreign direct investment (FDI) ßows are between industrialized countries and to a large extent multina- I wish to thank Markus Haavio, Erkki Koskela, Marko Lindroos, Niku Määttänen, Markku Ollikainen, and the participants in the FPPE Workshop for Microeconomics for useful comments. The usual disclaimer applies. This paper is part of the project Studies in Environmental and Resource Economics Þnanced by the Academy of Finland. The funding is gratefully acknowledged. Department of Economics, P.O. Box 54, FIN University of Helsinki, Finland; essi.eerola@helsinki.þ; Tel: ; Fax:

2 tional Þrms spread horizontally, that is, exports are at least partly replaced by production in foreign country. At the same time, there has been increasing pressure towards trade liberalization through international agreements, for instance, within the WTO and its predecessor the GATT. International free trade agreements prevent national governments from using direct trade policy measures, for instance, to enhance the competitiveness of domestic producers. Environmental policy, in turn, has largely remained at the discretion of national governments. It is well established that under certain conditions this may lead governments to use environmental policy both to correct negative externalities caused by emissions and to trade related purposes. In this paper, we study how free trade agreements shape national environmental policies and inßuence location strategies of polluting industries. The main questions of interest are the following: First, under what conditions does national environmental policy induce Þrms to invest in production capacity in a foreign country? Second, how does the change in the international institutional regime inßuence the incentives to invest abroad? A partial list of important papers analyzing the design of environmental policy in the absence trade policy would include Rauscher (1994), Barrett (1994), Kennedy (1994), and Ulph (1996). These papers show that environmental policy may be distorted in favor of domestic producers leading to so-called environmental dumping. Walz and Wellisch (1997), in contrast, consider both trade and environmental policy and analyze whether, given the tendency for environmental dumping, a trade agreement that prohibits the use of trade policy increases welfare. All these papers focus on a situation where polluting Þrms are purely national. The location decision of polluting Þrms is in turn analyzed in Markusen et al. (1995), Rauscher (1995), and Hoel (1997) in a setting where national governments Þrst choose environmental policies and after that a Þrm or Þrms decide where to locate. The governments are assumed to be able to commit to chosen policies after investment costs have been sunk, and unless transportation costs are high enough, Þrms invest in only one country. This literature reveals important insights about the behavior of governments, but it does not consider how attractiveness of FDI depends on national policies and the international regime. Following the emergence of multinational Þrms, a substantial theoretical and empirical literature analyzing the rationale for multinational production has developed. However, also this literature largely overlooks the inßuence of explicit changes in the international institutional regime on the emergence of multinational production. Brainard (1993) and Horstman and Markusen (1992) show in a strategic trade policy framework that multinational Þrms are more likely to arise if plant-level Þxed costs and barriers to investment are low and Þrm-level Þxed costs and transportation costs are high. Markusen and Venables (1998) further show that multinational production occurs when 2

3 countries are similar in size as well as in relative endowment. 1 In addition, investing in production capacity in two countries may be proþtable if a Þrm wishes to protect itself against time inconsistent governments that would otherwise conþscate all proþts after investment has been sunk (Janeba 2000), or if a Þrm wants to strengthen its position against national unions (Zhao 1995). In this paper we build a two country model with two polluting Þrms in the spirit of the strategic trade policy literature introduced by Brander and Spencer (1985). Initially each Þrm is established only in one country, has two production plants, and sells its product to a third market. The Þrms can invest abroad by closing one plant in their home country and opening a new plant abroad. We analyze the incentives of the Þrms to do that under two different regimes. Under the Þrst regime, national governments may freely use both trade policy and environmental policy. Under the second regime, a trade agreement prohibits the use of direct trade policy measures. When using environmental policy, the national governments are constrained to set a uniform tax rate on all polluting plants within their territory. 2 As we want to examine how national policies inßuence Þrms strategies under the two international regimes, we concentrate on a situation where the incentives for FDI are purely policy driven: The only potential advantage of being multinational is to be able to shift production from one country to another and thereby escape stricter environmental regulation. Our main result is that a change in the international regime that restricts the instrument set available to the national governments may lead to multinationalization of production. When the governments cannot use explicit trade policy measures, polluting Þrms are more likely to invest abroad. The reason is rather surprising. When the Þrms are national, governments use environmental policy not only to correct the negative externality caused by production but also to shift proþts to domestic shareholders. By opening a production plant abroad a Þrm is able to break this link between the domestic producer and the government. Foreign owned plants not only eliminate the possibility to use environmental policy to shift proþts to domestic shareholders but also create an incentive to tax proþts accruing to foreign shareholders. In contrast, when the governments can use both environmental and trade policy, an investment abroad does not have similar strategic effects. The paper is organized as follows. Section 2 presents the model. In sections 3 and 4 we solve the equilibrium policies and industry structure. Section 5 concludes. 1 For a survey of this literature, see e.g. Markusen (1995). 2 This would be the case, for instance, within the EU single market. 3

4 2 The Model We develop a model of two identical countries, home and foreign, and a polluting industry consisting of two Þrms, home (H) andforeign(f ). Each Þrm has two production plants and sells its product to a third market. When the Þrms are national, the two production plants of each Þrm are located in one country. When the Þrms are multinational, each Þrm has one plant in home and one in foreign country. Total output of Þrm i, y i,is y i = x i + x i, where i = H, F and x is plant level production. If Þrm i is multinational, x i refers to production in home country and x i refers to production in foreign country. Revenue of Þrm i is p (Y ) y i, where p (Y ) is the inverse demand function in the third market and Y = y H + y F aggregate supply. Throughout the paper we assume that: is Assumption 1 Demand function satisþes p 00 (Y ) y i + p 0 (Y ) < 0 for i = H, F. Same abatement technology is available for the Þrms in both countries. We normalize units of emission so that emissions generated by Þrm i are x i q i + x i q i, where q i and q i are the abatement levels of Þrm i in its two plants. In the case that Þrm i is multinational, (x i q i ) refers to emissions generated in home country and (x i q i ) to emissions generated in foreign country. Level of emissions in home country is denoted by E and in foreign country by E. Environmental damage caused by emissions is d (E) in home country and d (E ) in foreign country. Trade policy takes the form of export subsidies for domestic producers. 3 Environmental policy, in turn, consists of choosing emission tax levied on the Þrms. Each 3 This is a natural instrument to consider in a setting where polluting industry exports its product. In a setting where there are markets for the product of the polluting Þrms in the host countries, an import tariff for instance would have similar consequences. 4

5 government levies a tax on emissions generated in its own country and cannot discriminate between Þrms according to their ownership. Tax revenues are distributed back to consumers as lump sum transfers. We consider policy design under two different regimes. Under the Þrst regime, national governments may use both trade and environmental policies. Under the second regime, a trade agreement between the two countries prevents them from using export subsidies. There are four possible industry structures denoted by j {nn, mm, nm, mn}, where n denotes national and m multinational Þrm, and j = nm refers to the industry structure where home Þrm is national and foreign Þrm is multinational. As all results will be symmetric when j = nm and j = mn, we study policy design only under industry structures j = nn, j = mm, andj = nm. Investment abroad in the form of opening a production plant in a foreign country and closing a plant in home country must entail some cost for the investing Þrm. We represent all the costs related to this investment by a Þxed cost G for both Þrms. Under industry structure j, total variable costs of Þrm i are c i u; x i,x i,q i,q i ; j = c x i + c x i + g q i + g q i + γ i u; x i,x i,q i,q i ; j (1) where u =(t, t,s,s ) is the vector of policy variables. Terms t and t are the emission tax rates in home and foreign country. Terms s and s in turn are the export subsidies for the domestic Þrm in home and foreign country. The Þrst two terms in (1) denote the cost of production in each plant. The two following terms denote the cost of emission abatement in each plant. The last term in (1) is the total tax burden of each Þrm and depends on the location of its production plants. For home Þrm γ H u; x H,x H,q H,q H ; nn = γ H u; x H,x H,q H,q H ; nm = t x H q H + t x H q H s x H + x H and γ H u; x H,x H,q H,q H ; mm = γ H u; x H,x H,q H,q H ; mn = t x H q H + t x H q H s x H + x H. We will assume that: Assumption 2 All cost functions are strictly convex, i.e. c 0 ( ) > 0, c 00 ( ) > 0, g 0 ( ) > 0, g 00 ( ) > 0, d 0 ( ) > 0, andd 00 ( ) > 0. We use convex production costs in order to rule out the unrealistic event that when Þrms are multinational, a marginal change in environmental policy induces Þrms to shift all production from one country to another. 4 4 We assume that Þxed cost of investment when opening new production plants and the size of the export market are such that despite convex production costs, only two Þrms exist in the market. 5

6 Timing of the events is as follows. The Þrms Þrst decide whether to remain national or to invest abroad. Investing abroad implies opening a new plant in different country and closing one production plant in home country. The investment decisions of the Þrms determine the industry structure. The national governments of the two countries then choose their trade and environmental policies. Finally, after policies have been chosen Þrmsdecidehowmuchtoproduceandabateineachplant. 5 3 Output, Abatement, and Policy Choices In this section, we determine the output and abatement decisions of the Þrms and policy decisions of the governments under the different industry structures. We begin by considering the proþt maximization problem of the Þrms and then analyze policy design under the two regimes. 3.1 Output and Abatement Choices Given the industry structure and the policies chosen by the national governments, the two Þrms choose output and abatement levels in their two plants so as to maximize proþts. ProÞt ofþrm i is denoted by π i.proþtmaximization problem of Þrm i is max π i = p (Y ) y i c i u; x i,x i,q i,q i ; j ª x i,x i,q i,q i where c i is determined by equation (1). Assuming interior solutions, Þrst-order conditions for proþt maximization for Þrm i are 6 π i x i = p 0 (Y ) y i + p (Y ) c i x u; x i,x i,q i,q i ; j =0, i (2) π i x i = p 0 (Y ) y i + p (Y ) c i x u; x i,x i,q i,q i ; j =0, i (3) π i q i = c i q u; x i,x i,q i,q i ; j =0, i (4) π i q i = c i q u; x i,x i,q i,q i ; j =0. i (5) 5 This timing assumption is made to capture the idea that decisions on investing in production capacity in a foreign country occur less frequently than decisions on taxes and subsidies or output and abatement. 6 Derivatives of functions of one variable are denoted by primes. Partial derivatives of functions of several variables are denoted by subscripts. 6

7 Conditions (2) and (3) imply that proþt maximizing output level in each plant is such that marginal revenue equals marginal cost of production. Conditions (4) and (5) in turn require that marginal abatement cost in each plant equals the emission tax rate in the country where the plant is located. We denote the proþt maximizing output levels as a function of the policies and industry structure by y i (u; j). The corresponding maximum proþts are denoted by π i (u; j). From the Þrst-order conditions for proþt maximization we can derive the following comparative statics results: y H t (u; j) < 0 for j = nn, mm, nm, and y F t (u; nn) > 0, y F t (u; mm) < 0, andy F t (u; nm) < 0. Regardless of the industry structure, stricter emission control in home country increases the production costs of home Þrm. Therefore, when emission tax rate is increased in home country, home Þrm reduces supply. However, the reaction of foreign Þrm depends on the industry structure. When foreign Þrm is national, its production costs are not affected by the policy change, but its rival reduces supply. In this situation, foreign Þrm increases supply. When foreign Þrm is multinational, its production costs in the plant that is located in home country increase. As a result, it reduces output in the plant located in home country and increases output in the plant located in foreign country. Due to convex production costs, its overall output declines. In the same manner, we have that y H s (u; j) > 0 and y F s (u; j) < 0 for j = nn, mm, nm. Regardless of the industry structure, a higher export subsidy in home country reduces the production costs of home Þrm and induces it to expand output. As a result, foreign Þrm cuts down on supply. 3.2 Policy Design We now turn to analyze policy design under the two international regimes. Under both regimes, each government chooses its policies in order to maximize domestic aggregate welfare and takes the policies chosen by the other government and the equilibrium behavior of the Þrms as given. In order to be able to study policies when at least one Þrm is multinational, the ownership structure of multinational Þrms must be speciþed. For simplicity, we assume that the ownership structure of a Þrm remains unchanged when 7

8 it invests abroad. Thus, when home Þrm opens a plant in foreign country, ownership of the Þrm remains entirely in home country. We Þrst analyze policy design when the governments are free to use both environmental and trade policy measures. We call this a protectionist regime. We then consider a trade agreement regime where trade policy is not available to the governments. Unless the industry structure is asymmetric, the policy design problems of the two governments are symmetric. We will therefore mainly focus on home government Protectionist Regime In each country, domestic aggregate welfare consists of proþts for domestic shareholders, tax revenue net of export subsidies, and environmental damage caused by emissions. We denote domestic aggregate welfare in home country by w andinforeigncountryby w. Thus home government chooses t and s so as to maximize w (u; j) =π H (u; j) sy H (u; j)+te (u; j) d (E (u; j)), and foreign government chooses t and s so as to maximize w (u; j) =π F (u; j) s y F (u; j)+t E (u; j) d (E (u; j)), where E (u; j) and E (u; j) areemissionlevelsinhomeandforeigncountry. The Þrst-order conditions for welfare maximization for home government are and π H t (u; j) sy H t (u; j)+e (u; j)+te t (u; j) d 0 (E (u; j)) E t (u; j) =0 (6) π H s (u; j) sy H s (u; j) y H (u; j)+te s (u; j) d 0 (E (u; j)) E s (u; j) =0, (7) where the Þrst term denotes the effect of a policy change on proþts of the domestic Þrm. The three following terms refer to changes in tax revenue net of subsidies to the domestic Þrm. The last term is the welfare-effect of a change in emissions. First-order conditions for foreign government are analogous. Under the assumption that the second-order conditions are satisþed, these Þrst-order conditions implicitly determine a unique equilibrium emission tax policy and subsidy policy for each government. We denote these equilibrium policies under industry structure j by u j =(t j,t j,s j,s j ). Results on policy design under the different industry structures are summarized as 8

9 Characterization 1 When the governments can freely design both trade and environmental policy and both Þrms are national, each government sets emission tax rate so that the negative effect of emissions is fully internalized and give export subsidies to the domestic producer. When at least one Þrm is multinational, the government of the host country of the investment sets an emission tax rate that exceeds the marginal damage of emissions and gives export subsidies to the domestic Þrm. Proof. See Appendix A. Environmental policy is not an efficient way of inßuencing the market share and proþts of the domestic Þrm as lowering environmental standards yields a cost in terms of increased environmental damage. An export subsidy, in contrast, can be efficiently used to enhance the competitiveness of the domestic producer. Therefore, when possible, the governments use environmental policy to correct the negative externality caused by production and export subsidy to increase proþts of domestic shareholders. When one or both Þrms are multinational, in at least one country, there is a production plant that is owned by foreign shareholders. The government of the host country of the affiliate ignores the proþts accruing to foreign shareholders when choosing policies. As the governments cannot discriminate between the Þrms in terms of environmental regulation, they set emission tax rates above the level of full internalization of environmental damage and then compensate the domestic shareholders by giving export subsidies Trade Agreement Regime We now turn to the trade agreement regime. The interpretation of the trade agreement is that national governments cannot use export subsidies for domestic producers and are thus restricted to set s =0and s =0. We denote the restricted policy vector by bu =(t, t, 0, 0). Under the trade agreement regime, home government chooses t so as to maximize w (bu; j) =π H (bu; j)+te (bu; j) d (E (bu; j)), and foreign government chooses t so as to maximize w (bu; j) =π F (bu; j)+t E (bu; j) d (E (bu; j)), where E (bu; j) and E (bu; j) areemissionlevelsinhomeandforeigncountry. The Þrst-order condition for welfare maximization for home government is then π H t (bu; j)+e (bu; j)+(t d 0 (E (bu; j))) E t (bu; j) =0. (8) 9

10 First-order condition for foreign government is again analogous. Under the assumption that w tt (bu; j) < 0 and wt t (bu; j) < 0, theþrst-order ³ conditions determine unique equilibrium tax rates, which we denote by bu j = b t j, bt j, 0, 0. As under the protectionist regime, we are interested in policy design under the different industry structures given equilibrium behavior of the Þrms. We summarize the results as Characterization 2 Under the trade agreement regime, emission tax rates are lower than the marginal damage of emissions when both Þrms are national. When both Þrms are multinational, tax rates in both countries exceed marginal damage. When only one Þrm is multinational, tax rate exceeds marginal damage in the host country of the affiliate of the multinational Þrm and is lower than marginal damage in the home country of the multinational Þrm. Proof. See Appendix A. The reason for these results is twofold. First, when the Þrms are national, lowering emission tax rate enhances the competitiveness of the domestic Þrm. This effect has an important role in policy design as proþts accruing to domestic shareholders constitute part of the domestic aggregate welfare that each government seeks to maximize. In contrast, when the Þrms are multinational, relaxing environmental control induces both Þrms to increase output. Therefore, changes in environmental policy do not inßuence the market share of the domestic producer and consequently environmental policy is not useful in favoring shareholders of the domestic Þrm. Second, when the polluting Þrms are multinational each government has an incentive to tighten pollution control in order to tax proþts accruing to foreign owners of the multinational Þrms. 4 Equilibrium Industry Structure We now determine the equilibrium industry structure. Each Þrm faces a decision whether to remain national or to open a new production plant abroad. When taking this decision, the Þrms take into account what kind of policies will be chosen by the governments under each industry structure and what the corresponding proþts will be. In order to able to compare proþts under different industry structures, we assume that Assumption 3 Demand is p (Y )=A Y, where A>0, and cost functions are c (x) = 1 2 x2, g (q) = 1 2 q2,andd (E) = 1 2 de2. 10

11 By these assumptions the model becomes fully parametrized and provides closed form solutions. In order to keep the presentation clear, we locate the solutions for optimal output and abatement levels as a function of the policies as well as equilibrium policies in Appendix B. Thetwostrategies available totheþrms are to invest abroad and to remain national. If the Þrms do not have a dominant strategy, we have three Nash equilibria, two in pure strategies and one in mixed strategies. Throughout this section, we concentrate on the mixed strategy equilibrium. The interpretation of the results is similar when only pure strategies are considered. However, mixed strategies are convenient in that they can be thought of as a tendency towards multinationalization. When focusing on mixed strategies, it is possible to analyze in a tractable way, how changes in the parameters of the model inßuence the attractiveness of investing abroad. We represent all costs related to opening a new plant abroad and closing one plant in Þrm s initial home countrybyaþxed investment cost, G. In order to facilitate the discussion of the results, we deþne n and m to be the gain in operating proþts from investment under the protectionist regime when the rival Þrm is national and multinational, respectively. Thus and n = π H (u mn ; mn) π H (u nn ; nn) =π F (u nm ; nm) π F (u nn ; nn) m = π H (u mm ; mm) π H (u nm ; nm) =π F (u mm ; mm) π F (u mn ; mn). Under the trade agreement regime, we denote this gain by b n and b m.weanalyzethe equilibrium proþts and location choices of the Þrms Þrst under the protectionist regime and then under the trade agreement regime. 4.1 Protectionist Regime We Þrst determine how investing abroad inßuences the operating proþts of the Þrms under the protectionist regime. By using the equilibrium policies and corresponding supply and abatement strategies of the Þrms, one can show that there exist cut-off values d 0.30 and d 1.12 for the environmental damage parameter, such that the following relationships hold: m < 0 for all d>0, n > 0 if d d, d and n < 0 if d<dand d>d. 11

12 The Þrst inequality implies that it is never proþtable to invest abroad if the rival Þrm is multinational. The following two inequalities state that when the rival Þrm is national, investment abroad increases operating proþts only under certain values for the environmental damage parameter. Figure 1 illustrates these relations. It shows the relative change in proþts when investing abroad when the rival Þrm is national and multinational. This relative change is independent of the size of the export market, determined by parameter A in the demand function Relative change in profits Rival firm national Rival firm multinational Environmental damage parameter (d) Figure 1. Relative change in proþts when investing abroad under the protectionist regime. If foreign Þrm invests in home country while home Þrm remains national the following policy changes occur. First, the government of home country uses emission taxation both to control emissions and to tax proþts of foreign shareholders. When d is small, the latter incentive dominates and investment induces a large increase in emission tax rate. If d is higher, the relative incentive to tax foreign shareholders is less important compared to the incentive to control emissions, and the reaction of emission tax rate to investment is more modest. Second, home government increases the export subsidy for the domestic Þrm. The increase in export subsidy due to investment is large when d is small. This suggests that the aim of this policy change is to compensate the domestic shareholders for the negative effect of stricter emission taxation. If foreign Þrm is multinational and home Þrm invests in foreign country, home government relaxes 12

13 emission taxation. This beneþts both home and foreign Þrm. Foreign government, in turn, increases emission taxation and export subsidy for foreign Þrm. Lemma 1 When the governments can use both trade and environmental policy, investing abroad is never a dominant strategy. Proof. Straightforward since m < 0 for all d>0 and A>0. This implies that when d / d, d both Þrms remain national regardless of the strategy adopted by the rival Þrm. A necessary condition for at least one Þrm to invest abroad is therefore d d, d. However, even if investment abroad increases operating proþts, it may still be in the interest of the Þrms to remain national. This of course happens if the Þxedcostofinvestmentistoohigh,i.e. if n <G. The following proposition characterizes the situation where there are potential gains from investing abroad, that is, d d, d and in addition G is low enough. Proposition 1 If n >G, there exists a mixed strategy equilibrium where both Þrms remain national with probability β. This probability is increasing in G and decreasing in A and is always strictly larger than 1 2. Proof. Assume that one Þrm remains national with probability β. Then the rival Þrm is indifferent between investing abroad and remaining national if β = m G m, n where m < 0 and n > 0. Thus n >G β < 1. As the Þrms are symmetric there is a mixed strategy equilibrium where the Þrms remain national with probability β. Furthermore, β G = 1 β > 0 and m n A < 0. It is straightforward to show that n >Gimplies that m > n. Assume that G =0.Then β = m m n > 1 2. It is intuitive that investing abroad is more attractive when G is small. Firms are also more likely to invest abroad when the size of the export market, represented by parameter A, increases. This is because the larger the export market, the bigger is the 13

14 gain in operating proþts when investing abroad, which reduces the relative importance of the cost of investment. However, even in the extreme case where G =0each Þrm is more likely to remain national than to invest abroad. When the Þrms are national, each government uses emission tax to correct the negative externality generated by production and export subsidy to increase the proþts of the domestic shareholders. Assume then that foreign Þrm invests in home country. This investment induces home government to set a higher emission tax rate in order to tax part of the proþts accruing to foreign shareholders. Consequently, proþts generated by the plant that is moved from foreign country to home country decrease compared to the situation where both plants remain in foreign country. Foreign government in turn lowers its emission tax rate and as a result the proþts generated by the plant that remains in foreign country increase. However, when home government tightens emission taxation it also increases export subsidy in order to compensate the negative effect of a higher emission tax rate on the domestic Þrm. This mitigates the positive effect of a higher emission tax rate in home country and lower tax rate in foreign country on the market share of foreign Þrm and thereby reduces the proþtability of the investment. Due to this effect, in general, the increase in proþts generated by the plant that remains in foreign country is not large enough to offset the costs related to the investment. 4.2 Trade Agreement Regime We now turn to the regime where the governments are restricted not to use trade policy. As under the Þrst regime we can establish the following relationships: b m > 0 for all d>0, b n > 0 if d> d b and b n < 0 if d< d. b where b d The Þrst inequality implies that investment abroad always increases operating proþts when the rival Þrm is multinational. The following two inequalities state that for low values of the environmental damage parameter, d, investment abroad lowers proþts whereas for higher values it always increases proþts. As under the protectionist regime, the size of the export market affects the absolute but not the relative change in operating proþts when investing abroad. Figure 2 below shows this relative change under the trade agreement regime. 14

15 Relative change in profits Rival firm national Rival firm multinational Environmental damage parameter (d) Figure 2. Relative change in proþts when investing abroad under the trade agreement regime. If foreign Þrm invests in home country, emission taxation in home country is partly destined to tax foreign shareholders and partly to control emissions. When d is small, the Þrst effect is relatively more important and investment to home country induces a large increase in tax rate. Except for very small values of environmental damage, this policy change is proþtable for the investing Þrm as the domestic Þrm reacts by reducing supply. When the environmental damage caused by emissions is very small and both Þrms are national, the governments actually set negative emission tax rates. In that situation investing abroad would trigger policy changes that are not favorable to the investing Þrm. 7 Again, it may be the case that the Þxed cost of investment is high enough to deter all investment and each Þrm prefers to remain national regardless of what the rival Þrms does. However, Lemma 2 When the governments can use only environmental policy and d> b d,investing abroad is a dominant strategy for both Þrms if the Þxedcostofinvestmentis low enough. 7 We do not require the equilibrium emission tax rates to be non-negative. An alternative assumption would be that zero is the lower bound for emission tax rates, as under the trade agreement regime negative emission tax rates could well be prohibited. This assumption would make remaining national less proþtable under the trade agreement regime. 15

16 Proof. Straightforward, since d> d b implies that b n > 0 for all A>0. Since b m > 0 for all d>0 and A>0, there exist values of G>0 such that b n >Gand b m >G. When both Þrms are national, the governments use environmental policy not only to correct the negative externality caused by production but also to achieve trade policy goals. By Characterization 2, emission tax rates in both countries are lower than the marginal damage caused by emissions. Assume that in this situation foreign Þrm invests in home country. As a result, the national governments of the two countries will choose different emission tax policies. In particular, home government will set a tax rate that exceeds the marginal damage of emissions whereas the tax rate set by foreign government will be lower than the marginal damage of emissions. There are two reasons for this. First, in choosing its policy, home government ignores proþts accruing to foreign shareholders. Second, from the point of view of home government, relaxing environmental policy is no longer useful in shifting proþts to the domestic Þrm as adopting a more lax environmental policy induces foreign Þrm to expand supply. For foreign Þrm the motive to invest abroad is purely strategic: the proþts generated by the plant that is moved to home country decrease. However, by investing in home country foreign Þrm is able to inßuence the policy choices of home government and thus supply decisions of home Þrm in a way that leads to increased proþts generated by the plant that remains in foreign country. Figure 2 shows that except for very low values of the environmental damage parameter this positive effect dominates. We are now able to compare the incentives of the Þrms to invest abroad under the two regimes. The previous discussion is summarized in the following proposition: Proposition 2 Emergence of multinational Þrms is more likely when the governments are constrained not to use direct trade policy measures. Proof. By using b n, b m,and n derived in Appendix C we can show that Therefore, b n > n and b m > n for all A>0 and d>0. n >G= b n >Gand b m >G. ThegaininoperatingproÞts when investing abroad is always higher under the trade agreement regime than under the protectionist regime. Therefore, if gain in operating proþts when the rival Þrm is national exceeds the Þxed cost of investment under the protectionist regime, the Þrms remain national with probability β 1,

17 under the protectionist regime, but have a dominant strategy to invest abroad under the trade agreement regime. The Þrst implication of this result is that under a certain parameter range, signing an agreement that leads from the protectionist regime to the trade agreement regime, increases the probability that polluting Þrms invest abroad. The second implication is that this change in the international institutional setting may well trigger investment ßows that move the countries from a situation of entirely national production to a situation of full multinationalization of production. What is crucial for Proposition 2 is the different nature of the two instruments available for the national governments. The governments cannot differentiate the treatment of emissions generated by production plants according on their ownership. Therefore, environmental policy cannot be used to discriminate between the Þrms when they are multinational. In contrast, export subsidies can be efficiently targeted to domestic Þrms. This means that the relationship of each government and the domestic Þrm depends on the international regime. When the government cannot use trade policy to protect the domestic Þrm, it will use an instrument that attracts investment from foreign Þrm. 5 Conclusions We have studied the inßuence of international free trade agreements on the location strategies of polluting Þrms and shown that a change in the international regime that restricts the instrument set available to the national governments may lead to multinationalization of production. Under the trade agreement regime, Þrms invest abroad because they want to inßuence the policy decisions of the other government and output decisions of the rival Þrm. When the Þrms are national, governments use environmental policy not only to correct the negative externality caused by production but also to shift proþts to domestic shareholders. By opening a production plant abroad a Þrm is able to break this link between the domestic producer and the government. Namely, foreign owned plants not only eliminate the possibility to use lax environmental policy to trade related goals but also create an incentive to tax proþts accruing to foreign shareholders. This induces the government to tighten emission control which in turn leads to lower market share for domestic producers. From the point of view of the investor, the new plant abroad is less proþtable than the closed plant. Thus, investment abroad is purely strategic in nature. In contrast, when the governments can freely use both environmental and trade policy, investment abroad does not have the same strategic effect. When both Þrms are national, emission tax rates equal the marginal damage caused by emissions and Þrms receive export subsidies. The fact that the governments have two instruments at 17

18 their disposal and can use export subsidies to discriminate between the Þrms, makes investment abroad less attractive for the Þrms. If foreign Þrm invests in home country under the protectionist regime, home government raises the emission tax rate above marginal damage of emissions in order to capture part of the proþts accruing to foreign shareholders. However, at the same time it also raises export subsidy to the domestic Þrm so as to compensate domestic shareholders for the negative effects of stricter emission taxation. This makes investment abroad less proþtable. References [1] Barrett, Scott (1994), Strategic environmental policy and international trade, Journal of Public Economics 54, [2] Brainard, Lael (1993), An Empirical Assessment of the Factor Proportions Explanation of Multinationals Sales, NBER Working Paper No [3] Brander, James and Barbara Spencer (1985), Export subsidies and international market share rivalry, Journal of International Economics 18, [4] Hoel, Michael (1997), Environmental Policy with Endogenous Plant Location, Scandinavian Journal of Economics 99(2), [5] Horstman, Ignatius J. and James R. Markusen (1992), Endogenous market structure in international trade, Journal of International Economics 20, [6] Janeba, Eckhard (2000), Tax Competition When Governments Lack Commitment: Excess Capacity as a Countervailing Threat, American Economic Review 90(5), [7] Kennedy, Peter W. (1994), Equilibrium Pollution Taxes in Open Economies with Imperfect Competition, Journal of Environmental Economics and Management 27(1), [8] Markusen, James R. (1995), The Boundaries of Multinational Enterprises and the Theory of International Trade, Journal of Economic Perspectives 9(2), [9] Markusen, James R. and Anthony J. Venables (1998), Multinational Þrms and the new trade theory, Journal of International Economics 46,

19 [10] Markusen, James R., Edward R. Morey, and Nancy Olewiler (1995), Competition in regional environmental policies when the plant locations are endogenous, Journal of Public Economics 56, [11] Rauscher, Michael (1994), On Ecological Dumping, Oxford Economic Papers 46, [12] Rauscher, Michael (1995), Environmental Regulation and the Location of Polluting Industries, International Tax and Public Finance 2(2), [13] Ulph, Alistair (1996), Environmental Policy and International Trade when Governments and Producers Act Strategically, Journal of Environmental Economics and Management 30, p [14] Walz, Uwe and Dietmar Wellisch (1997), Is free trade in the interest of exporting countries when there is ecological dumping? Journal of Public Economics 66, [15] Zhao, Laixun (1995), Cross-hauling direct foreign investment and unionized oligopoly, European Economic Review 39(6),

20 A Appendix For the sake of notational clarity, we deþne z H,u (u; j) and z F,u (u; j) to be the price effect of a change in policy variable u {t, t,s,s } on the proþt ofhomeandforeign Þrm, respectively, under industry structure j. Then z H,u (u; j) = p 0 (Y (u; j)) yu F (u; j) yh (u; j),and z F,u (u; j) = p 0 (Y (u; j)) yu H (u; j) yf (u; j). Proof of Characterization 1. (7) become When the Þrms are national, equations (6) and z H,t s nn y H t +(t nn d 0 (E)) E t =0 (A1) and z H,s s nn y H s +(t nn d 0 (E)) E s =0, (A2) where all functions and partial derivatives are evaluated at (u nn ; nn). Asz H,t (u nn ; nn) = z H,s (u nn ; nn) and y H t (u nn ; nn) = y H s (u nn ; nn), we can deduce that t nn = d 0 (E) and s nn > 0. When both Þrms have one production plant in each country, (6) and (7) become and z H,t s mm y H t +(t mm d 0 (E)) E t + x F t mm =0 (A3) z H,s s mm y H s +(t mm d 0 (E)) E s =0, (A4) where all functions and derivatives are evaluated at (u mm ; mm). Solving (A4) for s mm and plugging into (A3) gives z H,t y H s (tmm d 0 (E)) E s y H t E t y H s z H,s y H t + x F t mm y H s =0. Solving for (t mm d 0 (E)) yields (t mm d 0 (E)) = zh,t y H s z H,s y H t + x F t mm y H s E s y H t E t y H s. The nominator is positive since z H,t (u;mm) > 0 and z H,s (u;mm) > 0 and y H s (u;mm) > 0 and y H t (u;mm) < 0. The denominator is also positive since yh t (u;mm) < yh s (u;mm) 20

21 and E s (u;mm) < E t (u;mm). Since (t mm d 0 (E (u;mm))) > 0 it directly follows from (A4) that s mm > 0. Finally, when foreign Þrm is multinational and home Þrm is national, (6) and (7) become and z H,t s nm y H t +(t nm d 0 (E)) E t + x F t nm =0 (A5) z H,s s nm y H s +(t nm d 0 (E)) E s =0, (A6) where all functions and derivatives are evaluated at (u nm ; nm). From proþt maximization it follows that y H t (u;nm) < 0, y H s (u;nm) > 0, y F t (u;nm) < 0, y F s (u;nm) < 0. Hence, z H,t (u;nm) > 0, z H,s (u;nm) > 0. Solving (A6) for s nm and plugging into (A5), it can be rewritten as where the RHS is positive since (t nm d 0 (E)) = zh,t ys H z H,s yt H + x F t nm ys H ³, ys H yt H E t yt H ³ E s y H s Et y H t = xh t qt H + x H t 0 < E s y H s x H t = xh s + x H s x H s < 0 as E s y H s Et y H t q H t + x F t q F t + x H t + x F s + x H s < 1. > 1 and Since (t nm d 0 (E)) it follows that s nm > 0. ProofofCharacterization2. When the Þrms are national, (8) becomes z H,t (bu;nn)+ bt nn d 0 (E (bu;nn)) E t (bu;nn) =0. (A7) Since z H,t (bu;nn) < 0 and E t (bu;nn) < 0 it directly follows that bt nn <d 0 (E (bu;nn)). When the Þrms are multinational, (8) becomes z H,t (bu;mm)+ bt mm d 0 (E (bu;mm)) E t (bu;mm)+ x F (bu;mm) bt mm =0. (A8) As z H,t (bu;mm) > 0, x F (bu;mm) bt mm > 0 and E t (bu;mm) < 0 we have that bt mm > d 0 (E (bu;mm)). 21

22 Finally, when foreign Þrm is multinational and home Þrm is national, the Þrst-order conditions of the governments are and z H,t (bu;nm)+ bt nm d 0 (E (bu;nm)) E t (bu;nm)+ x F (bu;nm) bt nm =0 z (bu;nm)+³ F,t t b nm d 0 (E (bu;nm)) Et (bu;nm) =0. (A10) (A9) Again E t (bu;nm) < 0 and E t (bu;nm) < 0. SincezH,t (bu;nm) > 0 and z F,t (bu;nm) < 0, we have that bt nm >d 0 (E (bu;nm)) and bt nm <d 0 (E (bu;nm)). B Equilibrium output, abatement, and policies under Assumption 3 Under Assumption 3, using Þrst-order conditions (2)-(5) and the corresponding Þrstorder conditions of foreign Þrm yields x H (u; nn) = x H (u; nn) = 1 7 A s 5 21 t t 2 21 s, (B1) x F (u; nn) = x F (u; nn) = 1 7 A s 5 21 t t 2 s, 21 (B2) E (u; nn) = 2 7 A s t 4 21 s t, (B3) E (u; nn) = 2 7 A s τ 4 21 s + 4 t, 21 (B4) and proþts for home Þrm are π H (u; nn) = 5 x H (u; nn) 2 + x H (u; nn) 2 + t 2. 2 (B5) 22

23 In the same manner when j = mm, x H (u; mm) = 1 7 A s 4 7 t 2 21 s t, (B6) x F (u; mm) = 1 7 A s 4 7 t 2 21 s t, (B7) x H (u; mm) = 1 7 A s 4 7 t 2 21 s + 3 t, 7 (B8) x F (u; mm) = 1 7 A s 4 7 t 2 21 s + 3 t, 7 (B9) E (u; mm) = 2 7 A 22 7 t t s s, (B10) E (u; mm) = 2 7 A t 22 7 t s s, (B11) and proþts for home Þrm are µ 3 π H (u; mm) = 2 xh (u; mm)+x H (u; mm) x H (u; mm)+ 1 2 t2 + (B12) µ 3 2 x H (u; mm)+x H (u; mm) x H (u; mm)+ 1 2 t 2. Finally, when j = nm, x H (u; nm) = x H (u; nm) = 1 7 A 4 21 t t 2 21 s + 5 s, (B13) 21 x F (u; nm) = 1 7 A t t s 2 s, (B14) 21 x F (u; nm) = 1 7 A t t s 2 s, (B15) 21 E (u; nm) = 3 7 A t t s + 8 s, 21 (B16) E (u; nm) = 1 7 A t t s 2 s, 21 (B17) and proþts for home and foreign Þrm are π H (u; nm) = 5 x H (u; nm) 2 + x H (u; nm) 2 + t 2. (B18) 2 and π F (u; nm) = µ 3 2 xf (u; nm)+x F (u; nm) x F (u; nm)+ 1 2 t2 + µ 3 2 x F (u; nm)+x F (u; nm) x F (u; nm)+ 1 2 t 2. (B19) 23

24 B.1 Protectionist Regime By plugging (B1)-(B4) into equations (A1), (A2) and the corresponding Þrst-order conditions for foreign government, we have that t nn = t nn A = 10d 72d +31 and s nn = s nn A =4(2d + 1) 72d +31. (B20) (B21) Similarly, plugging (B6)-(B11) into equations (A3) and (A4) yields t mm = t mm = 1 31d +7 A 6 37d +22 s mm = s mm = A52d 12 37d +22. (B22) (B23) Finally, plugging (B13)-(B17)) into equations (A5)-(A6) gives t nm Ad (282d + 79) = 2040d d +445, t nm = 1 A (1144d d + 77) d d + 445, s nm = A (256d2 +274d +73) 2040d d + 445,and s nm = 1 A (448d d + 111) d d (B24) (B25) (B26) (B27) B.2 Trade Agreement Regime Given that s =0and s =0by plugging (B1)-(B4) into equation (A7) and the corresponding Þrst-order condition for foreign government, we have that bt nn = bt nn =2A 13d 1 208d +89. (B28) Similarly, plugging (B6)-(B11) into (A8) gives bt mm = bt mm = 1 44d +9 A 4 88d +53. (B29) 24

25 Finally, plugging (B13)-(B17)) into (A9) and (A10) yields and bt nm = 1 2 A 4506d d d d 2 (B30) bt nm = A 2 (5576d d + 277) d d 2. (B31) C Appendix Under the protectionist regime, gain from investment when the rival Þrm is multinational is m = π H (u mm ; mm) π H (u nm ; nm), where π H (u mm ; mm) follows from (B12) using (B6)-(B9), (B22), and (B23). Similarly, π H (u nm ; nm) follows from (B18) using (B13)-(B15) and (B24)-(B27). Straightforward calculations (using for instance Mathematica) show that m < 0 for all d>0. Gain from investment when the rival is national is n = π H (u mn ; mn) π H (u nn ; nn) where π H (u mn ; mn) =π F (u nm ; nm) follows from (B19) using equations (B13)-(B15) and (B24)-(B27). In the same manner, π H (u nn ; nn) follows from (B5) using equations (B1), (B2), (B20), and (B21). Straightforward calculations show that n =0when d = d 0.30 and d = d Furthermore, n > 0 when d d, d and n < 0 when d/ d, d. Under the trade agreement regime, gain from investment when the rival Þrm is national is b n = π H (bu mn ; mn) π H (bu nn ; nn) where π H (bu nn ; nn) is follows from (B5) using (B1), (B2) and (B28) and π H (bu mn ; mn) = π F (bu nm ; nm) from (B19) using equations (B13)-(B15), and (B30) and (B31). We have that b n =0when d = b d 0.05 and that b n > 0 for all d> b d. Finally, gain from investment when the rival is multinational is b m = π H (bu mm ; mm) π H (bu nm ; nm) 25

26 where π H (bu mm ; mm) follows from (B12) using (B6)-(B9), and (B29), and π H (bu nm ; nm) from (B18) using (B13)-(B17), (B30), and (B31). It is again straightforward to show that b m > 0 for all d>0. Using the above expressions, we have that for all d>0 and A>0 b n n > 0 and b m n > 0. 26