Fiscal decentralization, governance, and. economic performance: a reconsideration

Transcription

1 Fisa deentraization, governane, and eonomi performane: a reonsideration In ountries with tax sharing systems, assigning oa governments a arge share of oay-generated tax revenues is often thought to motivate them to promote eonomi deveopment. The more oa offiias benefit from oa eonomi ativity, so the argument goes, the more supportive of business and the ess orrupt they shoud be, eading to higher eonomi output. Some shoars attribute China s rapid growth to its high oa tax retention rates and Russia s stagnation in the 1990s to the predatory awbak of oa revenues. I show that suh arguments ignore a ruia ator in the game the entra government. If inreasing the oa tax share improves inentives for oa offiias, it usuay worsens them for their entra ounterparts. The net effet on eonomi output is indeterminate. Keywords: fisa deentraization, orruption, inentives. JEL H11, H30, H71, H77 Danie Treisman Department of Poitia Siene University of Caifornia, Los Angees 489 Bunhe Ha Los Angees, CA Treisman@poisi.ua.edu Marh 006 Pubished in Eonomis and Poitis, Juy 006, 18,, pp I am gratefu to Hongbin Cai, J.R. DeShazo, Stephan Haggard, Andrei Sheifer, Mihe Treisman, Yi Zhang, anonymous referees, and partiipants in the 003 Amerian Poitia Siene Assoiation annua meeting for vauabe omments. I thank the UCLA Aademi Senate, the John Simon Guggenheim Foundation, and the German Marsha Fund of the US for support.

2 1 Introdution A growing body of iterature reates ountries eonomi performane to inentives reated by the way fisa authority is divided between eves of government. Some have studied how ompetition between oa and entra governments to tax or reguate the same base an ead to ineffiienty high tax or reguatory burdens. 1 Others have expored how fator mobiity might onstrain oa governments in desirabe or undesirabe ways. Sti other work has examined how subnationa governments might expoit a pre-ommitment advantage to extrat baiouts from higher eves. 3 Another ommon argument whih I address in this paper onerns inentives in a system of tax sharing. In suh systems, governments at two or more eves eah reeive a share of the revenues from taxes oeted in a given jurisdition. Tax sharing systems are extremey widespread. A very partia ist of ountries where at east one eve of subnationa government got more than haf its tax revenues from intereve tax sharing as of the ate 1990s woud inude Austria, Begium, Germany, Norway, Spain, Turkey, Mexio, Boivia, Nigeria, Hungary, the Czeh Repubi, Poand, Romania, Bugaria, Abania, Croatia, Estonia, Latvia, Lithuania, Russia, Ukraine, Armenia, Bearus, Georgia, Kazakhstan, Modova, and Tajikistan. 4 In many other ountries, tax sharing exists in a ess pronouned form. Even where tax sharing does not our de jure, if entra governments provide fisa transfers to oa budgets that vary with oa performane, then a kind of tax sharing often exists de fato. In suh systems, assigning arger shares of revenue to oa governments is often thought to 1 This is the iterature on vertia fisa externaities. For a reent anaytia survey, see Keen (1998). Contributions inude Fowers (1988), Dahby (1996), Boadway and Keen (1996), Boadway, Marhand, and Vigneaut (1998), Wrede (1999), and Dahby and Wison (003). Cassi ites inude Tiebout (1956) on mobiity of residents and Brennan and Buhanan (1980) on mobiity of apita. Widasin and Wison (004) review the iterature. 3 See anayses of soft budget onstraints in federa finane; for instane, Widasin (1997), Bordignon et a. (001), Rodden et a. (003). 4 See OECD (1999), Daba-Norris and Wade (00, p.8), Ebe and Yimaz (00, p.8), Ter-Minassian (1997), Turkish State Institute of Statistis (1999, p.64).

3 promote more business-friendy oa poiies, and onsequenty better eonomi performane. 5 If oa governments get to keep a arge share of margina revenues generated in their jurisdition, they shoud be more eager to invest in business-promoting infrastruture, to imit osty reguations, and where orruption is rampant to urb their demand for bribes. On the other hand, if most of the margina doar of oay-generated revenues is siphoned off to higher eve budgets, oa offiias wi pae ower priority on stimuating eonomi ativity, and more on other objetives suh as ining their own pokets. This argument has been appied to a variety of ountries. In China, some shoars ontend that the high and inreasingy seure share of revenues retained by subnationa governments heps expain the ountry s phenomena growth in the 1980s and 1990s (Jin, Qian, and Weingast 005). Aording to Roand (000, p.81), China s fisa deentraization has heped to aign the inentives of government authorities with eonomi effiieny. By ontrast, a ow ex post share of revenue oay retained is seen as one reason Russia stagnated in the 1990s (Zhuravskaya 000; Banhard and Sheifer 000). Simia r arguments are often made about other deveoping ountries. Kisubi (1999, p. 13) redits revenue deentraization in Uganda with reduing the inentives for distrit eve orruption. In Pakistan, inadequate revenue deentraization is said to have disouraged provinia governments from fuy deveoping their own tax bases (Shah 1998, p.141, Ahmad and Wasti 003, pp.196-8). In the industriaized word as we, this argument sometimes surfaes in poiy debates. The Eonomist reenty attributed Britain s reativey suggish post-war performane in part to over-entraization of tax revenues, whih weakens inentives for oa governments to grant panning permission to firms that wish to expand. The state of the oa eonomy makes itte differene to oa government offers, so in the baane between growth and beauty, growth does not get muh of a say (The Eonomist 001, p.0). The intuition, based on an anaogy to ageny probems in orporate finane, is appeaing. Equity given to a ompany exeutive heps to aign his interests with those of other share-hoders. He shoud therefore strive to inrease the ompany s profits. In a simiar way, assigning a oa government a arger share of revenue from oay-originating taxes might be expeted to enourage that government to 5 I use oa in this paper to refer to a eves of subnationa government. 3

4 support growth of its tax base, resuting in higher output. However, the ontext is different in one ruia respet. The argument appied to fisa federaism ignores one important ator in the game the entra government. Inreasing the margina revenue shares of oa governments may indeed improve their inentives. But suh reforms simutaneousy weaken inentives for the entra government by reduing its stake in eonomi deveopment. Any positive effet on the behavior of oa offiias shoud be aompanied by a negative effet on entra government behavior. If equiibrium output is muh more sensitive to oa than to entra government predation, the oa effet might dominate. But in most ases interventions by both eves of government infuene eonomi ativity, and there is no genera reason to think one effet wi be stronger than the other. Indeed, in ertain pausibe formuations the two effets exaty ane eah other out: deentraizing revenues merey shifts predation or ineffiieny from one eve of government to the other, without hanging output. As a basis for generaizations or poiy reommendations, the fisa inentives argument thus seems highy questionabe. 6 That hanging the division of revenues wi usuay have opposite effets on entra and oa governments is obvious one reognized, but it haenges a widey hed intuition that has been used to argue for deentraization reforms. I demonstrate the point in a simpe mode of tax sharing simiar to those used in previous treatments, first ignoring the entra government and then inuding it as a strategi ator. The resut is robust to a variety of reformuations and hanges of modeing assumptions. As I show in a working paper version of this artie (Author 005), the inentive effet on output is indeterminate whether governments intervene in negative ways (by extrating bribes) or in positive ones (providing infrastruture), whether bribes are proportiona to firms revenues or ump-sum, whether governments are 6 In a reated paper, Gordon and Li (1997) argue that governments whose revenue omes from taxing private firms wi be more supportive of private markets than those that tax state enterprises. They note the heavy dependene of Chinese oa governments on taxing non-state firms, and suggest that this may hep expain why China s eonomy performed better in the 1990s than East European eonomies. This might, indeed, expain why oa governments support non-state business in China. But to draw onusions about ountries reative performane, one woud need to omp are aso the inentives for entra governments in China and Eastern Europe. If the Chinese entra government were more dependent than East European entra governments on taxing state enterprises, it woud ikey be ess friendy toward private business, and the net effet on performane woud be unear. 4

5 purey predatory or provide pubi goods, whether oa governments are more or ess aountabe than entra government, whether governments prefer bribes to tax revenues or vie versa, whether or not bribery is more osty than taxation (for instane, beause of serey), whether oa or entra governments provide infrastruture more effetivey, and whether or not oa governments are onstrained by apita mobiity. The argument that I address in this paper assumes a system of intereve tax sharing, or some funtiona equivaent. My goa here is to haenge the view that deentraizing revenues within suh a system eads to higher output. Of ourse, many eonomists woud argue for eiminating tax sharing entirey. Assigning separate tax bases to different eves of government shoud give eah a strong inentive to onserve and grow its own base. However, separating tax bases ompetey is not ususay feasibe. Poiies of governments at one eve amost aways affet the tax bases of other eves (whether via effets of infrastruture investments, hanges in demand, reguations, or bribery). I examine suh ases in Setion 3 and iustrate how, in the presene of suh externaities, assigning separate tax bases to different eves an at times make things worse. To be ear, I do not pretend to offer here a fuy speified mode of the fisa interations between a entra and oa governments from whih genera resuts an be derived. Rather, I address one ommon and infuentia argument that has informed aademi anayses and poiy disussions. 7 The reevant question to ask about the modes I present is, therefore, not whether they auratey represent the word, in a its ompexity, but whether they auratey represent the argument about inentives in a system of tax sharing that I am addressing. Sine versions of the mode without a entra government are virtuay identia to those presented by proponents of this argument (Zhuravskaya 000, Jin et a. 005), this woud seem to be the ase. I aso do not deny that the standard inentive argument may hep to expain partiuar ases. Where output is more sensitive to oa than to entra poiy, the onventiona expetation may be right. Zhuravskaya and Jin et a. present onsiderabe empiria evidene that good 7 For an exeent review of many other arguments about fisa deentraization and the quaity of government, see Oates (1999). 5

6 fisa inentives motivated Chinese provinia governments to support growth in the 1980s and that bad inentives ed to poorer governane in Russian ities in the 1990s. I argue ony that there is no theoretia basis to derive genera expetations or poiy reommendations from the outomes of these ases. Fisa deentraization under tax sharing.1 The standard argument I begin with a simpe formaization of the intuition that inreasing oa governments revenue shares motivates those governments in ways that improve eonomi performane. Consider a two-eve state, and fous on a singe oaity, ontaining a singe itizen with a produtive business. Output and inome in the oaity is Y. There is no interregiona trade or investment, so a inome arues to this itizen. A singe tax is evied on output, at a fixed rate t (0,1) oa and entra governments, the oa government reeiving a share r (0,1). 8 The revenue is divided between the a share 1 r. Sine the argument is about omparative statis of r, I aso take this as fixed. 9, and the enter reeiving Various types of government intervention might affet eonomi performane. Suppose here that governments extrat bribes from business, proportiona to output. (I show in Author (005) that simiar resuts hod for ump-sum bribes or for interventions to provide infrastruture.) The oa government hooses a bribe rate on output, 0 b 1 t. 10 One might think of the oa offiias as demanding a share of the business s output in return for waiving some osty reguation. 11 To mode the argument that 8 In Author (005), I onsider endogenizing the tax rate and show that this wi not usuay hange the main resut. 9 Of ourse, the determinants of r are interesting in themseves. I return to this briefy in the Conusion. 10 As is ustomary, I assume here a form of bribery with diret distortionary effets. If the government s intervention is not distortionary, motivating oa governments to intervene ess wi not improve eonomi performane. I assume throughout that bribe rates are set after the tax rate is fixed. This seems reasonabe sine tax rates are usuay set by government in an annua aw that is hard to hange, whereas bribe rates an be atered by oa government offiias at their disretion. 11 Athough strit proportionaity may not our, it seems reasonabe to think the size of bribe demanded might hange with the size of the business. 6

7 deentraization redues output-depeting government interventions, we must assume that at east some interventions depete output. Thus, I assume that output in this oaity, Y, dereases in b : Y = ftb (, ) where f < 0 if b > 0 and f (0,0) > 0. 1 As in previous treatments, I start by assuming the government maximizes its tota revenues from taxation and bribery. Given t and r, the oa government sets b to maximize R = ( tr + b ) ftb (, ). Soving the mode is straightforward. The first order ondition yieds: ftb (, *) b* = tr f (, tb*) (1) where asterisks indiate equiibrium vaues. Now this might or might not represent a unique, interior maximum. However, to make the standard argument that higher r redues b * we must assume that it does. If, by ontrast, the ony equiibrium were a orner soution, then the bribe rate in this orner soution woud be independent of r. If there were mutipe equiibria, then it woud be unear how to draw any onusions from omparative statis. I therefore foow the iterature in assuming that ftb (, ) is a ontinuous, twie differentiabe, onave, dereasing funtion of t and b, and that ( tr+ b) ftb (, ) has a maximum at 0 < b* < 1 t (see Zhuravskaya 000, Jin et a. 005). 13 From (1), it is ear that inreasing r inreases equiibrium f. Totay differentiating (1) with regard to r, we get: b *(, tr) f = t < 0. So r f ff f b b * > 0. This is the standard argument in r a nutshe. Inreasing r dereases the oa government s equiibrium bribe rate, whih inreases equiibrium output. Fisa deentraization, by strengthening the inentives for oa governments to support oa business, improves eonomi performane. 1 In reaity, output might not be ontinuous in b, in whih ase there might be no equiibrium. This woud be another reason to doubt the generaity of the argument I am ritiquing. Subsripts on funtions denote derivatives. 13 The seond order ondition is f + ( tr + b*) f 0, whih is aways met given that f is dereasing and onave. 7

8 . Adding a entra government However, the mode makes one key omission: it does not inude a strategi entra government. Suppose instead that the entra government an aso extrat a rate of bribes, 0 b 1 t, from oa business and that it ats strategiay. To maintain the parae as strity as possib e, suppose output in the oaity, Y = ftbb (,, ), dereases onavey in both the entra and oa bribe rates, and that the entra government aso seeks to maximize its tota revenues, R = ((1 t r) + b ) ftbb (,, ). The oa government s maximand is now R = ( tr+ b ) f (, t b, b ). It is most reaisti to suppose the two governments set their bribe rates simutaneousy otherwise we woud have to beieve one oud ommit to a partiuar bribe rate, whih given the verifiation diffiuties seems unikey and so the appropriate soution onept is Cournot-Nash. It seems reasonabe to assume that if b + b + t 1, f = 0 (i.e., if governments together extrat 100 perent of output, no one wi produe), so this wi never be the ase in equiibrium. The first order onditions beome ftb (, *, b) b* = tr f (, tb*, b ) and f(, t b, b*) b* = t(1 r) () f (, tbb, *) 3 These define the reation funtions of the two governments, b *( btr ;,) and b *( btr ;,). 14 To avoid the trivia ases in whih one government s intervention drives the other ompetey out of the bribe market or in whih both wish to inrease their interventions without bound, I make ertain tehnia 1 assumptions ommon in simiar probems in industria organization: 0 < b *(0; tr,) < b * (0; tr, ); 1 0 < b *(0; tr,) < b * (0; tr, ); 3 3 f f ff < f ff ; and f f ff < f ff. The first two state that the bribe rate one government woud set if its riva set a rate of zero is positive and ower than the rate it woud have to set to drive its riva down to a zero bribe rate. The seond two require that 14 Sine f is onavey dereasing in both b and b, the seond order onditions, f + ( tr + b*) f 0 and f + ((1 t r) + b *) f 0, are aso met

9 the sopes (in absoute vaue) of the two reation urves are ess than one. 15 These aso ensure that the interior Nash equiibrium defined by the intersetion of the reation urves is unique. How do the equiibrium bribe rates, b * and b *, respond to hanges in r? One way to answer this woud be to totay differentiate both the first order onditions in () with regard to r, and then sove for the derivatives, db */ dr and db */ dr. An aternative method, whih heps to deveop intuitions, is to derive the partia derivatives b *( b ;,)/ t r r and b *( b ;,)/ t r r from (), hoding the other government s bribe rate onstant. These partia derivatives revea how the reation urves shift in response to hanges in r. One an then examine and demonstrate graphiay how the impied shifts woud affet the vaues of b * and b *, whih are defined by the intersetion of the reation urves. 16 Differentiating the reation urves with respet to r, we get: b *( b ;,)/ tr r= tf /( f ff ) < 0 and b *( b;,)/ t r r = tf /( f ff ) > 0. So, the oa government s reation urve, b *( btr ;,), shifts down in response to an inrease in r, whie the entra government s reation urve, b *( btr ;,), shifts up. The effet on equiibrium bribery then depends on whether entra and oa bribery are strategi substitutes or ompements, as we as the urvature of f. If entra and oa bribery are strategi substitutes (the reation urves sope down, and f f 3 ff 3 0 > ), inreasing r has opposite effets on b * and b * : it redues the equiibrium oa bribe rate, but simutaneousy inreases the entra bribe rate (see Figure 1a). The net effet on output is indeterminate. If, by ontrast, entra and oa bribery are strategi ompements, we annot say muh in genera about how equiibrium bribe rates woud hange in response to a hange in r: the ony 15 This aso rues out unstabe equiibria. If the absoute vaue sopes of the reation urves are greater than one, unstabe equiibria may sti exist. I disuss these in Author (005) and argue there that the net effet of hanging r wi sti be indeterminate. 16 This aso permits anaogies to reation urve anaysis in industria organization. 9

10 ombination of hanges that is rued out is an inrease in b * and a derease in b * (see Figure 1b). Equiibrium output might either rise, fa, or stay the same. 17 In short, without speifying more about the output funtion, there is no reason to think the inentive effets of fisa deentraization woud generay derease, rather than inrease, the aggregate bribe burden. Indeed, in one interesting and quite pausibe speia ase, the effets of fisa deentraization on entra government behavior preisey offset the effets on oa governments. So far I have assumed the redued form output funtion Y = ftbb (,, ), foowing the previous iterature. However, it is worth onsidering an exampe with a itte more struture. Suppose output is a onavey inreasing funtion of the abor suppied by the oaity s resident: Y = f (); f ' > 0, f '' < 0. As before, the governments share revenues from a tax on output with fixed rate, t, and eah an extrat bribes proportiona to output. The resident onsumes the remaining output. Taking the tax and bribe rates as given, the resident aoates a fixed time budget, normaized to one, between abor and eisure to maximize his utiity: U = ( 1 ( t + b + b )) f() + v(1 ), where ( ) 1 ( t + b + b ) f () is the resident s onsumption and v() is a onavey inreasing funtion measuring the vaue he paes on eisure. The first order ondition ( ) 1 ( t + b + b ) f '( *) v'(1 *) = 0 impiity defines equiibrium abor suppy, *( T ), as a dereasing funtion of the sum of the bribe and tax rates, T t + b + b. 18 The respetive governments simutaneousy hoose bribe rates to maximize R = ((1 t r) + b ) f( *( T)) and R = ( tr+ b ) f (*( T)). Their first order onditions are: f (*) + ( tr+ b ) f '( *) *'( T) = 0 and f (*) + ((1 t r) + b ) f '( *) *'( T) = 0, whih when ombined define the equiibrium tota tax and 17 The two governments ations are ikey to be strategi substitutes when, as in this exampe, the governments intervene by extrating bribes. An inrease in b wi derease f for a given b, whih one woud expet to inrease f, impying f < 0, and therefore f f ff > 0, the ondition for downward soping reation urves / b However, if governments intervene by providing infrastruture, the governments ations are ikey to be strategi ompements if entra and oa infrastruture are ompementary. 18 It is easy to see the seond order ondition wi aways be met. 10

11 Figure 1: Effets of fisa deentraization on oa and entra bribe rates b b *( b; r ) 0 E 0 E 1 b *( b ; r) 0 b 1. a Strategi Substitutes: b * dereases, b * inreases b b *( b; r ) 0 b *( b ; r) 0 E 0 E 1 b 1. b Strategi Compements: b *, b * may inrease, derease, or stay the same 11

12 bribe rate, * T, in a way that is independent of r: T* '( T*) f ( T ( *)) / f' ( T ( *)) =. Thus equiibrium abor suppy and output are aso independent of r. The effet of r on the oa government s bribe rate is preisey offset by its effet on the entra government s rate. In genera, inreasing the oa tax share wi ony redue the aggregate bribe burden and inrease output if b *( b ; tr, )/ r = tf /( f ff ) is arge reative to b *( b;,)/ t r r = tf /( f ff ), whih impies that / f f must be sma reative to f 33 / f3. So if in the initia equiibrium output is more sensitive to oa than to entra bribery and ess onave in the oa than in the entra bribe rate, inreasing the oa tax share wi inrease output. The basi indeterminay turns out to be robust to a variety of reformuations and hanges of assumptions (Author 005). First, governments might intervene in different ways, extrating bribes from firms (disouraging prodution), or providing business infrastruture (enouraging it). They might demand ump-sum bribes, the same for a, or payments proportiona to the firm s revenues. 19 They might provide pubi goods or onsume a revenues themseves. In a these variations, the effet of fisa deentraization remains indeterminate. Seond, one might mode governments with a variety of objetives from predatory revenue maximization to pure benevoene. For ease of exposition, I mode governments here as Leviathans, maximizing their revenues. But I show in Author (005) that the indeterminay remains even if governments are onsidered partiay or, under some assumptions, whoy benevoent, providing pubi goods at rates demanded by their residents. Loa governments might be more aountabe than entra ones beause fewer issues are bunded together in a oa eetion or beause yardstik ompetition faiitates effiieny omparisons (Bardhan and Mookherjee 000, Seabright 1996, Besey and Case 1995). Government payoffs might weight bribes more or ess highy than tax revenues. (Bribe revenues, sine unreorded, an be spent with greater disretion; on the other hand, they ome with a risk of detetion and proseution.) Varying the governments preferenes in these 19 One must assume ump -sum bribes are distortionary, or the argument that reduing bribery inreases output makes no sense. Athough ump -sum evies are often thought non-distortionary, paying a ump -sum bribe might impose indiret osts if it requires firms to engage in seretive behavior (Sheifer and Vishny 1993). 1

13 ways aso does not hange the resut. Third, different types of government intervention might have different eonomi effets. Bribery may be more harmfu than taxation beause of the osty serey it requires (Sheifer and Vishny 1993). Loa governments might be abe to invest in infrastruture more produtivey than entra government beause of oa knowedge or other oa advantages. Sine I assume ony that output dereases onavey in the tax and the bribe rates and inreases onavey in both oa and entra infrastruture, this woud not affet the onusions. I assume an exogenousy fixed tax rate, but endogenizing this wi not neessariy remove the indeterminay. Finay, oa governments may be more onstrained by fear of osing mobie apita than entra governments (Brennan and Buhanan 1980, Qian and Roand 1998). I show that even assuming this, the effet of a arger oa revenue share on output remains unear. 0 3 Tax sharing versus tax assignment Some shoars have argued that fewer inentive probems arise if different eves of government are assigned separate tax bases rather than reeiving a share of a joint tax (e.g., Sheifer and Treisman 000, Lavrov et a. 000). If a government hods exusive rights to a partiuar tax base, it wi simpy set its most preferred tax rate and have no need for bribery. If oetion and enforement are perfet, and poiies of eah eve of government affet ony that government s tax base, this wi, indeed, eiminate the probems this paper addresses. A perhaps more interesting question onerns settings where enforement is imperfet or effets of poiies spi over. Wi assigning separate tax bases in toto to different governments ead to better inentives and eonomi performane when governments an reguate or extrat bribes from eah other s tax-payers or when poiies have indiret effets? In pratie, suh overap seems hard to avoid. Export duties may arue to the federa budget. But the reguations of regiona governments on abor onditions, the environment, minimum wages, and so 0 Some suggest that entra governments may suffer from greater ageny probems in monitoring and motivating bureaurats (Bardhan and Mookherjee 000). Suh ageny probems may indeed be important. To fous on the fisa inentives argument, I abstrat from this one here. 13

14 on wi affet the profitabiity of exports in numerous ways. Thus, exporters wi have an inentive to bribe or obby regiona governments to waive requirements. Residentia rea estate taxes may be assigned entirey to oa governments. But federa aws on everything from disrimination to mortgages affet the prie that house seers an get. Whie overapping taxation and bribery may mosty be a probem in deveoping ountries, overapping effets of infrastruture investments may pay an anaogous roe in deveoped ones. Centra and oa government infrastruture investments are both ikey to affet many tax bases, regardess of whih eve of government has the right to tax them. In suh ases, assigning partiuar tax bases to different eves of government wi not avoid the externaities that render tax sharing probemati. Consider a simpe exampe with externaities in infrastruture investment rather than bribery. Suppose a given oaity ontains two non-overapping, immobie tax bases, α and β. One might think of these as stoks of fixed apita tied to partiuar setors say, automobie fatories and farmand. Under tax assignment, the entra government has the right to tax output produed using α, and keeps a the revenue, whie the oa government taxes and keeps a revenue from output produed using β. Denote the enter s tax rate t α and the oa government s rate t β. Centra and oa governments an aso make infrastruture investments, I and I respetivey, in, say, interregiona and oa roads. I assume that output produed with eah tax base dereases onavey in the tax rate and inreases onavey in the eves of both oa and entra infrastruture investment (better roads inrease the produtivity of both automobie prodution and farming.) Output subjet to entra taxation is Y = α f( t, I, I ) and output subjet to oa taxation is Y = β gt (, I, I ), where α β f, f, f, f, g, g, g, g < 0; f, f, g, g > 0; f f f 0; g g g 0. I assume that both tax bases are fuy empoyed eah period. As before, I suppose that the two governments set their tax and infrastruture investment rates simutaneousy, and ook for Nash equiibria. 14

15 Under tax assignment, the entra government sets t α and I to maximize tαα f( tα, I, I) I. The oa government sets t β and I to maximize tββ gt ( β, I, I) I. Obviousy, infrastruture investments wi be set beow the eve that woud maximize the sum of the two governments objetives. 1 This joint optimum is desribed by the first order onditions for infrastruture investments: Joint optimum: I : t f ( t, I, I ) + t g ( t, I, I ) = 1 j j αα α ββ β I : t f ( t, I, I ) + t g ( t, I, I ) = 1 j j αα 3 α ββ 3 β where the supersripts stand for joint optimum. Under tax assignment, when governments set their infrastruture investments non-ooperativey, the orresponding first order onditions are: Tax assignment I : t g ( t, I, I ) = 1 a ββ β I : t f ( t, I, I ) = 1 a αα 3 α where the supersripts stand for assignment. Ceary, for given tax rates, I a j < I and I a j < I. To ompare these outomes to those under tax sharing, suppose now that a tax rate t α is evied on output produed with base α, a tax rate t β is evied on output produed with base β, and revenues are divided so that the oa government reeives a share r. Now the oa government hooses I to maximize rt [ αα f( tα, I, I) + tββgt ( β, I, I)] I, and the entra government hooses I to maximize (1 r)[ t α f( t, I, I ) + t βgt (, I, I )] I. First order onditions for infrastruture investments are: α α β β Tax sharing I : t f ( t, I, I ) + t g ( t, I, I ) = 1/ r s s αα α ββ β I : t f ( t, I, I ) + t g ( t, I, I ) = 1/(1 r) s s αα 3 α ββ 3 β where the supersripts stand for sharing. We see two effets. First, under tax sharing the eft-hand side of eah equation has one more term than under tax assignment, whih in this ase reates an inentive for higher infrastruture investment. Beause both governments have some stake in both tax bases, this inines them to invest more in infrastruture than if they ony benefited from one. This effet pushes the 1 And the revenue-maximizing eve of infrastruture investment is itsef beow the soia optimum, sine the government in this simpe mode ignores the effet of infrastruture on private inomes. 15

16 outome under tax sharing oser to the joint optimum. However, the seond effet pus in the opposite diretion. The right-hand side of both equations is smaer under tax sharing than tax assignment. Beause eah government gets ony a fration of the benefit from its infrastruture investment, this reates an inentive to invest ess. Putting the two effets together, under tax sharing, a government that gets a arge revenue share is ikey to invest more in infrastruture than it woud under tax assignment. But a government that gets a sma share wi tend to invest ess. Given suh externaities, it is not possibe to say in genera whether tax assignment wi ead to a higher or ower eve of tota infrastruture. Depending on the partiuar output funtions, either may our. A simiar mode an be used to make the same point in the ontext of bribery. Athough there may be other reasons to favor tax assignment, given the diffiuty of isoating tax bases in pratie it is not ear that this wi generay improve inentives. 4 Conusion That government predation and orruption are major obstaes to growth in deveoping ountries is widey reognized. Reent iterature suggests a growing appreiation of the dangers of suh predation at oa and regiona eves (Bardhan and Mookerjee 000, Tanzi 1995, Prud homme 1995). Fisa deentraization in tax sharing systems has seemed to some a way to redue the appetite of oa governments for bribes and embezzed funds and to motivate them to invest in pubi infrastruture. The arge share of margina tax revenues retained by subnationa governments in China is said to expain why its eonomi performane has in the past ompared so favoraby to that of Russia. On examination, however, there is no ear theoretia foundation to expet that deentraizing revenues in this way wi generay improve eonomi performane. Inreasing the margina revenue share retained by oa governments may render them ess predatory. But one the responses of the entra government are taken into aount, the net effet might be negative rather than positive. As inentives It is straightforward to show that whether under tax sharing the tax rates are set by the oa government, the entra government, or by agreement between the two, they wi be set at the revenue-maximizing eves that are aso hosen under tax assignment. 16

17 improve in the oaities, they deteriorate at the enter. Even if oa governments are assumed to be more aountabe, honest, onstrained by ompetition, or effiient at providing infrastruture, deentraization has no determinate net effet. There are ertainy ases in whih a higher oa tax share eads to higher output; but these are speia ases, whih depend upon partiuar parameter vaues or funtiona forms. They are not a vaid basis for genera onusions or poiy reommendations. It might be possibe to redisover desirabe inentive effets of arger oa tax shares in a dynami setting, but it is not obvious how. The same issue of mutuay offsetting oa and entra behaviors shoud sti arise. It is definitey possibe to regenerate a deentraization benefit by means of ad ho assumptions. If entra offiias are assumed to be onstrained from responding to fisa inentives, then deentraization may have positive net effets. But why woud entra offiias be non-strategi or unabe to respond to fisa inentives? To date, no ompeing reason has been suggested. 3 That theory does not yied genera preditions does not, of ourse, mean that the standard argument about fisa inentives annot hep to aount for partiuar empiria ases. It may be true that provinia tax retention rates in China hep expain why some provines grew faster than others in the 1980s (athough the system hanged in 1994 from one based on fisa ontrats to one based mosty on tax sharing with no ear drop-off in market-friendy provinia poiies or growth.) It may aso be that weak fisa inentives hep expain Russia s eonomi depression in the 1990s (athough growth has surged sine 1999, argey fueed by high oi pries, despite ontinuing tax sharing and weak oa inentives.) The papers by Jin et a. (005) and Zhuravskaya (000) are primariy empiria, and they provide onsiderabe evidene for the aims they make. However, both suggest that broader essons an be derived from the ases they examine. Jin et a. (005, p.1740) argue that their perspetive yieds an important insight about how federaism works in promoting eonomi deveopment; namey, there exists a positive reationship between the strength of fisa inentives faed by ower-eve governments and oa eonomi performane. Countries with strong fisa inentives for oa governments are expeted 3 Some might argue that orruption is a greater probem at the oa eve beause fae-to-fae interations with oa offiias are more ommon. Even if this were true whih is open to question it woud not hange the resut (see Author (005) for a fuer demonstration). 17

18 to experiene heathier oa business deveopment whie those with ow fisa inentives are expeted to experiene the opposite. Zhuravskaya (000, pp.338, 365) ontends that her onusions aso have impiations for the theory of deentraization and that In a system with stronger oa fisa inentives, one shoud observe more benign reguation, and higher growth ompared to a system with weaker fisa inentives. Neither of these generaizations hods up if one takes into aount the ikey responses of entra government as we as oa governments. The modes examined here are extremey sparse. Severa assumptions, reprodued here beause they are ommon to the iterature, woud be interesting to reax. One perhaps questionabe feature is the treatment of governments as unitary ators. In fat, disorganization or heks-and-baanes within governments at eah eve may aso affet orruption. A seond feature worth note is the treatment of r, the division of revenues between eves, as exogenous. One might endogenize this in various ways. The forma division of revenues might be modeed as the outome of bargaining between oa and entra governments, or as the resut of deisionmaking in a egisature (in whih ase muh woud depend on the rues of proedure). However, the atua division of revenues may differ from the forma division beause of tax evasion. If oa governments oude with oa tax-payers to onea inome that woud otherwise be taxed by the entra authorities, the atua division may even be a funtion of the forma sharing rate. Finay, it is important to reiterate that this paper evauates ony one argument about deentraization, and that others might be more persuasive. I do not address here the arguments that ompetition between oa governments redues bribery, embezzement, and waste, or that deentraization enabes governments to better math bundes of pubi goods to itizen demands or to monitor bureaurats more effetivey. Many arguments for and against different types of deentraization exist besides the one examined here. 18

19 Referenes Ahmad, Nuzhat and Syed Ashraf Wasti Pakistan, in Yun-Hwan Kim and Pau Smoke, eds., Intergovernmenta Fisa Transfers in Asia: Current Pratie and Chaenges for the Future, Mania: Asian Deveopment Bank, pp Author 005. Bardhan, Pranab and Diip Mookerjee Corruption and Deentraization of Infrastruture Deivery in Deveoping Countries, Boston University, unpubished. Besey, Timothy and Anne Case Inumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstik Competition, Amerian Eonomi Review, 85, 1, Marh, pp.5-45 Banhard, Oivier and Andrei Sheifer Federaism With and Without Poitia Centraization: China versus Russia, NBER Working Paper Boadway, Robin and Mihae Keen Effiieny and the optima diretion of federa-state transfers, Internationa Tax and Pubi Finane, 3, pp Boadway, Robin, Maurie Marhand, and Marianne Vigneaut The onsequenes of overapping tax bases for redistribution and pubi spending in a federation, Journa of Pubi Eonomis, 68, pp Bordignon, Massimo, Paoo Manasse, and Guido Tabeini Optima Regiona Redistribution Under Asymmetri Information, Amerian Eonomi Review, 91, 3, June, pp Brennan, Geoffrey and James M. Buhanan The Power to Tax: Anaytia Foundations of a Fisa Constitution, New York: Cambridge University Press. Daba-Norris, Era and Pau Wade. 00. The Chaenge of Fisa Deentraization in Transition Countries, Washington DC: IMF, WP/0/103. Dahby, Bevery Fisa externaities and the design of intergovernmenta grants, Internationa Tax and Pubi Finane, 3, pp Dahby, Bevery and Leonard S. Wison Vertia fisa externaities in a federation, Journa of Pubi Eonomis, 87, pp Ebe, Robert D. and Serdar Yimaz. 00. On the Measurement and Impat of Fisa Deentraization, Washington: Word Bank. The Eonomist Does Britain Want To Be Rih? Apri 14-0, pp Edwards, Jeremy and Mihae Keen Tax Competition and Leviathan, European Eonomi Review, 40, pp Fowers, Mariyn R Shared Tax soures in a Leviathan mode of federaism, Pubi Finane Quartery, 16, 1, Janurary, pp

20 Gordon, Roger H. and David D. Li Taxes and Government Inentives: Eastern Europe vs. China, Wiiam Davidson Institute: Working Paper No.56. Jin, Hehui, Yingyi Qian and Barry R. Weingast Regiona Deentraization and Fisa Inentives: Federaism, Chinese Stye, Journa of Pubi Eonomis, 89, pp Keen, Mihae Vertia Tax Externaities in the Theory of Fisa Federaism, IMF Staff Papers, 45, 3, September, pp Kisubi, Mohammad M Invoving Civi Soiety in the Fight Against Corruption, in Rik Stapenhurst and Sahr J. Kpundeh, eds., Curbing Corruption: Toward a Mode for Buiding Nationa Integrity, Washington DC: Word Bank, pp Lavrov, A., J. Litwak and D. Sutherand Fisa Federaist Reations in Russia: A Case for Subnationa Autonomy, OECD: unpubished paper. Maomson, James M Some Anaytis of the Laffer Curve, Journa of Pubi Eonomis, 9, pp Oates, Waae E An Essay on Fisa Federaism, Journa of Eonomi Literature, XXXVII, September, OECD, Taxing Powers of State and Loa Government, Paris, OECD, Prud homme, Remy On the Dangers of Deentraization, Word Bank Researh Observer, 10,, pp Qian, Yingyi and Gérard Roand Federaism and the Soft Budget Constraint, Amerian Eonomi Review, Deember, 88, 5, pp Rodden, Jonathan, Gunnar S. Eskeand, and Jennie Litvak, eds Fisa Deentraization and the Chaenge of Hard Budget Constraints, Cambridge MA: MIT Press. Roand, Gérard Transition and Eonomis: Poitis, Markets, and Firms, Cambridge MA: MIT Press. Seabright, Pau Aountabiity and Deentraization in Government: An Inompete Contrats Mode, European Eonomi Review, 40. Shah, Anwar Indonesia and Pakistan: Fisa Deentraization an Eusive Goa? in Rihard M. Bird and Franois Vaianourt, eds., Fisa Deentraization in Deveoping Countries, New York: Cambridge University Press, pp Sheifer, Andrei and Danie Treisman Without a Map: Poitia Tatis and Eonomi Reform in Russia, Cambridge: MIT Press. Tanzi, Vito Fisa Federaism and Deentraization: A Review of Some Effiieny and Maroeonomi Aspets, in Word Bank, Annua Word Bank Conferene on Deveopment Eonomis 1995, Washington DC: Word Bank. Ter-Minassian, Teresa, ed Fisa Federaism in Theory and Pratie, Washington, DC: IMF. 0

21 Tiebout, Chares A Pure Theory of Loa Expenditures, Journa of Poitia Eonomy, 64, Otober, Turkish State Institute of Statistis Statistia Yearbook of Turkey, Istanbu: State Institute of Statistis. Widasin, David E Externaities and Baiouts: Hard and Soft Budget Constraints in Intergovernmenta Fisa Reations, Washington: Word Bank, Working Paper Widasin, David E. and John D. Wison Capita tax ompetition: boon or bane? Journa of Pubi Eonomis, 88, 6, June, pp Wrede Tragedy of the fisa ommon?: Fisa stok externaities in a Leviathan mode of federaism, Pubi Choie, 101, pp Zhuravskaya, Ekaterina V Inentives to Provide Loa Pubi Goods: Fisa Federaism, Russian Stye, Journa of Pubi Eonomis, 76, 3, June, pp