Accountability and fiscal equalization

Accountability and fiscal equalizationChristos Kotsogiannis a,1,Robert Schwager b,⁎a Department of Economics,School of Business and Economics,University of Exeter,Streatham Court,Rennes Drive,Exeter EX44PU,England,UKb Georg-August-Universität Göttingen and Zentrum für Europäische Wirtschaftsforschung Mannheim,Platz der Göttinger Sieben 3,D-37073Göttingen,GermanyReceived 10November 2006;received in revised form 22December 2007;accepted 29December 2007Available online 11January 2008AbstractA common feature of multi-jurisdictional systems is equalization programs.The implementation of such programs,that is based on some measurement of sub-national fiscal capacity and effort,is particularly complex.Within a political economy model,this paper analyzes the impact of such systems on accountability,identifying a positive and a negative effect.The positive effect arises because with equalized fiscal resources,a consequence of equalization,citizens attach more importance to any remaining variation in public good supplies and so punish rent-taking more severely.This induces politicians to restrain themselves and so accountability improves.The negative effect arises because the complexity of such programs reduces the informational content of observed public good supplies.This introduces a perverse fiscal incentive that reduces accountability.Thus,the overall impact of equalization programs on accountability depends on the balance of these effects.©2008Elsevier B.V .All rights reserved.JEL classification:D72;H77Keywords:Equalization transfer programs;Rent-seeking;Yardstick fiscal competition;Fiscal federalism1.IntroductionIntergovernmental transfers form the cornerstone of sub-national government financing in most countries around the world.Though such transfers can take many forms,2there is one particular category of transfers that has attracted a lot of attention recently:equalization transfers.As with all intergovernmental transfers,the rationale behind equalization transfers is the need for a reduction in horizontal fiscal imbalances that may exist between lower-level jurisdictions in a given country.The reduction in fiscal disparities,achieved by monetary payments from jurisdictions with above-average fiscal capacity to jurisdictions with lower-than-average fiscal capacity,ensures that have-not jurisdictions have Journal of Public Economics 92(2008)2336–/locate/econbaseCorresponding author.Tel.:+49551397293.E-mail addresses:c.kotsogiannis@ (C.Kotsogiannis),rschwag@uni-goettingen.de (R.Schwager).1Tel.:+441392264500.2They can be,for instance,conditional,unconditional or matching.0047-2727/$-see front matter ©2008Elsevier B.V .All rights reserved.doi:10.1016/j.jpubeco.2007.12.013the necessary fiscal capacity to guarantee themselves the national average level of public services per resident without imposing higher than average tax rates.The design and implementation of equalization transfers is crucial for the potential equity of public sector services provision.Though such transfers have a clear equity objective,their implementation is bound to be complicated for three reasons.Firstly,there is the measurement of fiscal capacity typically defined as the potential revenues that can be obtained from the tax bases assigned to the lower-level governments if some average level of effort is applied to those tax bases.But to infer potential tax revenues or even tax bases from observing actual revenues collections is problematic,since the typical equalization mechanisms,as noted earlier,compensate deviations from an average,thereby giving negative incentives to lower-level governments.3Secondly,in the absence of collection agreements for most tax bases jurisdictions can,and often do,define their own tax bases quite differently.4These considerations point to the view that an ‘actual ’equalization transfer system will most likely deviate from an ‘ideal ’one,defined as the one that is based on a precise measurement of potential fiscal capacity and fiscal needs.There is also a third consideration which may cause a deliberate deviation of the actual equalization transfer away from the ideal one and is based not on fiscal incentives but on political considerations.Given the complex interactions between the levels of government,it is conceivable that such deviations might be enacted by a political process the outcome of which is to favor some particular jurisdiction.5This,of course,does not mean to suggest that all equalization transfer systems suffer from some form of political favoritism,but all it does is to point out this possibility.Indeed,this is an issue that has been identified empirically.Inman (1998)and Johansson (2003),for example,provide a rationale for the observation that transfers allocated to jurisdictions cannot be explained by traditional concerns of equity and efficiency alone.Variables representing political considerations are additional and significant determinants of these transfers.The discussion in the preceding paragraph raises the issue of whether an equalization system that departs from the ‘ideal ’one distorts the political incentives of policy makers.This might arise because the specific nature of such deviations will not be easily understandable by the electorate,rendering the equalization system not fully transparent.It is conceivable,then,and indeed it will be shown here this to be the case,that such transfers,by equalizing fiscal capacity between jurisdictions,interact with the incentives of policy makers to divert resources away from public good provision and for personal gain.This possibility,though it has,implicitly or explicitly,appeared in policy discussions,has not attracted,to the best of our knowledge,any formal analysis.And this is the objective of this paper:to develop a model within which issues of accountability and equalization transfers,of the form discussed previously,can be articulated and investigated.This issue is analyzed using a political competition framework which features the idea of relative performance evaluation popularized in Public Finance by Besley and Case (1995).6This strand of literature considers the effect of ‘yardstick ’competition on rent extraction and,in particular,on the selection of ‘good ’incumbents.Like in these models,in the present paper voters of a typical jurisdiction can evaluate the incumbent of their jurisdiction using information obtained from observing the behavior of a neighboring jurisdiction.Unlike these contributions,the present analysis considers the interaction between equalization transfers and the incentives arising from elections.We explore this aspect by considering a simple two period model within which incumbents have career concerns and are subject to yardstick competition.7In this model,own resources,and thus the supply of public goods in a jurisdiction,are not only affected by the ‘competence ’and the extent of rent-seeking behavior of the local incumbent but also by a shock which is common across jurisdictions.Since voters neither observe competence and rent-seeking behavior nor the common shock,they assess the performance of the incumbent in their own jurisdiction by considering public good supplies in all jurisdictions.An incumbent who takes more rents will see her jurisdiction fare worse in this comparison and,thus,her chances of re-election are reduced.Into this framework we introduce a system of horizontal intergovernmental transfers where a fraction of the difference between the jurisdictions ’own tax revenues is equalized.It is shown that such a system has two effects on accountability:a direct and an indirect one.To see the working of the direct effect,which is conducive to less rent-3A theoretical analysis of such incentives is provided by Smart (1998).Baretti et al.(2002)show that the outflow of tax revenues caused by equalization reduces the efforts by German States to enforce and collect federal taxes.4See Boadway (1998,2004),and Smart (2005),for the Canadian equalization program.5Examples for political favoritism in the construction of equalization transfers are provided by Hofman et al.(2006).6This theory has been further developed by,among others,Bordignon et al.(2004),Belleflamme and Hindriks (2005),Revelli (2006),and Besley and Smart (2007).7As in Persson and Tabellini (2000),Chapter 9.2337C.Kotsogiannis,R.Schwager /Journal of Public Economics 92(2008)2336–23492338 C.Kotsogiannis,R.Schwager/Journal of Public Economics92(2008)2336–2349taking and therefore beneficial for the citizens,consider first an economy without an equalization scheme in place. Suppose further that citizens observe a reduction in public good supply.Since public good supply is determined by own resources which,in turn,are determined by the competence level of the incumbent,citizens naturally attribute the reduction in public good supply to a lower competence level.Now suppose that an equalization scheme is in place and that the same reduction in public good supply is observed.Since the equalization scheme partly compensates jurisdictions for the loss in own resources this,given the reduction in public good supply,reveals an even larger reduction in own resources than in the case without equalization.As a consequence,citizens attribute an even lower competence level to the incumbent.Rent-taking,which reduces public good supply,is thus punished more severely in the election.As a result,politicians restrain their rent-taking activity.8The indirect effect works through the lack of transparency of the equalization transfer,as emphasized above.To capture this,we introduce a random component in the assessment of the jurisdictions'fiscal capacities.The consequence of this is that,even knowing the equalization rate,citizens cannot perfectly derive fiscal capacities from the supplies of public goods observed in both jurisdictions. Hence,the informational content of observing public good supplies in both jurisdictions is reduced.As a consequence, the adverse effect of increased rent-seeking on voters'assessment of the incumbent's performance is mitigated by equalization transfers.Thus,the incumbent politician's trade-off between current rents and the probability of winning the elections is tilted towards more rent diversion.Based on these effects,we characterize the rents taken in a Nash equilibrium.It is shown that the amount of rents taken in equilibrium decreases in the equalization rate if the equalization scheme is fully transparent.Conversely,an equalization system that lacks transparency induces politicians to take more rents in equilibrium.Our work is part of the fast growing literature,termed the second-generation theory of fiscal federalism(SGTF),that focuses on the political processes and the behavior of political agents and their effects on fiscal outcomes in federal systems(see Oates(2005),and Weingast(2006)for comprehensive surveys on the SGTF literature).While identifying institutions and political incentives rather than preferences and technology as the driving forces,this line of research, like the first generation theory of fiscal federalism,aims at explaining the main characteristics of federal systems,such as the vertical allocation of powers or equalization transfers.Consequently,a first and major task has been to develop a political theory of the benefits and drawbacks of decentralization(see Seabright,1996;Persson and Tabellini,2000; Lockwood,2002;Hindriks and Lockwood,2005).Equalization transfers,which are at the heart of our analysis,have also received some attention in the SGTF.Careaga and Weingast(2003)show that the common pool problem created by revenue sharing induces lower-level governments to divert resources away from productive use,a prediction validated by empirical results from Mexican States.Boarnet and Glazer(2002)find that spending in U.S.States is lower when neighboring States obtain larger federal grants. According to this latter contribution,this occurs because politicians at the State level are considered to be incompetent when they fail to win federal grants and as a consequence a rational response by voters is to force them cut public spending.To this line of research,which is mainly empirical in nature,our work contributes by providing a formal model suitable for analyzing the impact of fiscal equalization on the political incentives provided by elections.The organization of the paper is as follows.Section2introduces the model while Section3presents its equilibrium. Section4briefly concludes.2.Description of the modelWe consider a model with two periods and two jurisdictions labeled i=1,2which are ex ante identical.There is electoral accountability in the sense that voters hold incumbents accountable ex post for incompetent behavior in office. This occurs in an election at the end of period1(described in Section2.3).There is a representative citizen in each jurisdiction who values public goods.The supply of the public good in jurisdiction i=1,2in period1is denoted by g i, whereas g i2denotes public good supply in period2.9The supply of public goods in each period is determined by the8This mechanism parallels the beneficial effect of equalization in the standard tax competition models.There,since jurisdictions are compensated for the loss in their tax base when they increase taxes,the damaging effect of tax competition is mitigated,see Wildasin(1989),Köthenbürger (2002),and Bucovetsky and Smart(2006).Empirical evidence for the impact of equalization on the tax setting behavior of lower level jurisdictions is provided by Dahlby and Warren(2003)for Australia,Büttner(2006)for Germany,and Karkalakos and Kotsogiannis(2007)for Canada.9We denote,throughout,second period variables by the superscript2.Also,for ease of notation,we drop the time index for variables relating to the first period.own source revenue of a given jurisdiction (introduced in Section 2.1),the fiscal equalization scheme that is in place (introduced in Section 2.2),and the amount of rents diverted.2.1.Determination of own source revenuesThe aggregate tax base in jurisdiction i =1,2depends on the ‘ability ’(equivalently ‘competence ’)of the incumbent politician in the given jurisdiction,denoted by ηi ,and the common economic environment of the federation,denoted by ε.This formulation reflects the fact that competent economic policy,as captured by ηi ,and positive shocks in the world economy,as captured by ε,raise production and therefore the tax base of jurisdiction i .The competence level ηi ,which is a permanent feature of incumbent i =1,2,and the economic environment of the federation ε,which is common to both jurisdictions,are both stochastic and unknown to both voters and incumbents.Thus,the incumbent politician,while fully informed about her ability in general,is new to the particular tasks associated with the incumbency and so does not know how well she performs them.The abilities of first period incumbents are identically and independently distributed normal random variables with mean μη=1and variance ση2.The common shock εis normally distributed with mean με=0and variance σε2and is independent from both competence levels η1and η2.Own source revenues in jurisdiction i ,denoted by τi ,are obtained by applying the tax rate τ¯,which is exogenously given,to the tax base and so are given bys i ¼g i þe ðÞs ¯;i ¼1;2:ð1ÞEq.(1)simply states that,ceteris paribus ,the higher the level of competence of the incumbent of a jurisdiction the higher the own source revenues of that jurisdiction.Similarly,the better the economic environment of the federation,all other things being equal,the better the own source revenues of both jurisdictions.The specification in Eq.(1)raises two issues,one relating to the tax base and the other to the tax rate.Firstly,a possible,but not the only,interpretation of Eq.(1)is that ηi +εrepresents the overall economic activity of jurisdiction i (the GDP of jurisdiction i ),and thus the tax base of the regional economy.Such economic activity naturally depends upon the competence of the incumbent that runs the regional economy but also upon the economic conditions that prevail in the national economy.Arguably,a natural extension to the formulation in Eq.(1)consists of allowing the overall economic activity of jurisdiction i to depend,in addition to those variables,on the idiosyncratic conditions that prevail in this jurisdiction.Such specification of interregional differences provides a rationale for an equalization system that is based on fiscal disparities and is completely unrelated to politics.Interestingly,however,it is straightforward to show 10that this specification does not affect the results.Thus,without loss of generality,one can interpret the variable ηi as capturing both the competence of the incumbent politician and an exogenous shock to the economy of jurisdiction i .Introducing such shocks into the formal model would,however,burden the exposition of the paper without adding any further insights to the current analysis.For this reason the analysis makes use of the specification in Eq.(1).Secondly,the assumption of a fixed tax rate is a simple way to account for the fact that in many countries lower-level jurisdictions have much less control over tax rates than over public spending.Take,for instance,Austria and Germany.In these countries,autonomous own tax revenues,that is revenues from taxes whose rates are set by the States themselves,account for a negligible share of the tax revenues accruing to those States.Scotland,where the new Scottish parliament does not have the right to legislate on quantitatively important taxes,is another example that illustrates the point.Although these cases are clearly extreme,there is,in general,a tendency for autonomous own tax revenues in sub-national governments in many countries to be less important than other sources of revenue.Over the period 1996–2001,for instance,the share of autonomous own tax revenue in total own revenue of consolidated local and regional governments was 45.4%in Belgium,24.7%in Ireland,26.8%in Italy,33.6%in Spain and 27.9%in the UK (see Stegarescu,2006:159,Table B.6).We turn now to a description of the fiscal equalization program.10The details are available upon request from the authors and are also available from the Journal's website.2339C.Kotsogiannis,R.Schwager /Journal of Public Economics 92(2008)2336–23492.2.The fiscal equalization programIn practice a typical tax-base-equalization program has the following structure.For each revenue source a base is chosen to represent,as closely as possible,the actual base of that revenue source.Total revenues for all jurisdictions from that source are then divided by the nationwide base to arrive at a ‘national average revenue rate ’.This rate is then applied to the base in a particular jurisdiction and the resulting tax is divided by the provincial population to obtain the per capita yield of the tax at the national average rate.The difference between the jurisdiction's per capita yield and the national per capita yield,multiplied by the jurisdiction's population,represents the base for calculating the equalization payments due to the jurisdiction with respect to that particular revenue source.If the difference is negative (positive),a certain fraction of the difference,called the equalization rate,is paid out to (collected from)the jurisdiction.11As already noted,the implementation of an actual equalization transfer program is problematic.To capture this feature of the program in a simple way we introduce the random variables γi ,i =1,2,and assume that the own source revenues of jurisdiction i =1,2are overestimated by the amount γi .It is thus the value of τi +γi ,instead of the true own source revenues τi ,that enters the equalization formula.To parallel the discussion of the introductory section,true own source revenues τi is the ideal measure of a region's resources which should be equalized on equity grounds,but the quantity actually used in the equalization formula deviates from this by τi .12The random variables γi ,i =1,2,are normally distributed with mean μγ=0and variance σγ2.They are unknown both to voters and incumbents,13independent from each other,and also independent from η1,η2,and ε.The equalization transfer to jurisdiction i =1,2is,then,given by z i ¼t Pj ¼1;2s j þg j ÀÁ2Às i þg i ðÞ"#;ð2Þwhere 1≥t ≥0is the federation's equalization rate.Naturally,since the budget of the federal economy must balance,we have that Σi z i =0.Total revenue in jurisdictions i =1,2is,then,given by s i þz i ¼s i þt =2ðÞs j Às i þg j Àg i ÀÁ;ð3Þwhere j ∈{1,2},j ≠i denotes the other jurisdiction.In period 1,the incumbent politician in jurisdiction i =1,2decides to take rents r i out of total revenues τi +z i .These choices are not observed by voters before the election.Rents cannot,of course,be negative and so r i ≥0,i =1,2.It is also assumed that rents satisfy 1N r ¯≥r i ,i =1,2.14The remaining revenues are transformed into public goods,denoted by g i ,i =1,2,as followsg i ¼s i þz i ðÞ1Àr i ðÞ:ð4ÞMaking now use of Eq.(2)in Eq.(4)for both jurisdictions and solving these equations simultaneously,one can recover,for i =1,2and j ∈{1,2},j ≠i ,the level of own source revenues given by s i ¼1Àt 2ÀÁg i 1Àr j ÀÁÀt 2g j 1Àr i ðÞ1Àt ðÞ1Àr i jÀÁÀh g j Àg i ÀÁ;ð5Þwhereh t ðÞ¼t =21Àt ðÞ½ z 0:ð6Þ11This is,for instance,a variant of the equalization systems in Canada and Germany.In Canada the equalization rate is constant (derived from using an all-province standard)and the ‘gross system ’is applicable whereby only positive equalization entitlements are paid.In Germany the equalization rate varies with the difference between the jurisdiction's own fiscal capacity and the average fiscal capacity in the federation but the ‘net system ’is applicable whereby both positive and negative transfers exist.12As noted in the introductory section,such a deviation from pure equalization might deliberately be enacted by the federal government so as to favor some particular jurisdiction.For the present analysis this interpretation would fit the model as long as this bias in the federal government's policy cannot be predicted by voters nor by local incumbents.13One,of course,may argue that interested parties may have the incentive (if they have the capability of doing so)to precisely calculate fiscal capacities and inform voters,Smart (2005).Though this is a possibility it does not seem to be a perfectly convincing one.For interested parties,typically,have opposing incentives in the calculation of fiscal capacities giving scope for unlimited conflict over transfers.This conflict,as far as the true fiscal capacities of the jurisdictions are concerned,is unlikely to be very informative and,therefore,some uncertainty will still linger.14A possible,and arguably convincing,reason for this restriction is the possibility that a zero provision of public goods triggers an immediate investigation by an independent authority,such as the constitutional court,into the workings of the government.2340 C.Kotsogiannis,R.Schwager /Journal of Public Economics 92(2008)2336–2349The inequality in Eq.(6)follows from the restriction on the equalization rate.Notice now,for later use,following Eq.(6)it is the case that 15h V t ðÞ¼1=21Àt ðÞ2h i N 0;ð7Þand so θis a monotonically increasing function of the equalization rate t .However,since citizens are not informed about γ1nor γ2they can neither infer own source revenues nor equalization payments from g 1and g 2.Instead,they must form expectations about these revenues and the underlying competence levels of the incumbents.2.3.Payoffs and second-period decisionsIn period 2,own source revenues,the equalization scheme,and rent-taking behavior determine public good supplies g i 2,i =1,2,just as in period 1,by equations analogous to Eqs.(1)–(4).For the own source revenues in jurisdiction i =1,2,however,now the competence of the government in the second period is relevant.This is either the competence ηi of the first period incumbent,if the latter is re-elected,or,if she is defeated,the competence of a challenger which is drawn from the same normal distribution with mean μη=1and variance ση2.Moreover,the second-period rent r i 2satisfies the same restrictions as the first period rent,that is,r i 2≥0and 1N r ¯≥r i 2,i =1,2.Politicians are interested in expropriating rents collected in both periods and in an exogenous additional rent from winning the elections,denoted by R N 0.Denoting by δthe discount factor,and by p I ,i the probability that the incumbent of jurisdiction i =1,2is re-elected for office in the second period,the payoff to the incumbent of jurisdiction i is given byr i s i þz i ðÞþp I ;i d R þr 2i s 2i þz 2iÀÁÂÃ:ð8ÞIn the second period there is no re-election motive anymore and thus every government will take the maximal rentr i 2=r ¯,i =1,2.Nevertheless,given that r ¯b 1,there always remains some tax revenue which is used for public good provision.Thus,for given maximal rent-taking behavior,a more competent incumbent still produces higher own source revenues.Now,as can be seen from Eq.(4),for all equalization rates 0≤t ≤1,the supply of public goods in a jurisdiction is increasing in the own source revenues of this jurisdiction.Therefore,a more competent government in a jurisdiction will deliver a higher quantity of the public good to that jurisdiction's citizens.Since citizens value public goods,they have an incentive to elect the most competent incumbent.Consequently,in the election at the end of the first period,voters in jurisdiction i =1,2vote for the incumbent if their estimate of the incumbent's ability η~i exceeds the expected ability of the challenger,which is given by μη=1.3.Equilibrium analysisThe model is analyzed using the Nash equilibrium concept under which the decisions by voters and incumbents in the first period are simultaneously optimal,given a correct assumption on the other players'behavior.Following this,the optimal voting behavior of citizens in jurisdiction i =1,2is determined by the estimate η~ithey form about the competence of the incumbent in this jurisdiction.This estimate is based on the information citizens have at that moment and on an assumption about the rent-taking behavior of both incumbents,denoted by r ~i ,for i =1,2.16In Section 3.1,the formation of the expectation η~i ,i =1,2,conditional on r ~i and r ~j,for j ∈{1,2},j ≠i ,is analyzed.The incumbent of jurisdiction i =1,2decides about how much rents r i to expropriate,anticipating the impact of this decision on the estimate η~i and,hence,on the probability of winning the election.This is described in Section 3.2.An equilibrium requires that the actual decisions coincide with the assumptions used by the voters,that is,r ~i =r ifor i =1,2.In Section 3.3,the rent taken in such an equilibrium is calculated.The analysis is then completed by deriving the impact of an increase in the equalization rate on this rent.15A prime denotes the derivative of a function of one variable.16The supposed strategies r ~i ,i =1,2–just as the rents r i ,i =1,2,that are actually chosen –do not depend on the levels of competence η~i since when the rents are chosen competence is not known to the incumbents.2341C.Kotsogiannis,R.Schwager /Journal of Public Economics 92(2008)2336–2349。