Are we getting value for our tax money? Improving the transparency of subnational government performance

Transcription

1 Are we gettng value for our tax money? Improvng the transparency of subnatonal government performance Maarten A. Allers Center for Research on Local Government Economcs (COELO) Unversty of Gronngen The Netherlands March, 2014 Abstract Ctzens pay taxes n order to enoy publc servces. But because they do not know the publc producton functon, t s hard for them to assess whether they are gettng value for money. Poltcal yardstck competton, based on a comparson of publc servces and tax rates wth those n nearby ursdctons, can provde voters wth a useful nstrument to help solve ths asymmetrc nformaton problem. However, s has been shown that fscal dspartes bas ths yardstck. A poltcan n a fscally advantaged ursdcton can perform badly and stll compare favorably, even f hs neghbors perform well. An ncumbent n a fscally dsadvantaged ursdcton may be unable to avod a bad reputaton, even when performng well. Ths paper derves the characterstcs of a fscal equalzaton scheme that removes ths yardstck bas. It turns out that currently used fscal equalzaton systems do not remove the yardstck bas except under restrctve assumptons. Keywords: rent-seekng, yardstck competton, fscal dspartes, equalzaton, transparency JEL classfcaton: D72, H77 Acknowledgements The author thanks partcpants of the 2013 Copenhagen workshop on Interacton between local expendture responsbltes and local tax polcy, and n partcular Bernard Dafflon. Beuw Geertsema and Jacob Veenstra gave useful comments on an earler verson of ths paper. 1

2 1 Introducton Wth some exceptons (e.g., natonal securty), the provson of publc servces s best left to subnatonal governments. An mportant reason for ths s that these can talor publc servces to local needs (Oates, 1999). In every ursdcton, local ctzens can then decde how much tax money they are wllng to pay to receve publc servces. Another advantage s that subnatonal voters can compare ther own ursdcton s tax rates and publc servce levels wth those of nearby ursdctons (yardstck competton). By comparng ther ncumbent s performance wth the performance of admnstrators n smlar ursdctons, voters can re-elect good poltcans and send non-performers packng. Ths n turn gves admnstrators an ncentve to perform better. Decentralzaton of government, however, creates the problem of fscal dspartes. In order to provde a certan servce level, some subnatonal governments must, for reasons outsde ther control, spend more money per nhabtant than others. The frst reason for ths s that the demand for certan servces may dffer. In some communtes, e.g., the proporton of schoolchldren s hgher than elsewhere. The second reason s that, because of adverse geography, geology, clmate, etcetera, some servces are more costly to produce n some regons than n others. Publc transport, e.g., wll be more costly n mountanous areas. For both these reasons, spendng needs of subnatonal governments may dffer sgnfcantly. On the other hand, the ablty to rase revenues may dffer as well. Some ursdctons have an affluent populaton and many successful busnesses. In ths case, low tax rates suffce to generate substantal revenues. 1 Jursdctons wth a lower revenue capacty need hgher tax rates n order to keep up. It has been shown that fscal dspartes make t dffcult to compare the performance of local governments (Allers, 2012). Poltcans n dsadvantaged ursdctons seem to perform worse than they actually do, whle the performance of poltcans n ursdctons wth low costs or a hgh revenue capacty s overestmated. Ths makes yardstck competton based. 1 Revenue capacty may nclude other ncome sources besdes taxaton, whch wll be gnored n ths paper. 2

3 In many countres, fscal dspartes are equalzed to some extent through a system of ntergovernmental grants. Tradtonally, equalzaton s advocated on the grounds that t mproves locatonal effcency, as t removes an ncentve to move to ursdctons wth favorable fscal condtons (Buchanan, 1950, 1952; Buchanan and Goetz, 1972; Boadway and Flatters, 1982); on equty grounds (Le Grand, 1975; Bramley, 1990; Cappelen and Tungodden, 2007); or as an nsurance aganst regonal shocks (Bucovetsky, 1998; Von Hagen, 2006; Konrad and Setz, 2003). 2 Allers (2012) argues that a case can be made for equalzaton n order to mprove the decson-makng process of subnatonal governments. If fscal dspartes are equalzed to the extent that every ursdcton s able to provde the same servce level at the same tax sacrfce, subnatonal government output levels, combned wth tax rates, provde an unbased ndcator of subnatonal government performance. Note that fscal equalzaton does not mean that all subnatonal governments wll have dentcal servce levels. Equalzaton apples to the capacty of subnatonal ursdctons to provde an attractve combnaton of servces and taxes. It s up to local admnstrators to use the avalable means effcently and effectvely. Voters can compare performance n dfferent ursdctons n order to assess ther elected admnstrators (yardstck competton). In the presence of fscal dspartes, yardstck competton s hampered by the fact that rent-takng poltcans n ursdctons wth a large revenue capacty relatve to spendng needs are less lkely to be found out, whereas admnstrators who do not take rent may stll compare unfavorably f ther ursdcton suffers from adverse crcumstances. The mpact of fscal dspartes on accountablty has not yet attracted much analyss. Kotsoganns and Schwager (2008) argue that yardstck competton s more effectve f dfferences n revenue capactes are equalzed. However, ths s not because equalzaton helps voters to mprove ther estmate of ncumbents renttakng. On the contrary, n ther model, voters are not nterested n rent-takng: because there are only two perods, every admnstrator they choose after the frst perod wll take maxmum rent n the second. Ths paper studes possble remedes for the yardstck bas, n partcular equalzaton of revenue capacty and spendng need. The paper s organzed as follows. In secton 2, we dscuss the theoretcal background and related lterature. 2 For a revew of the arguments for equalzaton, see Boadway (2004; 2006). 3

4 We proceed by dervng an equalzng grant that would entrely elmnate the yardstck bas (secton 3). We demonstrate that equalzaton systems exstng n practce do not remove the bas n the yardstck, except under restrctve assumptons (secton 4). Secton 5 argues that the problems attached to a transparency-mprovng equalzaton system make t dffcult and costly to mplement n real-world crcumstances. As an alternatve, we suggest that nformaton on fscal dspartes could be made avalable to the publc, n order to allow voters to form a true pcture of ther admnstrators performance. Secton 6 summarzes and concludes. 2 Background The tradtonal arguments for equalzaton mplctly assume a benevolent government, whch ams to maxmze the electorate s welfare. The poltcal economy lterature challenges ths assumpton, and stresses that poltcans and bureaucrats maxmze ther own welfare nstead. Here, decentralzaton s often seen as a strategy to reduce the monopolstc character of government and therefore to mprove accountablty (e.g., Brennan and Buchanan, 1985). Accountablty can be defned as the extent to whch voters can hold ncumbents responsble for ther performance. Decentralzaton may ntroduce two forms of competton between subnatonal governments. The frst one works through moblty: competton for moble tax bases (fscal competton) or to avod hgh cost ctzens (welfare competton). The second form nvolves poltcs: competton for comparatve performance (yardstck competton). Voters have two optons when they are dssatsfed: voce or ext (Hrschman, 1970; Tebout, 1956). The most powerful way to voce dsapproval s through the vote. People can ether move away, or send the ncumbents packng. Competton based on the ext opton may lmt ncumbents freedom to collect rents (e.g., Edwards and Keen, 1996). The ext opton s characterzed by hgh transacton costs, as people have to fnd a new home, move house, and perhaps fnd a new ob. Therefore, ths opton only becomes attractve f dfferences between ursdctons performance are substantal. Wth only the ext opton, poltcans would have 4

5 consderable leeway. 3 Vote, on the other hand, s relatvely cheap. However, to be effectve, ths nstrument requres that voters are able to dentfy good poltcans, that s, poltcans who gve them value for ther tax money. If voters are able to dentfy good poltcans from bad ones, they can re-elect good ones and dsmss the bad ones. Besley and Smart (2007) call ths the selecton effect. Moreover, poltcans wll have an ncentve to perform well n order to be re-elected (ncentve effect). Because of asymmetrc nformaton, voters are usually unable, at reasonable cost, to determne how much servce an effcent government s able to supply at a gven tax rate (Bradford et al., 1969). Only the bureaucrats themselves know the governments producton functon. Because promses cannot be trusted n ths settng, past performance s the best ndcator of future performance (Downs, 1957). Retrospectve votng can remove poltcans who do not perform well from power. The problem, however, s to assess performance. As Salmon (1987) ponts out, n a world wth only one government, the only way to do ths s to compare government output and tax rates over tme. In a statonary world, ths could be suffcent. Of course, the world s n fact far from statonary. As a result of the frequent occurrence of exogenous shocks, output s an mperfect ndcator of performance. The retrospectve vote s a blunt nstrument. Ths changes fundamentally f government s decentralzed. If there are comparable ursdctons, subect to the same exogenous shocks, voters can use tax rates and servce levels n other ursdctons to create a yardstck for assessng the performance of ther admnstrators. Thus, decentralzaton may work as an ncentve scheme. If ncumbents try to compare favorably to admnstrators n other ursdctons, they engage n polcy competton. Ths poltcal yardstck competton may dscplne poltcans. Although ths has been recognzed by earler wrters (e.g., Parks and Ostrom, 1981), Salmon (1987) s the frst to systematcally descrbe ths mechansm. Several theoretcal papers study the effectveness of yardstck competton to mprove accountablty (e.g., Wrede, 2001; Bordgnon et al., 2004; Belleflamme and Hndrks, 2005; Besley and Smart, 3 Epple and Zelentz (1981) show that, even wth costless mgraton, ext wthout vote s nsuffcent to prevent ursdctons from exercsng monopoly power f ursdctonal boundares are fxed. Because land s mmoble, bureaucrats can share n the rents accrung to land. 5

6 2007). A steadly ncreasng number of emprcal studes confrm the occurrence of yardstck competton (e.g., Besley and Case, 1995; Bordgnon et al., 2003; Allers and Elhorst, 2005; Revell, 2006). Ths paper s concerned wth an aspect of yardstck competton that has receved lttle attenton. Yardstck competton needs the exstence of comparable ursdctons. However, ursdctons, even f they operate n the same nsttutonal settng, have the same servce responsbltes, and are susceptble to common exogenous shocks, dffer wth respect to fscal capacty and spendng need. In order for poltcal yardstck competton to work optmally, dfferences n subnatonal government output and tax rates should reflect only dfferences n polces, not fscal dspartes. It would be sub-optmal to punsh or to credt ncumbents for factors outsde ther control (Allers, 2012). We nvestgate what knd of equalzaton system would remove the transparency loss resultng from fscal dspartes. As t turns out, such a system does exst, but t s not normally used for equalzaton purposes. Therefore, we analyze the effects of two dfferent equalzaton schemes whch are actually used n varous countres. 3 Equalzaton and transparency Yardstck bas To model the way fscal dspartes bas yardstck competton, we buld on Allers (2012). There s a central government and there are two subnatonal ursdctons 4. Subnatonal ursdctons provde publc servces and fnance ths through tax revenues and, n the case of fscal equalzaton, equalzaton transfers. Publc servce levels are chosen at the subnatonal level. 5 The central government decdes whether equalzaton s appled, and how. Each ursdcton must balance ts budget. The ursdcton s budget constrant s E B G (1) where E s ursdcton s per capta expendtures; G ts per capta equalzaton grant; B the average per capta tax base; β the relatve per capta tax base, defned 4 Ths s for ease of exposton only. It s straghtforward to extend the analyss to a greater number of ursdctons. 5 In practce, the central government usually delegates some (or many) tasks to subnatonal governments. Here, we abstract from that. 6

7 B as, where B s the per capta tax base of ursdcton ; and θ the tax rate, B defned as the share of the tax base that the ursdcton collects (0 < θ <1). Thus, β B s ursdcton s per capta tax base, and θ β B s ts tax revenue. The admnstrator n knows β and B; voters do not. We assume that relatve tax bases are not affected by rent levels. 6 Box 1. Assumptons In order to keep the analyss smple, a number of assumptons have been made. Here s a lst of the most mportant ones. 1) Subnatonal government spendng s fnanced from own tax revenues and equalzaton grants (no borrowng). 2) Subnatonal taxes are borne by the ursdctons own resdents (no tax exportng). 3) Benefts of subnatonal publc servces are enoyed by resdents only (no spllovers). 4) Subnatonal governments choose local servce levels (no mandated tasks). 5) Poltcans strve for re-electon (no term lmts). 6) Subnatonal ursdctons face dentcal exogenous shocks. 7) Admnstrators n ursdcton observe B, β, γ and ρ, voters do not. Everyone observes S, S, and. Each ursdcton s governed by an elected poltcan. After beng elected, he or she chooses a tax rate (and thus an expendture level, see equaton 1), n a way that s exogenous to our model, and then a fracton ρ of publc expendtures that s extracted as rent (0 ρ <1). Admnstrators are assumed to strve for re-electon. Thus, we abstract from term lmts. Ths can be motvated by the fact that poltcans are often organzed n partes, whch can compete n electon after electon. Moreover, poltcans often try for a dfferent offce when a term lmt prevents them from beng re-elected. E.g., mayors may want to run for governor later n ther career. Lke re-electon, ths requres a good reputaton wth voters and support from ther party. As a result of common exogenous shocks ω to the economy, the servce level correspondng to a certan amount of spendng vares. As a result, past performance s a weak ndcator of future performance, whch lmts the usefulness of 6 Ths s probably a smplfcaton. There s some evdence (Hlber et al., 2011, Allers and Vermeulen, 2013), that servce levels and tax rates are captalzed nto property values. 7

8 retrospectve votng. Followng the lterature (e.g., Besley and Case, 1995), we assume that both ursdctons experence dentcal shocks. Apart from ω, the per capta servce level S depends on per capta spendng on the publc servce (1-ρ )E, and on spendng need, whch may be expressed as the ursdcton s need ndex γ : S ( 1 ) E. (2) γ reflects both demographc and other factors outsde the control of the subnatonal government that determne the amount of spendng on the publc servce needed to supply a certan servce level n ursdcton. Lke β, γ s expressed n relatve terms; γ > 0, wth average value one. The ncumbent knows γ ; voters do not. The exstng yardstck competton lterature generally assumes γ to be the same for all ursdctons (e.g., Besley and Case, 1995). That s clearly unrealstc. Table 1. Explanaton of symbols β γ λ π ρ θ θ ω B B E G S S t Jursdcton 's relatve per capta tax base Jursdcton 's need ndex Relatve fscal advantage of ursdcton Relatve performance yardstck used by voters to udge the ncumbent of Rent as fracton of ursdcton 's spendng Natonwde standard tax rate Jursdcton 's tax rate Common exogenous shock Average per capta tax base Jursdcton 's per capta tax base Jursdcton 's per capta expendtures Jursdcton 's per capta equalzaton grant Natonwde standard servce level Jursdcton 's per capta servce level Equalzaton rate Voters do not observe ρ. Instead, they observe servce levels and tax rates. Voters value hgh servce levels and low tax rates. They maxmze value for money: the rato of servces provded to tax sacrfce S. Regularly, voters choose a poltcan to govern ther ursdcton. They ether re-elect the ncumbent, or elect a challenger. Voters use a relatve performance yardstck π to udge the ncumbent. If π > 1, ursdcton s ncumbent s performance s consdered superor to that of hs or her counterpart n the other ursdcton. If π < 1, s ncumbent s consdered 8

9 nferor. Voters base ther udgment on an ncumbents performance n the entre perod snce the prevous electons. 7 Gven voters preferences, the benchmark for ursdcton s ncumbent s relatve performance π s other ursdcton: S, value for money, relatve to the correspondng rato n the where S (3) S. Substtutng (2) n (3), the performance benchmark becomes (1 ) E. (4) E (1 ) Note that ω s cancelled out of the equaton. That s because we have made the assumpton that ursdctons and experence dentcal shocks. It s convenent to defne. Ths s the relatve fscal advantage of ursdcton, compared wth that of ursdcton. Substtutng (1) n (4), settng G to zero as there s as yet no equalzaton, yelds (1 ). (5) (1 ) Equaton 5 demonstrates that the relatve fscal advantage of a ursdcton,, affects the voters udgment of hs or her performance. Now consder the case wthout fscal dspartes: λ =1. The performance yardstck (5) reduces to 7 If voters only take the stuaton nto account as t s durng electon tme, a poltcal busness cycle may result. 9

10 * (1 ), (6) (1 ) where π * s the yardstck wthout fscal dspartes. Voters approve of ncumbent f * * 1. It follows from (6) that ths requres ρ < ρ. Thus, the yardstck gves a true pcture of the ncumbent s relatve performance. Poltcans can mprove ther chances to be re-elected by takng less rent. Equalzaton to remove the yardstck bas In order to remove the bas n the comparatve performance yardstck, we need an equalzaton grant that ensures that the yardstck used by voters (Equaton 4) equals the optmal yardstck (Equaton 6). Ths requres E E. (7) The equalzaton grant s by defnton equal to expendtures mnus tax revenue (see Equaton 1). Combnng ths wth the expendture level derved from (7) yelds G E B. However, n such an equalzaton system, a grant would depend on tax and spendng levels chosen n the other ursdcton. Ths may be avoded by settng both elements of Equaton 7 equal to a common value: E E C. (8) Here, C can have any postve value, provded t s the same for all ursdctons. Condton (8) ensures the yardstck bas s zero. Combnng (8) and the budget constrant (1), the optmal equalzaton grant becomes G * ( C B) (9) It s easy to demonstrate how grants accordng to (9) affect the relatve performance yardstck. The budget constrant (1) can now be wrtten as E B ( C B). (10) Substtutng (10) n (4) yelds (1 ) C *. (11) (1 ) C 10

11 Thus, equalzaton accordng to (9) completely removes the yardstck bas caused by fscal dspartes. A specal case of the grant n (9), whch has been used n practce (Ladd and Ynger, 1994), s derved by settng C = B: G PEG E t B( ) t ( ). (12) where t (0 t 1) s the program s equalzaton rate: the extent to whch fscal dspartes are equalzed. Complete equalzaton s characterzed by t=1, but, n practce, t mght be lower. The grant n (12) s called a power equalzaton grant. Essentally, these are matchng grants, where the match rate depends on the ursdcton s relatve needs and on ts relatve tax capacty. If the need ndex matches the tax base ndex (γ = β ), the grant s zero. For ursdctons wth hgh needs relatve to tax capacty (γ > β ), the grant s postve, and hgher expendtures translate nto hgher grants n order to offset the negatve effect of hgh needs relatve to tax base. For ursdctons where γ < β, the grant s negatve, and more spendng leads to bgger negatve grants. 8 Thus, recpents can nfluence ther power equalzaton grant by changng spendng behavor. In practce, power equalzaton grants are not used to equalze fscal dspartes, although they are used sometmes to fnance specfc servces. 9 4 Exstng equalzaton systems We now analyze the effect on the relatve performance yardstck of two equalzaton approaches whch underpn many exstng equalzaton systems. 10 We wll see that these generally do not remove the yardstck bas caused by fscal dspartes. 8 Ths s the case where grants are transfers between subnatonal governments: G = -G (horzontal equalzaton). Grants may nstead be fnanced through tax revenues collected by the central government. In that case, for ursdctons where γ < β, the grant s lower than for ursdctons where γ > β. 9 Ladd and Ynger (1994) report the use by several Amercan states to help fnance local educaton. 10 Other equalzaton schemes are concevable. See Musgrave (1961) for a useful taxonomy. 11

12 Need-capacty equalzaton A relatvely ambtous equalzaton system ams at closng or narrowng the gap between spendng need and revenue capacty, both defned n absolute per capta terms (Bradbury et al., 1984). Ths obectve s at the root of the equalzaton schemes n, e.g., the UK, the Netherlands, and Australa. These countres have made consderable efforts to estmate both revenue capactes and spendng needs of subnatonal governments. Possbly the most ambtous program exsts n the Netherlands (Hugsloot, 2007), where equalzaton grants to muncpaltes are allocated usng no less than sxty dfferent local characterstcs. Moreover, the equalzaton system s assessed annually, and changes are made regularly. 11 Although fscal dspartes are estmated and used to base the equalzaton grant on, they are not common knowledge. The allocaton system of the equalzaton grant s complcated, and grants receved by local governments nclude non-equalzng parts. It would take a consderable effort to derve the relatve fscal poston of a partcular ursdcton. Therefore, we mantan our assumpton that voters do not know β or γ, even n the case of equalzaton. 12 Here, spendng need s the spendng necessary for a ursdcton to provde standard-qualty servces S, and equals Sγ, where γ s the ursdcton s need ndex, as before. S can be defned as the average per capta servce level. Revenue capacty s now defned as the tax revenue gven a standard or average tax rate, θ, and equals θβ B. A ursdcton s need-capacty grant dfference between spendng need and revenue capacty: 13 NCG G s then gven as (a fracton of) the G NCG t( S B). (13) 11 In Belgum, Swtzerland and elsewhere, much smpler versons are used, based on a few demographc or geographc characterstcs whch are not derved from an extensve study of spendng needs. 12 Even f voters would know β or γ, ths would not be enough to remove the yardstck bas, for they now would have to take the equalzaton grant nto account as well. The dfference between the actual equalzaton grant and the deal one from a transparency pont of vew s what determnes the yardstck bas now. We can safely assume that voters would generally not know ths. Emprcal evdence supports ths. Recall from secton 3 that Allers and Elhorst (2005) found that voters n the Netherlands seem to use raw tax and expendture levels to compare local government performance, wthout takng fscal dspartes nto account. 13 In the lterature, such grants are also known as foundaton grants. 12

13 Jursdctons wth favorable fscal condtons (revenue capacty exceedng spendng need) would receve a negatve grant, less favored ursdctons a postve grant. 14 Note that a ursdcton s need-capacty grant depends nether on ts expendture level nor on ts tax rate, but only on γ, β and on natonal standards. 15 Wth full equalzaton (t=1), ursdctons wshng to supply a standard servce level can do so by levyng the standard tax rate. 16 However, they are free to choose a dfferent servce level, and hgher or lower taxes to match. Thus, subnatonal government autonomy s preserved. Stated dfferently, servce capactes, not servce levels, are equalzed. Note, however, that f ursdctons choose to dffer from the standard tax rate, fscal dspartes wll not be completely equalzed, because the servce level ncrease that can be fnanced by rasng the tax rate depends on β /γ. Need-capacty equalzaton ensures every ursdcton can have standard servce levels at the standard tax rate. Ths s an mportant dfference wth power equalzaton, whch ensures equal servce levels at equal tax rates (see Cappelen and Tungodden (2007) for a comparson of both grants). From (1) and (13) we obtan the budget restrcton wth NCG-grants E ts B( t ) (14) To derve π, we substtute (14) n (4). After rearrangng, ths yelds (1 ) (1 ) ts B ts B t. (15) t Now, the yardstck bas, the quotent of the bracketed terms n (15), may or may not decrease wth ncreasng t, dependng on S, B, θ and θ. Consder the case where t=1. If tax rates are smlar, both and wll be small, neutralzng the bas caused by and S. In addton, B n the numerator wll tend to cancel 14 Negatve grants would only occur n the case where grants are transfers between subnatonal governments (horzontal equalzaton). Negatve grants may be avoded by fnancng grants through tax revenues collected by the central government. 15 If natonal standards are based on averages, an ndvdual ursdcton does exert some nfluence on them, dependng on ts share. Wth few ursdctons, ths share s large. 16 Ths s smlar to Boadway s (2004) untary state benchmark. 13

14 S out B n the denomnator. Indeed, f θ = θ = θ and t=1, then π n Equaton (15) s reduced to the optmal yardstck π * n (6). Thus, need-capacty equalzaton mght mprove transparency, but whether t actually does so s far from certan. That depends on (relatve) tax rates, and thus on ncumbents choces. 17 Revenue capacty equalzaton A less ambtous equalzaton scheme ams only at equal revenue capactes, gnorng dfferences n spendng need. Ths s used n, e.g., Canada. Reasons for not equalzng spendng need may be that spendng need s hard to estmate correctly, or the wsh to avod perverse ncentves of equalzaton. A common approach to measurng revenue capacty s the representatve tax system (RTS) developed by the Advsory Commsson on Intergovernmental Relatons (ACIR, 1962). Under ths approach, full equalzaton mples that ursdctons levyng the average tax rate have average per capta spendng power E: E B. (16) Jursdctons spendng E face the budged restrant E RTS G B. (17) Combnng (16) and (17), and ntroducng t as before, yelds the equalzaton grant under RTS: G RTS tb 1 ). (18) ( Ths s smlar to the equalzaton program analyzed by Kotsoganns and Schwager (2008). There are two dfferences compared wth the power equalzaton grant n (12). In de frst place, RTS G depends on θ not θ. Lke the NCG-grant, the RTS grant cannot be nfluenced by ndvdual ursdctons, except n an ndrect way by nfluencng natonal averages (θ and B). The second dfference s that spendng need dfferences are gnored n (18): the need ndex γ does not enter nto the equaton. 17 In our model, tax rates are exogenously determned. For future research, t could be nterestng to make tax rates endogenous. Eq. (15) shows that the choce of the tax rate would then nteract wth the choce of the rent level. 14

15 Combnng (1) and (18) yelds the budget restrcton wth RTS-grants E tb B( t). (19) Substtutng (19) n (4) yelds the comparatve performance yardstck: t (1 ) t (1 ) ( t ). (20) ( t ) Comparng (20) wth the performance yardstck under need-capacty equalzaton (Eq. 15), we see that only the frst terms between square brackets dffer. Consder the case wth t=1 (full equalzaton). Lke (15), (20) devates from the optmal yardstck n (6). The bas n yardstck (20) wll be zero only f the terms between square brackets cancel out. Ths s the case f θ = θ = θ and γ = γ. As we have seen, under need-capacty equalzaton, for the yardstck bas to be zero t suffces to have θ = θ = θ. The condton θ = θ = θ and γ = γ s unlkely to be met n practce. Lke needcapacty equalzaton, equalzaton of revenue capacty usng the RTS-approach results n a based yardstck. Whether ths bas s smaller than the bas wthout equalzaton (Equaton 5) depends on (relatve) tax rates. 5 Dscusson We have shown that the yardstck bas created by fscal dspartes may, n theory, be removed entrely through equalzaton. However, equalzaton may not be the deal nstrument. In the frst place, n the equalzaton system needed to remove the yardstck bas completely, grants depend, apart from tax capacty and spendng need, on recpents spendng. Ths s often not consdered desrable. Equalzaton systems where grants are ndependent from local spendng are avalable, e.g., the needcapacty system and the RTS system analyzed above. However, these do not remove the yardstck bas, except n specal cases. Secondly, equalzaton may create neffcences. Equalzaton of spendng need may lead to an neffcently large populaton n hgh-cost area s (e.g., Oakland, 15

16 1994). Grants must be fnanced through natonal taxaton, whch s usually dstortve, or they may be transfers between subnatonal governments, n whch case they dstort tax prces of local publc servces (Dafflon, 2007). Equalzng tax capacty, on the other hand, may elmnate or greatly reduce ursdcton s ncentves to attract or preserve ther tax base (e.g., Büttner, 2006). Fnally, even f an approprate equalzaton system wthout harmful sde effects could be found, poltcans would not necessarly mplement t. Several studes document poltcal nfluence on exstng ntergovernmental transfers (Kheman, 2007; Allers and Ishemo, 2011). In order to crcumvent the dsadvantages of equalzaton, yardstck bas may perhaps be reduced by mprovng nformaton avalablty nstead. Our analyss shows that, to ths end, voters need to know λ. If voters augment ther relatve performance ndcator (5) by dvdng t by λ, they obtan the unbased yardstck (6). The value of γ and β must be estmated for each ursdcton n order to establsh an equalzaton system amed at reducng dspartes n spendng need and revenue capacty. However, nstead of usng them to create an equalzaton system, they may be used to calculate λ for each ursdcton. In order to avod poltcal nfluences, ths should be done by ndependent authortes, or cvc organzatons whch use ndependent research nsttutes or unverstes to do the actual numbercrunchng. Whether voters would actually understand, trust and be able to use ths knd of nformaton effectvely s an open queston. In practce, voters may well contnue to base ther vote on ther own percepton of servce levels and tax rates. The avalable emprcal evdence seems to be lmted to Revell (2006), who fnds that the ntroducton, n the UK, of a natonal performance ndcator of locally provded socal servces, consderably reduced the degree to whch local ursdctons mmc the socal care polces of ther neghbors. Ths suggests that local admnstrators assume voters wll no longer look at neghbors to assess local government performance, but nstead use the newly ntroduced performance ndcator. Both solutons equalzaton and makng fscal dspartes or relatve performance known suffer from the practcal problem that fscal dspartes are hard to measure accurately. Especally spendng need s hard to quantfy satsfactorly (e.g., Duncan and Smth, 1996). 16

17 6 Conclusons Ctzens pay taxes n order to enoy publc servces. But because they do not know the publc producton functon, t s hard for them to assess whether they are gettng value for money. Increasngly, poltcal yardstck competton s seen as an nstrument helpng voters get a grp on elected admnstrators at relatvely low cost. By comparng ther ncumbent s performance wth the performance of admnstrators n smlar ursdctons, voters can derve nformaton helpng them to re-elect good poltcans and send non-performers packng. Ths n turn gves admnstrators an ncentve to perform better. The key to yardstck competton s transparency. If admnstrators performance cannot be derved from subnatonal government output and tax rates n a straghtforward manner, yardstck competton s lkely to be based. Ths s the case when fscal dspartes exst. Then, poltcans n ursdctons wth a large revenue capacty relatve to spendng needs can take more rent than ther counterparts n less favored fscal crcumstances, and stll keep a good reputaton. Admnstrators of ursdctons sufferng from adverse fscal crcumstances may acqure a bad reputaton even f they do not take any rent at all. We show how fscal dspartes bas the relatve performance yardstck avalable to voters, and how ths bas may be reduced or removed through fscal equalzaton. We also show, however, that equalzaton schemes exstng n practce are less successful n mprovng transparency. Although t s possble, n theory, to remove the yardstck bas entrely through equalzaton, the problems attached to ths remedy make t uncertan that ths wll ever be accomplshed satsfactorly. Moreover, even f ths would be feasble, the costs arsng from perverse ncentves may well exceed the benefts. Such costs have not stopped countres from ntroducng equalzaton systems amed at equty or effcency, however. When an equalzaton system s set up or when an exstng one s evaluated, the effects on transparency should at least be taken nto consderaton. If fscal dspartes can be dentfed, t may not be necessary to use them to set up an equalzaton system. They may nstead be used to provde voters wth a readymade relatve performance measure. In ths case, the remanng challenge would be 17

18 how to dssemnate ths nformaton n a way that would lead to actual and effectve use by voters. References Advsory Commsson on Intergovernmental Relatons (ACIR), Measures of State and Local Fscal Capacty and Tax Effort. Washngton, D.C. Allers M.A., 2012, Yardstck competton, fscal dspartes, and equalzaton, Economcs Letters, 117, 4-6. Allers, M.A., Elhorst, J.P., Tax Mmckng and Yardstck Competton Among Local Governments n the Netherlands. Internatonal Tax and Publc Fnance 12, Allers, M.A., Ishemo, L.J., Do formulas reduce poltcal nfluence on ntergovernmental grants? Evdence from Tanzana, Journal of Development Studes, 47, Allers, M.A., Vermeulen, W., Fscal Equalzaton and Captalzaton: Evdence from a Polcy Reform, CPB Dscusson Paper 245. Belleflamme, P., Hndrks, J., Yardstck competton and poltcal agency problems. Socal Choce and Welfare, Besley, T., Case A., Incumbent Behavor: Vote-Seekng, Tax-Settng, and Yardstck Competton. Amercan Economc Revew 85, Besley, T., Smart, M., Fscal Restrant and Voter Welfare. Journal of Publc Economcs, 91, Boadway, R., The theory and practce of equalzaton. CESfo economc studes, 50, Boadway, R., Intergovernmental redstrbutve transfers: effcency and equty. In: Ahmad, E., Broso, G. (Eds.), Handbook of Fscal Federalsm. Edward Elgar, Cheltenham. Boadway, R., Flatters, F., Effcency and equalzaton payments n a federal system of government: a synthess and extenson of resent results. Canadan Journal of Economcs, 15, Bordgnon, M., Cerngla, F., Revell, F., In Search of Yardstck Competton: A Spatal Analyss of Italan Muncpal Property Tax Settng. Journal of Urban Economcs, 54, Bordgnon, M., Cerngla, F., Revell, F., Yardstck Competton n Intergovernmental Relatonshps: Theory and Emprcal Predctons. Economcs Letters 83, Bradbury, K.L., Ladd, H.F., Perrault, M., Reschovsky, A., Ynger, J., State Ad to Offset Fscal Dspartes Across Communtes. Natonal Tax Journal 37,

19 Bradford, D.J., Malt, R.A., Oates, W.E., The rsng costs of local publc servces: Some evdence and reflectons. Natonal Tax Journal, 22, Bramley, G., Equalzaton Grants and Local Expendture Needs. The Prce of Equalty. Avebury, Aldershot. Brennan, G., Buchanan, J.M., The Reason of Rules. Consttutonal Poltcal Economy. Cambrdge Unversty Press, Cambrdge. Buchanan, J.M., Federalsm and fscal equty. The Amercan Economc Revew, 40, Buchanan, J.M., Federal grants and resource allocaton. The Journal of Poltcal Economy, 60, Buchanan, J., Goetz, C., Effcency lmts of fscal moblty: An assessment of the Tebout model. Journal of Publc Economcs 1, Bucovetsky, S., Federalsm, equalzaton and rsk averson. Journal of Publc Economcs, 67, Büttner, T., The ncentve effect of fscal equalzaton transfers on tax polcy. Journal of Publc Economcs, 90, Cappelen, A.W., Tungodden, B., 2007, Local autonomy and nterregonal equalty. Socal Choce and Welfare, 28, Dafflon, B., 2007, Fscal Capacty Equalzaton n Horzontal Fscal Equalzaton Programs, n: Robn Boadway and Anwar Shah (eds.), Intergovernmental Fscal Transfers: prncples and practce, The World Bank, Washngton, DC, pages Downs, A., An Economc Theory of Democracy. Harper & Row, New York. Duncan, A., Smth, P., Modellng local government budgetary choces under expendture lmtaton. Fscal Studes, 16, Edwards, J., Keen, M., Tax competton and Levathan. European Economc Revew, 40, Epple, D., Zelentz, A., The Implcatons of Competton among Jursdctons: Does Tebout Need Poltcs? Journal of Poltcal Economy, 89, Hlber, C.A.L., Lyytkänen, T., Vermeulen, W., Captalzaton of central government grants nto local house prces: Panel data evdence from England, Regonal Scence and Urban Economcs 41, pp Hrschman, A.O., Ext, voce and loyalty: Responses to declne n frms, organzatons and states. Harvard Unversty Press, Cambrdge, Mass. Hugsloot, P., The equalsed allocaton of local expendture needs n the Netherlands: an optmsed mxture of obectvty and poltcs. In: Km, J., Lotz, J. (Eds.), Measurng Local Government Expendture Needs: The Copenhagen Workshop The Korea Insttute of Publc Fnance and the Dansh Mnstry of Socal Welfare. 19

20 Kheman, S., The Poltcal Economy of Equalzaton Transfers. In: Martnez-Vazques, J., Searle, B. (Eds.), Fscal Equalzaton. Challenges n the Desgn of Intergovernmental Transfers. Sprnger, New York. Konrad, K.A.,. Setz, H., Fscal federalsm and rsk sharng n Germany: The role of sze dfferences. In: Cnossen, S., Snn, H.-W. (Eds.), Publc fnance and publc polcy n the new century. MIT Press, Cambrdge, MA. Kotsoganns, C., Schwager, R., Accountablty and Fscal Equalzaton. Journal of Publc Economcs, 92, Ladd, H.F., Ynger, J., The case for equalzng ad. Natonal Tax Journal, 47, Le Grand, J., Fscal equty and central government grants to local authortes. The Economc Journal, 85, Musgrave, R.A., Approaches to a Fscal Theory of Poltcal Federalsm, n: Publc Fnances: Needs, Sources, and Utlzaton, Prnceton Unversty Press. Oakland, W.H., Fscal equalzaton: an empty box? Natonal Tax Journal, 47, Oates, W.E., The Effects of Property Taxes and Local Publc Spendng on Property Values: An Emprcal Study of Tax Captalzaton and the Tebout Hypothess. The Journal of Poltcal Economy, 77, Oates, W.E., 1999, An Essay on Fscal Federalsm, Journal of Economc Lterature, 37, No. 3. (Sep., 1999), pp Parks, R.B., Ostrom, E., Complex models of urban servce systems. In: Clarck, T.N. (Ed.), Urban polcy analyss. Sage, Beverly Hlls and London. Revell, F., Performance Ratng and Yardstck Competton n Socal Servce Provson. Journal of Publc Economcs, 90, Salmon, P., Decentralsaton as an Incentve Scheme. Oxford Revew of Economc Polcy, 3, Tebout, C.M., A Pure Theory of Local Expendtures. Journal of Poltcal Economy, 64, Von Hagen, J., Achevng economc stablzaton by sharng rsk wthn countres. In: Boadway, R., Shah, A. (Eds.), Intergovernmental fscal transfers. Prncples and practce. The World Bank, Washngton, D.C. Wrede, M., Yardstck Competton to Tame the Levathan. European Journal of Poltcal Economy, 17,