Special Interest Groups and Trade Policy in the EU Marianna Belloc and Paolo Guerrieri∗ CIDEI Working paper No. 69 - September 2006

Abstract The aim of this work is to employ theoretical and empirical analysis on the role of special interest groups in the determination of the EU trade policy. We build a two-stage game model of trade policy formation in a multisector-multicountry framework. We obtain the level of protection as a function of industry characteristics, in addition to political and economic factors at member state and European levels. The model is then tested by 2SLS estimation using data for 15 countries and 41 sectors. The econometric output suggests empirical support to model’s predictions as it highlights an important role for both national and European groups in trade policy making. Keywords: lobbying; policy making; trade policy; European Union JEL classification: D71; D78; F13; F15

∗ Marianna

Belloc (corresponding author): Department of Economics and CIDEI, Sapienza University of Rome

(Italy). Email: [email protected]. Paolo Guerrieri: Department of Economics, and CIDEI, Sapienza University of Rome (Italy). We would like to thank Alessandro Antimiani, Carl-Johan Belfrage, Betina V. Dimaranan and Luca Salvatici for useful suggestions. This work is part of the Research Program of National Scientific Relevance (PRIN) on “The new multilateral trade negotiations within the World Trade Organisation (Doha Round): liberalisation prospects and the impact on the Italian economy”. Financial support by the Italian Ministry of Universities and Scientific Research is gratefully acknowledged. Usual disclaimers apply.

1

1

Introduction

The aim of this work is to employ theoretical and empirical analysis on the role of special interest groups in the determination of the EU trade policy. The starting point is given by Grossman and Helpman (1994) (GH hereafter). The fundamental idea of their model is to microfound the behavior of organized lobbies and politicians, in order to derive a clear-cut expression for the level of endogenous protection as a function of industry characteristics. Three main predictions are obtained: (i) Protection is higher in organized sectors than in unorganized ones; (ii) Protection is higher in industries with lower import elasticity; (iii) Protection is a decreasing (increasing) function of the import penetration ratio in the organized (unorganized) sectors. The validity of these predictions has been verified for single-country studies: Gawande and Bandyopadhyay (2000) and Goldberg and Maggi (1999) - United States-; Mitra et al. (2002) -Turkey-; and McCalman (2004) -Australia. In the more recent literature, the GH framework has offered the basis to a number of extensions and adaptations for the analysis of related topics: trade blocs formation, trade bargaining, liberalization reforms, and so on (e.g. Grossman and Helpman 1995a, 1995b; Karacaovali and Limão 2005). Yet, a common denominator for much of the literature on special interest groups (SIGs hereafter) and trade policy is that it is highly suited for the institutional features of the U.S. political system (Persson 1998). In this paper we tailor the GH model to fit the EU institutional environment and develop a framework meant to describe trade policy formation in the EU. In so doing we seek to unravel the role of SIGs at the different stages of the policy-making process and to distinguish lobbying at the national level from lobbying at the European level. We borrow from Grossman and Helpman (1994) the microfoundations to the behavior of lobbies and politicians in a multi-sector framework. Nonetheless our model is different from theirs in two major respects. First, policy-making is a two-stage game: it includes two macro-stages, respectively state-level and European-level; the former consisting of two substages, lobbying activity and government decision making. Second, we allow pressure groups to resort to other means of political influence besides financial contributions to politicians,

2

namely, selective information provision. Grounding on this theoretical framework we derive the level of protection as a function of industry characteristics, in addition to political and economic factors at the member state level and at the European level. Despite the increasing interest on European policy formation and interest groups in the literature (see for instance: Broscheid and Cohen 2003; Cohen 1997; Crombez 2002), surprisingly few works have been addressed to analyze such an issue with a special regard to trade policy. Belfrage (2004) generalizes the GH model to account for the possibility of downstream lobbying and optimum tariff concerns and tests it on a number of OECD countries and regions including the EU. Since however, the author’s main concern is not on the special EU institutional environment, the model is maintained single-stage without considering the two levels of policy formation in a custom union. As a consequence, Belfrage’s model fits very well all the countries in his sample but the EU, in which case the estimated coefficients are weakly significant or with the wrong sign. Furthermore, due to the lack of data on lobbying activity, Belfrage (2004) assumes that all sectors are organized. While it is not hard to believe that this hypothesis has some empirical support, in our paper we preferred to find out some criterion to distinguish organized from unorganized groups as well as European lobbying from national one. Balaoing and Francois (2005) focus on trade policy formation in the special EU institutional environment without entirely grounding on the GH model. As a consequence, while Balaoing and Francois’ paper has many points in common with our work, the framework adopted differs remarkably. The authors use a computational general equilibrium model to estimate the marginal impacts of a set of import policies in the EU. They obtain the implied weights of the various industries in the policy process. Accordingly, rather than classifying industries as either organized or unorganized as standard in the “protection for sale” literature, the authors aim to assess the relative importance of individual sectors against overall economic welfare. Moreover, linking lobbying ability to industry characteristics they are able to indirectly derive the relation between weights attached to different sectors and the respective amount of lobbying. Finally, specific national characteristics are analyzed to derive their effects on the amount of protection.

3

A further work relevant to our concern is offered by Tavares (2006). The author investigate on whether the European trade policy making reflects a “deep” process of integration or it is only the effect of “shallow” integration resulting from the bargaining between national governments (Cadot et al. 1999). To accomplish this goal, Tavares (2006) develops a twostage model: in the first stage, interest groups lobby national governments; in the second, voting in the Council of Ministers defines the common tariff. In the empirical part of the paper, the first stage is then used to predict the preferred national tariff, while in the second one a model for common trade policy determination is estimated to test different decision making rules in the Council of Ministers (unanimity, qualified majority, simple majority). Results are supportive the deep integration hypothesis. The main reason of departure of this paper from ours is that, while it offers an interesting analysis of the decision making process at the EU level, it lacks microfoundation of government and lobbies’ behavior. No guess, as a consequence, can be provided on the value of the structural parameters that drive agents’ decisions. Finally, Karacaovali and Limão (2005) model the political economy weights attached to each sector in the economy, in order to convey information about the ranking of industries in governments’ preferences in terms of structural parameters rather than policy outcomes. Under the hypothesis that all sectors are organized, they obtain that industrial sectors with higher share of employment and higher regional concentration receive higher tariff protection. Moreover, industries with lower wages are associated to slightly larger weights; however the latter effect is never significant. The remainder of our paper is as follows. Section 1 presents a brief analysis of the European institutional environment where trade policy formation takes place. Section 2 presents an exposition of the theoretical model and derives the equilibrium policy. Grounding on the theoretical basis so provided, in the same section we also derive the model to estimate. Section 3 describes the econometric strategy and the empirical results. Finally section 4 draws the conclusions.

4

2

The EU institutional environment

The major actors in the formation of EU trade policy are the Commission, the Council, and the European Parliament (articles 133, 300, 308 and 310 of the Treaty Establishing the European Community). The inter-institutional procedure for the common trade policy consists of two stages: the proposal and the approval. In the first stage the Commission drafts the proposal under authorization of the Council and in consultation with special committees appointed by the Council (art. 300). In the second stage, the Council of Ministers approves (rejects/amends) the draft generally by qualified majority. The Council has the duty to inform the European Parliament of any decision taken by the Commission, although consultations on trade agreements are not mandatory. The general aim of the common commercial policy is “to contribute, in the common interest, to the harmonious development of world trade, the progressive abolition of restrictions on international trade and the lowering of customs barriers, instruments and scope” (art. 133 of the Treaty). Therefore we assume that “in principle” the European institutions are called to represent the interests of the citizens of the EU as a whole. However, each member state’s representatives tend to pay a special attention to their own country’s interests. Furthermore, community policy makers (as well as national ones) are susceptible to pressure by lobby groups. SIGs can affect policy decisions by operating either at the national or at the international level. Generally speaking, the measures available to lobbies to sway EU government’s favor are similar to those used by national lobbying: political pressure, economic contributions, information provision, and so on. Yet the policy making environment that characterizes the EU is much more complex than the national one: it is a focal point of different and contrasting interests, and relies on a very open and decentralized decision process. In such an uncertain surrounding business groups have especially to gain from alliances and coalitions (Mazey and Richardson 1993). Before the adoption of the Single European Act in 1986, most of the EU lobbying was wielded by national groups through political and administrative channels. This was due to the fact that the decision making process was mainly in the hands of the Council of

5

Ministers. Later on, when the Single European Act became effective, the Commission has been endowed with the power to initiate EU policies and has started to play a crucial role in the formulation of policy proposals. This implied a particular need for SIGs coalitionbuilding at the EU level at the early stages of the lobbying process. Unlike the Council, the Commission is meant to be a supranational body. As a consequence, commissioners are supposed to be independent of member states in taking their decisions. Furthermore the Commission is essentially a technical bureaucracy that counts in large measure on private actors to gather the necessary information to draft legislation; while it does not need funds for re-election. It follows that the Commission is not very exposed to political contribution lobbying but is very dependent on information provision by interest groups. Lobbies have indeed become an important source for grass-roots information and have started to play a very active role during the legislative process. They bring issues to the policy makers’ attention, provide information, and often take part in the committees (Directorates-General) that assist the Commission in preparing its proposals (Broscheid and Cohen 2003). On the contrary, the Council is at most susceptible to direct lobbying from groups that operate at the national level and by the means of the traditional channels of political pressure. To represent this complicated picture and also in view of empirical tractability for the following application, we need to introduce some simplifying hypotheses in order to stress as much as possible the different roles of the various players and the distinct interests at stake. As we have said, the final decision maker is the Council, which is composed of each member state’s government representatives but works on the basis of Commission’s proposal. Then the final decision must reflect the following interests: general welfare at the European level (that is the sum of the general welfare of every single member state), SIGs at the European level (that mainly wield pressure through the Commission), and SIGs at the national level (that tend to lobby directly the Council). Hence the policy making process can be pictured in two stages: the choice of the preferred tariff level for each member state’s representative that is part of the Council, and the formation of a common tariff level that summarizes somehow the individual preferences. National governments define the preferred tariff level that maximizes national welfare.

6

However, in so doing, they are susceptible to political pressures from two directions. First, since the Council decides after the initiative taken by the Commission, each member state forms her preferences on the basis of the information included in the proposal (that reflects the action of European lobbies to sway the Commission’s favor). Second, the Council is subject to the pressure wielded directly by national lobbies. As a consequence, we adopt the following modelling strategy. The European government is considered as a unitary actor composed of the representatives of the various member countries. Each member state’s representative forms her preferred tariff level maximizing the weighted sum of general national welfare, contributions (financial, informative or other) by European lobbies and contributions (financial, informative or other) by national lobbies. This maximization leaves us with K preferred tariff rates (where K is the number of member states at the time considered). To close the model we need to choose a rule, which obtains a single common tariff from those K. Because voting in the Council is not publicly observable, we do not know exactly how decisions are taken (see Tavares 2006, for theoretical and empirical analysis of the possible collective decision-making). Since, however, disentangling this point is not the main goal of the present paper and in order to significantly simplify the theoretical model that follows, we resort to an indirect method: the equilibrium collective decision is the tariff level that minimizes the penalty in terms of welfare loss that is associated to the misalignment between the common trade policy defined by the European government and the rate of protection that is optimal for each single country. The subsequent section describes the theoretical model in more detail.

3 3.1

Theoretical model Economy

We consider a region (the EU), whose member states agree to form a single market, and the rest of the world. The region consists of K countries (members of the EU), but it negotiates with its trade partners as a single entity with a single common trade policy. Countries, taken one by one, are considered small with respect to the rest of the world, while the EU P is deemed to be a large economy. Each country is populated by Nk (with N = k∈K Nk ) 7

individuals identical in every relevant respect but factor endowments. There are M + 1 consumption goods, out of which good 0 is the numeraire. Good 0 is produced using only labor, while the non-numeraire goods are produced using labor and a sector-specific factor. Individuals own labor and possibly a specific factor (at most one). In each country k ∈ K, individuals seek to maximize their utility given by the quasi-linear function: M X uk = x0k + uik (xik )

(1)

i=1

where u is differentiable, increasing and strictly concave, x0k is consumption of the numeraire good and xik is consumption of good i with i = 1, 2...M . dik (pi ) = xik = 1/u0ik (xik ) is P individual demand of the non-numeraire good i, whereas x0k = Ek − i pi dik (pi ) is demand

of the numeraire good. Indirect utility is:

vk (p, E) = Ek + sk (p)

(2)

where E is expenditure, s the consumer surplus, and p the vector of domestic prices of non-numeraire goods (bold stands for vector). Consumer surplus in country k (∈ K) is defined as: M M X X uik [dik (pi )] − pi dik (pi ) sk (p) ≡ i=1

(3)

i=1

All technologies exhibit constant returns to scale and specific inputs are available in inelastic supply. While the numeraire good is produced only with labor, any good i 6= 0 is produced with labor and a sector-specific factor. Denoting by π ik (pi ) the reward for the use of the specific factor i, domestic output of sector i (∈ M ) in country k (∈ K) is given by: yik (pi ) = π 0ik (pi )

(4)

The only instrument available to the policy makers is trade taxes. Calling p∗i the exogenous world price, ti = (pi − p∗i ) > 0 (< 0) is trade tariff (subsidy). The common pool of tax revenues is used to provide an array of public goods whose benefits are enjoyed by each individual in the whole region. Net government transfer to each individual in country k (∈ K) is: rk (p) =

P

i∈N

(pi − p∗i )

∙ ¸ 1 dik (pi ) − yik (pi ) N

8

(5)

Thus, individual income consists of three elements: labor income (wage, equal to one), government transfer (public good consumption), and, possibly, rewards from the ownership of the specific factor. Total welfare in country k (∈ K) can hence be defined as: Wk (p) = lk +

P

π ik (pi ) + Nk [rk (p) + sk (p)]

(6)

i∈N

where l is total labor supply (and labor income, given that wage is equal to unity). Finally denoting by mik (pi ) = Nk dik (pi ) − yik (pi ) the net import of good i (∈ M ) by P country k (∈ K), by mi (pi ) = k∈K mik (pi ) the total net import of good i by the EU, and by m∗i (pi ) = N d∗i (pi ) − yi∗ (pi ) the net import of good i by the rest of the world, we have: mi (pi ) + m∗i (p∗i ) = 0

3.2

(7)

Lobbying

L interest groups manage to get organized at either state member level (S) or European level (E). While in practice, national and European lobbies adopt different measures to sway political favor, we only pay attention to the fact that an amount Ci of resources is spent by group i to wield influence on policy making. It is worth noting that while Ci in the previous literature is constrained to be the contribution function as the vector of funds for re-election offered by lobby groups to policy makers, here we adopt a broader definition including all the resources spent for either political contributions or information provision. In deciding how much of their budget to devote to lobby action, SIGs seek to maximize net joint welfare of group members that is given by gross welfare less expenditure for lobbying activity. Gross-of-contribution joint welfare for the organized sector i (∈ L) in country k (∈ K) is: Wik (p) ≡ lik + π ik (pi ) + αik N [r (p) + s (p)]

(8)

where αi is the fraction of voting population that owns factor i (for simplicity we assume symmetry among countries so that αi = αik ∀k ∈ K). For the sake of generality and of realism, we imagine that interest groups do not need to choose between national and supranational action. It follows that a certain group i, let us say producers of leather apparel, can indeed be organized in country A and not in country B, and at the same time take part in a transnational organization of European producers 9

of leather apparel. Therefore, denoting by E the number of European lobbies and by S the number of national lobbies, we have that L = S ∪ E. Net lobby welfare for lobby i (∈ L) and country k (∈ K) is: Vik = Wik − Cik

(9)

where Cik is total amount of resources spent in lobbying activity.

3.3

Trade policy formation

As we have explained in section 1, each member state’s representative in the European government aims at maximizing the weighted sum of aggregate general welfare and the contribution from all the lobbies (either national or European). We can then formulate the objective function as follows: Gk (p) =

P

Cik + b

i∈S

where Wk (p) = lk +

P

P

i∈E

Cik + aWk (p)

a, b ≥ 0

(10)

π ik (pi ) + N [rk (p) + sk (p)]

i∈M

is total welfare in country k (∈ K); a is the weight each country attaches to aggregate welfare (for simplicity we assume that a = ak ∀k); and b is the political weight that denotes how much more (or less) importance each national government places on European lobbying with respect to national lobbying. The level of protection that satisfies the maximization of Gk (p) defines the platform that member states’ representatives will regard as point of reference to defend while bargaining for the common European trade policy. We denote this policy platform for the generic sector i by e tik (with k = 1, 2, ..K).

We need now to synthesize these K platforms into a single common tariff. This is how-

ever not a simple task. The European institutional environment is indeed a patchwork of different, sometime contrasting, interests, represented by national government, national lobbies, European lobbies and social non-government actors. To represent such a complicated picture, we inevitably need to resort to some simplification to describe the collective policy making. We choose a solution that is consistent with the present theoretical framework and, at the same time, suitable for the empirical analysis carried out in the second part of the paper. We assume that the European government minimizes the penalty in terms 10

of welfare loss that is associated to the misalignment between the common trade policy and each national platform. As a consequence, the EU government maximizes its utility minimizing the weighted sum of squares of those penalties as perceived by each member state. The weights are given by the number of votes for each country in the Council of Ministers. We have: Qj (p) = −

¢2 P 1 ¡ EU θ k tj − e tjk k∈K 2

(11)

is the where we have chosen the quadratic formulation for mathematical tractability, tEU j common policy in sector j (∈ M ), e tjk is the optimal tariff platform in sector j and country

k (∈ K), and θk is the weight of country k within the EU political arena.

3.4

Equilibrium

We can now find the equilibrium proceeding by backward induction. The equilibrium concept adopted is subgame-perfect Nash equilibrium, that is: given the information available, the expectations on other agents’ behavior and the specified rules of the game, agents define and follow the best strategy such that they cannot gain from deviations at each stage. Stage two In the second stage, the European government minimizes (11), i.e. she maximizes −Qj (p). After setting: tEU = arg max j tEU j

¢2 P 1 ¡ EU θ k tj − e tjk k∈K 2

it is straightforward (see the mathematical appendix) to obtain: ³ ´2 P EU − e t θ t ∂ k jk ∂p∗j ¡ EU ¢ j P P k∈K 1 e = θk tj − tjk + θk EU 2 ∂tEU ∂tj k∈K k∈K j

(12)

(13)

which, for tEU that satisfies (12), equals zero, and we can write: j tEU j

P e ∂p∗j k∈K θ k tjk = P + EU . ∂tj k∈K θ k

(14)

where ∂p∗j /∂tEU is the terms of trade effect. j Stage one

To find the national platform e tjk is more elaborate. Following Grossman and Helpman

(1994), an equilibrium trade policy in country k must be an equilibrium for each lobby and 11

the member state k’s government. Given Cjk , a feasible lobby spending function ∀j ∈ L = E ∪ S, this occurs if the following two conditions are met: 1. e tk = arg max (Wik − Cik ) ∀i ∈ L h tk

2. e tk = arg max (Wjk − Cjk ) + b h tk

P

i∈E

Cik +

P

i∈S

Cik + aWk

∀j ∈ L.

As explained in the mathematical appendix, the tariff rate that simultaneously satisfies

1. and 2. can be expressed as: e tjk = −

S + bI E − α Ijk L yjk (pj ) j , αL + a m0jk (pj )

(15)

where j ∈ M and k ∈ K. In (15) symbols’ definitions are as follows: IjE is an indicator that equals one if j ∈ E (European lobby) and zero otherwise; IjS is an indicator that equals one if j ∈ S (state level lobby) and zero otherwise; αL is the fraction of total population of voters represented by an organized (either national or European) group. Finally, we remark that a is the weight attached to general welfare, b is the weight placed on European level pressure, and m0jk (pj ) stands for first derivative of mjk (pj ). To get the common tariff rate, we substitute e tjk for (15) in (14) and obtain:

∂p∗j ¡ S ¢ yjk (pj ) P 1 E + =− φ I + bIj − αL , αL + a k∈K k jk m0jk (pj ) ∂tEU j P where j ∈ M and φk = θk / k∈K θk . tEU j

(16)

´ ³ Finally, after dividing both sides of (16) by pj and defining τ j = tj /pj = pj − p∗j /pj ,

we can also write:

= τ EU j

∂p∗j 1 ¡ S ¢ zjk (pj ) P 1 + EU , φk Ijk + bIjE − αL αL + a k∈K ejk (pj ) ∂tj pj

(17)

where again j ∈ M , τ EU is a construct based on the ad-valorem rate of protection for the j EU , zjk = yjk /mjk is the inverse of the import penetration ratio, and ejk = −pj m0jk /mjk is the absolute elasticity of import demand. (17) is very similar to the equation for the protection structure in Grossman and Helpman (1994) and so are the derived predictions. Since (1 + b − αL ) is positive, the model predicts that, if sector j is organized, the level of protection is the higher, the larger the inverse of the import penetration and the smaller the elasticity of import demand (in absolute terms). Indeed:

12

• zjk (= yjk /mjk ) relates to the stakes from protection. The larger domestic output in a certain sector (numerator of zjk ), the more specific-factor owners have to gain from protection (higher price) in that sector. The larger domestic import of a certain good (denominator of zjk ), the less the economy as a whole has to lose from larger protection. • The elasticity of import demand (ejk ) is associated to the deadweight loss due to deviations from free trade. The higher the import elasticity, the larger the deadweight loss, the less the government is willing to grant protection at the total population’s expenses. • Finally, the terms of trade (∂p∗j /∂tEU × 1/pj ) reflects the large country effect of trade j policy. We expect the protection rate to be the higher, the stronger the large country effect.

4

Empirical analysis

4.1

Econometric model and strategy

In the remainder of this paper, we estimate a stochastic version of (17). To simplify the Sz Ijk IjE zjk P P P zjk jk S = E = notation we set: Ωj = φ , Ω φ , Ω φ and j j k∈K k k∈K k k∈K k ejk ejk ejk ∂p∗j 1 Tj = EU . Then we end up with the following equation regression: ∂tj pj τ EU = j

β × ΩSj | {z }

National lobbying

+

γ × ΩE j | {z }

European lobbying

+

δ × Ωj | {z }

Unorganized sectors

+

μ × Tj | {z }

+ εj

(18)

Large country effect

where j is any sector, β = 1/ (αL + a) , γ = b/ (αL + a) , δ = −αL / (αL + a), and εj is the residual term that has been included additively for simplicity. β, γ, δ, and μ are the parameters to be estimated. According to the theoretical model, we expect the following signs and relations to hold: β > 0, γ > 0, δ < 0, μ > 0 and β + γ + δ > 0. Equation (18) cannot be estimated by ordinary least squares. Indeed the variables on the right hand side are likely to be correlated with the residuals. This case is due to possible endogeneity of the regressors with respect to the left hand side variable (the level of protection, τ EU j ). As a consequence, ordinary least squares would produce biased and

13

inconsistent estimates that turn out unreliable. To cope with this problem we implement a 2SLS (two-stage least squares or instrumental variables) estimation procedure. This strategy is essentially an econometric device for capturing the exogenous source of crossS sector variation in the right hand side variables (ΩE j , Ωj , Ωj , and Tj ) that influences the

degree of protection. Accordingly, we choose a number of instrumental variables (at least as many as the endogenous regressors) that are not included in model (18) and meet the following two requirements: a. Relevance: The instruments must be significantly correlated with the endogenous regressors; following the literature we say that they must be not weak. In the presence of weak instruments the estimated coefficients from the identification tests turn out non-standard and tests can be misleading (Staiger and Stock 1997). b. Exogeneity: The instruments must affect the dependent variable only through the endogenous regressor(s) and not directly. Technically, what we do is then estimating the model in two stages. In the first stage, the right hand variables are regressed on a constant and the instruments; the fitted values are hence predicted and saved. In the second stage, regression (18) is estimated using the fitted values from the first stage(s). The estimates of the regression coefficients obtained are now consistent and unbiased. Following Mitra et al. (2002), as instrumental variables we choose variables that include different forms of domestic (non-trade) support and represent sectorspecific characteristics. They are likely to be strongly correlated with sectoral performance such as the level of output, imports, market power and organization ability, but not to exert direct effects (besides those produced through the just mentioned channels) on the rate of protection (these hypothesis are verified in subsection 4.3). Finally, in the estimation, we allow for errors to be correlated through clustering across sectors by using heteroskedasticity robust standard errors.

4.2

Data and variables’ construction

To estimate model (18) we need data on output and demand, prices, sectoral characteristics and protection rates, behavioral parameters for the member countries forming the EU as 14

well as for the EU as an aggregate. To satisfy our data requirements we mainly use the Global Trade Analysis Project (GTAP) dataset version 6, which provides data on trade, production and demand for several countries and regions, and for 57 sectors. In what follows we briefly describe the variables used. is the ad-valorem equivalent rate of protecProtection. The dependent variable τ EU j tion. It includes tariff as well as non-tariff barriers. Following Belfrage (2004), it is obtained by the ratio of the sum of import tax revenue and export subsidy to the sum of imports and exports. This measure includes both tariff and non-tariff barriers as the non-tariff barriers have been converted to trade tax/subsidy equivalents in the GTAP database. Import penetration ratio, demand elasticity and political weights. To construct P φ zjk Ωj = k∈K k we need data on national value output, volumes of domestic imports and ejk private demand (that is, in turn, the sum of total private domestic purchases and import) that are directly provided by the GTAP database. ejk is obtained following Hertel (1997) and Belfrage (2004) to whom we refer the reader for further details. The weights φk are equal to the share of votes available to each member country in the Council of Ministers (reported in the appendix). Terms of trade effect. Departing from the previous literature, our model includes the large country effect of tariff formation. This is analytically obtained from the maximization of (12) and the assumption that the EU as a whole is indeed large with respect to the rest of the world. The large country effect, Tj , is proxied by the share of European imports in world trade for each sector j (see also Belfrage 2004). Instrumental variables. Resorting again to the GTAP database, we use different forms of domestic (non-trade) support that represent sector-specific characteristics, namely: output subsidies, tax on primary factors (labor), tax on primary factors (capital), private domestic consumption taxes, government domestic consumption taxes, government’s domestic purchases taxes, firms’ purchases of primary factors, spending for unskilled labor, dummy variables for manufacturing and agricultural sectors. S and I E , where the former represents a dummy Political dummies. To construct Ijk j

for lobbies at the national level, while the latter stands for lobbies at the European level, is not straightforward. Indeed contributions data in the European countries simply do not

15

exist, as they would denote illegal activities. This is maybe the major reason why the literature on lobbying and trade policy has so far privileged the US. In this paper, we try to circumvent the problem by constructing proxy variables. In particular, we rely on a heuristic approach and assume that industries that are most likely to be organized in pressure groups are those characterized by large firms. This choice is coherent with the literature on organized groups and collective action for three main reasons. First, bigger industry size is associated with larger stakes involved in cooperation among producers to ask for protection. As emphasized by Tavares (2006), the larger the industry asking for protection, the greater is the incentive to take part in the tariff-setting process. Second, larger firms are associated to a larger number of votes in the elections, then to more effective lobbying. Finally, big size enterprises tend to concentrate a large number of people in the same environment so leading to cooperation. Although collective action and free-riding problems can also arise (see, in particular, Olson 1965), we believe the above motivations sufficient to assume that larger firms are better able to exert pressure on policy makers (see also Balaoing and Francois 2005, Karacaovali and Limão 2005, Tavares 2006). Thus, we collect data on value added and number of firms present in each sector from the Eurostat S is set equal to one if the average value Industry, Trade and Services (2006) database. Ijk

added per firm in the considered industry is in the 70th percentile of the sector-specific and country-specific distribution, and zero otherwise. IjE is defined in an analogous way with respect to the European aggregates’ distribution. This reasoning is not valid, however, for agricultural sectors that, as well known, enjoy a privileged position with regard to protection in the EU, regardless farms’ size. Consequently, rather than constructing the proxy as explained above, we always set the European political dummy variable equal to one if the sector is agricultural. The methodology explained above presents two major problems. First, it suffers from arbitrariness since the criterion chosen to identify organized sectors is only one among several possible alternatives, all coherent with theory. Second, our identification strategy distinguishes organized from unorganized sectors regardless the specific policy issue considered. In facts we are unable to establish the reason why a sector is said to be organized, and to separate trade policy concerns from any other, for instance from domestic policies not

16

related to trade (for a discussion on this limitation of the identification strategy with regard to previous empirical works on the political economy of trade see Gawande and Krishna 2003). To verify the proposed classification, we implement a validation procedure suggested by Mitra et al. (2002). Accordingly, we run a probit regression of the (ex-ante) political dummy, constructed as explained above, on a number of variables that denote industry characteristics and are likely to be correlated with the ability of a given sector to sheer political favor, some explicitly relating to trade policy. They are: percentage of value of total purchase of a certain good over total expenditures, elasticity of import demand, firms’ domestic purchases and imports of intermediate goods, value of imports, spending for unskilled labor, dummies for agricultural and manufacturing (versus extraction) sectors. Then we construct a new dummy (ex-post) that is equal to one if the predicted probability that a sector is organized exceeds the 60% and zero otherwise. As a last step, we compare pairwise the political dummies constructed with the two alternative procedures (respectively ex-ante and ex-post classification). With regard to national lobbies we obtain that ex-ante classification is validated in the 84% of cases; whereas with regard to European lobbies the 90% of cases are correctly predicted. These percentages are fairly large (also considering the small size of the sample used) to support the ex-ante classification that will be used in the econometric analysis that follows. We are aware, however, that severe problems of missclassification may affect our econometric analysis, and this is why we warn the reader to take results with prudence. Further effort to find a more appropriate identification strategy is required in the future, in order to single out organized sectors with more accuracy and to focus on trade policy concerns only. This seems a fertile field for research. Sample. A last notice concerns the sample of sectors and countries used in the estimation. The GTAP database covers 57 industries out of which we extract only those that are tradable and for whom market prices comparisons among countries is sensible. This leaves us with 41 sectors. The list of included sectors is provided in the appendix. Data are referred to 2001. We thus refer to the 15 countries belonging to the EU in that year (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxem-

17

burg, Netherlands, Portugal, Spain, Sweden, and the UK). Aggregation is implemented as required by the model and explained in the text.

4.3

Results

Table 1 reports our results from the 2SLS estimation of model (18). As one can notice, our econometric output strongly supports the predictions suggested by the theoretical model: protection to (either national or European) organized sectors is positively related to the level of weighted average of the inverse of the import penetration ratio in the 15 EU countries and negatively related to the weighted average of absolute values of national import elasticity. The opposite holds for unorganized sectors. This result is consistent with the predictions of the original GH model. To confirm the reliability of our conclusions, some checks on instruments’ reliability are compelling. First (condition a.), we want to be sure that instruments are not weak.

Table 1: Model estimation 2SLS Estimation (second stage) − Dependent variable: τ EU j Variable (parameter)

Coefficient

t-student

p-value

Expected sign

ΩSj

(β)

0.0120

2.30∗∗

0.027

>0

ΩE j

(γ)

0.0070

2.02∗∗

0.050

>0

Ωj

(δ)

-0.0087

-2.31∗∗

0.026

0

5.03∗∗∗

F -stat (p-val) = 7.75∗∗∗ (0.000)

Number of obs = 41

Correlation protection/fitted values = 0.559 (0.000) First stage information: Dep variable: ΩSj

F -stat (p-val) = 33.22∗∗∗ (0.000)

R2 = 0.92

Dep variable: ΩE j

F -stat (p-val) = 15.22∗∗∗ (0.000)

R2 = 0.85

Dep variable: Ωj

F -stat (p-val) = 30.05∗∗∗ (0.000)

R2 = 0.92

Dep variable: Tt

F -stat (p-val) = 20.60∗∗∗ (0.000)

R2 = 0.88

J-stat (p-val) = 12.1 (0.100) NOTES: Significance at: ∗∗ 5%; ∗∗∗ 1%

18

To this purpose, we need to control that: (i) the first stage F -statistics are significantly different from zero and, as a rule of thumb, larger than 10 (Stock, Wright and Yogo 2002); (ii) the first stage R2 is greater than 30% (Shea 1997). As one can notice, these requirements are always met by our econometric output. Second (condition b.), instruments must be exogenous. Accordingly, we implement the J-test for overidentifying restrictions. It detects departures from the assumption that instrumental variables can be excluded from the second stage, i.e. are uncorrelated with εj . The J-statistics is given by J = Fh × h, where Fh is the joint F -statistics for the excluded instruments in the first stage regression, and h is the number of excluded instruments. J is distributed as a ChiSqare with h − m degrees of freedom (m being the number of endogenous regressors). As table 1 shows the null hypothesis of exogeneity cannot be rejected at least at the 10% level. Finally, as a general check, we compute the correlation coefficient between actual and fitted values of the degree of protection, and find a positive value that is statistically significant at any confidence level. Once sensible quantitative results have been obtained, we are now ready to test the major qualitative implications of the model. First, we want to control that the sum of the coefficients β, γ and δ is larger than zero. Given that: 0.0120 + 0.0070 − 0.0087 > 0, we can test the null hypothesis H0 : β + γ + δ = 0 against the alternative H1 : β + γ + δ > 0. Results are in table 2, from which we can notice that the null is rejected at least at the 5% level. Table 2: Hypothesis testing Null hypothesis

F -stat

p-value

β+γ+δ =0

4.41∗∗

0.042

1/a = 0

5.39∗∗

0.026

b=1

34.96∗∗∗

0.000

1/a = 0 and b = 1

29.73∗∗∗

0.000

αL = 0

341.72∗∗∗

0.000

αL = 0, 1/a = 0 and b = 1

241.56∗∗∗

0.000

NOTES: Significance at: ∗∗ 5%; ∗∗∗ 1%

19

Second, using the estimated coefficients and the structural model, we can compute the implied value of the relevant behavioral parameters. The political weights turn out to be: a = 83 and b = 0.58. We observe that the estimated value of a is consistent with previous studies already mentioned in this paper (Gawande and Bandyopadhyay 2000, Goldberg and Maggi 1999, and Mitra et al. 2002). A large value of this parameter implies that the government does not substantially differentiate between the relative weights on (either national or European) specific-factor owners’ and total population’s welfare. As noticed by Gawande and Bandyopadhyay (2000), this would be in conflict with the empirical literature on computational general equilibrium models which suggests that what lobbies obtain from protection is much larger than what they spend (see for instance Hufbauer et al. 1986, and Stern 1988). Furthermore, a value of b very close to unity hints that the policy maker places the same weight upon national and European lobbying. Therefore, we test the hypotheses that 1/a = 0 (i.e. a → ∞) and b = 1, either jointly and separately taken. As table 2 shows we are led to reject the null hypotheses in both cases at any standard confidence level. Finally, our econometric output suggests that the percentage of the voting population organized in interest groups is αL = 0.75. While this value is quite large, we remind that, as it turns out from the theoretical model, it includes both national and European lobbies. Again, as expected, we are induced to strongly reject the hypothesis that it is zero as table 2 indicates.

5

Conclusions

Both theoretical and empirical analysis of trade policy and lobbying activity in the EU is very scanty in the literature. This is for several reasons. First of all the EU presents a sui generis institutional environment, as it negotiates with its trade partners as a single entity with a single common trade policy but, at the same time, member states’ interests maintain an important and autonomous role in the policy making process. Second, the cobweb of interests at stake is very complex as national and supranational issues meet continuously and often conflict, at different stages of the policy making process. Third, lobby formation at the European level is a quite new experience as it has become more intense only after the Single European Act ratification. Only since then, a lively debate has arisen in the 20

literature on whether European transnational groups stick to their national patterns of representation or evolve mostly relying on transnational bases. Fourth, traditional means of political pressures (contributions for re-election) are being substituted for new forms of lobby activity such as information provision during the legislative process. Finally, and as a consequence of the above points, data on political contributions are simply not available for European countries. Our study has been addressed to cope with the mentioned difficulties. The chosen strategy has consisted in taking a well theoretically founded model of trade policy formation in the presence of lobby groups (Grossman and Helpman 1994), and tailoring it to fit, as much as possible, the EU institutional environment. To accomplish this goal we resorted to a game model where policy formation at the European level is the result of two stages: in the first stage national representatives’ preferences are formed, and in the second one they are aggregated in a common trade policy measure. We have then derived an estimable model and tested it with data from the Global Trade Analysis Project and the Eurostat. Our results are interesting. We obtain that the model fits very well the data as all the theoretical predictions are largely confirmed. Lobby groups, at both national and international levels, manage to obtain higher protection the higher the inverse of the import penetration ratio and the lower the deadweight loss for protection (price effect on imports). The terms of trade exerts a further positive effect on the level of trade protection. While these results seem very promising a note of caution is due. Indeed, the absence of data on political contributions (or on sensible proxies) represents a strong constraint to the research on the political economy of trade in the EU. As already noticed, this has strongly affected our identification strategy that had to rely on indirect and approximate selection criteria. On the other side, the increasing role acquired by supranational interest groups in the European arena suggests that the issues presented in this paper open up the door to a rich and fruitful area for inquiry. We believe that further research effort should be directed, in the future, to data collection in order to obtain more robust and reliable empirical evidence in this field.

21

References Balaoing, Annette and Joseph Francois (2005) “The Political Economy of Protection in a Customs Union: What Drives the Tariff Structure of the EU?”, mimeo, Tinbergen Institute, Erasmus University Rotterdam Belfrage, Carl-Johan (2004) “Special Interest Politics and Trade Policy - An Empirical Challenge”, mimeo, Department of Economics, Lund University Broscheid Andreas and David Cohen (2003) “Insider and Outsider Lobbying of the European Commission An Informational Model of Forum Politics”, European Union Politics 4(2), 1465-1165 Cadot, Olivier, de Melo Jaime and Marcelo Olarreaga (1999) “Regional Integration and Lobbying for Tariffs Against Non-Members”, International Economic Review 40(3), 635-57 Cohen, David (1997) “The European Business Lobby”, Business Strategy Review 8(4), 17-25 Crombez, Christopher (2002) “Information, Lobbying and the Legislative Process in the European Union”, European Union Politics 3(1), 7-32 Eurostat

(2006)

Industry,

Trade

and

Services,

http://epp.eurostat.cec.eu.int/pls/portal/url/page/SHARED/PER_INDCOM Gawande, Kishore, and Usree Bandyopadhyay (2000) “Is Protection for Sale? Evidence on the Grossman-Helpman Theory of Endogenous Protection”, Review of Economics and Statistics 82(1), 139-152 Gawande, Kishore, and Pravin Krishna (2003) “The Political Economy of Trade Policy: Empirical Approaches”. In E. Kwan Choi and James Harrigan (eds) Handbook of International Trade, Vol. I. Malden (MA): Blackwell Publishing, pp. 213-250 Goldberg, Penelopi, and Giovanni Maggi (1999) “Protection for Sale: An Empirical Investigation”, American Economic Review 89(5), 1135-1155 Grossman Gene M. and Elhanan Helpman (1994) “Protection for Sale”, American Economic Review 84(4), 833-50 Grossman Gene M. and Elhanan Helpman (1995a) “Trade Wars and Trade Talks”, Journal of Political Economy 103(4), 675-708

22

Grossman Gene M. and Elhanan Helpman (1995b) “The Politics of Free Trade Agreements”, American Economic Review 85, 667-90 Global Trade Analysis Project (2005) GTAP 6 Data Base, West Lafayette (IN): Purdue University Press Hertel, Thomas W. (1997) Global Trade Analysis: Modeling and Applications, Cambridge: Cambridge University Press Hufbauer, Gary C., Berliner Clyde and Kimberly A. Elliott (1986) Trade Protection in the US: 31 Case Studies, Washington D.C.: Institute for International Economics Karacaovali, Baybars and Nuno Limão (2005) “The Clash of Liberalizations: Preferential vs. Multilateral Trade Liberalization in the European Union”, World Bank WP 3493 Mazey, Sonia and Jeremy Richardson (1993) “Transference of Power, Decision Rules, and the Rules of the Game”. In Sonia Mazey and Jeremy Richardson (eds) Lobbying in the European Community. Oxford: Oxford University Press: pp. 3-26 McCalman, Phillip (2004) “Protection for Sale and Trade Liberalization: an Empirical Investigation”, Review of International Economics 12(1), 81-94 Mitra, Devashish, Thomakos, Dimitrios D., and Mehmet A. Uluba¸so˘ glu (2002) “Protection for Sale in a Developing Country: Democracy vs. Dictatorship”, Review of Economics and Statistics 84(3), 497-508 Olson, Mancur (1965) The Logic of Collective Action. Cambridge (MA): Harvard University Press Persson, Torsten (1998) “Economic Policy and Special Interest Politics”, Economic Journal 108(447), 310-327 Shea, John (1997) “Instrument Relevance in Multivariate Linear Models: A Simple Measure”, Review of Economics and Statistics 2(49), 348-352 Stern, Phillip M. (1988) The Best Congress Money Can Buy, New York: Pantheon Staiger, Douglas and James H. Stock (1997) “Instrumental Variable Regression with Weak Instruments”, Econometrica 5(65), 557-586 Stock, James H., Wright, Jonathan H. and Motohiro Yogo (2002) “A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments”, Journal of Business and Economic Statistics 20(4), 518-529

23

Tavares, Samia (2006) “Deeper Integration and Voting on the Common European External Tariff” MPRA Paper 960, University Library of Munich (Germany)

A

Mathematical appendix

A. 1. Proof of (14): To obtain (14), we maximize (∀j ∈ M ): ¤2 P 1 £ EU θk tj − e tjk k∈K 2

(A.1)

The first order condition is: h i2 P EU − e ∂ t θ t jk ∂p∗j £ EU ¤ P P k∈K k j 1 e = θ − t θ =0 t + k j jk k 2 ∂tEU ∂tEU k∈K k∈K j j Whence:

tEU j

P e ∂p∗j k∈K θ k tjk = P + EU ∂tj k∈K θ k

(A.2)

(A.3)

A. 2. Proof of (15): To find (15) for each member country, we need to find e tjk such that

the following two conditions are met: 1. e tk = arg max (Wik − Cik ) , h tk

∀i ∈ L;

2. e tk = arg max (Wjk − Cjk ) + b h tk

P

Cik +

i∈E

P

i∈S

Cik + aWk ,

∀j ∈ L.

From 1., we get (∀j ∈ L and k ∈ K): ∇Cjk = ∇Wjk

(A.4)

and, from 2.: ∇ (Wjk − Cjk ) + b

P

i∈E

∇Cik +

P

i∈S

∇Cik + a∇Wk = 0

(A.5)

Summing (A.4) over j, and combining with (A.5), it follows: b

P

i∈E

∇Wik +

P

i∈S

∇Wik + a∇Wk = 0

(A.6)

´ ³ where Wik (p) ≡ lik + π ik (pi ) + αik N [r (p) + s (p)]. Since tj = pj − p∗j and p∗j is exoge-

nous, we can derive with respect to pj (notice that some mathematical details are skipped

24

for reasons of space). We have: ∂sik = pej d0jk (pj ) − pej d0jk (pj ) − djk (pj ) = −djk (pj ) , j ∈ M ∂pj ¢ ∂rik 1 1 ¡ = mjk (pj ) + pj − p∗j m0jk (pj ) , j ∈ M ∂pj N N ¡ ¢ ∂Wik = yjk (pj ) − αi N djk (pj ) + αi mjk (pj ) + αi pej − p∗j m0jk (pj ) , ∂pj

(A.7) (A.8) i ∈ L, j ∈ M (A.9)

where we have used the fact that: mjk (pj ) − N djk (pj ) = −yjk (pj ) , αi = αik ∀k ∈ K and u0 [djk (pj )] = pj . Separating national from European lobbies, it follows that: ¡ ¢ ∂Wik = (δ jS − αiS ) yjk (pj ) + αiS pj − p∗j m0jk (pj ) , i ∈ S, j ∈ M ∂pj ¢ ¡ ¢ ∂Wik ¡ = η jE − αiE yjk (pj ) + αiE pj − p∗j m0jk (pj ) , i ∈ E, j ∈ M ∂pj

(A.10) (A.11)

where δ jS = 1 if i ∈ S and = 0 otherwise, whereas ηjE = 1 if i ∈ E and = 0 otherwise. Then, aggregating equations (A.10) and (A.11) over i ∈ L, the first two terms of the summation in equation (A.6) turn out to be (with j ∈ M ): ¢ ¡ ¢ P ∂Wik ¡ S P ∂Wik +b = Ijk + bIjE − αL yjk (pj ) + αL pj − p∗j m0jk (pj ) i∈S ∂pj i∈E ∂pj

where IiE ≡

P

S ≡ η iE = 1 if i ∈ E and = 0 otherwise, Iik

(A.12)

P

= 1 if i ∈ S in P country k and = 0 otherwise. For simplicity we set αL = αS + bαE , where αS ≡ i∈S αiS P (αE ≡ i∈E αiE ) is the fraction of total population of voters represented by an organized i∈E

i∈S δ iS

national (European) group. We derive now Wk with respect to pj and get (with k ∈ K and

j ∈ M ):

∙ ¸ ¢ 0 1 1 ¡ ∂Wk ∗ = yjk (pj ) + N −djk (pj ) + mjk (pj ) + pj − pj mjk (pj ) ∂pj N N ¢ ¡ = pj − p∗j m0jk (pj )

(A.13)

Thus, combining (A.12) and (A.13) in equation (A.6), we have: ¡

¡ ¢ ¡ ¢ ¢ IjS + bIjE − αL yjk (pj ) + αL pj − p∗j m0jk (pj ) + a pj − p∗j m0jk (pj ) = 0 ¡ ¢ tjk m0jk (pj ) + ae tjk m0jk (pj ) + IjS + bIjE − αL yjk (pj ) = 0 αLe

e tjk = −

IjS + bIjE − αL yjk (pj ) , αL + a m0jk (pj )

25

where IiS + bIiE − αL > 0,

(A.14)

that is the tariff platform for country k. From (A.3) and substituting e tjk for (A.14), we can finally write:

IjS + bIjE − αL P ∂p∗j yjk (pj ) + EU θk 0 αL + a mjk (pj ) ∂tj k∈K k∈K θ k

tEU = −P j

1

(A.15)

Dividing both sides of (A.15) by pj , defining ejk = −m0jk pj /mjk , and using the fact that: Ã ! P θk yjk 1 θK yjK mjK θ2 yj2 mj2 θ1 yj1 mj1 1 1 −P = −P + + ... + 0 0 m0j1 pj mj1 m0j2 pj mj2 mjK pj mjK k∈K θ k k∈K mjk pj k∈K θ k µ ¶ yjK /mjK yj1 /mj1 yj2 /mj2 1 =P + θ2 + ... + θK θ1 ej1 ej2 ejK k∈K θ k µ ¶ P zjk 1 =P θk , (A.16) ejk k∈K θ k k∈K we obtain equation (17).

B

Appendix to the empirical analysis

B. 1. List of sectors included in the analysis1 : 1 - Paddy rice; 2 - Wheat; 3 - Cereal grains nec; 4 - Vegetables, fruit, nuts; 5 - Oil seeds; 6 - Sugar cane, sugar beet; 7 - Plantbased fibers; 8 - Crops nec; 9 - Bovine cattle, sheep and goats, horses; 10 - Animal products nec; 12 - Wool, silk-worm cocoons; 13 - Forestry; 14 - Fishing; 15 - Coal; 16 - Oil; 17 - Gas; 18 - Minerals nec; 19 - Bovine cattle, sheep and goat, horse meat prods; 20 - Meat products nec; 21 - Vegetable oils and fats; 22 - Dairy products; 23 - Processed rice; 24 - Sugar; 25 - Food products nec; 26 - Beverages and tobacco products; 27 - Textiles; 28 - Wearing apparel; 29 - Leather products; 30 - Wood products; 31 - Paper products, publishing; 32 Petroleum, coal products; 33 - Chemical, rubber, plastic products; 34 - Mineral products nec; 35 - Ferrous metals; 36 - Metals nec; 37 - Metal products; 38 - Motor vehicles and parts; 39 - Transport equipment nec; 40 - Electronic equipment; 41 - Machinery and equipment nec; 42 - Manufactures nec. Source: http://www.gtap.agecon.purdue.edu/. B. 2. List of countries and relative political weights: 1 - Austria (4); 2 - Belgium (5); 3 - Denmark (3); 4 - Finland (3); 5 - France (10); 6 - Germany (10); 7 - Greece (5); 8 1 All

the services are excluded. Furthermore we do not include in the analysis those sectors where market prices

comparisons among countries would be misleading. The criterion (suggested by Belfrage 2004) is not to include in the analysis sectors for whom concordance with the international classification system for tradable goods (SITC/HS ) is not available. This leads us to exclude the raw milk sector.

26

Ireland (3); 9 - Italy (10); 10 - Luxemburg (2); 11 - Netherlands (5); 12 - Portugal (5); 13 Spain (8); 14 - Sweden (4); and the 15 - UK (10). Source: http://europa.eu.int/

27