Protection without Discrimination

Transcription

1 Protection without Discrimination Vincent REBEYROL and Julien VAUDAY August 14, 2013 Abstract In this paper, we show that Non Tari Measures NTMs), such as product standards or domestic regulations, may be designed so as to fully respect the non-discrimination principle of the WTO, and still act as a protectionist device. Our argument is based on a prot shifting eect between rms, within sectors. A NTM, by increasing production costs of all operating rms, forces the least ecient rms to exit. This however increases market shares of surviving rms that therefore may benet from such a measure. We show that this eect can generate protectionism at the aggregate level, by forcing relatively more foreign exporters to exit. To highlight the conicts of interests that arise between domestic rms, we draw on Grossman and Helpman's 1994) lobbying model. In this context, we show that trade liberalization can lead to a rise in NTMs. Finally, we also highlight that the protectionist nature of such NTMs may be asymmetric between countries. Hence, the adoption of international standards, as promoted by the WTO, may induce some international income redistribution, which would contradict the principle of reciprocity in trade negotiations. Keywords: Trade protection, Non tari measures, Non-discrimination, Prot shifting J.E.L: F12, F13, D72 Toulouse School of Economics. vincent.rebeyrol@tse-fr.eu CNRS, Université Paris XIII-CEPN UMR 7115) 1

2 1 Introduction Trade protection is usually and broadly dened as government action or inaction that discriminates in favor of home producers against foreign producers Anderson, 2011). In this paper, we argue that a protectionist policy may challenge this denition. Specically, we want to highlight that a policy, to be protectionist, does not need to discriminate between domestic and foreign producers. As a result, some domestic producers may well suer from a protectionist policy while some foreign producers may benet. 1 Over the last 60 years, substantial eorts have been made to liberalize trade. Through the succession of multilateral negotiation rounds at the GATT and later at the WTO, taris, the most explicit protectionist policy tool, have been reduced substantially. Despite this success, a clear suspicion remains that some national policies may still result in protectionism through the use of other, less explicit, policy instruments. This possibility led to important concerns in trade negotiations about the rise of Non Taris Measures NTM). To tackle this problem, WTO redactors explicitly introduced the Technical Barrier to Trade TBT) agreement and the Sanitary and Phyto-Sanitary SPS) agreement. 2 These agreements stipulate that any domestic regulation or technical standard should fully respect the non-discrimination principle, i.e. they should not favor domestic at the expense of foreign producers. 3 promote the adoption of international standards. Additionally, these agreements In this paper, we show that NTMs may be designed so as to fully respect the non-discrimination principle of the WTO, and still act as a protectionist device. Further, we show that the rise in 1 Other standard denitions of trade protection may be stated as policies which are intended to insulate domestic producers from competition with imported goods Deardor's glossary). Alternatively, policies that raise the cost of imported goods or otherwise restrict their entry into a market, and thus strengthen the competitive position of domestic goods UNCTAD). Regarding these alternative denitions that do not explicit a discriminative nature of trade protection, we challenge the view that all domestic goods/producers need to be shielded from import competition to generate protection. 2 According to the World Trade Report 2012), the number of SPS notications was around 200 in 1995 to reach a record level of just under 1100 in As for the TBT notications, the number was less than 200 in 1995 to reach a record level of 1500 in The non-discrimination principle has actually two components. The Most Favored Nation principle states that countries should not discriminate between foreign producers of dierent origins, while the National Treatment principle states that they should not discriminate between domestic and foreign producers GATT article I and III). In this paper, we will focus on the latter that, as noticed by Sykes 1995), includes the former. 2

3 NTMs can be explained by trade liberalization. Finally, we also highlight that the protectionist nature of such NTMs may be asymmetric between countries. Hence, the adoption of international standards may induce some international income redistribution. Our main argument is based on a prot shifting eect between rms, within sectors. We start from the assumption that a NTM a domestic regulation or technical standard) raises production costs for rms operating in the domestic market. The direct eect is to force the least ecient rms to leave the market. However, this exit induces another indirect eect: it increases market shares of surviving rms. It follows that this indirect eect may dominate for some rms the most ecient), which will therefore benet from the NTM implementation. Crucially, this mechanism is at work even if the cost increase is strictly the same for all rms, i.e. without discriminating between rms. In an open economy, the introduction of any costly regulation or standard forces the least ecient domestic and foreign rms to exit, and generates gains for the most ecient domestic and foreign rms. We show that, even when considering perfectly symmetric countries, this exit of the least ecient rms may be ercer for foreign rms. It follows that in aggregate, a NTM can generate a prot shifting from foreign towards domestic rms: the domestic industry taken as a whole may be protected by a non discriminatory measure. Note however that this is only true in aggregate: some domestic rms will suer from this policy while some of the most ecient foreign rms will gain. The basic reason for this result is that, while foreign and domestic rms pay in absolute terms the same additional cost to access the domestic market, the cost increase may be larger in relative terms for foreign rms i.e. for the exporters). This may be the case because domestic rms, in contrast to exporters, bear the cost of building their production facility in order to serve the domestic market, while exporters have already incurred this cost on their local market. Further, we show in the last part of the paper that beyond this relative cost increase eect, the relative distribution of rm eciency between domestic and foreign rms also matters crucially because it determines the number mass) of winners and losers among those two groups. As a consequence, if countries are asymmetric in this specic way, i.e. they dier in their rm 3

4 eciency distribution, a policy aiming at harmonizing standards internationally, as promoted by the TBT and SPS agreements, may have international redistributive eects. Clearly, this prot shifting eect between rms within sectors is not new and has been studied in the literature, addressing various other issues. Nonetheless, and to the best of our knowledge, we are the rst to consider it as a possible protectionist mechanism. Before describing our framework and main results in more detail, we rst compare our argument to the trade policy literature, in order to clarify where our main contribution stands. While trade opening is supposed to be welfare enhancing, the literature has put forward a variety of reasons justifying trade protection. Most probably on top of the list is the incentive for large countries to manipulate in their favor their terms of trade. Since the early contribution of Johnson ), this general equilibrium eect is recognized to put countries in a traditional prisoner's dilemma when choosing trade policy unilaterally, justifying the need for multilateral negotiations. In a very inuential paper, Bagwell and Staiger 1999) show that the two key principles of the GATT/WTO, non-discrimination and reciprocity 4, are well designed to neutralize the terms of trade eect and thus multilateral negotiations respecting these principles reach ecient outcomes. 5 However, Staiger and Sykes 2011) show that, under certain circumstances, large countries may have the incentive to set non-discriminatory domestic regulations at an ineciently high level in order to manipulate their terms of trade. Recently, Ossa 2011) has reinterpreted these two principles in light of another motive for protection, related to our argument. He develops a "new trade" model, where he shuts down the terms of trade eect, and shows that countries have an incentive to raise import taris in order to attract a larger share of the world manufacturing production. This relocation eect may be alternatively understood as a prot shifting eect if we assume away free entry. Protected domestic rms then capture prots of foreign rms Ossa, 2012). 6 The prot shifting we put forward in this paper is in a 4 Following the denition of Bagwell and Staiger 1999), the principle of reciprocity requires countries to exchange equivalent concessions in terms of import volumes. 5 Bagwell and Staiger 2009, 2012) show that we can extend this result to settings incorporating imperfect competition and rm delocation. 6 Mrázová 2011) makes a similar argument in an oligopolistic framework, showing that import taris may be motivated by both a terms of trade eect and a prot shifting eect. As Ossa 2011), she gives a new interpretation 4

5 way similar: in an imperfect competition setting, the motive for protection is to capture some of the rents accruing to foreign producers on the domestic market. It is however quite dierent and new, because, by considering another policy tool than taris, our prot shifting goes from inecient towards ecient rms and not from foreign to domestic rms. It is only indirectly, in aggregate, that it can generate protection. In this paper, we develop a simple new trade model of monopolistic competition with heterogeneous rms, where we assume away free entry and where a non manufacturing sector absorbs all trade imbalances, shutting down the terms of trade eect. In order to shed light on the way NTMs may be turned away from their ocial social objective, we assume they have no enhancing eect on social welfare. They only consist in an additional and mandatory cost for active rms. In the following, we will thus refer to this specic) policy instrument as an entry tax. Further, we will simply assume that it takes the form of an additional xed cost. 7 Assuming away any positive eect on welfare implies that the entry tax has only an indirect detrimental eect on social welfare, by forcing some rms to drop out of the market. As such, a welfare maximizing government would thus never have the incentive to implement it. Second, as underlined by Ossa 2012), studying trade protection in new trade models emphasizes a major role for producers' interests. In contrast to the previous literature on trade protection, our contribution stresses a case where these interests dier among domestic producers in the same industry. In order to highlight these conicts of interests within sectors, we endogenize the entry tax decision by introducing political economy motives for protection in our model. Specically, we draw on the Protection for Sale PFS) framework developed by Grossman and Helpman 1994), assuming that a government that has social and private concerns receives contribution schedules from the active lobbies in the economy. This allows us to underline the political motives for implementing an entry tax and how they dier from those that arise with the implementation of taris, as in the PFS framework. Given that this prot shifting eect is at work even in the absence of any competition from to the principles of non-discrimination and reciprocity, which still ensure an ecient outcome. 7 The entry tax could additionally aect the variable cost of rms without changing our qualitative results. This extension is available upon request. 5

6 abroad, we rst develop the model in a closed economy, as a benchmark case. This benchmark shows that in absence of foreign rms, the most ecient rms should be overrepresented by lobbies to get a positive entry tax in equilibrium. If all rms were represented, their conicting interests would just neutralize each other and the choice of the government would be to implement no entry tax. This implies a major dierence with the lobbying game over taris: the level of the policy instrument is not necessarily increasing in the amount of contributions. This also shows that our framework provides a rationale for the formation of lobbies based on rm size. 8 We next consider the situation where two perfectly symmetric countries trade with each other. We focus on the case where the entry tax distorts the relative xed cost of accessing the domestic market in favor of domestic rms. It follows that, in contrast to the closed economy case, an entry tax may be implemented even if all rms are represented, because it generates aggregate gains at the industry level. Further, we derive the equilibrium level of the entry tax and show that it is increasing in the level of trade openness, measured by the level of trade costs. The basic intuition for this result is that trade liberalization increases the size of the prot shifting for a given entry tax. This may explain why NTMs have grown steadily in the last decades. The optimal response of an industry facing a ercer competition from abroad may therefore be to ask to be taxed more. But this unilateral policy is clearly inecient, as the mirror country chooses the exact same policy, withdrawing the gains on the domestic market by losses on the foreign market. Countries are therefore trapped in a prisoner's dilemma and the non-discrimination principle cannot preclude this situation. In a two country world, the principle of reciprocity allows trade negotiations to reach an ecient outcome. This is also the case in our framework. In the last part of the paper, we stress that the principle of reciprocity may nevertheless be in contradiction with the implementation of international standards, as promoted by the TBT and SPS agreements of the WTO. To do so, we introduce an asymmetry in the rm eciency distribution between the two 8 There is large evidence for lobbies based on rm size, a feature that cannot be rationalized in standard political economy models of lobbying over taris such as the PFS framework. For instance, in France, the CGPME is an association of small and medium size rms whereas the AGREF represents some large rms. While large rms tend to lobby more individually, various lobbies exist representing small rms, for example the COSE or the NFIB in the USA, the FSB in the UK. 6

7 countries. An asymmetry in the eciency distribution, even if only at the margin, plays an important role because it aects the mass of inecient relative to ecient rms in each country, and thus the mass of winners and losers in each country. In this situation, a common entry tax has international redistributive eects. 9 To ensure no international income redistribution, as reciprocity would require, the entry tax should dier between countries, which is at odds with the implementation of international standards. Though we do not test empirically our predictions in this paper, we argue that there is some empirical evidence in favor of the prot shifting we put forward. Some papers have tested the heterogeneous impact of new regulations on rms. For instance, Bartel and Thomas 1987) have assessed the impact of the implementation of a new regulation in the late 70s in the US, OSHA, that aimed at increasing the safety on workplaces. They estimate the direct negative eect and the indirect positive eect of this regulation and nd evidence for both. They show that large rms have beneted from the introduction of this new regulation at the expense of small rms. More recently, Suzuki 2013) shows that an increase of one standard deviation in land use regulation in Texas increases the entry cost in the lodging industry by 10%, decreases the number of operating rms hotels) by 15% and increases the revenue per room by 6%. Again, surviving rms benet from the regulation because of the less competitive environment induced by the regulation. Further, Fontagné et al. 2013) have tested the impact of the SPS measures that raised concerns at the WTO on the intensive and extensive margin of French exports. They nd that, on top of discouraging exports, these SPS measures have a heterogeneous impact on rms. Precisely, they nd that big exporters suer less than small ones and might even benet from these measures because of the weaker competition they imply. Their empirical evidence is thus in line with our results. Our paper is related to dierent strands in the literature. The role of product standards as a possible protectionist instrument has been previously studied theoretically in the literature. Closely related to our paper, Fischer and Serra 2000) 9 In our framework, it can even have a protectionist eect in one country and a anti-protectionist eect in the other. 7

8 also examine the protectionist eects of non-discriminatory standards. Akiko and Verdier 2002) study the political economy determinants of the implementation of a product standard in a similar setup. In these papers, the government employs standards to transfer rents from the foreign to the domestic rm. They however assume that the standard implies an extra adaptation cost for the foreign rm, which in turn generates the protectionist eect. In contrast, we focus on the case where the domestic regulation or standard is fully equivalently costly for domestic and foreign rms. Hence, we respect the spirit of the non-discrimination principle as there is no additional hidden costs for foreign competitors. As mentioned above, the prot shifting eect we put forward has been studied in the literature, addressing other issues. For instance, the IO literature has extensively studied the possibility of raising rivals' costs strategies Salop and Scheman, 1983), that include our main mechanism. Rogerson 1984) provides an early contribution stressing the possibility of a raising rivals' cost strategy based on the increase of xed costs for all competitors. Bliss and Di Tella 1997) analyze the relationship between competition and corruption. Corruption, by increasing the xed cost active rms have to pay to corrupt ocials, may endogenously generate surplus by concentrating prots in a smaller number of active rms. Do and Levchenko 2009) develop a heterogeneous rm trade model where they interpret the xed cost of production as the inverse of institutional quality. They show that trade liberalization may deter institutional quality, since trade liberalization increases the relative size of the most productive rms compared to the least productive rms, and hence the political demand for high xed costs. Their political game, which is a reduced form of the median-voter model of Benabou 2000), assumes that larger rms have higher political power larger number of votes). We show that this is not necessarily the case in the more general setting we use: we show that large rms may contribute more, but these larger contributions do not necessarily lead to a greater inuence. Finally, a growing trend in the literature focuses on endogenous trade policy when rms are heterogeneous in productivity. Bombardini 2008) extends the PFS framework by assuming that each sector is composed of several rms of dierent sizes. She shows that it is ecient for a lobby to be formed by the largest rms in a sector. Therefore the size distribution of rms plays an 8

9 important role: sectors with a higher share of rms above a given size exhibit higher intensity of political activity. Firms however lobby for an import tari so there is no conicts of interests between rms within sectors. Chang and Willman 2007) introduce lobbying into a Melitz 2003) model and show that there is some conict of interests within sectors between domestic rms and exporters. Exporters lobby against a reciprocal import tari because it would reduce their prots made abroad, while domestic rms sell on the local market only and thus lobby for an import tari that protects them from foreign competition. Finally, we already developed our main argument in a previous working paper where we detail the political determinants of a protectionist entry tax in a closed economy Rebeyrol and Vauday, 2008). Abel-Koch 2010) extends our framework to study the political economy determinants of border versus behindthe-border protectionist measures, with the assumption that only the most ecient rms are represented. She however considers that behind-the-border measures are more costly for foreign rms and are thus discriminatory. The paper is organized as follows. In section 2, we describe the structure of the closed economy and how the implementation of an entry tax generates conicts of interest among domestic producers. We detail the political game and the properties of the equilibrium in section 3. In section 4, we present the open economy version of our model in the symmetric case. We extend it to the case of asymmetric countries in section 5 and discuss the consequences for trade negotiations based on reciprocity and non-discrimination. Section 6 concludes. 2 Closed Economy 2.1 Structure of the economy We rst develop our model in a closed economy, as a benchmark case. This economy is composed by two sectors: M and A. Labor l) is the only factor of production. The M sector is characterized by increasing returns to scale in the production of a continuum of varieties and is subject to monopolistic competition à la Dixit-Stiglitz. The A sector produces a homogeneous good under 9

10 perfect competition and constant returns to scale, and serves as a numeraire. This numeraire good is produced with one unit of labor per unit of output and the wage rate is normalized to one. Firms are owned by workers. Demand Following Grossman and Helpman 1994), the preferences of the representative consumer are depicted by a quasi-linear utility function U. We introduce a CES sub-utility function over the continuum of manufacturing varieties: U = A + µ ln C M C M = i Θ ) 1 c σ σ i di 1, with σ > 1 1) C M and A denote consumption for the M composite good and the numeraire good, respectively. σ is the constant elasticity of substitution between any two varieties. Θ is the set of all available varieties in this economy to be determined in equilibrium), and µ the preference parameter over manufactured goods. We restrict our analysis to the case where consumer income is large enough to ensure a positive consumption of the numeraire good. Utility maximization yields the following demand for any variety i: c i = µl p σ P 1 σ i 2) where P = i Θ p1 σ i di ) 1 1 σ is the perfect price index and p i is the price of variety i. Production In order to operate in the market, any rm has to pay two types of xed overhead production costs. First, rms have to pay F P, the cost of setting up the production facility. On top of it, rms also have to bear a xed cost F d associated with the domestic distribution of production, that is aected by the standards and the regulations in force in the country. To alleviate notations, we label F = F P + F d, the total amount of xed costs any rm has to pay to operate in the market. Any rm i also incurs a constant marginal production cost a i. The cost of producing q units of good i with marginal cost a i is thus: C i q) = a i q + F. Given the demand function 2), the optimal price charged by a rm i is a constant mark-up over its marginal cost. Hence, rm i will 10

11 charge a price p i = σ a σ 1 i. It follows that the pure prots of rm i are: π i = µ σ σ ) 1 σ σ 1 P σ 1 a 1 σ i F 3) Firm heterogeneity As informally explained in the introduction, the possibility of some conicts of interest between rms within a sector comes from the fact that some rms are better than others. We thus assume that rms dier in their productivity. Although the existence of conicts of interests do not depend on a specic distribution of productivity among rms, we assume here that rms' marginal costs are drawn from a Pareto distribution. This assumption is made because, as we show below, it ensures that the aggregate prot in this economy is independent of the number of operating rms. It thus allows us to focus on a specic case where the motive for lobbying by rms exclusively comes from a prot shifting eect. Specically, we assume that marginal costs a are comprised between 0 and a 0, and that these marginal costs are drawn from a Pareto distribution F a) with a shape parameter κ: ) a κ. F a) = a 0 We further dene xi a σ 1 i. The ratio of any two rms revenues is therefore rx 1 ) rx 2 ) = x 2 x 1, where x may be understood as an index of the inverse of rm size. We will thus consider x rather than a and assume that x is drawn from a Pareto distribution Gx), with a shape parameter ρ > ) ρ x Gx) =, with 0 < x x 0 4) x 0 We further normalize x 0 = 1. Finally, as we assume away free entry, we simply consider as in Chaney 2008) that there is a group of entrepreneurs proportional to country size. Hence, the total mass of entrants is proportional to l = It can be easily shown that ρ = κ κ σ 1) > 0. σ 1. Therefore, ρ > 1 when the standard regularity condition is satised: 11

12 2.2 Market equilibrium Without loss of generality, we normalize labor endowment such that l = 1. The pure prot of rm i is given by: π i = µ p1 σ σp 1 σ i F 5) The computation of the perfect price index ultimately depends on the eciency index of the least ecient rm able to enter the market, since all rms with an x below this threshold are active in this market and make positive prots. Let x E denote the index of the least ecient active rm. We get: xe 0 p x) 1 σ dgx)dx = ) σ 1 σ σ 1 λx ρ 1 E with λ ρ ρ 1 To get the simplest possible benchmark i.e. without any lobbying activity), we further normalize the value of F : F = µ σλ This last normalization pins down the eciency index of the least ecient active rm: x E = 1. In this very simple economy, rms rst observe their eciency index drawn from the distribution Gx) and then decide whether to produce or not. All rms with an eciency index x i x E are able to cover the total xed cost F with their operational prots and thus decide to produce and make positive pure prots. Conversely, all the rms with an eciency index x i > x E do not enter the market, as they would make negative prots in that case. 2.3 Intra-sectoral conicts of interests Any change in the level of the xed cost F aects the mass of operating rms in equilibrium. We suppose that any additional xed cost reects the implementation of a new regulation or technical standard rms have to comply with. Formally, we assume that the implementation of an entry tax raises the xed cost F d by an amount T βf with β > 0, such that any operating 12

13 rm has to pay a total xed cost of 1 + β) F. The crucial variable in this model and the only endogenous instrument for a policy maker is thus β. We interpret a larger β as a more stringent standard or regulation. Impact of setting an entry tax The implementation of a positive entry tax splits the mass of rms into three groups. The rst group is composed of rms that are forced to exit the market. The second is composed of rms that can bear the larger xed cost but make smaller prots than in the benchmark case. Finally, the last group is composed of rms that make larger prots. Suppose that an entry tax is set-up which raises the total entry xed cost to F +T = 1 + β) F. The prots of an active rm i thus becomes: π i = σλ σ σ 1 µ ) 1 σ x ρ 1 E p 1 σ i 1 + β) F 6) All rms characterized by a x i > x E exit the market and compose the rst group of rms. The new least ecient rm makes no pure prots. Its operational prots are just sucient to cover the entry xed cost 1 + β) F. This allows to dene x E : π E = µ σλx ρ 1 E x 1 E 1 + β) F = 0 x E = 1 + β) 1/ρ < 1 7) The rm that makes exactly the same prots with or without the entry tax denes the separation between the two remaining groups. We call the eciency index of this indierent rm x C. Formally, x C is given by: π xc = πx β>0 C πx β=0 C = 0 ) x C = β β) ρ 1 ρ 1 The conicts of interests among producers induced by the entry tax generates however only two 13

14 groups of rms. One is composed of rms that are hurt by the new regulation and those that benet from it. The conicts of interests arise as soon as the positive eect on per rm demand) overcomes the negative eect on xed cost) for some rms: π i = µ ρ β) 1 ρ σλ ρ }{{} positive eect x 1 }{{} negative eect i F 8) Finally, note that the quasi-linear utility function implies that the total amount of expenditure spent over manufactured goods is constant and does not depend on consumer income. As a result, market shares lost by small rms are exactly equal to market shares gained by large rms. Besides, the Pareto distribution induces the convenient feature that the sum of xed costs saved due to the non entry of some small rms is exactly equal to the extra xed entry costs paid by the remaining active rms. Consequently, the sum of pure prots in this economy is constant, no matter the level of the tax entry implemented: xe 0 πx)dgx)dx = µ σλ 1 x ρ 1 E xe 0 x 1 dgx)dx xe β) F dgx)dx = µ σρ 9) We get: Lemma 1 i Θβ) π i di = 0 β. This ensures that the aggregate pure prot variation of any group of rms due to any entry tax is equal to the opposite of the aggregate pure prot variation of all other rms. In this closed economy case, the industry taken as a whole is therefore indierent with respect to any entry tax implementation. This guarantees that the prot shifting eect between rms is the unique reason for private producers to lobby for the implementation of an entry tax, as the aggregate pure prot is constant. 14

15 3 The political game In this section we introduce the political economy game, that determines the level of the entry tax in equilibrium with solid micro-foundations. 11 Specically, we follow Grossman and Helpman 1994) henceforth GH) and consider a common agency game under complete information, with transferable utility, where the decision of the agent here the government) aects its well-being as well as the well-being of the L set of principals here the lobbies), when each of whom oers a menu of payments contingent on the action chosen by the government. By using the same setup as in GH, we can show how the properties of the equilibrium of this political game dier when the endogenous policy instrument is an entry tax rather than a tari. We rst characterize each player, namely the government and the lobbies. We then present the equilibrium. 3.1 Government The government maximizes an objective function, namely G, composed by the aggregate social welfare and the contributions eectively paid by the exogenous L organized lobbies. The coecient η is a measure of the relative weight of social welfare compared to private revenue. The objective function of the government is: Gβ) = j L C j β) + ηw β) 10) Where j L C jβ) represents the sum of contributions paid by the L active lobbies. The aggregate social welfare, W β), gross-of-contributions, is dened as the sum of aggregate income and consumer surplus S: W β) = xe 0 ) π i β) µ ln µ µ + µ ln λ ) ) σ 1 σ 1 ρ σ 1 σ β) ρ 11 Alternatively, we could have introduced a positive eect of NTMs on social welfare in order to endogenize the level of the entry tax. However, this would have somewhat hidden the heterogeneous impact of the entry tax on rms, as the social planner would only consider the industry as a whole. Besides, we would have had to make an assumption, necessarily ad hoc, about the functional form of the relationship between the cost of the NTM and its welfare enhancing eect. 11) 15

16 In contrast to taris, we assume here that the entry tax does not generate any revenue. This assumption has no impact on our qualitative results. 12 According to Lemma 1, the aggregate pure prots, x E 0 π i β), is constant in this framework and independent of β. The aggregate income thus remains unaected by the implementation of an entry tax. An entry tax has only a detrimental eect on social welfare. It forces the least ecient rms to exit, reducing the number of varieties available for consumers and thus has a negative impact on consumer utility. From a pure social perspective, there is thus no reason to implement a positive entry tax. The government has a direct interest for social welfare but is also concerned by its private revenue the lobbies' contributions). The government evaluates all lobbies' proposals included in their contribution schedules and chooses the entry tax level that maximizes its objective function Gβ). Assuming that political contributions are dierentiable around the equilibrium, this maximization implies that in equilibrium: j L where β denotes the equilibrium value of the entry tax. C j β ) + η W β ) = 0 12) 3.2 Lobbies We assume that there are L exogenous organized lobbies. Any lobby j j L) maximizes its objective function G j, which is simply the sum of the joint welfare of the lobby members, W j β), net of the contributions paid to the government, C j β): G j = W j β) C j β) 12 The possible entry tax revenues would raise consumers' income and thus reduces the negative eect of the entry tax on consumers' utility and social welfare. This would however never overturn the negative impact of the entry tax on social welfare see Rebeyrol and Vauday 2008) for details). Following the ndings of Facchini et al. 2006), we could also assume that the implementation of a regulation implies some sunk costs for rms that cannot be captured by the government, which therefore only receives a share of all entry taxes paid by rms. This extension is available upon request. 16

17 where W j β) is dened in the same way as the aggregate social welfare: ) W j β) = π i β) + α j i j + α j µ ln µ µ + µ σ 1 ln λ σ σ 1 ) ) 1 σ 1 ρ 1 + β) ρ The objective function of any lobby j is composed of a producer interest which depends on the composition of its ownership and a consumer interest that has the same shape for all active lobbies. α j represents the fraction of the total population represented by lobby j and denes the relative weight of the producer and consumer interests in the lobby's objective function. No assumption is made on the eciency or the number of rms represented by lobby j. As shown above, large rms have opposed interests to small rms over the implementation of an entry tax. However, the denition of large and small rms, depends in ne on the level of the entry tax, which ultimately depends on the government's decision and cannot be given ex ante. We therefore make no assumption on the shape of i j π iβ), which depends on the size of rms represented by lobby j. It is worth noting that this already contrasts with a tari, since the producer interest of the lobby is not necessarily strictly increasing in the level of the entry tax. 3.3 Closed Economy Equilibrium In this simple setup, the incentive to lobby for the implementation of a positive entry tax only comes from the possibility for some large rms to capture market shares of other smaller rms that would be forced to exit. Competition among lobbies therefore arises within a given sector, which again contrasts with the exogenous convergence of interests of rms within sectors with respect to the implementation of a tari. The chronology of the game is the following. First, rms draw their eciency index from the distribution Gx). The L exogenous lobbies are then formed. The lobbying activity is free and costless. Once formed, each lobby proposes a contingent monetary contribution to the government for each possible level of the entry tax. The government collects all contribution schedules, chooses β that maximizes Gβ) and receives all contributions associated with β. 17

18 Finally, rms with an index x i < x E β ) produce and serve the market. Bernheim and Whinston 1986) have characterized the equilibrium conditions of this game. As shown by GH, these conditions ensure that the equilibrium entry tax respects the following rst order condition: 13 j L W j β ) + η W β ) = 0 13) The equilibrium of the game may be interpreted as the government maximizing the aggregate social welfare with weighting individuals represented by a lobby by a parameter 1 + η) and the other unorganized individuals by a simple weight η. This equilibrium condition may be rewritten so as to isolate the total marginal gain and the total marginal loss of the lobbying activity in our specic set-up: j L i j π iβ) = }{{} marginal gain σ σ 1)) ) ) α j + η F j L } {{ } marginal loss Any marginal increase in j L α j or η reduces the equilibrium entry tax, everything else being equal, since they both increase the marginal loss induced for a given β. These eects are those expected as a higher η increases the concern of the government for social welfare. Similarly, a larger j L α j implies that the L lobbies represent a larger share of the total population. While these eects are similar to what is obtained by GH, 14) highlights an important dierence when considering an entry tax as the endogenous policy instrument: assuming that lobbies do not have any interest as consumers α j = 0 for any j L) does not suppress the rivalry among them. 14 Indeed, as long as the shape of their prots with respect to β dier, because the composition of their ownership dier, their preferred policy will dier as well. Firm heterogeneity is here the sole motivation for proposing positive contributions. however that larger contributions at the sectoral level do not necessarily lead to a larger entry tax. Lobbies with ownership biased towards large rms would contribute for a positive and potentially 13 See Rebeyrol and Vauday 2008) for a detailed derivation when the game applies to an entry tax. 14 In the framework developed by GH, the conicts of interests among lobbies come solely from their interest as consumers. Their interest as producers, while dierent, are never conicting. See example 3 in GH for details ) Note

19 large entry tax. But lobbies with ownership biased towards small rms would contribute so as to make the entry tax as small as possible. It follows that the level of the entry tax only comes from the relative strength of these two groups of lobbies. As an illustration, note that the denition of the equilibrium entry tax in 14) shows that if all operating rms were represented in equilibrium by a lobby, the equilibrium level of entry tax would be β = 0, due to the fact that j L i j π iβ) = i Θ π i β) di = 0, β. Therefore, large rms must be over-represented by lobbies for a positive entry tax to be possible. It follows: Proposition 1 β > 0 only if j L i j π iβ) > 0, i.e. only if active lobbies are biased towards large rms. The larger the bias of active lobbies towards large rms, the larger the entry tax. We refer to the bias of active lobbies towards large rms as the average productivity of the representative active rm in the lobby. From 8), we get: i j π i = i j [ x 1 i µ σλ ρ 1 ρ 1 + β) 1 ρ F ] Further label n j β) the number of rms represented in lobby j that are operating in the market when β is implemented. We obtain: j L i j π iβ) = F xe n j β) λ j L ) x 1 j 1, This expression is increasing in x 1 j i j x 1 i n j, the weighted) average productivity of operating β) rms in each lobby. It follows that, everything else equal, the higher the average productivity of 1 organized rms x j, what we refer to as the bias of active lobbies towards large rms, the larger the equilibrium entry tax. This result may be compared to Bombardini 2008). In a set-up similar to GH, she shows that a greater heterogeneity among rms implies larger contributions, because more large rms enter the lobbying activity, which results in a larger protection. Our result presents some similarities: 19

20 the larger the organized rms, the higher the entry tax. It is however worth noticing that in her model, there is a demand for implementing a tari even in the absence of any rm heterogeneity. In contrast, the incentive to lobby is uniquely based on the presence of heterogeneous rms in our framework. Further, there is no direct relationship between the level of contributions and the size of the entry tax, because some rms may want to limit the size of the entry tax. Finally, one point is worth noticing. According to Bombardini 2008), large rms are more likely to make positive contributions, which is an argument used by Do and Levchenko 2009) to justify their assumption that larger rms have higher political power. 15 The more general framework used here induces however a discrepancy between the level of lobbies' contributions and their political power, i.e. their weight in the determination of policies. Assume for instance that at the equilibrium β, there are two active lobbies p and u, such that Wpβ ) W uβ ) > 0 and < 0. Lobby p thus gathers some large productive rms while lobby u gathers smaller rms. Lobby p would benet from a marginal increase in β while lobby u would benet from a marginal decrease in β. Note that around β, there is a convergence of interest between lobby u and the government since W β ) < 0). It follows, according to 14), that lobby p should compensate the welfare losses of the government and lobby u to marginally increase β, while lobby u should compensate lobby p welfare loss minus the government welfare gain to marginally decrease β. This asymmetry implies that the eectiveness of lobbies contributions dier as long as, around the equilibrium, some have a closer convergence of interests with the government than others. It follows that large rms may contribute more, but this does not imply that they will have a more political power. 15 Do and Levchenko 2009), p12: Bombardini 2008) documents that larger rms are more involved in lobbying activity, and thus one would expect them to have a higher weight in the determination of policies. Rather than assuming a specic bargaining game, this paper [...] modies the basic median voter setup to allow for a connection between income and the eective number of votes. 20

21 4 Open economy: Symmetric Case We now turn to a simple open economy framework where two perfectly symmetric economies, identical in all respects to the closed economy described above, trade with each other. We rst describe the trade equilibrium for any possible β. We then highlight the properties of the equilibrium entry tax in this framework. 4.1 Additional Assumptions Following the recent literature introducing rm heterogeneity into trade models e.g. Melitz 2003), we suppose that trade is costly and entails two types of costs. In order to serve the foreign market, rms have rst to incur an additional xed cost F x. This reects the costs implied by the regulations and standards in the foreign country, as well as the costs associated with managing remotely the distribution of the rm's product. Second, for each unit of good to be delivered abroad, τ units have to be shipped with τ > 1): rms incur an iceberg variable trade cost. For expositional convenience, we further dene γ > 0 such that F x = γf and φ τ 1 σ, the so called freeness of trade. Note that we focus here on the case where the two trade costs are exogenous: in contrast to the entry tax, the policymaker cannot manipulate them. Finally, we also make the standard assumption in this literature: φf < F x. This ensures that only a fraction of active rms choose to export see Melitz 2003). 4.2 Market Equilibrium The economy is now described by two cutos. x E has the same denition as in the closed economy: the index of the least ecient domestic) active rm. We further label x X the index of the least ecient rm engaging in international trade. Note that, due to the symmetry between countries, these two cutos are the same in the two countries. We label π D the prots made on the domestic market and π X the prots made on the foreign market that are identical to the prots of foreign exporters on the domestic market). Firms with an eciency index x i < x E 21

22 enter the domestic market and rms with an eciency index x i < x X serve the foreign market as well. It follows that x E and x X are given by: π D x E ) = 0 σλ x ρ 1 E µ + φxρ 1 X )x 1 E = 1 + β)f 15) and: π X x X ) = 0 σλ x ρ 1 E With equations 15) and 16), we obtain : µ + φxρ 1 X )φx 1 X = γ + β)f 16) x X = 1 + β x E γ + β φ 17) Note that, under the assumptions above, x X < x E β: only the most productive domestic rms engage in international trade. Plugging equation 17) into 15) and 16) gives x E and x X : x ρ E = 1 ) 1+φ )φ ) ρ 1), x ρ X = φ ) φ+ )φ) ρ 1) 18) We can now assess the impact of an entry tax on the sharing of pure prots between domestic and foreign competitors. To do so, we compute the aggregate pure prots made by domestic rms on the domestic market: x E 0 π D x)dgx)dx and the aggregate pure prots made by foreign rms on the domestic market: x X 0 π X x)dgx)dx. We get: xe 0 π D x)dgx)dx = µ σρ 1 1+φ )φ ) ρ 1, xx 0 π X x)dgx)dx = µ σρ φ φ+ )φ) ρ 1 19) Note that Lemma 1 still holds in this open economy: x E 0 π D x)dgx)dx+ x X 0 π X x)dgx)dx = µ. It follows that the variation of aggregate pure prot made on the domestic market by domes- σρ tic rms due to any entry tax is equal to the opposite of the variation of aggregate pure prot made on the domestic market by foreign rms. However, in contrast to the closed economy, the sum of pure prots earned by domestic rms 22

23 on the domestic market now depends on β as long as γ 1, i.e. as long as F x F. Precisely: x E 0 π D β)dgx)dx = ) ρ 1 γ) F φ ) ) ρ ) 1 + β) φ φ The direction of the prot shifting between domestic and foreign rms depends only on the sign of 1 γ). It follows: Lemma 2 If γ < 1, The aggregate pure prot variation of domestic rms on the domestic market due to any entry tax is positive: a non-discriminatory entry tax has an aggregate protectionist eect. Though non discriminatory, the entry tax produces an aggregate prot shifting between domestic and foreign rms. The domestic industry as a whole is thus no more indierent with respect to the entry tax: there is not only a prot shifting among domestic rms but also between domestic and foreign rms such that the aggregate prot of domestic rms is also impacted. The intuition behind this result is the following. The entry tax T βf ) rms have to pay to serve the domestic market is the same for domestic and foreign rms, so it is not discriminatory. However, the entry tax distorts the relative xed cost domestic and foreign rms have to pay in order to operate in the domestic market, as long as γ 1. The percentage increase of the total xed cost paid is indeed not the same. It is equal to β for domestic rms and to β γ for foreign rms. Therefore, as the change in demand µp σ 1 ) is the same for all rms, the group that faces a larger percentage increase in xed cost will experience a ercer selection due to the entry tax. It follows that, if γ < 1, the entry tax implementation induces a tougher selection for foreign rms than for domestic rms x X decreases relatively more than x E ) and generates an aggregate prot shifting from foreign to domestic rms. In the following, we will restrict our analysis to the case where γ < 1, i.e. F x < F P + F d F. We have some reasons to focus on this situation. 23

24 First, this seems more natural when considering our framework. Indeed, if γ > 1, foreign rms would have to pay more xed and variable costs to access the domestic market than domestic rms, though those costs do not cover the costs associated with setting up the production facility that domestic rms have to pay. In that case, there would be an incentive to serve the foreign market through FDI. For instance, in our symmetric framework, F x F d > 0 may be interpreted as the extra cost of managing remotely the distribution from the foreign country. Following this interpretation, we need F P > F x F d, i.e. γ < 1, to get trade. Otherwise, all rms wanting to serve the foreign market would do so by paying again F P + F d in the foreign country, choosing FDI rather than exports, in order to save on variable trade costs, to serve the foreign market. Assuming γ < 1 is also in line with recent calibrations of trade models with heterogeneous rms. For example, Di Giovanni and Levchenko 2012) have calibrated a model much in line with our framework and nd that, over a sample of 50 countries, xed costs associated with entry in a foreign market are on average about 40% of the xed cost associated with domestic production, which would lead to γ = 0.4. Melitz and Redding 2013) also calibrate a heterogeneous rm trade model and choose γ = in order to match the average fraction of U.S. manufacturing rms that export. For completeness, we discuss in appendix D some implications of our model in the alternative case γ > Open Economy Equilibrium The political game is identical to the one described in the closed economy analysis. The introduction of international trade adds an important eect however. The social part of Gβ) see eq. 11)) is now aected by the possible prot shifting between domestic and foreign rms on the domestic market. The aggregate income is therefore no more constant and depends on β. Note 24

25 that the consumer surplus also changes due to the presence of foreign rms. W β) becomes: W β) = xe 0 π D x)dgx)dx + xx 0 π X x)dgx)dx ) +µ ln µ µ + µ ln λ ) σ 1 σ σ 1 σ β) 1 ρ + φ ρ γ + β) 1 ρ) 1 ρ While the prots of domestic rms on the domestic market x E 0 π D x)dgx)dx) depend on β, note that the prots of domestic rms on the foreign market x X 0 π X x)dgx)dx) are not impacted by the entry tax decision of the domestic government. Clearly, these prots from exporting are aected by the entry tax decision made by the foreign government. But in contrast to a tari decision that would aect the terms of trade, the decisions of the two governments are here independent. To ease exposition and without loss of generality, we only report results in the case where j L α j = 0: lobbies do not have any consumer interest. As underlined above, in contrast to GH, focusing on this case does not suppress conicts of interests between lobbies. 16 is still described by 13), which now writes: j L i j π iβ) = ηf Equilibrium ) ρ ) ) ρ γ) σ σ 1 ) ) ρ β) 1 ) ) + φ ρ ρ β) φ } {{ } social marginal loss 22) The RHS of the above expression is simply the social marginal loss implied by a marginal increase in the entry tax: W. It has two components: the rst term in the RHS shows that an entry tax reduces social welfare by forcing some rms to exit, which increases the price index and thus decreases consumer's surplus S: S > 0. The second term shows that this marginal loss can be alleviated as the entry tax produces a prot shifting from foreign to domestic rms: x E 0 π D β)dgx)dx > 0. However, this positive aggregate prot shifting can never compensate the negative eect on the price index: W < 0 proof in appendix A). The social marginal loss is 16 Detailed computations and proofs in the general case, i.e. j L α j 0) are available upon request. 25