forthcoming, American Economic Review

Federalism and the Soft Budget Constraint

By Yingyi Qian and Gérard Roland

The government's incentives to bail out inefficient projects are determined by the tradeoff

between political benefits and economic costs, the latter depending on the

decentralization of government. Two effects of federalism are derived: First, fiscal

competition among local governments under factor mobility increases the opportunity

costs of bailout and thus serves as a commitment device (the "competition effect").

Second, monetary centralization, together with fiscal decentralization, induces a conflict

of interests and thus may harden budget constraints and reduce inflation (the "checks and

balance effect"). Our analysis is used to interpret China's recent experience of transition

to a market economy. (JEL E62, E63, H7, L30, P3)

Soft budget constraints represent a major incentive problem. They were a key characteristic of

socialist economies and remain an important concern in transition economies. There are also instances of soft

budget constraints in market economies such as bailouts of banks (e.g., the S&Ls) and corporations (e.g.,

Chrysler). Following Janos Kornai (1980, 1986), an enterprise or any organization is said to have a soft

budget constraint when it expects to be bailed out in case of financial trouble. This creates an incentive

problem because the manager of the organization could fail to observe financial discipline. According to

Mathias Dewatripont and Eric Maskin (1995), the fundamental reason for the soft budget constraint problem

is the inability of rescuers to commit to no bailout ex ante.1

In reality, soft budget constraints mostly involve a government, through the bailout of enterprises and

banks (public and private) or of lower level governments. This paper studies the relationship between forms

of federalism, that is, organizations of government involving some fiscal decentralization, and the degree of

2

the soft budget constraint. Based on the tradeoff between political benefits and (endogenous) economic costs,

it analyzes how the government's incentive to soften budget constraints depends on the extent of

decentralization of fiscal and/or monetary authority. Although our study is motivated by the recent Chinese

experience where devolution of authority from the central to local governments is a major feature of reform,

the theme of the paper is more general. Indeed, in the U.S. and Europe, and in many developing and

transition economies, there is a growing interest in studying the relationship between federalism and

government incentives.

We build a model of the soft budget constraint in a three-tier hierarchy with a "central government"

at the top, multiple "local governments" in the middle, and "enterprises" (state and non-state) at the bottom.

The model has the following features. First, the soft budget constraint game between government bodies and

enterprises is sequential: government bodies face sequential bailout decisions in the presence of sunk costs

and enterprises behave strategically in securing subsidies. Second, the local governments compete with each

other in allocating their own budgets to attract mobile factors and grants from the central government in a

simultaneous-move game. And third, the central government plays a sequential game against local

governments in deciding on the allocation of grants and, if monetary financing is possible, on total money

supply. The study of incentive problems in a three-tier hierarchy is generally complicated and difficult, and

this difficulty is only reinforced by the introduction of fiscal competition and factor mobility. Nevertheless,

we are able to derive several results on the effect of federalism in hardening budget constraints.

We first show that decentralization of fiscal authority to local governments together with mobility of

non-state capital across regions may be effective in hardening the budget constraints of enterprises under

local governments' control. This is because competition among local governments in attracting capital to

their region creates an externality which increases the opportunity cost of subsidizing inefficient enterprises,

which in turn reduces the incentives for bailouts. Fiscal competition can thus be viewed as a commitment

device (the "competition effect"). However, this comes at the cost of distortions such as excessive

infrastructure investment.

When fiscal decentralization is not complete and the central government has the power to allocate

part of the fiscal revenues, local governments will also compete for grants. Interestingly, if the central

3

government earmarks grants for local public goods and subsidies, strategic distortions by local governments

further increase the opportunity cost of bailout, thus achieving harder budget constraints, provided the grant

size is not too large.

We then consider the possibility of monetary financing with the central government creating money

and allocating seigniorage revenue while local governments continue to have full authority over their

expenditures. Fiscal competition together with competition for monetary grants works in a similar way

towards hardening the budget constraint of enterprises. However, inflation will be higher than under fiscal

centralization if the central government allocates earmarked grants hoping to "correct" distortionary decisions

by local governments. But if the central government has no discretionary powers in the allocation of

seigniorage revenue between local governments, inflation will be lower than under fiscal centralization. The

reason is that the difference in spending preferences between the central and local governments due to fiscal

competition induces endogenously a conflict of interests, which may achieve both harder budget constraints

and monetary restraint (the "checks and balance effect").

In the extreme case of monetary decentralization, with the externality of inflation, each local

government receives the full benefits from its monetary creation but shares the costs of inflation with other

regions. This creates high inflation and generally softens budget constraints.

Our analysis can be used to interpret the unique features of China's transition process to markets. In

Eastern Europe and Russia, privatization is the main focus of transition. In contrast, China's transition was

first and foremost associated with the devolution of government authority from the central to local level.

Recently, local governments in China took the initiative to lay off excess workers and even privatize the

enterprises under their supervision. This happened at a time when monetary authority was recentralized and

inflation was lowered. According to our analysis, decentralization together with competition has a profound

impact on the incentives of local governments. It induces harder budget constraints because it increases the

local governments' costs of subsidies, and furthermore, this effect requires the support of monetary

centralization. From this perspective, reform of government organization, such as fiscal decentralization, is

an important dimension of the transition from a plan to a market economy.

Our paper presents the first macroeconomic model of the soft budget constraint viewed as a dynamic

4

commitment problem in the framework of a federal government. Even though our concept of the soft budget2

constraint is the same as that of Dewatripont and Maskin (1995), there are a number of important distinctions

between our model and theirs. First, in their model, firms are bailed out by profit-maximizing banks, and in

ours by a government which also cares about employment rents. Second, our model endogenizes sources of3

revenue for bailouts. This allows us to endogenize the opportunity cost of bailouts which depends on the

organization of government. Finally, crucial conditions leading to hard budget constraints in their model are

the limited liquidity of the bank and the costly information transfer between banks. In our model, the size of

the budget matters less than its allocation among various expenditures. In fact, government budgets are

larger when budget constraints are hard since a higher efficiency yields more tax revenues.

This paper contributes to the literature on local public finance and fiscal federalism (e.g., Charles

Tiebout, 1956, Wallace Oates, 1972, Roger Gordon, 1983, and David Wildasin, 1988) by analyzing the link

between fiscal competition and government incentives for bailouts. Thus, it adds to the growing literature on

incentive problems in the context of the organization of government. For example, Jean-Jacques Laffont and

David Martimort (1994) investigate the role of duplication of regulators in preventing collusion, and Torsten

Persson, Gérard Roland, and Guido Tabellini (1997) study the role of separation of powers in the

organization of government in improving the accountability of government. Our results are also in

accordance with recent historical analysis (e.g., Barry Weingast, 1995) and studies on transition economies

(e.g., Yingyi Qian and Barry Weingast, 1996; 1997) emphasizing the commitment effects of federalism.

The paper is organized as follows. Section I presents a three-tier hierarchy model involving the soft

budget constraint in a macroeconomic context. Section II demonstrates the role of fiscal competition among

local governments as a commitment device to harden the budget constraints of enterprises. Section III

introduces monetary creation as an additional source of financing and analyzes the implications of alternative

monetary arrangements under fiscal decentralization on the budget constraints and inflation. Section IV uses

our theory to interpret the reform experience of China. Section V concludes with some general implications

of our analysis.

I. The Model

5

We consider an economy with N identical regions. In each region, there are state enterprises and new

non-state enterprises. The economy can be viewed as a three-tier hierarchy composed of a central

government at the top, N local governments in the middle, and enterprises at the bottom. Although the setup

of the model is tailored to the situation of the Chinese economy, the logic of the model can be applied to other

federal states. The model has three dates.4

State Enterprises

At date 0, there are n (n is large) state enterprises, each having one project in place. The return on a

project has two components: a verifiable (taxable) return R and a non-verifiable (non-taxable) private benefiti

B (net of effort) accruing to employees (workers and managers). A type 1 project, present in proportion ",i

yields (R , B ) at date 1 independent of worker/manager effort (Figure 1). A type 2 project, present inq q

proportion (1 - "), yields (R , B ) at date 1 only if the enterprise chooses action (effort) e , which can beq q h

interpreted as "restructuring" or "privatization." If, however, action (effort) e is chosen, which meansl

maintaining the "status quo," then the project yields (0, 0) at date 1. In the latter case, the government and the

enterprise can engage in efficient renegotiation: if 1 unit of funds is injected (i.e., bailout), a type 2 project

will yield (R , B ) at date 2. For the sake of simplicity, we assume no discounting and R = 0. We assumes s s

further that the private benefits are ranked such that B > B > 0. This implies that if all verifiable revenuess q

are taxed away, the manager with a type 2 project prefers e to e if and only if bailout is expected.l h

[place Figure 1 here]

Non-State Enterprises

Non-state enterprises emerge at date 1 and their activities in region i are described by a production

function f(K , I ), where K represents non-state capital and I represents public infrastructure investmenti i i i

which is financed from the government budget at date 1. We make standard assumptions

f (K , I ) > 0, f (K , I ) > 0, f (K , I ) < 0, f (K , I ) < 0, and f (K , I ) > 0.K i i I i i KK i i II i i KI i i

6

Thus, public infrastructure investment raises the marginal productivity of non-state capital. For analytical

simplicity, we assume that the total amount of non-state capital is fixed (E K = K) and non-state capital isi

perfectly mobile across regions. All we really need for our qualitative conclusions, however, is the

assumption that non-state capital is less mobile across countries than across regions within a country.

We offer two interpretations of non-state capital. The first interpretation is domestic non-state or

private capital. With this interpretation, the assumption of inelastic supply of non-state capital is realistic

because formally domestic capital in China is not allowed to leave the country (even though some capital

flight is inevitable in practice). On the other hand, despite the fact that allocation of state capital remains

under state control, domestic non-state capital in China has a considerable degree of mobility across

provinces.

The second interpretation is foreign capital. In such a case, the assumption that foreign capital is less

mobile across countries than across regions within a country can be justified on empirical grounds in general

and in particular for the case of China. Following the influential paper by Martin Feldstein and Charles

Horioka (1980), extensive evidence has accumulated indicating substantial capital immobility across

countries in general (e.g., Roger Gordon and Lans Bovenberg, 1996). This includes a high correlation

between domestic savings and investment; real interest differentials across countries; and a lack of

international portfolio diversification. In contrast, capital mobility within countries is high. For example,

Tamim Bayoumi and Andrew Rose (1993) found strong evidence of capital mobility across regions in the

U.K. using the Feldstein-Horioka methodology.

But international capital mobility to China is particularly imperfect for idiosyncratic reasons. Unlike

most OECD and newly industrializing countries, China has maintained formal control over its capital

account. Officially, capital is not free to flow into or out of China: Capital inflow to China is controlled as

well as repatriation of profits. In fact, it was not until December 1996 that the Chinese government

announced the current account convertibility of its currency, but it has no plans to make capital accounts

convertible in the near future. These imperfections in international capital mobility need to be contrasted5

with much more perfect foreign capital mobility within China. Once a foreign firm manages to invest in

China, it can freely choose its location. The only consideration is profitability. Moreover, international

7

investors must incur some costs before investing in China to overcome language barriers and acquire

knowledge of Chinese laws, etc. Once these costs are sunk, it is much less costly to move capital from, say,

Guangdong to Shanghai.

The Government's Budget

We consider the government budget at date 1. In each region, on the revenue side, T - J is the neti i

tax revenue exclusively from state enterprises, where T is total taxes and J is transfers to state employees. i i

On the expenditure side, the budget is used for three purposes: subsidies (or bailout) S for type 2 projects;i

public infrastructure investment I and local public goods provision z :i i6

T - J = S + I + z .i i i i i

Actual tax revenues and expenditures will depend on whether the budget constraints of enterprises

are hard (action e and no bailout of type 2 projects) or soft (action e and bailout of type 2 projects). We canh l

rewrite the government budget as:

J + I + z = T - S = Ei i i i i

where E = E = nR or E = E = "nR - (1-")n (E > E ), depending on whether budget constraints are hard orH S H Sq q

soft.

We assume no tax on returns to non-state capital. This can be justified by the difficulty of such

taxation in practice. It is difficult to tax domestic non-state capital in transition economies due to the lack of

good tax institutions. Indeed, the great majority of taxes in China were collected from the state sector. It is

also well known that foreign firms are able to evade income taxes in host countries by transfer pricing

schemes. This is true in developed countries and more so in developing countries like China. However, in

Appendix A we show that our analysis goes through and our results continue to hold even with taxation of

returns to non-state capital.

8

The Government's Objective Function

We assume a welfare maximizing government: The local government in region i maximizes W andi

the central government maximizes W = E W . We assume that the population in region i consists of twoi

groups: employees in state enterprises and in non-state firms. Let the total private benefits of the employees

in non-state firms be

x = x(K , I ) = f(K , I ) - K f (K , I ),i i i i i i K i i

where x is assumed to be an increasing and concave function of I . The total private benefits of thei i

employees in state enterprises are y , where y = nB + J under hard budget constraints and y = "nB + (1-i i q i i q

")nB + J under soft budget constraints. The total utility of the two groups derived from local public goodss i

is u(z ), which is assumed to be concave in z . i i

Under the interpretation of K as domestic non-state capital, the objective function of local

government in region i is expressed as

W = f(K , I ) + y + u(z ),i i i i i

assuming the government puts an equal weight between rents to non-state employees and returns to non-state

capital. Under the interpretation of K as foreign capital, the objective function of local government in region i

becomes

W = x(K , I ) + y + u(z ),i i i i i

because the government is not concerned with returns to foreign capital. The two specifications will give

qualitatively similar results; for concreteness we will use the interpretation of foreign capital.

Equilibrium

9

We define the game played under federal institutions as follows:

The interaction between governments and state enterprises. In this sequential interaction an

equilibrium is defined by: a) a choice of action e of state enterprises at date 0 to maximize private benefits

given the expected choice of J , S , I and z by the government; and b) the government's choice of J , S , I andi i i i i i i

z at date 1 in order to maximize social welfare, given the choice of action by state enterprises. We define softi

budget constraint equilibria as subgame perfect equilibria in which enterprises with type 2 projects choose el

and the government chooses bailout, and hard budget constraint equilibria as equilibria in which enterprises

with type 2 projects choose e and the government chooses no bailout.h

The interaction among local governments. Except for the case of complete centralization, there is a

competition between local governments at date 1 to attract foreign capital and grants from the central

government. Local governments move simultaneously where strategies are choices of budget allocation. The

equilibrium is a Nash equilibrium in which each local government takes the budget allocation of other local

governments as given.

The interaction between the central and local governments. The interaction between the central

and local governments at date 1 concerns allocation of grants and monetary creation (except for the cases of

complete centralization and complete decentralization). This is a sequential interaction between the central

government and local governments. We will consider the case in which the central government makes

decisions (such as allocating grants and creating money) after the local governments' decisions, which reflects

the fact that the central government often accommodates the local governments' needs. We will also consider

the case of no accommodation in which the central government moves before local governments.

We define an equilibrium of the above game as strategy profiles of state enterprises, local

governments and the central government, such that none of the parties has incentives to deviate unilaterally.

Our analysis is considerably simplified because whether an enterprise has a soft or a hard budget constraint is

determined by a simple comparison between the benefits of bailout (B ) and the opportunity cost of doing so. s

Soft budget constraints are obtained when the benefits exceed the cost of doing the best alternative. Hard

budget constraints are obtained otherwise.

10

II. Fiscal Competition As a Commitment Device

This section demonstrates a main theme of our paper: Fiscal competition among local governments

may serve as a commitment device to harden budget constraints of enterprises.

A. The "First-Best" Allocation

In the "first-best" allocation, all inefficient projects are terminated (S = 0), and infrastructureiFB

investment (I ) and local public goods provision (z ) are chosen to maximize social welfare W = E W . i i iFB FB 7

The first best optimum (I , z ) satisfiesi iFB FB 8

(1) Mx(K , I )/MI = u'(z )i i i iFB FB

(2) I + z = E .i iFB FB H

B. Fiscal Centralization

In the benchmark case of fiscal centralization, the central government is entitled to all tax revenues,

has full authority over its budget, and has control rights over public investment and subsidies. Here the three-

tier hierarchy degenerates into a two-tier one in which local governments play no role. Define (I , z ) as ai iC C

solution to

(3) Mx(K , I )/MI = u'(z )i i i i

(4) I + z = E .i iS

The following proposition provides a condition under which hard budget constraints cannot be

achieved with fiscal centralization. Proofs of all propositions are in Appendix B.

Proposition 1: Under fiscal centralization, there exists a soft budget constraint equilibrium with I and zi iC C

but no hard budget constraint equilibrium, provided B > Mx(K , I )/dI . Furthermore, there iss i i iC

11

underinvestment in infrastructure I < I and underprovision of local public goods z < z .i i i iC FB C FB

There is a soft budget constraint equilibrium because, once enterprises with type 2 projects have

chosen e , the marginal benefit from bailing out an enterprise is greater than the marginal cost, measured byl

the welfare loss from less infrastructure investment and local public goods. Notice that because E > E , hardH S

budget constraints would yield more tax revenues and thus would allow a higher welfare level. Thus, the

government would prefer a commitment to no bailout.9

Comparative statics on ", the proportion of type 1 projects, yield interesting results which provide a

reason for the increased number of bailouts after the reforms of expanding enterprise autonomy in China and

other transition economies. To see this, assume that the government monitors enterprises with probability p,

thereby forcing enterprises with type 2 projects to choose e . Because monitored enterprises with type 2h

projects behave as if they had type 1 projects, the effective proportion of type 1 projects becomes "' = " + p(1

- "), which increases with p. Reforms for expanding enterprise autonomy reduce government monitoring and

have the effect of reducing ". A decrease in " worsens the average quality of the projects and thus increases

the number of bailouts under a soft budget constraint equilibrium.

C. Fiscal Decentralization

We now examine the effect of fiscal competition under the conditions of Proposition 1. With fiscal

decentralization, each local government is entitled to all tax revenues generated in its region. Moreover, it has

full authority over its own budget and has control over public investment and subsidies in its jurisdiction.

Because the number of projects n is large, each region seen in isolation is identical in structure to a centralized

nation. Because local governments are assumed to have the same objective function as the central

government, it is therefore not self-evident that fiscal decentralization should be superior in achieving harder

budget constraints.10

The crucial element is that fiscal decentralization induces competition among jurisdictions under

factor mobility. In addition to the game between government and enterprises, there is a game among local

governments. In our model, local governments compete with each other by making infrastructure investment

12

I to attract foreign capital K into their region. Given the choices of any other region j (j … i), locali i

government i chooses S , I , and z so as toi i i

max W = x(K , I ) + y + u(z )i i i i i

s.t. I + z = E.i i

Given any choices (I , ..., I ), perfect mobility of foreign capital implies that1 N

f (K , I ) = f (K , I )K i i K j j

for all i and j. This, together with equation E K = K, determines foreign capital allocation (K (I , ..., I ), ...,i 1 1 N

K (I , ..., I )) for any given (I , ..., I ).N 1 N 1 N

Since all regions are identical, we will look at symmetric Nash equilibria. Complementarity between

infrastructure investment and foreign capital inflow gives the following standard result:

Lemma 1: If f (K, I) < 0 and f (K, I) > 0, then dK /dI > 0 and dK /dI < 0 for all j … i.KK KI i i j i

Because of the fiscal competition, the first order condition for local government i becomes:11

(5) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) = u'(z ).i i i i i i i i i

Define I and z as a solution to (5) above and (6) belowi iD D

(6) I + z = E .i iH

Proposition 2: If Mx(K , I )/MI < B < Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ), then budget constraints arei i i s i i i i i i i iC D D

hard under fiscal decentralization but soft under fiscal centralization. However, hard budget constraints under

13

fiscal decentralization come at the cost of an allocative distortion: I > I and z < z .i i i iD C D C

The intuition is as follows. Competition among local governments in our model generates

externalities under intranational capital mobility: It increases the marginal regional value of infrastructure

investment above its marginal social value. This creates an allocative distortion with too much infrastructure

investment and too few local public goods for a given budget. This distortion increases the marginal utility

from local public goods provision and thus increases the opportunity cost of subsidizing type 2 projects in the

state sector. When this cost outweighs the benefit of bailout (B ), terminating type 2 projects becomess

credible. To the extent that under-provision of local public goods is a general feature of fiscal competition,

our results about fiscal competition hardening budget constraints should be robust.

The above results depend crucially on the difference between international and intranational mobility

of capital. Without this difference, there would be no difference between centralization and decentralization

because the central and local governments would compete in the same way for capital from the rest of the

world. On the other hand, our results do not depend on the assumption that foreign capital is not taxed. We

show in Appendix A that the results continue to hold for any given tax rate on foreign capital (or domestic

non-state capital). Indeed, taxation of foreign capital will reinforce the government's incentive to increase

infrastructure investment because an increase in foreign capital not only increases the employment benefits in

the non-state sector but also increases tax revenue.

Hard budget constraints under decentralization yield more tax revenues but the cost of obtaining

commitment is an under-provision of public goods, not only compared to the first-best solution but even

compared to fiscal centralization. Figure 2 shows a possible welfare comparison (in terms of the total utility

derived from infrastructure investment and local public goods, excluding the private benefits of the

employees in state enterprises) among the first-best optimum, the soft budget constraint equilibrium under

fiscal centralization, and the hard budget constraint equilibrium under fiscal decentralization. If the income

effect from hard budget constraints dominates the distortionary effect in the choice of I and z , fiscali i

decentralization yields a higher total utility from infrastructure investment and local public goods compared

to fiscal centralization.

14

[place Figure 2 here]

It is important to note the difference between commitment to hard budget constraints and

commitment to decentralization. Interestingly, they differ for the same reason -- irreversibility of earlier

decisions. The lack of commitment to hard budget constraints is due to the irreversibility of effort decisions

of enterprises at date 0, which gives rise to a bailout possibility at date 1. In contrast, because of the

irreversibility of expenditure decisions of local governments, commitment to fiscal decentralization is self-

enforced. Indeed, after observing the allocative choices of local governments at date 1, the central

government would find it impossible to bail out enterprises since revenues would have already been spent.

But the central government would have no incentives for recentralization before date 1 if the total welfare is

higher under fiscal decentralization. Recentralization is then either ineffective ex post, or undesirable ex ante.

D. Partial Fiscal Decentralization and Competition for Grants

In federal states, the central and local governments generally share fiscal revenues. One motivation

for fiscal transfers between the central and local governments is to "correct" distortions created by fiscal

competition between regions. In our model, this could happen through earmarked grants for local public

goods and/or subsidies. However, such "corrections" may have adverse consequences for incentives. One

may in particular wonder to what extent hard budget constraints can still be obtained when there are inter-

governmental transfers.

Consider a situation of partial fiscal decentralization where each local government transfers G to the

central government which then reallocates NG back to N local governments. Suppose the central government

earmarks grants for local public goods and subsidies. Partial fiscal decentralization introduces the12

possibility of soft budget constraints of local governments since the fiscal grants they receive ex post depend

on the expenditure decisions they make ex ante. But at the same time it also introduces a competition among

local governments for grants from the central government, in addition to competition for foreign capital.

Given the choices of any other region j (j … i), local government i chooses S , I , and z toi i iP P P

max W = x(K , I ) + y + u(z + z )i i i i i iG

15

s.t. I + z = E - G,i i

with the expectations of the rule determining z and S from the central government and K by foreign firms. i i iG G

Given the choices of local governments, the central government chooses (z , S ), i = 1, ..., N, toi iG G

max E {x(K , I ) + y + u(z + z )}i i i i iG

s.t. E (z + S ) = NG.i iG G

Foreign capital is allocated to equalize the marginal returns: f (K , I ) = f (K , I ), for all i and j.K i i K j j

Because foreign capital allocation K in a particular region depends only on the allocation of locali

infrastructure investment (I , ..., I ), the central government's allocation of grants does not have a direct1 N

impact on foreign capital allocation. Hence the central government's optimal allocation of grants must satisfy

u'(E - G - I + z ) = u'(E - G - I + z )i i j jG G

for all i and j. This, together with E z = NG (assuming S = 0), determines the allocation of grants (z (I ,i i 1 1G G G

..., I ), ..., z (I , ..., I )) for any given (I , ..., I ). We have an analogue to Lemma 1:N N 1 N 1 NG

Lemma 2: If u" < 0, then dK /dI > 0 and dK /dI < 0 for all j … i.i i j i

The mechanisms underlying competition for grants and for foreign capital are in fact the same: By

distorting more in the direction of higher I , a local government can obtain more grants from the centrali

government, given the choice of other regions. In any symmetric Nash equilibrium with S = 0, we will haveiG

z = G for all i. Let (I , z ) satisfyi i iG P P

(7) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) + (Mu/Mz)(dz /dI ) $ u'(z + z )i i i i i i i i i i i iP P G P G

(8) I + z = E - Gi iP P H

with a strict inequality holding under z = 0.iP

16

Proposition 3: Under partial fiscal decentralization with the earmarking of grants for local public goods and

subsidies and under the conditions of Proposition 2, there is a hard budget constraint equilibrium and no soft

budget constraint equilibrium if and only if G < z, where u'(z) = B .s

The competition for grants pushes local governments toward even more infrastructure investment

and keeps the budget constraints of enterprises hard as long as the grant size is not too large. When the size13

of the grant is small (i.e., G < z - z ), the central government is unable to provide a sufficient amount ofi iD P

local public goods. Competition for grants then leads to more distortions. Total local public goods provision

is even smaller than under complete decentralization, and so is total welfare. When the size of the grant is

moderate (i.e., z < G < z), local public goods provided by the central government through grants are larger,iD

but hard budget constraints still prevail because the central government prefers allocating grants for local

public goods rather than for bailouts. Welfare is also improved because the distortion by local governments

is partly corrected while maintaining hard budget constraints.

III. Fiscal Decentralization and Monetary Finance

In reality, the central government also has access to monetary sources of revenue. By incorporating

the possibility of monetary financing by the central government, we can analyze the relationship between

fiscal and monetary arrangements under federalism on the one hand, and soft budget constraints and inflation

on the other. Two effects will occur: (1) monetary finance will increase budgets and thus tend to soften

budget constraints; and (2) monetary finance may induce more distortions towards infrastructure investment

due to competition for monetary grants. To obtain a hard budget constraints equilibrium, the latter effect

must dominate the former.

We assume that R is the level of real resources from inflationary taxation available for additional

government expenditures. Because inflationary finance is distortionary, we assume that the total welfare loss

is equal to the amount of resources withdrawn R plus a quadratic deadweight loss (a/2)R . Therefore the2

central government maximizes

17

W = E{x(K , I ) + y + u(z )} - (R + (a/2)R ).i i i i2

A. The "First Best" Allocation

In the "first best" allocation, all inefficient projects are terminated (S = 0) and total monetaryFB

financing (R ), infrastructure investment (I ), and local public goods provision (z ) are chosen toFB FB FBi i

maximize total social welfare. We therefore have:

(9) Mx(K , I )/MI = u'(z ) = 1 + aR , i i i iFB FB FB

and

(10) z + I = E + R /N.i iFB FB H FB

B. Fiscal and Monetary Centralization

In the benchmark case of fiscal and monetary centralization, the central government maintains both

fiscal and monetary authority, but termination of inefficient projects is determined endogenously. Let (I ,iCC

z , R ) be the solution toiCC CC

(11) Mx(K , I )/MI = u'(z ) = 1 + aR ,i i i iCC CC CC

and

(12) I + z = E + R /N.i iCC CC S CC

Proposition 4: If soft budget constraints prevail under fiscal centralization without monetary finance, they

will also prevail under fiscal and monetary centralization. Moreover, soft budget constraints have

inflationary effects: R > R .CC FB

The possibility of monetary finance increases the government budget and thus reduces the

opportunity cost of subsidies. Therefore, the budget constraint of enterprises can only be made softer, the

softer the lower the marginal cost of inflation (1 + aR ). In the following discussions, we take as the caseCC

18

where budget constraints under fiscal and monetary centralization are soft as the benchmark case.

C. Fiscal Decentralization and Monetary Centralization

We first consider a situation which we call "accommodation," where the central government intends

to use monetary grants to "correct" the distortionary decisions of local governments. In such a game, the local

governments first decide on their expenditures, then the central government chooses total monetary finance

and allocates monetary grants to local governments earmarked for local public goods and subsidies.

Given the choices of I and z of local governments, the central government determinesi iDA DA

simultaneously aggregate monetary finance R and allocation of grants z (and possible subsidies S ) forDA G Gi i

each region to solve:

max E {x(K , I ) + y + u(z + z )} - (R + (a/2)R )i i i i iDA DA G 2

s.t. E (z + S ) = R,i iG G

The first order conditions of the central government are given by

(13) u'(z + z ) = u'(z + z ) = 1 + aR , i i j jDA G DA G DA

and

(14) E (z + S ) = R .i iG G DA

Anticipating the central government's decisions, local governments behave strategically and make

their choice (I , z ) according toi iDA DA

(15) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) + (Mu/Mz)(dz /dI ) $ u'(z + z )i i i i i i i i i i i iDA DA G DA G

(16) I + z = E,i iDA DA

with a strict inequality holding in (15) if z = 0.iDA

19

Proposition 5: There exists a hard budget constraint equilibrium under fiscal decentralization with monetary

accommodation if a > (B -1)/R . However, it is associated with higher inflation than that under fiscal andsDA

monetary centralization: R > R .DA CC

Proposition 5 shows that, due to the fiscal competition and competition for grant effects, hard budget

constraints may still be obtained under monetary centralization with accommodation when inflation is costly

enough (a is large). In such a case, the distortion towards infrastructure investment due to the competition for

monetary grants, which hardens budget constraints, will not be offset by the expansion of local budgets due to

inflationary finance. This situation in fact resembles closely that under partial fiscal decentralization without

monetary finance as presented in Proposition 3. Nevertheless, hard budget constraints are necessarily

associated with higher inflation than under fiscal and monetary centralization. Indeed, as local governments

behave strategically to invest less in local public goods, this increases the marginal benefit of inflation for the

central government. The latter then uses monetary creation to correct the distortions of the local

governments, which increases inflation. Clearly, when inflation is not too costly (a is small), budget

constraints can become soft.

Under what conditions is it possible to obtain lower inflation than under fiscal and monetary

centralization, together with hard budget constraints? To answer this question we examine the situation

opposite to the preceding one ("no accommodation"). The central government commits ex ante to a rule of

money creation and seigniorage allocation between regions (which eliminates competition for grants), and

local governments have full control over the use of revenues including the allocated seigniorage. More

precisely, the central government first decides total monetary creation R and allocates monetary grants to each

region. Afterwards, each local government decides on I , z and S , given their grant. For a given grant R/N, ai i i

Nash equilibrium I (R/N) and z (R/N) satisfies:i iDN DN

(17) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) = u'(z ).i i i i i i i i iDN DN DN

Anticipating the local governments' decisions, the central government thus chooses R to solve:

20

max E{x(K , I (R/N)) + y + u(z (R/N))} - (R + (a/2)R ).i i i iDN DN 2

The first order conditions of the central government are given by

(18) (Mx(K , I )/MI )(dI /dR) + u'(z )(dz /dR) = 1 + aR .i i i i i iDN DN DN DN DN

Proposition 6: Compared with fiscal and monetary centralization, budget constraints can only be hardened

under fiscal decentralization with no monetary accommodation. Moreover, inflation will be lower (R <DN

R ) provided dI /dR is sufficiently high.CC DNi

Because of fiscal competition, the central and local governments exhibit (endogenously) different

spending preferences since the marginal regional value of infrastructure investment is higher than the

marginal national value. Therefore the local governments will use an increased regional budget in a different

way (more for infrastructure investment) from the one the central government prefers (more for local public

goods). When the central government has no power to correct the distortion created by the local governments

towards infrastructure investment, total expenditures on local public goods under fiscal decentralization and

monetary centralization can only be lower than those under fiscal and monetary centralization. Therefore,

budget constraints will be harder since the marginal utility of local public goods is higher. Furthermore, the

conflict of interest between the central and local governments also reduces the marginal benefit from inflation

for the central government, and thus the central government has fewer incentives for monetary creation,

provided that fiscal competition induces significant responses from local governments (i.e., dI /dR isiDN

sufficiently high).

Comparing the results of Propositions 5 and 6 with those of Proposition 4, we see that higher

inflation is obtained under fiscal decentralization when the central government has the power to use monetary

creation to correct distortions made by local governments. In contrast, when the local governments have full

control over their expenditures and the central government has no discretion over the allocation of seigniorage

revenues between regions, both lower inflation and hard budget constraints can be achieved. The system of

21

separation of powers underlying Proposition 6 is thus better structured to achieve both objectives.

D. Fiscal and Monetary Decentralization

Although the case of monetary decentralization in which the central government completely loses

control over the money supply to local governments is extreme, it is empirically relevant in several occasions.

In the 1980s, some local governments in China (mainly provincial governments) managed to secure easy

credit from local branches of the central bank, which amounts de facto to a decentralization of the money

supply (Qian and Weingast, 1996). Similarly, Brazil had a recent experience where the money supply policy

was also de facto decentralized from the central bank to heavily indebted states, which controlled state-owned

banks (Wildasin, 1997b). Another striking example is the case of the Commonwealth of Independent States

where, after the break-up of the Soviet Union, independent republics continued to print rubles that circulated

in the former Soviet territory.

Decentralized monetary creation creates an inflationary externality as each local government gets the

full benefit from money creation but shares the costs of inflation with other regions, which leads to an

extreme form of a beggar-thy-neighbor policy, a "tragedy of the commons." In our model, because regions

are all identical, we assume that when region i secures R from inflationary finance, it only bears 1/N of thei

total cost. The local government in region i

max x(K ,I ) + y + u(z ) - (R + (a/2)R )/Ni i i i2

s.t. S + I + z = E + R ,i i i i

by taking other regions' decisions (including R , j … i) as given. The first order conditions are:j

(19) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) = u'(z ) = (1 + aR )/N.i i i i i i i i iDD DD DD DD

Proposition 7: Inflation is always higher under fiscal and monetary decentralization: R > R . DD CC

Furthermore, soft budget constraints also prevail as long as R # NR + (N-1)/a.DD CC

22

Because of the externality of inflation, inflation is always higher. Concerning budget constraints,

there are two offsetting effects: Fiscal competition tends to harden them while monetary expansion tends to

soften them. The latter effect tends to dominate because the inflation externality reduces the marginal cost of

inflation, leading to higher budgets with high I and z and thus with a lower opportunity cost of bailouts.i i14

IV. Interpreting the Reform Experience of China

The above analysis sheds light on China's reform experience and its unique features. It is well known

that in Eastern Europe and Russia, privatization is the main focus of transition and it is also considered a

major instrument in hardening the budget constraints of enterprises. Privatization is also largely driven by the

central government. In contrast, one of the most distinct features of China's transition has been associated

with devolution of authority from the central to local levels of government (Qian and Weingast, 1996).

Decentralization in China started as early as 1980 and has continued as a fundamental component of reform

since then. As a result, local governments in China at all levels have acquired authority as well as

responsibility over their own local economies. To see the extent of decentralization in China, consider, for

example, the relative importance of local government expenditures. In 1994, local government expenditure

represented 60% of government expenditures in China, compared with 34% in industrialized countries and

22% in developing countries (World Bank, 1996).

Decentralization induces fiscal competition among local governments. Indeed, local governments in

China compete vigorously in investing in infrastructure and establishing development zones in order to attract

foreign capital and domestic business into their regions. Harry Broadman and Xiaolun Sun (1997) show that

the level of infrastructure development, along with the market size, the educational level of the labor force,

and the geographical location, accounts for 80% of the variation in the stock of foreign direct investment

across provinces.

Decentralization also entails devolution of the supervision power of state-owned enterprises from the

central to local governments. During the 1980s, state-owned enterprises supervised by local governments

accounted for about three-quarters of the total industrial output in the state sector. Using a data set of 769

23

state enterprises covering this period, David D. Li (1997) found that state enterprises with directors

appointed by local governments were more likely to reduce workers' wage and bonuses in cases of poor

performance than the enterprises with directors appointed by the central government, where "poor

performance" is measured by an annual decrease in two alternative indicators: real net output and real total

returns to the state (tax, profits, plus interests). These results indicate that local government supervision over

state enterprises, as opposed to central government supervision, has an effect of hardening the enterprise's

budget constraints.

In recent years between 1994 and 1997, local governments in China took the initiative to massively

lay off excess workers from state-owned enterprises and even to privatize some enterprises under their

supervision (Yuanzheng Cao, Yingyi Qian, and Barry Weingast, 1997). Layoffs were mostly organized by

the local governments at the city level. In 1996 alone, about 10 million workers were laid off (most of them

were from state enterprises), and even more in 1997. Privatization, on the other hand, was most significant at

the county level. By 1996, some pioneering counties, such as Yibin of Sichuan, Shunde of Guangdong, and

Zhucheng of Shandong, had privatized almost all of the state-owned enterprises under local supervision; and

by 1997, most counties in the country had progressed to privatize more than half of the state-owned

enterprises under local supervision.

What is more interesting are the institutional changes behind (and, relatedly, the timing of) such local

government-led restructuring and privatization. Before 1993, China's monetary institutions were

problematic. On some occasions China experienced a de facto monetary decentralization when some local

governments forced the central bank branches in their regions to extend credit and print money; and on other

occasions the central bank made a monetary expansion hoping to correct distortions made by the local

governments (Wang, 1991; Qian and Weingast, 1996). Since 1993, China has centralized its monetary

authority and the central bank has refused to accommodate monetary policy to the local governments'

behavior (World Bank, 1996). As a result, a nationwide credit crunch was imposed and inflation has declined

drastically, from over 20 percent in 1993 to about 2 percent in 1997.

The local government-led restructuring and privatization of state-owned enterprises in China is an

important and interesting phenomenon. The existing theories on government-owned firms and privatization

24

(e.g., Shleifer and Vishny, 1994) cannot account for it, because they do not address the issue of how the

organization of government matters in providing government itself with incentives for reform. According to

our analysis, the incentives of local governments for restructuring and privatization can be induced

endogenously by decentralization and competition, reflecting harder budget constraints for local governments

and enterprises. Furthermore, the effect of hardening budget constraints of state-owned enterprises

supervised by local governments should be stronger under monetary centralization. Our analysis fits the fact

that local governments' initiatives of restructuring and privatization took place at the time when fiscal

decentralization was maintained together with monetary recentralization.

V. Conclusions

Until recently, the microeconomic literature on transition economies has focused largely on the

privatization and liberalization of markets in isolation from the reform of the organization of government in

these economies. In this paper, we focus on the relationship between the reform of the organization of

government and the incentive of government, and its effect on enterprises. We analyzed specifically how

fiscal and monetary arrangements under federalism limit the government's behavior and affect the degree of

the soft budget constraints of enterprises. Therefore, reform of the organization of government such as fiscal

decentralization is an important dimension of the transition process, and can have a profound impact on many

other dimensions of the transition.

Although our model was tailored to the Chinese economy, the themes we analyzed can apply to more

than China. The first general insight relates to the commitment effects of fiscal competition which tends to

change the incentives of government and limit its behavior. The literature on fiscal federalism has shown

how fiscal competition can inefficiently reduce government activity. Our paper has shown that in a second

best world fiscal competition may serve as a commitment device to prevent inefficient government spending.

The issue is important in the context of European unification, where the question of the adequate level of

fiscal centralization or decentralization is high on the policy agenda. It is also important in the U.S. context

where there are renewed debates over the advantages of fiscal expenditures being taken away from the federal

25

government and given to the states.

Second, the paper sheds light on the effects of fiscal decentralization on the composition of

government expenditures in general. Recent empirical evidence on developing and industrialized countries

alike shows that decentralization of fiscal authority leads to a change in the composition of public

expenditures, and, in particular, to an increase in the share of infrastructure expenditures (Antonio Estache

and Sarbajit Sinha, 1994). Our model precisely predicts such a change due to fiscal competition.

Third, the paper shows the importance of the link between monetary institutional arrangements and

the degree of fiscal decentralization. In our model, low inflation and hard budget constraints are associated

with the decentralization of fiscal expenditures and the centralized rules of money supply and allocation of

seigniorage revenues to local governments. Fiscal centralization would lead to higher inflation and softer

budget constraints. The issue is of particular concern for Europe from the perspective of monetary

unification. Because the issue of how to allocate seigniorage revenue to European Union members in a

monetary Union has not been settled, competition among national governments for monetary grants may lead

to higher inflation.

26

Appendix A

We show that even if we allow for taxation of returns to non-state capital, our results continue to

hold. First, consider the interpretation of K as foreign capital. Suppose the government imposes an

exogenously given tax rate t on the returns to foreign capital (our original model corresponds to the case of t

= 0). Under centralization, the central government receives the tax revenue, and under decentralization, the

local governments receive the tax revenue. We want to compare centralization and decentralization for any

level of t, 0 # t # 1.

Under fiscal centralization, by symmetry, the central government, taking K =K/N as given,i

max x(K , I ) + u(z )i i i

s.t. I + z = E + tK f (K , I ).i i i K i i

The first order conditions become

(A1) Mx(K , I )/MI = u'(z )[1 - tK f (K , I )],i i i i i KI i i

I + z = E + tK f (K , I ).i i i K i i

When t > 0, as compared with the case of t = 0, the government wants to spend more on I and less on

z in order to get more tax revenue for any given budget. Therefore, introducing taxation tends to harden the

budget constraint. On the other hand, the income effect tends to increase z, thus softening budget constraints.

The net effect is not clear. But our main interest is to compare centralization with decentralization for the

same t.

Under fiscal decentralization, perfect capital mobility implies that

(1-t)f (K , I ) = (1-t)f (K , I ),K i i K j j

27

which is the same as t = 0. Thus, foreign capital allocation (K (I , ..., I ), ..., K (I , ..., I )) for any given (I ,1 1 N N 1 N 1

..., I ) remains the same.N

However, the first order conditions for local governments are changed to

(A2) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) = u'(z )[1-td(K f (K , I ))/dI ],i i i i i i i i i i K i i i

I + z = E + tK f (K , I ).i i i K i i

Note that

td(K f (K , I ))/dI = t[(f (K , I )+Kf (K , I ))(dK /dI ) + K f (K , I )]i K i i i K i i KK i i i i i KI i i

which is greater than

tK f (K , I )i KI i i

provided

(A3) f (K , I ) + Kf (K , I ) > 0.K i i KK i i

We assume (A3) holds (this is satisfied for Cobb-Douglas function f = K I )." 1-"

Comparing (A1) and (A2), starting with the optimal solution under centralization, then under

decentralization, the left hand side becomes bigger, and the right hand side becomes smaller. Therefore,

adding taxation reinforces the fiscal competition effect because the local government has more incentives to

increase infrastructure investment. Therefore, with taxation t, under condition (A3), our results on

decentralization are strengthened. The intuition is that, under decentralization, more investment in

infrastructure attracts more foreign capital, which not only increases employment rents but also increases tax

revenues. This double benefit makes the local governments spend more of their budget on infrastructure and

less on other local public goods, which thus raises the opportunity cost for bailout.

28

The benefit of bailout is still B , and the opportunity cost of bailout is still u'(z ). Then ifs i

u'(z ) < B < u'(z ),i s iC D

or equivalently,

{Mx(K , I )/MI } / {1 - tK f (K , I )} < Bi i i i KI i i sC C

< {Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI )} / {1-td(K f (K , I ))/dI },i i i i i i i i i K i i iD D D

budget constraints are hard under fiscal decentralization but soft under fiscal centralization.

To conclude, our result in comparing centralization and decentralization continues to hold,

independent of the tax rate t.

Second, consider the interpretation of K as domestic non-state capital. The objective function

government in region i becomes:

W = x(K , I ) + (1-t)K f (K , I ) + y + u(z )i i i i K i i i i

where t is the tax on non-state capital.

When t = 0, it is easy to see that our results go through, because in all the proofs we simply replace

x(K , I ) by f(K , I ). For any general t>0, we show that our analysis still goes through. Under centralization,i i i i

the central government,

max x(K , I ) + (1-t)K f (K , I ) + y + u(z )i i i K i i i i

s.t. I + z = E + tK f (K , I ).i i i K i i

The first order conditions become

29

(A4) Mx(K , I )/MI + (1-t)K f (K , I ) = u'(z )[1 - tK f (K , I )],i i i i KI i i i i KI i i

I + z = E + tK f (K , I ).i i i K i i

Under decentralization, with the same tax rate t, the first order conditions for local governments are

changed to

(A5) Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) + (1-t)d(K f (K , I ))/dIi i i i i i i i i K i i i

= u'(z )[1 - td(K f (K , I ))/dI ],i i K i i i

I + z = E + tK f (K , I ).i i i K i i

Comparing (A4) and (A5) under assumption (A3), starting with the optimal solution under

centralization, then under decentralization, the left-hand side becomes bigger, and the right-hand side

becomes smaller. Therefore, our results on the hardening of budget constraints under decentralization remain

valid.

30

Appendix B

Proof of Proposition 1: Because the cost of not bailing out m (m$1) projects is mB but the benefit is lesss

than mu'(z ), "no bailout" is not a worthwhile deviation under the specified condition. Comparing (1) andiC

(3), E > E together with concavity of u(z ) implies that I < I and z < z . Concavity of x(K , I ) withH S C FB C FBi i i i i i i

respect to I ensures thati

B > Mx(K , I )/MI > Mx(K , I )/MI .s i i i i i iC FB

Hence there cannot be a hard budget equilibrium (which has to be I = I ).�i iFB

Proof of Proposition 2: The condition implies that bailout is not worthwhile. If there were a soft budget

constraint equilibrium with I , we would have, considering (5) and (6) and E >E ,i* H S

Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) = u'(E - I ) > u'(E - I ) > B ,i i i i i i i i i i S* * S * H D

then bailout is not preferred, a contradiction. Comparing (3) and (5), we have u'(z ) < u'(z ), hence z > z . i i i iC D C D

Finally, E > E implies that I < I .�H S C Di i

Proof of Proposition 3: When (7) holds with equality, comparing (5) and (7), dz /dI > 0 implies for thei iG

same budget E = z + G = z + z < z . Hence, u'(z + z ) > u'(z ) > B by the condition of Proposition 2. H P P G D P G Di i i i i i i s

Then neither the central government nor local governments have incentives for bailout. When (7) holds with

inequality, if z + z = G < z , then the left hand side of (7) is greater than u'(z + z ), which is greater thani i i i iP G D P G

u'(z ) and B . If z + z = G > z , we must have I < I , then the left hand side of (7) is greater thani s i i i i iD P G D P D

Mx(K , I )/MI +(Mx(K , I )/MK)(dK /dI ),i i i i i i i iD D

which is in turn greater than B . In both cases, the local governments have no incentives for bailout. Unders

the condition G < z, u'(G) > B , the central government has no incentives for bailout either.s

31

To show there is no soft budget equilibrium, use (7) and apply the proof of proposition 2.�

Proof of Proposition 4: If R < R , E +R /N < E + R /N as E < E . The smaller budget would implyCC FB S CC H FB S H

u'(z ) = 1 + aR > u'(z ) = 1 + aR , or R > R .�i iCC CC FB FB CC FB

Proof of Proposition 5: Take a high enough so that R > 0 is arbitrarily small. This implies that (15) holdsDA

with equality since z > 0. Suppose z > z . Since I + z = I + z + R /N and R is small, theni i i i i i i i iDA DA D DA DA D D DA DA

I < I . This would imply that (15) holds with inequality, a contradiction. Hence z # z < z and z +i i i i iDA D DA D DA

R /N < z. Furthermore, under a hard budget constraint, u'(z + z ) > B > u'(z ), using (11) and (13). DA DA G CCi i s

This implies R > R .�DA CC

Proof of Proposition 6: z > z would imply u'(z ) < u'(z ), the budget constraint would remain soft. i i i iDN CC DN CC

By (17), we would obtain Mx(K , I )/MI < u'(z ) < u'(z ). Using dI /dR + dz /dR = 1, we derivei i i i i i iDN DN CC DN DN

1 + aR = (Mx(K , I )/MI )(dI /dR) + u'(z )(dz /dR) < u'(z ) = 1 + aR .DN DN DN DN DN CC CCi i i i i i i

Therefore, R < R , which would imply z # z , a contradiction. Therefore, z # z , that is, theDN CC DN CC DN CCi i i i

budget constraint can only be made harder. If u'(z ) > B > u'(z ), the budget is hard.i s iDN CC

Given u'(z ) > u'(z ), we establish, using dI /dR + dz /dR = 1, (17) and (18),i i i iDN CC DN DN

R < RDN CC

if and only if

[Mx(K/N, I )/MI - u'(z )](dI /dR) < u'(z ) - u'(z )i i i i i iDN DN DN CC DN

if and only if

dI /dR > {u'(z ) - u'(z )}/[(Mx(K/N, I )/MK)(dK /dI )].�i i i i i iDN DN CC DN

Proof of Proposition 7: R > R would imply 1 + aR > (1 + aR )/N, thus u'(z ) > u'(z ) and z <CC DD CC DD CC DD CCi i i

z . As B > u'(z ) > u'(z ), budget constraints are always soft. But R > R and z > z imply I >i s i i i i iDD CC DD CC DD DD CC CC

I , which gives iDD

Mx(K , I )/MI + (Mx(K , I )/MK)(dK /dI ) > Mx(K , I )/MI , i i i i i i i i i i iDD DD CC

32

that is, u'(z ) > u'(z ), a contradiction.i iDD CC

By (11) and (19), R < NR + (N - 1)/a if and only if u'(z ) < u'(z ). Then the budget constraintDD CC DD CCi i

is soft if R < NR + (N - 1)/a.�DD CC

33

REFERENCES

Bayoumi, Tamim A. and Andrew K. Rose, "Domestic Savings and Intra-National Capital Flow," European

Economic Review, 1993, 37, 1197-1202.

BBroadman, Harry, and Xiaolun Sun, "Distribution of Foreign Direct Investment in China," Working Paper

1720, World Bank, Washington, DC, February, 1997.

Cao, Yuanzheng, Yingyi Qian, and Barry Weingast, "From Federalism, Chinese Style, to Privatization,

Chinese Style," Working Paper No. 16, Center for Research on Economic Development and Policy

Reform, Stanford University, 1997.

Dewatripont, Mathias and Maskin, Eric, "Credit and Efficiency in Centralized and Decentralized Economies,"

Review of Economic Studies, October, 1995, 62, 841-555.

Estache, Antonio and Sarbajit Sinha, "Comparing the Effect of Decentralization on Public Infrastructure

Expenditure in Developing and Industrialized Countries", World Bank, 1994.

Feldstein, Martin and Charles Horioka, "Domestic Savings and International Capital Flows", Economic

Journal, 1980, 90, 316-329.

GGordon, Roger H., "An Optimal Taxation Approach to Fiscal Federalism," Quarterly Journal of Economics,

1983, 98, 567-586.

Gordon, Roger H., and A. Lans Bovenberg, "Why Is Capital So Immobile Internationally? Possible

Explanations and Implications for Capital Income Taxation," American Economic Review,

December, 1996, 86(5), 1057-75.

Kornai, Janos, Economics of Shortage, Amsterdam: North Holland, 1980.

Kornai, Janos, "The Soft Budget Constraint," KYKLOS, 1986, 39(1), 3-30.

Laffont, Jean-Jacques, and David Martimort, "Duplication of Regulators Against Collusive Behavior,"

mimeo, IDEI, 1994.

Li, David D., "Government Control of State Enterprises: Evidence from China," mimeo, University of

Michigan, 1997.

Oates, Wallace, Fiscal Federalism, New York: Harcourt Brace Jovanovich, 1972.

34

Persson, Torsten, Gérard Roland, and Guido Tabellini, "Separation of Powers and Political Accountability,"

Quarterly Journal of Economics, November, 1997, CXII(4), 1163-1202.

Qian, Yingyi, and Barry R. Weingast, "China's Transition to Markets: Market-Preserving Federalism,

Chinese Style," Journal of Policy Reform, 1996, 1, 149-185.

Qian, Yingyi, and Barry R. Weingast, "Federalism As a Commitment to Preserving Market Incentives,"

Journal of Economic Perspectives, Fall, 1997, 11(4), 83-92.

Segal, Ilya, "Monopoly and Soft Budget Constraint," mimeo, Harvard University, 1993.

Shleifer, Andrei and Robert Vishny, "Politicians and Firms," Quarterly Journal of Economics, November,

1994, CIX, 995-1025.

Tiebout, Charles, "A Pure Theory of Local Expenditures," Journal of Political Economy, 1956, 64, 416-24.

Wang, Yijiang, "Economic Reform, Fixed Capital Investment Expansion, and Inflation: A Behavior Model

Based on the Chinese Experience," China Economic Review, 1991, 2(1), 3-27.

Wildasin, David E., "Nash Equilibria in Models of Fiscal Competition," Journal of Public Economics, 1988,

35, 229-240.

Wildasin, David E., "Externalities and Bailouts: Hard and Soft Budget Constraints in Intergovernmental

Fiscal Relations," mimeo, Vanderbilt University, 1997a.

Wildasin, David E., "Fiscal Aspects of Evolving Federalisms: Issues for Policy and Research," in D.E.

Wildason (ed.) Fiscal Aspects of Evolving Federalisms, Cambridge University Press, forthcoming,

1997b.

Weingast, Barry R., "The Economic Role of Political Institutions: Market-Preserving Federalism and

Economic Growth," Journal of Law, Economics and Organization, 1995, 11, 1-31.

World Bank, The Chinese Economy: Fighting Inflation, Deepening Reforms, A World Bank Country Study,

1996.

35

1 Subsidization of loss-making enterprises is often an indicator of soft budget constraints. However,

subsidies are not identical to soft budget constraints. There are cases in which firms receive subsidies but do

not expect to be bailed out in cases of bad financial performance.

2 In a recent paper, David Wildasin (1997a) studied the issue of hard vs. soft federal grants to local

governments. He argues that, due to an externality of local public goods, large local jurisdictions are likely to

face softer budget constraints than small jurisdictions. This is because the failure to bail out the former

would have far more severe consequences for the economy as a whole, leading to a "too big to fail" effect.

3 Both the models of Andrei Shleifer and Robert Vishny (1994) and Ilya Segal (1993) on soft budget

constraints are based on the difference in objectives between the government and the private sector. In the

former, the government cares about employment and in the latter about the total social surplus.

4 By focusing on a two-period model we leave aside the role of reputation in sustaining hard budget

constraints. As is known from the theory of repeated games, reputation is an effective mechanism only if

there is a long enough time horizon and a low enough discount rate. In this paper, we prefer to focus on the

role of institutions in hardening budget constraints.

5 In this regard, it is useful to make a comparison between China and some other newly industrializing

countries such as Mexico, Thailand, Malaysia, Indonesia, and South Korea, which have experienced financial

crises recently. Unlike those countries, foreign capital is not allowed to go into China freely, and thus foreign

investors are not able to respond to China's higher domestic interest rate in the same way as they do in those

countries. Indeed, short-term and portfolio investment opportunities in China are not available to foreign

Footnotes

* Qian: Department of Economics, Stanford University, Stanford, CA 94305-6072; and Roland: ECARE,

Université Libre de Bruxelles and CEPR. The authors are grateful to Erik Berglöf, Patrick Bolton, Jiahua

Che, Mathias Dewatripont, David D. Li, Eric Maskin, Ronald McKinnon, Barry Naughton, Yijiang Wang,

Barry Weingast, David Wildasin, Chenggang Xu, three referees, and seminar participants in Brussels,

Stanford, Gerzensee and the CEPR annual workshop in macroeconomics in Tarragona for helpful discussions

and comments.

36

investors. Such restrictions on foreign capital are responsible, in part, for China's ability to avoid financial

crises like those as in Mexico and Southeast Asian countries.

6 The difference between z and I is that the choice of the latter exerts an externality on other regions whilei i

the former does not. Whether z is a form of public investment or consumption is irrelevant for our results.i

7 The "first best" is in reference to domestic welfare.

8 We assume that u'(z ) is greater than 1. This ensures that at the optimum J = 0, which simplifies ouri iFB

analysis. The model can, however, be accommodated to the case where u'(z ) = 1 and J > 0.i iFB

9 In our model soft budget constraints are a consequence of a distortionary taxation (all verifiable revenues

are taxed away). If enterprises were allowed to keep (R - 1) of their verifiable revenues, then enterprisesq

with type 2 projects would get (R - 1) + B under e and only B under e . However, because the two types ofq q h s l

projects are indistinguishable, in order to induce e from type 2 projects, the government would have toh

concede B - B to all enterprises. If B - B > (1 - ")(R + 1), this leads to a fall in tax revenues and thus to as q s q q

lower welfare level as compared to a soft budget constraint equilibrium. In such a case, the government

prefers to tax away all verifiable revenues.

10 The fact that local governments do not receive grants from the central government (i.e., the budget

constraints of local governments are hard) is not really relevant here. Indeed, under centralization, the fiscal

budget of the central government is also hard but enterprises have soft budget constraints as shown in

Proposition 1.

11 For this first order condition to make sense, we assume that the left-hand side, i.e., the marginal value of

infrastructure investment to region i, is decreasing in I . One can verify that this will be the case if f(K , I )i i i

takes the form of a Cobb-Douglas function.

12 The general case where the central government can also earmark grants for infrastructure investment is

technically more complicated and does not yield additional economic insight in the current set-up, given that

local governments already overspend in infrastructure investment.

13 Yijiang Wang (1991) demonstrates similar distortionary effects in a model with a Cobb-Douglas

production function (with fixed investment and circulating capital as two inputs).

37

14 A pathological case of hard budget constraints under very high inflation may prevail when R > NR +DD CC

(N-1)/a. In that case, there will be both very high inflation and a high marginal utility of local public goods.

This is a case where the marginal value of infrastructure investment is very high and decreases slowly so that

the local public goods provision is always very small for any budget size.

"

1-"

eh

el

(Rq, Bq)

(Rq, Bq)

(0, 0)

(Rs, Bs)

date 0 date 1 date 2

termination

bailout

type 1

type 2

38

Figure 1.

The Mechanism of the Soft and Hard Budget Constraint

0 ES EH

I i

zi

( I , z )FB FB

( I , z )C C

( I , z )D D_z

39

Figure 2.

Comparison among the First-Best Optimum, the Soft Budget Constraint Equilibrium under FiscalCentralization, and the Hard Budget Constraint Equilibrium under Fiscal Decentralization