This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Transcription

1 This rticle ppered in journl published by Elsevier. The ttched copy is furnished to the uthor for internl non-commercil reserch nd eduction use, including for instruction t the uthors institution nd shring with collegues. Other uses, including reproduction nd distribution, or selling or licensing copies, or posting to personl, institutionl or third prty websites re prohibited. In most cses uthors re permitted to post their version of the rticle (e.g. in Word or Tex form) to their personl website or institutionl repository. Authors requiring further informtion regrding Elsevier s rchiving nd mnuscript policies re encourged to visit:

2 Journl of Public Economics 92 (2008) The effects of tx competition when politicins crete rents to buy politicl support Wolfgng Eggert,c, Peter Birch Sørensen b,c,d, Fkultät für Wirtschftswissenschften, Universitität Pderborn, Wrburger Strsse 100, Pderborn, Germny b Deprtment of Economics, University of Copenhgen, Studiestrede 6, 1455 Copenhgen K, Denmrk c CESifo, Germny d EPRU, Denmrk Received 9 Mrch 2007; received in revised form 27 November 2007; ccepted 10 December 2007 Avilble online 17 December 2007 Abstrct We set up probbilistic voting model to explore the hypothesis tht tx competition improves public sector efficiency nd socil welfre. In the bsence of tx bse mobility, distortions in the politicl process induce vote-mximising politicins to crete rents to public sector employees. Allowing tx bse mobility my be welfre-enhncing up to point, becuse the ensuing tx competition will reduce rents. However, if tx competition is crried too fr, it will reduce welfre by cusing n underprovision of public goods. Strting from n equilibrium where tx competition hs eliminted ll rents, coordinted rise in cpitl txtion will lwys be welfre-improving. For plusible prmeter vlues it will even be welfre-enhncing to crry tx coordintion beyond the point where rents to public sector workers strt to emerge Elsevier B.V. All rights reserved. JEL clssifiction: D72; H73; H87 Keywords: Tx competition; Rent-seeking; Probbilistic voting 1. Tx competition nd Levithn The globlistion of economic ctivity hs shrpened the interntionl debte on the costs nd benefits of tx competition. Critics rgue tht such competition will led to n underprovision of public goods s governments undercut ech others' tx rtes in n ttempt to ttrct mobile tx bses. The theoreticl foundtions for this view were lid by Otes (1972), Zodrow nd Mieszkowski (1986), Wilson (1986), Wildsin (1989) nd Jneb nd Peters (1999), mong others. In the opposite cmp it is rgued tht tx competition helps to reduce government wste nd to discipline rent-seeking politicins nd bureucrts. According to Public Choice theorists such s Brennn nd Buchnn (1977, Corresponding uthor. Deprtment of Economics, University of Copenhgen, Studiestrede 6, 1455 Copenhgen K, Denmrk. E-mil ddresses: Wolfgng.Eggert@uni-pderborn.de (W. Eggert), Peter.Birch.Sorensen@econ.ku.dk (P.B. Sørensen) /$ - see front mtter 2007 Elsevier B.V. All rights reserved. doi: /j.jpubeco

3 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) ), government is n ever-expnding Levithn tht needs to be tmed, nd one wy of strving the best is to llow interjurisdictionl competition for mobile tx bses. 1 This scepticl view of government which welcomes tx competition seems to hve gined ground in recent yers. A populr version of the rgument tht tx competition increses public sector efficiency ssumes tht self-serving politicins nd bureucrts re somehow ble to divert the txpyer's money wy from uses tht benefit the generl public nd into uses tht re pure wste from society's viewpoint. It is then rgued tht tx competition hmpers this diversion of resources wy from beneficil public use, since growing mobility of tx bses rises the mrginl cost of public funds, thereby hrdening voter resistnce to government wste. For exmple, the World Bnk (2004, p. 53) vividly rgues tht decentrlistion permits degree of institutionl competition between centers of uthority tht cn reduce the risk tht governments will exproprite welth. Edwrds nd Keen (1996) ttempted to synthesize the conflicting views on tx competition. In their nlysis politicins mximise n objective function of the form V(R,U), where R is the rent pproprited by the politicins themselves (which is modelled s pure wste), nd U is the welfre of the representtive citizen. Thus politicins trde off the interests of voters ginst rents to themselves. Combining this objective function with stndrd model of tx competition where the mrginl source of public funds is source-bsed tx on mobile cpitl, Edwrds nd Keen demonstrted tht tx competition will hve two offsetting effects on consumer welfre. On the one hnd it will tend to rise welfre by reducing the volume of rents pproprited by politicins. On the other hnd it will tend to cuse n underprovision of public goods by rising the mrginl cost of public funds. On blnce, Edwrds nd Keen found tht if the elsticity of the tx bse with respect to the tx rte is lower thn the politicins' mrginl propensity to spend public funds on wste, tx competition will be preferble to tx coordintion, nd vice vers. Severl other uthors including Otes nd Schwb (1988), Fuest (2000), Ruscher (2000), Eggert (2001), Keen nd Kotsoginnis (2003), Sto (2003), Wilson nd Gordon (2003) nd Wilson (2005) hve nlysed the effects of tx competition in Levithn models where policy mkers pproprite prt of the tx revenue for their own purposes. Jneb nd Schjelderup (2002) hve studied how tx competition ffects the bility of politicins to pproprite rents under lterntive politicl institutions, nd Besley nd Smrt (2007) hve investigted the effects of vrious constrints on fiscl policy (including tx competition) when imperfectly informed voters fce the chllenge of distinguishing Levithn-type politicins from benevolent politicl cndidtes. Wheres these contributions hve tended to find tht tx competition my ply n efficiency-enhncing role, Ci nd Treismn (2005) show in Levithn-type model with symmetric regions tht tx competition my ctully generte more government wste in poorly endowed regions. While ll of these studies hve generted vluble insights, the positive nd normtive nlysis of public policy in trditionl Levithn models hs severl problemtic fetures. First, the modelling of rents s pure wste goes ginst the fundmentl normtive principle tht the welfre of ll citizens (including rent-seekers) should be llowed to count in the socil welfre function. Second, in Levithn models rent cretion typiclly reduces politicl support for the policy mker becuse rents re chieved t the expense of the welfre of voters. This my be resonble wy of modelling the kind of rent-seeking tht tkes the form of unnecessry nd wsteful luxury for top government officils, but in most western democrcies this type of rent is probbly of minor quntittive importnce reltive to totl income. Insted it ppers tht rents re typiclly creted with the purpose of obtining politicl support from the recipients. Thus, wheres the Levithn literture ssumes tht rent cretion lwys reduces the policy mker's politicl bcking, it seems more relistic to ssume tht rents re generted becuse they increse the likelihood tht those responsible for creting them will remin in government office. 2 Third, vrible such s the frction of public revenue tht is wsted which plys crucil role in the Levithn literture is not very opertionl from n empiricl perspective. The concept of government wste is very subjective; wht seems wste to one person my be useful government ctivity in the eyes of nother person. To be ble to subject politicl economy models of tx competition to empiricl 1 This rgument for tx competition is very different from those offered in the clssicl contributions by Hyek (1939) nd Tiebout (1956). Tiebout did not ssume Levithn government, but rgued tht fiscl competition would llow mobile households to locte in those jurisdictions tht offered their most preferred level nd mix of public services. 2 Ansolbehere nd Snyder (2006) provide evidence from the United Sttes tht governing politicl prties t the stte level do in fct skew the distribution of public funds in fvour of res tht provide them with the strongest electorl support.

4 1144 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) testing, it seems desirble to develop mesures of politicl distortions tht re more objective nd hence esier to identify empiriclly. 3 In this pper we present politicl economy frmework llowing n nlysis of the effects of tx competition nd tx coordintion on rent-seeking nd socil welfre in setting where rents re creted s prt of politicl strtegy to mximise the probbility of winning the election. Insted of considering rents s pure wste, we thus tret them s mens of redistributing income in fvour of politiclly influentil groups. Our model llows for politicl distortion in fvour of public sector workers, sy, due to the existence of strong public sector trde unions. As n empiricl mtter, we do not ctully postulte tht public sector employees lwys hve disproportionte influence on the politicl process, but we explore the implictions of this ssumption becuse it is implicit in the resoning of mny dvoctes of tx competition. Our study offers synthesis of the trditionl Public Finnce view of tx competition nd the view of the Public Choice school by embedding probbilistic voting model of the type proposed by Persson nd Tbellini (2000, ch. 3) in tx competition model similr to the one used by Edwrds nd Keen. 4 In our model the often vgue concept of politicl distortion hs very precise mening. Our indictor of the degree of politicl distortion depends on the size of the public sector lobby nd on the reltive politicl influence of n individul lobby member, mesured by the derivtives of the voting function mximised by politicins. The greter the sensitivity of voting behviour to chnge in economic benefits offered to lobby member, the greter is his politicl influence reltive to the influence of voter outside the lobby. 5 In our frmework tx competition is powerful institutionl device which my completely wipe out rents to public sector workers, but only t the cost of n underprovision of public goods. Unlike the Levithn literture, this pper offers theory of the politicl trnsmission mechnism through which tx competition leds to reduced rent cretion s well s reduced public goods provision. In our set up politicins my cpture more votes from public sector workers by pying them higher wges, nd they my lso ttrct votes by creting dditionl high-pying public sector jobs. At the sme time politicins my gin votes by offering higher privte consumption opportunities through lower txes. In politicl equilibrium, politicl cndidtes strike blnce between these competing wys of gining votes, ccounting for the government budget constrint. When tx competition is llowed, the mount of privte consumption tht must be scrificed to rise unit of tx revenue goes up, so finncing the public sector wge bill becomes costlier in economic nd politicl terms. The vote-mximising politicl strtegy will then involve lower rents to public sector workers nd fewer public sector jobs (nd hence less public service provision). A min point in this explntion of the impct of tx competition on the politicl equilibrium is tht the public sector workers erning rents re prt of the voting popultion, so cut in rents comes t cost to vote-mximising politicins. Obviously, this politicl cost rises from the fct tht the fll in rents induced by tx competition reduces the utility of public sector workers. By recognizing this, our frmework llows us to identify n optiml degree of tx competition, ccounting for the welfre of ll citizens. The stndrd Levithn models of tx competition re unsuited for this purpose becuse they postulte tht rents do not generte welfre for ny citizen, nd becuse they do not ccount for the fct tht tx competition by chnging the size of the public sector will lso chnge the composition of the voting popultion, thereby ffecting the politicl equilibrium nd the distribution of welfre. Qulittively, tx competition ffects socil welfre through three different chnnels in our model: 1) By driving wedge between the mrginl rte of substitution nd the mrginl rte of trnsformtion between public nd privte 3 In n interesting recent pper Angelopoulos et l. (2006) incorporte rent-seeking into n otherwise stndrd Dynmic Stochstic Generl Equilibrium model of the Europen economies. The clibrted version of this model llows n estimte of the frction of time spent on unproductive rent-seeking ctivities imed t diverting public revenues into privte hnds. In the model of Angelopoulos et l. this frction turns out to be lrge. However, s the uthors recognize themselves, the bility of stndrd DSGE model to generte relistic employment fluctutions (with plusible lbour supply elsticity) is improved whenever one introduces third use of time in ddition to leisure nd mrket work. In the uthors' model rentseeking represents such third use of time, but similr improvement in the model's bility to fit the dt might hve been chieved by introducing nother lterntive use of time such s home production (see, e.g., Greenwood et l. (1995)). 4 Our pper my lso be seen s n extension of some ides in Wilson (1989) who studies optiml constrints on the tx bse in world where the tx rte is controlled by policy mker who diverts resources from spending on public goods towrds fvoured group of consumers. Such behviour by the policy mker could be interpreted s n ttempt to buy votes from n influentil interest group, but unlike us, Wilson (op. cit.) does not explicitly model the politicl process, nd he does not consider the effects of tx competition. 5 To limit the scope of the pper, we do not consider whether tx competition leds to less corruption nd whether it cn be used to generte vluble informtion to voters. Much empiricl work on the efficiency effects of fiscl federlism hs focused on the reltionship between fiscl decentrlistion nd corruption. In cross-country pnel study using the Interntionl Country Risk Guide's corruption index, Fissmn nd Gtti (2002) find significnt negtive reltionship between corruption nd decentrlistion. Huther nd Shh (1998) obtin similr results using similr indictors of corruption constructed by the World Bnk.

5 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) goods, it tends to reduce ggregte welfre. This is the welfre-reducing effect of tx competition emphsized in the literture ssuming tht policy is mde by benevolent socil plnner. 2) By reducing rents to public sector workers, tx competition equlizes the mrginl rte of substitution between public nd privte goods cross privte nd public sector workers. Ceteris pribus, this results in n outwrd shift of the utility possibility frontier which tends to increse socil welfre. 3) By curbing rents to public sector workers, tx competition lso equlizes the mrginl utility of income for ll citizens. In society concerned bout equlity, this likewise tends to increse socil welfre. By including the mechnisms 2) nd 3), our nlysis ccounts for both of the stndrd criticisms ginst rent-seeking, i.e., the objections tht it inefficient s well s unfir. We first consider the cse where governments non-coopertively choose the level of cpitl txtion, public sector rents nd employment. We show tht n increse in tx bse mobility will initilly tend to be welfre-incresing, but beyond certin point which depends inter li on the politicl strength of the public sector lobby, further increse in tx bse mobility will reduce welfre s the underprovision of public goods becomes more serious. In quntittive version of the model we trce the impct of incresing the number of competing tx jurisdictions nd rnk different scenrios in welfre terms to identify the economic nd politicl conditions under which tx competition is preferble to utrky. Our findings suggest tht tx competition is bdly trgeted remedy ginst politicl distortions. We then nlyse the effects of interntionl tx coordintion on socil welfre. When individul countries re too smll to ffect the world interest rte, some mount of interntionl tx coordintion will be welfre-improving under very mild conditions. Indeed, we find tht it my be welfre-enhncing to crry tx coordintion beyond the point where rents to public sector workers strt to emerge. In Section 2 we set up our model. Section 3 nlyses how the politicl equilibrium is influenced by tx competition while Section 4 studies how interntionl tx coordintion ffects rent-seeking nd socil welfre. Section 5 summrises our min conclusions, nd three technicl ppendices document the results reported in the text. 2. The model We consider world economy consisting of n symmetric countries. Residents in ech country cn either work in the privte or in the public sector, nd they consume privte goods s well s pure public good. Lbour is the only input into the production of the public good, while privte goods re produced by mens of cpitl nd lbour. Cpitl is perfectly mobile cross countries, wheres lbour is immobile interntionlly. There re no interntionl spillovers from the supply of public goods, but since public expenditure is finnced by source-bsed tx on cpitl, there is fiscl externlity rising from interjurisdictionl competition for the mobile tx bse. All countries produce the sme good, so ntionl tx policies hve no effects on the commodity terms of trde. Politicins choose the level of txtion, the level of public service provision nd the public sector wge rte with the purpose of mximising the probbility of being voted into office. Voters re split into group of well-orgnised insiders employed in the public sector nd group of non-orgnised outsiders minly employed in the privte sector. By incresing the economic welfre of the members of prticulr group, politicins cn increse the expected number of votes from tht group. The model enbles us to specify the exct conditions under which rents to public sector employees will rise. A centrl issue to be explored is whether tx competition will tend to reduce such rents nd move public sector employment closer to its socilly optiml level. Below we present the detils of the model Tstes nd technology We use the subscript g for vribles relting to government sector employee nd the subscript p for vribles referring to privte sector employee. All gents hve identicl preferences with respect to consumption (but not with respect to non-economic spects of public policy, see Section 2.4), nd the totl economic welfre U j of worker employed in sector j is U j ¼ uc j þ gg ð Þ; j ¼ g; p; ð1þ uvn 0; uw b 0; g VN 0; gw b 0;

6 1146 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) where C j is privte consumption nd G is the non-rivl consumption of the public good. Note tht since individul working time is ssumed to be institutionlly fixed, there is no need to llow for the disutility of work in the utility function (1). The totl popultion nd lbour force is normlised to unity nd the frction of the lbour force employed in the public sector is denoted by α, 0bαb1. Totl cpitl input into privte sector production is (1 α)k, where k is the cpitl lbour rtio, nd the totl output of privte goods (Y) is given by the linerly homogeneous production function Y ¼ Fðð1 Þk; 1 Þ; ð2þ implying tht the verge productivity of privte sector worker is yu Y 1 ¼ Fk; ð 1Þuf ðþ; k f VN 0; f W b 0: ð3þ The public good is produced by simple liner technology with lbour s the only input: G ¼ : ð4þ At the strt of the period considered, ech country in the world is endowed with fixed totl cpitl stock k. All countries re ssumed to be symmetric, with identicl lbour forces, cpitl endowments, tstes nd technologies The first-best lloction For lter reference it will be useful to chrcterize the first-best lloction of resources in our simple world economy, ssuming tht the socil plnner in the representtive country wishes to mximise the utilitrin socil welfre function SW ¼ uc g þ gðþ þ ð1 Þ ucp þ gðþ : ð5þ One condition for globl optimlity is globl production efficiency which requires tht cpitl's mrginl product be equlized cross countries. With identicl countries this is chieved when investment in ech country equls the country's fixed cpitl endowment. Hence optimlity is ttined when the socil welfre function (5) is mximised with respect to C g, C p, nd α, subject to the resource constrint C g þ ð1 ÞC p ¼ Fð P k; 1 Þ: ð6þ Denoting the mrginl product of privte sector lbour input by F L, the first-order conditions for the solution to this problem cn be shown to imply uv C g ¼ uv Cp Z Cg ¼ C p ¼ C; ð7þ g Vð Þ uvðc Þ ¼ F P L ðk; 1 Þ: ð8þ Eq. (7) sttes tht privte consumption levels must be equlized so s to equlize the mrginl utility of consumption cross the two groups of workers. This condition my be sid to reflect policy concerns bout equity. Eq. (8) is the Smuelson condition for the optiml supply of public goods, stting tht the sum of the mrginl rtes of substitution between privte nd public goods should equl the mrginl rte of trnsformtion (recll tht the totl popultion is normlised to unity, so the left-hnd side of (8) is the sum of the mrginl rtes of substitution). Clerly, (8) cptures policy concerns bout efficiency. We will now study whether the mrket-bsed lloction will differ from this first-best optimum The mrket economy Competitive profit-mximising firms invest up to the point where cpitl's mrginl product equls the cost of cpitl, implying f VðÞ¼r k þ s; ð9þ

7 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) where r is the fter-tx interest rte nd τ is source-bsed unit tx on cpitl. From (9) it follows tht cpitl intensity is given by k ¼ krþ ð sþ; k V¼ 1=f W b 0: ð10þ Moreover, (9) nd the liner homogeneity of the production function imply tht the privte sector rel wge (w) is wrþ ð s Þ ¼ fðkðrþ sþþðr þ sþkrþ ð sþ; wv¼ k: ð11þ Cpitl is perfectly mobile cross countries. With source-bsed cpitl txtion, this mens tht ll the n countries in the world fce the sme fter-tx interest rte r. A globl cpitl mrket equilibrium is ttined when ð1 Þkrþ ð sþþ ðn 1Þð1 Þ krþ ð sþ ¼ n P k; ð12þ where (1 α)k(r+τ) is cpitl demnd in the domestic country under considertion, nd (1 αˆ)kˆ (r+τˆ) is cpitl demnd in ech of the n 1 identicl foreign countries. Thus the left-hnd side of (12) mesures the globl demnd for cpitl which must equl the fixed globl cpitl supply, nk. By implicit differentition of (12) we my find the isolted effects of domestic tx nd spending policies on the fter-tx interest rte, exploiting the symmetry ssumption tht ll countries end up choosing the sme policies in equilibrium 6 : Ar As ¼ ð1 Þk V ð1 Þk Vþ ðn 1Þð1 Þ k V ¼1 n ; ð13þ Ar A ¼ k ð1 Þk Vþ ðn 1Þð1 Þ k V ¼ k nð1 Þk V : ð14þ When choosing their fiscl policy pltforms, politicins ccount for these policy effects on the interest rte. To focus on the potentil conflicts of interest between privte nd public sector employees, we ssume tht cpitl endowments re eqully distributed cross the working popultion. Reclling tht the totl lbour force is normlised t unity, this mens tht ech worker owns the mount of cpitl k. Denoting the public sector wge rte by W, the privte consumption of the two types of workers is then given by C g ¼ W þ r P k; C p ¼ w þ r P k: ð15þ 2.4. The politicl economy of fiscl policy The policy vribles in our model re W, G nd τ. We wish to provide simple frmework in which these vribles re chosen by politicins competing for votes. Inspired by Persson nd Tbellini (2000, ch. 3), we describe the politicl process by probbilistic voting model with lobbyism. In our prticulr version of this model, voters re split into insiders nd outsiders. The insiders re ll employed in the public sector nd ll belong to lobby (sy, trde union) which enforces the wge rte W throughout the public sector in order to prevent underbidding from outsiders. The outsiders re those voters who do not belong to the lobby. These individuls re employed either in the public or in the privte sector. Thus the mrginl workers in the public sector re outsiders lthough they re pid the sme wge s the insiders. As we shll see below, in the bsence of tx competition the public sector wge rte will generlly exceed the privte sector wge. The mrginl high-pying public sector jobs tht re not lredy filled by the insiders re llocted to some of the outsiders. Flexible wge djustment in the privte lbour mrket ensures tht those outsiders who do not get public sector job re ll ble to find privte sector employment. Our ctegoriztion of public sector workers into insiders nd outsiders is motivted by the observtion tht some groups of civil servnts re often employed on long-term contrcts providing high degree of job security wheres other public sector workers re ppointed on short-term contrcts offering less job protection. Our distinction between public sector insiders who hve full job security nd the mrginl public sector workers who cn esily be dismissed 6 The symmetry ssumption implies tht α=αˆ nd τ=τˆ in equilibrium so tht (1 α) k (r+τ)=(1 αˆ) kˆ (r+τˆ ).

8 1148 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) cptures this observed difference in the terms of employment in stylised wy. Note tht the stronger ttchment of insiders to the public sector could explin why this group hs formed lobby to protect their interests wheres the mrginl workers with looser link to the public sector do not enter the lobby. Visser (2006) documents tht public sector workers in the OECD re re in fct better orgnised thn workers in the privte sector, s reflected in much higher degree of unionistion in the public sector. Our ssumption tht only public sector insiders hve formed lobby seeks to cpture this mrked difference cross sectors in simple wy. Whether the higher union density implies tht public sector workers ctully ern rents is ultimtely n empiricl issue. As we show in Appendix A, if privte sector workers re very tightly knit group in terms of ideologicl preferences, they could be politiclly more influentil thn public sector workers even if they hve not formed lobby. However, the probbilistic voting model considered below does imply tht public sector workers generlly ern rents under utrky. Our ssumption tht (orgnised) public sector voters constitute strong interest group cpble of extrcting rents is mde becuse it seems to be implicit in the resoning of mny of those who dvocte tx competition s remedy ginst rent-seeking. 7 Our purpose is to investigte whether fiscl competition could indeed be n pproprite mens of curbing n excessive politicl influence of public sector workers. While we do not wish to pss verdict on whether such n excess influence ctully exists, we note tht most of the empiricl studies surveyed by Bender (1998) do find positive centrl government privte sector wge differentil in the U.S. nd Western Europe. 8 Let us now describe the detils of the politicl process, drwing hevily on the frmework proposed by Persson nd Tbellini (2000, sec. 3.5). There re two politicl prties (A nd B) competing for government office. Ech prty chooses fiscl policy pckge consisting of level of public service provision, public sector wge rte nd cpitl tx rte. The two prties differ in some ideologicl dimension nd voters hve different individul preferences regrding this non-economic spect of policy. Let U i P denote the economic welfre of public sector insider in cse the policy of prty P is implemented, P=A,B. An individul member j of the public sector insider lobby will then vote for prty A if U A i N Ui B þ q j i þ x; ð16þ where ρ i j is n individul ideologicl bis in fvour of prty B, with zero men vlue cross ll lobby members, nd ω is generl (stochstic) ideologicl preference in fvour of tht prty, cpturing ny underlying politicl mood ffecting ll voters. Similrly, if U o P is the expected economic welfre of n outsider in cse prty P's economic policy is implemented, voter v in the group of outsiders will prefer prty A if U o A NU o B +ρ o v +ω. The generl ideologicl bis is ssumed to be given by x ¼ x þ h ð i Z B i Z A Þ; h N 0; 0 b i b 1: ð17þ Here ω is stochstic term with men zero, Z P is the public sector lobby's cmpign effort in support of prty P, mesured per member of the lobby, nd α i is the predetermined frction of voters belonging to the lobby so tht α i Z P is the totl lobby support for prty P. The cmpign effort could tke the form of lobby officils working in the politicl cmpign. 9 According to (17), the more the cmpign effort in support of prty B exceeds the effort in fvour of prty A, the stronger is the voter bis in fvour of prty B on election dy. We ssume tht cmpign efforts cnnot be negtive, i.e. the lobby cnnot extrct rel resources from politicins, only politicl concessions. Let p A denote the probbility tht prty Awins the election so tht 1 p A is the probbility tht prty B crries the election dy. Lobby officils choose their 7 If privte sector workers were politiclly more influentil thn those employed in the public sector, the model set up below implies tht n unconstrined politicl cndidte would idelly wnt to keep the public sector wge rte below tht in the privte sector in order to keep txes low. However, this scenrio would not illustrte the interdependence between public sector rents nd tx competition which is the focus of the present pper. Moreover, with flexible wge djustment preventing involuntry unemployment in the privte sector, the public sector would fce recruitment problem if it offered lower wge rte thn the privte sector. 8 Flch nd Strøm (2005) lso find evidence from Norwy tht vrious indictors of the politicl strength of public sector employees hve positive impct on public sector wge rtes. However, these uthors do not investigte whether public sector workers re generlly better pid thn corresponding groups of workers in the privte sector. Whether public sector workers ern rents is n issue tht is probbly hrd to settle. In our model rents tke the form of reltively high public sector wge rte, but the wge rtes in the model should be interpreted s wges per unit of effort for wge differentils to be n pproprite indictor of rents. Thus, even if empiricl studies were to revel tht public sector wge rtes do not exceed the wges for similr groups of workers in the privte sector, this would not necessrily imply the bsence of rents in the public sector. 9 The Z P -vribles could lso be interpreted s monetry cmpign contributions. The exct interprettion is unimportnt since Z P will be zero in politicl equilibrium, s we demonstrte below.

9 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) cmpign efforts to mximise the following objective function (L), representing the lobby members' expected utility from the election outcome, net of the cost of lobby effort: L ¼ p A U A i þ ð1 p A ÞUi B 1 2 Z2 A þ Z2 B ð18þ The negtive third term in (18) ssumes convex costs of cmpign efforts, reflecting incresing mrginl disutility of effort. Note tht the individul ideologicl preferences hve cncelled out in the lobby's objective function, since the preference vrible ρ j i hs zero men vlue nd the lobby mximises the verge utility of its members. The timing of politicl events is s follows: 1) Ech prty nnounces fiscl policy pckge, tking the number of insiders nd the policy pltform chosen by the other prty s given. 2) Lobby officils choose their cmpign efforts. 3) Nture chooses the vlue of the stochstic voter preference vrible ω. 4) Elections re held. 5) The prennounced policy of the winning prty is implemented, nd the mrginl public sector jobs re llocted mong j outsiders (by process to be specified below). For simplicity we ssume tht the individul ideologicl preferences ρ i nd ρ v o follow n identicl huniform distribution i nd tht the generl ideologicl preference vrible ω is uniformly distributed on the intervl 1 2w ; 1 2w. On these ssumptions Appendix A demonstrtes tht the optiml cmpign efforts re Z A ¼ mx 0; i whui A Ui B ; ZB ¼ mx 0; i whui B Ui A : ð19þ Thus the lobby will only support prty tht offers its members higher level of economic welfre thn the other prty. Moreover, Appendix A shows tht the probbility tht prty A will win the election is p A ¼ 1 2 þ ip i Ui A U B i þ ð 1 i Þp o Uo A U o B ; 0 V i b 1; p i ¼ w þ i w 2 h 2 ; p o ¼ w: ð20þ In other words, the greter n insider's economic welfre implied by the policy of prty A, the greter is the likelihood tht he will vote for tht prty, given the economic policy pckge offered by prty B. In similr wy, prty A cn increse its voter support from outsiders by choosing fiscl policy pltform tht increses the (expected) economic welfre of members of tht group of voters. Mximistion of p A, given U B i nd U B o, gives prty A's best response to the policy chosen by prty B. The ltter prty fces the symmetric problem of mximising 1 p A, yielding similr first-order conditions nd n identicl bestresponse function. In Nsh equilibrium the two prties therefore end up choosing the sme fiscl policy pltforms implying U A i U B i =0, so ccording to (19) the public sector lobby will not wnt to offer ny cmpign contributions in politicl equilibrium. 10 Thus the politicl influence of the lobby derives from the potentil rther thn from the ctul politicl support tht it offers. The economic welfre of n insider is simply equl to the utility of public sector worker (U g ), tht is, U i = U g =u(w + rk )+g(g) (since both prties choose the sme policy in equilibrium, we no longer ttch prty superscripts to ny vribles). The expected economic welfre of n outsider, U o, depends on the probbility tht he will be ble to get one of the high-pying public sector jobs tht hve not lredy been reserved for the lobby insiders. For simplicity we ssume tht ll outsiders fce the sme probbility of getting one of the mrginl public sector jobs, s if these jobs were llocted by lottery. The totl number of public sector jobs is α of which α i b α re reserved for the lobby members. Hence (α α i ) is the number of public sector jobs offered to outsiders, nd 1 α i is the number of outsiders competing for those jobs. Thus, t the time of voting, (α α i )/ (1 α i ) is the probbility tht n outsider will gin ccess to public sector job, so the expected utility of n outsider is U o ¼ i 1 i U g þ 1 i 1 i U p ¼ i uwþr P k 1 i ð Þþ 1 i 1 i uwþ ð r P kþþgg ð Þ ð21þ 10 This is why our model specifiction in Section 2.3 did not explicitly llow for lobby ctivities s one possible use of the economy's resources.

10 1150 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) where we recll tht U p is the utility of privte sector employee nd tht those outsiders who do not get public sector job (the number of which is 1 α) ll end up finding employment in the privte sector, due to flexible djustment of the privte sector wge rte w. When choosing fiscl policy pckge (W,G, τ), politicins fce the technologicl nd mrket constrints (4), (13) nd (14) plus the government budget constrint which requires tht the revenue from cpitl txtion must cover the cost of the wges to public sector employees: sð1 Þkrþ ð sþ ¼ W: ð22þ Moreover, in order to be ble to ttrct workers to the public sector, these workers must be offered utility level t lest s high s tht enjoyed by workers in the privte sector. This recruitment constrint in turn requires tht W z w: ð23þ Our ssumption tht public sector insiders hve full job security lso implies tht fiscl policy must stisfy the nonfiring constrint α α i. In the nlysis below we ssume tht this constrint is never strictly binding. 11 Our prsimonious model obviously relies on strong simplifictions. First, in more elborte politicl economy frmework politicins might try to dole out the mrginl high-pying public sector jobs in return for politicl support. Second, the model feture tht cmpign contributions re zero in equilibrium derives from n implicit ssumption tht ll voters re eqully well informed. As shown by Bron (1994), when voters hve different informtion sets it my be optiml for lobbies to offer positive cmpign contributions in equilibrium to influence uninformed voters. 12 Third, like other sttic probbilistic voting models, our model neglects the potentil time inconsistency problem in the politicl process, implicitly ssuming tht reputtion mechnisms keep politicins from defulting on their cmpign promises. In ddition to ensuring nlyticl trctbility, the bove simplifictions llow precise definition of the populr concept of politicl distortion. In the nlysis below we shll thus mesure the degree of politicl distortion by the following prmeter (using the specifictions from (20)): du i p i p o p o ¼ 2 i h2 w: The politicl distortion is the product of the predetermined size of the public sector lobby (α i ) nd the reltive increse in votes politicl cndidte my expect to gin by ctering to the economic interests of insiders rther thn outsiders, (p i p o )/p o. The more δ exceeds zero, the greter is the politicl influence of public sector insiders reltive to tht of other voters. It is intuitive tht the politicl distortion is greter the lrger the lobby nd the greter the impct of lobby efforts on voter preferences (the higher the vlue of h). We lso see tht smller dispersion of ideologicl preferences ( higher vlue of ψ which reduces the intervl over which ideologicl preferences re distributed) increses the politicl distortion. When ideologicl preferences re firly similr cross lrge number of lobby members, n increse in the economic benefits offered by one prty to lobby members will induce mny of them to shift their vote in fvour of tht prty, nd hence the lobby becomes more influentil. Note how our politicl set-up tries to ccount for the views of those dvoctes of tx competition who rgue tht the public sector tends to employ too mny people on overly generous conditions: First, becuse economic benefits offered to public sector insiders generte more votes thn benefits offered to outsiders (s reflected in the fct tht p i Np o ), our model includes n incentive for politicins to offer rents to public sector workers. Second, when politicl cndidte offers high public sector wges, he my lso be inclined to promise more jobs in the public sector, since (20) nd (21) imply tht this will increse n outsider's expected economic gin from voting for tht cndidte. 13 On the other hnd, reltively high public sector wge rte mkes the cretion of public sector jobs more expensive by requiring higher tx ð24þ 11 If it were binding, we would hve reltively uninteresting scenrio with n exogenous lloction of lbour between the public nd the privte sector. 12 Lorz (1998) lso offers politicl economy model with positive lobbying ctivity in equilibrium. In his setting tx competition cuses welfreincresing drop in lobby ctivity becuse it reduces the bility of the government to redistribute income, thereby diminishing the expected gin from lobbyism. However, unlike the present pper, Lorz (op. cit.) does not provide n explicit description of the voting process. 13 When WNw, we hve U g NU p, so from (2.21) we get AUo A ¼ N 0. It then follows from (20) nd (21) tht Ap A A ¼ U g U o N 0. UgUp 1 i

11 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) rte. Hence politicins must trde off the politicl gin from high public sector wges nd public sector job cretion ginst the politicl cost of hving to rise txes. The next section nlyses the resulting politicl equilibrium. 3. Politicl equilibrium, tx competition nd rents 3.1. Politicl equilibrium In politicl equilibrium the fiscl policy vribles W, G, nd τ re set so s to mximise the probbility of election victory (20), subject to the government budget constrint (22) nd the recruitment constrint (23). The first-order conditions for the solution to this problem re derived in Appendix B. When the constrint W w is not strictly binding, these conditions cn be shown to imply tht uv g ¼ g VðÞ uv g þ d 1 n 1 n e þ u g u p uv g ð1 þ dþ ¼ þ d þ d uv p ; e u n 1 n sk V k ; 1 þ ðn 1Þ W ð1 Þðn Þ w where u g u(w+rk ) nd u p u(w+rk ) re the totl utilities of privte consumption for public nd privte sector workers, respectively; u g u (W+rk ) ndu p u (w+rk ) re the corresponding mrginl utilities; nd ɛ is the numericl elsticity of the tx bse with respect to the tx rte. 14 To understnd the effects of tx competition on public sector efficiency, it is useful to strt by considering the benchmrk cse of utrky where no interntionl cpitl mobility is llowed. The world economy will then function like closed economy which we my model by setting the number of countries n=1. Suppose for moment tht there is no politicl distortion, i.e. δ=0. According to (25) the politicl equilibrium under utrky then implies u g =u p which in turn implies W=w nd u g =u p.forn=1 nd δ=0 Eq. (26) then reduces to the Smuelson condition g /u =F L.In other words, politicl equilibrium without cpitl mobility will gurntee first-best lloction without rents when there is no politicl distortion. In this cse politicins cnnot cpture more votes by offering prticulr benefits to one group t the expense of nother, so vote-mximising politicins hve n incentive to ct like utilitrin socil plnner who ttches n equl weight to the welfre of ech individul citizen. Wht hppens if we llow politicl distortions in fvour of public sector workers while mintining the utrky ssumption? In tht cse we obtin Proposition 1. Strting from n undistorted politicl equilibrium under utrky, the introduction of smll politicl distortion in fvour of public sector workers will drive the public sector wge rte bove the wge rte in the privte sector. It will lso drive up the tx rte but will leve public sector employment unffected. Proof. See Appendix C. According to Proposition 1 the formtion of lobby for (some of the) public sector workers will induce politicins to crete rents to civil servnts. Not surprisingly, the tx rte will hve to rise to finnce the increse in public sector wges. However, the number of public sector jobs will sty the sme becuse of two offsetting politicl incentives. On the one hnd, the emergence of rents to public sector employees provides n incentive for politicl cndidte to boost public sector employment, since he cn thereby cpture more votes from outsiders by incresing their chnces of getting n ttrctive public sector job (see footnote 13). On the other hnd, the emergence of the lobby mkes public goods more expensive by driving up the public sector wge rte. Ceteris pribus, this rise in the cost of public goods provision induces politicins to offer fewer public sector jobs. When there is no lobby initilly, it turns out tht these two counterviling politicl incentives exctly neutrlize ech other. F L ; ð25þ ð26þ 14 Note tht ɛ is generl-equilibrium elsticity, llowing for the impct of chnge in the domestic tx rte on the world interest rte. Specificlly, the tx bse elsticity is defined s eu dðkrþs ð ÞÞ s V ð dsþar As ds k ¼k dsþ s n1 ds k ¼ sk V n k ; where we hve used the symmetry ssumption plus Eq. (13) to derive the lst equlity.

12 1152 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) Since prt of the tx increse needed to finnce the rise in public sector wge rtes is pid by privte sector workers, the disposble income nd privte consumption of public sector workers must go up. With n unchnged vlue of α = G, it follows tht the mrginl rte of substitution g (G)/u (W + rk ) is driven bove the mrginl rte of trnsformtion F L (α), so public goods become underprovided from the viewpoint of public sector workers. However, since the higher cpitl tx rte cuses drop in r (see (13)) nd resulting fll in the privte consumption of privte sector workers, their mrginl rte of substitution g (G)/u (w+rk ) will fll below F L (α), so public goods will become overprovided from the perspective of privte sector workers Tx competition, rent destruction nd public goods provision Consider next the role of tx competition, i.e. the cse where nn1 so tht the tx bse elsticity becomes positive, due to interntionl cpitl mobility. From (25) we cn show Proposition 2. Tx competition will completely eliminte rents to public sector employees if the politicl equilibrium under tx competition implies d e N 1 : ð27þ þ d n Proof. See Appendix C. The condition in (27) is very intuitive: the higher is the tx bse elsticity ɛ, the higher is the mrginl cost of public funds, so the more costly (in economic nd politicl terms) it is for politicins to rise txes to finnce rents to public sector employees. Hence, if the politicl distortion in fvour of public sector insiders is not too high (so tht the frction d þd in (27) is not too big), tx competition will prevent rent cretion. However, while tx competition my provide n institutionl defence ginst rent-seeking, s emphsized by the Public Choice school, it will lso distort the supply of public goods, s climed by the trditionl Public Finnce school. This is reflected in Proposition 3. When tx competition mong smll jurisdictions is sufficiently strong to eliminte ll rents so tht u g =u p =u, public goods will be underprovided nd the supply of public goods will stisfy the condition g Vð Þ uv ¼ 1 1 F L : ð28þ 1 1 eð1 Þ Proof. See Appendix C. Under tx competition public goods re underprovided (i.e. g (α)/u NF L ) since the interntionl mobility of cpitl cuses the tx bse to be elstic from the individul country's perspective wheres from the viewpoint of the world economy s whole it is in fct inelstic. However, under utrky we sw tht the supply of public goods is lso distorted, due to the bis in the politicl process. To evlute which regime is likely to generte the biggest distortions, it is useful to consider quntittive version of our model Is tx competition good or bd? A numericl generl-equilibrium nlysis To illustrte how rents, public goods provision nd socil welfre my evolve s the number of competing jurisdictions increses, we simulte clibrted version of our model, ssuming Cobb Dougls production function y ¼ Ak b ; A N 0; 0 b b b 1; ð29þ nd preferences of the form uc ð Þ ¼ C1r c ; gðþ ¼ h1r g ; r c N 0; r g N 0; h N 0; ð30þ 1 r c 1 r g

13 W. Eggert, P.B. Sørensen / Journl of Public Economics 92 (2008) Tble 1 Simulted effects of tx competition n ɛ W w α s r þ s SW SW jsw j Clibrtion: δ=0.12, σ c =σ g =5, β=0.25, θ=1, k =0.1, A=0.1. where the prmeter θ reflects the preference for public goods. Assuming β=0.25 nd σ c =σ g =5 15 ; postulting politicl distortion δ=0.12; setting θ=1, nd clibrting the prmeters A nd k to ensure relistic reltive size of the public sector (nd relistic effective cpitl income tx rte τ/(r + τ)), we obtin the simultion results reported in Tble The lst column shows the chnge in the level of socil welfre reltive to the welfre level SW ttined under utrky, nd the first row in the tble shows the sitution previling under utrky. The second column in the tble shows tht the elsticity of the tx bse grdully increses with the number of competing jurisdictions. As tx competition grows more intense, the reltive public sector wge rte W/w grdully declines, nd when the number of jurisdictions becomes sufficiently lrge, rents re completely eliminted, i.e., the recruitment constrint W w becomes binding. Tx competition lso reduces the size of the public sector, but not drmticlly so, since our ssumed vlues of σ c nd σ g imply reltively low degree of substitutbility between public nd privte goods. Notice the interesting profile of the welfre chnge in the lst column in Tble 1: s the number of countries rises from one to some smll number, socil welfre rises bove the utrky level, becuse the positive effect of rent destruction domintes the negtive effect of lower public goods provision. However, s the number of countries increses from eight to nine, implying n increse in the tx bse elsticity from to , the welfre gin from tx competition is turned into slight loss, s the negtive efficiency effect of reduced public service provision strts to dominte. Indeed, in this prticulr exmple the mximum welfre gin from tx competition is ttined lredy when the number of countries is two, t tx bse elsticity of bout 0.3. Given our clibrtion, this tx bse elsticity represents the optiml intensity of tx competition. Of course these results re sensitive to the choice of prmeter vlues. One criticl prmeter is the degree of politicl distortion, δ. As the vlue of this prmeter increses, it tkes higher intensity of tx competition reflected in the number of countries nd the ssocited elsticity of the tx bse before the negtive welfre effect of reduced public goods provision strts to dominte the positive welfre effect of rent destruction. This is illustrted in Figs. 1 nd 2 which show the combintions of the politicl distortion nd the number of competing jurisdictions (nd the implied tx bse elsticity) tht will led to exctly the sme level of welfre s tht ttined under utrky, given the other prmeter vlues stted in the note to Tble 1. For prmeter combintions bove the grphs in the two figures, tx competition is welfre-improving, wheres in the re below the grphs it is welfre-reducing. As one would expect, the figures illustrte tht tx competition is more desirble the greter the politicl distortion in fvour of public sector voters. The clibrted version of our model suggests tht even lrge politicl distortion cn only justify moderte intensity of tx competition. For exmple, if the public sector lobby is trde union comprising 10% of the totl work 15 In n intertemporl context, our prmeter r c would be identicl to the inverse of the intertemporl elsticity of substitution in privte consumption. Bsed on the estimtes of the ltter prmeter by Hll (1988), r c should be t lest 5, wheres the estimtes presented in Attnsio nd Weber (1995) imply vlues of r c between 2.2 nd The complete model implied by the specifictions (29) nd (30) is documented in supplementry ppendix vilble from the uthors.