This action might not be possible to undo. Are you sure you want to continue?

# Designing domestic institutions for international monetary

**policy cooperation: A Utopia?
**

Florin O. Bilbiie

a, b, c,

*

a

Paris School of Economics, Paris, France

b

Université Paris 1 Panthéon-Sorbonne, Paris, France

c

CEPR, London, United Kingdom

JEL classiﬁcaion:

F33

F42

E58

E61

D62

Keywords:

International monetary policy cooperation

Institutions

Equilibrium optimal delegation

Inﬂation contracts

Credible

Subgame perfect contracts

a b s t r a c t

In a wide variety of international macroeconomic models mone-

tary policy cooperation is optimal, non-cooperative policies are

inefﬁcient, but optimal policies can be attained non-cooperatively

by optimal design of domestic institutions/contracts. We show that

given endogenous institutional design, inefﬁciencies of non-

cooperation cannot and will not be eliminated. We model the

delegation stage explicitly and show that subgame perfect, cred-

ible contracts (chosen by governments based on individual ratio-

nality) are non-zero, but are different from optimal contracts and

hence lead to inefﬁcient equilibria. Optimal contracts require

cooperation at the delegation stage, which is inconsistent with the

advocated non-cooperative nature of the solution. A general

solution method for credible contracts and an example from

international monetary policy cooperation are considered. Our

results feature delegation as an equilibrium phenomenon, explain

inefﬁciencies of existing delegation schemes and hint to a poten-

tially stronger role for supranational authorities in international

policy coordination.

Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction

A large body of literature deals with optimal delegation of macroeconomic policy in an international

context (see Persson and Tabellini, 2000 for a comprehensive review). In this framework, optimal

contracts or targeting regimes over some macroeconomic variable are viewed as panacea for solving

* Centre d’Economie de la Sorbonne, 106-112 Boulevard de l’Hopital, Paris 75013, France. Tel.: þ33 1 44078202.

E-mail address: ﬂorin.bilbiie@parisschoolofeconomics.eu.

URL: http://ﬂorin.bilbiie.googlepages.com

Contents lists available at ScienceDirect

Journal of International Money

and Finance

j ournal homepage: www. el sevi er. com/ l ocat e/ j i mf

0261-5606/$ – see front matter Ó 2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.jimonﬁn.2011.01.011

Journal of International Money and Finance 30 (2011) 393–409

inherent inefﬁciencies of non-cooperative (and discretionary) policymaking. Notably, much of the

work concerning monetary policy institutions adopts this line of reasoning. The inefﬁciencies that

optimal delegation is supposed to ‘ﬁx’ in this case are problems due to non-cooperative policymaking

in the presence of policy spillovers in a multi-country world (and/or ‘credibility’ problems like the

inﬂation bias). A recurrent result is that the cooperative optimum can be achieved in a decentralized,

non-cooperative manner by delegating through optimal inﬂation contracts (i.a. Persson and Tabellini,

1995, 1996). This is done by assuming that before the actual policy game takes place, there is an

‘institutional design stage’, where governments choose the appropriate delegation scheme for their

central banks that implements the optimum.

This paper starts from the observation that these delegation schemes are not subgame perfect, i.e.

not credible: indeed, they implicitly assume cooperation (or some form of coordination) at the dele-

gation stage, which is hard to reconcile with the alleged ‘purely non-cooperative implementation of the

cooperative optimum’. We develop this argument analytically by explicitly modeling the institutional

design stage and studying the (credible, subgame perfect) contracts that are consistent with govern-

ments’ incentives and hence occur in equilibrium. Speciﬁcally, at the delegation stage governments

choose the delegation parameters in a non-cooperative manner by backward induction, taking into

account the reaction functions of the central banks at the policy stage. These credible, subgame perfect

contracts turn out to be non-zero (hence delegation is always an equilibrium) whenever there is

strategic complementarity or substitutability. However, they are always different from the optimal

contracts, which would instead require cooperation of governments (or some form or coordination) at

the delegation stage. But then, if binding agreements were possible, one wonders why would dele-

gation be needed in the ﬁrst place.

Intheinternational policycontext, it has beenlongrecognizedthat cooperativepolicymaking

1

is Pareto

optimal when sovereign policymaking has externalities on the other countries (see e.g. Hamada, 1976).

Typically, externalities take the form of conﬂicts over shock stabilization or over preferred levels of

macroeconomic outcomes. The Pareto optimum is not enforceable for various reasons (individual

incentives to deviate, suboptimality of cooperation when commitment with respect to the domestic

private sector is impossible, uncertainty regarding models, loss functions, etc.) - all these issues are

extensively reviewed in Canzoneri and Henderson (1991) or Ghosh and Masson (1994). Given individual

incentives to deviate fromthe optimal cooperative policies, the literature has moved towards identifying

mechanisms that sustainthe collusive outcome. We focus onthe‘institutional design’ approachpioneered

by Persson and Tabellini (1995) and extended by Jensen (2000) and Persson and Tabellini (1996, 2000).

This focus is reinforced in the international context by an observation of Rogoff (1985): in the presence of

domestic credibility problems as the ones reviewed above cooperation itself might even be welfare-

reducing.

2

But institutional design, or delegationtoanindependent monetaryauthority, couldinprinciple

act as a solution to correct inefﬁciencies coming fromboth discretion and non-cooperative policymaking.

The state of the art in the literature on optimal monetary policy delegation in an international

context can be summarized as follows. Persson and Tabellini (1995, 1996) analyze performance

contracts written by the governments before the game is played, at an ‘institutional design’ stage and

showhowthese contracts can be designed such that the inefﬁciencies related to both discretionary and

non-cooperative policymaking are eliminated.

3

The optimal linear contracts hence found are state-

contingent, which is a non-desirable feature as it makes them difﬁcult to implement (for example

1

We adopt the game-theoretical deﬁnition of cooperation as joint optimization by a group of players of their payoffs,

implying a ‘pregame’ and the possibility of binding agreements. Coordination would by contrast mean choosing one particular

equilibrium in the Nash Equilibrium set of the non-cooperative game (this might imply the presence of an external enforcing

mechanism). Exchange of information is captured by the non-cooperative policymaking case.

2

Canzoneri and Henderson (1991) interpret this insight as a particular case of a more general result: coalitions of only

subsets of players are inefﬁcient. See also Kohler (2002).

3

Persson and Tabellini also look at non-linear discontinuous performance contracts with state-dependent parameters

written directly over welfare functions that can implement the cooperative optimum. This is an application of a Folk Theoremin

Delegation Games by Fershtman et al. (1991), where it is argued that in a two-player game the principals can obtain every

Pareto optimal outcome as the unique subgame perfect Nash Equilibrium of the delegation game via such contracts written on

target compensation form, as long as these contracts become common knowledge. However, the authors move to analyzing

linear contracts, arguing that these are highly non-realistic and difﬁcult to implement.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 394

because they imply that the institution changes each time a shock occurs). However, Jensen (2000)

addresses this issue by ﬁnding state-independent transfer functions that implement the cooperative

outcome. These functions penalize quadratically inﬂation deviations froma certain level (chosen by the

government) as well as inﬂation differentials between the two countries. He also provides interpre-

tations of these contracts in terms of real-life institutions. A general criticism of this line of research is

that welfare conclusions and prescriptions cannot be properly addressed in a model that lacks

microfoundations (Obstfeld and Rogoff, 1996). However, recent research shows that the insights of

optimal design of institutions carries over to more realistic setups in the new open-economy macro-

economics tradition. In a recent insightful contribution, Benigno and Benigno (2005) use a micro-

founded, general equilibrium two-country model and show that targeting rules can be designed that

implement optimal cooperative policies, and that optimal contracts exist that could make these tar-

geting rules occur in a non-cooperative equilibrium.

4

The remainder of the paper proceeds as follows. Section 2 sets the stage and develops a general

version of our argument in a simple linear-quadratic two-country model with spillovers/externalities.

We derive the optimal contracts that implement the cooperative optimum under non-cooperative

(Nash) policymaking, and introduce the notion of ‘credible’ (subgame perfect) contracts; Section 2.2

extends this framework to a setup where there is a domestic credibility (commitment) problem.

Section 3 applies the general results of Section 2 to a simple model of international monetary policy

cooperation due to Persson and Tabellini (1996, 2000); it shows that, and explains why, credible

subgame perfect contracts are different from optimal contracts. Section 4 concludes and points out

some implications for the design of supranational institutions.

2. Credible, subgame perfect contracts in a general linear-quadratic framework

In this sectionwe describe a general solution method for credible contracts (as a shorthand notation

for subgame perfect, non-cooperative contracts) as opposed to optimal contracts in a two-country

model with policy spillovers.

5

We start with a simplest setup in which there is no domestic credibility

problem and prove formally three propositions. The ﬁrst states that delegation always occurs in

equilibrium when there is strategic complementarity or substitutability (in the sense that one coun-

try’s equilibrium strategy is increasing/decreasing in the other country’s strategy). The second states

that optimal contracts – that implement the cooperative optimum in the non-cooperative game-

require cooperation at the delegation stage, i.e. are a solution to government’s delegation problemonly

under cooperation. The third shows that these cooperative contracts (optimal contracts) are always

different from subgame perfect contracts - that are chosen by governments based on their individual

rationality. We then describe the solution method for a more general setup in which we allow for

domestic credibility problems.

2.1. A simple but general version of the argument

Following most of the literature on policy coordination and institutional design reviewed above,

suppose that there are two countries, home and foreign, and distinguish the foreign country by an

asterisk. In each country a policymaker (the government) minimizes a quadratic aggregate loss

function

6

deﬁned over deviations of some macroeconomic variables, which are related linearly to the

policy instruments, i and i* respectively. The loss functions are therefore L(i,i*) and L*(i,i*), and their

being quadratic implies that second-order derivatives are constant.

7

We assume that externalities

4

See also Benigno (2002), for an earlier effort in this direction, using the framework of Corsetti and Pesenti (2001).

5

Speciﬁc (parametric) examples of models most related to this one are i.a. Persson and Tabellini (1995, 2000). We provide an

example in the next section.

6

This loss function is usually quadratic and directly postulated, although possible to derive as a quadratic approximation of

an aggregate welfare function describing society’s preferences (Woodford, 2003; see Benigno and Benigno, 2005 for the open-

economy case).

7

We denote a derivative with respect to an argument by appending the corresponding argument as a subscript, i.e. L

ii

Ã is the

cross-derivative of L.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 395

(spillovers) are present in the sense that L

i

Ã s0; L

Ã

i

s0; and their sign depends on the sign of these

derivatives.

2.1.1. The Nash, non-cooperative equilibrium

The Nash, non-cooperative equilibriumconsists of strategies i and i* that solve the following system

of linear ﬁrst order conditions (they are linear because the loss functions are quadratic):

L

i

ði; i

Ã

Þ ¼ 0 and L

Ã

i

Ã

ði; i

Ã

Þ ¼ 0: (1)

Solving this system one can ﬁnd the (linear) reaction (or best response) functions of each policy-

maker as a function of the other policymaker’s strategy, i

N

(i*) and i*

N

(i) By implicit differentiation of (1)

we obtain the slopes of these reaction functions as:

vi

N

vi

Ã

¼ À

L

ii

Ã

L

ii

;

vi

ÃN

vi

¼ À

L

Ã

i

Ã

i

L

Ã

i

Ã

i

Ã

:

Note that strategic complementarity ðvi

N

=vi

Ã

> 0; vi

ÃN

=vi > 0Þ or substitutability (<0) will depend

on the sign of the second derivatives L

ii

Ã and L

Ã

i

Ã

i

; as in Cooper and John (1988). Under usual concavity

assumptions on L and L* existence and uniqueness of equilibrium require that these slopes, in the (i,i*)

space, be different or equivalently, that the Jacobian of the (1) systembe non-singular. This condition is:

DhL

ii

L

Ã

i

Ã

i

Ã

À L

ii

Ã L

Ã

i

Ã

i

s0: (2)

2.1.2. The cooperative equilibrium

The cooperative equilibrium is obtained by joint minimization of an aggregate, global loss function

Lði; i

Ã

Þ ¼ Lði; i

Ã

Þ þ L

Ã

ði; i

Ã

Þ with respect to both instruments. The strategies that implement this

cooperative equilibrium (denoted by a C superscript) i

C

and i*

C

will solve the linear system:

L

i

ði; i

Ã

Þ ¼ L

i

ði; i

Ã

Þ þ L

Ã

i

ði; i

Ã

Þ ¼ 0 and

L

i

Ã ði; i

Ã

Þ ¼ L

i

Ã ði; i

Ã

Þ þ L

Ã

i

ði; i

Ã

Þ ¼ 0:

(3)

The slopes of the optimal reaction functions can be found by implicit differentiation as before and

compared to the Nash reaction functions. The condition for existence and uniqueness of the equilib-

rium is in this case: L

ii

L

i

Ã

i

Ã ÀL

ii

Ã L

i

Ã

i

s0: The non-cooperative Nash equilibrium is inefﬁcient in the

presence of externalities since the terms L

Ã

i

and L

i

Ã are non-zero.

2.1.3. Delegation

Delegation to independent policy authorities is a prominent solution to solve these inefﬁciencies

while preserving non-cooperative policymaking, i.a. in the pioneering work of Persson and Tabellini

(1995).

8

Consider that before the non-cooperative game is played, each government delegates policy

to an independent policy authority by imposing a certain transfer function T(i); assuming further that

these contracts or transfer functions are linear (since the model is linear-quadratic), the delegated loss

functions L

D

become:

L

D

¼ L þ ti

L

ÃD

¼ L

Ã

þ t

Ã

i

Ã

;

(4)

where t and t* are the marginal penalties/rewards. Policy authorities minimize these newloss functions

in a non-cooperative manner, and the ‘delegated’ strategies i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ, contingent upon the

contracts chosen by the governments previously, solve the ﬁrst order conditions:

L

i

ði; i

Ã

Þ þ t ¼ 0 and L

Ã

i

Ã

ði; i

Ã

Þ þ t

Ã

¼ 0: (5)

8

Other solutions include repeated game mechanisms sustaining the equilibrium in (9) or (2.2.1) as a unique subgame perfect

equilibrium in the repeated version of the game described above (Canzoneri and Gray, 1985; Ghosh and Masson, 1994).

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 396

Under usual concavity assumptions on L and L* this linear systemof equations has a unique solution

if and only if the condition (2) is satisﬁed, as in the Nash equilibrium case. Under the same condition,

the implicit function theorem can be applied to the system (5) in order to study the sensitivity of the

equilibrium in terms of policy authorities’ strategies to changes in the contracts (which are treated as

parameters in the policy authorities’ problem). These can be found as:

_

vi

vt

vi

vt

Ã

vi

Ã

vt

vi

Ã

vt

Ã

_

¼ À

_

L

ii

L

ii

Ã

L

Ã

i

Ã

i

L

Ã

i

Ã

i

Ã

_

À1

¼ D

À1

_

ÀL

Ã

i

Ã

i

Ã

L

ii

Ã

L

Ã

i

Ã

i

ÀL

ii

_

; (6)

where D has been deﬁned in (2) above.

2.1.4. Optimal versus subgame perfect, credible contracts

The optimal contracts that implement the desired cooperative equilibrium in the non-cooperative

game can be easily found by comparing (5) with (3) as:

t

O

¼ L

Ã

i

_

i

C

; i

ÃC

_

; t

ÃO

¼ L

i

Ã

_

i

C

; i

ÃC

_

; (7)

where the strategies are evaluated at their ‘cooperative equilibrium’ values found in (3). This is the

result in Persson and Tabellini (1995, 1996). The argument of this paper is that these contracts are not

a non-cooperative way of achieving cooperation through delegation. Intuitively, there is nothing to

ensure that the precise marginal penalties implementing the optimum will in fact be chosen at the

delegation stage: these contracts are not incentive-compatible.

We solve for the subgame perfect, credible contracts that governments will choose by backward

induction based on their individual incentives, taking into account the reaction functions of the policy

authorities i

D

,i*

D

. These contracts are a solution to min

t

Lði

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

ÞÞ and min

t

Ã

L

Ã

ði

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

ÞÞ

respectively, where i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ is the solution found in (5). First order conditions of this problem

constitute a system of linear equations in t and t*:

L

i

_

i

D

; i

ÃD

_

vi

D

vt

þ L

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

¼ 0

L

Ã

i

_

i

D

; i

ÃD

_

vi

D

vt

Ã

þ L

Ã

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

Ã

¼ 0

(8)

Using (5) and (6) into (8), subgame perfect contracts t

P

; t

ÃP

are a solution to (the ﬁxed point of):

t ¼ À

L

Ã

i

Ã

i

L

Ã

i

Ã

i

Ã

L

i

Ã

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

_

t

Ã

¼ À

L

ii

Ã

L

ii

L

Ã

i

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

_

:

(9)

We are now in a position to state our main results.

Proposition 1. Delegation always occurs in equilibrium(subgame perfect contracts t

P

; t

ÃP

are non-zero) if

and only if there is strategic complementarity/substitutability.

Proof. The proof is by contradiction; suppose that t

P

¼ t*

P

¼ 0, which implies

i

D

ð0; 0Þ ¼ i

N

; i

ÃD

ð0; 0Þ ¼ i

N

and further from (9) that

L

Ã

i

Ã

i

L

Ã

i

Ã

i

Ã

L

i

Ã

_

i

N

; i

ÃN

_

¼ 0;

L

ii

Ã

L

ii

L

Ã

i

_

i

N

; i

ÃN

_

¼ 0:

Since L

i

Ã ði

N

; i

ÃN

Þ and L

Ã

i

ði

N

; i

ÃN

Þ are non-zero (otherwise the Nash equilibrium would be efﬁcient,

hence no need for delegation), this can hold if and only if L

Ã

i

Ã

i

¼ L

ii

Ã ¼ 0; i.e. if and only if there is no

strategic complementarity or substitutability.

Intuitively, incentives for delegation occur because each government/principal recognizes it can

inﬂuence the other player’s strategy by delegating. Whether the equilibrium contract requires

a penalty or a reward depends on the sign of the cross-derivative (i.e. on whether there is comple-

mentarity or substitutability) and on whether there are positive or negative spillovers.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 397

Proposition 2. Optimal contracts t

O

; t

ÃO

are equivalent to ‘cooperative contracts’ t

C

; t

ÃC

found by solving

min

t;t

Ã

_

L

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

_

þ L

Ã

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

__

:

Proof. See Appendix.

Intuitively, Proposition 2 shows that optimal contracts can only be credible (subgame perfect) if

governments cooperate at the delegation stage; however, the cooperation technology needed is

equivalent to the one needed to implement optimal policies i

C

;i*

C

. If the possibility of binding agree-

ments existed, there would be no need for delegation in the ﬁrst place. In addition, Proposition 2 serves

as an intermediary result for Proposition 3.

Proposition 3. Subgame perfect, credible contracts t

P

; t

ÃP

are always different from cooperative contracts

t

C

; t

ÃC

and hence also from optimal contracts t

O

; t

ÃO

.

Proof. See Appendix.

To summarize, we have introduced subgame perfect, credible contracts in a general linear-quadratic

model with spillovers. We have shown that subgame perfect contracts are always non-zero (delegation

is an equilibrium) whenever there are strategic complementarities or substitutabilities; however, they

are always different from optimal contracts (which are instead equivalent to cooperative contracts,

chosen by minimizing an aggregate, ‘global’ loss function). The position of the equilibrium occurring

under credible contracts i

D

ðt

P

; t

ÃP

Þ; i

ÃD

ðt

P

; t

ÃP

Þ relative to the Nash equilibrium i

N

; i

ÃN

and the cooper-

ative equilibrium i

C

; i

ÃC

depends on two factors, as can be seen by inspection of (9): the sign and

magnitude of spillovers/externalities (ﬁrst derivatives of loss functions) and the sign and magnitude of

complementarity (substitutability) given by the second (cross-)derivatives.

Finally, we relate our results to a general argument regarding delegation in Persson and Tabellini

(1995), which is based on the Folk Theorem for Delegation Games in Fershtman et al. (1991). This

theorem provides conditions under which in a two-principal-two-agent game every Pareto optimal

outcome of the principals’ game can become the unique subgame perfect equilibriumof the delegation

game; this is done if each principal delegate to an agent and both contracts are public information.

However, contracts consistent with the condition of this theorem are of a form that does not seem

easily mapped into real-world policy institutions; for instance in our example (for the home country):

Tði; i

Ã

Þ ¼

_

Y; if ½Lði; i

Ã

Þ; L

Ã

ði; i

Ã

Þ

_

L

C

; L

ÃC

_

Y þ c; c > 0 otherwise

¼

_

i

C

; iff ½Tði; i

Ã

Þ; T

Ã

ði; i

Ã

Þ ðY; YÞ

i

D

; otherwise

(10)

The strategies in (10), together with mirroring strategies for the foreign country, implement the ﬁrst

best once contracts are public information,

9

but require that each penalty be written over both payoffs

directly. Although interesting theoretically, transferring this idea to linear contracts is dangerous. While

this theorem shows that strategies of the form (9) implementing the cooperative optimum in

a decentralized manner do exist, it says nothing about their implementability or their being chosen in

equilibrium (obvious issues related to this are observability of payoffs and inconsistency of sovereign

policymaking with making contracts depend on the other country’s strategy). Moreover, if one is to

think about governments choosing a delegation scheme and facing a decision problem that can be

reduced to choosing a set of parameters of the transfer function T, then choice of the optimal contracts

is by no means insured. In the ‘linear contracts’ example, each government has a choice parameter (t)

and as we have shown, while it will always choose some contract, it will never choose the optimal

contract t

C

. Moreover, it is unclear why, if the government had the ability to commit to and actually

implement the optimal contract, does it need to delegate policy rather than choose directly the optimal

policy rule i

C

.

9

For details see Persson and Tabellini (1995); for the game-theoretical argument see Fershtman et al. (1991).

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 398

2.2. Solution method for model with a credibility problem

We now brieﬂy outline the solution method for a model in which there is a credibility problem. The

main reason for doing so is that it has been recognized (since the work of Rogoff, 1985) that policy

cooperation may be suboptimal when a credibility problem exists. Therefore, optimal contracts are

usually designed to correct for both distortions (one coming fromnon-cooperation, the other one from

the lack of commitment). Suppose as before that in each country a policymaker (the government)

minimizes an aggregate, quadratic loss function LðX; sÞ deﬁned over deviations of some macroeco-

nomic variables stacked in the vector X from some target (socially desirable) levels s (X would include

e.g. inﬂation, the output gap, the exchange rate, etc.). Suppose that in each country the policymaker has

at its disposal one policy instrument (such as the interest rate or growth in a monetary aggregate for

monetary policy) and denote this by i. Additionally, assume the model is stochastic, hence each variable

will be hit by a stochastic shock and let the vector of such shocks be denoted by e. Apart from the

policymaker, in each country there is a private sector forming expectations over some relevant subset

of variables of X and hence ultimately over the policy instruments conditional on some information

available one period in advance (U

À1

): i

e

¼ E½irU

À1

; i

eÃ

¼ E½i

Ã

rU

Ã

À1

: As the two countries are inter-

dependent, we also assume that the instrument in one country inﬂuences at least one of the macro-

economic variables of the other, either directly or indirectly (e.g. through a variable such as the

exchange rate). With these assumptions, the relevant macroeconomic variables can ultimately be

expressed as a linear function of the instruments, expectations and shocks in both countries:

X ¼ X

_

i; i

Ã

; i

e

; i

eÃ

; e; e

Ã

_

; X

Ã

¼ X

Ã

_

i; i

Ã

; i

e

; i

eÃ

; e; e

Ã

_

(11)

Using (11), the loss functions can be expressed as functions of instruments, expectations, shocks and

target levels:

L ¼ L

_

i; i

Ã

; i

e

; i

eÃ

; e; e

Ã

; s

_

; L

Ã

¼ L

Ã

_

i; i

Ã

; i

e

; i

eÃ

; e; e

Ã

; s

Ã

_

(12)

Strategic interactions in this model result from heterogeneity of targets (sss

Ã

) and from different

preferences for the stabilization of shocks, when there are spillovers. Assuming further L, L* are

differentiable, the policies have positive or negative externalities depending on whether

vLð:Þ

vi

Ã

;

vL

Ã

ð:Þ

vi

X0:

In addition, the presence of a private sector forming rational expectations of some variable(s)

combined with a real distortion in the economy gives rise to domestic incentives to deviate from

optimality (deﬁned below) that are not related to cross-country spillovers.

10

Suppose that in choosing

the policies, the policymaker faces the following timing in each period: (i) targets s,s* are revealed; (ii)

expectations are formed, i

e

,i

eA

* are determined; (iii) shocks e,e* are realized; (iv) policy instruments i,i*

are chosen simultaneously; (v) macroeconomic variables X,X* are fully determined.

2.2.1. The cooperative and commitment equilibrium

A policy regime whereby the two policymakers decide before stage (i) to cooperate (i.e. to minimize

a joint loss function) and commit to an optimal rule with respect to the private sector will be Pareto

optimal (see e.g. Persson and Tabellini, 1995, 2000). We label the equilibrium occurring under this

benchmark regime ‘the cooperative and commitment equilibrium ’ and denote it with superscript C

as in the previous section. The policy instruments at this equilibrium obey:

ði; i

Ã

Þ

C

¼ arg min

i;i

Ã

_

E

_

L

_

i; i

Ã

; i

e

; i

Ãe

; :

_

þ L

Ã

_

i; i

Ã

; i

e

; i

Ãe

; :

_¸

s:t:

i

e

¼ E½irU

À1

; i

Ãe

¼ E

_

i

Ã

rU

Ã

À1

¸

_

(13)

The optimal policy rules i

C

(e,e*),i*

C

(e,e*) can be found by solving for

11

the (linear, since loss functions

are quadratic) ﬁrst order conditions, rewritten after eliminating the Lagrange multipliers of the rational

expectations constraints:

10

This is the case in the ‘dynamic inconsistency’ literature.

11

Throughout we assume that certain properties of loss functions, policy sets, etc. are met so that the considered equilibria do

exist and are unique, which is the case in most models considered in the literature.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 399

vE½Lð:ÞþL

Ã

ð:Þ

vi

þ E

_

vE½Lð:ÞþL

Ã

ð:Þ

vi

e

rU

À1

_

¼ 0

vE½Lð:ÞþL

Ã

ð:Þ

vi

Ã

þ E

_

vE½Lð:ÞþL

Ã

ð:Þ

vi

Ãe

rU

Ã

À1

_

¼ 0

(14)

Upon specifying functional forms for loss functions and for the models determining the macro-

economic variables, the policy rules are obtained by taking conditional expectations of the system(14),

hence determining expected variables, and then substituting the latter in the original system (14).

However, this equilibrium is unrealistic: since real-world policymaking is best described in a non-

cooperative setup (i.e. binding agreements of any sort are not possible) then the appropriate

equilibrium concept to use is discretionary Nash equilibrium. This equilibrium will obviously be

inefﬁcient, due to two reasons: ignoring the spillovers of policy to the other countries’ loss function (as

in the previous section) and ignoring externalities on the own-country private sector (being more

speciﬁc about the source of inefﬁciencies would require a parametric example which we postpone to

the next section). We study again delegation as a possible solution to both these inefﬁciencies.

2.2.2. Discretionary Nash equilibrium under delegation

Consider that at stage (0), before stage (i) above, each government delegates the policy to an

independent policy authority by imposing a certain transfer function, or contract TðXÞ; where X would

be a subset of the relevant macroeconomic variables (e.g. only inﬂation for inﬂation contracts). Ulti-

mately, this function can also be deﬁned over instruments and shocks, hence delegation would mean

assigning loss functions of the form (where superscript D stands for ’delegated’):

L

D

ð:; TÞ ¼ Lð:Þ þ Tði; i

Ã

; e; e

Ã

; s; s

Ã

Þ (15)

L

DÃ

ð:; T

Ã

Þ ¼ L

Ã

ð:Þ þ T

Ã

ði; i

Ã

; e; e

Ã

; s; s

Ã

Þ

The independent authorities face these loss functions when choosing their policy instruments

simultaneously at stage (iv), in a non-cooperative and discretionary manner. The corresponding

discretionary Nash equilibrium policy rules under delegation will be given by:

i

D

¼ arg min

i

_

L

_

i; i

Ã

; i

e

; i

Ãe

; :

_

þ Tði; i

Ã

; :Þ

¸

i

D

Ã

¼ arg min

i

Ã

_

L

Ã

_

i; i

Ã

; i

e

; i

Ãe

; :

_

þ T

Ã

ði; i

Ã

; :Þ

¸

(16)

The policy instruments can be solved starting from the ﬁrst order conditions given below, taking

expectations of these to pin down expected variables and substituting these back in the original

system:

v½Lði;i

Ã

;i

e

;i

Ãe

;:ÞþTði;i

Ã

;:Þ

vi

¼ 0

v½Lði;i

Ã

;i

e

;i

Ãe

;:ÞþT

Ã

ði;i

Ã

;:Þ

vi

Ã

¼ 0

(17)

2.2.3. Discretionary Nash equilibrium without delegation

Consider ﬁrst the case without delegation, i.e. Tði; i

Ã

; :Þ ¼ T

Ã

ði; i

Ã

; :Þ ¼ 0; leading to the pure Nash

equilibrium choices, say i

N

; i

ÃN

. The two sources of inefﬁciencies mentioned before are obvious by

comparing the systems (14) with (17) evaluated at T ¼ T* ¼ 0; in the latter, two terms are absent that

come from ignoring externalities on (i) the other policymaker and (ii) the private sector. Solutions to

this inefﬁciency usually considered in the literature consist of governments choosing the functions

Tði; i

Ã

; :Þ; T

Ã

ði; i

Ã

; :Þ such that the solutions to the systems coincide.

2.2.4. Optimal versus credible contracts

It is easily seen by direct comparison of (14) and (17) that optimal contracts (making the equi-

librium under delegation identical with the Pareto optimum) should fulﬁll:

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 400

_

vTði;i

Ã

;:Þ

vi

_

C

¼

vE½L

Ã

ð:Þ

vi

þ E

_

vE½Lð:ÞþL

Ã

ð:Þ

vi

e

jU

À1

_

_

vT

Ã

ði;i

Ã

;:Þ

vi

Ã

_

C

¼

vE½Lð:Þ

vi

Ã

þ E

_

vE½Lð:ÞþL

Ã

ð:Þ

vi

Ãe

¸

¸

U

Ã

À1

_

;

(18)

where loss functions are evaluated at the cooperative and commitment optimum i

C

,i*

C

found in (13).

Specifying functional forms for the transfer functions usually results in a solvable system for the

delegation parameters. To choose the most prominent example, linear inﬂation contracts, suppose T ‘s

are linear functions of the policy instruments and can be expressed as T(i) ¼k þti, where k and t are the

delegation parameters to be chosen, and vT=vi ¼ t. The system (18) fully determines the ’optimal

contracts’ t and t*, which we label by t

C

and t

CA

*. In most cases, these marginal contracts are state-

dependent (i.e. dependent on the realization of the shocks), which makes themhard to implement and

undermines their credibility. Jensen (2000) addresses this problem by showing how the ﬁrst best in

(13) can nevertheless be implemented through state-independent delegation by choosing a quadratic

form for the transfer function T.

Finally, when we model the delegation stage (0) as a separate stage of the game -whereby

governments choose their ‘strategies’ (the parameters determining the transfer functions)- the timing

is: (0) governments delegate policies to independent authorities by imposing the transfer functions T

(.), T*(.); (i)-(v): same as before. The solution method is based on backward induction: policy authorities

choose their policy instruments independently and discretionarily taking delegation as given and

governments choose delegation parameters taking into account the choice of policy instruments made

previously by the delegated authorities. The policy rules governments face at stage (0) are ði

D

; i

DAÃ

Þ

given by (16) and fulﬁll the ﬁrst order conditions (17). For the sake of simplicity and for their wide-

spread use, we restrict the functional form of the transfer functions to linear contracts

TðiÞ ¼ k þ ti; T

Ã

ði

Ã

Þ ¼ k

Ã

þ t

Ã

i

Ã

: The subgame perfect, credible contracts are determined by:

t

P

ð:Þ ¼ arg min

t

_

EL

_

i

D

ðt; t

Ã

; :Þ; iD

Ã

ðt; t

Ã

; :Þ; :

__

t

ÃP

ð:Þ ¼ arg min

t

Ã

_

EL

Ã

_

i

D

ðt; t

Ã

; :Þ; iD

Ã

ðt; t

Ã

; :Þ; :

__

(19)

The other parameters k,k* can be chosen such that the participation constraint of the policy

authority is met. The ﬁrst order conditions that credible contracts fulﬁll

12

are:

vE½Lð:Þ

vi

vi

D

ðt; t

Ã

Þ

vt

þ

vE½Lð:Þ

vi

Ã

vi

DÃ

ðt; t

Ã

Þ

vt

¼ 0

vE½L

Ã

ð:Þ

vi

vi

D

ðt; t

Ã

Þ

vt

Ã

þ

vE½L

Ã

ð:Þ

vi

Ã

vi

DÃ

ðt; t

Ã

Þ

vt

Ã

¼ 0

(20)

The objects in (18) and (19) are different in most situations (i.e. when externalities are present,

which is why one considers delegation in the ﬁrst place). This implies that optimal contracts are not

consistent with individual rationality of the governments, the cooperation problem being not solved

but merely relocated to the delegation stage. To sustain optimal contracts as an equilibrium

phenomenon, cooperation or some form of coordination of governments/principals is unequivocally

necessary. Note again the difference with the Folk Theorem in delegation games: here, by delegating

the principal modiﬁes the reaction functions of the agent in a linear way (or else, if contracts non-

linear) instead of ‘forcing’ the Nash equilibrium to overlap with the desired Pareto optimum. Evenwith

this formof delegation, the Nash equilibriumcould be made identical to the Pareto optimum, but this is

not compatible with individual incentives of governments. This is illustrated in the following example,

where credible and optimal contracts are clearly different in an intuitive way.

12

Note that this can be done more generally for a certain functional form of T as long as it is differentiable; the only

modiﬁcation would be that the number of parameters, and hence the number of ﬁrst order conditions to solve, would increase

(for example, for Jensen quadratic contracts there would be three parameters and three ﬁrst order conditions).

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 401

3. An example: credible vs. optimal inﬂation contracts in international monetary policy

We use a parameterized version of the model in the previous section for an illustrative example. The

model is an adapted version of Persson and Tabellini (1996, 2000) and consists of directly postulated

reduced-form equations. The world consists as before of two countries, each one being specialized in

producing a consumer good, which is an imperfect substitute for the other country’s good. This

generates the main spillover of policy through the real exchange rate. Each country has a monetary

policy instrument, which it uses for short-run stabilization (for instance, because there is some form of

nominal rigidity). The policy is also subject to a credibility problem generated by a real distortion

making the natural rate of output (employment) suboptimally low. The model parameters are

symmetric for simplicity but the shocks hitting the economy are arbitrarily correlated. All variables are

in log-differences, a star denotes a foreign country variable (for brevity just the home country’s model

is presented) and time subscripts have been suppressed:

y ¼ gðp À p

e

Þ À e (21)

p ¼ m (22)

zhs þ p

Ã

À p (23)

z ¼ dðy À y

Ã

Þ (24)

p ¼ p þ bz (25)

Deviations of output growthy fromthe natural rate(normalizedtozero) are deﬁnedin(21) bya usual

expectations-augmented Phillips curve, where inﬂation surprises in producer price inﬂation p matter.

For simplicity, in(22) we suppose the growth rate of money is the same as that of producer inﬂation, and

therefore we abstract from velocity shocks.

13

Real exchange rate appreciation z is deﬁned in (23) as

nominal depreciation s plus the differential of producer inﬂation. (24) relates the relative prices z of the

two goods to their relative demand, hence deﬁning an inverse demand equation, where d > 0 is the

inverse relative demand elasticity of outside goods. A higher supply of foreign goods reduces z (real

appreciation) by inducing a relative excess demand for home goods. Finally, consumer price index

inﬂation p is producer inﬂation plus inﬂation induced by the consumption of foreign goods, where b is

theshareof the latter inthedomestic consumptionbasket. Observe that the onlysource of uncertaintyin

the economy is given by adverse supply shocks (e,e*) with zero mean, different variances and arbitrary

covariance (s

2

e

,s

2

e

Ã ; s

ee

Ã ,0). The private sector forms rational expectations of producer inﬂation (and

hence money growth), as the expectation of the latter over the distribution of shocks, conditional upon

the information set U

À1

of previous realizations of macroeconomic variables and model parameters:

p

e

¼ m

e

¼ E½pjU

À1

¼ E½mjU

À1

(26)

Social welfare in each country is deﬁned over variability of output and inﬂation from some socially

desirable levels. For simplicity, normalize the socially optimal inﬂation to zero and suppose the

desirable output q is greater than the natural rate due to some real distortion (monopolistic compe-

tition, for instance). Then the policymakers’ task is to minimize the expected value of the following

conventional period loss function,

14

using as instruments the money growth rates m:

Lð:Þ ¼

1

2

_

p

2

þ lðy À qÞ

2

_

(27)

We assume that q >0 giving rise to the domestic inﬂation bias described in the previous section. The

timing is as in the previous section: just substitute X with (y,p,p,z), i with m, e with e and s with (0,q).

13

In fact, in contrast to Persson and Tabellini, we abstract from all shocks other than supply shocks for simplicity. Having one

shock that creates incentives to deviate from cooperative policy is sufﬁcient for our point.

14

When the game is repeated over time, the policymakers minimize the expected present discounted value of this loss

function. Since the stage game is always identical, this is equivalent to period-by-period minimization.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 402

3.1. The cooperative optimum, non-cooperative inefﬁciency and optimal contracts

For the sake of brevity, we shall apply directly the solution method described in the last section,

without getting into computational details. The cooperative and commitment equilibrium(as in (13)

and (14)) is attained in this example for the optimal state-contingent policy rules:

m

C

ðe; e

Ã

Þ ¼

b

1 þ bg

e þ

dð1 þ2a À bgÞ

ð1 þ bgÞ

_

ð1 þ2aÞ

2

þbg

_ ðe À e

Ã

Þ;

m

ÃC

ðe; e

Ã

Þ ¼

b

1 þ bg

e

Ã

À

dð1 þ2a À bgÞ

ð1 þ bgÞ

_

ð1 þ2aÞ

2

þbg

_ ðe À e

Ã

Þ;

(28)

where we used the change of notation b ¼ lg; a ¼ bdg and d ¼ bd: At the optimum, the policymaker

stabilizes domestic supply shocks (due to their inﬂuence on output and inﬂation). She also stabilizes

relative shocks e À e

Ã

due to their indirect impact on welfare through real exchange rate appreciation/

depreciation.

15

The responses are optimal due to the cooperative features of the equilibrium. Note also

the absence of the inﬂation bias due to commitment (expected policies are zero). Each policymaker

internalizes the effects of its instruments on both the other country’s welfare and its domestic private

sector.

The non-cooperative and discretionary equilibrium is also solved by the corresponding method

described in the previous section ((17)). Suppose delegation to an independent central bank has taken

place before the stage game is played, at stage (0). For the moment we assume this takes the form of

linear inﬂation contracts of the type considered by Persson and Tabellini

16

: each government imposes

a transfer function on its central bank of the form T ¼ k þ tp; T

Ã

¼ k

Ã

þ t

Ã

p

Ã

: The marginal penalties t

and t* are allowed to be state-contingent. Given the linear nature of the model and linearity in

stochastic shocks we model this by assuming that each marginal penalty is additively separable in

a state-independent and a state-dependent component, namely:

t ¼ t þ

~

tðe; e

Ã

Þ where E½t ¼ t and E

_

~

t

_

¼ 0

t

Ã

¼ t

Ã

þ

¯

t

Ã

ðe; e

Ã

Þ where E½t

Ã

¼ t

Ã

and E

_

¯

t

Ã

_

¼ 0

(29)

Given these linear penalties, each central bank will minimize its loss function (27) modiﬁed as in the

system(15), e.g. Lð:Þ ¼

1

2

fp

2

þ lðy À qÞ

2

g þ Tð:Þ. The discretionary Nash equilibriumpolicy instruments

given delegation are found as in the system (16) in the previous section as (where the same change of

notation as before was used and additionally A ¼ ð1 þ aÞ

2

þ bg; B ¼ að1 þ aÞ):

m

D

ðq; e; e

Ã

; t; t

Ã

Þ ¼ Àt þ

b

1 þ a

q À

ð1 þ aÞA

A

2

À B

2

~

t À

ð1 þ aÞB

A

2

À B

2

¯

t

Ã

þ

b

A À B

e þ

d

_

1 þ a

2

_

A

2

À B

2

ðe À e

Ã

Þ

m

ÃD

ðq; e; e

Ã

; t; t

Ã

Þ ¼ Àt

Ã

þ

b

1 þ a

q À

ð1 þ aÞB

A

2

À B

2

~

t À

ð1 þ aÞA

A

2

À B

2

¯

t

Ã

þ

b

A À B

e

Ã

À

d

_

1 þ a

2

_

A

2

À B

2

ðe À e

Ã

Þ

(30)

The purely non-cooperative discretionary equilibrium without delegation ðt ¼

~

t ¼ t

Ã

¼

¯

t

Ã

¼ 0Þ

features two inefﬁciencies, as expected. First, a familiar inﬂation bias (the q term) is present in each

country (and is the same in both due to the assumption on homogeneity of targets) due to discretion.

Secondly, the responses to both domestic supply shocks and relative shocks are different from the

15

See e.g. Bilbiie (2000) or Persson and Tabellini (1995, 1996, 2000) for details and the solution of slightly more general

models.

16

In Bilbiie (2000) we show equivalence of linear inﬂation contracts and inﬂation targets in this framework. We shall focus

only on contracts for the sake of exposition but the results apply equally to delegating with a non-zero inﬂation target.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 403

optimal ones due to not internalizing of policy externalities when acting non-cooperatively. The exact

nature of the distortions will depend on the shocks and the values of parameters but as a general rule

the policies would have a contractionary bias when a favorable shock hits (positive externalities) and

would be too expansionary, at the other country’s cost, when an adverse shock is realized. Following for

instance Persson and Tabellini (1996) we shall call this a stabilization bias.

17

Following the analysis in the general case we can easily ﬁnd the marginal penalties that implement

the cooperative and commitment optimum when policy m; m

Ã

is chosen in a non-cooperative and

discretionary manner. Mutatis mutandis, system (18) in this case translates to:

t

C

ðq; e; e

Ã

Þ ¼

1

1þa

_

bq À ap

ÃC

ðe; e

Ã

Þ

_

¼

b

1 þ a

q À

ab

ð1 þ aÞð1 þ bgÞ

e

Ã

À

2a

2

b

ð1 þ bgÞ

_

ð1 þ2aÞ

2

þbg

_ ðe À e

Ã

Þ

t

C

ðq; e; e

Ã

Þ ¼

1

1þa

_

bq À ap

C

ðe; e

Ã

Þ

_

¼

b

1 þ a

q À

ab

ð1 þ aÞð1 þ bgÞ

e þ

2a

2

b

ð1 þ bgÞ

_

ð1 þ2aÞ

2

þbg

_ ðe À e

Ã

Þ

(31)

The marginal penalties are intuitive. The ﬁrst terms are the familiar ones correcting for the domestic

inﬂation bias in each country. The other terms correct for suboptimal stabilization of shocks. The

penalty is weaker if the foreign country suffers an adverse supply shock (e

Ã

> 0) or a less severe supply

shock as compared to the home country (e À e

Ã

). In these two cases foreign inﬂation is positive and the

real exchange rate appreciates at home. In this case the home policy is too contractionary and a reward

(or lower penalty, depending on the q term) for additional inﬂation is needed to correct for that.

3.1.1. State-independent optimal delegation

While the marginal penalties’ being state-contingent is intuitive, the design of ‘state-dependent

institutions’ in practice is not (for instance, they would require changes in institutions for every

realization of shocks).

18

If only state-independent contracts are feasible, then only the domestic

incentives are corrected for, leaving suboptimal shock stabilization unaltered. This problemis solved by

Jensen (2000) by proposing a delegation scheme based on quadratic contracts with targets. Jensen

argues that since there are three distortions that optimal delegation should correct for, corresponding

to the three terms in (31) (the average inﬂation bias, the suboptimal response to symmetric and

asymmetric shocks, respectively), the central bank’s incentives should be corrected in all three

dimensions. To achieve this, the particular form of delegation should feature three parameters, one for

each distortion. In our model, this implies that the contract should be, as in Jensen, a quadratic contract

with targets of the form: Tð:Þ ¼

1

2

faðp À p

B

Þ

2

þ mðp À p

Ã

Þ

2

g , where a, p

B

; m are decision variables of the

government when delegating. Following the same solution method as above it is readily shown that

the optimum is implemented for:

a ¼ À

a

1 þ2a

; p

B

¼

b

a

q; m ¼

a

1 þ a

:

Beyond familiar terms purported to correct for the average inﬂation bias (p

B

) and the suboptimal

response to symmetric shocks (by requiring a relatively more liberal central bank, a < 0), it is worth

noting that delegation works by penalizing the cross-country inﬂation differential between home and

foreign (via m < 0) in order to correct for the suboptimal response to asymmetric shocks featured in the

non-cooperative equilibrium. Note that, compared the state-contingent contracts that make the

marginal penalties depend directly upon the realization of home and foreign shocks, state-independent

delegation makes the loss function depend upon realization of the other country’s macroeconomic

17

More details on interpretation of incentives in this equilibrium can again be found in Bilbiie (2000), Jensen (2000) or

Persson and Tabellini (1996).

18

For an elaborated critique of state-dependent delegation see Jensen (2000).

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 404

outcomes (in this case inﬂation). But the logic of the two mechanisms is consistent; indeed, rewriting

the state-contingent contracts derived in (31) for the home country:

t

C

ðq; e; e

Ã

Þ ¼

b

1 þ a

q À

ab

ð1 þ aÞð1 þ bgÞ

e þ

ab þ ab

2

g þ2a

2

bð1 þ aÞ

ð1 þ aÞð1 þ bgÞ

_

ð1 þ2aÞ

2

þbg

_ðe À e

Ã

Þ;

one can notice that domestic inﬂation is rewarded when there is an adverse symmetric supply shock

and penalized when there is an asymmetric shock, consistently with the intuition in the state-inde-

pendent case. The advantage of state-independent delegation, however, is that institutions need not

change every time a shock hits the economy; we refer the interested reader to Jensen (2000) for a more

detailed discussion of state-independent delegation.

3.2. Credible contracts: the linear inﬂation contracts case

Although linear contracts à la Persson and Tabellini implement the optimum (albeit with state-

contingent parameters), and the quadratic contracts with targets à la Jensen also solve the problem of

state-contingency, they are both subject to the problemwe identiﬁed in the previous section. We shall

now see an example of this at work.

19

To see what contracts government will choose (and hence implement) based only on their indi-

vidual rationality and their perception of rationality of the agents to which they delegate (central

banks) we follow the solution method outlined in the general case. By backward induction, at the

delegation stage (0), governments face the policy rules contingent on contracts that we solved for

previously in (30). They then minimize the expected values of the social losses given by (27) (and its

foreign counterpart), where inﬂation and the output gap are evaluated at the delegated Nash Equi-

librium. We treat the state-independent part of the contract as control variables of the government.

Hence, e.g. the ‘home’ government will only choose t , and for ﬁnding the equilibrium state-contingent

part of the contract

~

tðe; e

Ã

Þ we use (29). Substituting m

D

ð:; t; t

Ã

Þ; m

Ã

Dð:; t; t

Ã

Þ found in (30) into

E½Lð:Þ; E½L

Ã

ð:Þ and minimizing the latter two with respect to t and t

Ã

respectively yields the two ﬁrst

order conditions:

p

D

_

t;

~

t; t

Ã

;

¯

t

Ã

; q; e; e

Ã

_

¼ 0 p

ÃD

_

t;

~

t; t

Ã

;

¯

t

Ã

; q; e; e

Ã

_

¼ 0 (32)

Substituting the delegated Nash equilibriummoney growth rates from(30), we get two equations in

four unknowns t; t

Ã

;

~

tðe; e

Ã

Þ;

¯

t

Ã

ðe; e

Ã

Þ: Using state independence of the ﬁrst two and zero mean of the last

two one gets a solution for credible contracts as (where a and b are deﬁned as before):

t

P

ðe; e

Ã

Þ ¼

b

1 þ a

q þ

b

1 þ a

e

Ã

þ

b

1 þ2a

ðe Àe

Ã

Þ

t

ÃP

ðe; e

Ã

Þ ¼

b

1 þ a

q þ

b

1 þ a

e À

b

1 þ2a

ðe À e

Ã

Þ

(33)

We are now ready to compare these non-cooperative credible contracts with the optimal contracts

implementing the ﬁrst best, focusing on the home country. A ﬁrst thing to note is that, may be

surprisingly so, the state-independent termleading to elimination of the systematic inﬂation bias is the

same in t

C

ðe; e

Ã

Þ and t

P

ðe; e

Ã

Þ: Another way to read this is that if only state-independent delegation were

possible, the two contracts would coincide, although they would then be both suboptimal in that they

would not affect stabilization of shocks.

The inefﬁciency of credible contracts comes fromsuboptimal responses to shocks (second and third

term). Consider again the case where the foreign country is hit by an adverse supply shock e

Ã

> 0

and this is less severe than in the home country (or equivalently, there is a larger favorable shock), i.e.

e À e

Ã

> 0: In equilibrium p

Ã

is greater than zero and there is a contractionary bias of the home

19

Jensen recognizes this problem himself in the last paragraph of the mentioned paper ‘[.] incentives causing policymakers

to deviate from cooperative policies would also cause governments to deviate from cooperative institutions’, but he focuses on

identiﬁcation of optimal institutions and not their implementability.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 405

country’s monetary policy. In the optimal contract t

C,

both coefﬁcients on shock stabilization are

negative: the optimal penalty in the home country decreases to correct for the deﬂationary bias. On the

contrary, in the credible contract t

P

both coefﬁcients are positive: the penalty is increasing in e and e À

e

Ã

; aggravating the contractionary bias of home policy. Equivalently, note from (32) that the credible

contract is always imposed so that producer inﬂation is zero, whereas in the considered case the

optimal response would be a positive producer inﬂation. To achieve the zero inﬂation in the perfect

equilibrium an increasing marginal penalty is needed. In the converse case, where the foreign country

faces a favorable supply shock e

Ã

< 0 relatively smaller than in the home country (e À e

Ã

< 0Þ there is an

expansionary bias of home monetary policy. This negative spillover would be eliminated through

optimal delegation: the penalty becomes larger (see (31)) to reduce inﬂationary incentives. As the

spillovers are ignored in the perfect equilibrium, the penalty will be smaller, tailored to achieving

a zero inﬂation consistent with the ﬁrst order condition (32). But the optimal response should in fact

target deﬂation.

The two contracts are different even when shocks are perfectly correlated, in the symmetric case

whereby e ¼ e

Ã

: The optimal marginal penalty is t

C

¼

b

1þa

q À

ab

ð1þaÞð1þbgÞ

e , whereas the credible one is

t

P

¼

b

1þa

q þ

b

1þa

e. An adverse common supply shock generating a contractionary bias is optimally

corrected by a decrease in the penalty for additional inﬂation. Not recognizing the positive externality

that would result fromboth countries inﬂating more, at the delegation stage governments increase the

penalty, therefore aggravating the deﬂationary bias.

The different contracts can be directly compared by substituting them in the best response func-

tions (30).

20

The linear optimal contracts t

C

; t

ÃC

implement the ﬁrst best optimal money growth rates

m

C

; m

ÃC

; and so do the Jensen quadratic contracts and the contracts consistent with the Folk Theorem

in delegation games presented in (10). Without delegation, the Nash equilibrium policies would

feature an inﬂation bias and suboptimal shock stabilization. Delegation with credible contracts would

mean elimination of the inﬂation bias but still suboptimal shock stabilization, so they will not lead to

implementation of m

C

; m

ÃC

: By substituting t

P

; t

ÃP

in the best response functions one gets

m

D

ð:; t

P

; t

ÃP

Þ ¼

d

1þ2a

ðe À e

Ã

Þ; m

ÃD

ð:; t

P

; t

ÃP

Þ ¼ À

d

1þ2a

ðe À e

Ã

Þ: Whether these will be higher or lower

than m

C

; m

ÃC

depends on the nature of the shocks and the parameters.

Hence, two governments seeking to delegate policy in a manner consistent with their individual

rationality (or with their mandate, that is maximizing social welfare) would fail to achieve a ﬁrst best

equilibrium. In order to delegate with the scheme that would insure that optimal policies are followed

they would need to cooperate at the delegation stage. Alternatively, a supranational institution able to

coordinate the two governments on the ‘right’ institutions would do the same job. However, this is far

fromthe non-cooperative setup one wishes to describe in the ﬁrst place. If compromises are to be made

in terms of allowing for the possibility of binding agreements at the level of governments, it is hard to

understand why then wouldn’t governments cooperate directly without any need for delegating

policies.

4. Conclusions

Various inefﬁciencies associated with policymaking, whether at a domestic or international level,

can allegedly be solved by delegation of policy to independent monetary authorities. In a prominent

example, monetary policy, delegation schemes have been viewed as panacea for both domestic

credibility problems and inefﬁciencies coming fromcross-country spillovers. Givenpolicy externalities,

a policy regime where governments cooperate (and commit with respect to the private sectors) is

unequivocally Pareto optimal, but there are strong incentives to deviate from it. Some simple and

intuitive delegation schemes easily mappable into real-life institutions have been found to ‘ﬁx’ both

these incentives: i.a. the linear inﬂation contracts proposed by Persson and Tabellini (1995, 1996,

2000), quadratic contracts with targets of Jensen (2000), or targeting rules in a new open-economy

model as in Benigno and Benigno (2005). In these cases, each government delegates to an independent

policy authority by imposing to the latter a certain transfer function. The governments can then in

20

In Bilbiie (2000) we provide a more detailed welfare comparison of equilibria.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 406

theory design the contract to ensure that the delegated authorities choose the policy instruments that

implement the desired equilibrium.

The argument of this paper is that this implementation mechanism hides an implicit assumption

about governments being actually able to sign binding agreements in order to coordinate on exactly

those delegation parameters that ‘do the job’. But if this were the case, it is hard to see why

governments need to delegate instead of committing themselves to the optimal policy rules. The way

out from this dilemma is, in our view, an explicit modeling of the delegation decision of the

governments. We do this by supposing that each government chooses the (subgame perfect, cred-

ible) contracts based on its individual rationality, taking into account the agents’ choices at a future

stage.

21

First, we provide a general solution method and deﬁne the newequilibriumconcept in a non-

parametric model of policymaking with spillovers. We ﬁrst show that delegation always occurs in

equilibriumwhen there is strategic complementarity or substitutability. However, we then showthat

optimal contracts occur in equilibrium only under cooperation at the delegation stage (the same sort

of cooperation needed to implement optimal policies): optimal contracts are equivalent to ‘coop-

erative contracts’ found by minimizing the global loss function. We then arrive at our main point:

subgame perfect, credible contracts are always different from cooperative contracts, and hence also

from the optimal ones.

22

We extend the analysis to a model featuring a domestic credibility problem

and present an example from international monetary policy cooperation, where it turns out that the

contracts governments would actually choose are different in an intuitive way from the optimal ones

for any correlation of shocks.

23

Our analysis raises a normative question: what could then insure implementation of the ‘right’

institutions, preserving the non-cooperative assumption about policymaking? One possible solution

consists of creating of a supranational institution that is able to ‘coordinate’ governments on the

optimal contracts at the delegation stage. An alternative would be strengthening the role of some

existing supranational institutions (such as the International Monetary Fund), which reinforces

arguments made by Canzoneri and Henderson (1991) in a different setup. There, such an institution

helped governments choose, i.e. coordinate on, a best equilibrium among a multiplicity of feasible

equilibria. The equilibrium (and hence coordination by the international principal), however, is in

terms of policies directly, which seems hard to map into real-life practice. In our context, the supra-

national institution would help design the appropriate incentive schemes of countries’ policy

authorities and monitor their implementation over time. While this might seem akin to centralization

or cooperation on policies directly (which would be a solution by assumption), we think it is indeed

more realistic to assume that national governments agree to coordinate on some institutional features

than to systematically pursue cooperative policies that are not consistent with their incentives. By the

presence of such an international institution, sovereignty of policymaking is preserved. However,

further institutional challenges would be raised by this setup, for such a supranational institution

would simultaneously be a common principal to and a common agency of the sovereign states; this

raises again the issue of optimal design of the delegation scheme of the supranational organization

itself, when viewed in its role of common agency. Further research, perhaps along the lines of the

‘common agency’ problem analyzed in Dixit and Jensen (2003), is needed in order to analyze the

incentives of such supranational institutions, their design and the mechanisms by which they could

implement and monitor globally optimal policy regimes.

21

It is important to note that once we consider transfer functions we are no longer in the conditions of the Folk Theorem in

Delegation Games of Fershtman et al., which dealt with ‘take-it-or-leave-it’ target compensation functions by which the

equilibrium in the agents’ game can be made identical to a Pareto optimum. Once we make the transfer functions linear (e.g.),

the equilibrium in the agents’ game will depend on the delegation in no simple way as this will change their reaction functions.

In order to pin down the equilibrium one has to pin down some values for the delegation parameters.

22

Note that our results are different from McCallum’s (1995) critique concerning closed-economy monetary institutions or

more speciﬁcally enforceability of inﬂation contracts. Rather than studying enforceability (and sustainability over time) of

optimal contracts, we solve for incentive-compatible, credible contracts in an open-economy framework.

23

The two are identical only if shocks are absent and/or only state-independent institutions are feasible, but the latter case

features again inefﬁciency.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 407

Acknowledgments

A previous, different version of this paper has been circulated under the title ‘Perfect versus Optimal

Contracts: an Implementability-Efﬁciency Trade-off’. I thank without implicating the Oesterreichische

Nationalbank for conferring me the Olga Radzyner Award for an earlier version of this paper, and Banque

de France for ﬁnancial support through the Chaire Banque de France at the Paris School of Economics.

I am grateful to an anonymous referee and Ben Lockwood in particular and to Mike Artis, Gianluca

Benigno, Henrik Jensen, Marcus Miller, Paul Mizen, Roberto Perotti, Guido Tabellini and participants at

the Bank of England Seminar and EUI Monetary Economics Workshop and Informal Theory Group for

comments on earlier versions. All errors are mine.

Appendix: A Proofs of Propositions 2 and 3.

A.1. Proof of Proposition 2

The ﬁrst order conditions for cooperative contracts are:

L

i

_

i

D

; i

ÃD

_

vi

D

vt

þ L

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

þ L

Ã

i

_

i

D

; i

ÃD

_

vi

D

vt

þ L

Ã

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

À0

L

i

_

i

D

; i

ÃD

_

vi

D

vt

Ã

þ L

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

Ã

þ L

Ã

i

_

i

D

; i

ÃD

_

vi

D

vt

Ã

þ L

Ã

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

Ã

(34)

From (5), we know L

i

ði

D

; i

ÃD

Þ ¼ Àt and L

Ã

i

Ã

ði

D

; i

ÃD

Þ ¼ Àt

Ã

; which substituted in the above yield:

_

L

Ã

i

_

i

D

; i

ÃD

_

À t

_

vi

D

vt

þ

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_

vi

ÃD

vt

¼ 0

_

L

Ã

i

_

i

D

; i

ÃD

_

À t

_

vi

D

vt

Ã

þ

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_

vi

ÃD

vt

Ã

¼ 0

(35)

The second equation in (35) implies:

_

L

Ã

i

_

i

D

; i

ÃD

_

À t

_

¼ À

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_ vi

ÃD

vt

Ã

vi

D

vt

Ã

;

which substituted in the ﬁrst equation in (35) delivers:

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_

¼

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_ vi

ÃD

vt

Ã

vi

D

vt

Ã

vi

D

vt

vi

ÃD

vt

¼

_

L

i

Ã

_

i

D

; i

ÃD

_

À t

Ã

_

L

ii

L

Ã

i

Ã

i

Ã

L

ii

Ã L

Ã

i

Ã

i

; (36)

where the second equality uses the result in (6). As long as the necessary and sufﬁcient condition for

existence and uniqueness of equilibrium (2) is satisﬁed (L

ii

L

Ã

i

Ã

i

Ã

sL

ii

Ã L

Ã

i

Ã

i

), (36) can be satisﬁed if and only

if L

i

Ã ði

D

; i

ÃD

Þ À t

Ã

¼ 0; which implies that cooperative contracts t

C

; t

ÃC

obey:

t ¼ L

Ã

i

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

_

t

Ã

¼ L

i

Ã

_

i

D

ðt; t

Ã

Þ; i

ÃD

ðt; t

Ã

Þ

_

Under our assumptions on L

i

(notably, condition (2) holds) this is a linear system with a unique

solution, t

C

; t

ÃC

. On the other hand, we know from (7) above that t

O

; t

ÃO

; is a solution, so we conclude

that cooperative contracts and optimal contracts coincide t

C

¼ t

O

; t

ÃC

¼ t

ÃO

.

A.2. Proof of Proposition 3

The proof is by contradiction. Suppose that t

P

; t

ÃP

and t

C

; t

ÃC

are identical: comparing the ﬁrst order

conditions in each case, (34) with (8), this can happen if and only if:

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 408

L

Ã

i

_

i

D

; i

ÃD

_

vi

D

vt

þ L

Ã

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

¼ 0

L

i

_

i

D

; i

ÃD

_

vi

D

vt

Ã

þ L

i

Ã

_

i

D

; i

ÃD

_

vi

ÃD

vt

Ã

¼ 0

(37)

We substitute the partial derivatives of i

D

and i

ÃD

using the result in (6) to obtain:

ÀL

Ã

i

Ã

i

Ã

L

Ã

i

_

i

D

; i

ÃD

_

þ L

Ã

i

Ã

i

L

Ã

i

Ã

_

i

D

; i

ÃD

_

¼ 0

L

ii

Ã L

i

_

i

D

; i

ÃD

_

À L

ii

L

i

Ã

_

i

D

; i

ÃD

_

¼ 0

(38)

From (5) we know that L

i

ði

D

; i

ÃD

Þ ¼ Àt and L

Ã

i

Ã

ði

D

; i

ÃD

Þ ¼ Àt

Ã

, which substituted above give:

t ¼ À

L

ii

L

ii

Ã

L

i

Ã

_

i

D

; i

ÃD

_

; t

Ã

¼ À

L

Ã

i

Ã

i

Ã

L

Ã

i

Ã

i

L

Ã

i

_

i

D

; i

ÃD

_

However, we knowthat perfect contracts have to satisfy also (9). Equating the expressions for either

t or t

Ã

we obtain (using that there are externalities L

i

Ã ði

D

; i

ÃD

Þs0; L

i

Ã ði

D

; i

ÃD

Þs0 by assumption):

L

ii

L

ii

Ã

¼

L

Ã

i

Ã

i

L

Ã

i

Ã

i

Ã

;

which is a contradiction since violates the condition for existence and uniqueness of equilibrium (2).

Therefore, subgame perfect contracts are always different from cooperative contracts, and hence also

from optimal contracts (using Proposition 2).

References

Benigno, P., 2002. A simple approach to international monetary policy coordination. Journal of International Economics 57 (1),

177–196.

Benigno, G., Benigno, P., 2005. Designing targeting rules for international monetary policy cooperation. Journal of Monetary

Economics 53 (3), 473–506.

Bilbiie, F.O., 2000. Inﬂation Contracts, Targets and Strategic Incentives for Delegation in International Monetary Policy Games.

European University Institute. Working Paper ECO 2001/16.

Canzoneri, M., Gray, J.A., 1985. Monetary policy games and the consequences of non-cooperative behaviour. International

Economic Review 26, 547–564.

Canzoneri, M., Henderson, D., 1991. Monetary Policy in Interdependent Economies: A Game-theoretic Approach. MIT Press,

Cambridge MA.

Cooper, R., John, A., 1988. coordinatingcoordinationfailures inKeynesianmodels. The Quarterly Journal of Economics 103(3), 441–

464.

Corsetti, G., Pesenti, P., 2001. Welfare and macroeconomic interdependence. The Quarterly Journal of Economics 116 (2), 421–445.

Dixit, A., Jensen, H., 2003. Common agency with rational expectations: theory and application to a monetary union. Economic

Journal 113, 539–549.

Fershtman, C., Judd, K., Kalai, E., 1991. Observable contracts: strategic delegation and cooperation. International Economic

Review 32, 551–559.

Ghosh, A., Masson, P., 1994. Economic Cooperation in an Uncertain World. Blackwells, Oxford, UK.

Hamada, K., 1976. A strategic analysis of monetary interdependence. Journal of Political Economy 84, 677–700.

Jensen, H., 2000. Optimal monetary policy cooperation through state-independent contracts with targets. European Economic

Review 44, 517–539.

Kohler, M., 2002. Coalition formation in international monetary policy games. Journal of International Economics 56 (2), 371–385.

McCallum, B., 1995. Two fallacies concerning central-bank independence. American Economic Review, Papers and Proceedings

85, 207–211.

Obstfeld, M., Rogoff, K., 1996. Foundations of International Macroeconomics. MIT Press, Cambridge, MA.

Persson, T., Tabellini, G., 1995. Double-edged incentives: institutions and policy coordination. In: Grossman, G., Rogoff, K. (Eds.),

Handbook of International Economics, vol. III (North-Holland).

Persson, T., Tabellini, G., 1996. Monetary cohabitation in Europe. American Economic Review, Papers and Proceedings, 111–117.

Persson, T., Tabellini, G., 2000. Political Economics: Explaining Economic Policy. MIT Press, Cambridge, MA.

Rogoff, K., 1985. International monetary policy coordination may be counterproductive. Journal of International Economics 18,

199–217.

Woodford, M., 2003. Interest and Prices. Princeton University Press, Princeton, NJ.

F.O. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 409

In the international policy context. the authors move to analyzing linear contracts. loss functions. which is a non-desirable feature as it makes them difﬁcult to implement (for example 1 We adopt the game-theoretical deﬁnition of cooperation as joint optimization by a group of players of their payoffs. as long as these contracts become common knowledge. which would instead require cooperation of governments (or some form or coordination) at the delegation stage. Notably. one wonders why would delegation be needed in the ﬁrst place. Exchange of information is captured by the non-cooperative policymaking case. they implicitly assume cooperation (or some form of coordination) at the delegation stage. However. they are always different from the optimal contracts. 1996). Given individual incentives to deviate from the optimal cooperative policies.394 F. externalities take the form of conﬂicts over shock stabilization or over preferred levels of macroeconomic outcomes.e. 1976). A recurrent result is that the cooperative optimum can be achieved in a decentralized. could in principle act as a solution to correct inefﬁciencies coming from both discretion and non-cooperative policymaking.a. See also Kohler (2002). Coordination would by contrast mean choosing one particular equilibrium in the Nash Equilibrium set of the non-cooperative game (this might imply the presence of an external enforcing mechanism). 2000). Speciﬁcally. where governments choose the appropriate delegation scheme for their central banks that implements the optimum. We focus on the ‘institutional design’ approach pioneered by Persson and Tabellini (1995) and extended by Jensen (2000) and Persson and Tabellini (1996.all these issues are extensively reviewed in Canzoneri and Henderson (1991) or Ghosh and Masson (1994). which is hard to reconcile with the alleged ‘purely non-cooperative implementation of the cooperative optimum’.) . . 3 Persson and Tabellini also look at non-linear discontinuous performance contracts with state-dependent parameters written directly over welfare functions that can implement the cooperative optimum. The Pareto optimum is not enforceable for various reasons (individual incentives to deviate.2 But institutional design. This is done by assuming that before the actual policy game takes place. where it is argued that in a two-player game the principals can obtain every Pareto optimal outcome as the unique subgame perfect Nash Equilibrium of the delegation game via such contracts written on target compensation form. Typically. not credible: indeed. non-cooperative manner by delegating through optimal inﬂation contracts (i. Persson and Tabellini (1995. uncertainty regarding models.O. But then. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 inherent inefﬁciencies of non-cooperative (and discretionary) policymaking. This paper starts from the observation that these delegation schemes are not subgame perfect. We develop this argument analytically by explicitly modeling the institutional design stage and studying the (credible. etc. The state of the art in the literature on optimal monetary policy delegation in an international context can be summarized as follows. This is an application of a Folk Theorem in Delegation Games by Fershtman et al. However. Persson and Tabellini. at the delegation stage governments choose the delegation parameters in a non-cooperative manner by backward induction. (1991). there is an ‘institutional design stage’. The inefﬁciencies that optimal delegation is supposed to ‘ﬁx’ in this case are problems due to non-cooperative policymaking in the presence of policy spillovers in a multi-country world (and/or ‘credibility’ problems like the inﬂation bias). arguing that these are highly non-realistic and difﬁcult to implement. 1996) analyze performance contracts written by the governments before the game is played. or delegation to an independent monetary authority. These credible. it has been long recognized that cooperative policymaking1 is Pareto optimal when sovereign policymaking has externalities on the other countries (see e. Hamada. if binding agreements were possible. suboptimality of cooperation when commitment with respect to the domestic private sector is impossible.g.3 The optimal linear contracts hence found are statecontingent. much of the work concerning monetary policy institutions adopts this line of reasoning. at an ‘institutional design’ stage and show how these contracts can be designed such that the inefﬁciencies related to both discretionary and non-cooperative policymaking are eliminated. the literature has moved towards identifying mechanisms that sustain the collusive outcome. subgame perfect contracts turn out to be non-zero (hence delegation is always an equilibrium) whenever there is strategic complementarity or substitutability. This focus is reinforced in the international context by an observation of Rogoff (1985): in the presence of domestic credibility problems as the ones reviewed above cooperation itself might even be welfarereducing. 2 Canzoneri and Henderson (1991) interpret this insight as a particular case of a more general result: coalitions of only subsets of players are inefﬁcient. i. subgame perfect) contracts that are consistent with governments’ incentives and hence occur in equilibrium. implying a ‘pregame’ and the possibility of binding agreements. taking into account the reaction functions of the central banks at the policy stage. 1995.

that are chosen by governments based on their individual rationality. general equilibrium two-country model and show that targeting rules can be designed that implement optimal cooperative policies. non-cooperative contracts) as opposed to optimal contracts in a two-country model with policy spillovers. recent research shows that the insights of optimal design of institutions carries over to more realistic setups in the new open-economy macroeconomics tradition. 2. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 395 because they imply that the institution changes each time a shock occurs). Credible. 2005 for the openeconomy case).i*). it shows that. see Benigno and Benigno. although possible to derive as a quadratic approximation of an aggregate welfare function describing society’s preferences (Woodford. However. The loss functions are therefore L(i. 2000).7 We assume that externalities See also Benigno (2002). and distinguish the foreign country by an asterisk. We provide an example in the next section. 7 We denote a derivative with respect to an argument by appending the corresponding argument as a subscript. 2003. 6 This loss function is usually quadratic and directly postulated. subgame perfect contracts in a general linear-quadratic framework In this section we describe a general solution method for credible contracts (as a shorthand notation for subgame perfect.i*) and L*(i. credible subgame perfect contracts are different from optimal contracts. Section 2 sets the stage and develops a general version of our argument in a simple linear-quadratic two-country model with spillovers/externalities. Benigno and Benigno (2005) use a microfounded.1.F. 2. 2000).e. i. Jensen (2000) addresses this issue by ﬁnding state-independent transfer functions that implement the cooperative outcome. and that optimal contracts exist that could make these targeting rules occur in a non-cooperative equilibrium.2 extends this framework to a setup where there is a domestic credibility (commitment) problem. In each country a policymaker (the government) minimizes a quadratic aggregate loss function6 deﬁned over deviations of some macroeconomic variables.5 We start with a simplest setup in which there is no domestic credibility problem and prove formally three propositions. The ﬁrst states that delegation always occurs in equilibrium when there is strategic complementarity or substitutability (in the sense that one country’s equilibrium strategy is increasing/decreasing in the other country’s strategy). for an earlier effort in this direction. Section 3 applies the general results of Section 2 to a simple model of international monetary policy cooperation due to Persson and Tabellini (1996. 1996). Section 4 concludes and points out some implications for the design of supranational institutions.4 The remainder of the paper proceeds as follows. suppose that there are two countries. LiiÃ is the cross-derivative of L. A general criticism of this line of research is that welfare conclusions and prescriptions cannot be properly addressed in a model that lacks microfoundations (Obstfeld and Rogoff. The second states that optimal contracts – that implement the cooperative optimum in the non-cooperative gamerequire cooperation at the delegation stage. He also provides interpretations of these contracts in terms of real-life institutions. home and foreign. The third shows that these cooperative contracts (optimal contracts) are always different from subgame perfect contracts . These functions penalize quadratically inﬂation deviations from a certain level (chosen by the government) as well as inﬂation differentials between the two countries. are a solution to government’s delegation problem only under cooperation. Section 2. which are related linearly to the policy instruments. A simple but general version of the argument Following most of the literature on policy coordination and institutional design reviewed above.O. However. Persson and Tabellini (1995. and their being quadratic implies that second-order derivatives are constant.a. and introduce the notion of ‘credible’ (subgame perfect) contracts. In a recent insightful contribution. i. using the framework of Corsetti and Pesenti (2001). We derive the optimal contracts that implement the cooperative optimum under non-cooperative (Nash) policymaking. Speciﬁc (parametric) examples of models most related to this one are i. and explains why. 5 4 .e. i and i* respectively. We then describe the solution method for a more general setup in which we allow for domestic credibility problems.

i. iÃ Þ ¼ Lði. viÃN =vi > 0Þ or substitutability (<0) will depend on the sign of the second derivatives LiiÃ and LÃÃ i . contingent upon the contracts chosen by the governments previously. global loss function Lði. Policy authorities minimize these new loss functions in a non-cooperative manner. in the pioneering work of Persson and Tabellini (1995). iÃ Þ ¼ Li ði.i*) space. Ghosh and Masson.2. that the Jacobian of the (1) system be non-singular. (4) where t and t* are the marginal penalties/rewards. the delegated loss functions LD become: LD ¼ L þ ti LÃD ¼ LÃ þ t Ã iÃ . non-cooperative equilibrium consists of strategies i and i* that solve the following system of linear ﬁrst order conditions (they are linear because the loss functions are quadratic): Li ði. iÃ Þ ¼ 0 and LÃÃ ði.396 F. iÃ Þ with respect to both instruments. ¼ À Ã i: Ã vi Lii vi LiÃ iÃ Note that strategic complementarity ðviN =viÃ > 0. Delegation Delegation to independent policy authorities is a prominent solution to solve these inefﬁciencies while preserving non-cooperative policymaking. iÃ Þ þ LÃ ði. i 2.1) as a unique subgame perfect equilibrium in the repeated version of the game described above (Canzoneri and Gray. iÃ Þ ¼ 0 and i LiÃ ði. The strategies that implement this cooperative equilibrium (denoted by a C superscript) iC and i*C will solve the linear system: Li ði. solve the ﬁrst order conditions: Li ði.1. iÃ Þ þ t Ã ¼ 0: i (5) 8 Other solutions include repeated game mechanisms sustaining the equilibrium in (9) or (2. each government delegates policy to an independent policy authority by imposing a certain transfer function T(i). iÃ Þ þ t ¼ 0 and LÃÃ ði. t Ã Þ. 1985. assuming further that these contracts or transfer functions are linear (since the model is linear-quadratic). and the ‘delegated’ strategies iD ðt. iÃ Þ þ LÃ ði. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 (spillovers) are present in the sense that LiÃ s0. iÃ Þ ¼ LiÃ ði.O.8 Consider that before the non-cooperative game is played. iN(i*) and i*N(i) By implicit differentiation of (1) we obtain the slopes of these reaction functions as: LÃÃ viN L Ã viÃN i ¼ À ii .1.a. t Ã Þ.1. The cooperative equilibrium The cooperative equilibrium is obtained by joint minimization of an aggregate. non-cooperative equilibrium The Nash. iÃ Þ ¼ 0: i (3) The slopes of the optimal reaction functions can be found by implicit differentiation as before and compared to the Nash reaction functions. be different or equivalently. 2.3. iÃD ðt. and their sign depends on the sign of these i derivatives. iÃ Þ ¼ 0: i (1) Solving this system one can ﬁnd the (linear) reaction (or best response) functions of each policymaker as a function of the other policymaker’s strategy.1. iÃ Þ þ LÃ ði. The condition for existence and uniqueness of the equilibrium is in this case: Lii LiÃ iÃ À LiiÃ LiÃ i s0: The non-cooperative Nash equilibrium is inefﬁcient in the presence of externalities since the terms LÃ and LiÃ are non-zero. The Nash. in the (i. This condition is: DhLii LÃÃ iÃ À LiiÃ LÃÃ i s0: i i (2) 2. Under usual concavity i assumptions on L and L* existence and uniqueness of equilibrium require that these slopes. LÃ s0. . as in Cooper and John (1988). 1994).2.

1. t ÃP are non-zero) if and only if there is strategic complementarity/substitutability. Intuitively. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 397 Under usual concavity assumptions on L and L* this linear system of equations has a unique solution if and only if the condition (2) is satisﬁed. iÃD þ LiÃ iD . iÃC . iÃN ¼ 0: LÃÃ iÃ i Lii i i Since LiÃ ðiN . iÃD i i vt Ã vt Ã LÃ i i t ¼ ÀLÃÃÃ Ã LiÃ iD ðt. t Ã Þ. iÃN ¼ 0. Intuitively. t Ã Þ i i t Ã ¼ ÀLLiiiiÃ LÃ iD ðt.4.e. t Ã Þ. i. t Ã Þ. iÃC . iÃD ðt. iÃN Þ are non-zero (otherwise the Nash equilibrium would be efﬁcient. t Ã ÞÞ and min LÃ ðiD ðt. t ÃO ¼ LiÃ iC . t ÃP are a solution to (the ﬁxed point of): (9) Proposition 1. (8) Using (5) and (6) into (8). 0Þ ¼ iN and further from (9) that that tP ¼ t*P ¼ 0. credible contracts that governments will choose by backward induction based on their individual incentives. incentives for delegation occur because each government/principal recognizes it can inﬂuence the other player’s strategy by delegating. there is nothing to ensure that the precise marginal penalties implementing the optimum will in fact be chosen at the delegation stage: these contracts are not incentive-compatible. where iD ðt. iÃD ð0. iÃD ðt. t Ã Þ.F. First order conditions of this problem constitute a system of linear equations in t and t*: viD viÃD Li iD . on whether there is complementarity or substitutability) and on whether there are positive or negative spillovers. subgame perfect contracts t P . i (7) where the strategies are evaluated at their ‘cooperative equilibrium’ values found in (3). 0Þ Proof. ii LÃ iN . which implies LÃÃ i L Ã i L Ã iN . iÃD ðt. ÀLii (6) where D has been deﬁned in (2) above. iÃD ðt. Optimal versus subgame perfect. t Ã Þ is the solution found in (5). . Whether the equilibrium contract requires a penalty or a reward depends on the sign of the cross-derivative (i. this can hold if and only if LÃÃ i ¼ LiiÃ ¼ 0. iÃD ¼ 0 vt vt viD viÃD þ LÃÃ iD . the implicit function theorem can be applied to the system (5) in order to study the sensitivity of the equilibrium in terms of policy authorities’ strategies to changes in the contracts (which are treated as parameters in the policy authorities’ problem). suppose ¼ iN . Delegation always occurs in equilibrium (subgame perfect contracts t P .e. t Ã Þ : i We are now in a position to state our main results. if and only if there is no i strategic complementarity or substitutability. iÃD ðt. Under the same condition. The proof is by contradiction. 2. t Ã Þ. iÃN Þ and LÃ ðiN .O. t Ã ÞÞ t tÃ respectively. iD ð0. We solve for the subgame perfect. taking into account the reaction functions of the policy authorities iD. This is the result in Persson and Tabellini (1995. These contracts are a solution to min LðiD ðt. iÃD ¼ 0 LÃ iD . 1996). These can be found as: vi vt viÃ vt vi vt Ã viÃ vt Ã ! Lii ¼ À Ã LiÃ i LiiÃ LÃÃ iÃ i À1 ¼ DÀ1 ÀLÃÃ iÃ i LÃÃ i i LiiÃ . The argument of this paper is that these contracts are not a non-cooperative way of achieving cooperation through delegation. credible contracts The optimal contracts that implement the desired cooperative equilibrium in the non-cooperative game can be easily found by comparing (5) with (3) as: t O ¼ LÃ iC .i*D. i hence no need for delegation). as in the Nash equilibrium case.

398 F. Proposition 2 shows that optimal contracts can only be credible (subgame perfect) if governments cooperate at the delegation stage. We have shown that subgame perfect contracts are always non-zero (delegation is an equilibrium) whenever there are strategic complementarities or substitutabilities. Intuitively. See Appendix. Although interesting theoretically. Moreover. LÃC (10) Y þ c. this is done if each principal delegate to an agent and both contracts are public information. does it need to delegate policy rather than choose directly the optimal policy rule iC. iÃ Þ ¼ ¼ & ( Y. however. LÃ ði. otherwise ðY. 9 For details see Persson and Tabellini (1995). If the possibility of binding agreements existed. Subgame perfect. iÃD ðt. then choice of the optimal contracts is by no means insured. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 Proposition 2. . contracts consistent with the condition of this theorem are of a form that does not seem easily mapped into real-world policy institutions.i*C. they are always different from optimal contracts (which are instead equivalent to cooperative contracts. While this theorem shows that strategies of the form (9) implementing the cooperative optimum in a decentralized manner do exist. it is unclear why. however. iÃ Þ iD . iÃD ðt P . In the ‘linear contracts’ example. YÞ The strategies in (10). it will never choose the optimal contract tC. t ÃO are equivalent to ‘cooperative contracts’ t C . for instance in our example (for the home country): Tði. t ÃP Þ relative to the Nash equilibrium iN . there would be no need for delegation in the ﬁrst place. iÃ Þ. as can be seen by inspection of (9): the sign and magnitude of spillovers/externalities (ﬁrst derivatives of loss functions) and the sign and magnitude of complementarity (substitutability) given by the second (cross-)derivatives.t Proof. iÃN and the cooperative equilibrium iC . t Ã Þ. t Ã Þ. Optimal contracts t O . transferring this idea to linear contracts is dangerous. we relate our results to a general argument regarding delegation in Persson and Tabellini (1995). chosen by minimizing an aggregate. while it will always choose some contract. if the government had the ability to commit to and actually implement the optimal contract. iÃC depends on two factors. Proof. To summarize. T Ã ði. which is based on the Folk Theorem for Delegation Games in Fershtman et al. together with mirroring strategies for the foreign country. credible contracts t P . t Ã Þ þ LÃ iD ðt.9 but require that each penalty be written over both payoffs directly. Moreover. t ÃC found by solving h i min L iD ðt. See Appendix. we have introduced subgame perfect. each government has a choice parameter (t) and as we have shown. iff ½Tði. (1991). the cooperation technology needed is equivalent to the one needed to implement optimal policies iC. t ÃP are always different from cooperative contracts t C . if ½Lði. if one is to think about governments choosing a delegation scheme and facing a decision problem that can be reduced to choosing a set of parameters of the transfer function T. (1991). However. t ÃO . iÃ Þ. for the game-theoretical argument see Fershtman et al. Finally. t ÃP Þ. t ÃC and hence also from optimal contracts t O . iÃD ðt. iÃ Þ LC . it says nothing about their implementability or their being chosen in equilibrium (obvious issues related to this are observability of payoffs and inconsistency of sovereign policymaking with making contracts depend on the other country’s strategy). t Ã Þ : Ã t. credible contracts in a general linear-quadratic model with spillovers. Proposition 3. Proposition 2 serves as an intermediary result for Proposition 3. ‘global’ loss function). The position of the equilibrium occurring under credible contracts iD ðt P .O. implement the ﬁrst best once contracts are public information. In addition. c > 0 otherwise iC . This theorem provides conditions under which in a two-principal-two-agent game every Pareto optimal outcome of the principals’ game can become the unique subgame perfect equilibrium of the delegation game.

The policy instruments at this equilibrium obey: ði. iÃ . the policies have positive or negative externalities depending on whether vLð:Þ. (iv) policy instruments i. the exchange rate. The cooperative and commitment equilibrium A policy regime whereby the two policymakers decide before stage (i) to cooperate (i.i* are chosen simultaneously.e* are realized. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 399 2. eÃ (11) Using (11).1. the output gap. Throughout we assume that certain properties of loss functions. 11 10 . e. Suppose that in each country the policymaker has at its disposal one policy instrument (such as the interest rate or growth in a monetary aggregate for monetary policy) and denote this by i. (iii) shocks e. 2000).X* are fully determined. iÃ .O. eÃ .g. sÞ deﬁned over deviations of some macroeconomic variables stacked in the vector X from some target (socially desirable) levels s (X would include e.2. in each country there is a private sector forming expectations over some relevant subset of variables of X and hence ultimately over the policy instruments conditional on some information available one period in advance (UÀ1 ): ie ¼ E½irUÀ1 . expectations and shocks in both countries: À Á À Á X ¼ X i.e*).10 Suppose that in choosing the policies. through a variable such as the exchange rate). The main reason for doing so is that it has been recognized (since the work of Rogoff. vLvið:ÞX0: viÃ In addition. Solution method for model with a credibility problem We now brieﬂy outline the solution method for a model in which there is a credibility problem. the presence of a private sector forming rational expectations of some variable(s) combined with a real distortion in the economy gives rise to domestic incentives to deviate from optimality (deﬁned below) that are not related to cross-country spillovers. iÃ . either directly or indirectly (e. : Ã s:t: ie ¼ E½irUÀ1 . L* are Ã differentiable. expectations. Therefore. hence each variable will be hit by a stochastic shock and let the vector of such shocks be denoted by e.F. i .Âie . Additionally. the relevant macroeconomic variables can ultimately be expressed as a linear function of the instruments. the other one from the lack of commitment). 1995. i . ie . eÃ . Apart from the policymaker. 2.ieA* are determined. since loss functions are quadratic) ﬁrst order conditions. assume the model is stochastic.iÃ ' & Â À Ã e Ãe Á ÁÃ À E L i. ie. ieÃ . which is the case in most models considered in the literature. inﬂation.g. s . Assuming further L. (ii) expectations are formed. iÃ . quadratic loss function LðX. e. ie . Suppose as before that in each country a policymaker (the government) minimizes an aggregate. With these assumptions. sÃ (sssÃ ) (12) Strategic interactions in this model result from heterogeneity of targets and from different preferences for the stabilization of shocks. we also assume that the instrument in one country inﬂuences at least one of the macroeconomic variables of the other. to minimize a joint loss function) and commit to an optimal rule with respect to the private sector will be Pareto optimal (see e. policy sets. Persson and Tabellini. ieÃ . ie . eÃ . 1985) that policy cooperation may be suboptimal when a credibility problem exists. the policymaker faces the following timing in each period: (i) targets s. optimal contracts are usually designed to correct for both distortions (one coming from non-cooperation. e. (v) macroeconomic variables X.g.). iÃ Þ ¼ arg min C i. : þ LÃ i. ieÃ . i . ie .e. are met so that the considered equilibria do exist and are unique. iÃe . etc. e.e*) can be found by solving for11 the (linear. etc.i*C(e. XÃ ¼ XÃ i. the loss functions can be expressed as functions of instruments.s* are revealed. shocks and target levels: À Á À Á L ¼ L i. iÃe ¼ E iÃ rUÃ À1 (13) The optimal policy rules iC(e. LÃ ¼ LÃ i. iÃ . We label the equilibrium occurring under this benchmark regime ‘the cooperative and commitment equilibrium ’ and denote it with superscript C as in the previous section. rewritten after eliminating the Lagrange multipliers of the rational expectations constraints: This is the case in the ‘dynamic inconsistency’ literature.2. ieÃ . ieÃ ¼ E½iÃ rUÃ : As the two countries are interÀ1 dependent. when there are spillovers.

2.iÃe . say iN . or contract TðXÞ.ie . Tði. This equilibrium will obviously be inefﬁcient. : þ T Ã ði. iÃ . e.2.iÃ . iÃ . where X would be a subset of the relevant macroeconomic variables (e. Discretionary Nash equilibrium under delegation Consider that at stage (0). 2. :Þ such that the solutions to the systems coincide.400 F. : þ Tði. sÃ Þ LDÃ ð:. TÞ ¼ Lð:Þ þ Tði.3. iÃe .O. before stage (i) above. this function can also be deﬁned over instruments and shocks.iÃ .:ÞþTði.2. hence determining expected variables. iÃe .ie . iÃ .2.e. We study again delegation as a possible solution to both these inefﬁciencies. Solutions to this inefﬁciency usually considered in the literature consist of governments choosing the functions Tði. :Þ iÃ (16) The policy instruments can be solved starting from the ﬁrst order conditions given below.:ÞþT Ã ði. e. s.2.iÃe . and then substituting the latter in the original system (14). taking expectations of these to pin down expected variables and substituting these back in the original system: v½Lði. sÃ Þ (15) The independent authorities face these loss functions when choosing their policy instruments simultaneously at stage (iv). iÃ . only inﬂation for inﬂation contracts). :Þ ¼ T Ã ði. binding agreements of any sort are not possible) then the appropriate equilibrium concept to use is discretionary Nash equilibrium.iÃ . eÃ . Optimal versus credible contracts It is easily seen by direct comparison of (14) and (17) that optimal contracts (making the equilibrium under delegation identical with the Pareto optimum) should fulﬁll: . this equilibrium is unrealistic: since real-world policymaking is best described in a noncooperative setup (i. :Þ. in a non-cooperative and discretionary manner. iÃ . the policy rules are obtained by taking conditional expectations of the system (14). each government delegates the policy to an independent policy authority by imposing a certain transfer function.4. iÃ . in the latter. The corresponding discretionary Nash equilibrium policy rules under delegation will be given by: Á Ã Â À iD ¼ arg min L i. leading to the pure Nash equilibrium choices.:Þ viÃ ¼ 0 ¼ 0 (17) 2. iÃN . i. T Ã ði.:Þ vi v½Lði. Ultimately. iÃ . iÃ . hence delegation would mean assigning loss functions of the form (where superscript D stands for ’delegated’): LD ð:. T Ã Þ ¼ LÃ ð:Þ þ T Ã ði.g. The two sources of inefﬁciencies mentioned before are obvious by comparing the systems (14) with (17) evaluated at T ¼ T* ¼ 0. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 Ã vE½Lð:ÞþLÃ ð:Þ þ E vE½Lð:ÞþL ð:ÞrUÀ1 vi vie i h Ã vE½Lð:ÞþLÃ ð:Þ þ E vE½Lð:ÞþL ð:ÞrUÃ À1 viÃ viÃe h i ¼ 0 ¼ 0 (14) Upon specifying functional forms for loss functions and for the models determining the macroeconomic variables.iÃ . Discretionary Nash equilibrium without delegation Consider ﬁrst the case without delegation. due to two reasons: ignoring the spillovers of policy to the other countries’ loss function (as in the previous section) and ignoring externalities on the own-country private sector (being more speciﬁc about the source of inefﬁciencies would require a parametric example which we postpone to the next section). two terms are absent that come from ignoring externalities on (i) the other policymaker and (ii) the private sector. :Þ ¼ 0. :Þ i Â À Á Ã Ã iD ¼ arg min LÃ i. iÃ . eÃ .e. ie . However. iÃ . ie . s.

). t Ã . This is illustrated in the following example. :Þ. :Þ. credible contracts are determined by: io n h t P ð:Þ ¼ arg min EL iD ðt. iDAÃ Þ given by (16) and fulﬁll the ﬁrst order conditions (17). t Ã . The ﬁrst order conditions that credible contracts fulﬁll12 are: vE½Lð:Þ viD ðt. For the sake of simplicity and for their widespread use. cooperation or some form of coordination of governments/principals is unequivocally necessary. Specifying functional forms for the transfer functions usually results in a solvable system for the delegation parameters. iDÃ ðt. To sustain optimal contracts as an equilibrium phenomenon. which we label by tC and tCA*. To choose the most prominent example. t Ã .F. we restrict the functional form of the transfer functions to linear contracts TðiÞ ¼ k þ ti. t Ã Þ þ ¼ 0 vi vt viÃ vt vE½LÃ ð:Þ viD ðt. the cooperation problem being not solved but merely relocated to the delegation stage. the only modiﬁcation would be that the number of parameters. which is why one considers delegation in the ﬁrst place). by delegating the principal modiﬁes the reaction functions of the agent in a linear way (or else.O.the timing is: (0) governments delegate policies to independent authorities by imposing the transfer functions T (. and vT=vi ¼ t.:Þ C viÃ Ã vE½Lð:Þ þ E vE½Lð:ÞþL ð:ÞUÃ À1 viÃ viÃe where loss functions are evaluated at the cooperative and commitment optimum iC. t Ã Þ vE½Lð:Þ viDÃ ðt. Even with this form of delegation. but this is not compatible with individual incentives of governments. T*(. T Ã ðiÃ Þ ¼ kÃ þ t Ã iÃ : The subgame perfect. (i)-(v): same as before.:Þ C vi ¼ ¼ Ã vE½LÃ ð:Þ þ E vE½Lð:ÞþL ð:ÞjUÀ1 vi vie h h i i (18) . where credible and optimal contracts are clearly different in an intuitive way.k* can be chosen such that the participation constraint of the policy authority is met. The solution method is based on backward induction: policy authorities choose their policy instruments independently and discretionarily taking delegation as given and governments choose delegation parameters taking into account the choice of policy instruments made previously by the delegated authorities. t Ã . In most cases.). for Jensen quadratic contracts there would be three parameters and three ﬁrst order conditions). the Nash equilibrium could be made identical to the Pareto optimum.iÃ . The policy rules governments face at stage (0) are ðiD . vT Ã ði.e. 12 Note that this can be done more generally for a certain functional form of T as long as it is differentiable. where k and t are the delegation parameters to be chosen. : t n h io t ÃP ð:Þ ¼ arg min ELÃ iD ðt. This implies that optimal contracts are not consistent with individual rationality of the governments. which makes them hard to implement and undermines their credibility. suppose T ‘s are linear functions of the policy instruments and can be expressed as T(i) ¼ k þ ti. iDÃ ðt.e. :Þ. t Ã Þ vE½LÃ ð:Þ viDÃ ðt. Finally. The system (18) fully determines the ’optimal contracts’ t and t*. linear inﬂation contracts. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 401 vTði. dependent on the realization of the shocks). : Ã t (19) The other parameters k. Note again the difference with the Folk Theorem in delegation games: here. would increase (for example. Jensen (2000) addresses this problem by showing how the ﬁrst best in (13) can nevertheless be implemented through state-independent delegation by choosing a quadratic form for the transfer function T. when we model the delegation stage (0) as a separate stage of the game -whereby governments choose their ‘strategies’ (the parameters determining the transfer functions).i*C found in (13). . if contracts nonlinear) instead of ‘forcing’ the Nash equilibrium to overlap with the desired Pareto optimum. when externalities are present. these marginal contracts are statedependent (i. t Ã Þ þ ¼ 0 vi vt Ã viÃ vt Ã (20) The objects in (18) and (19) are different in most situations (i.iÃ . and hence the number of ﬁrst order conditions to solve. :Þ.

13 In fact. (24) relates the relative prices z of the two goods to their relative demand. 2000) and consists of directly postulated reduced-form equations. this is equivalent to period-by-period minimization. e with e and s with (0. . 14 When the game is repeated over time. each one being specialized in producing a consumer good. For simplicity.402 F. The private sector forms rational expectations of producer inﬂation (and e e hence money growth). Then the policymakers’ task is to minimize the expected value of the following conventional period loss function. because there is some form of nominal rigidity). Having one shock that creates incentives to deviate from cooperative policy is sufﬁcient for our point. a star denotes a foreign country variable (for brevity just the home country’s model is presented) and time subscripts have been suppressed: y ¼ gðp À pe Þ À e p ¼ m zhs þ pÃ À p z ¼ dðy À y Þ Ã (21) (22) (23) (24) (25) p ¼ p þ bz Deviations of output growth y from the natural rate (normalized to zero) are deﬁned in (21) by a usual expectations-augmented Phillips curve. we abstract from all shocks other than supply shocks for simplicity. The model parameters are symmetric for simplicity but the shocks hitting the economy are arbitrarily correlated. the policymakers minimize the expected present discounted value of this loss function. where inﬂation surprises in producer price inﬂation p matter. All variables are in log-differences.0). where d > 0 is the inverse relative demand elasticity of outside goods.13 Real exchange rate appreciation z is deﬁned in (23) as nominal depreciation s plus the differential of producer inﬂation. The model is an adapted version of Persson and Tabellini (1996. A higher supply of foreign goods reduces z (real appreciation) by inducing a relative excess demand for home goods.q). For simplicity. normalize the socially optimal inﬂation to zero and suppose the desirable output q is greater than the natural rate due to some real distortion (monopolistic competition. i with m. which it uses for short-run stabilization (for instance. Observe that the only source of uncertainty in the economy is given by adverse supply shocks (e. as the expectation of the latter over the distribution of shocks.p.O. where b is the share of the latter in the domestic consumption basket. in contrast to Persson and Tabellini. conditional upon the information set UÀ1 of previous realizations of macroeconomic variables and model parameters: pe ¼ me ¼ E½pjUÀ1 ¼ E½mjUÀ1 (26) Social welfare in each country is deﬁned over variability of output and inﬂation from some socially desirable levels.s2Ã .e*) with zero mean. which is an imperfect substitute for the other country’s good. hence deﬁning an inverse demand equation.z). different variances and arbitrary covariance (s2 . in (22) we suppose the growth rate of money is the same as that of producer inﬂation. This generates the main spillover of policy through the real exchange rate. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 3. The timing is as in the previous section: just substitute X with (y. and therefore we abstract from velocity shocks. seeÃ .p. The policy is also subject to a credibility problem generated by a real distortion making the natural rate of output (employment) suboptimally low. consumer price index inﬂation p is producer inﬂation plus inﬂation induced by the consumption of foreign goods. optimal inﬂation contracts in international monetary policy We use a parameterized version of the model in the previous section for an illustrative example. Finally. The world consists as before of two countries. An example: credible vs. Since the stage game is always identical.14 using as instruments the money growth rates m: Lð:Þ ¼ o 1n 2 p þ lðy À qÞ2 2 (27) We assume that q > 0 giving rise to the domestic inﬂation bias described in the previous section. Each country has a monetary policy instrument. for instance).

t Ã Þ ¼ Àt Ã þ t þ ðe À eÃ Þ tÀ 2 1þa AÀB A À B2 A À B2 A À B2 mD ðq.15 The responses are optimal due to the cooperative features of the equilibrium. eÃ Þ where E½t ¼ t and E ~ ¼ 0 tð t h i t t t Ã ¼ t Ã þ eÃ ðe. a familiar inﬂation bias (the q term) is present in each country (and is the same in both due to the assumption on homogeneity of targets) due to discretion. First.O. the responses to both domestic supply shocks and relative shocks are different from the 15 See e. each central bank will minimize its loss function (27) modiﬁed as in the system (15). namely: hi t ¼ t þ ~ e. eÃ . eÃ Þ ¼ mÃC ðe. t. without getting into computational details. eþ 1 þ bg ð1 þ bgÞ ð1 þ 2aÞ2 þbg b dð1 þ 2a À bgÞ ðe À eÃ Þ. we shall apply directly the solution method described in the last section. B ¼ að1 þ aÞ): À Á d 1 þ a2 b b ð1 þ aÞA ~ ð1 þ aÞB eÃ eþ 2 ðe À eÃ Þ qÀ 2 t þ ¼ Àt þ tÀ 2 AÀB A À B2 1þa A À B2 A À B2 À Á b ð1 þ aÞB ~ ð1 þ aÞA eÃ b Ã d 1 þ a2 qÀ 2 e À 2 mÃD ðq. at stage (0). The cooperative optimum. 16 In Bilbiie (2000) we show equivalence of linear inﬂation contracts and inﬂation targets in this framework. She also stabilizes relative shocks e À eÃ due to their indirect impact on welfare through real exchange rate appreciation/ depreciation. 1996. the policymaker stabilizes domestic supply shocks (due to their inﬂuence on output and inﬂation). eÃ Þ ¼ b dð1 þ 2a À bgÞ ðe À eÃ Þ.1. Note also the absence of the inﬂation bias due to commitment (expected policies are zero). eÃ . non-cooperative inefﬁciency and optimal contracts For the sake of brevity. The cooperative and commitment equilibrium (as in (13) and (14)) is attained in this example for the optimal state-contingent policy rules: mC ðe. eÃ À 1 þ bg ð1 þ bgÞ ð1 þ 2aÞ2 þbg (28) where we used the change of notation b ¼ lg. Secondly.g. e. as expected. e. Bilbiie (2000) or Persson and Tabellini (1995. 2000) for details and the solution of slightly more general models. Each policymaker internalizes the effects of its instruments on both the other country’s welfare and its domestic private sector. For the moment we assume this takes the form of linear inﬂation contracts of the type considered by Persson and Tabellini16: each government imposes a transfer function on its central bank of the form T ¼ k þ t p. Suppose delegation to an independent central bank has taken place before the stage game is played. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 403 3. Lð:Þ ¼ 1fp2 þ lðy À qÞ2 g þ Tð:Þ. t.g. e. a ¼ bdg and d ¼ bd: At the optimum.F. The discretionary Nash equilibrium policy instruments 2 given delegation are found as in the system (16) in the previous section as (where the same change of notation as before was used and additionally A ¼ ð1 þ aÞ2 þ bg. We shall focus only on contracts for the sake of exposition but the results apply equally to delegating with a non-zero inﬂation target. eÃ Þ where E½t Ã ¼ t Ã and E eÃ ¼ 0 (29) Given these linear penalties. t Ã Þ (30) The purely non-cooperative discretionary equilibrium without delegation ðt ¼ ~ ¼ t Ã ¼ eÃ ¼ 0Þ t t features two inefﬁciencies. T Ã ¼ kÃ þ t Ã pÃ : The marginal penalties t and t* are allowed to be state-contingent. The non-cooperative and discretionary equilibrium is also solved by the corresponding method described in the previous section ((17)). Given the linear nature of the model and linearity in stochastic shocks we model this by assuming that each marginal penalty is additively separable in a state-independent and a state-dependent component. .

In our model. Following for instance Persson and Tabellini (1996) we shall call this a stabilization bias. mÃ is chosen in a non-cooperative and discretionary manner. then only the domestic incentives are corrected for.404 F. it is worth noting that delegation works by penalizing the cross-country inﬂation differential between home and foreign (via m < 0) in order to correct for the suboptimal response to asymmetric shocks featured in the non-cooperative equilibrium. m ¼ : 1 þ 2a a 1þa Beyond familiar terms purported to correct for the average inﬂation bias (pB ) and the suboptimal response to symmetric shocks (by requiring a relatively more liberal central bank. e. they would require changes in institutions for every realization of shocks). eÃ Þ ¼ 1þa bq À apC ðe. eÃ Þ b ab 2a2 b i ðe À eÃ Þ h qÀ eþ ¼ 1þa ð1 þ aÞð1 þ bgÞ ð1 þ bgÞ ð1 þ 2aÞ2 þbg t C ðq. To achieve this. a quadratic contract with targets of the form: Tð:Þ ¼ 1faðp À pB Þ2 þ mðp À pÃ Þ2 g . the design of ‘state-dependent institutions’ in practice is not (for instance. state-independent delegation makes the loss function depend upon realization of the other country’s macroeconomic 17 More details on interpretation of incentives in this equilibrium can again be found in Bilbiie (2000). this implies that the contract should be. The penalty is weaker if the foreign country suffers an adverse supply shock (eÃ > 0) or a less severe supply shock as compared to the home country (e À eÃ ). leaving suboptimal shock stabilization unaltered. The ﬁrst terms are the familiar ones correcting for the domestic inﬂation bias in each country. eÃ Þ ¼ 1 1þa (31) The marginal penalties are intuitive. Following the same solution method as above it is readily shown that the optimum is implemented for: a¼À a b a . Jensen argues that since there are three distortions that optimal delegation should correct for. when an adverse shock is realized. The exact nature of the distortions will depend on the shocks and the values of parameters but as a general rule the policies would have a contractionary bias when a favorable shock hits (positive externalities) and would be too expansionary. the particular form of delegation should feature three parameters. In these two cases foreign inﬂation is positive and the real exchange rate appreciates at home.1. respectively). compared the state-contingent contracts that make the marginal penalties depend directly upon the realization of home and foreign shocks. one for each distortion. a < 0).17 Following the analysis in the general case we can easily ﬁnd the marginal penalties that implement the cooperative and commitment optimum when policy m. m are decision variables of the 2 government when delegating. Jensen (2000) or Persson and Tabellini (1996). system (18) in this case translates to: É È bq À apÃC ðe. State-independent optimal delegation While the marginal penalties’ being state-contingent is intuitive. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 optimal ones due to not internalizing of policy externalities when acting non-cooperatively. as in Jensen. . where a. depending on the q term) for additional inﬂation is needed to correct for that. at the other country’s cost. pB ¼ q. In this case the home policy is too contractionary and a reward (or lower penalty. pB . 3. the central bank’s incentives should be corrected in all three dimensions.1. corresponding to the three terms in (31) (the average inﬂation bias. eÃ Þ b ab 2a2 b i ðe À eÃ Þ h qÀ eÃ À ¼ 1þa ð1 þ aÞð1 þ bgÞ ð1 þ bgÞ ð1 þ 2aÞ2 þbg É È 1 t C ðq.O. 18 For an elaborated critique of state-dependent delegation see Jensen (2000). Mutatis mutandis. the suboptimal response to symmetric and asymmetric shocks. Note that. e. The other terms correct for suboptimal stabilization of shocks. This problem is solved by Jensen (2000) by proposing a delegation scheme based on quadratic contracts with targets.18 If only state-independent contracts are feasible.

governments face the policy rules contingent on contracts that we solved for previously in (30).2. may be surprisingly so. . where inﬂation and the output gap are evaluated at the delegated Nash Equilibrium. Credible contracts: the linear inﬂation contracts case Although linear contracts à la Persson and Tabellini implement the optimum (albeit with statecontingent parameters). 3. e.g. t (32) Substituting the delegated Nash equilibrium money growth rates from (30). e. But the logic of the two mechanisms is consistent. E½LÃ ð:Þ and minimizing the latter two with respect to t and t Ã respectively yields the two ﬁrst order conditions: pD t. t. The advantage of state-independent delegation. h qÀ eþ 1þa ð1 þ aÞð1 þ bgÞ ð1 þ aÞð1 þ bgÞ ð1 þ 2aÞ2 þbg one can notice that domestic inﬂation is rewarded when there is an adverse symmetric supply shock and penalized when there is an asymmetric shock. By backward induction. ~ t Ã . indeed. t Ã Þ. eÃ Þ and t P ðe. e. but he focuses on identiﬁcation of optimal institutions and not their implementability. t Ã . q. Substituting mD ð:. eÃ Þ ¼ ðe À eÃ Þ 1þa 1þa 1 þ 2a (33) We are now ready to compare these non-cooperative credible contracts with the optimal contracts implementing the ﬁrst best. consistently with the intuition in the state-independent case. eÃ Þ: Another way to read this is that if only state-independent delegation were possible. they are both subject to the problem we identiﬁed in the previous section. We treat the state-independent part of the contract as control variables of the government. t. focusing on the home country. rewriting the state-contingent contracts derived in (31) for the home country: t C ðq. ~ t Ã . we refer the interested reader to Jensen (2000) for a more detailed discussion of state-independent delegation. however. A ﬁrst thing to note is that.e. eÃ .] incentives causing policymakers to deviate from cooperative policies would also cause governments to deviate from cooperative institutions’. We shall now see an example of this at work. t Ã Þ found in (30) into tð E½Lð:Þ. eÃ ¼ 0 t. eÃ Þ we use (29). eÃ Þ. there is a larger favorable shock). and for ﬁnding the equilibrium state-contingent part of the contract ~ e. the two contracts would coincide. i. although they would then be both suboptimal in that they would not affect stabilization of shocks. at the delegation stage (0).19 To see what contracts government will choose (and hence implement) based only on their individual rationality and their perception of rationality of the agents to which they delegate (central banks) we follow the solution method outlined in the general case. The inefﬁciency of credible contracts comes from suboptimal responses to shocks (second and third term). eÃ Þ ¼ b ab ab þ ab2 g þ 2a2 bð1 þ aÞ iðe À eÃ Þ. eÃ ¼ 0 t. e À eÃ > 0: In equilibrium pÃ is greater than zero and there is a contractionary bias of the home 19 Jensen recognizes this problem himself in the last paragraph of the mentioned paper ‘[. mÃ Dð:.F. Hence. eÃ Þ: Using state independence of the ﬁrst two and zero mean of the last two one gets a solution for credible contracts as (where a and b are deﬁned as before): t P ðe. the state-independent term leading to elimination of the systematic inﬂation bias is the same in t C ðe. e. q. is that institutions need not change every time a shock hits the economy. we get two equations in tð t four unknowns t. the ‘home’ government will only choose t . ~ e. eÃ Þ ¼ b b Ã b qþ e þ ðe À eÃ Þ 1þa 1þa 1 þ 2a b b b qþ eÀ t ÃP ðe. Consider again the case where the foreign country is hit by an adverse supply shock eÃ > 0 and this is less severe than in the home country (or equivalently. eÃ ðe. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 405 outcomes (in this case inﬂation).O. and the quadratic contracts with targets à la Jensen also solve the problem of state-contingency. They then minimize the expected values of the social losses given by (27) (and its foreign counterpart). eÃ . t pÃD t.

t ÃP Þ ¼ 1þ2aðe À eÃ Þ. or targeting rules in a new open-economy model as in Benigno and Benigno (2005). at the delegation stage governments increase the penalty. tailored to achieving a zero inﬂation consistent with the ﬁrst order condition (32). The two contracts are different even when shocks are perfectly correlated. whereas the credible one is b b t P ¼ 1þaq þ 1þae. 4. On the contrary. t ÃP in the best response functions one gets d d mD ð:. 2000).20 The linear optimal contracts t C . mÃD ð:. monetary policy. To achieve the zero inﬂation in the perfect equilibrium an increasing marginal penalty is needed. delegation schemes have been viewed as panacea for both domestic credibility problems and inefﬁciencies coming from cross-country spillovers. Given policy externalities. it is hard to understand why then wouldn’t governments cooperate directly without any need for delegating policies. each government delegates to an independent policy authority by imposing to the latter a certain transfer function. t ÃP Þ ¼ À1þ2aðe À eÃ Þ: Whether these will be higher or lower C . but there are strong incentives to deviate from it. the penalty will be smaller. two governments seeking to delegate policy in a manner consistent with their individual rationality (or with their mandate. Delegation with credible contracts would mean elimination of the inﬂation bias but still suboptimal shock stabilization. The different contracts can be directly compared by substituting them in the best response functions (30). Alternatively. t ÃC implement the ﬁrst best optimal money growth rates mC . In a prominent example. In these cases. In the optimal contract tC. the Nash equilibrium policies would feature an inﬂation bias and suboptimal shock stabilization. and so do the Jensen quadratic contracts and the contracts consistent with the Folk Theorem in delegation games presented in (10). this is far from the non-cooperative setup one wishes to describe in the ﬁrst place. As the spillovers are ignored in the perfect equilibrium. Conclusions Various inefﬁciencies associated with policymaking. than m Hence.a. whether at a domestic or international level. mÃC depends on the nature of the shocks and the parameters. where the foreign country faces a favorable supply shock eÃ < 0 relatively smaller than in the home country (e À eÃ < 0Þ there is an expansionary bias of home monetary policy. in the symmetric case b ab whereby e ¼ eÃ : The optimal marginal penalty is t C ¼ 1þaq À ð1þaÞð1þbgÞe .O. a policy regime where governments cooperate (and commit with respect to the private sectors) is unequivocally Pareto optimal. In the converse case. t P . The governments can then in 20 In Bilbiie (2000) we provide a more detailed welfare comparison of equilibria. t P . whereas in the considered case the optimal response would be a positive producer inﬂation. the linear inﬂation contracts proposed by Persson and Tabellini (1995. 1996. If compromises are to be made in terms of allowing for the possibility of binding agreements at the level of governments. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 country’s monetary policy. An adverse common supply shock generating a contractionary bias is optimally corrected by a decrease in the penalty for additional inﬂation. a supranational institution able to coordinate the two governments on the ‘right’ institutions would do the same job. both coefﬁcients on shock stabilization are negative: the optimal penalty in the home country decreases to correct for the deﬂationary bias. Some simple and intuitive delegation schemes easily mappable into real-life institutions have been found to ‘ﬁx’ both these incentives: i. In order to delegate with the scheme that would insure that optimal policies are followed they would need to cooperate at the delegation stage. But the optimal response should in fact target deﬂation. mÃC . note from (32) that the credible contract is always imposed so that producer inﬂation is zero. that is maximizing social welfare) would fail to achieve a ﬁrst best equilibrium. aggravating the contractionary bias of home policy. therefore aggravating the deﬂationary bias.406 F. in the credible contract tP both coefﬁcients are positive: the penalty is increasing in e and e À eÃ . so they will not lead to implementation of mC . . However. Not recognizing the positive externality that would result from both countries inﬂating more. quadratic contracts with targets of Jensen (2000). can allegedly be solved by delegation of policy to independent monetary authorities. Without delegation. mÃC : By substituting t P . Equivalently. This negative spillover would be eliminated through optimal delegation: the penalty becomes larger (see (31)) to reduce inﬂationary incentives.

In our context. where it turns out that the contracts governments would actually choose are different in an intuitive way from the optimal ones for any correlation of shocks. preserving the non-cooperative assumption about policymaking? One possible solution consists of creating of a supranational institution that is able to ‘coordinate’ governments on the optimal contracts at the delegation stage.e.21 First.F. There. However. the equilibrium in the agents’ game will depend on the delegation in no simple way as this will change their reaction functions. coordinate on. However. however. in our view.22 We extend the analysis to a model featuring a domestic credibility problem and present an example from international monetary policy cooperation. we then show that optimal contracts occur in equilibrium only under cooperation at the delegation stage (the same sort of cooperation needed to implement optimal policies): optimal contracts are equivalent to ‘cooperative contracts’ found by minimizing the global loss function. taking into account the agents’ choices at a future stage. Further research. the supranational institution would help design the appropriate incentive schemes of countries’ policy authorities and monitor their implementation over time. for such a supranational institution would simultaneously be a common principal to and a common agency of the sovereign states. a best equilibrium among a multiplicity of feasible equilibria. i. such an institution helped governments choose. credible contracts in an open-economy framework. But if this were the case. . we provide a general solution method and deﬁne the new equilibrium concept in a nonparametric model of policymaking with spillovers. An alternative would be strengthening the role of some existing supranational institutions (such as the International Monetary Fund). and hence also from the optimal ones. credible) contracts based on its individual rationality. it is hard to see why governments need to delegate instead of committing themselves to the optimal policy rules. The equilibrium (and hence coordination by the international principal). which dealt with ‘take-it-or-leave-it’ target compensation functions by which the equilibrium in the agents’ game can be made identical to a Pareto optimum. credible contracts are always different from cooperative contracts.).. In order to pin down the equilibrium one has to pin down some values for the delegation parameters. 21 It is important to note that once we consider transfer functions we are no longer in the conditions of the Folk Theorem in Delegation Games of Fershtman et al. By the presence of such an international institution. sovereignty of policymaking is preserved. this raises again the issue of optimal design of the delegation scheme of the supranational organization itself. further institutional challenges would be raised by this setup. perhaps along the lines of the ‘common agency’ problem analyzed in Dixit and Jensen (2003). We do this by supposing that each government chooses the (subgame perfect. but the latter case features again inefﬁciency. is in terms of policies directly. We ﬁrst show that delegation always occurs in equilibrium when there is strategic complementarity or substitutability.23 Our analysis raises a normative question: what could then insure implementation of the ‘right’ institutions. their design and the mechanisms by which they could implement and monitor globally optimal policy regimes. The argument of this paper is that this implementation mechanism hides an implicit assumption about governments being actually able to sign binding agreements in order to coordinate on exactly those delegation parameters that ‘do the job’. an explicit modeling of the delegation decision of the governments. We then arrive at our main point: subgame perfect. The way out from this dilemma is. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 407 theory design the contract to ensure that the delegated authorities choose the policy instruments that implement the desired equilibrium. is needed in order to analyze the incentives of such supranational institutions.g. we think it is indeed more realistic to assume that national governments agree to coordinate on some institutional features than to systematically pursue cooperative policies that are not consistent with their incentives. which reinforces arguments made by Canzoneri and Henderson (1991) in a different setup. which seems hard to map into real-life practice.O. Rather than studying enforceability (and sustainability over time) of optimal contracts. 23 The two are identical only if shocks are absent and/or only state-independent institutions are feasible. when viewed in its role of common agency. we solve for incentive-compatible. Once we make the transfer functions linear (e. While this might seem akin to centralization or cooperation on policies directly (which would be a solution by assumption). 22 Note that our results are different from McCallum’s (1995) critique concerning closed-economy monetary institutions or more speciﬁcally enforceability of inﬂation contracts.

iÃD Þ ¼ Àt and LÃÃ ðiD . t C . iÃD À t Ã i vt iviD h i h LÃ iD . iÃD Ã vt viÃD þ LÃ iD . iÃD À t ¼ À LiÃ iD . Appendix: A Proofs of Propositions 2 and 3. Paul Mizen.O. Proof of Proposition 3 The proof is by contradiction. which substituted in the above yield: i h iviD h i LÃ iD . is a solution. t ÃC . viD vt Ã which substituted in the ﬁrst equation in (35) delivers: h i h i LiÃ iD . t Ã Þ. I thank without implicating the Oesterreichische Nationalbank for conferring me the Olga Radzyner Award for an earlier version of this paper. t ÃO . and Banque de France for ﬁnancial support through the Chaire Banque de France at the Paris School of Economics. LiiÃ LÃÃ i i (36) where the second equality uses the result in (6). Guido Tabellini and participants at the Bank of England Seminar and EUI Monetary Economics Workshop and Informal Theory Group for comments on earlier versions. (36) can be satisﬁed if and only i i if LiÃ ðiD . iÃD þ LiÃ iD .1. t Ã Þ i t Ã ¼ LiÃ iD ðt. iÃD Þ ¼ Àt Ã . t ÃC are identical: comparing the ﬁrst order conditions in each case. iÃD Li iD . iÃD i Ã vt viÃD À0 vt viÃD vt Ã (34) From (5). we know from (7) above that t O . Marcus Miller. we know Li ðiD . different version of this paper has been circulated under the title ‘Perfect versus Optimal Contracts: an Implementability-Efﬁciency Trade-off’. iÃD À t Ã þ LiÃ iD . iÃD i Ã vt viD þ LÃÃ iD . As long as the necessary and sufﬁcient condition for existence and uniqueness of equilibrium (2) is satisﬁed (Lii LÃÃ iÃ sLiiÃ LÃÃ i ). Gianluca Benigno. this can happen if and only if: . Suppose that t P . t Ã Þ Under our assumptions on Li (notably. iÃD vt viD þ LiÃ iD . iÃD À t Ã i viÃD vt Ã . Henrik Jensen. iÃD Þ À t Ã ¼ 0. iÃD À t Ã ¼ LiÃ iD . iÃD À t Ã viÃD vt Ã viD vt Ã viD vt viÃD vt i L LÃ h ii iÃ iÃ ¼ LiÃ iD . iÃD À t þ LiÃ iD . iÃD ðt. (34) with (8).408 F. which implies that cooperative contracts t C . t ÃC obey: t ¼ LÃ iD ðt. t ÃC ¼ t ÃO . t Ã Þ. condition (2) holds) this is a linear system with a unique solution. t ÃP and t C . On the other hand. iÃD ðt. iÃD À t Ã i vt The second equation in (35) implies: viÃD ¼ 0 vt viÃD ¼ 0 vt Ã (35) h i h i LÃ iD . A. iÃD À t Ã . iÃD i vt viÃD þ LÃ iD . A. All errors are mine. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 Acknowledgments A previous. I am grateful to an anonymous referee and Ben Lockwood in particular and to Mike Artis. so we conclude that cooperative contracts and optimal contracts coincide t C ¼ t O . iÃD i vt viD þ LÃÃ iD . Proof of Proposition 2 The ﬁrst order conditions for cooperative contracts are: viD Li iD .2. Roberto Perotti.

1991.1988. 2005. Economic Cooperation in an Uncertain World. C. t ¼ À iÃi LÃ iD . MIT Press. Handbook of International Economics. T. A.. 1991. Interest and Prices.. and hence also from optimal contracts (using Proposition 2). Observable contracts: strategic delegation and cooperation. Benigno. 1994. 473–506. Dixit. Judd. Gray.. Two fallacies concerning central-bank independence. MA. American Economic Review. B. Cambridge MA. Canzoneri. Designing targeting rules for international monetary policy cooperation. A simple approach to international monetary policy coordination.. iÃD ¼ 0 i i i i LiiÃ Li iD . Masson. P. iÃD ¼ 0 Li iD . Inﬂation Contracts. A. M. American Economic Review. 2000. 551–559. E. P. Princeton. Kalai. iÃD Þs0. 2002. LiiÃ LiÃ iÃ which is a contradiction since violates the condition for existence and uniqueness of equilibrium (2). Tabellini.O.F.. M. iÃD ¼ 0 i i vt vt viD viÃD þ LiÃ iD .. Double-edged incentives: institutions and policy coordination... . Obstfeld... which substituted above give: i (38) t ¼ À LÃÃ Ã Lii D ÃD Ã . 1985. Benigno. III (North-Holland). A strategic analysis of monetary interdependence. 1985. i i LiiÃ Li Ã i However. Rogoff. Blackwells. Journal of International Economics 56 (2). G.. References Benigno.. G. Tabellini. Jensen.. MIT Press.). 2000. Cambridge. The Quarterly Journal of Economics 103 (3). iÃD ¼ 0 From (5) we know that Li ðiD . Papers and Proceedings 85. 1995. Woodford. Pesenti. Optimal monetary policy cooperation through state-independent contracts with targets. 111–117. 199–217. M.. K. iÃD À Lii LiÃ iD . Political Economics: Explaining Economic Policy. Princeton University Press. H. International Economic Review 32. F. A. K. Papers and Proceedings.. Equating the expressions for either t or t Ã we obtain (using that there are externalities LiÃ ðiD . G. H. 677–700. M. NJ. iÃD LiÃ i . Persson. 441– 464. European Economic Review 44. we know that perfect contracts have to satisfy also (9).. MIT Press. Journal of Political Economy 84. 371–385. K. Bilbiie. 2001. Corsetti. T. R. K.. Fershtman. Economic Journal 113. iÃD Þ ¼ Àt Ã . T. Bilbiie / Journal of International Money and Finance 30 (2011) 393–409 409 viD viÃD LÃ iD . Coalition formation in international monetary policy games.. Journal of International Economics 18. Henderson. G. 421–445.A. Working Paper ECO 2001/16. 517–539.. Monetary Policy in Interdependent Economies: A Game-theoretic Approach. J. Persson.. iÃD Þs0 by assumption): LÃÃ Lii i ¼ Ã i.. coordinating coordination failures in Keynesian models. Foundations of International Macroeconomics... 539–549. Journal of International Economics 57 (1). UK. (Eds. International Economic Review 26.. iÃD Þ ¼ Àt and LÃÃ ðiD . 1995.. Persson. Welfare and macroeconomic interdependence. Cambridge. Canzoneri.. K. Cooper.. Therefore. Targets and Strategic Incentives for Delegation in International Monetary Policy Games.. MA. 2000. G.. Rogoff.. vol. Monetary policy games and the consequences of non-cooperative behaviour. M.. 207–211.. G. 2003. D. Ghosh.O.. iÃD Ã vt vt Ã We substitute the partial derivatives of iD and iÃD using the result in (6) to obtain: (37) ÀLÃÃ iÃ LÃ iD . Monetary cohabitation in Europe. 547–564. Jensen. International monetary policy coordination may be counterproductive. European University Institute. 1996. Journal of Monetary Economics 53 (3).. Tabellini. In: Grossman. Hamada. McCallum. 1996. Common agency with rational expectations: theory and application to a monetary union. P. Oxford. 177–196. subgame perfect contracts are always different from cooperative contracts. P. 2002. 2003. LiÃ ðiD . Kohler. John. iÃD þ LÃÃ i LÃÃ iD . Rogoff. iÃD þ LÃÃ iD . 1976. The Quarterly Journal of Economics 116 (2).