Decentralization of Government Functions Introduction to Public Economics A.Estache

Introduction Increasing political pressure worldwide to rethink how to assign the responsibilities for expenditure and revenue raising across government levels ….or across regions as in Belgium, Australia, Canada, Germany, Spain, India, Brazil, Argentina, Mexico, … For example, who should be responsible for education??? the national government the linguistic communities the provinces the local governments (communes in Belgium) ???

Centralized Unitary Federal System Decentralized Unitary System NOTE: This is a really broad concept: Different Forms of Decentralization Centralized Unitary Federal System Decentralized Unitary System Central GVT State/regional/provincial Central Unit Local Management Units Subntnl gvt GVT Unit Local Management Units

How has the world chosen to organize its fiscal constitutions? Unitary (148 countries) Federal (45 countries)…usually large ones but not always… Australia, Belgium, Brazil,Canada, Germany, India, Mexico, Malaysia, Russia, Pakistan, Russia, South Africa, USA Confederal : European Union Not all federal countries are decentralized and not all unitary countries are centralized!!!

Context Critical to understand because strong debates about how much subnational governments can actually deliver This … in spite of the sense that it increases accountability by bringing political decisions closer to the voters =>Increased Decentralization Worldwide Some leading (e.g. high income countries) Some lagging (e.g. low income countries) Some sensitivity with respect to definition of decentralization used Expenditure vs. revenue Relevance of mismatch between revenue and expenditure assignments across government levels Note also relevant in the EU context… …debate on …the Subsidiarity principle How do you aggregate the preferences of sovereign nations into a European set of preferences? The analysis is “cash-based” only (no accrual based).

Basic Questions to Answer Which level of gvt should be responsible to deliver which expenditure? Which level has to pay for which specific state functions? Which level should control which sources of fiscal (tax and non tax) revenue? Should there be any transfers or grants from the federal level to the states/provinces or from the states/provinces to the local governments/municipalities? How should transfers across government levels be designed? Should there be a control of sub-national government borrowing?

It boils down to... …looking for: a theory of how the efficiency of public goods provision may differ at different levels of government (which level should deliver what at the highest efficiency level)… but recognizing that we have concerns to deal with: equity, fiscal constraints accountability 7 7

But … what exactly do we mean by decentralization? Fairly mechanical view =“old fashion” public finance definitions 2 measures Decentralization in terms of revenue Definition: share of subnational tax and non-tax revenues excluding grants in total national revenue What does it measure? Use of autonomy over resources? Decentralization in terms of expenditure Definition: share of subnational spending over total national spending What does it tell us? Compliance with responsibilities over service obligations? Decentralization Definitions: - Political Decentralization Administrative Decentralization Deconcentration Delegation Devolution Fiscal Decentralization** (our definition!) Co-Financing/Co-Producing Loan Guarantee Revenue and Expenditure Autonomy Market Decentralization Privatization Deregulation

Decentralization in the OECD Table 1 Decentralization in the OECD Spending % Revenue % Greece 5.0 3.7 Portugal 12.8 8.3 France 18.6 13.1 Norway 38.8 20.3 United States 40.0 40.4 Denmark 57.8 34.6 OECD Average 32.2 21.9 The average state/local government collects 22% of total government revenue, while those in the U.S. collect 40%. On the spending side, the differences are also huge…and the match between spending and revenue…modest at best!.

Example: Evolution of expenditure decentralization across key sectors Developed countries Developing countries 1990 2004 Education 70.8 93.0 65.2 69.2 Health 57.1 89.6 61.1 48.8 Infrastructure 71.1 76.5 50.4 57.0

Centralized vs decentralized Spending Case for DECENTRALIZED Local participatory planning Diversity consistent with local preferences Incentives for competitive and innovative service delivery Better quality, quantity and access of public services Lower agency costs Bottom-up Fiscal and political accountability (“the voice of the people”) Restraints against corruption and rent seeking Easier to constraints on government size because better accountability??? Case for CENTRAL Central planning Redistribution across regions for regional equity Economies of scale Regional risk sharing Internal common market Harmonization of expenditures across regions Avoidance of race towards bottom to attract investment or people Rules driven top-down accountability Ensure subnational budget constraints are not soft and hurt the country

Centralization vs Decentralization of Taxation Case for Centralized Efficiency in collection Lower compliance costs Encourages factor mobility for internal common market Wasteful tax competition avoided Discourages shifting tax burden to non-residents Fiscal inequity avoided Case for Decentralized Fiscal accountability Better administration of taxes on immobile factors Better use of fees and charges Reduced corruption (?)

Theoretical Foundations of Decentralized Fiscal Constitutions (1) Always useful to conceptualize stylized facts… Traditionally, public economics has looked at decentralization as a simple optimization problem which deals with public goods, economies of scale and externalities with various government levels The decentralization theorem (all the work by Oates (1972) and his many co-authors): decentralization can improve (allocative) efficiency of public service delivery by matching expenditures with constituents’ preferences, provided there are no economies of scale and no externalities/spillovers 13

Theoretical Foundations of Decentralized Fiscal Constitutions (2) More recently some economists looked at decentralization as an incentive issue and deals with it with agency models Multiple principals, multiple agents (see Laffont and his followers on this) Also some economists looked at decentralization as in the optimal level of vertical disintegration of firms Application of the new theory of the firm (Cremer, Estache and Seabright) Also lots of good research in public administration, politics, …that needs to be recognized if you do research here

Expenditure Assignment Principles Essential to keep in mind when thinking about how to design decentralization that the policy goals include: Economic Efficiency driven by Spatial externalities and spillovers Economies of scale and scope Administrative and compliance costs Proximity to users…and voters Fiscal efficiency Regional equity Redistribution Correction of market failures provision of quasi-private goods Preservation of internal common market Economic stabilization Recognition of relevance of spending power/capacity

Tax Assignment Principles Similarly, essential for policymakers to have sound economic criteria for distributing revenue sources: Expenditure responsibilities of sub-levels=> fiscal needs Availability of local taxes for local services Avoidance of tax-induced movements Stability for sub-level governments Fair distribution of national wealth Efficiency of internal common market Efficiency in tax administration

So…what is the optimal fiscal federalism? Given all the concerns we just saw…how can we come up with a formal model to answer the basic question? “What is the optimal division of responsibilities across different levels of government?” As hinted earlier, the answer is offered by a theory of optimal fiscal federalism which is essentially a theory of how the efficiency of public goods provision may have to differ at different levels of government

Optimal fiscal federalism? (2) We already know that two of the major problems in public goods provision are: Preference revelation: Difficult to design democratic institutions to cause individuals to reveal their preferences honestly. Preference aggregation: Difficult to aggregate individual preferences into a social decision.

The Tiebout Model Decentralization can be viewed as a solution to these problems Tiebout (1956) showed that the key to understand inefficiency in a centralized public goods provision came from the recognition of two missing factors which are relevant when looking at the opportunities offered by decentralized provision. Shopping induces efficiency in private markets. Competition induces the right prices and quantities in private markets. So the main message of Tiebout is: competition across local jurisdictions places competitive pressures on the provision of local public goods… …and in spite of some heroic assumptions needed to develop the model…it turns out to be quite consistent with empirical evidence

The Tiebout Model (2) With public goods provided at the local level, competition naturally arises because individuals can vote with their feet by moving to another city or commune or province…without much disruption. This induces fiscal discipline for local governments, thus, decentralization can be seen as a new preference revelation device and preferences are revealed by mobility. Tiebout argued that the threat of exit and hence loss of tax revenue or other local income can induce efficiency in local public goods production. Under certain (unrealistic) conditions public goods provision will be fully efficient at the local level.

The Tiebout Model (3) Tiebout’s formal model assumes the following: Large number of individuals, who divide themselves up across towns that provide different levels of public goods. Town i has Ni residents who all demand Gi of the public good. Uniform tax of Gi/Ni.

The Tiebout Model (4) In this model, individuals will divide themselves up so that each resident in any town has the same taste for public goods Since they have the same taste, they will demand the same level of public good spending Gi

The Tiebout Model (5) Tiebout’s model solves our two problems: Preference revelation: There is no incentive to lie. With a uniform tax on all residents, the consumer saves 1/Ni in tax but receives 1/Ni less of the public good. Preference aggregation: it is solved because everyone in the town wants the same level of public goods, Gi.

The Tiebout Model (6) => Lindhal pricing can work in this model! each individual report own true valuation of the PG the valuations are added up and each individual is billed for the total cost of PG divided by the population size of the local community => this is an equilibrium since each person is happy to pay his/her share of the tax to get the PG and the condition for optimal good provision is met because the level of PG is determined by the sum of individual benefits!

Problems with the Tiebout Model (1) Main problems with the model, however are related to: Tiebout competition Tiebout financing Spillovers

Problems with the Tiebout Model (2) Tiebout competition may not hold because: It requires perfect mobility. It requires perfect information on the benefits individuals receive and the taxes they pay. It requires enough choice of towns so that individuals can find the right levels of public goods.

Problems with the Tiebout Model (3) Tiebout financing is problematic because: It requires lump-sum taxes that are independent of a person’s income => viewed as highly inequitable. YET, it is more common for towns to finance public goods through proportional taxes on homes or home related services Note that proportional taxation leads to the problem of the poor chasing the rich when the rich too large a share of the services mostly used by the poor The use of zoning—limits on use of land-- can address this problem by pricing the poor out of the market For example, a town that prohibits multifamily dwelling such as apartments lowers the available amount of housing, and thus inflates the value of existing housing, keeping the poor out.

Problems with the Tiebout Model (4) Tiebout model is also problematic because of the assumption of no externalities or spillovers: Model assumes public goods only have effects in a given town, and that they do not spill over to neighboring towns. Some public goods, like a public park, probably violate this assumption.

Evidence on the Tiebout Model In spite of these problems, Tiebout model’s basic intuition that individuals vote with their feet is still a strong one. Two types of tests reveal this: Resident similarity across areas Capitalization of fiscal differences into house prices

Evidence on the Tiebout Model (2) Residents similarity across areas A clear prediction of the Tiebout model is that residents in a local community will have similar preferences for local public goods. The more local communities and choices there are, the more residents can sort themselves into similar groupings. Gramlich and Rubenfeld (1982) found greater sorting in larger metropolitan areas (where mobility costs would be smaller), and greater satisfaction with public goods provision.

Evidence on the Tiebout Model (3) look at the prices of houses across areas Very little actual mobility is required for the Tiebout mechanism to operate because people not only vote with their feet. They also vote with their pocketbook!! Tiebout model predicts that any differences in fiscal attractiveness will be capitalized into house prices. For instance, access to top local schools drives the prices of houses!..coming soon in Belgium with the Arena rule on school access

Evidence on the Tiebout Model (4) look at the prices of houses across areas (2) Overall, the price of any house reflects the costs (including local property taxes) AND benefits (including local public goods) of living there. Holding taxes constant, higher or better levels of public goods raise housing prices. Hold public good levels constant, raising taxes lowers housing prices. =>Housing prices are a reflection of people voting with their pocketbook.

So what can we do with the Tiebout model So what can we do with the Tiebout model? ...towards a theory of optimal fiscal federalism? More precisely, what are the normative implications of the Tiebout model? That is, what should be the principles that guide the provision of public goods at different levels of government? The extent to which public goods should be provided at the local level is determined by: Tax-benefit linkages Positive externalities or spillovers Economies of scale

Optimal Fiscal Federalism (1) Tax-Benefit Linkages: The model implies that the extent to which public goods should be provided at the local level is determined by tax-benefit linkages. IF Strong linkages (such as local roads)=> most residents benefit, => the good should be provided locally. IF Weak linkages (such as welfare payments) => most residents do NOT benefit, => the good should be provided at a higher level. If residents can see directly the benefits they are buying with their property tax dollars, they will be willing to pay local taxes. Otherwise, they may “vote with their feet.”

Optimal Fiscal Federalism (2) Externalities: The 2nd factor to determine the optimal level of decentralization is the extent of positive externalities. If the local public good has spillovers to other communities, the LPG will be underprovided. With spillovers, higher levels of government have a role in promoting the provision of these public goods.

Illustrating Positive Spillovers Examples: Environmental policies. Tax/VAT competition. If decentralised, each region chooses a level of public good that is too low. e.g. Qd2 for Region 2. Two-region gain from centralisation is area A. Similar conclusion with negative spillovers: Q too high under decentralization.

Optimal Fiscal Federalism (3) Economies of scale: The 3rd factor driving the optimal level of decentralization is the economies of scale in production. Public goods with large economies of scale, like national defense, are best provided by the central government (not efficiently provided by many competing local jurisdictions). Public goods without large economies of scale, like police protection, may be provided more efficiently in Tiebout competition.

Illustrating Economies of Scale Costs of providing public good may fall with scale E.g. single national rail or bus network may be more efficient than many regional ones. At EU level: foreign and security policies…even if strong differences in preferences Different marginal costs apply in centralized and decentralized case. Different demand curve as well (“standardization of demand) Region 1: welfare gain C vs loss D  centralization may be preferred.

Optimal Fiscal Federalism (4) The Tiebout model therefore predicts that local spending should focus on broad-based programs with few externalities and relatively low economies of scale. Examples include road repair, education, garbage collection, and street cleaning.

Redistribution across communities The Tiebout model allows us to consider one of the most important problems in fiscal federalism: Should there be redistribution of public funds across communities? There is currently enormous inequality in the ability and perhaps desire for communities to finance public goods. For instance, gaps in per-pupil spending between Flanders and Wallonia arise because of differences in the local tax rate and revenue, but more importantly, from differences in tax bases across regions.

Evidence of within country inter-regional inequality around the World

So how does this redistribution issue come about? Often decentralization of expenditure and revenue responsibilities do not match => built-in deficits, at least for some decentralized governments So how can subnational governments close the gaps between their own revenue sources and their assigned expenditure responsibilities? Many countries practice fiscal equalization, whereby the national government distributes grants to sub-national government in an effort to equalize differences in wealth across regions. => need to look at this in some details

Should We Care? The question then becomes: Should higher levels of government mandate redistribution across lower levels to offset these differences in spending? In a perfect Tiebout world, communities would have formed for the efficient level of public goods, redistribution would not be good since it would impede that efficiency. For Tiebout, redistribution should be across individuals WITHIN communities, NOT across communities Yet there is a political demand for interregional redistribution

Should We Care? (2) To the extent that Tiebout does not perfectly describe reality, however, there are two arguments for redistribution. The first is the need to fix the failures of the Tiebout mechanism. For example, even if one desires to be in a high benefit community, a household may be priced out of it by zoning restrictions, etc. The second is the need to address externalities. It is possible that local public goods, like education, have spillovers to other communities=> case for higher level subsidies to local primary education

What are the tools of Redistribution in a decentralized gvt structure? When higher levels of government redistribute, they do so through grants Grants are cash transfers from one level of government to another. They do so to fix two types of imbalances: Horizontal imbalances Vertical imbalances

Reasons for intergovernmental transfers? Horizontal imbalances? Different capacities of the sub-levels to raise revenues. Different costs of services to be provided (expenditure responsibilities). $$$$$$ Income Expenditure $$ $$$$ $ State A State B State C State D $$$$$$ $$$$ $$$ $$$

Reasons for Intergovernmental transfers (2) Vertical imbalances? Revenue responsibilities lie with national level while expenditure responsibilities are with sub- level $$$$$$$$$$$ National government $$ $ $ $ State A State B State C $$$ $$$ $$$$ $$ $$$

Types of Grants used for Redistribution Unconditional vs conditional transfers Unconditional: preserving local autonomy and enhancing inter-jurisdictional equity Block grants –a fixed amount of money with no mandate on how it is to be spent. Conditional: providing incentives to undertake specific activities Conditional block grants–a fixed amount of money with a mandate that it be spent in a particular way.

Types of Grants used for Redistribution Conditional Transfers matching vs non-matching: Matching grants –which ties the amount of funds transferred to the community to the amount of spending it currently allocates to public good open-ended vs. closed-ended matching Input based conditionality vs output based conditionality Input based conditionality often intrusive and unproductive. Output based conditionality can advance grantor’s objectives while preserving local autonomy

The effects of different government grants Private spending (in thousands) $1,375 $1,000 $800 $625 $500 IC3 IC2 IC1 $575 $500 $750 $0 $1,375 $2,000 Education spending (in thousands) $375 $1,000 Figure 2 a The effects of different government grants

The city could still choose to consume $750,000 of education. Private spending (in thousands) Such a grant acts as an income effect, but keeps the price ratio at 1 rather than ½. Imagine instead that the city was given an identical amount – $375,000 – in a block grant. $1,375 Initially, the city faces a tradeoff between education and private goods. The price ratio is 1. Such a grant might mandate that the city can spend receive up to $375,000 if it is spent on education. A one-for-one “matching grant” changes the price of education and the price ratio to ½. A final alternative is a conditional block grant, a fixed amount of money that can only be spent on education. As long as the city is already spending more than $375,000 on education, it is equivalent to a block grant and has no effect on behavior. But the block grant also allows other choices, and utility is higher at IC3, which entails less education. With this lower price for education, the income and substitution effects lead to an increase in educational spending. The city pays only half of the $750,000 cost, however. The state pays the remainder, $375,000. Total spending on both education and private goods has increased with the matching grant. $1,000 The city could still choose to consume $750,000 of education. $800 The voters in the city have preferences over public and private goods. They choose $500,000 of education. $625 $500 IC3 IC2 IC1 $0 $375 $500 $575 $750 $1,000 $1,375 $2,000 Education spending (in thousands) Figure 2 b The effects of different government grants

Goals matter to the optimal Grants/transfers design choice Objective/Goal Grant Design Better Practices Practices to Avoid Closing subnational Fiscal Gap Reassign, tax base sharing Canada Deficit grants, tax by tax sharing Closing Regional fiscal disparities Fiscal capacity equalization (FCE) FCE with an explicit standard as in Canada, Germany, Denmark General revenue sharing with multiple factors, Fiscal eq. with a fixed pool as in Australia, China Setting national minimum standard Block transfers, conditions on service standards Ex-Indo. roads and education, Chile, Brazil Conditions on spending Benefit spillovers Matching grant S. Africa teaching hospitals Influencing local priorities Open-ended matching Canada social assistance Ad hoc grants Stabilization Capital grant with upkeep requirements Political and policy risk guarantee Stabilization without upkeep

So what is the evidence on the effects of decentralization? Lots of evidence on country specific experiences Not that much on cross-country In on on-going research project (Goicoechea et al.) generate some insights on the effects of decentralization on the total level of expenditure on various key expenditure categories.

Summing up the international comparison Decentralization: Reduced total expenditures in: Clearly in transport and all infrastructure in very poor countries Increased total expenditures in: Power and housing in HICs Changed the composition of expenditures in: At the total level: Clearly in all LICs and MICs (reading down the columns in Summary (1) since not the same sign for all sectors) Not much in HICs At the sub-national level Not much for LICs Somewhat more for MICs and HICs But generally, except for transport in LICs, always preference for more expenditure than with less decentralization Had statistically unclear effects in the other cases

How does this all matter in the context of the design of the EU? According to fiscal federalism theory, which decisions should be made at the EU level? Thinks of the E In practice? How does the EU make its decisions? How likely is a decision-making gridlock?

The core: Qualified Majority Voting Most EU decisions made by co-decision procedure Proposal adopted in the Council of Ministers by QMV and in the EP by majority voting Voting in the Council reflects States’ national interests QMV requires more than a simple majority to approve a decision This makes it easier for small Member States to block decisions

Pre-2004 QMV (i.e. Pre-Nice) Number of votes not perfectly proportional to population: Total number of votes in the EU15: 87 Threshold for winning majority: 62 votes  ‘qualified majority’: about 71% of votes required to adopt proposal. Relatively large coalition required to win a vote. Relatively small coalition of countries can block a vote.

Voting Post-Nice Treaty Reforms Two main changes: 1. QMV rules more complex: two new criteria in addition to votes votes: 255 votes out of 345 Council votes in EU27 (74% of votes required to adopt proposal) number of members: half of the member states, i.e. 14 out of 27 population: 62% of EU population. 2. Votes reallocated in favour of big nations

Constitutional Treaty (CT) Nice Treaty QMV rules: relatively small coalition can block important decisions. Risk that Council’s becomes deadlocked CT QMV rules: a proposal wins if backed by member states with At least 65% of EU population At least 55% of member states. At least 15 member states (irrelevant if the EU has 27 or more members: 15/27=56%) Reallocation of vote shares: large nations gain (except Spain and Poland)

Blocking-coalition analysis Ability of ‘likely’ coalitions to block EU decisions. Example: “Newcomers” and “Poor” in EU27 Newcomers: 12 new member states Poor: Newcomers+4 ‘cohesion’ members (ES, PT, GR & IE) ‘Poor’ exceed the Nice Treaty blocking thresholds of votes and member states ‘Newcomers’ exceed the votes threshold only

A last word… This (last) lecture presented a theory (and its tests) of why some spending and some revenue collection should be done locally or regionally The theory has provided lots of insights… some case for redistribution across regions optimal redistribution tool depends on goals But it does not really address all incentives issues associated with the decentralization choice Agency problems with unclear assignments and undefined property rights (in terms of taxing, spending and regulation) Information asymmetry matters a lot as well when looking within government structures! Intergovernmental bargaining can have high transaction costs…Look at Belgium! Scope for lobbying for inefficient outcomes can be huge (Tragedy of commons with pork-barrel politics)