Fiscal Rules, Bailouts, and Reputation in Federal Governments

Transcription

1 Fiscal Rules, Bailouts, and Reputation in Federal Governments Alessandro Dovis University of Pennsylvania and BER Rishah Kirpalani Pennsylvania State University Septemer 207 Astract Expectations of ailouts y central governments incentivize overorrowing y local governments. In this paper, we ask if fiscal can correct these incentives to overorrow when central governments cannot commit and if these will arise in equilirium. We address these questions in a reputation model in which the central government can either e a commitment or a no-commitment type and the local governments learn aout this type over time. We find that if the central government s reputation is low enough, then fiscal can lead to even more det accumulation relative to the case with no. This is ecause the punishment for violating the fiscal rule worsens the payoffs of preserving reputation. Despite eing welfare reducing, inding fiscal will arise in the equilirium of a signaling game due to the incentives of the commitment type to reveal its type. First version: January 207. We thank Marco Bassetto, Charlie Brendon, Pierre-Olivier Gourinchas, Marina Halac, Boyan Jovanovic, Patrick Kehoe, Ramon Marimon, Leonardo Martinez, Diego Perez, Deraj Ray, Thomas Sargent, and Pierre Yared for valuale comments.

2 Introduction There are numerous examples throughout history in which excessive spending and det accumulation y sunational governments led to ailouts y central governments. Examples include provinces in Argentina, states in Brazil, länders in Germany, and most recently countries Greece, Ireland, and Portugal in the European Union. One view of such events is that the lack of commitment of central governments to not ail out leads to profligating fiscal policies ex-ante, which in turn justifies the ailouts ex-post. This idea has een formally studied y Chari and Kehoe 2007, Chari and Kehoe 2008, and Cooper et al in the economics literature and Rodden 2002 in political science. See also Sargent 202. A commonly held view is that fiscal can correct these incentives to overorrow when central governments lack commitment. In practice, fiscal take the form of limits to det-to-gdp or deficit-to-gdp ratios along with some penalty if these are violated. For example, the Staility and Growth Pact SGP calls for all EU memer countries to keep udget deficits elow 3% of GDP and pulic det to elow 60% of GDP. EU memer countries are liale to financial penalties of up to 0.5% of GDP if they repeatedly fail to respect these limits. When thinking aout the design of fiscal, a natural question that arises is why central governments can commit to enforcing these if they cannot commit to not ail out. In this paper, we ask if fiscal can e eneficial if central governments cannot commit and if these will arise in equilirium. We address these questions in a reputation model in which the type of the central government is uncertain: it can e either a commitment type or a no-commitment type. The reputation of a central government is the proaility that local governments assign to it eing a commitment type. In the tradition of Kreps and Wilson 982 and Milgrom and Roerts 982, we focus on the case in which there is a small initial proaility that the central government is the commitment type. Our first main result is that if the reputation of the central government is low enough, then fiscal are welfare reducing and lead to even more det accumulation relative to the case with no. This is ecause the punishment associated with the fiscal rule enforcement makes it more attractive for the no-commitment type to reveal its type earlier relative to an environment without. This early resolution of uncertainty makes overorrowing more attractive for the local governments. Our second main result is that despite eing welfare reducing, inding fiscal can arise in an equilirium of a signaling game ecause the commitment type wants to signal its type and it is optimal for the no-commitment type to initially mimic and then not enforce the rule once violated. See Rodden et al. 2003, Rodden 2006, and Bordo et al. 203 for further documentation. 2

3 We show these results in a stylized three-period model populated y local governments and a enevolent central government. Local governments choose the provision of a local pulic good and have access to local tax revenues. They can also orrow from the rest of the world at a given interest rate. We first consider the case in which local governments are homogenous and then assume that they can e one of two types: the orth and the South. In the latter case, we assume that the orth has access to a larger period 0 tax revenue, which leads to a non-degenerate distriution of det holdings in period along the equilirium path. The central government does not have tax revenues, ut it can impose transfers from one state to another. We consider an institutional setup in which the constitution requires the central government to not impose such transfers no-ailout clause and local governments to keep their det elow some level or face an output cost if they violate this rule fiscal rule. The central government can either e a commitment type that enforces the fiscal constitution or a no-commitment type that chooses its policy sequentially. This type is initially unknown to the local governments, which learn aout it through the actions of the central government. In period the intermediate period, the enevolent no-commitment central government faces a trade-off etween not enforcing the constitution and preserving its reputation, which incentivizes the local governments to keep future det accumulation in check. We first consider the case in which the constitution contains only a no-ailout clause and no fiscal. We show that when the central government s initial reputation level is low enough, there is a unique equilirium in which the no-commitment type central government does not make transfers to the local governments in the intermediate period and so there is no revelation of uncertainty until the terminal period. The central government prefers to delay the revelation of its type ecause for low enough reputation levels, the costs of early information revelation are first order, while the enefits of equalizing the provision of the local pulic good in the interim period via a ailout are second order. When local governments are homogenous, these costs are exactly zero on path. When local governments are heterogeneous, the distriution of det inherited in the interim period is non-degenerate and so these costs are positive. However, if the proaility of facing the commitment type is close to zero, the provision of the local pulic good in the orth and the South is almost identical even without a ailout in the interim period, ecause the South orrows against the ailout transfer it anticipates in the final period. We next consider a constitution with oth a no-ailout clause and a fiscal rule. If the central government s reputation and discount factor are low enough, there exists a unique equilirium in which fiscal are violated in period 0 y the local governments and are not enforced ex-post y the no-commitment type central government. Therefore, in this equilirium there is early resolution of uncertainty i.e., the central government reveals 3

4 its type in period. The intuition ehind this result is that with fiscal, the value of preserving reputation is lower, since the enforcement of the constitution now requires the no-commitment type central government to impose costly penalties on the local governments that violate the rule. In particular, unlike in the case without fiscal, the costs of enforcing the constitution are no longer second order. We then compare the det levels in the equilirium outcomes with and without. Having fiscal in the constitution leads to even more det accumulation relative to the case without. The key driver for this result is that the type of the central government is revealed in the interim period with early resolution of uncertainty, and only in the terminal period without late resolution of uncertainty. Knowing the type of the central government in period allows the local governments to condition their new det issuances on the government type. This in turn lowers the cost of servicing the det inherited in period ; hence, the local governments will issue more det in period 0. We next consider a planner tasked with designing the optimal fiscal rule taking into account this lack of commitment. We show that if the prior of the central government eing the commitment type is low enough, it is strictly optimal to not have fiscal. The previous result raises the question of why we would ever see fiscal eing instituted in practice if they were welfare reducing. We study a signaling game in which are chosen at the eginning of time y the central government. We show that for intermediate values of the central government s discount factor, in the equilirium of this game, the commitment type chooses to announce a fiscal rule, which is mimicked y the no-commitment type. However, in this equilirium the rule is not enforced in period y the no-commitment type, leading to early resolution of uncertainty and even more det accumulation. This result sheds light on historical and contemporary episodes when fiscal were instituted ut were not enforced ex-post. A leading example is the SGP in the Eurozone. The SGP was instituted for the newly formed monetary union, under the pressure of Germany, with the intent of constraining fiscal policy in memer countries to insulate the European Central Bank ECB from the pressure to inflate or monetize the det of memer countries. However, the enforcement of the SGP has een very lax. For example, in 2003 oth Germany and France violated it and sanctions were not imposed. Moreover, the sanctionary powers of the European Commission were susequently weakened. Through the lens of our theory, this corresponds to the case in which the central government reveals its type in the intermediate period. Consistent with our theory, after 2003, the power of the SGP in disciplining fiscal policy was argualy weakened. According to several commentators, this was a major factor in the current European det crisis in which Greece, Ireland, and Portugal received ailout packages from the European Union and the ECB the central government, as our theory predicts. 4

5 Argualy, after the ailouts to peripheral memer countries, the reputation and crediility of the central European institutions were very low. EU memer countries and European institutions agreed to impose tough fiscal y strengthening the SGP y introducing the so-called Six-Pack and Fiscal Compact consistent with the prediction of our signaling game. The provisions of the Six-Pack were soon violated y Spain and Portugal without any sanction eing levied. 2 In 206 the governor of the Bundesank, Jens Weidmann, accused the Commission of not enforcing the fiscal : My perception is that the European Commission has asically given up on enforcing the of the Staility and Growth Pact. 3 Another leading example of federal governments with poor fiscal discipline among sunational governments is Brazil, the most decentralized state in the developing world. The fiscal ehavior of the states and large municipal governments in Brazil were a major source of macroeconomic instaility and resulted in sunational det crises in 989, 993, and 997. The federal government took a variety of measures to control state orrowing in the 990s, and at a first glance it would appear to have had access to an impressive array of hierarchical control mechanisms through the constitution, additional federal legislation, and the central ank. Most of these mechanisms have een undermined however, y loopholes or ad incentives that discourage adequate enforcement Rodden et al page 222. In 997, the federal government assumed the dets of 25 of the 27 states that were unale to service their det an amount equivalent to aout 3% of GDP. By Septemer 200, 84% of state det was held y the national treasury see Rodden et al. 2003, page 234. After the ailouts in 997, the Cardoso administration approved the Fiscal Responsiility Law, which instituted a rule-ased system of decentralized federalism that leaves little room for discretionary policymaking at the sunational level. It has een motivated y the recognition that market control over sunational finances should e replaced, or strengthened, y fiscal as well as appropriate legal constraints and sanctions for noncompliance, Afonso and De Mello So, in a manner similar to Europe, the central government in Brazil imposed stringent fiscal when its reputation was argualy low. Related literature Our paper is related to several strands of literature. First, it is related to the literature that studies the free-rider prolem in federal governments when the central government cannot commit e.g., Chari and Kehoe 2007, Chari and Kehoe 2008, Cooper et al. 2008, Aguiar et al. 205, Chari et al. 206, and Rodden The main result in this literature is that the inaility of the central government or monetary 2 See 3 See 5

6 authority to commit not to ail out ex-post leads to overorrowing ex-ante. In such settings, it is often argued that fiscal can improve outcomes y lowering the amount of det issued e.g., Beetsma and Uhlig 999. Our paper contriutes to this literature y analyzing the effects of fiscal when the government cannot commit to enforcing them. Fiscal have een studied in several environments as the solution to time inconsistency prolems. See for instance Athey et al. 2005, Amador et al. 2006, Halac and Yared 204, and Halac and Yared 207 in the context of delegation, and Hatchondo et al. 205 and Alfaro and Kanczuk 206 in the context of sovereign default. All these papers assume that the agents can commit to and do not analyze the enforcement prolem, which is the main focus of our paper. The aseline model uses a reputational setup similar to Kreps et al. 982, Kreps and Wilson 982, and Milgrom and Roerts 982 with uncertainty aout the type of the central government. It also relates to papers that try to account for several features of policy outcomes y studying models in which a government with a hidden type interacts with a continuum of private agents e.g., Cole et al. 995, Phelan 2006, and D Erasmo In contrast, in our paper the local governments are strategic and can incentivize the central government to reveal its type via its actions. In addition, we also study the optimal policy in this environment and how varying the costs of maintaining good reputation affects outcomes. Uncertainty aout the type of the central government plays a key role in the provision of incentives to local governments. osal and Ordoñez 203 also consider an environment in which uncertainty can mitigate the time inconsistency prolem when a central government cannot commit not to ail out anks. The mechanism is very different: here uncertainty aout the type of the central government curs det issuances y the local governments, while in their paper it is the uncertainty aout anks local governments that restraints the central government to not intervene ex-post. The rest of the paper is organized as follows. In Section 2 we present the model, and in Section 3 we show that fiscal promote fiscal indiscipline when local governments are homogeneous. Section 4 demonstrates that our results extend to the case with heterogenous local governments. Section 5 discusses the role of having large strategic local governments. In Section 6, we show that imposing no is optimal under the veil of ignorance, and in Section 7 we show that can arise in the equilirium of a signaling game. Section 8 concludes the paper. 6

7 2 Model Environment Let t = 0,, 2. 4 Consider a small open economy consisting of states or regions indexed y i {, 2,..., }. The representative citizen in region i has preferences over the local pulic good provision {G it } 2 U i = β t u G it. t=0 Throughout we make the following assumptions: Assumption. The period utility function u is strictly increasing, strictly concave, u C, lim c 0 u c =, and u 0 finite. The local pulic good provision is decided y a enevolent local government with local tax revenues {Y it }. We start y characterizing the simplest possile case, Y it = Y for all i and t. We will later show that our results generalize to cases when tax revenues are heterogeneous across regions and vary over time. The local government can orrow from the rest of the world at a rate + r. Let q = / + r e the price of a ond that promises to pay one unit of the consumption good next period. There is also a central government. The central government does not have tax revenues, ut it can impose transfers from one region to another suject to a udget constraint T it 0, i= where T it is the transfer to region i in period t. Efficient allocation As a enchmark, we consider the efficient allocation in this environment. An allocation is efficient if for some set of Pareto weights {λ i } it solves max {G it } λ i i= t=0 2 β t u G it suject to 2 q t G it Y it ] 0. t=0 i= 4 Our main results extend to any finite horizon economy. 7

8 Any efficient allocation must satisfy qu G it = βu G it+ 2 and the consolidated udget constraint with equality. Institutional setup and equilirium Consider an institutional setup in which the central government is suject to a fiscal constitution. The fiscal constitution contains two clauses. The first clause states that the central government should not ail the regions out i.e., T it = 0 for all i, t. We call such a provision the no-ailout clause. The second clause requires the local governments to keep their det issued in period 0 elow a cap. In case i >, the central government must impose a penalty ψy on the region that violated the rule. We assume that the resources collected from penalties are thrown away. 5 We call this constitutional provision a fiscal rule. A fiscal rule is then fully descried y, ψ. To simplify notation, we astract from a cap on det issued in period and its associated penalty. All our propositions will extend to the case with a cap on det issued in period 2. The central government can e one of two types: a commitment type, which follows the prescriptions in the constitution, and a no-commitment type, which is not ound to follow the prescriptions of the constitution, as it chooses policies sequentially to maximize an equally weighted average of the utility of citizens in oth regions: 6 W r = 2 t r β t u G it. i= The type of the central government is drawn at the eginning of period 0 and is not known to the local governments. They have a common prior π that the central government is the commitment type. Throughout the paper we consider the proaility of facing the commitment type as eing close to zero. Local governments cannot rule out the possiility that the central government might always enforce the constitution, no matter what the state of the world is. This is consistent with Kreps and Wilson 982 and Milgrom and Roerts 982, who assume that players are ehavioral with an aritrarily small proaility. The timing is as follows: 5 This assumption ensures that the cost of imposing the fiscal rule is nonzero for the central government even if π = 0. 6 The redistriution motive generates an incentive for the central government to ail out the local government with higher det. We would otain similar results if ailouts were motivated y spillovers, as in Tirole

9 At t = 0, the local governments choose the local pulic good provision G i0 and det i suject to the udget constraint G i0 Y i0 + q i. At t =, if the central government is the no-commitment type, it decides whether to make transfers {T i } or not and whether to enforce the penalty if the fiscal rule is violated y a local government. After oserving the central government s actions, the local governments update their prior aout the central government type and decide the provision of the local pulic good G i and new det issuance i2 suject to G i + i Y + T i + q i2 ψ YI {i > and central government enforces fiscal rule}. At t = 2, if the central government is the no-commitment type, it decides whether to make a transfer {T i2 } or not. ext, the local governments choose G i2 suject to udget constraints G i2 + i2 Y + T i2. We assume that the local government can commit to repaying its det. This can e motivated y the existence of high default costs, which makes repayment always optimal for the local government. We now define the states, payoffs, and eliefs at each node of the game tree. Period 2 The state in the last period is the distriution of det among local governments, 2 = i2 i {,2,...,}. If the central government is the no-commitment type, it will choose transfers T i2 2 such that the consumption of the local pulic good is equalized etween regions 7 : T i2 2 = i2 j= j2 so that G i2 = Y j= j2, and it will not impose the penalty if the fiscal rule is violated. We refer to this situation as det mutualization. The value for the central government is W 2 2 = i= u Y j= j2, 7 ote that there is no enefit to preserving reputation, since the world ends after period 2. 9

10 and the value for a local government is j= j2 V i2 2 = u Y. If instead the central government is the commitment type, each region will consume G i2 = Y i2. The value for the local government is then Vi2 c 2 = u Y i2. Period The state in period is the distriution of det among the local governments, = i i {,2,...,} and the prior on the type of the central government, π. Let σ e the equilirium strategy of the central government in period. The central government can either enforce the fiscal constitution or not. 8 We consider equiliria where the law of motion for eliefs follows Bayes rule and is given y π π π+ π σ, ζ, π; σ =,π if ζ = 0, 3 0 if ζ = where ζ = if the central government does not enforce the fiscal constitution in period, and σ denotes the enforcement strategy for the central government and is defined y σ 0 W, π, ψ; π e, π, 0, π; σ, ψ > W ne = W e, π, 0, π; σ, ψ < W ne, 4 0 < σ < W e, π, 0, π; σ, ψ = W ne where σ = means that the constitution is not enforced, while σ = 0 denotes enforcement; W e is the value for the no-commitment type central government if it enforces the fiscal constitution in period, and W ne is the value for the no-commitment type central government if it does not enforce the fiscal constitution in period. We will descrie these value functions in detail in what follows. We now analyze the decision of the local governments. Suppose first that there is enforcement so that the posterior of the central government s type remains constant at π, π, 0, π; σ = π. In this case, the local governments choose G i, i2 to solve Vi e, π = max u G i + βπvi2 c i2 + β π V i2 i2, j2, π G i, j i i2 8 To ease notation, we exclude the case in which the central government enforces only one of the provisions of the fiscal constitution. This is without loss of generality, since it will never e optimal for the central government to do so. 0 5

11 suject to G i + i Y i + q i2 ψyi {i > } taking as given the strategy j2, π followed y the other local governments. For later reference, the equilirium outcome at this node will e given y { i2, π} i=, which solves for all i u Y qu Y i + ψyi {i > } + q i2 = βπu Y i2 + β π j= j2. 6 If local government i exceeds the det limit and the punishment is implemented, the continuation outcome is equivalent to one in which local government i enters the period with det i + ψy, so the det it issues is i2 i, i + ψy, π. Moreover, unless the proaility of facing the commitment type is one, the optimality condition 6 differs from the Euler equation 2 that characterizes the efficient allocation. In particular, if π <, there is overorrowing ecause each local government internalizes only of the marginal cost of repaying its det if it anticipates a ailout when the central government is the no-commitment type. ote, however, that in any symmetric equilirium there is never a ailout on path. If a local government were to deviate and orrow a larger amount, it would trigger a transfer in period 2 if the central government is the no-commitment type. 9 ext, suppose that the fiscal constitution is not enforced, which implies that π,, π; σ = 0 so that the central government reveals its type. In this case, the value for the local government given a set of transfers T is suject to V ne i, 0, T = max G i, i2 u G i + βv i2 i2, i2, 0, T 7 G i + i Y i + T i + q i2 taking as given the strategy i2, 0, T followed y the other local governments. To simplify the exposition, note that if π,, π; σ = 0, whether there is det mutualization in period and 2 or only in period 2 is irrelevant in that the equilirium consumption outcomes are identical. Lemma. If π,, π; σ = 0, the continuation values and pulic good provisions for the local governments are independent of transfers in period. In particular, for all udget feasile 9 This is similar to the split-the-ill prolem.

12 T = T i i=, the value of a violation of the no-ailout clause is independent of T in that V ne i, 0, T = V e i, 0 = V ne i B, 8 where B = i= i. The proof of this lemma is provided in the Appendix. The intuition is that when π = 0, a form of Ricardian equivalence holds: when the local governments are certain that they are facing the no-commitment type central government, the timing of transfers is irrelevant. Asent transfers, the local governments with inherited det aove average will simply orrow more to keep current consumption constant, expecting a ailout in the second period. On the other hand, the local governments with inherited det elow average, asent transfers, will reduce new det issuances ecause they anticipate a negative transfer in period 2. As a result of Lemma, the value for a local government in the event of non-enforcement depends only on the aggregate level of det rather than the distriution of det. Moreover, we can then drop T as a decision variale and assume without loss of generality that transfers equal zero in period. The value for the no-commitment type central government is then W, π, ψ = σ, π, ψ] W e, π, 0, π, ψ + σ, π, ψ W ne B, 9 where the value of not enforcing is W ne B = i= Vne i B, 0 and the value of enforcing is 0 W e, π, ψ = i u Y i + q i2, π, ψ ψyi {i > } + βu Y B ] 2, π, ψ, where B 2 = i i2, and the equilirium enforcement strategy is given y 4. Similarly, the value for the commitment type in period is W c, π, ψ = i ] u Y i + q i2, π, ψ ψyi {i > } + βu Y i2, π, ψ. 0 ote that W e i Ve i since the no-commitment type central government knows that it will mutualize det in period 2, while the local governments have uncertainty aout the central government s type. 2

13 Period 0 The state in period 0 is the prior on the type of the central government, π the realization of Y i0 is incorporated y indexing the value functions y t and i. In period t = 0, each local government chooses the local pulic good provision and det to solve V i0 π, ψ = max G i0, i u G i0 + β σ, π, ψ] V e i i, i, π, ψ 2 + βσ, π, ψ πv c i i + β π V e i i, i, 0] suject to the udget constraint G i0 Y i0 + q i, taking as given the strategies i π, ψ followed y other local governments, and σ, π, ψ followed y the central government. For later reference, we also define the value for the no-commitment type central government in period 0, W 0 π, ψ = i= u G i0 π, ψ + β σ π, ψ, π, ψ] W e π, ψ, π, ψ 3 + βσ π, ψ, π, ψ W e π, ψ, 0, 0, where G i0 π, ψ and π, ψ are the decision in 2. The value for the commitment type is W c 0 π, ψ = i= u G i0 π, ψ + β σ π, ψ, π, ψ] W c π, ψ, π, ψ 4 + βσ π, ψ, π, ψ W c π, ψ,, ψ. Equilirium definition We can now define a Perfect Bayesian Equilirium for this institutional setup. Definition. A Perfect Bayesian Equilirium is a set of strategies and eliefs for the local governments, i π, ψ, π, ζ, π, i2, π, ψ, a strategy for the no-commitment type central government, σ, π, ψ, and associated value functions, such that i given i π, ψ and σ, π, ψ, i π, ψ solves 2; ii given i2, π, ψ, i2, π, ψ solves 5; iii π, ζ, π satisfies 3; and iii σ, π, ψ satisfies 4. 3

14 3 Fiscal promote fiscal indiscipline In this section we present the first main result of the paper: if the reputation of the central government is low enough, then fiscal lead to even more det accumulation relative to the case with no. This is ecause the punishment associated with the fiscal rule enforcement makes it more attractive for the no-commitment type to reveal its type earlier relative to an environment without. This early resolution of uncertainty makes overorrowing more attractive for the local governments. Equilirium outcomes without fiscal : Enforcement in period We start y characterizing the equilirium when the fiscal constitution contains only a no-ailout clause and no fiscal. The main result in this section is that without fiscal, if the central government s initial reputation is low enough, there exists a unique equilirium outcome in which the central government type is not revealed in period. Proposition o revelation of central government type. Suppose the constitution has no fiscal. Then, for π sufficiently small ut positive and sufficiently large, there exists a unique symmetric equilirium in pure strategies in which the type of the central government is not revealed in period. Moreover, the det issuances { no-, no- } 2 satisfy qu Y + q no- = βu Y no- + q i2 no-, π 5 + β 2 π u Y j2 no- i2 no-, π, π i and no- 2 = i2 no-, π. The proof of this and other propositions is provided in the Appendix. Since all regions are homogenous, if a pure strategy symmetric equilirium exists, it must have no ailout on path, as each local government enters period with the same amount of det. However, off equilirium, a local government could potentially increase the det issued in period 0 to induce the central government to ail it out in period. In the proof, we show that such a deviation is not profitale provided the reputation of the local government is low enough i.e., π sufficiently close to zero. The central government does not reveal its type in period. So, when the local governments choose their det issuance in period, they are still uncertain aout the type of the central government and aout the proaility of receiving a transfer in the terminal period 2. Given these expectations, det issuances along the equilirium path While this trivially holds with identical local governments, we show in a later section that the argument extends to the case with heterogeneity. 4

15 are characterized y equation 5 and no- 2 = i2 no-, π. The first two terms of condition 5 resemle those in a standard intertemporal Euler equation, while the last term on the right hand side captures strategic effects in the det issuance decision. Each local government understands that its choice of det issuance in period 0 will affect the det issuance decisions of the other local governments in period, which in turn affects the utility of the local government in period 2 in case of det mutualization which happens with proaility π. otice that this term vanishes as since i2 no-, π / i /q, as shown in Lemma 3 in the Appendix. Equilirium outcomes with fiscal : o enforcement in period We now consider the case in which the fiscal constitution has a no-ailout clause and a fiscal rule. We say that fiscal are inding if the det limits are lower than the equilirium outcome without fiscal, < no-. When the central government s reputation is low and the fiscal rule is inding, there is a unique equilirium in which the local government violates the fiscal rule in period 0 and the no-commitment central government does not enforce the punishment in period. This leads to the revelation of the central government s type in period. Proposition 2 Early revelation of central government type. Suppose the constitution has inding fiscal. Then, for π and β sufficiently small ut positive and sufficiently large, there exists a unique symmetric equilirium in pure strategies in which the fiscal rule is violated in period 0 and not enforced y the no-commitment type in period{ so that the type of the } central government is revealed in period. Moreover, the det issuances,,c 2,,nc 2 satisfy qu Y + q = βπu Y + ψ + q i2 + ψ, + β π u Y + q i2, 0 + β 2 π u Y j2, 0,c 2 = i2 + ψy,, and,nc 2 = i2, 0. i2, 0 i, The key step to estalish the proposition is to show that in period, the no-commitment type central government when faced with det i = > for all i prefers to not enforce the punishment ψ and reveal its type π = 0 thereafter than to enforce the punishment and enjoy the reputation gain, 2 and therefore lower distortions in the local govern- 2 The posterior jumps to one as the local governments expect only the commitment type to enforce the fiscal rule. 6 5

16 ment s Euler equations. In the Appendix, we show that this is true when β is sufficiently low. Hence, when local governments choose the new det levels in period, they know with certainty the type of the central government they are facing. This shows up in equation 6, where the right side of the Euler equation is contingent on the type of the central government. ote for future reference that the strategic term in equation 6, the last term on the i2,0 right side, does not vanish as. This is ecause while i 0 as in the case without, u Y j2, 0 ecause local governments exhaust their det capacity in period if they know for sure they face the no-commitment type. We can show that the product converges to a finite negative numer. However, we will show that the sum of the second and third terms on the right hand side converges to zero as. Comparing det levels We next show that when the central government s reputation is low, inding fiscal promote more fiscal indiscipline than a constitution without fiscal that is, the det levels in this equilirium are higher than in the equilirium without fiscal. The key driver for this result is that the type of the central government is revealed in period with and so the local governments can condition their new det issuances on the government type. Proposition 3 Fiscal promote fiscal indiscipline when reputation is low.. Under the assumptions of Proposition and 2, the level of det issued in period 0 is higher with inding fiscal than without. Moreover, contingent on facing the no-commitment type, the det issued in period is higher with inding fiscal than without. Consider first the det issued in period 0. For the case without fiscal, we can comine 5 with 6 to otain a condition that characterizes the det issuance in period 0: u Y + q q = β2 π q u Y i2, π + β2 π q u Y i2, π 7 + β2 π u Y i2, π j i j2, π i. 6

17 For the case with fiscal, we can comine 6 with 6 to otain u Y + q q = β2 π q u Y i2 + ψ, + β2 π q u Y i2, 0 8 β 2 π u Y i2, 0 j i j2, 0 i. These two optimality conditions are identical with the exception that with no fiscal condition 7, det issued in period is not conditional on the type of the central government, i2 = i2, π. With inding fiscal condition 7, det issued in period is conditional on the type of the central government, and is either i2, if the central government is the commitment type with proaility π or i2, 0 if the central government is the no-commitment type. We next show that the early revelation of the central government s type in the equilirium with fiscal induces the local governments to issue more det. Taking the limit as goes to infinity, since lim u Y i2, π <, as shown in Lemma 2 in the Appendix, condition 7 reduces to u Y + q q = β2 π q u Y i2, π 9 as the second and third terms on the right side converge to zero. Condition 8 instead reduces to ecause lim βu Y i2, 0 u Y + q q = β2 π q u Y i2 + ψy,, 20 q = lim u Y i2, 0 j i j2, 0 i = u Y + q i, as shown in Lemma 2 and 3 in the Appendix. We can then compare the right hand side of 9 and 20. We know that for a small enough π, i2, π > i2 + ψy,, ecause as π 0, i2, π Y ut i2 + ψy, is ounded away from Y see Lemma 2 for details. This oservation along with the concavity of u implies that β 2 π q u Y i2 + ψy, < β2 π q u Y i2, π. Therefore, from 9 and 20 we see that the expected marginal cost of issuing det in period 0 is lower when there is early revelation of the central government s type. Hence, local governments will issue more det in period 0 ecause of the lower expected marginal 7

18 cost. Intuitively, if the central government reveals its type only in period 2, even if a local government is confident it will receive a ailout in period 2, it does not orrow a lot in period 0 ecause it knows that if the central government is the commitment type, consumption in period 2 will e very low. If instead the central government reveals its type in period, the local government will orrow more ecause in the unlikely event that the central government is the commitment type, the local government can spread the losses associated with not receiving a ailout over period and period 2. The latter is preferale and so the government has a higher incentive to orrow more in period 0 ecause it can etter insure the risk of facing the commitment type. Consider now det issuances in period if the central government is the no-commitment type. In this case, det issued in period is higher with than without for two reasons: first, the inherited det is larger; second, the local governments face no uncertainty aout the type of central government and therefore internalize only / of the cost of issuing det, while with they internalize the full cost with proaility π and / the cost with proaility π. This argument concludes the proof of Proposition 3. Strategic effects By considering the limit as the numer of regions goes to infinity, we astract from strategic effects in det issuances. Strategic effects will further incentivize local governments to orrow more when there is early revelation of the central government s type. The strategic interaction in det choices etween the two local governments is captured y the third term on the right side of 8 and 7. The elasticity of the det issuance y the other local governments in period to the det issued y local government i in period 0 is decreasing in π. Formally, i2, π i < i2, π i < 0 2 for π > π. The intuition is straightforward. If π = 0, as is the case when the central government s type is revealed at t =, then a given local government has a high incentive to adjust its period det issuance in response to the inherited det of the other local governments. This is ecause at π = 0, the local governments know there will e det mutualization with proaility one next period. If instead there is no early revelation and π > 0, then there is det mutualization in period 2 only with proaility π and so a local government s det issuance will e less sensitive to det issued in the previous period y the other local governments. The increased sensitivity implies that if a local government orrows an additional unit in period 0, it will receive a larger transfer in period 2 conditional on facing the no-commitment type when π = 0 than when π > 0. This strategic channel contriutes to the lower cost of servicing det internalized y local 8

19 Figure : Equilirium outcomes: Det issued in period and 2 o Rule Rule governments when there are fiscal. umerical example We illustrate the proposition in Figure. 3 The two panels of Figure 2 display the det issued y a representative local government along the equilirium path without lue line and with red line as a function of the prior in period 0 that the central government is the commitment type. When π is low enough, det issuances in period 0 are higher with. The same is true in period conditional on facing the no-commitment type. When the initial prior π is not close to zero, we cannot characterize the equilirium analytically. umerically, we show that when instead π is aove a threshold, there exists an equilirium in which are followed, the central government does not reveal its type in period, and total indetedness is lower than in the case without. Hence fiscal may e effective in reducing det only when the central government s reputation is sufficiently high. But when the central government s reputation is high, the gains from reducing indetedness are smaller: det is decreasing in π ecause the local governments expect that they will not receive a ailout with a high proaility. Therefore, fiscal are detrimental exactly when the prolem of overorrowing is most severe, while they are effective only when the gains from enforcement are relatively low. Given the assumptions in our model, the non-enforcement of any one of the clauses in the constitution implies full revelation of the central government s type. As a result, the 3 To construct the figure, we compute the equilirium outcome with u c = c α 2 c2 and = 2. 9

20 no-commitment type will never choose to enforce one of the clauses ut not the other. However, the arguments that estalish the results aove will extend to a situation in which the central government can somehow commit to the no-ailout clause ut not to the fiscal rule, so long as the non-enforcement of the fiscal rule lowers reputation and increases the likelihood of a ailout. Our characterization is consistent with the experience of several federal states in which fiscal were instituted and often violated y sunational governments. It also provides a rationale for why sunational governments kept on orrowing excessively after the central governments deviated from the fiscal constitution. Argualy, this is what happened in the European Monetary Union EMU after the violation of Maastricht treaty in 2005 and the susequent relaxation of the and penalties. This is also consistent with the experience in Brazil where d]et urden continued to grow in the 990s. Despite the previous crises and ailouts - or perhaps ecause of them - the states continued to increase spending. Rodden et al Heterogeneous local governments So far we have assumed that the local governments are identical. As a result, in any symmetric equilirium outcome, the central government is never tempted to impose transfers etween regions, ecause all governments have the same det position. In this section, we show that our arguments extend to the case in which the local governments are heterogeneous and so the central government might find it optimal to ail out the poorer local governments along the equilirium path. Suppose that = 2M with M. We partition the local governments in two groups: the orth, n = {,..., M}, and the South, s S = {M +, M + 2,..., 2M}, and for all n and s S we let 4 Y n0 > Y s0, Y nt = Y st = Y for t =, 2 and so the orth is richer at time 0 relative to the South. This creates incentives for the no-commitment type central government to ail out the South and reveal its type. While the ulk of our analyses focuses on the effects of the choice of fiscal constitution on the equilirium det holdings of the local governments, it is worth noting that from a utilitarian perspective, commitment to the fiscal constitution need not always e optimal when local governments are heterogeneous. The reason for this is that a utilitarian planner values redistriution and as a result ailouts can e valuale as a means of equaliz- 4 Adding heterogeneity in tax revenues Y it for t > 0 leaves the results unchanged. 20

21 ing consumption. evertheless, this comes at a cost of distorting the local government s Euler equation, which leads to overorrowing as compared to the efficient enchmark. However, for Y s0 and Y n0 sufficiently close to each other, the second effect will always dominate and the ex-ante welfare associated with the commitment type is strictly larger than that of the no-commitment type. This is the region of the parameter space we will restrict our attention to. ext, we show that without fiscal, provided that π is small enough and Y n0 Y s0 is not too large, it is still optimal for the central government to delay revealing its type and not provide a ailout in period : Proposition 4 o ailout in period when crediility is low.. Suppose the constitution has no fiscal. Then, for π and Y n0 Y s0 sufficiently small ut positive and sufficiently large, there exists a unique symmetric equilirium in pure strategies in which the type of the government is not revealed in period. Moreover, the det issuances { no- i, no- } i2 i=n,s satisfy qu Y i0 + q no- i = βu Y no- i + q i2 no-, π + β 2 π u Y i2 no-, π and no- i2 = i2 no-, π. i2 no-, π i The proof of this proposition mainly follows the same steps as the proof of Proposition with one exception: since the local governments in the orth and in the South will enter period with different levels of det, we now have to show that the central government wants to enforce the constitution in period, when its reputation is sufficiently low. To understand this step, let us consider the costs and enefits of enforcing the fiscal constitution in period. By enforcing the no-ailout clause, the central government preserves its reputation. A higher π in turn promotes fiscal responsiility, ecause the local governments expect to repay their det without a ailout from the central government with higher proaility. Hence the enefits of enforcing are associated with a reduction of the distortions in the local government s Euler equations 6 relative to the efficient one 2. The costs of enforcing the no-ailout clause are associated with the high inequality in the provision of the local pulic good. A higher π will induce the South government to orrow less in period and cut the consumption of the local pulic good relative to the orth. This dispersion in the consumption of the local pulic good across regions orth and South is costly from the perspective of the enevolent central government. For π close to zero, if the central government enforces the constitution and does not ail out, there is essentially no inequality of the local pulic good consumption, since 2 22

22 the local governments expect a ailout with high proaility in period 2 and so the costs of not redistriuting are second order. However, the enefits from inducing more fiscal discipline are first order, since the Euler equation is distorted relative to the efficient allocation. Hence, it is optimal for the central government to not ail the regions out or to enforce the constitution when its reputation is very low. In our model, we do not allow the central government to make transfers in period 0. If we did allow for transfers, notice that the no-commitment type might not want to, since it would reveal its type in period 0, thus leading to large distortions in the Euler equations for oth periods. Moreover, for π close to zero, we can use a similar argument to to show that in period 0, local governments orrow so that there is very little inequality in consumption. As a result, the central government will not choose to reveal its type until the last period. ext, we state the analogue of Proposition 2 and 3 for the economy with heterogeneous local governments: Proposition 5 Early revelation of central government s type.. Suppose the constitution has inding fiscal. Then, for π, Y n0 Y s0, and β sufficiently small ut positive and sufficiently large, there exists a unique symmetric equilirium in pure strategies in which the fiscal rule is violated in period 0 and not enforced y the no-commitment type in period { so that the type of the } central government is revealed in period. Moreover, the det issuances,,c 2,,nc 2 satisfy qu Y i0 + q = βπu Y + ψ + q i2 + ψ, + β π u Y + q i2, 0 + β 2 π u Y j j2,c 2 = i2 + ψ,, and,nc 2 = i2, 0., 0 j2, 0, j i i Proposition 6 Fiscal promote fiscal indiscipline when reputation is low.. Under the assumptions of Proposition 4 and 5, the level of det issued in period 0 is higher with inding than without. Moreover, contingent on facing the no-commitment type, the det issued in period is higher with inding fiscal than without. As in the case without heterogeneity, we show that along the equilirium outcome with fiscal the rule is violated and not enforced ex-post y the no-commitment type central government that reveals its type in period. The early revelation in turn promotes fiscal indiscipline. The proofs of the two propositions essentially follow from continuity of the equilirium outcome in Y n0 Y s0 given the results in Proposition 2 and

23 Figure 2: Equilirium outcomes: Det issued in period and 2 with heterogeneous governments o Rule Rule We illustrate Proposition 6 with a numerical example. The four panels of Figure 2 display the det issued y the South and orth along the equilirium path without lue line and with red line as a function of the prior in period 0 that the central government is the commitment type. To make the two graphs comparale, we assume that in the equilirium with fiscal no transfers are made in period, so the different det levels issued in period do not reflect the different pattern of transfers. As shown in Lemma, this does not affect pulic good provision or the aggregate amount of det. ext, as shown in Proposition 6, for π close to zero the det issued in period 0 y all local governments is higher with than without. The same is true in period. As in the case with homogenous local governments, when the central government s reputation is aove a threshold, the South respects the fiscal rule and the central government does not reveal its type in period. In this case, total indetedness is lower than in the case without. ote that the orth still orrows more with ecause it now anticipates that the South will orrow less, which implies that it will have to transfer less in the event of a ailout and so its expected marginal utility of consumption is lower. So, again, fiscal may e effective in reducing det only when the central government s reputation is sufficiently high. 23

24 5 Large vs. small local governments In this section we riefly discuss the role of having large local governments for our results. We show that when =, there exists another equilirium in addition to the one that is the limit of finite economies in which fiscal can e eneficial. We now show that if = and all local governments are small, in addition to the equilirium we characterized aove the limit of finite, there exists another equilirium with fiscal where all local governments oey the rule and therefore fiscal achieve their intended role of curing det issuance y local governments. Proposition 7. Under the same assumptions of Proposition 2, if =, there exists another equilirium in which fiscal are followed in period 0. The key intuition for Proposition 7 is that with non-atomistic local governments, there are no costs for the central government to enforce the penalty for a violation of the fiscal rule y an individual local government that has measure zero. Hence, if one local government expects that other local governments will respect the fiscal rule, it is optimal for it to respect the rule as well and so there is an equilirium in which fiscal can cur indetedness and where the local governments internalize the free-rider prolem. This result is fragile: there is always an equilirium where the rule is ignored y all the local governments and not enforced. In particular, if a government expects the other governments to violate the rule, it will find it optimal to violate the rule as well since it anticipates that the rule will not e enforced ex-post. This type of multiplicity is similar to the one in Farhi and Tirole 202 and Chari and Kehoe 205. Proposition 7 suggests that the forces we emphasize in this paper are more likely to e relevant when the local governments are relatively large. This result may help to rationalize why when two large countries such as Germany and France violated the SGP in 2003, no sanctions were imposed y the European institutions. More generally, Eyraud et al. 207 provides suggestive evidence that compliance with the SGP has een lower among the largest countries. However, it may e possile for institutions such as the IMF to enforce penalties on a small country to preserve their reputation. 6 Optimal fiscal We now turn to analyzing whether the optimal fiscal constitution should have fiscal. By optimal we mean the fiscal constitution that induces the maximal average welfare for the citizens who elieve that the central government is the commitment type with proaility π and the no-commitment type with proaility π. We show that if the central government s reputation is low, it is optimal to have no fiscal. 24