Contributions to Economic Analysis & Policy

Transcription

1 Contributions to Economic Analysis & Policy Volume 3, Issue Article 1 The Pollution Haven Hypothesis Pollution Havens and the Regulation o Multinationals with Asymmetric Inormation Xiaodong Wu University o North Carolina at Chapel Hill, wux@ .unc.edu Copyright c 2004 by the authors. All rights reserved. No part o this publication may be reproduced, stored in a retrieval system, or transmitted, in any orm or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission o the publisher, bepress. Contributions to Economic Analysis & Policy is one o The B.E. Journals in Economic Analysis & Policy, produced by The Berkeley Electronic Press (bepress).

2 Pollution Havens and the Regulation o Multinationals with Asymmetric Inormation Xiaodong Wu Abstract This paper develops a common agency model to analyze the strategic interaction between governments in regulating polluting multinationals. We show that when a irm has private inormation about its production technology relating output to pollution that is diicult to monitor, the inormation rent extraction behavior o non-cooperative governments will work against the pollution haven hypothesis in a Nash equilibrium with or without pooling. The pollution haven result is more likely to be reversed in a separating equilibrium than in a pooling equilibrium as a irm s output is urther away rom the most eicient outcome. This result provides an explanation or why many empirical studies do not support the pollution haven hypothesis even ater controlling or private non-environmental cost dierentials. KEYWORDS: multinational corporations, environmental regulation, common agency I would like to thank Gene Grossman, Avinash Dixit, Claudio Mezzetti, Eric Bond, and the editor Don Fullerton or their valuable comments and suggestions. I would also like to thank Gary Biglaiser, Patrick Conway, Robert Shimer, the conerence participants at the Mid-West International Economics Meetings at University o Illinois and at the Econometric Society 8th World Congress at University o Washington, and the seminar participants at various universities and institutions. Financial support rom the Ford Foundation is grateully acknowledged. Address: Gardner Hall 300B, CB#3305, UNC- CH, Chapel Hill, NC Phone: (919) wux@ .unc.edu.

3 Wu: "Pollution Havens" and the Regulation o Multinationals 1 One version o the Pollution Haven Hypothesis (PHH) is that less developed countries care more about economic growth than environmental protection and hence set lower environmental standards than developed countries. 1 Under reer trade, the hypothesis predicts that multinational irms will relocate to less developed countries to take advantage o their lax environmental standards. Over time, these countries will develop a comparative advantage in pollution-intensive industries and become the havens or the world s polluting industries. In this paper, transboundary pollution can occur and decrease the welare o the home country i a dirty industry is allowed to move to a oreign country with lax environmental controls so as to reduce its tax obligations. Hence, the creation o a pollution haven could reduce world welare rom the home country s point o view. However, the movement o the dirty industry to the oreign country could also raise world welare because the residents o the oreign country are the ones more directly aected by the pollution but care more about jobs than about pollution (Oates and Schwab, 1988). Indeed, this paper inds that cooperation between governments increases welare even as the dirty industry moves toward the oreign country with more pollution. Moreover, because o the inormation asymmetries, the concern that the dirty industry can and will over-pollute the oreign country is not justiied. Grossman and Krueger (1993) and Jae et al. (1995) demonstrate that there is little empirical evidence or the existence o pollution havens due to actor endowment and economies o scale considerations. Thus, reer trade may not harm the environment. To investigate this issue, Antweiler et al. (2001) and Grossman and Krueger (1993) divide trade s impact on pollution into a scale eect (i.e. aggregate demand or the polluting good), a technique eect (i.e. the pollution intensity o the dirty industry), and a composition eect (i.e. the share o the dirty industry in total output). They estimate the magnitude o these three eects and conclude that reer trade can be neutral or even beneicial to the environment. This paper ocuses on the technique eect under imperect inormation and turns attention to the role o regulatory agencies (later reerred to as governments). A key actor in regulating pollution is asymmetric inormation that arises when a government observes neither a multinational irm s pollution level nor its production technology to iner pollution rom level o production. The paper demonstrates that the technique eect dominates, as in Antweiler et al., but due to inormational reasons rather than a lower cost o capital. Although there have been many improvements in measuring pollution emissions directly and accurately with reasonable costs, incomplete inormation remains an important obstacle in regulating pollution. 2 1 Empirical studies on the relationship between pollution and income include Grossman and Krueger (1995), Komen et al. (1997), Selden and Song (1994), and the World Bank (1992). 2 The World Bank (1992, p. 78) report also states that, Ideally, regulators would attempt to change the behavior o resource users by means o direct policies or instance, by taxing or regulating emissions. But these measures involve a heavy administrative burden because they target individual polluters or resource users. Produced by The Berkeley Electronic Press, 2004

4 2 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 Given that multinational irms have better inormation about the pollution intensity o their technology than do regulatory agencies, an eiciency-seeking government needs to design a regulatory regime that induces a irm to reveal its technology. Such a scheme will invariably generate inormational rents to irms, which governments seek to minimize. In an open economy, what regulatory regime is optimal depends on the regulatory policies o other countries as well because a polluting multinational irm s output and pollution in one country aect both countries welare. By developing a common agency model with ootloose multinationals as agents o separate governments, this paper demonstrates that the rent extraction behavior between governments can provide another rationale or why pollution havens may not exist and hence why reer trade may be good or the environment. The rent extraction incentive also provides an explanation to the ollowing empirical indings. Eskeland and Harrison (2003) and OECD (1995) ind that, even ater discounting actors such as transportation costs, actor costs, economies o scale, and country risks, there is still no signiicant production surge in the developing countries resulting rom multinational irms incentive to lee environmental costs at home. Javorcik and Wei (2004) extend this empirical pursuit by taking into account the domestic regulatory environment such as corruption. They ind only weak evidence that supports the PHH. In testing the PHH in terms o revealed comparative advantages, the indings are mixed. Low and Yeats (1992) ind that dirty industries account or a growing share o exports in some developing countries. Xu (1999) inds that export perormance o dirty goods or most countries remained unchanged between the 1960s and 1990s. Since the major concern o the PHH is the production reallocation rom a developed country to a developing country, the ollowing analysis will consider the case where all multinational irms originate in the home (developed) country that cares more about pollution. Also, the home government knows each irm s technology while the oreign government does not. This assumption is consistent with the inding in Caves (1996), the World Bank (1992) report, and UNCTAD (1993) that the developed countries in general have a lower administrative burden in regulating emissions directly than the developing countries. More generally, as discussed in Bond and Gresik (1997 and 2004), the case with only one uninormed government can capture the strategic interaction between governments just as well as the case where both governments are uninormed. Without knowing a irm s technology, the oreign government either has to let all irms produce the same output (a pooling output scheme) or to internalize the externality o pollution by assigning dierent output levels to irms with dierent reported technologies (a separating output scheme). Thus, the home government can use its superior inormation to inluence a irm s reported technology to the oreign government and hence to manipulate oreign regulatory policies so as to extract rents rom the oreign government. On the other hand, since only the home government knows the irm s technology, the oreign government cannot do the same to the home government. The analysis in this paper inds two main conclusions.

5 Wu: "Pollution Havens" and the Regulation o Multinationals 3 First, i the oreign government chooses a separating output scheme, the home government has an incentive to encourage the irm to cheat the oreign government by increasing its production at home to compensate or its loss o production abroad due to cheating. This not only shits inormation rent via tax collection rom the oreign government to the home government, but also reduces the home government s welare loss rom transboundary pollution (i there is any). Foreseeing this, the oreign government reduces every irm s production abroad to induce truth telling. Alternatively, i the oreign government chooses a pooling output scheme, then a irm s output no longer depends on its reported technology and hence the home government can no longer induce cheating by increasing its domestic production. Thereore, the oreign government preempts any rent extraction rom the home government. However, by doing this, the oreign government has to give up internalizing the externality o pollution. Second, since the home (developed) country cares more about pollution, the above inormation rent extraction behavior between governments works against the PHH in both a pooling and a separating equilibrium. Moreover, the deviation rom the cooperative most eicient outcome is bigger, and hence the pollution haven result is more likely to be reversed in a separating than in a pooling equilibrium. 3 At equilibrium, most irms ace a separating output scheme in the oreign country. The paper also shows that this rent extraction eect is reinorced by proit repatriation and by transboundary pollution. There are other cases where governments act strategically in regulating polluting multinationals, but under perect inormation. In Dockner and Long (1993), Folmer et al. (1993), and Hoel (1997), this strategic consideration is to reduce transboundary pollution. Copeland (1996) analyzes how the home government aected by transboundary pollution can extract rents rom oreign pollution quotas. Markusen et al. (1995) ocus on the competition between governments in levying environmental taxes on multinational irms that generate local pollution. As in this paper, a government s major trade-o in regulating polluting multinationals is between the disutility o more pollution and the competition or additional tax revenues. For Markusen et al., the tax revenues come rom monopoly proits. In this paper, tax revenues come rom inormation rents. Moreover, this paper ocuses on the asymmetric inormation problem in Spulber (1988) and Dasgupta et al. (1980), and extends their analysis on regulating local irms in a single principal-agent problem to regulating multinational irms in a common agency problem. The analysis demonstrates that there can be over-production in the home country and under-production in the oreign country compared to the most eicient outcome. This asymmetry is not ound in the previous studies and is due to the inormation rent extraction between the asymmetrically inormed governments. This study also allows us to examine how irms with dierent technologies (private inormation) may respond dierently to a policy change. The literature on common agency models and regulation o multinationals with asymmetric inormation has been burgeoning since the early 1990s. Bernheim and Whinston 3 Bond and Gresik (2004) also shows that, or irms with two discrete types, a pooling equilibrium can be more eicient than a separating equilibrium in a tax competition between governments over one inal output. Produced by The Berkeley Electronic Press, 2004

6 4 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 (1985 and 1986), Laont and Tirole (1991), Martimort (1992), and Stole (1991) were the irst to introduce the common agency model with two uninormed principals and one common agent with either private inormation or unobservable action. In their models, the two principals have the same inormation set and the agent s type has the same eect on principals. Many studies have extended this homogeneity assumption. For examples, Bond and Gresik (1997 and 2004) study the case where there is one inormed and one uninormed government rather than two uninormed governments, while Biglaiser and Mezzetti (1993) and Mezzetti (1997) extend the study o homogeneous principals to vertically and horizontally dierentiated principals, respectively. Other extensions include Biglaiser and Mezzetti (2000), Bond and Gresik (1996), Gal-Or (1991), Ivaldi and Martimort (1994), Martimort (1996), and Olsen and Osmundsen (2001 and 2003). To analyze the environmental regulation o multinationals by asymmetrically inormed governments, this paper develops a common agency model that integrates Stole (1991) and Bond and Gresik (1997 and 2004). In the literature on common agency models, most studies ocus on intrinsic common agencies only. 4 This paper shows that whether a irm has to produce in both or neither countries (an intrinsic common agency) or has the reedom to produce in only one country (a delegated common agency) aects only the tax payments at home and abroad, but leaves the equilibrium output levels unchanged. Thus, the rent extraction under asymmetric inormation works against the PHH regardless o the type o equilibrium. This analysis o the relationship between an intrinsic and a delegated common agency in a separating and a pooling equilibrium is applicable to other common agency problems as well. The remainder o the paper is organized as ollows. Section 1 describes the model. Section 2 shows that cooperation leads to the most eicient outcome, where the PHH always holds. Section 3 derives the non-cooperative Nash equilibrium with and without pooling and discusses the reversal o the pollution haven result under asymmetric inormation. Section 4 concludes and draws some empirical implications. 1. The Model The model includes two countries, home and oreign, and a competitive industry with a continuum o irms. The production o this competitive industry generates a negative externality either by producing polluting by-products or by using polluting inputs. The product o the polluting industry is homogeneous and is sold at world price. All irms are price takers. Furthermore, the total output in each country is only a small proportion o world production so that each country s government cannot inluence the world price by changing its domestic policy. Thus, we can single out distortions o asymmetric inormation in regulating polluting multinationals. All irms are identical except or their production technologies indexed by θ. A higher θ means more pollution per unit o production. It is assumed that a irm s technology is 4 Bernheim and Whinston (1986) provide a ormal deinition o an intrinsic and a delegated common agency.

7 Wu: "Pollution Havens" and the Regulation o Multinationals 5 determined by its technology know-how so that a irm is endowed with a given θ. 5 Also, each irm uses the same technology at home and abroad. 6 There is no strategic interaction between irms. As discussed in the introduction, all multinational irms originate in the home (developed) country, whose residents have a higher marginal disutility o pollution. The home government also has superior inormation than the oreign government. In particular, the home government knows the exact value o each irm s technology parameter θ while the oreign government does not, but knows only the distribution o θ or the whole industry. Hence, θ is a random variable or the oreign government. The industry s technology θ ranges rom θ to θ with θ > 0. Let F(θ) be the cumulative distribution unction, and (θ) be the density unction o θ. F(θ) and (θ) are continuous and dierentiable with bounded derivatives. Since regulation does not change an existing irm s θ, regulation does not change F(θ) and (θ). It is urther assumed that (θ) =0 orallθ so that the reciprocal o the hazard rate, 1 F(θ),isdeined or all θ. This inverse hazard rate is (θ) d dθ ( 1 F(θ) (θ) assumed to be nonincreasing in θ,i.e. ) 0orallθ. Similar to the quadratic unction used in Spulber (1988), a irm s cost o producing y units o total output in the home country is: (1) C(y,θ)= 1 2 c 1 y 2 + (c 2 θ + ξ)y + S, where c 1 is the change o marginal cost with respect to a change in output, c 2 is the change o marginal cost with respect to a change in technology, ξ is a constant component o marginal cost, and S > 0 is a small sunk cost o setting up a plant. 7 It is assumed that there is a social welare trade-o between pollution reduction and private cost reduction so that c 2 < 0. 5 This is true, or example, i the pollution reduction technology is tied to human capital (e.g. R&D) or physical assets, which are either too costly to change ater a irm sets up its plant or are not equally available to all irms. Alternatively, i pollution reduction is a matter o a irm s continuous eort or i a irm can choose its technology ater knowing a country s environmental regulation, then it is more o a moral hazard problem than an adverse selection problem. In general, these two problems need to be analyzed separately. Under some circumstances, as discussed in Grossman and Hart (1983), a moral hazard problem can be transormed to an adverse selection problem discussed in this paper. The inormation cost o implementing a given eort under moral hazard is then equivalent to that o enorcing truth telling under adverse selection. 6 Empirical studies suggest that irms usually use the same technology across countries (Bhagwati and Hudec, 1996; Caves, 1996; Ulph, 1998; UNCTAD, 1993). One reason is that it is more eicient or a irm to specialize in one technology than to operate under dierent technologies. The other reason is that, when a irm sets up a new plant in a developing country, the irm tends to make its production complying to the higher environmental standards in the developed countries just to avoid any sunk costs that may occur in the uture as regulators in the developing countries tighten their requirements. 7 I there are transportation costs, then the constant marginal cost should be ξ plus a per unit transportation cost i part or all o the output in one country is shipped to consumers in another country. For tractability, the ollowing analysis assumes zero transportation cost. However, this assumption should not change the main results in this paper. Produced by The Berkeley Electronic Press, 2004

8 6 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 Also, it is assumed that production in the home country exhibits decreasing returns to scale so that c 1 > 0. I the sunk cost abroad S is not signiicant, then irms (originated in the home countries) have an incentive to open a new plant in the oreign country with the lower environmental standard. 8 However, irms will not stop producing in the home country because, as derived later, the tax scheme in the oreign country is non-linear and is increasing in a irm s total output to keep pollution under control. Finally, or the private marginal cost to be non-negative or all y and θ, ξ c 2 θ>0. Let y 0 denote home production and y 0 denote oreign production. With technology θ, a multinational irm s total cost o producing in both countries is: (2) TC(y, y ) = C(y,θ)+ C(y,θ)+ cyy, where c 0 represents the additional operation cost o producing at two dierent locations rather than one, or example, coordination costs between plants to standardize their products. 9 However, c 1 > c so that it pays irms to become multinationals. Equation (2) ensures that home and oreign production are perect substitutes so that a irm s cost is minimized by choosing the same production in both countries i their environmental regulations were the same. This rules out the economies o scale eect already studied in Antweiler et al. (2001) and Grossman and Krueger (1993). Usually, governments use only pure strategies to regulate pollution. Suppose the home government assigns each irm a pair o output and transer, {y(θ), t(θ)}, according to a irm s true technology θ. Without knowing each irm s θ, the oreign government uses a direct mechanism to assign each irm a pair o output and transer, {y (ˆθ), t (ˆθ)}, accordingtoairm s reported technology ˆθ. 10 I the oreign government chooses a pair o pooling schemes, then y and t are constants so that y = t = It is assumed that the governments oer output and transer schemes to irms and irms only have the choice o stay or exit. A irm receives a zero proit i it chooses exit. Thus, when its individual rationality constraint is satisied so that it produces in both countries, a irm s objective unction is its proit plus any transers received rom the governments, i.e. (3) u(ˆθ,θ) = π(y(θ), y (ˆθ),θ)+ t(θ) + t (ˆθ), 8 A superscript o a variable is hereater used to denote its corresponding variable in the oreign country. 9 The ollowing discussion will ocus on the case where c > 0 and the special case where c = 0 will be discussed in the appendix. 10 Alternatively, as in Peters (2003), a direct mechanism can also speciy transer t as a unction o y or each given technology. By the taxation principle developed by Guesnerie (1981 and 1995) and Rochet (1986) or a single agent problem and later extended to a common agency problem by Martimort and Stole (2002), we can implement y (ˆθ) and t (ˆθ) derived in this paper by speciying t as a unction o y at each ˆθ. Thus, the two deinitions are compatible. Although direct mechanisms may not be the best mechanisms to model a common agency game as shown in Martimort and Stole (2002), this paper uses the direct mechanism approach in order to derive the equilibrium output levels across countries and hence compare non-cooperative output with cooperative output directly and avoid the complication in deriving explicitly the tax schemes to implement this equilibrium outcome, which is non-essential in this analysis. 11 Hereater, a prime denotes the irst order derivative with respect to θ andanysubscripttoavariable denotes the partial derivative o the variable with respect to the subscript.

9 Wu: "Pollution Havens" and the Regulation o Multinationals 7 where (4) π(y, y,θ)= p (y + y ) 1 2 c 1(y 2 + (y ) 2 ) (c 2 θ + ξ)(y + y ) cyy S S, and p is the world price, which is assumed to be greater than or equal to the private average total cost o production so that all irms produce in both countries in the absence o regulation. Otherwise, regulation becomes redundant. Under a direct mechanism, all irms with the same θ have the same output and receive the same transer. Since consumers always buy at the world price p, ollowing Copeland and Taylor (1995), it is assumed that pollution in each country only aects a representative consumer s utility through damage to health, but not his/her choice among goods. A irm s damages to the home and oreign consumers, respectively, are given by the ollowing equations: (5) (6) B(y(θ), y (θ), θ) = γ bθ(y(θ) + αy (θ)), B (y(θ), y (θ), θ) = γ bθ(y (θ) + αy(θ)), where γ isthehomeandγ is the oreign marginal disutility o pollution. bθ >0isthe rate o pollution emission per unit o output so that total emission increases not only as output rises but also as a more polluting technology is used. A more polluting irm causes more damage even with the same amount o output. Thus, damage depends on a irm s technology as well as its output. α [0, 1] is the rate at which pollution is transerred rom one country to another. By assumption, γ > γ > 0. Thus, the PHH holds i all irms produce less at home than abroad. Suppose each government s objective is to maximize social welare, i.e. a weighted average o consumer surplus, producer surplus, and tax revenue. In order to compare with the most eicient outcome, the ollowing discussion assigns the same weight to the part o proit that remains in the country (i.e. producer surplus), tax revenue, and the negative damage (i.e. consumer surplus as the rest o it stays the same). Shares (claims on ater-tax proits) o multinational irmsareassumedtobesoldonlytohomeandoreignresidents. 12 Hence, the objective unctions o the home and the oreign governments are (7) and (8) W = W = = θ θ θ θ θ θ [β(π(s) + t(s) + t (s)) t(s) B(y(s), y (s), s)] (s)ds [βπ(s) (1 β)t(s) + βt (s) B(y(s), y (s), s)] (s)ds, [(1 β)(π(s) + t(s) + t (s)) t (s) B (y(s), y (s), s)] (s)ds 12 This assumption can be relaxed i the home government has ull inormation. The proit accrued to a third country works as i a third party levied a proit tax beore home and oreign residents split the proit. However, i both governments do not know a irm s technology, then this proit tax to the third country causes the home and oreign governments to deviate rom the most eicient outcome even under cooperation as discussed in the next section (see ootnote 15). Produced by The Berkeley Electronic Press, 2004

10 8 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 = θ θ [(1 β)π(s) βt (s) + (1 β)t(s) B (y(s), y (s), s)] (s)ds, where β (0, 1) is the percentage o shares purchased by home residents and hence is the share o a irm s retained proit contributing to the home government s welare. 13 The timing o the model is as ollows. First, each irm is endowed with technology θ. Second, the home and oreign governments announce their joint output and transer schemes i they cooperate. Otherwise, in an intrinsic common agency problem, the home and oreign governments announce their individual output and transer schemes as unctions o a irm s true and reported technology, respectively. 14 In a delegated common agency problem, each government, in addition, announces another set o output and transer schemes i a irm chooses not to produce in the other country. Ater knowing the output and transer schemes, irms decide whether to produce in a country and what technology to report to the oreign government. Finally, production and transers occur. 2. Cooperative Governments I the home and oreign governments set their environmental regulations cooperatively, they not only share inormation, but also choose production jointly. From equations (7) and (8), the home and oreign governments choose output y and y or each irm with technology θ to maximize their joint welare, i.e. (9) max y(θ),y (θ) W + W = π(y, y,θ) B(y, y,θ) B (y, y,θ). By the small country assumption, price is determined in the international market. The irst order conditions are: (10) (11) p = c 1 y + c 2 θ + ξ + cy + bθ(γ + αγ ), p = c 1 y + c 2 θ + ξ + cy + bθ(γ + αγ). 13 Since this paper examines the competition between governments in extracting irms proits, each government s welare depends on the other s tax scheme as well as its own. This externality invalidates the revelation principal so that there may be other kind o equilibria as illustrated in Peters (2003) and also Martimort and Stole (2002), Epstein and Peters (1999) and Peters (2001). However, the rent extraction incentive studied in this paper should still play an important role in the other equilibria. Moreover, according to Theorem 2 in Peters (2003), the pure strategy equilibrium in direct mechanisms derived later in the paper is robust (stays an equilibrium) against the possibility that principals might oer ar more complicated communication schemes. Hence, the analysis in this paper is still relevant as it captures the properties o at least one kind o equilibrium no matter what kind o communication mechanisms are easible. 14 Since the home government s schemes will not reveal to the oreign government any additional inormation about a irm s technology and the home government knows all the inormation, which government announces its schemes irst does not change the inormation structure and hence is not crucial as long as the oreign government announces its schemes beore each irm reports its technology. In any case, the home government always takes into consideration the eect o its schemes on the oreign government s schemes and chooses the best as to be discussed in section 3. On the other hand, the oreign government always tries to maximize its utility expecting any collusion between its inormed opponents. As both the home government and irms have ull inormation, and will make right expectations o the oreign government s schemes in equilibrium, the home government s optimal choice is renegotiation proo.

11 Wu: "Pollution Havens" and the Regulation o Multinationals 9 The second order condition is satisied as c 1 > c by assumption. From the irst order conditions (10) and (11), the equilibrium output in each country achieves the most eicient outcome, where price p equals the private marginal cost o local production in that country plus its induced marginal damage o pollution to both home and oreign residents (the last term on the right hand side o each condition). 15 For a given technology θ, solving y asaunctionoy rom condition (10) gives the home government s output level that maximizes joint welare or a given oreign output. To serve as a benchmark to compare with the non-cooperative case, this optimal choice o y at a given y is represented by the home government s cooperative irst order condition, F c, as shown in Figure 1. Similarly, solving y asaunctionoy rom (11) gives the oreign output level that maximizes joint welare at a given home output. This optimal choice o y at a given y is represented by the oreign government s irst order condition, F c.sincec 1 > c, F c is steeper than F c. Hence, the intersection o F c and F c deines a stable cooperative equilibrium, E c. Figure 1: Cooperative Equilibrium (or irms with a given θ) y * F c 45 y * c E c F * c y c y 15 This result hold even i neither government observes a irm s technology as long as all proits are distributed among home and oreign residents. Without proit outlow, it is easy to see rom equation (9) that the joint welare is independent o transers. Thus, the governments can internalize the inormation cost and implement the most eicient outcome by letting each multinational irm be the residual claimant o all consumer surpluses and hence eliminating any incentive to cheat. This ails i part o a irm s proit goes to residents in a third country. Produced by The Berkeley Electronic Press, 2004

12 10 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 Subtracting condition (10) rom (11) gives (12) (c 1 c)(y y ) = bθ(1 α)(γ γ). Since γ > γ,0 α 1andc 1 > c, y y or all θ. Allirms produce less in the country with a higher marginal disutility o pollution, i.e. E c lies above the 45 line, unless pollution is perectly mobile (α = 1) so that y = y. Moreover, although transboundary pollution reduces the production gap between countries compared with the case without transboundary pollution (α = 0), transboundary pollution by itsel cannot reverse the pollution haven result. Proposition 1. Cooperative governments implement the most eicient outcome that internalizes national as well as international externalities via inormation sharing. The PHH holds under cooperative governments regardless o transboundary pollution: all irms produce more in a country with a lower marginal disutility o pollution than in a country with a higher marginal disutility. 3. Non-Cooperative Governments In the non-cooperative case, governments no longer maximize joint welare. Instead, each maximizes individual welare. As a result, each cares how proits are distributed between them, which depends on both a country s proit share β and its tax revenue relative to the other government s. With asymmetric inormation, the home government has no incentive to reveal a irm s true θ to the oreign government. The oreign government has to design a pair o output and transer schemes, {y (ˆθ),t (ˆθ)},basedonairm s reported technology ˆθ. Thus, the crucial points o competition between governments in this common agency problem are the home government s attempt to extract inormation rent rom the oreign government and the oreign government s eort to counteract such an attempt. In a separating equilibrium, the ully inormed home government has an incentive to encourage irms to cheat the oreign government so as to make the oreign government pay a higher inormation rent to irms, which eventually becomes the home government s tax revenue. As home and oreign production are substitutes in a irm s proit unction and total production exhibits decreasing returns to scale, the home government wants to increase domestic production above a irm s most eicient level so that each irm has a bigger incentive to report a alse θ to cut both its tax and production abroad. This orces the oreign government to pay a higher inormation rent in order to induce truth telling. Because this(irst order) revenue gain outweighs the additional (second order) damage rom more pollution, home welare rises at the expense o oreign welare. Meanwhile, to counteract the home government s rent extraction attempt, the oreign government urther reduces each irm s production in the oreign country to discourage cheating. By increasing its local output to extract inormation rents, the home government reduces production in the oreign country. Hence, i pollution is transboundary, asymmetric

13 Wu: "Pollution Havens" and the Regulation o Multinationals 11 inormation provides the home government an opportunity to use its superior inormation not only to internalize the externality o home production, but also to reduce transboundary pollution rom abroad. 16 Thus, the pollution haven result is more likely to be reversed i pollution crosses national boundaries. To demonstrate the above arguments, we solve the common agency problem. A irm s net proit as a unction o its true type (θ) and its reported type (ˆθ)is (13) U(ˆθ,θ), u(y(θ), y (ˆθ),t(θ), t (ˆθ),θ). In equilibrium, the oreign government s output and tax schemes, y (ˆθ) and t (ˆθ),must be implementable and easible so that U(ˆθ,θ) is maximized at ˆθ = θ. Thus, the local irst order necessary condition or implementability is (14) t (θ) = π y (y(θ), y (θ), θ)y (θ). Applying Theorem 12 in Chapter 2 o Hurewicz (1958), there exists a transer scheme t (θ), which satisies the above condition at all points where y and y are dierentiable. 17 From (A.3) in the appendix, the local second order necessary condition or implementability is (15) (c 2 + cy )y 0. In an intrinsic common agency problem, the home government (with ull inormation) extracts all retained proits rom irms and leaves them with zero utility. 18 At a given pair o output and tax schemes (y (ˆθ),t (ˆθ)) set by the oreign government satisying condition (14), the home government s maximization problem is: (16) (17) max y(θ),t(θ) βπ(y(θ), y (ˆθ),θ) (1 β)t(θ) + βt (ˆθ) B(y(θ), y (ˆθ),θ) s.t. π(y(θ), y (ˆθ),θ)+ t(θ) + t (ˆθ) = 0, where ˆθ is a unction o y. Hence, when choosing y, the home government also considers how y aects ˆθ and thus y as discussed later in deriving the irst order conditions. Substituting t(θ) = (π(y(θ), y (ˆθ),θ)+ t (ˆθ))rom the constraint into the objective unction gives the home government s simpliied maximization problem: (18) max y(θ) = π(y(θ), y (ˆθ),θ)+ t (ˆθ) B(y(θ), y (ˆθ),θ). 16 In this case, although the world price stays the same, irms still have an incentive to reallocate their production across countries to take advantage o the dierences in the tax schemes. Hence, as the governments change their environmental policies, they can aect the output and hence transboundary pollution across countries. This incentive is more apparent with the alternative direct mechanism speciication discussed in ootnote In a pooling equilibrium, y = t = 0 so that (14) still holds. Firms will report their true θ to the oreign government as they cannot gain rom cheating. Meanwhile, the oreign government chooses not to use this inormation to charge a more polluting irm a higher tax. 18 The intuition o this policy is similar to that o capital export neutrality. Since the home government has ull inormation so that taxation does not alter domestic production, extracting all retained proits minimizes the gain to oreign shareholders. Produced by The Berkeley Electronic Press, 2004

14 12 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 In a delegated common agency problem, a irm can choose to produce only in the oreign country. The above analysis stays the same except that t(θ) = (π(y(θ), y (ˆθ),θ)+ t (ˆθ))+ π(0, y0 (θ), θ) + t 0 (θ),wherey 0 (θ) and t 0 (θ) are the oreign government s output and tax schemes when y = Since y0 (θ) and t 0 (θ) are independent o y(θ) at equilibrium, this will only change the home government s transer scheme, but not output as the irst order condition o the maximization problem (18) stays the same. Footnote 20 shows a similar property or the oreign government s maximization problem. Hence, whether a irm has to produce in both or neither countries (an intrinsic common agency) or can choose to produce in only one country or both (a delegated common agency) will only aect the home and oreign transer schemes, but will have no eect on the output levels at home and abroad. The oreign government s maximization problem is (19) (20) (21) max y (θ),t (θ) s.t. θ θ ((1 β)π βt + (1 β)t B ) (θ)dθ (IC) u(y(θ), y (θ), t(θ), t (θ), θ) u(y(θ), y (ˆθ),t(θ), t (ˆθ),θ) (IR) u(y(θ), y (θ), t(θ), t (θ), θ) 0. Following Mirrlees (1971), let U(θ) be a irm s indirect utility unction. Thus, (IC) implies that U(θ) = maxˆθ u(ˆθ,θ) = π(y(θ), y (ˆθ),θ) + t(θ) + t (ˆθ), which yields equation (14). From the envelope theorem, du dθ = u θ = π θ + π y y + t so that U(θ) = θ θ u θ(y(s), y (s), s)ds + u(θ). Since it is never proitable or the oreign government to give any inormation rent to the lowest type (the highest cost) irm, u(θ) = We can then write (22) (23) U(θ) = θ θ [π θ (y(s), y (s), s) + π y (y(s), y (s), s)y (s) + t (s)]ds. Since t = U π t, the oreign government s objective unction becomes (1 β)π β(u π t)+(1 β)t B = π +t B β θ θ [π θ +π y y +t ]ds. 19 I y = 0, the problem degenerates to a single principal-agent problem with adverse selection. Following Mirrlees (1971) and Fudenberg and Tirole (1991), it is easy to prove that y 0 (θ) satisying p = c 1y + c 2 θ + ξ + γ bθ β 1 F(θ) (θ) c 2 and t0 (θ) satisying (14) with t 0 (θ) = π(0, y 0 (θ), θ) are the solution to the single principal-agent problem. Clearly, the monotonicity condition, y > 0, is satisied. 20 This holds in both an intrinsic and a delegated common agency problem. For an intrinsic common agency, the home and oreign tax schemes or a irm with θ are jointly determined by the condition, t(θ) + t (θ) = π(y(θ), y (θ), θ) so that u(θ) = 0. For a delegated common agency, u(θ) = π(0, y0 (θ), θ) + t0 (θ) + π(y 0(θ), 0,θ) + t 0 (θ), wherey 0 (θ) = p (c 2θ+ξ+γ bθ) and t 0 (θ) = π(y 0 (θ), 0,θ) i β<1. I β>0, then at y = 0, t0 (θ) = π(0, y 0 (θ), θ). Hence, or 0 <β<1, u(θ) = 0. In act, rom equations (14) and (17), U θ = 0sothatallirms get zero utility at equilibrium as the ully inormed home government taxes away all irms residual surpluses. c 1

15 Wu: "Pollution Havens" and the Regulation o Multinationals 13 Integration by parts gives = θ θ θ θ θ θ [π θ (y(s), y (s), s) + π y (y(s), y (s), s)y (s) + t (s)] (θ)dsdθ 1 F(θ) [π θ (y(θ), y (θ), θ) + π y (y(θ), y (θ), θ)y (θ) + t (θ)] (θ)dθ. (θ) Under the monotonicity condition, y 0, proved in the appendix, the oreign government s maximization problem at a given pair o home output and tax schemes, (y(θ), t(θ)), reduces to the ollowing: (24) max y (θ) = θ θ θ + [π(y, y,θ) β 1 F (π θ (y, y,θ)+ π y (y, y,θ)y )] (θ)dθ ( t β 1 F ) t B (y, y ) (θ)dθ. θ It is easy to see that the transer schemes are dierentiable and the appendix proves that and are strictly concave in y and y, respectively, and that both have an interior maximum. Given the oreign government s tax scheme derived rom (14), the irst order conditions o (18) and (24) by piecewise dierentiation are (25) (26) p = c 1 y + c 2 θ + ξ + γ bθ + cy + αγbθ y d ˆθ dy, p = c 1 y + c 2 θ + ξ + γ bθ + cy β 1 F(θ) (c 2 + cy ), (θ) where y d ˆθ dy is 0 i y = 0andis cy c 2 +cy i y This term, calculated in the appendix, relects a strategic interaction between governments. It gives how much the home government can aect the oreign government s output by changing a irm s production at home to encourage the irm to report a alse θ to the oreign government. As to be explained in the ollowing subsections, this strategic eort o the home government becomes ruitless (y d ˆθ dy = 0) i the oreign government chooses a pooling output scheme (y = 0) or i there is no transboundary pollution (α = 0) There is no loss o generality here in restricting analysis to piecewise dierentiation as the home government has ull inormation and all the terms in the oreign government s objective unction have second order derivatives. 22 Obviously, the last term also becomes zero i the production at home and abroad are unrelated in a irm s proit unction (c = 0). This case is discussed in the appendix. Clearly, i α = β = c = 0, then the non-cooperative outcome collapses to the most eicient cooperative outcome. In this extreme case, there is no incentive or the home government not to convey its inormation to the oreign government. Produced by The Berkeley Electronic Press, 2004

16 14 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 Since all unctions are continuous and have bounded derivatives to the third order, the Lipschitz condition is satisied uniormly. From Theorem 11 and 12 in Chapter 2 o Hurewicz (1958), there exists a unique solution satisying the above conditions rom any initial point (y 0, y0,θ). Thus, although dierent initial conditions lead to dierent equilibrium output schemes, (25) and (26) characterize the common property o all output schemes that maximize the non-cooperative individual welare o home and oreign governments. 23 These schemes are locally and globally implementable i and only i the monotonicity condition (y 0) and the local second order necessary condition or implementation ((c 2 + cy )y 0) are satisied. The appendix proves that the monotonicity condition holds i c 2 γ b, which ensures that the production cost advantage o a more polluting irm is big enough or both home and oreign governments to let this irm produce more than a cleaner irm Separating Equilibrium From the monotonicity and local implementability conditions, the appendix proves that the home and oreign outputs are strategic substitutes. I the oreign government chooses a separating output scheme so that y 0, then y d ˆθ dy 0andc 2 + cy < 0. Since the last term in condition (25) is negative unless y = 0orα = 0orc = 0, as discussed at the beginning o this section, the home government s rent extraction incentive raises the home non-cooperative output above the output that maximizes joint welare at any given oreign output. Hence, the home government s non-cooperative irst order condition, F nc, shits to the right o F c as shown in Figure 2. I there is no transboundary pollution (α = 0), then the last term in (25) becomes zero so that F nc stays at F c. Thisisbecausetheinormedhomegovernmentcollectsairm s 23 A common agency problem with asymmetrically inormed principals usually results in multiple equilibria. This is because the uninormed principal s incentive compatible constraint depends on the inormed principal s schemes (as in this case, U θ depends on y(θ) and t(θ)), which in turn, depend on how the uninormed principal responds. Bond and Gresik (2004) illustrate this possibility o multiple equilibria as the inormed principal chooses dierent transer schemes. This multiplicity usually disappears as we move to a common agency with symmetrically uninormed principals. As studied in Martimort and Stole (2003), less inormation imposes more restrictions on the set o reasonable conjectures and hence can lead to a unique robust equilibrium. In this paper, since the relation o output levels between countries is derived rom the irst order necessary conditions, which must hold in all equilibria, multiplicity does not matter too much in deriving the eect o asymmetric inormation on the pollution haven hypothesis. 24 Otherwise, the optimal output should be decreasing in θ. Any output scheme decreasing in θ will not be implementable when the marginal cost o production decreases in θ, i.e.c 2 < 0. This is because y 0is necessary and suicient or implementability under the single crossing condition π yθ = c 2 > 0(Chapter 7, Fudenberg and Tirole, 1991). However, i c 2 > 0, then there will be a second-best implementable output scheme which is decreasing in θ. In this case, a cleaner irm with a lower θ also has a lower marginal cost. There will be no social trade-o between cost reduction and pollution reduction. Hence, this case is less interesting.

17 Wu: "Pollution Havens" and the Regulation o Multinationals 15 residual proit in equilibrium. Without transboundary pollution, the home government s preerences over output and tax abroad coincide with a irm s private preerences. The oreign government s incentive compatibility constraint ensures that the last term in condition (26), β 1 F(θ) (θ) (c 2 + cy ), is positive or all irms with θ< θ. This demonstrates that, at each given home output, the oreign government chooses a lower output than the output that maximizes joint welare to induce truth telling. cy in the last term relects the impact o the home government s rent extraction attempt on the oreign government s equilibrium output scheme as discussed at the beginning o this section. Asymmetric inormation shits the oreign government s non-cooperative irst order condition, Fnc,tothe let o Fc. Transboundary pollution has no eect on F nc as the oreign government does not have the inormation advantage to aect the home government s output decision. Figure 2: Non-Cooperative Equilibrium (or irms with a given θ< θ) y * F nc 45 F c E c y * c y * nc E nc F * nc F * c y c y nc y In sum, asymmetric inormation in a common agency model has two eects. One is to decrease the output in a country with inerior inormation urther below its output in a single principal-agent problem with asymmetric inormation. The other is to increase the output in a country with superior inormation. This latter rent extraction eect is absent in a single principal-agent problem. In this paper, both o these two eects weaken the pollution haven eect and can even reverse it as shown in Figure 2 or irms with θ< θ For irms with θ = θ, both governments choose their cooperative best responses i α = 0. This is consistent with the standard result that there is no distortion at the top, i.e. or those irms with the most Produced by The Berkeley Electronic Press, 2004

18 16 Contributions to Economic Analysis & Policy Vol. 3 [2004], No. 2, Article 1 Moving rom cooperation to non-cooperation, the governments not only stop sharing inormation as discussed above, but also exclude the pollution damage imposed on consumers in another country rom their individual welare unctions. Compared with the cooperative case, transboundary pollution urther shits F c out by γ αbθ and Fc up by γαbθ. This reinorces the previous eect on the home government s irst order condition, but makes the overall eect on the oreign government s irst order condition ambiguous, depending on the relative size o β 1 F(θ) (θ) (c 2 + cy ) and γαbθ. We can address this ambiguity mathematically and analyze the eects o asymmetric inormation more rigorously by subtracting (25) rom (26). I y 0, (27) (c 1 c)(y y ) = bθ(γ γ) β 1 F (c 2 + cy ) αγbθcy. c 2 + cy From condition (27), a country s relative output depends on the country s relative marginal disutility o pollution (the irst term on the right hand side), the rent extraction eect (the second term), and the transboundary pollution eect resulting rom the home government s attempt to use its superior inormation to reduce transboundary pollution (the third term) by adjusting its transer scheme. Obviously, the third term becomes zero i pollution has local eects only. Since c 2 + cy < 0, the rent extraction eect works against the PHH as long as some proits accrue to home residents, i.e. β>0, so that the oreign government does lose part o its inormation rent payments to irms. This rent extraction eect increases with the proit share accruing to home residents as the second term is increasing in β. The transboundary pollution eect reinorces the rent extraction eect as y > 0and αγbθcy c 2 +cy < 0 in a separating equilibrium. This is consistent with the empirical result in Bui (1998) that the gain rom cooperation is large when there is severe transboundary pollution. Compared with the cooperative outcome, asymmetric inormation under non-cooperation always decreases oreign output and increases home output. This distorts the pollution haven result and can completely reverse it i β 1 F (c 2 + cy ) αγbθcy c 2 +cy > bθ(γ γ ) Pooling Equilibrium I the oreign government chooses a pooling output scheme, then the last term in (25) and that in (27) become zero. F nc stays at F c or all irms and the oreign government pays no inormation rent. The pollution haven result is reversed only i β 1 F (c 2 + cy )> polluting technology. 26 I the home and oreign governments are both uninormed o a irm s production technology, then the rent extraction incentive will shit each government s non-cooperative irst order condition to the right while the usual incentive compatibility constraint will shit each government s non-cooperative irst order condition to the let. Thus, the overall eect o asymmetric inormation on a country s output and hence the pollution haven hypothesis in a common agency problem depends on the relative eectiveness o the rent extraction behavior between governments as analyzed in Wu (2000).