fiscal federalism and lobbying

Journal of Public Economics 92 (2008) 2288–2301

Contents lists available at ScienceDirect

Journal of Public Economics

j ourna l homepage: www.e lsev ie r.com/ locate /econbase

Fiscal federalism and lobbying

Massimo Bordignon a, Luca Colombo a,⁎, Umberto Galmarini b

a Istituto di Economia e Finanza, Università Cattolica, Largo Gemelli 1, 20123 Milano, Italyb Dipartimento di Diritto ed Economia, Università dell'Insubria, Viale Cavallotti 5, 22100 Como, Italy

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +39 02 7234 2637; faxE-mail addresses: [email protected] (

0047-2727/$ – see front matter © 2008 Elsevier B.V.doi:10.1016/j.jpubeco.2008.05.003

a b s t r a c t

Article history:Received 18 April 2005Received in revised form 20 March 2008Accepted 7 May 2008Available online 24 May 2008

Which government functions should be decentralized (resp. centralized) once lobbyingbehavior is taken into account?We find that the answer largely depends on how the interests ofthe regional lobbies are positioned with respect to the function to be decentralized (resp.centralized). When regional lobbies have conflicting interests, then lobbying is less damagingfor social welfare under centralization than under decentralization. On the contrary, whenregional lobbies have aligned interests, then lobbying is less damaging for social welfare underdecentralization, provided that policy spillovers on the non-organized groups are not toostrong.

© 2008 Elsevier B.V. All rights reserved.

JEL classification:D70H23H77

Keywords:Fiscal federalismLobbyingCommon agencyGames played through agents

1. Introduction

One of themost fundamental questions in the theory of fiscal federalism concerns the correct allocation of functions to differentlevels of government. This question has not only theoretical appeal. Given the recent and widespread tendency towardsdecentralization within countries, and centralization (of some functions) across countries, it also has a deep policy content.Economists are not completely devoid of answers. According to Oates's (1972) celebrated ‘decentralization theorem’, we shouldcentralize (decentralize) functions with more (less) spillover effects and less (more) heterogeneity of preferences acrossjurisdictions. In its simplicity, this is a recipe that can carry one someway (see, for instance, Alesina et al., 2005; Tabellini, 2003, onthe European Union). However, an important limitation of Oates's analysis is that it assumeswelfaremaximizing governments, andit is not clear how far its insights could go in more realistic political environments.

As anexample, consider the current debate on the role that the EuropeanUnion (EU) should playon the supply side of the economy,in fields such as labormarkets institutions, pension policy, infrastructures, competition and regulation policy. In these fields, currentlylargely under the control of national governments,many observers would agree that themost important policy distortions come fromthe pressure of powerful organized interest groups on governments (e.g. Tabellini andWyplosz, 2006). The important policy questionis thenwhether these pressures are likely to becomemore or less powerful once these functions were centralized at the EU level. Butthis is not the issue that has been considered in the attempts to extend Oates' analysis to a political economy framework (see Besleyand Coate, 2003; Lockwood, 2002). Moreover, despite the large economic literature on lobbying (see Grossman and Helpman, 2001,for a comprehensive survey), very few studies have concentrated on the specific relationship between interest groups and(de)centralization. Andwhen they have done so, theymostly focused on the higher heterogeneity of preferences under centralizationas the main discriminating factor (e.g. De Melo et al., 1993, and Redoano, 2002). However, the differences in preferences among EUcountries do not seem to play an important role in the cases discussed above (see again Alesina et al., 2005, for empirical evidence).

: +39 02 7234 2781.M. Bordignon), [email protected] (L. Colombo), [email protected] (U. Galmarini).

All rights reserved.

mailto:[email protected]



http://dx.doi.org/10.1016/j.jpubeco.2008.05.003

http://www.sciencedirect.com/science/journal/00472727

2289M. Bordignon et al. / Journal of Public Economics 92 (2008) 2288–2301

A more extensive discussion of the relationship between decentralization and lobbying is offered by Bardhan and Mookherjee(2000, 2002). They consider a probabilistic model of voting, where rich voters may decide to form an interest group which offerscampaign contributions to Downsian parties before the elections, in order to influence the voting behavior of poorer and lessinformed voters. Comparing national versus local elections, they find several arguments supporting the idea that interest groupsare more powerful at the local level, and many others that pull in the opposite direction, so that the question rests on empiricalgrounds.1 Other studies are also somewhat connected to the topic, although they do not explicitly focus on the relationshipbetween decentralization and lobbying. For example, Treisman (2006) criticizes the usual argument in favor of assigning a largeshare of locally generated revenue to local authorities in tax sharing systems, by noting that the incentives given to local politiciansfor improving the local economy (that may take the form of bribes from local businesses) may be counterbalanced by the oppositeincentives given to the central government. Brou and Ruta (2006) consider the effect of political integration when lobbies areasymmetrically distributed among integrating countries, arguing that the country with more organized groups should fare betterin the ensuing political game. Endogeneizing lobbying formation, they then conclude that political integration (e.g. centralization)should promote more lobbying formation.

While we do not question the potential importance of all these factors, we argue that there is another important channel whichhas so far escaped attention. This has to do with the simple idea, widespread in the political debate, that local governments caremore for local interests than for foreign ones, as supporting a local interest generates additional benefits for the local politiciansthan supporting a foreign one (e.g. Prud'homme, 1994). As an example, advocates of the EU traditionally argue for more power tobe assigned to the Union to reduce each country's incentives to defend its own ‘national champions’. Loosely, the argument is thatnational governments do not internalize the effects of their policies on foreign interests and, as a result of this failure, they makeinefficient choices that a federal government would avoid. While interesting, this argument is however incomplete. It fails torecognize that interest groups would attempt to influence policy even under centralization, so that the true question to ask, asargued above, is whether these pressures are likely to become less important once those functions were centralized.

To address this issue, we build a simple model that encompasses several possible realistic interactions among national lobbiesand national residents. We consider two identical regions and, in each region, one organized group which may lobby politicians tothe detriment of non-organized interests. Organized groups may have either conflicting interests – when the successful lobbyingactivities of a regional lobby hurt the other regional organized group – or aligned interests – when the lobbying activities of aregional lobby benefit also the other. Similarly, non-organized groups may be damaged either only by the lobbying activitiesoccurring in their region of residence, or by those occurring in the other region as well. Lobbying equilibria are studied under bothcentralization, in which a national policy maker is in charge of all decisions, and decentralization, in which local policy makers(simultaneously) set local policies. Following Grossman and Helpman (1994, 2001) and Dixit et al. (1997), we model lobbying byusing the common agency approach of Bernheim and Whinston (1986a), extending it to multiple principals and multiple agentswhen considering the decentralized case (see Prat and Rustichini, 2003). In this approach, politicians maximize a weighted utilityfunction which takes into account both social welfare and money contributions from the interest groups.2 This is of course a‘reduced form’ of political behavior that lacks many details emphasized by the more recent literature. However, in our view, thissimpler approach has the advantage of maintaining a larger generality.

We get very sharp results. In our basic model, lobbying is more damaging for social welfare under decentralization than undercentralization when the lobbies have conflicting interests; while it is more damaging for welfare under centralization when theinterests of the lobbies are aligned, provided that policy spillovers on non-organized groups are not too large and that policymakers do not attach too much (or too little) weight on lobbies' contributions. The latter result is somehow surprising as in thebenchmark allocations in the absence of lobbying decentralization is always worse than centralization in terms of social welfare.The intuition is very simple and rests on an externality argument. When the interests of the two lobbies are aligned,decentralization means that the local policy makers do not take into account the beneficial effect of their policy on the non-resident interest group. Thus, the resulting policy under-provision reduces the distortionary effect of lobbying. In the opposite caseinwhich the interests of the two lobbies are conflicting, local policymakers do not internalize the interests of the group hurt by thepolicy, thus strengthening lobbying distortions under decentralization. Hence, our model offers one important insight. It suggeststhat the answer to the normative question of the allocation of functions to different levels of government under lobbying behaviormay also depend on the specific function under consideration, and in particular on how the interests of the national lobbies arepositioned with respect to that particular function.

Naturally, in order to keep the model tractable we do not explicitly model many other relevant features that could make adifference between centralization and decentralization. For example, we do not explicitly account for within groups heterogeneityof preferences, agents' mobility, intergovernmental transfers, tax competition, or differences in the preferences of local versusnational politicians.3 We consider however some extensions of the basic model in Section 7, showing that our results survive, with

1 We are not aware of any specific empirical work studying the relationship between decentralization and lobbying. There are, however, empirical worksdiscussing the relationship between corruption and decentralization, usually finding a negative correlation between the two. For recent examples, see Treisman(2000), and Fisman and Gatti (2002).

2 We focus exclusively on lobbying activities aimed at buying influence by means of contributions offered to the policy maker. However, lobbying activities maybe directed at a variety of other goals, such as providing information to policy makers. See Grossman and Helpman (2001) and Spiller and Liao (2006) forcomprehensive surveys.

3 These features have already been scrutinized by a large literature. For instance, see Persson (1998) and Persson and Tabellini (2000) for ‘common pool’ effectswhichmay arise out of transfers from the central level to local ones,Wilson (1999) for ‘fiscal competition’ effects whichmay arise out of themobility of the tax base,Keen and Kotsogiannis (2002) for ‘spillover effects’ in taxation, and Seabright (1996) for honesty effects induced by the larger accountability of local politicians.

2290 M. Bordignon et al. / Journal of Public Economics 92 (2008) 2288–2301

some qualifications, in these more complex settings. In particular, we discuss situations in which organized interests have adifferent capacity to influence foreign and home policy makers, we consider the case in which discriminated minorities uselobbying to voice their interests, and finally we extend our framework by allowing for a richer structure of policy spillovers.

The rest of the paper is organized as follows. In Section 2 we set up the basic model. In Section 3 we derive the benchmarkallocations in the absence of lobbying and introduce lobbying behavior. In Sections 4 and 5 we derive the equilibria under lobbyingin centralized and decentralized settings, respectively. The welfare comparison between the two institutional regimes is carriedover in Section 6. Section 7 considers the extensions of the basic model we referred to above. Finally, Section 8 concludes bysummarizing the results and suggesting avenues for further research. The proofs of all propositions and the characterization of thefirst stage of the lobbying game are contained in two Appendices.

2. The model

Consider an economy composed of two identical regions (or jurisdictions), labeled r={α, β}. Residents in each jurisdiction aredivided into two groups: a ‘majority’ group, of mass 1, and a ‘minority’ group, of mass θ, θ ∈ (0, 1). Let A and B be the two majoritygroups, and a and b the minority groups, resident in regions α and β, respectively. The choice of residence is exogenously given,and there is no mobility of individuals across groups. In each region, a single policy instrument is set by the policy maker in power.Let pr≥0 denote the level of policy set in region r.

Members of each group inside each region have identical preferences, but there is heterogeneity across groups. In particular, theper capita welfare of members belonging to group i, i={a, b}, is represented by the linearly additive utility functions:

wherethe po

where

Ua ¼ pα + λpβ ; ð1Þ

Ub ¼ pβ + λpα ; ð2Þ

λ ∈ [−1, 1]. The value of the parameter λ accounts for regional policy spillovers. If λ=0 then each group i benefits only fromlicies implemented in its own region. If λN0 (resp. λb0) then the policy implemented in one region generates a positivenegative) externality on the group resident in the other region.
(resp.
Members of group I, I={A, B}, have identical preferences represented by a strictly concave utility function in the policyvariables:

UA ¼ 1 − σð ÞF pα;ωð Þ + σF pβ;ω� �

; ð3Þ

UB ¼ 1 − σð ÞF pβ;ω� �

+ σF pα;ωð Þ; ð4Þ

ω≥0, σ ∈ 0; 12� �

, and

F pr;ωð Þ ¼ ω − pr=2ð Þpr :

arameter ω denotes the preferred policy levels by members of groups I, whereas the parameter σ captures the extent of
The ppolicy spillovers for the majority groups. Specifically, if σ =0 then members of groups I are only affected by policies set in the‘home’ region, while if σ N0 then also ‘foreign’ policies affect the members of these groups.
Themodel may capturemany real world situations inwhich policy spillovers are relevant. For instance, let p represent antitrustpolicy, with high levels of p being ‘soft’ policies. In this case, a natural interpretation of the model would have groups a and brepresenting firms, and groups A and B representing consumers. Consumers desire a strong antitrust policy (and thus ω should beset at zero), while firms would prefer a soft policy (i.e. p as large as possible). If the relevant setup is one of open markets then λwould be close to one, whereas if firms mostly operate within their own region then λwould be close to zero. In turn, σ could takedifferent values depending on the extent inwhich consumers engage in cross-border shopping. As another example, high levels ofpmay represent labor legislations that offer little protection to workers (groups A and B) to the benefit of firms (groups a and b). Inthis case, it is ω=0 and σ equal to or greater than zero, depending onwhether workers are hired mostly in the home region or alsoin the foreign region, respectively. Moreover, if firms' production is mainly based in their own region but trade occurs in bothregions, each local firm is positively affected by a mild labor legislation at home and a restrictive legislation in the other region,meaning that λ should be negative. As a final example, consider cash transfer programs financed through general taxation on theentire population, but mainly targeted to members of groups a and b. In this case, by setting ω=0 the utility functions (3)–(4) formembers of groups A and B can be interpreted as a reduced form representation of the welfare loss incurred for having to pay thetax bill while receiving no direct benefits from transfer programs. The parameter σ could instead capture the structure of the taxsystem. For instance, σ=0 may be appropriate if members of groups A and B pay residence-based taxes, while σ N0 may representthe case in which regional public expenditure is financed out of a common pool of national taxes.

Other cases may be more difficult to study in this simple setting, such as those in which the policy spillovers on groups a and bdepend not only on location, but also on the heterogeneity of preferences over policy. Section 7.3 extends our framework to thiscase.


3. Policy making, social optima, and lobbying

Policy makers are assumed to use a Benthamite utilitarian criterion to aggregate groups' welfare into a social welfare function.4

In jurisdiction r per capita social welfare, Wr=θUi+UI, is then equal to

Simila

4 See5 We

when p6 As c

limitati

Wα ¼ θ pα + λpβ� �

+ 1−σð ÞF pα;ωð Þ + σF pβ;ω� �

; ð5Þ

Wβ ¼ θ pβ + λpα� �

+ 1−σð ÞF pβ;ω� �

+ σF pα;ωð Þ: ð6Þ

rly, national social welfare, W=Wα+Wβ, is equal to

W ¼ θ 1 + λð Þ pα + pβ� �

+ F pα;ωð Þ + F pβ;ω� �

: ð7Þ

consider two types of institutional setting. Under the regime of policy centralization, a single policymaker is in charge of public
Wepolicy in both regions. Under decentralization, an independent regional policy maker chooses local public policy in each region. As anormative benchmark, let us consider first how policies would be set by benevolent policy makers whose only goal is social welfaremaximization. Under centralization, a benevolent policy maker maximizes national social welfare W. Therefore, policy choices by abenevolent policy maker under centralization correspond by definition to the utilitarian optimum. Under decentralization, the policymaker in region r, r={α, β}, independently from the policy maker in region j ( j=α, β, j≠r) maximizes regional social welfare Wr. Incontrast with centralization, benevolent local policy makers do not necessarily implement the utilitarian optimum, as they do notinternalize the spillover effects of their policies on the residents living in the other region. More precisely5
Proposition 1. In the utilitarian optimum, pr⁎=θ (1+λ)+ω, r={α, β}.

• Under Centralization, a benevolent policy maker implements pr⁎C=pr⁎; social and groups welfare are the same as in the utilitarian

optimum.• Under Decentralization, benevolent policy makers implement pr⁎

D=θ/(1−σ)+ω. Let σ⁎≡λ/(1+λ). The comparison between theutilitarian optimum and the decentralized outcomes is then:(D.1) if λ=σ=0, or if λN0 and σ=σ⁎, then pr⁎

D=pr⁎; social and groups welfare are the same as in the utilitarian optimum;(D.2) if λb0 and σ ∈ 0; 12

� �, or if λN0 and σ ∈ σ⁎; 12

� i, then pr⁎

DNpr⁎; social welfare is W⁎DbW⁎ and per capita groups' welfare isUI⁎DbUI

⁎, I={A, B}, U i⁎DNUi

⁎ for λ≠−1 and Ui⁎D=Ui

⁎ for λ=−1, i={a, b};(D.3) if λN0 and σ ∈ 0;σ⁎Þ½ , then pr⁎

Dbpr⁎; social welfare is W⁎DbW ⁎ and per capita groups' welfare is U I⁎DNUI

⁎, I={A, B}, U i⁎DbUi

⁎,i={a, b}.

Under decentralization, as the benevolent local policy makers ignore both types of externalities when making their policychoice, the resulting allocation is different, in general, from that of the utilitarian optimum. Whether there is policy under-provision (pr⁎Dbpr⁎) or over-provision (pr⁎DNpr⁎) depends on the sign and the strength of the policy spillovers. While a positiveexternality on the non-resident minority group (λN0) induces policy under-provision, a negative externality (λb0) induces policyover-provision. Instead, the externality on the non-resident majority group always induces policy over-provision (unless σ=0),since part of the cost for setting the policy above the preferred level (ω) of the majority groups is borne by the non-resident group.Therefore, decentralization always leads to policy over-provision if λb0, no matter the size of σ, since the two externalities push inthe same direction. Decentralization results in policy over-provision also if λN0, provided that σ Nσ⁎, so that the negativeexternality outweighs the positive one. Otherwise, if λN0 and σbσ⁎, there is policy under-provision.

To study lobbying behavior, we now assume that in each region the minority group is organized as a lobby, whereas themajority group is not organized.6 Notice that in our setting organized and non-organized groups always have conflictinginterests: each minority group i strictly prefers the largest possible level of the policy instrument set in the home region, whileeach majority group I has a bliss point at pr=ω. In contrast, the interests of the two lobbying groups may interact in differentways, depending on the sign of λ. Lobbies have aligned interests when λ ∈ (0, 1] and conflicting interests when λ ∈ [−1, 0),while they are obviously independent when λ=0. When λN0, if (say) lobby a successfully induces the (national or regional)policy maker(s) to increase the level of policy p, then the benefits of the lobbying effort also accrue to the members of group b.The opposite occurs when λb0.

As anticipated in the Introduction, we model lobbying behavior using the common agency framework proposed by Bernheim andWhinston (1986a,b), and further developed by Grossman andHelpman (1994, 2001) and Dixit et al. (1997), inwhich a set of principals(the lobbies) try to influence the actions of one agent (the policy maker). The case of policy centralizationwe consider below falls intothis category. However, the case of policy decentralization, inwhich there are twoprincipals (lobbies a andb) lobbying two agents (localpolicy makers in α and β), falls into the more general category of the so called Games Played Through Agents investigated by Prat andRustichini (2003). We begin by analyzing policy centralization.

Section 7.2 for an extension in which politicians discriminate minorities.use an asterisk to denote the allocation in the utilitarian optimum, and the superscript ⁎C (⁎D) to denote the allocation under centralization (decentralization)olicy makers are benevolent.ustomarily in this type of models, we take as exogenously given the decision of some groups to organize themselves as lobbies. This is certainly aon of this literature. We refer to Mitra (1999) and Felli and Merlo (2006) for contributions on the endogenous formation of lobbies.


4. Lobbying under centralization

Under centralization, in both regions public policy is set by a single policy maker caring about both social welfare andcontributions from the lobbies. Politicians care about social welfare, presumably because they want to be re-elected, but they alsocare for lobbyists' contributions, either because the latter increase their chances of being re-elected, or simply because lobbies paybribes that increase the policy maker's private consumption. Each lobbying groupmaximizes its welfare net of the contributions tothe policy maker. As for the timing, lobbies move first, by independently and simultaneously offering the policy maker acontribution schedule defining a monetary contribution contingent on the vector of policy instruments. Upon acceptance of thelobbies' contributions, the policy maker chooses public policy. The game is solved by backward induction.

Following Dixit et al. (1997), we focus on truthful subgame perfect Nash equilibria, inwhich each lobby offers the policymaker anon-negative compensating (or truthful) contribution schedule, shaped along its indifference curve. Formally, in the second stage ofthe game, given the per capita contribution schedules Sa and Sb offered by members of groups a and b respectively, the policymaker chooses pα and pβ to maximize her objective function

The cowelfar

ClearlycontricontriAppen

7 Sincworse o

V ¼ W + mθ Sa + Sbð Þ: ð8Þ(8), the parameter m, m≥0, measures the relative weight that the policy maker attaches to social welfare, W, and to
In Eq.
contributions, Si. In what follows, we will label the policy maker as ‘benevolent’ if m is close to zero, and as ‘greedy’ if m is large.Using Eqs. (1) and (2), the per capita compensating contribution functions take the following form:

Sa ¼ max 0;Ua − πaf g ¼ max 0;pα + λpβ−πa� �

; ð9Þ

Sb ¼ max 0;Ub−πbf g ¼ max 0;pβ + λpα−πb� �

: ð10Þ

ntribution function Si of lobby i is said to be truthful relative to the constant πi since, given the net welfare level πi (grosse Ui net of the contributions si paid), the change in the compensation offered to the policy maker for any change of policy isto the welfare change for lobby i.
equal
Since common agency games typically present multiple equilibria, truthful equilibria are a useful refinement. Bernheim andWhinston (1986b, Theorem 1) and Dixit et al. (1997, Proposition 2) show that the set of best reply strategies of a principal to thecontribution functions of the other principals (truthful or not) always contains a truthful contribution function. Moreover, truthfulequilibria are easier to characterize than other types of equilibria, since they can be computed by solving a set of simultaneousequations in terms of the equilibrium welfare numbers πi (see Proposition 3 in Dixit et al., 1997).

To solve our lobbying game, we make use of Corollary 1 to Proposition 4 in Dixit et al. (1997), that applies to cases (like ours) inwhich the utility functions of both the lobbies and the policy maker are linear in the contributions. In terms of our game, theCorollary states that, given a truthful equilibrium with policy choices pα

C and pβC and with positive contributions sa

C and sbC, the

equilibrium policy choices are obtained by maximizing

V ¼ W + mθ Ua + Ubð Þ ¼ θ + mθð Þ 1 + λð Þ pα + pβ� �

+ F pα;ωð Þ + F pβ;ω� �

: ð11Þ

, maximizing Eq. (11) is an easier task than maximizing Eq. (8), since the former ‘ignores’ the non-negativity constraints onbutions. We thus derive the optimal choices of the policy maker in the second stage of the lobbying game, assuming thatbutions are non-negative. We then characterize the equilibrium contributions and check for their non-negativity indix B, where we solve for the first stage of the game. By maximizing (11) with respect to pα and pβ, we obtain:

Proposition 2. Under Centralization, in the truthful equilibrium of the lobbying game, the policy choice in region r, r={α, β}, isprC=(θ+mθ) (1+λ)+ω.

(CL.1) If m=0, or if mN0 and λ=−1, then prC=pr⁎; social and groups' gross welfare are the same as in the utilitarian optimum.

(CL.2) If mN0 and λ≠−1 then prCNpr⁎; per capita social welfare isWCbW⁎ and groups' gross welfare is UI

CbUI⁎, I={A, B}, Ui

CNUi⁎, i={a, b}.

Proposition 2 shows that lobbies are in general able to induce an upward distortion in policy choices if the policy maker caresfor lobbyist's contributions (mN0). This increases lobbyists' welfare (gross of the contributions paid) but reduces aggregate socialwelfare as the gain for the lobbies does not compensate for the loss suffered by non-organized citizens. Note further that when theinterests of the two lobbies are totally conflictual (λ=−1), the two lobbies end up neutralizing each other, a well known result inthis literature. In this case, social welfare is the same as under a pure benevolent government, and lobbying just induces a transferof resources from the organized interest groups to the politician.7

5. Lobbying under decentralization

The game under decentralization is identical to the one considered in Section 4, except for the fact that the lobbies now face adistinct policy maker in each region. In stage one, the lobbies simultaneously and independently present a truthful contribution

e the gains from lobbying in terms of gross welfare may be smaller than the contributions paid to the policy makers, interest groups may end up beingff under lobbying than in the absence of lobbying.


schedule contingent on policy choices to each local policy maker, and then – in stage two – the policy makers simultaneously andindependently choose public policy in their own region, given the contribution schedules. To ensure comparability betweeninstitutional settings, we assume that the preferences of politicians are the same (i.e.m is the same) both at the decentralized andat the centralized level.8

As already noted above, while the lobbying game under centralization falls into the common agency framework (manyprincipals, one agent), the corresponding game under decentralization falls into the more general one of Games Played ThroughAgents, or GPTA (many principals, many agents). To deal with this case, we follow Prat and Rustichini (2003). In particular, weassume that the contribution from a lobby to a local policy maker is action-contingent in the sense that it depends only on thepolicy choices taken by that policy maker.9

In the second stage of the game, given the contribution function offered by the ‘home’ lobby, Sa,α, and that offered by the ‘foreign’lobby, Sb,α,10 the policy maker α, independently from the policy maker β, chooses pα to maximize her objective function

respecwelfartoken,

where

8 See9 Prat

actionsmore gethe utilithe acti10 In ddenotes11 In grestrictisubstituthe prin

Vα ¼ Wα + mθ Sa;α + Sb;α� �

: ð12Þ

oblem of policy maker β is defined in a similar way, so that she maximizes the objective function Vβ=Wβ+mθ(Sa,β+Sb,β).
The prUsing Eq. (1), the per capita compensating contribution functions offered by lobby a to the policy makers of regions α and β are
defined as

Sa;α ¼ max 0; pα−πa;α� �

; ð13Þ

Sa;β ¼ max 0;λpβ−πa;β� �

; ð14Þ

tively. Observe that the contribution (13) offered to the ‘home’ policy maker is truthful relative to the net welfare level (grosse minus contributions) achieved in region α, πa,α, and depends only on the policy chosen by that policy maker. By the samethe contribution (14) offered to the policy maker of the other region is truthful relative to the net welfare level achieved inβ, πa,β, and depends only on the policy chosen by that policy maker.11
region
In a similar manner, the per capita compensating contribution functions offered by lobby b to the policy makers of regions βand α are defined as

Sb;β ¼ max 0; pβ − πb;β� �

; ð15Þ

Sb;α ¼ max 0;λpα − πb;α� �

; ð16Þ

tively.
respecTo characterize the equilibrium policy choices in the second stage of the game, we proceed as in Section 4 by maximizing the
objective function of the policy maker, while ignoring the non-negativity constraints on contributions. In Appendix B, we thensolve for the first stage of the game and characterize the equilibrium contributions checking for non-negativity. Inwhat follows, wefocus on region α only, as the problem for region β is symmetric. Given the contributions (13) and (16), the problem of the policymaker of region α is that of choosing pα to maximize her objective function

Vα ¼ Wα + mθ 1 + λð Þpα ¼ θ + mθ 1 + λð Þ½ �pα + θλpβ + 1−σð ÞF pα;ωð Þ + σF pβ;ω� �

; ð17Þ

we have substituted Wα from Eq. (5). By solving this problem, we get:

Proposition 3. Under Decentralization, in the truthful equilibrium of the lobbying game, the policy choice in region r, r={α, β}, is prD=[θ+mθ(1+λ)]/(1−σ)+ω.

Let m⁎≡ [λ− (1+λ)σ]/(1+λ). The comparison between the decentralized policy choice and the social optimum is as follows:

(DL.1) If λN0, σ ∈ [0, σ⁎] and m=m⁎, or if λ=σ=m=0, then prD=pr⁎.

(DL.2) If λb0 and σ ∈ 0; 12� �

, or if λN0 and σ ∈ σ⁎; 12

� i, then pr

DNpr⁎ for m≥0.(DL.3) If λN0 and σ ∈ [0, σ⁎), then pr

Dbpr⁎ for m ∈ [0, m⁎), and prDNpr⁎ for m ∈ (m⁎, ∞).

(DL.4) If λ=σ=0, then prDNpr⁎ for mN0.

Seabright (1996) and Bardhan and Mookherjee (2000) for opposite arguments suggesting different values of m at the two levels.and Rustichini (2003) consider also the more general case of outcome-contingent contributions, in which the contribution to an agent depends also on thetaken by other agents. However, in the latter case no general characterization of the equilibrium outcomes of the game is available. Notice that our GPTA isneral than the one examined by Prat and Rustichini (2003), although their results fully apply to our framework as well. They consider the case in whichty function of each agent depends only on her actions. In our case, the utility (regional social welfare) of each agent (local policy maker) depends also onons (policies p) of the other agent, like in Segal (1999).enoting contributions under decentralization, the first subscript, a or b, denotes the group offering the contribution, whereas the second subscript, α or β,the policy maker to which the contribution is offered.eneral, as argued by Prat and Rustichini (2003, p. 1008), truthful contributions like those defined in common agency games “would impose too many equalityons on the transfer matrix” of the GPTA. They therefore introduce the weaker condition of weakly truthful contributions, that in terms of our model wouldte conditions (13)–(14)with the condition Sa,α+Sa,β≥max {0, pα+λpβ − πa}. In our setting, however, given the linearity and separability of the utility functions ofcipals, the games played in each region are independent of each other, which allows us to apply the condition of truthful contributions defined in Eqs. (13)–(14).


If prD=pr⁎, social and groups welfare are the same as in the utilitarian optimum. If pr

D≷pr⁎, per capita social welfare is WDbW⁎, groups'gross welfare is UI

D≶UI⁎,I={A, B}, Ui

D≷Ui⁎ if λ≠−1 and Ui

D=Ui⁎ if λ=−1, i={a, b}.

As in the case of centralization, under decentralization policies are an increasing function of the weight m that policy makersattach to lobbyists' contributions, unless λ=−1. However, there is an important difference. Under centralization, benevolent policymakers (those with m=0) always implement the utilitarian optimum, so that both the upward distortion in policy choices causedby lobbying activities and the associated social welfare loss increase with m. With decentralization, this is only true when thebenevolent local policymakers implement either the utilitarian optimum (pr⁎D=pr⁎), or policies above it (pr⁎DNpr⁎). In the other cases,lobbying plays a beneficial role, as it forces policy choices to be closer to the utilitarian optimum itself. Proposition 3 indeed showsthat, if λN0 and σ ∈ [0, σ⁎) so that pr⁎Dbpr⁎, as m increases prD gets closer to pr⁎ up to the point in which pr

D=pr⁎ for m=m⁎. As mfurther increases – i.e. politicians become relatively more greedy – social welfare declines and the social welfare loss reaches thesame value at m=2m⁎ as at m=0 by concavity and symmetry of the social welfare function. Finally, even in the case ofdecentralization, complete opposition of interests by lobbies (e.g. λ=−1) eliminates all the effects of lobbying on social welfare,although the equilibrium under decentralization would still be inefficient as long as σ≠0.

6. Centralization vs. decentralization

In order to compare social welfare under lobbying in the two institutional regimes, we need first to compare the correspondingpolicy choices with reference to the utilitarian optimum. This is done through Lemma 1 in Appendix A. Using the lemma, we canstate the following

Proposition 4. Let m^ uλ− 1 + λð Þσ1 + λð Þ 2−σð Þ and m̃u

λ− 1 + λð Þσ1 + λð Þσ :

• Social welfare under Centralization is higher than social welfare under Decentralization if λb0, σ ∈ 0; 12� �

and mN0; if λN0,σ ∈ σ⁎; 12

� �and mN0; if λN0, σ ∈ 0;σ⁎Þ½ and m ∈ (0, m̂), or m ∈ (m̃, ∞).

• Social welfare under Centralization is lower than social welfare under Decentralization if λN0, σ ∈ [0, σ⁎) and m ∈ (m̂, m̃).• The two regimes are equivalent if λN0, σ ∈ [0, σ⁎) and m=m̂; if λN0, σ ∈ (0, σ⁎) and m=m̃; if λ=σ=0 and mN0.

Differently from the case of benevolent governments, for which centralization dominates decentralization, Proposition 4 showsthat in the presence of lobbying no institutional setting unambiguously welfare dominates the other. On the one hand,centralization is a better regime when lobbies have conflicting interests (λb0), no matter the size of σ and policy makers'preferences. On the other hand, decentralization dominates centralization when lobbies interests are aligned (λN0), provided thatpolicy spillovers on non-organized groups are not too strong (σbσ ⁎), and that policy makers are neither very benevolent nor toogreedy (i.e. m̂bmbm̃); centralization still dominates decentralization in all other cases.

In the simpler case in which there are no spillovers on non-organized groups (σ=0) and policy makers are sufficiently greedy(mNm̂), the ranking in terms of social welfare only depends on the sign of the parameter λ, as it is shown in the following corollary:

Corollary 1. Assume σ=0 and mNm̂ . When the interests of lobbies are aligned (λN0), social welfare is higher under Decentralization thanunder Centralization.When lobbies interests are conflicting (λb0), social welfare is higher under Centralization than under Decentralization.

To understand the economic rationale of these results, note that the incentives of the policy makers come from two sources:social welfare (with weight 1) and lobbies' contributions (with weight m). As for the latter, given that the lobbies offer truthfulcontributions by assumption, the policy makers have always the right incentives to maximize the aggregate welfare of the twolobbies, no matter whether the lobbies have aligned or conflicting interests, and no matter whether the institutional setting iscentralized or decentralized. In particular, with aligned interests, lobbies have no incentives to free ride on each other's activity,since the actions are taken by the policy makers under the incentives provided by truthful incentive schemes. With conflictinginterests, the lobbies do not harm each other (in terms of gross welfare) since the policy makers, again by means of truthfulincentive schemes, fully internalize the aggregate benefits accruing to the lobbies from public policy. What makes the differencebetween the two regimes is thus the first component of the policy maker's objective function, i.e. social welfare. While undercentralization the policy maker fully internalizes (into the aggregate social welfare) the effects that the policy variables have on thetwo lobbying groups, under decentralization each local policy maker takes into account only the impact of policies on the residentlobby. With aligned interests, since policy in one region determines a positive externality on the non-resident organized group,under decentralization the policy instruments are set at lower levels than under centralization. Therefore, lobbying is less harmfulfor social welfare under decentralization; it is indeed beneficial as long as, for low levels ofm, it corrects for policy under-provision.The opposite is true with conflicting interests. In this case, the policy choice in one region determines a negative externality on thenon-resident organized group, and this implies that under decentralization the policy instruments are set at higher levels thanunder centralization, with the result that lobbying is more harmful for social welfare under decentralization.

Policy spillovers on non-organized groups (σ N0) add a further reason why decentralization may turn out to be welfaredominated by centralization even when lobbies' interests are aligned. This is because local policy makers cause larger distortionsthan those that would be caused by a central policy maker, as they do not internalize the negative externality imposed by theirpolicy choices on the non-organized citizens of the other region. Hence, for given λN0, centralization is a better regime in terms ofwelfare if policy spillovers on non-organized groups are sufficiently large (σ Nσ ⁎) and, provided that policymakers are very greedy(mNm̃), also when spillovers are weak (0bσbσ⁎).


7. Extensions

The results of the previous section suggest that the sign of regional spillovers matters in determining the relative merits ofcentralization and decentralization under lobbying behavior. In this section, we study the effect on the comparison betweencentralization and decentralization of more general assumptions about the structure of the model. We first consider the case inwhich regional lobbies have more difficulties in capturing the ‘foreign’ politician than the ‘home’ politician. Second, we consider asituation in which discriminated minorities use lobbying to voice their preferences. Finally, we discuss the case in which locallobbies have heterogeneous preferences on policies themselves, rather than just on their location.12 For the sake of simplicity, in allextensions we focus on the case in which there are no policy spillovers on non-organized groups; i.e. σ=0.

7.1. Heterogeneity in political capture under decentralization

In our basic model we assume that each group is equally effective when lobbying ‘domestically’ and when lobbying ‘abroad’.However, one could reasonably argue that lobbying of the home policymaker ismore effective than lobbying of the foreign one. Forinstance, local interests may have a better knowledge of their own cultural and political environment. Moreover, if contributionstake the form of (illegal) bribes, it might be easier to hide them when lobbying a home politician than a foreign one.

To capture this feature in a simple way, suppose that under decentralization one unit of contributions out of the lobbyists'pocket is worth one unit in the pocket of the ‘home’ politician but only ϕ units, ϕ ∈ [0, 1], in the pocket of the ‘foreign’ one. Undercentralization, instead, as there is a single policy maker, one unit of contributions is always worth one unit for the politician.

Under centralization lobbying behavior and policies choices are the same as in Section 4, with prC defined in Proposition 2.

Under decentralization, it is straightforward to extend the results in Proposition 3 to obtain that policy choices become

12 In tavailabl13 Obs

pDr ¼ θ + mθ 1 + λ/ð Þ + ω: ð18Þ

th respect to our previous analysis forϕ=1, a reduction inϕ decreasesprD ifλN0, and increases it ifλb0. If lobbying abroad becomes
Wiless effective, the impact onpolicy choices isweaker (resp. stronger)when the home and the foreign lobby have aligned (resp. conflicting)interests, since the foreign lobby is unable to reinforce (resp. countervail) the pressure for policy expansion by the home lobby.
The welfare effects stemming from the comparison between the two institutional regimes are then easily derived from theexpression for pr

D in Eq. (18) and from Proposition 4. On the one hand, if the lobbies have conflicting interests (λb0), so thatcentralization dominates decentralization (WC−WDN0) for all mN0, the welfare gap between the two regimes widens as ϕ getssmaller, since the foreign lobby becomes less able to offset the policy distortion caused by the home lobby. The same result holdstrue if the lobbies have aligned interests (λN0) and policy makers are sufficiently benevolent (mbm̂), since also in this casecentralization dominates decentralization.13 On the other hand, if the lobbies have aligned interests and policy makers aresufficiently greedy (mNm̂), so that decentralization dominates centralization (WD−WCN0), the impact of a reduction of ϕ on thewelfare gap is less clear-cut. In fact, the welfare gap between the two regimes diminishes as ϕ becomes smaller if prDbpr⁎bprC,whereas it increases if pr⁎bprDbprC, since the distance between pr

D and the social optimum level pr⁎ becomes larger in the former caseand smaller in the latter case. To summarize, a reduction in ϕ increases the welfare distortions under decentralization whencentralization is better than decentralization, whereas it may increase or reduce them depending on policy makers' preferenceswhen decentralization dominates centralization.

7.2. Policy makers with biased social welfare valuations

It is often argued that lobbying may be beneficial, besides of the reasons we focused on above, because it allows minorities tohave some ‘voice’ on a political market whichwould otherwise be dominated by thewishes of themajority. We add this element toour basic model, and investigate whether minorities would have a better chance under centralization or decentralization. Weassume that policy makers may under-represent the minority groups into the social welfare function. Specifically, let per capitasocial welfare in region r be defined asWr=θμUi+UI, where μ, μ∈ [0, 1], is the social welfare weight of the minority groups i. Noticethat, for μb1, the social welfare weight of the minority group falls short of their mass size θ. We say that policy makers are neutralwhen μ=1 and biased when μb1.

To avoid unnecessary complications, we assume that the lobbies do not have entirely conflicting objectives (i.e. λ≠−1), so thatin the utilitarian optimum policies are set above the bliss point of non-organized groups, i.e. pr⁎Nω (see Proposition 1).

When policy makers' are biased, by extending the results in Propositions 2 and 3, policy choices with lobbying undercentralization and decentralization become

pCr ¼ θμ + mθð Þ 1 + λð Þ + ω and pDr ¼ θμ + mθ 1 + λð Þ + ω; ð19Þtively. It is immediate from Eq. (19) that the distance in terms of policy provision, and thus also in terms of welfare levels,
respec
between centralization and decentralization diminishes when μ decreases, and it vanishes when μ=0. When lobbies have

he discussion below we omit all analytical details that are not strictly required for the illustration of the results. Formal propositions and proofs aree on request.erve that m̂ increases as ϕ becomes smaller.


aligned interests (λN0), the qualitative results of Proposition 4 for our basic model carry over to a situation in which μ ∈ (0, 1):centralization is better than decentralization if the policy makers are sufficiently benevolent (i.e.mb m̂), whereas the opposite istrue formN m̂. Instead, when lobbies have conflicting interests (λb0), the result in Proposition 4 that centralization is better thandecentralization for all m does not immediately extend to the case in which μ ∈ (0, 1). In particular, centralization turns out towelfare dominate decentralization for all m if and only if μ is sufficiently large, whereas for smaller μ decentralization is betterthan centralization whenever m is below a threshold level. The intuition for these results rests essentially on the relationshipbetween policy choices by biased but benevolent policy makers (i.e. μb1 and m=0) and the utilitarian optimum. Undercentralization, policy choices fall short of the utilitarian optimum since spillovers are internalized but the minority groups areunder-represented in the social welfare function. Under decentralization, the position of policy choices with respect to theutilitarian optimum results from the interplay between the effect of the under-representation of minorities (inducing policyunder-provision) and that of the non-internalization of the negative externalities on the non-resident lobby (determining policyover-provision). When minorities have little weight in social welfare (low μ) the first effect is such that policy distortions underdecentralization are smaller than those under centralization. Hence, if policymakers are sufficiently benevolent, decentralizationwelfare dominates centralization.

7.3. Lobbies with heterogeneous policy preferences

Our basic model captures only one dimension of the possible interregional spillovers of local policies on interest groups. Apolicy that is ‘good’ when implemented at home may be ‘good’ or ‘bad’ when implemented abroad, depending on the sign of theparameter λ. However, there are cases inwhich interregional spillovers depend also on the policy itself, rather than just onwhere itoriginates. As an example, suppose that firms supplying equipments to the railways industry – represented by group a – arelocated in region α, while firms supplying the aviation sector are located in region β and are represented by group b. Consider thenpolicies pα and pβ representing, respectively, public investments in the railways sector in regions α and β. It is then likely thatgroup a benefits (and group b is damaged) from higher levels of public investments in railways, no matter the region, α or β, inwhich the policy is implemented. This is a situation that is not captured by our basic model. It is thus interesting to study whetherthemore general framework allowing for both location and policy externalities, of which the example above is an illustration, leadsto different results in terms of the lobbies' influence under centralization and decentralization.

A simple way for extending our setting to encompass these cases is to assume that the utility functions (1)–(2) of the lobbyinggroups are no longer symmetric, but take the form:

whereexampwith rrespec

respecSectiocentramNm̂)whendominare hochoice

14 It iscontext15 In E

Ua ¼ 2−δð Þ pα + λpβ� �

; ð20Þ

Ub ¼ δ pβ + λpα� �

; ð21Þ

δ ∈ [−1, 1]. It is immediate to see that for δb0 and λN0 group a benefits whereas b is harmed from both pα and pβ, as in thele given above.14 In the extendedmodel (20)–(21), the sign of the parameter λ captures the alignment or conflict of interestsespect to the geographical source of policy, and that of the parameter δ captures the alignment or conflict of interests witht to the type of policy that is considered. If δ∈ (0,1], groups a and b have aligned interests on policies; if δ∈ [−1, 0), they haveting interests.
conflic
To understand how this affects our previous results, we focus on the cases in which lobbies' interests on policy are either fullyconflicting, i.e. δ=−1, or perfectly aligned, i.e. δ=1. By solving the second stage of the lobbying game using the utility functions (20)and (21), one obtains that policy choices under centralization and decentralization are

pCα ¼ θ + mθð Þ 2−δ + δλð Þ + ω; pCβ ¼ θ + mθð Þ δ + 2−δð Þλ½ � + ω; ð22Þ

pDα ¼ θ 2−δð Þ + mθ 2−δ + δλð Þ + ω; pCβ ¼ θδ + mθ δ + 2−δð Þλ½ � + ω; ð23Þ

tively.15 To compare the two institutional settings, note first that when δ=1 we revert to the situation already studied inn 6 (see Corollary 1), which suggests a neat welfare ranking of the two institutional systems depending on the sign of λ, withlization dominating decentralization for λb0, and vice versa for λN0 provided that politicians are sufficiently greedy (i.e.. No such a simple ranking is available for the case inwhich δ=−1. Centralization tends to welfare dominate decentralizationgeographical spillovers are positive and policy makers are sufficiently benevolent, whereas decentralization welfareates centralization in the opposite case inwhich geographical spillovers are negative andpolicymakers greedyenough. Therewevermany situations inwhich thewelfare ranking is ambiguous. Themain reasonwhy this occurs is that for δ=−1 the policys in the two regions induce externalities of opposite sign on the ‘foreign’ lobbying group, while externalities always have the

straightforward to extend the model to accommodate for two types of policies in each region, one valued by group a and one valued by group b. In theof our example, this amounts to consider regional investments not only in the railways sector but also in the aviation industry.qs. (22)–(23), to ensure interior solutions for m=0, we assume that ω is large enough (i.e. ωN4θ).

same sfrom p

16 From


ign for δ=1.16 Therefore, our results are quite sensitive onwhether the interregional spillovers between lobbies arise mainlyolicies themselves, or rather from their geographical source.

8. Concluding remarks

Is decentralization more conducive to lobbying? This paper offers a simple answer to this important question. When theinterests of the lobbying groups are aligned, provided that policy spillovers on non-organized groups are sufficiently weak andpoliticians are neither too greedy nor too benevolent, then decentralization is better than centralization because in the formerinstitutional setting local governments do not take into account the benefits of their policies on the non-resident lobbying groups,thus reducing the influence of lobbying. Vice versa, when the interests of local lobbies are in conflict, centralization is better thandecentralization because in the latter regime lobbying behavior is reinforced by policy over-provision by local governments thatfail to internalize the negative policy spillovers on non-resident groups. These results appear to be robust to more generalframeworks allowing for heterogeneity of political capture under decentralization, for the possibility of under-representation ofminorities in social welfare, and for a richer structure of policy spillovers.

Overall our findings suggest that, in deciding whether a given function should be decentralized (resp. centralized) in thepresence of significant lobbying, one should consider how the interests of local lobbies are positioned with respect to thatparticular function. For instance, considering the EU example cited in the Introduction, one notes that in fields such as consumersand environmental protection, foreign and domestic producers would have the same interests to lobby for low levels of protectionif these policies were decided at the EU level. Of course, theywould do the same if the policies remained at local level, but then eachcountry would have no interest to internalize the effects of these policies on the profits of foreign firms. Ceteris paribus, ourargument would then suggest to decentralize these functions. Vice versa, in regulatory fields such as production subsidies, or theprotection of the market shares of incumbents and ‘national champions’, national lobbies have conflicting interests, andcentralization at the EU level would force the policy maker to take into account also the interests hurt by a protectionist policy.Hence, ceteris paribus, our argument would suggest to centralize these functions.

Our model focuses on a single feature of the relationship between (de)centralization and lobbying, and we make no claim thatthis is the only possible channel. The literature surveyed in the Introduction discusses alternative models of lobbying, with a richerpolitical structure, which may produce different results. In particular, lobbying may provide useful information to politicians andcitizens, and as better information on policies and politicians is often quoted as one of themain advantages of decentralization (e.g.Bennedsen and Feldmann, 2002, 2006; Besley and Smart, 2007; Bordignon et al., 2004), discussing the link between informationallobbying and decentralization may offer further useful insights.

Acknowledgments

Previous versions of this paper have been presented at the European and North American Meetings of the Econometric Societyin 2003, at the IIPF Conference in 2003, at the Public Economic Theory Conference in 2004, at the IFIR-CESifo Conference inLexington – Kentucky in 2006, as well as in seminars at Barcelona, Berlin, Bielefeld, Exeter, Ferrara, Munich, Osaka, Pavia, Roma,Venezia. The authors are grateful to seminars and conferences participants for providing useful comments, as well as to twoanonymous referees and a co-editor. Financial support from MIUR is gratefully acknowledged.

Appendix A. Proofs

Proof of Proposition 1. The utilitarian optimum is characterized by maximizingW in Eq. (7). From the first order condition withrespect to pr,

AWApr

¼ θ 1 + λð Þ + ω−pr ¼ 0;

tains pr⁎. Under centralization, a benevolent policy maker maximizes W in Eq. (7) and therefore implements the utilitarian
one oboptimum. Under decentralization, the policy maker of region r sets pr that maximizes Wr in Eqs. (5)–(6). From the first ordercondition,
AWr

Apr¼ θ + 1−σð Þ ω−prð Þ ¼ 0;

ts pr⁎D. By comparing pr⁎ and pr⁎

D one obtains that pr⁎⋛pr⁎D if (1+λ)⋛1/(1− σ). Since 1/(1− σ)≥1 for σ ∈ 0; 12

� �, it follows� �
one ge
that if λb0 then pr⁎bpr⁎D for all σ ∈ 0; 12 . When λN0 then pr⁎⋚pr⁎

D if σ⋛λ/(1+λ)≡σ⁎. If pr⁎D≷pr⁎ it is W⁎DbW⁎ since (pα⁎, pβ⁎)=arg maxpα,pβ

W. If pr⁎D≷ pr

⁎ it is U I⁎D≶U I

⁎ since pr⁎DNω, pr

⁎Nω, with (ω, ω) = arg maxpα,pβUI, UI strictly concave. If pr

⁎D≷pr⁎it is U i⁎

D≷Ui⁎ for λ≠−1 and U i

⁎D=U i⁎ for λ=−1, since Ui= (1+λ)p for pα= pβ= p, and pα⁎

D= pβ⁎D, pα⁎= pβ⁎. □

the utility functions (20)–(21) it is immediate to see that∂Ua/∂pβ=(2 −δ)λ and∂Ub/∂pα=δλ. Thus, forδ=−1,∂Ua/∂pβ≷0 and∂Ub/∂pα≶0 dependingonwhetherλ≷0.


Proof of Proposition 2. The expression for prC follows from the maximization of Eq. (11) with respect to pr, which gives the firstorder condition:

AVApr

¼ θ + mθð Þ 1 + λð Þ + ω−pr ¼ 0:

m the definition of pr⁎ in Proposition 1, one can see that prC=pr⁎ for m=0 or mN0 and λ=−1, and prCNpr⁎ for mN0 and λ≠−1.
Fro
Therefore, for mN0 and λ≠−1 one has (i) WCbW⁎ since (pα⁎, pβ⁎)=arg maxpα,pβW; (ii) UI

CbU I⁎ since pr

CNpr⁎Nω, with (ω, ω)=arg maxpα,pβ

UI and UI strictly concave; and (iii) UiCNU i⁎ since Ui is monotonically increasing in p for pα=pβ=p and pα

C =pβC ,pα⁎=pβ⁎. □

Proof of Proposition 3. By differentiating Vα in Eq. (17) with respect to pα, one has the first order condition:

AVa

Apα¼ θ + mθ 1 + λð Þ + 1−σð Þ ω−pαð Þ ¼ 0;

hich one obtains prD. As for the comparison of the decentralized outcome and the utilitarian optimum, observe that
from w
prD=pr⁎D for m=0, with pr⁎

D defined as in Proposition 1, and that prD is monotonically non-decreasing in m. Let m⁎ be the uniquevalue of m such that prD=pr⁎, and recall that the conditions under which pr⁎

D⋛pr⁎ are stated in Proposition 1. Case (DL.1) thenfollows directly from (D.1) in Proposition 1, and from the definition of m⁎ and (D.3) in the same proposition. Similarly, case (DL.2)results from (D.2), and case (DL.3) from (D.3), in Proposition 1. Finally, (DL.4) follows immediately from (D.1) in Proposition 1.

If prD≷pr⁎ it is (i) WDbW⁎ since (pα⁎, pβ⁎)=arg maxpα , pβ

W; (ii) UID≶U I

⁎ since prDNpr⁎Nω if pr

DNpr⁎, and pr⁎NprDNω if pr

Db

pr⁎, where (ω,ω)=arg maxpα ,pβ UI and UI strictly concave; (iii) UiD≷U i

⁎ for λ≠−1 since ∂Ui/∂pN0 and UiD=U i

⁎ for λ=−1 since ∂Ui/∂p=0,given that pα=pβ=p and pα

D=pβD, pα⁎=pβ⁎. □The following Lemma is instrumental in proving Proposition 4.

Lemma 1. Let m̂uλ− 1 + λð Þσ1 + λð Þ 2−σð Þ ;m⁎u

λ− 1 + λð Þσ1 + λð Þ and m̃u

λ− 1 + λð Þσ1 + λð Þσ .

(CD.1) If λb0 and, σ ∈ 0; 12� �

or if λN0 and σ ∈ σ⁎; 12� �

, then prDNpr

CNpr⁎ for mN0.(CD.2) If λN0 and σ∈ (0, σ⁎), then

− pDr < p⁎r < pCr ; pCr −p⁎r < p⁎r −pDr for m ∈ 0; m̂

� ;

− pDr < p⁎r < pCr ; pCr −p⁎r ¼ p⁎r −pDr for m ¼ m̂;

− pDr < p⁎r < pCr ; pCr −p⁎r > p⁎r −pDr for m ∈ m̂;m⁎

� ;

− pDr ¼ p⁎r < pCr for m ¼ m⁎;

− p⁎r < pDr < pCr for m ∈ m⁎; m̃�

;

− p⁎r < pDr ¼ pCr for m ¼ m̃;

− p⁎r < pCr < pDr for m ∈ m̃;∞�

:

) If λN0 and σ=0, then
(CD.3
− pDr < p⁎r < pCr ; pCr −p⁎r < p⁎r −p

Dr for m ∈ 0; m̂

� ;

− pDr < p⁎r < pCr ; pCr −p⁎r ¼ p⁎r −pDr for m ¼ m̂;

− pDr < p⁎r < pCr ; pCr −p⁎r > p⁎r −p

Dr for m ∈ m̂;m⁎

� ;

− pDr ¼ p⁎r < pCr for m ¼ m⁎;

− p⁎r < pDr < pCr for m ∈ m⁎;∞�

:

If λ=σ=0, then prD=pr

CNpr⁎ for all mN0.
(CD.4)
Proof. Case (CD.1) follows from (D.2) in Proposition 1 given that for the values of λ and σ considered in (CD.1) it is pr⁎DNpr⁎C, and

∂prD/∂m≥∂prC/∂m≥0 for all mN0. Case (CD.2) results from (D.3) in Proposition 1 and (DL.3) in Proposition 3, once one observes thatfor λN0 and σ∈ (0, σ⁎) it is pr⁎CNpr⁎D, and 0b∂prC/∂mb∂prD/∂m. Therefore, there exist a unique value m̂ such that prC− pr⁎=pr⁎− pr

D, and aunique value m̃ such that prC=prD. Together with m⁎ defined in Proposition 3, these values determine the ranking between pr

C, prD

and pr⁎. Case (CD.3) follows again from (D.3) in Proposition 1 and (DL.3) in Proposition 3, when considering that for λN0 and σ=0 itis pr⁎

CNpr⁎

D and 0b∂prC/∂m=∂prD/∂m, so that there exist two thresholds m̂ andm⁎ as defined above permitting to rank prC, prD and pr⁎.

Finally, case (CD.4) results immediately from (D.1) in Proposition 1. □


Proof of Proposition 4. Recalling that (pα⁎, pβ⁎)=arg maxpα, pβW and thatW is strictly concave and symmetric with respect to pr⁎,

if |prC− pr⁎|≥ |prD− pr⁎|≥0 then WC≤WD≤W⁎; while, if |prD−pr⁎|≥ |prC− pr⁎|≥0 then WD≤WC≤W⁎. The results in the proposition thenfollow immediately from Lemma 1.

Appendix B. Contributions

In this appendix we solve the first stage of the lobbying game by making use of Proposition 3 in Dixit et al. (1997).

B.1. Centralization

Consider the case of policy centralization. In the first stage of the game, the strategy space of lobby i is given by the per capitanet welfare level πi that determines the ‘position’ of its truthful contribution function Si in the policy space. Therefore, we solve forthe Nash equilibrium in the per capita net welfare levels πa and πb. Consider, without loss of generality, lobby b. Its best responsefunction to any given strategy πa chosen by lobby a is obtained by solving

VC−mθ πa + πbð Þ¼VC að Þ−mθπa; if VC að Þ−mθπa≥WC nð Þ; ðB:1Þ

VC−mθ πa + πbð Þ¼WC nð Þ; otherwise: ðB:2Þ

is Eq. (11) evaluated at the equilibrium policies prC and qrC chosen in the second stage of the game by the policy maker under
VC
the influence of both lobbies offering the truthful contributions (9)–(10); see Proposition 2. VC ðaÞ is defined in a similar way for thecase in which only group a is lobbying by offering the truthful contribution (9), with group b abstaining from lobbying. Eqs. (B.1)and (B.2) show that the maximum amount of per capita net welfare πb that lobby b can get is constrained by the policy maker'sparticipation constraint. Given πa, if the policy maker is better off under lobbying by group a than under no lobbying (i.e. if VC(a)−mθπa≥WC(n), whereWC(n) denotes social welfare in the absence of lobbying), then Eq. (B.1) states that group b can increase its percapita net welfare πb (and correspondingly reduce contributions) up to the point in which the policy maker is indifferent betweenbeing lobbied by both groups and being lobbied only by group a. Otherwise, if the policymaker is better off under no lobbying thanunder lobbying by group a, then Eq. (B.2) states that group b can increase πb up to the point inwhich the policy maker is indifferentbetween being lobbied by both groups and not being lobbied.

From Eqs. (B.1) and (B.2), the best response function of lobby b can be written as

πb πað Þ¼VC−VC að Þ

mθ; if πa V

VC að Þ−WC nð Þ

mθ;

VC−WC nð Þ

mθ− πa; otherwise:

8>><>>:

ðB:3Þ

symmetry, VC(a)=VC(b), so that the best response function of lobby a is:
By
πa πbð Þ¼VC−VC bð Þ

mθ; if πb V

VC bð Þ−WC nð Þ

mθ;

VC−WC nð Þ

mθ− πb; otherwise:

8>><>>:

ðB:4Þ

e first stage of the lobby game under centralization is characterized in the following:
Th
Proposition 5. Under Centralization, in the first stage of the lobbying game, the Nash equilibrium in the per capita net welfare levels is:

– If VC ≤2VC (i)−WC(n), then πCi ¼

VC−VC ið Þmθ .

– If VC N2VC (i)−WC(n), then πCa + πC

b¼VC−WC nð Þ

mθ ; VC ið Þ−WC nð Þ

mθ V πCa ;π

Cb V VC−VC ið Þ

mθ :

Equilibrium contributions siC=Ui

C−πiC are non-negative.

Proof. By the symmetry of the best response functions (B.3) and (B.4), there is a unique Nash equilibrium, πCi ¼

VC−VC ið Þmθ ; if VC−VC ið Þ

mθ ≤VC ið Þ−WC nð Þ

mθ , i.e. if VC ≤2V C(i)−WC(n). The non-negativity of the equilibrium per capita contributions, siC=UiC−πiC, requires that VC−VC ið Þ

mθ ≤UCi .

Observing that VC=WC+2mθUiC and VC (i)=WC(i)+mθUi

C(i), we can write WC+mθUiC≤WC(i)+mθUi

C(i). The latter inequality holds true asWC(i)+mθUi

C(i)=arg maxpα ,pβW+mθUi.By inspection of the best response functions (B.3) and (B.4), it also follows that if VC−VC ið Þ

mθ ≥ VC ið Þ−WC nð Þmθ (i.e. V C≥2V C(i)−WC(n)) there is a

multiplicity of Nash equilibria: πCa + πC

b¼VC−WC nð Þ

mθ ; VC ið Þ−WC nð Þmθ ≤ πC

a ;πCb ≤

VC−VC ið Þmθ . As for the non-negativity of contributions, given the

symmetry of equilibria it is enough to show that they are non-negative in the equilibrium in which the net welfare of one lobby ismaximumand thatof theother isminimum, i.e. V

C−VC ið Þmθ ≤ UC

i and VC ið Þ−WC nð Þmθ ≤ UC

i .Wehave shownabove that thefirst inequalityholds. As forthe second one, it can bewritten asWC(i)+mθUi

C(i)≤mθUiC, that combinedwithWC+mθUi

C≤WC(i)+mθUiC(i) implies thatWC≤W C(n), which

is true since under centralization social welfare is maximized at the policy levels in the absence of lobbying (i.e.WC(n)=W⁎). □


B.2. Decentralization

Under decentralization, the first stage of the lobbying game is solved in a similar way. Consider, without loss ofgenerality, region α. In the first stage of the lobbying game, we solve for the Nash equilibrium in the per capita net welfarelevels πa,α and πb,α. Consider lobby b. Its best response function to any given strategy πa,α chosen by lobby a is obtained bysolving

VDα−mθ πa;α + πb;α

� �¼VD að Þ

α −mθπa;α ; if VD að Þα −mθπa;α �WD nð Þ

α ; ðB:5Þ

VDα−mθ πa;α + πb;α

� �¼WD nð Þ

α ; otherwise: ðB:6Þ

is Eq. (17) evaluated at the equilibrium policies prD and qrD chosen in the second stage of the game by the local policy makers
Vα
D

under the influence of both lobbies offering the truthful contributions (13)–(16); see Proposition 3. VαD(a) is defined in a similar way

for the case in which only group a is lobbying by offering the truthful contribution (13), with group b abstaining from lobbying.Finally,Wα

D(n) is social welfare in region αwhen there is no lobbying in region α; there can instead be no lobbying, lobbying by onegroup only, or lobbying by both groups, in region β; it is immediate to show that the best response functions (B.7) and (B.8) beloware independent of what goes on in region β. Eqs. (B.5) and (B.6) show that themaximum amount of per capita net welfare πb,α thatlobby b can get in region α is constrained by the participation constraint of the policy maker of region α.

From Eqs. (B.5) and (B.6), the best response function of lobby b can be written as

πb;α πa;α� �

¼

VDα−VD að Þ

α

mθ; if πa;α V

VD að Þα −WD nð Þ

α

mθ;

VDα−W

D nð Þα

mθ−πa;α ; otherwise:

8>><>>:

ðB:7Þ

e best response function of lobby a is obtained in a similar way:
Th
πa;α πb;α� �

¼VDα−VD bð Þ

α

mθ; if πb;α V

VD bð Þα −WD nð Þ

α

mθ;

VDα−W

D nð Þα

mθ−πb;α ; otherwise:

8>><>>:

ðB:8Þ

one can see from the best response functions (B.7) and (B.8), the lobbying game in regionα is not symmetric, sinceVαD(a)≠ Vα

D(b)
Asfor λ≠0. In region α, the first stage of the game under decentralization is characterized by the following:
Proposition 6. Under Decentralization, in region α, in the first stage of the lobbying game, the Nash equilibrium in the per capita netwelfare levels is:

– If VαD≤Vα

D(a)+VαD(b)−Wα

D(n), then πDa;α ¼ VD

α−VD bð Þα

mθ ; πDb;α¼

VDα−VD að Þ

αmθ :

– If VαDNVα

D(a)+VαD(b)−Wα

D(n), then πDa;α + πD

b;α¼VDα−W

D nð Þα

mθ ;VD að Þα −WD nð Þ

αmθ V πD

a;α V VDα−VD bð Þ

αmθ ;


αmθ V πD

b;α V VDα−VD að Þ

αmθ :

Equilibrium contributions si,αD =Ui,α

D −πi,αD , i={a, b} are non-negative.

Proof. From the best response functions (B.7) and (B.8), there is a unique Nash equilibrium, πDa;α¼

VDα−VD bð Þ

αmθ ; πD

b;α¼VDα−VD að Þ

αmθ

if VDα V VD að Þ

α + VD bð Þα −WD nð Þ

α . Equilibrium per capita contributions, siαD=Ui,αD −πi,αD , are non-negative if VD

α−VD bð Þα

mθ VUDa;α and VD

α−V D að Þα

mθ VUDb;α. The

first inequality implies that WαD+mθUb,α

D ≤WαD(b) +mθUb,α

D(b), that holds true as WαD(b)+mθUb,α

D(b)=arg maxpα Wα+mθUb,α. Analogously,the second inequality implies Wα

D+mθUa,αD ≤Wα

D(a)+mθUa,αD(a), which is satisfied since Wα

D(a)+mθUa,αD(a) =arg maxpα

Wα+mθUa,α.When Vα

DNVα

D(a)+VαD(b)−Wα

D(n), it follows immediately by inspection of the best response functions (B.7) and (B.8) that thereare multiple equilibria: πD

a;α + πDb;α¼

VDα−W

D nð Þα

mθ ;VD að Þα −WD nð Þ

αmθ V πD

a;α V VDα−VD bð Þ

αmθ ;


αmθ V πD

b;α V VDα−VD að Þ

αmθ . To show that contributions are non-

negative in all equilibria, it is sufficient to show that they are non-negative in the equilibrium inwhich the net welfare of lobby a ismaximum and that of b is minimum (i.e. V

Dα−VD bð Þ

αmθ VUD

a;α and VD bð Þα −WD nð Þ

αmθ VUD

b;α), as well as in the equilibrium in which the net welfare ofa is minimum and that of b is maximum (i.e. V

D að Þα −WD nð Þ

αmθ VUD

a;α and VDα−VD að Þ

αmθ VUD

b;α). We have already shown above that the first inequalityfor each of the two equilibria is satisfied. As for the second inequality of the first equilibrium, it implies that Wα

D(b) + mθUb,αD(b)≤

WαD(n) +mθUb,α

D , which combined with WαD+mθUb,α

D ≤WαD(b) +mθUb,α

D(b) implies that WαD≤Wα

D(n). The latter inequality is true sincesocial welfare is always greater in the absence of lobbying than when both lobbies are active. Analogously, for the secondinequality of the second equilibrium: Wα

D(a) + mθUa,αD(a) ≤ Wα

D(n) + mθUa,αD , that combined with Wα

D+mθUa,αD ≤ WαD(a) +mθUa,α

D(a)

gives WαD≤Wα

D(n). □

References

Alesina, A., Angeloni, I., Schuknecht, L., 2005. What does the European Union do? Public Choice 123, 275–319.Bardhan, P., Mookherjee, D., 2000. Capture and governance at local and national levels. American Economic Review 90 (2), 135–139.


Bardhan, P., Mookherjee, D., 2002. Relative capture of local and central governments: an essay in the political economy of decentralization. UC Berkeley WorkingPaper. Center for International and Development Economics Research, vol. 1013.

Bennedsen, M., Feldmann, S.E., 2002. Lobbying legislatures. Journal of Political Economy 110, 919–946.Bennedsen, M., Feldmann, S.E., 2006. Informational lobbying and political contributions. Journal of Public Economics 90, 631–656.Bernheim, B.D., Whinston, M.D., 1986a. Common agency. Econometrica 54, 923–942.Bernheim, B.D., Whinston, M.D., 1986b. Menu auctions, resource allocation, and economic influence. Quarterly Journal of Economics 101, 1–31.Besley, T., Coate, S., 2003. Centralized versus decentralized provision of local public goods: a political economy analysis. Journal of Public Economics 87, 2611–2637.Besley, T., Smart, M., 2007. Fiscal restraints and voter welfare. Journal of Public Economics 91, 755–773.Bordignon, M., Cerniglia, F., Revelli, F., 2004. Yardstick competition in intergovernmental relationships: theory and empirical predictions. Economics Letters 83,

325–333.Brou, D., Ruta, M., 2006. Special interests and the gains from political integration. Economics and Politics 18, 191–218.De Melo, J., Panagariya, A., Rodrik, D., 1993. The new regionalism: a country perspective. In: de Melo, J., Panagariya, A. (Eds.), New dimensions in Regional

Integration. Cambridge University Press, Cambridge.Dixit, A., Grossman, G.M., Helpman, E., 1997. Common agency and coordination: general theory and application to government policy making. Journal of Political

Economy 105, 752–769.Felli, L., Merlo, A., 2006. Endogenous lobbying. Journal of the European Economic Association 4, 180–215.Fisman, R., Gatti, R., 2002. Decentralization and corruption: evidence across countries. Journal of Public Economics 83, 325–345.Grossman, G.M., Helpman, E., 1994. Protection for sale. American Economic Review 84, 833–850.Grossman, G.M., Helpman, E., 2001. Special Interest Politics. The MIT Press, Cambridge, MA.Keen, M.J., Kotsogiannis, C., 2002. Does federalism lead to excessively high taxes? American Economic Review 92, 363–370.Lockwood, B., 2002. Distributive politics and the costs of centralization. Review of Economic Studies 69, 313–337.Mitra, D., 1999. Endogenous lobbies formation and endogenous protection: a long run model of trade determination. American Economic Review 89, 1116–1134.Oates, W., 1972. Fiscal Federalism. New York: Harcourt Brace and Jovanovich.Persson, T., 1998. Economic policy and special interest politics. Economic Journal 108, 310–327.Persson, T., Tabellini, G., 2000. Political Economics. The MIT Press, Cambridge, MA.Prat, A., Rustichini, A., 2003. Games played through agents. Econometrica 71, 989–1026.Prud'homme, R., 1994. On the dangers of decentralization. World Bank Policy Research WP, vol.1252.Redoano, M., 2002. Does centralization affect the number and size of lobbies? mimeo.Seabright, P., 1996. Accountability and decentralization in government: an incomplete contracts model. European Economic Review 40, 61–89.Segal, I., 1999. Contracting with externalities. Quarterly Journal of Economics 114, 337–388.Spiller, P.T., Liao, S., 2006. Buy, lobby or sue: Interest groups' participation in policy making — a selective survey. NBER Working Paper, vol. 12209.Tabellini, G., 2003. Principles of policy making in the European Union: an economic perspective. CESifo Economic Studies 49, 75–102.Tabellini, G., Wyplosz, C., 2006. Supply-side policy coordination in the European Union. Swedish Economic Policy Review 13, 101–156.Treisman, D., 2000. The causes of corruption: a cross-national study. Journal of Public Economics 76, 399–457.Treisman, D., 2006. Fiscal decentralization, governance, and economic performance. Economics and Politics 18, 219–235.Wilson, J.D., 1999. Theories of tax competition. National Tax Journal 52, 269–304.

fiscal federalism and lobbying

Documents