Studies in

Contemporary Economics
Didier Laussel William Marois
Antoine Soubeyran (Eds.)

Monetary Theory and Policy

Proceedings of the Fourth International
Conference on Monetary Economics and Banking
Held in Aix-en-Provence, France, June 1987

Spri nger-Verlag
Berlin Heidelberg New York
london Paris Tokyo

Editorial Board
D. BOs G. Bombach

B. Gahlen

K. W. Rothschild

Editors

Didier Laussel
Professor of Economics
University of Aix-Marseille II, Faculty of Economics
13621 Aix en Provence, Cedex, France
William Marois
Professor of Economics
University of Orleans, Faculty of Economics
BP 6749, 45067 Orleans, Cedex 2, France
Antoine Soubeyran
Professor of Economics
University of Aix-Marseille II, Faculty of Economics
13621 Aix en Provence, Cedex, France

ISBN-13: 978-3-540-50322-4
DOl: 10.1007/978-3-642-74104-3

e-ISBN-13: 978-3-642-74104-3

This work Is subject to copyright All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, re-lJse of Illustrations, rocitation, broadcasting, reproduction on microfilms or in other ways, and storage In data banks. Duplication of this
publication or parts thereof is only permitted under the provisions of the German Copyright Law of
September g, 1965, In its version of June 24, 1985, and a copyright fee must always be paid. Violations
fall under the prosecution act of the German Copyright Law.

@ Springer-Verlag Berlin Heidelberg 1988

2142/3140 - 54321 0

INTRODUCTION
Didier LAUSSEL, William MAROIS and Antoine SOUBEYRAN

The papers collected in this volume were presented at

the "4th International Conference on Monetary Economics and
Banking" held in Aix-en-Provence on June 1987 and organized by
the C.E.F.I. (Center for International Economics and Finance of
the University of Aix-Marseille II) and the GRECO "Monnaie et
Financement" of the C.N.R.S.
(National Center for Scientific
Research). They concern two main topics: monetary theory and
monetary policy.
In the first one, the contributions provide
new insights
in some
important problems
like rational
expectations, monetary optimizing models or portfolio choices.
In the second one, almost all the texts are devoted to the
game-theoretic approach of monetary policy which is a research
area mainly developped since about ten years.
I. MONETARY THEORY
In their well known article on "Recent Developments
in Monetary Theory", R. BARRO and S. FISCHER (1976) discussed
seven main topics among which "the theory of money demand",
"money,
inflation
and growth",
"disequilibrium
theory",
"rational expectations and the Phillips Curve".
Twelve years after, papers of this volume give some
new results in these areas or explore new paths of research
like a monetary theory of an innovative process of production
or the application of the theory of contracts to financial
problems.
BARRO and FISCHER were concluding : "It is clear from
this paper that many of the major outstanding questions in
monetary theory and macroeconomics more generally are related
to the Phillips Curve and concern the short-run dynamics and
costs of inflation and unemployment".
This is the heart of J. STEIN's paper. He examines
the well known controversies of the 60's and 70's between
keynesians, monetarists and new classics in the light of the
recent developments of new financial instruments in speculative

IV

markets.

The

central to

formalization

of

expectations

hypotheses

is

understand those controversies and consequently the

role of monetary policy.

STEIN compares

two hypotheses

: the

Muth

Rational

Expectations (MRE) and the Asymptotically Rational Expectations

(ARE). He

develops a macro model of inflation and unemployment

based on

the second

describe the
micro

hypothesis and

post 1980

markets,

hypothesis is

he

shows its best ability to

deceleration of

also

shows

inflation. Turning to

empirically

that

the

ARE

consistent with the microeconomic datas on three

markets : the nominal interest rates and inflation, the pricing

in the

futures markets and the forward foreign exchange rates.

His demonstration
i.e. the
on the

relies on

the existence

differences between
true model

and the

of bayesian errors

the objective expectations based

subjective one held by the market.

Those errors must be equal to zero for the MRE hypothesis.

So the
micro datas

concept of

and to

ARE appears to be consistent with

provide the

best macroeconomic support to

explain the relationships between inflation and unemployment.

Those relationships
work but

in a

are also

neo-keynesian and

HICKS (1974),

he develops

two sectors

intermediate

disequilibrium tradition. As

a 'Fix-price-Flex-price" model with

good and final good. The wage rate

is perfectly

indexed on

the price

flexible but

the

of

price

sluggish process
As in

the aim of J.P. AZAM's

the

of adjustment

of the final good which is

intermediate

good

shows

depending on costs and demand.

the disequilibrium theory he proves the existence of two

regimes (a

glut and

between inflation
employment when

a shortage

regime) and

and unemployment
steady state

with a

finds a relation
maximum

inflation is

nil.

level

of

Inflation and

deflation have then real costs in terms of jobs.

The ideas
of M.

AMENDOLA and

restore the
the heart
an

of HICKS

(1974) constitute also one basis

J.L. GAFFARD's

analysis of

contribution. They want to

changes in the productive capacity at

of monetary theory and develop a sequential model of

economy

which

highlights

technical intertemporal
production.
is essential

In

the

time

structure

complementarities of

the

and

process

the
of

this model, the character of liquidity of money

as in HICKS (1974)

: an accrued demand for liquid

v
assets can

be understood

formulating short
on its

either as the

of a mistake in

~esult

term expectations and hence the economy stay

"routine path"

structural change

or

in the

as the

signal of

process of

the search of a

production and hence the

economy goes on an innovative path.

The problem
following an
the output

is then

to

analyse

innovative choice
of the

early phase

the

three

phases

the preparatory phase before

new process has appeared on the market, the

when the

output of

both routine

and

innovative

processes is on the market, the late phase when only the output
of the innovative process is still on the market.
At the

beginning of all this, money matters and then

is essential to understand a process of change in an economy.

In their
BROCK model
SIDRAUSKI

survey, BARRO

as one
model.

In

separable utility

and FISCHER (1976) cited the

interesting and promising extension of the

monetary

optimizing

function, BROCK

(1974)

models

and

using

other

authors

provide some contradictory results on the possibility of ruling

out

divergent

price

paths

hyperinflationary).

In

the case

function, OBSTFELD

(1984) has

(hyperdeflationary

of a

non

shown that

or

separable

utility

price paths

may be

stable.
D. LAUSSEL
uniqueness of

SOUBEYRAN examine

equilibrium in

separable utility
demonstrate a

and A.

a general

function and

the problem of

framework with

a non

non zero monetary growth. They

simple condition to rule out stable and cyclical

price paths and give a set .of necessary and a set of sufficient
conditions

to

different and

have

divergent

price

complements-substitutes relationships
consumption is
the speed

paths.

They

provide

more general results than in the literature. The

between real

money

and

here important as, for long-run hyperdeflation,

of decrease

of the implicit rate of return on money

when real balances tend to infinity.

The next

three papers

are more

concerned with

the

detention of money or credit. J.H. ROUSSEAU's paper examines an

important question of macroeconomic theory : why agents want to
hold money
CLOWER

when there are bonds yielding a positive interest?

(1967)

constraints.

gave

an

answer

based

on

cash-in-advance

VI

J.M.
where the

ROUSSEAU

elaborates

choice between

theoretical

framework

assets is made on the basis of their

characteristics, as in the seminal work of LANCASTER (1966) for

consumption goods.
and the

The portfolio is composed of money and bond

two characteristics,

liquidity function

and store of

value, are random variables.

In some

cases, money

could be

totally dominated by the bond because,

held although

it is

in this random world,

it

represents a protection against risk.

X. FREIXAS
market, which
studying the
features of

examines the

is an

characteristics of
this market

asymmetry in

functionning of

important matter
a

are an

the credit

for monetary theory, by

debt

contract.

The

imperfect information

the relationships

between the

main
and an

borrower and

the

lender.
In a

debt contract

bankruptcy of
either a

the firm

there

In both

some

probability

of

with two consequences for the lender:

partial reimbursement

the firm.

is

or the

cases FREIXAS

possibility to control

examines the optimality of a

debt contract (linked with cash-flows).

Moreover,
information,the
i.e. the

in

repeated

firms must

effect on

game

consider the

one player

imperfect

with

reputation

problem

strategy of taking account the

consequences of his actions into the other player beliefs.

The theory
understanding of
interest when

of contracts provides new insights in the

the

one

instruments with

credit

knows

market

the

the financial

which

actual

are

of

development

deregulation or

peculiar
of

credit

the amount of

the international credit.

In his
the ,dynamic

contribution H. ANTAO provides an analysis of

relationships

monetary aggregate
popular VAR

between

the

counterparts

of

the

for the portuguese economy. He uses the now

model developped

initially in

the SIMS's

(1980)

seminal work. This empirical procedure does not need to have "a
priori" strong
variables and

differences between
imposes less

econometric models.
box without

endogenous

constraints

than

and
in

exogenous
traditional

One can discuss this procedure as a "black

theory" but

it gives some preliminary interesting

results concerning the relations between the studied variables,

VII
in ANTAO'

study the

that. in

sources of

Portugal. during

out effect

was strong

reserves were

monetary creation.

He

shows

the 1965-1986 period. the crowding-

but does

not influenced

not last

by the

and

that

external

credit component

of the

counterparts.
II. MONETARY POLICY CAMES
In the last ten years the analysis of economic policy
has

made

an

concepts.

increasing

There

were

developments which
KYDLAND and

use

of

two

game

basic

were initiated

PRESCOTT (1977)

theoretic

tools

motivations

for

and
such

by the pioneering papers of

on one

hand. of

HAMADA 1974).

(1976)1 on the other.

If the

private sector

has rational expectations. as

it has become fashionable to suppose.

it is likely to react not

only to the present rate of growth of the money supply but also
to its

future value

as announced

initially by

the

monetary

authorities. The Central Bank has then an incentive to use pure

"announcement effects"
agents' behavior.
problem" first
to. will

This gives

not

announced since

if

PRESCOTT (1977)

monetary

: the

policy

initially

for its

announcement

implementation date.

As argued by

one may

an open-loop

private

reoptimize at any time

partly designed

its planned

the

the "time inconsistency

to

the

monetary authority

framework of

manipulate

rise to

allowed

SALMON (1985)

between the

to

KYDLAND and

implement

it was

effects before
MILLER and

order

analyzed by

monetary authority.
t >

in

analyze these relationships

and the

private sector in the

Stackelberg dynamic

game (see also

WHITMAN (1986.
More recently (see TABELLINI (1986
between monetary
been modelled
Treasury"

and fiscal

as a

each

policies in the same country have

game between

endowed

the interactions

with

the "Central
a

different

Bank" and "the

quadratic

loss

function.
In the
no more
place in

increasingly open economies of nowadays it is

possible to

model economic policymaking as if it took

autarkic countries. Variations of the money supply in

one country

affects the

rest of

the

world

through

various

VIII

channels which
(1983):

are for instance analyzed in CANZONERI and GRAY

these are the external effects of monetary policy. On

the other

hand the

current account)
policymakers

same variables

may enter

in

way

simultaneous attainment
modelling of

such

as

of their

international

coordination in

(the exchange

the utility
it

rate or the

functions of
makes

several

impossible

the

purposes. This motivated the

economic

policy

conflict

and/or

a game theoretic framework initiated in a non-

cooperative and static framework by HAMADA ((1974,(1976.

Papers in

this part

distinguished with
definition of
dynamic or

analyze

respect to

which

are

their basic

the

repeated) of

~Q1jcy

players,

sa~

and may be

constituents: the

the

nature

(static,

the game, the solution concept (Nash,

Stackelberg, Cooperative) ...

K. CLINTON
literature, both
questions of
of the

and J.C. CHOURAQUI discuss and review the

theoretical and

empirical, related

to

both

time-consistency of optimal policy and reputation

monetary authority

and of

policy coordination between

countries : they try to see how much the efficiency of monetary

policy may

be increased

They conclude

through

that coordination

inconsistency-credibility

problem

PRESCOTT

closed

(1977)

in

inconsistent policies

international
may help

coordination.

to solve

raised

by

economy

the time-

KYDLAND

context.

and

If

time

are cooperatively designed and announced

by countries and if the gains from cooperation are large enough

(and hence

the costs

rationaly expect
that

these

of reneging

are large), the private may

that these policies will be implemented. Note

conclusions

do

not

contradict

SALMON's

(1987)

results according to which international policy cooperation may

be counterproductive

since SALMON constrained countries to use

only time-consistent

policies:

CHOURAQUI's argument

is precisely

cooperative policies

may be credible because reneging on these

policies implies
solution (one

the essence
that time

simultaneously reverting

has then

of

CLINTON

and

inconsistent but

to

non-cooperative

to compare countries welfare under the

cooperative time-inconsistent

policies to

their welfare under

the non-cooperative time-consistent ones).

Both CARRARO's
with

the

interactions

paper and BASAR and SALMON's one deal

and

conflict

between

the

monetary

IX

the private sector,

i . e. the so called
policymaker and
"Monetary Policy Game", and with the question first raised by
KYDLAND and PRESCOTT (1977)
: the optimal monetary policy is
time-inconsistent while the time-consistent is suboptimal.
C. CARRARa tries to solve this problem by transposing
in the framework of CUKIERMAN's (1986) model a repeated game
approach in order to prove an analogous for monetary policy of
the Folk Theorem in the Industrial Organization literature.
However he has to study a hierarchical (Stackelberg) repeated
game whereas
the results in the industrial organization
literature were conceived for symetric (Nash) players. Hence
the results he obtains are new in this respect and not only
with regard to the monetary policy field where the repeated
game approach is first applied. CARRARa proves that,
in
infinite games where one considers the possibility of using
both continuous and discontinuous state dependent Stackel~erg
Trigger strategies, all equilibria are possible, from the zero
growth inflationary one to the zero inflation cum growth one.
He shows that this Folk Theorem is eq~ally true for finite
Monetary Policy
Games with incomplete information and/or
bounded rationality. A sequential equilibrium concept is also
used to show the indeterminacy of equilibrium in the finite
Monetary Policy Game.
T. BASAR and M. SALMON are also concerned with the
"Monetary Policy Game" . Their framework is borrowed fr.om a
in which there is
paper by CUKIERMAN and MELTZER (1986)
asymmetric information between the private sector and the
monetary authority. Contrary to CUKIERMAN and MELTZER, BASAR
and SALMON solve the optimization problem without imposing
certainty equivalence. They find that the policymaker may
deliberately act so as to lower its credibility and that this
effect is explicitly accounted for when considering the effect
of its decision today on the private sector's forecasting
problem. Even in the certainty equivalent case BASAR and SALMON
show that the optimal policy does not imply an inflationary
bias
the reason why CUKIERMAN and MELTZER found an opposite
result is shown to be their use of an assumed (misspecified)
model of
how the private sector forms its expectations
regarding monetary policy.

x
A. LAVIGNE

and P.

interactions between
attempt to
Each of

monetary and

reduce the

these

WAECHTER

has

the

strategic

fiscal authorities in their

public debt

authorities

analyze

stock within

different

one country.

quadratic

loss

function

both want to minimize deviations of the public debt

stock from

zero but,

stabilize the

in addition,

monetary growth

minimize deviations
conflict between

of

the

the Central

while
budget

the

Bank wants to

Treasury

deficit

tries

from

zero.

to
The

these two authorities is studied in a dynamic

game framework

first without

private sector

and then

analyzing the

modelling a

behavior

of

private sector

characterized by

its demand

for money

and

behavior. Closed

Loop Nash,

Open Loop

Nash

its

the

which is

anticipative

and

Cooperative

equilibria are studied. The main results are as follows: under

different assumptions the stock of public debt is reduced. This
reduction is

accelerated with precommitment and even more with

cooperation.

When

explicitly

sector's behavior

slows down

accounted

for,

the

private

the reduction of the public debt

stock.
AOKI, GIRARDIN

and PICHT's papers are concerned with

international economic policy coordination.

H. AOKI's
monetary

rules

paper shows how to implement decentralized

in

model

of

three

large

interdependent

economies where policy actions not only affect each own economy
but also

affect

other

decompose policy
shocks when

economies.

spillover or

other countries
AOKI exhibits
functions that

The

main

problem

is

to

objectives, dynamic interactions and external

can't be

such

cross effect

actions

on

ignored as for small open economies.

recursive

allows for

of policy

decomposition

decentralized

policy

of

objective

actions.

He

adopts his average-difference coordinate system: the objective

choosen is

the weighted

variances. The

sum of

individual

author decomposes

this

country's

objective

output

into

three

other individual country's objectives, one country, the largest

one, minimize

the world

output variance

and the

two

others

minimize their d-output variance; d-output of country j is the

difference between
output). After
equation for

its output

and the

world output

(the

a-

building the objectives equations and the state

each country

he exhibits the dynamic multipliers

XI

summarizing all the information regarding how exogenous changes

in the

monetary instruments

or exogenous

disturbances affect

the output of each country.

The end

of the

paper analyses empirical evidence of

interdependence between the USA, West Germany and Japan.

Interdependence,
investigation. is
author estimates
dollar period

through

aim of

current account

for the

determinant is
by the

seen

also the

empirical
paper.

equations over the

seven main

the anticipated

an

E. GIRARDIN's
OECD countries.

The

floating
The

major

fiscal impulse measured either

surplus of the preceding period or by the estimation of

a reaction function of fiscal policy. Those functions depend on

exchange rates.

inflation, unemployment.

real rate

of growth

and the share of public debt in GOP.

In

the

current

foreign impulses

are both

account

equations,

considered and

domestic

and

GIRARDIN shows

the

influence of the US defioit on the current account of all other

countries whereas
affect foreign

the non

American budget

current accounts.

surpluses

do

not

In these equations, monetary

impulses are less significant and for only three countries.

This study
countries and

stresses the interdependence between OECD

the importance

of US

fiscal

policies

in

the

determination of current accounts during the eighties.

During

this

period.

other

countries,

the

LDC

countries, were confronted to their growing debt problem.

H. PICHT's
repudiation
(1981)

problem

when the

paper) debt

paper is
first

about the
studied

by

debt-with-potentialEATON

and

GERSOWITZ

debtor is a foreign country (LDC in PICHT's

servicing is

not exogenously

enforceable.

PICHT

shows that the relationships between creditors and debtors must

be analyzed

in a game theoretic framework as a principal-agent

problem. He

tries to

techniques the

establish with

the 1980's

: first

the 1970's

had changed

finance

in

favor

relative share

the help

of statistical

following explanation of the LDC debt crisis of

the emergence
of

of cross-default clauses in

the relative
debt

of direct

and,

price of debt and equity

consequently,

investment in

reduced

the

external LDC finance.

Then PICHT goes on to argue that this has led to a reduction of

economic performances

of the

capital importing countries, and

XII
especially of
itself was
lenders

the return

the cause

had

strong

strategies (voluntary

on invested

of the

capital. This reduction

"debt crisis" : both debtors and

to

incentives

lending became

choose

non-cooperative

marginal

on

one

hand,

recurrent reschedulings were chosen, on the other hand).

In publishing
some additional

all those

research in

texts, we

hope to raise up

monetary theory

and policy along

the paths of recent developments outlined here.

We are
the organization

very grateful
of the

to D. PEGUIN and J. JOYEUX for

Conference and

typing all the texts of this volume.

to C.

FONTENEAU

for

REFERENCES
BARRO R.J.

and S.

FISCHER (1976)

Monetary Theory",

- "Recent

Developments

in

Journal of Monetary Economics, 2, pp.

133-167.
BROCK, V.A.

(1974) -

"Money and Growth: the Case of Long-Run

Perfect-Foresight",

International

Economic Review,

IS,

pp. 750-77.
CANZONERI, H.B.
Policy

and J.A. GRAY (1983) - "Two Essays on Monetary

in

an

Interdependent

Finance Discussion

World",

International

Papers, n0219, Board of Governors of

the Federal Reserve System, February.

CLOVER, R.

(1967) - " A Reconsideration of the Microfoundations

of Monetary Theory", Western Economic Journal, December.

CUKIERHAN, A.

(1986) - "Central Bank Behavior and Credibility:

Some Recent

Theoretical Developments",

Federal Reserve

Bank of St. Louis Review, pp. 5-17.

CUKIERHAN, A. and A.H. HELTZER (1986) - "A Theory of Ambiguity,
Credibility

and

under

Inflation

Discretion

and

Asymmetric Information", Econometrica, vol. 54, nOS, pp.

1099-1128.
EATON, J.

and H.

GER50VITZ

Repudiation

"Debt

(1981)

Theoretical

and

with

Empirical

Potential
Analysis",

Review of Economic Studies, vol. 48(2), pp. 289-309.

HAHADA, K.

(1974) - "Alternative Exchange Rate Systems and the

Interdependence

Monetary

International

National

System,

(R.Z. Aliber ed.), University of Chicago Press

(1976)

the

in

Policies

K.

and

Policies",

Monetary

Ch i cago, pp.
HAMADA,

of

Financial

13-33.
A

Strategic

Analysis

of

Monetary

XIV

Interdependence", Journal of Political Economy, vol. 84,

n04, August, pp. 677-700.
HICKS, J.R.

(1974) - The Crisis in Keynesian Economics, Oxford,

Blackwell.
KYDLAND, P.W.

and E.C.

Discretion
Journal of

PRESCOTT (1977)
the

Inconsistency

Potitical Economy,

- "Rules
of

Rather than

Optimal

85, n03,

Plans",

June, pp. 473-

491.
LANCASTER, K.

(1966) -

"A New

Approach to

Consumer Theory",

Journal of Political Economy, 74, pp. 132-157.

HILLER, H.

and H. SALHON (1985) - "Policy Coordination and the

Time

Inconsistency

of

Optimal

Policy

in

an

qpen

Economy", Economic Journal, supplement, pp. 124-135.

OBSTPELD, H.

(1984) - "Multiple Stable Equilibria in a Perfect-

Foresight Model", Econometrica, 52, pp. 223-28.

SALHON, H.

(1986) -

"Rational Consumption Behaviour", Working

Paper, Department of Economics, University of Warwick.

SIHS, C.
(1980) "Macroeconomics and Reality", Econometrica,
vol. 48, nOlo
TABELLINI, G.

(1986) -

"Money, Debt and Deficits in a Dynamic

Game", Journal of Economic Dynamics and Control, 10, pp.

427-442.
WHITHAN, C.

(1986) -

"Analytical Policy Design under Rational

Expectations", Econometrica, vol. 54, n06.

LIST OF AUTHORS
AHENDOLA Hario
ANTAO Hario
AOKI Hasanao
AZAH Jean-Paul
BASAR Tamer
CARRARO Carlo
CHOURAQUI Jean-Claude
CLINTON Kevin
FREIXAS Xavier
GAFFARD Jean-Luc
GIRARDIN Eric
LAUSSEL Didier
LAVIGNE Anne
HAROIS William
PICHT Hartmut
ROUSSEAU Jean-Harie
SALHON Hark
SOUBEYRAN Antoine
STEIN Jerome
WAECHTER Philippe

University of Roma
University of Lisboa
University of Cal ifornia
University of Clermont-Ferrand I
University of Illinois
University of Venice
OECD
OECD
University of Toulouse I
University of Nice
University of Bordeaux I
University of Aix-Marseille II
University of Paris X-Nanterre
University of Orleans
Indiana University
University of Brest
University of Warwick
University of Aix-Marseille II
Brown University
University of Paris I

CONTENTS

PART I - MONETARY THEORY

Speculative Markets and Macroeconomic Controversy

J. L. STEIN
A Neo-Structuralist Model of Inflation and Unemployment

39

J.P. AZAM
Towards a Monetary Theory of a Process of Change

55

M. AMENDOLA and J.L. GAFFARD

Ruling

out Multiple Equilibrium Paths in Monetary Optimizing

Models: Necessary and Sufficient Conditions

D. LAUSSEL and A. SOUBEYRAN
The

73

Optimal Sharing Money-Bond in the Portfolio

the Random

Characteristics Approach

93

J.M. ROUSSEAU
Debt Contract under Imperfect Information

a Survey

X. FREIXAS
Causal

Relations Among

103

the Sources

of Money

Supply

the

Portuguese Case
M. ANTAO

PART II - MONETARY POLICY GAMES

117

149

Monetary Policy Credibility and Coordination

K. CLINTON and J.C. CHOURAQUI

151

A Folk Theorem of Monetary Policy

C. CARRARO

173

XVIII

On

the Convergence

of Beliefs

and

Policy

to

Rational

Expectations Equilibrium in a Dual Policy Problem

T. HASAR and H. SALHON
Public

Debt. Inflation

207

and the

Coordination of

Fiscal and

Honetary Policies
A. LAVIGNE and P. VAECHTER
Decentralized

Monetary Rules

in a

225
Three-Country Model

and

Time Series Evidence of Structural Dependence

H. AOKI
Fiscal

Expectations and

269
Current Account Surplus of the Main

OECD Countries
E. GIRARDIN
The

Political Economy

297
of Debt Repudiation and Expropriation

in LDCs
H.R. PICHT

329

Part 1 : MONETARY THEORY

SPECULATIVE MARKETS AND MACROECONOMIC CONTROVERSY

Jerome L. STEIN
Department of Economics, Brown University
Providence, Rhode Island 02912
During the
speculative

characterized by
the controversy
found

past decade

markets

to

and

very sharp
is that

be

the areas of macroeconomics,

international

finance

controversy. The

the "conventional

inconsistent

with

the

have

been

main reason for

wisdom"

has

empirical

been

evidence.

Consequently, the challenge is to formulate a more satisfactory

theory. The basic problems in all these areas are interrelated,
because of the crucial role of anticipations hypotheses. Recent
work in

speculative markets

futures markets

in

general

in commodities,

and,

in

particular

financial instruments,

stock

indexes and foreign exchange, has profound implications for the

microeconomic foundations of macroeconomic controversy.
The
purpo~e

of this paper is to show how the study of microeconomic

markets can

evaluate the

controversy surrounding

the various

macroeconomic models.
First, the
Monetarists,
discussed

macroeconomic controversies

Keynesians

and

evaluated

disagreements;

and

New

Classical

determine

results

of

the

the studies

the

Economists
sources

simultaneous

are summarized.

derived from

are brought

Second:

of
tests

are
the
of

the theories

of speculative markets

to bear upon the macro-economic controversies. The

evidence from

speculative markets,

part, concerning
with the

to

the

alternative hypotheses
and evidence

and

between

the anticipations

discussed
hypotheses,

in

the

is

second

consistent

empirical results on the macro level discussed in the

first part.
I. MACROECONOMIC THEORY AND POLICY
A. The Polarization of Macroeconomics
Disenchantment
during the

post 1968

declined, the

rate of

with

Keynesian

economics

developed

period when the rate of growth of output

unemployment

rose,

and

the

rate

of

inflation increased. This paradox, called stagflation, was

inconsistent with the tenet of Keynesian economics that
cyclical movements in prices and output relative to their
respective trends
are
positively
correlated.
Keynesian
economics assumed that as long as there was excess capacity, or
as long as unemployment was in excess of an equilibrium rate,
demand management could be used to increase the growth rate and
lower
the unemployment rate without increasing the rate of
inflation. The
standard tools of demand management were
monetary and fiscal policy.
The Keynesian view (Tobin; Modigliani-Papademos) is
that the change in the rate of inflation pitt) - pi(t-l) from
period t-l to period t depends negatively upon the deviation
u(t-l) = U(t-l) - Ue of the unemployment rate at the initial
date from the "natural" or equilibrium rate of unemployment Ue,
plus nonsystematic factors v(t). This view is expressed, as
equation (1).
(1) pitt) - pi(t-l)

-hu(t-l) + v(t).

pi
rate of inflation
u = U - Ue = unemployment rate less
its equilibrium value ; v(t) = non-systematic variable with
zero mean.
The period of stagflation was inconsistent with the
Keynesian point
of view. A search occurred for a more
satisfactory theory of macroeconomics which could explain the
paradox of stagflation and the observed economic phenomena.
The stagflation paradox has not only been observed in
the United States during the 1970's, but also in other
countries. In Germany, the "Wirtschaftstwunder" of the 1960's
turned into the "Wirtschaftsfrage" of the 1980's where sluggish
output growth
and record
unemployment rates
have been
accompanied by moderate inflation.
In Argentina, the rate of
inflation rose drastically from the first half of the 1970's to
the second half of the decade but the growth rate of output
declined.
The New Classical Economics (NeE) developed as the
total rejection of Keynesian economics. The Keynesians claimed
that their
demand management
policy contributed to the

5
obsolescence of the business cycle and successfully eliminated
the gap between full employment (potential) output and actual
output. The NCE argued just the opposite: there is no way that
the monetary authority can follow a systematic activist policy
that would achieve a rate of output that is, on average, higher
over the business cycle than would occur if the monetary
authority did not respond to varying business conditions. Their
main tenet is that the unemployment rate or rate of growth of
real output is insensitive to demand management policy choi~es.
(Lucas; Sargent and Wallace).
Crucial

to

their

analysis

is

the

MUTH

RATIONAL

EXPECTATIONS HYPOTHESIS (MRE), which states that the subjective

(or market) anticipation of a variable is equal to the
objective expectation based upon the true model. This means
that forecast errors made by the market have zero expectations
and are independent of any variables whose values are available
at the time the forecast is made. This is a very controversial
hypothesis. Their proponents take the MRE hypothesis as a
postulate akin to expected utility maximization. Others, such
as Alan Blinder argue that: " ... the weight of evidence -both
from directly
observed expectations
and
from
indirect
statistical tests of rationality (usually in conjunction with
some other hypothesis)- is overwhelmingly against the RE
(rational expectations) hypothesis".
The microeconomic analysis and evidence discussed
below is directly concerned with the MRE relative to an
alternative that I call ASYMPTOTICALLY RATIONAL EXPECTATIONS
(ARE).
The driving force in producing variations in the
growth of output relative to a long run tr~nd, or in deviations
of the unemployment rate from the "natural rate",
is the
difference between the actual and what the NCE call the
"rationally anticipated" rate of price change. The rationally
anticipated rate of price change is the rate that would be
predicted by the market if they knew the true model of the
economy and of the behaviour rule of the monetary authorities.
In their

framework, the

rate of unanticipated price

change is proportional to "unanticipated" money growth (Barro).

The forecast errors would be random variables with zero

expectations, because
publicly

the market

available

is alleged

information

to be

efficiently.

using all

The

monetary

authority can control the actual rate of inflation, but not the
so the monetary policy cannot be used for

unanticipated rate
demand management.

Milton Friedman
NCE.

has never

been a

supporter of

the

In the MONETARY HISTORY OF THE UNITED STATES, Friedman and

Schwartz

stated

their

theme

main

concerning

the

Great

Contraction as follows.
"The

monetary

consequence of
factor which

collapse

other forces,
exerted a

events. The

was

powerful influence

failure of

the

no~

inescapable

but rather a largely independent

the Federal

on the

course

of

Reserve System to prevent

the collapse reflected not the impotence of monetary policy but

rather

the

particular

authorities and,

policies

in smaller

followed

degree, the

by

the

monetary

particular

monetary

arrangements in existence ... (D)ifferent and feasible actions by

the monetary
the stock

authorities could

of money-

have prevented

indeed, could

the decline in

have produced

almost

any

desired increase in the money stock ... Prevention or moderation

of

th'e

dec line

substitution of

in

the

stock

of

monetary expansion,

contraction's severity

and almost

money,
could

let

have

as certainly

alone

the

reduced

the

its duration.

The contraction might still have been relatively severe. But it

is hardly

conceivable that money income could have declined by

over one-half

and prices

by over

one third

in the course of

four years if there had been no decline in the stock of money".

(1963 : 300-301).
The NCE

unemployment

rate,

or

growth

of

output,

equation has the form:

(2) u(t)

= au(t-l)

where mu(t)
is the

- b[mu(t) - Emu(t;t-l)]

is the actual rate of money growth, and Emu(t;t-l)

rationally expected

rate; conditional upon information

available when the forecasts are made.

The NCE
decline in
severity of
in the

disagree with

the stock

of money,

Friedman, and
per se,

argue that the

was irrelevant; the

the Great Depression did not result from a decline

stock of

money, but just depended upon the slower than

anticipated growth of money. As will be noted later, the NCE

models fail completely to explain the severity of the Great
Contraction.
The Muth
Rational Expectations
point of
view
underlying the NCE has logical beauty, since anticipations are
endogenous. This makes it an extremely attractive research
strategy. The Muth Rational Expectations hypothesis was alleged
by the NCE to be the complement to utility maximization. This
is an arbitrary assumption. One must show that its beauty and
simplicity are grounded in micro-economic theory and also that
it is consistent with evidence.
The polarization of the profession into the Keynesian
and NCE camps produced a landscape where Milton Friedman seemed
like a moderate between these poles.
What I call the Monetarist position is in between
poles of
Keynesianism and the NCE. The main Monetarist
propositions are as follows. The trend rate of growth of the
money stock, or the monetary base,
is the only systematic
factor determining the trend rate of inflation. A rise in the
trend rate of monetary expansion temporarily reduces the
unemployment rate, or temporarily raises the growth of output,
and permanently raises the inflation rate. A deceleration of
the rate of growth of the monetary base, following several
years of high rates of monetary expansion, exerts its effects
directly upon the unemployment rate or rate of growth of
output. However, the rate of inflation is kept high for a while
by the inertia of past rates of monetary expansion. Contrary to
the NCE, there is a significant cost in terms of unemployment
and lost output to reducing the rate of inflation. To be sure,
other people who call themselves Monetarists have different
points of view. What I stated is my view of Monetarism. Milton
Friedman's position is partially, but not fully, contained in
these propositions.
The theories underlying each school of thought used
different variables and communication between them was most
limited. The econometric testing of hypotheses was done without
simultaneously comparing the three different points of view in
terms of the same set of data. It was no surprise that one
group was unimpressed with another group's econometrics. There

was no

consensus

amon~

economists which

theory could best be

used to evaluate policy.

The disagreements
the steady

state,

but

in macroeconomics

rather

the

did not concern

dynamics

steady

the unemployment

deviates from

its respective "natural rate". The three schools

agree that
is at

in the

rate or

between

states, where

steady state:

its natural

rate, which

policy

(2) the

rate of

monetary

expansion

growth rate of output

(1) the rate of unemployment

is not

affected

inflation is

less

the

long

by

equal to
run

the rate of

growth

rate.

disagreements concern the path to the steady state.

an excess

monetary
The

If there is

unemployment u(t) > 0, can its convergence to u=O be

accelerated by

monetary policy

? What

will be

the resulting

effects upon the trajectory of the rate of inflation?

Figure 1
and will

facilitate an

upper half

of the

diagram. On
y(t) to

simple way

axis is

figure is

the horizontal

demand schedule
is

the

the

familiar

equations.
aggregate

The

demand

axis is the ratio of actual output

f[k(t)l, where

effective labor.

real aggregate

to see the disagreements,

understanding of

capacity output

capital to

M(t)

is a

k(t) is

the ratio of

On the upper part of the vertical

demand. The

height of

the

aggregate

depends upon a parameter M(t)/p*(t;t-1), where

the money stock at time "t" and p*(t;t-1) is the price

level anticipated
are held

to prevail

in the

at time

previous period

"t", when anticipations

t-1. The higher is M/p*, the

higher will be aggregate demand at any level of output.

There is
curve.

Its

better way

height

depends

to look at the aggregate demand

upon

the

real

measured in wage units, M(t)/W(t) where W(t)

But the

money wage

factors

(a)

the

anticipated price
the height
by more

W(t) is

of the

than W(t),

set at

unemployment

value

of

money,

is the money wage.

time (t-l) based upon two

rate

level p*(t;t-1).

u(t-l)

and

Monetary policy

(b)

the

can raise

aggregate demand curve if it can raise M(t)

where the

latter depends

upon p*(t;t-1),

given the initial unemployment rate.

The

lower

unemployment rate

part
U(t) to

of

the

the ratio

diagram
of

actual

relates
to

the

capacity

output. Full employment occurs when this ratio is unity, or the

unemployment rate is at the "natural" rate Ue.

Suppose that
curve [M/p]
at Oy.

the aggregate

demand is

described

by

so that the ratio of actual to capacity output is

An increase

aggregate demand

in the

to curve

money supply
[M/p]'

if

will

the

only

increase

anticipated

price

level p. does not increase by the same proportion.

The NCE claim that the rationally anticipated rate of
inflation

is

the

rationally

expansion. Therefore,
shift upwards,
if the

the aggregate

as a

rise in

The MRE

rate

demand

of

monetary

curve

will

only

result of the rise of monetary expansion,

the money supply was greater than anticipated.

hypothesis states

variables with

anticipated

that the forecast errors are random

zero expectations

serial correlation).

and no

Consequently, on

structure

average,

(such

as

rate

of

the

anticipated inflation will equal the rate of monetary expansion

; and

monetary policy

aggregate demand
the NCE

will not

curve. That

is unanticipated

which is

be successful in shifting the

is why

the crucial variable,for

money growth

[mu(t) -

Emu(t;t-1)],

a serially uncorrelated term with a zero expectation,

and is independent of any information available at t-1.

The Keynesians
variable which

p(t;t-1)

is

"sticky"

does not respond quickly to changes in monetary

variables. Therefore,
change M/p

argue that

monetary policy,

and shift

the aggregate

which changes

M,

can

demand curve. Keynesians

use a variety of arguments to explain the stickiness of prices,

but these

argumentsare

not

based

upon

explicit

optimizing

models and so fail to convince those who disagree with them.

B. Research Strategy to Resolve the Issues
This unsatisfactory

state of

affairs induced

me to

write MONETARIST, KEYNESIAN AND NEW CLASSICAL ECONOMICS (OXFORD

: BLACKWELL,

1982). The

general macro-dynamic
schools

of

thought

as

parameter specifications.
by the
stock of

three schools
capital and

research strategy

model which
special

was to

can imply
cases,

develop

depending

upon

that each

the

The model is broader than those used

of thought since prices, quantities, the

the stock

of assets are endogenous. The

macro model is part of a growth model in a monetary economy.

is shown

any of the three

school of

thought is

It

a special case of a

10

general model, and that the disagreement among the three

schools can be resolved by testing alternative statistical
hypotheses concerning parameter specifications.
The
Monetarist
position
relies
upon
the
ASYMPTOTICALLY RATIONAL EXPECTATIONS (ARE) hypothesis, which
will be
discussed in detail in part II. The effective
anticipated rate of inflation [p*(t;t-1)
p(t-1)l/p(t-1),
which affects the growth of nominal wages [WIt) - W(t-1)l/W(t1) with a unit coefficient, is the composition of several
functions. There are frictions in each link, thereby producing
a differential (lagged) response of the effective anticipated
rate of inflation to the current rate of monetary expansion.
Consequently, p* changes by a smaller proportion than does M ;
and this proportion depends upon the characteristics of an
economy, in a specific way.
[WIt) - W(t-1)l/W(t-1)

<--

E(pi)

<--

E(mu)

<--

mu(t-1)

First: there is the connection between the current

rate of monetary expansion mu(t-1) and the average or trend
rate of monetary expansion E(mu) that the public believes will
be pursued by the monetary authorities. This lag is often
referred to as the credibility problem. An examination of the
time series of mu the rate of monetary expansion shows that it
is highly volatile. Given the current mu relative to its
previous average rate, how long will it take for the public to
believe that there has been a change in the trend rate E(mu)?
This is the same problem that occurs in the quality
control literature. A factory produces a product with a given
machine
and the rate of defective products is E(mu). A new
machine is introduced, and suppose that it is heralded as a
significant improvement over the old machine in terms of
reliability. A sample of products is tested. Several batches
are run off the machine. The sample mean rate of defective
parts is mu and the variance of the rate of defective parts is
var(mu). How large a sample size is required for the firm to be
able to estimate the new rate of defective products with a
given degree of confidence? The sample size corresponds to the
number of quarters that the public will wait before its changes

11

its view

concerning the

new trend rate of monetary expansion.

The length of this lag is positively related to three factors:

(i) the

volatility of

the growth

of the

supply

(the

the rate

the desired

accuracy between the sample mean and the true mean

and (iii)

the level

its estimate

of defective

money

variance of

of confidence

size corresponds

Second, there
expected trend

required by the public in

economists have

to the

lag between

mu

and

is a lag between, E(pi) <-- E(mu), the

rate of

inflation E(pi) and the expected trend

monetary expansion

rate of

(ii)

(i.e., the narrower the confidence interval). The

required sample
E(mu) .

rate of

parts among batches),

E(mu), because economic agents and

different models

concerning how

quickly

the

inflation converges to the rate of monetary expansion.

The relation

between money

growth and inflation occurs in the

long run, but not in the short run.

This is

stressed by

Monetarists Milton
lags between

money and

simple relation
quarter and

the rate

speaks of

prices. They

between the

relation between
this well

the intellectual

Friedman who

variable

stress that -there is no

rate of

of inflation

money and

leader of the

long and

monetary expansion
next quarter.

this

Instead, the

prices refers to the long run. But

established empirical relation seems to be neglected

in the theoretical work done by the NeE, who argue (equation 2)

that:

on average,

and independent

of the

horizon, monetary

changes should be reflected not in output, but in prices.

Third is
betwee~

the

the lag

expected

anticipated rate
Wet).
risk

It

[Wet) -

trend

rate

of inflation

is shown

W(t-1)]/W(t-1) <--

used in

[Stein, 1982

of

inflation

E(pi)
and

setting nominal

the
wages

: 50-51] that, with a rise in

and risk aversion, the magnitude of the relation between

the price

expectation and the response by the firm in terms of

output or

prices is

reduced;

that is, the length of the lag

between the expectation and the action is increased.

All of
equation (3)

these three

whereby the

converges asymptotically

lags are

subsumed under the ARE

anticipated rate
to the

of

inflation

pi-

rate of monetary expansion mu

at a speed, described by coefficient "c" in equation (3).

12

pi
mu

percentage change in the certainty equivalent price p

percentage change in M
Figure 2

describes this

monetary expansion
of inflation
expansion;
affected by

of

much

as

the

rate

are obtained.

1 (parameterized

of

monetary

The aggregate

by M/p),

cannot be

monetary policy.

If "c" is finite, as described by

figure 2,

Monetarist and Keynesian results

curves in

are obtained

as

NCE results

in figure

the rate

If "c" is infinite, the expected rate

changes by
and the

demand curve
the two

rise.

equation. Let

whereby monetary

policy can

temporarily

affect

M/p.
The statistical
specification of

the

equations (4) and (5).

be labelled

a,b, ...

hypotheses implied

complex

dynamic

by my monetarist

structural

model

are

[In each equation, the coefficients will

but the coefficient "a" of one equation

is not necessarily the same as the "a" in another equation. The

relations among

the coefficients

are derived in Stein,

1982 :

chapter 31.
The first

states that

inflation between
upon the

years t

growth of

the change

and t-l

real balances

in

[pitt) [mu(t-l)

rate

of

pi(t-l)l,

monetary aggregate less the rate of inflation,

A strong

the

pi(t-l)l depends
the

in the year t-l.

version of the hypothesis is that : the change in the

rate of

inflation is

independent of

the unemployment rate in

the initial year. Contrary to the Keynesian view,

unemployment can

inflation and

both be high ; and high unemployment does not

imply that inflation will decelerate.

The logic
figures

and 2.

The anticipated
finite, as
in figure

equation can

be understood

from

Let the growth of the money supply increase.

rate of

inflation rises by less, since "c" is

described in

figure 2. As a result, parameter M/p

rises.

Keynesian excess
the shift

of this

in the

Say

demand of

that

output

was

initially

Oy.

AS is now produced, as a result of

aggregate demand

curve.

The

rise

in

the

Keynesian excess demand raises the rate of inflation above what

it was

initially, by

balances.

a multiple

"a" of

the growth

of

real

13

aggregate
demand

(M/p*)' (M/'Il) ,
/p*)

________

y
~

____

(M/W)

1
___________________
y/f(k)

Figure 1

rates of
monetary
expansion

anticipated and actual inflation

mu/2
Figure 2

T'

time

14

(4) pitt) - pi(t-I)

It has

a[mu(t-I) - pi(t-I)] + v(t)

been pointed

example, Benderly

and Zwick

restrictive equation
be obtained

real balances
the

several

(4a) which

strong Monetarist

the acceleration

by

authors

this

is

[for

an

unduly

; and basic Monetarist results could also

from equation

Keynesian and

out

I985b) that

of inflation

is a

synthesis of

equations. Equation

the

(4a) for

contains the previous growth of

and also the lagged unemployment rate. Call this

Monetarist equation.

(4a) pitt) - pi(t-I) = a[mu(t-I) - pi(t-I)] - bu(t-I) +v(t)

This statistical
fundamentally from
accelerate even

semi-reduced form

the Keynesian

if there

is

equation differs

view, because

high

rate

of

inflation,can
unemployment.

Stagflation becomes a possibility in this formulation,

The monetarist

unemployment (or

statistical semi-reduced
from the

NCE view.

form equation

growth
differs

Monetarist equation

of

output)

fundamentally

(5) states

that

the

unemployment rate deviation in year t depends upon its value in

year t-I

and the

growth of real balances in year t-I, defined

as mu(t-I) - pi(t-I).
(5) u(t)

au(t-I) - b[mu(t-I) - pi(t-I)] + v(t)

Since the
negatively related

growth rate
to the

of

the

change in

economy

the

G[y(t)]

unemployment

is
rate,

equation (5) can be written in terms of real GNP.

The logic of this unemployment or output equation can
be understood

from figures

depends upon

real unit

depends upon

the growth

price change.
initial

of the

The growth

unemployment

and

2.

labor cost

The

unemployment

rate

YIp, so its rate of change

nominal wage less the rate of

of the nominal wage depends upon the

rate

and

the

certainty

equivalent

expected rate of price change. It follows that the change in the

unemployment

rate

depends

negatively

upon

the

initial

15

unemployment

and

rate,

negatively

upon

the

unexpected

inflation.
Suppose that
same rate,

money and

there is

is unemployment

prices were

growing at

the

no unanticipated inflation but that there

in excess of the equilibrium rate u>O. That is

output is Oy. Then wages will grow at a slower rate than prices
and money.

The real

balances M/W

aggregate demand

curve in figure

aggregate demand

of AB

reduces the

in wage

unit rise, and the

rises. The resulting excess

leads to

an increase

in output which

unemployment rate. This effect is the au(t-1) term

in equation (5).
The second
balances,

is

term, which

explained as

expansion rise.

follows. Let

change in real

the rate

of monetary

According to figure 2, the rise in the rate of

monetary expansion
a smaller

contains the

raises the anticipated rate of inflation by

amount ("c"

is finite).

Therefore,

the

KeyneSian

excess demand

curve in figure 1 shifts up and produces a total

excess demand

of AB

rise at

a faster

at the initial level of output Oy. Prices

rate than

nominal wages,

when there

excess demand. This is implied by the model,

105]. The

resulting

decline

increases employment
rate. A

in

real

is an

[see Stein, 1982 :

unit

labor

cost

W/p

and output and decreases the unemployment

Keynesian could

accept

this

argument,

because

the

parameter M/W in the aggregate demand curve has increased.

No concept
here

because

of "unanticipated"

the

underlying

money growth

theory

of

is used

anticipations

is

ASYMPTOTICALLY RATIONAL as described by the curves in figure 2,

rather than the limiting case of MUTH RATIONAL.
The basic
inflation

and

statistical strategy has been to formulate

unemployment

variables stressed
simultaneously

by each

which

important. This

school of

sets

is the

equations
of

which
thought,

variables

nested tests

contain
and

are,

or

the

to

test

are

not,

of hypotheses, which are

implied by the general dynamic model.

In

the

process

alternative hypotheses
shown that

statistical

during the

testing

period 1953-86

of

the

it has been

(A) the NCE always fails to explain unemployment,

whereas the
with the

of

Monetarist unemployment

evidence;

(B) the

hypothesis is

weak Monetarist

consistent

hypothesis is a

16

better

explanation

hypothesis. This

of

inflation

than

is

the

Keynesian

is another way of saying that both the strong

Monetarist and the Keynesian hypotheses have elements of truth.

C. Statistical testing of the unemployment equation
Unemployment

(6)

equation

contains

the

NCE

and

Monetarist statistical hypotheses

(6) U(t)

= constant + aU(t-1)

- b(L)[mu(t) - Emu(t)1 - c[mu(t)

- pi(t-1)1 + v(t)
L

lag operator; mu(t) - Emu(t)

mu(t-1)

pi(t-1)

change in

unanticipated money growth

real balances

in the previous

in

(6) ,

period
The

NeE

hypothesis,

equation

that

is

unanti c i pated money growth matters, but not the previous year's
growth

of

real

hypothesis.

balances

Thus

the

as

NeE

implied

claims

by

the

Monetarist

vector

that

"b n

is

significantly positive and coefficient "c" is not significantly

different from

zero. The

Monetarist hypothesis

is

just

the

opposite.
The NeE
nested tests
their equation
such

as

do not

follow

of hypotheses
(2),

general

estimates of

in

this

research

but simply

strategy

consider

variants

of
of

isolation from alternative hypotheses,

equation

non-observable

(6).

They

construct

anticipations,

which

arbitrary
cannot

be

subject to verification by an independent observer.

It is
admit that

interesting that

they cannot

19861. They

even the

explain the

Great

NeE are

forced to

Depression

[Rush,

claim that this period is just an aberration which

does not affect the validity of their theory. The NeE school of
thought

cannot

particularly the

explain

the

event that

major

macro-economic

events,

gave rise to Keynesian economics.

In every' country and in every period the NeE fails when tested
against

the

"unanticipated"

Monetarist
money

hypothesis.

growth,

following

The
the

measures

of

NeE

of

ways

17

measuring it
growth of
(Stein,
Wohar,

are never

real

balances

1982
1984

significant;

ch.

are

but the previous year's

always

4). Several

significantly

authors 1983

negative

Turnovskyand

; Benderly and Zwick,1985 ; Rangazas and Abdullah,

1986) have used this research stategy of testing nested sets of

hypotheses. They
before 1929,

have considered

the interwar

period. Others
results are

always the

output equation
The NCE

period and

examined foreign

1952:3-1980:4 (Domenech,

several periods
the post

countries such

1987) and

in the US :
World War

II

as

Argentina

Germany (Sauer,

1987). The

same for the unemployment or growth of

(6), as were obtained in Stein [1982 : ch. 4].

always fails

in every period and country, when tested

simultaneously against the alternative schools of thought.

The Keynesians

are not

involved here,

because they

could accept the Monetarist formulation insofar as the monetary

variables are concerned.
D. Inflation equation
The inflation equation (7)
Keynesian versus
only indirectly

hypotheses;

involved. The

coefficient of
negative, and

the Monetarist
the lagged

is primarily a test of the

Keynesian

view

unemployment rate

the coefficient

and the
is

NCE is

that

the

is significantly

of the previous period's growth

of real

balances is not significant. The basic Monetarist view

is that

(i) the

coefficient of the change in real balances is

significantly positive.
hypothesis is that,

The stronger version of the Monetarist

in addition to (i),

(ii) the coefficient of

the lagged unemployment rate is not significant.

(7)

[pi(t) - pi(t-l)]

a[mu(t-l) - pi(t-l)] + b. U(t-l) + v(t)

or

(7a) pi(t) = (i-a). pi(t-l) + a.mu(t-l) + c.U(t-l) + constant +

v(t)

The logic
figure 1.

When

aggregate demand

of this

the

real

equation can
balances

curve shifts

be conveyed through

rise

upwards. The

in

year

t-l,

Keynesian

the

excess

demand AB reduces inventories and also raises the rate of price

18

change

in

auction

markets,

which

then

get

transmitted

throughout the economy. Given the rate of inflation at point A,

the Keynesian
inflation.

excess

Viewed

demand
in

of

this

equation is not that unusual.

inflation

will

follow

AB

way,

increases
the

the

Monetarist

rate

of

inflation

In terms of figure 2, the rate of

trajectory

or

c'

and

converge

asymptotically to the rate of monetary expansion.

For the
[Stein. ch.

United States

41;

[1957-76. Carlson,

[1958-80, Domenech),
strong Monetarist

the Keynesian

view is

considered jointly
1929, and

is

in equation

not

19781

(7). For

to 1953,

demonstrably

(1962-79),

and

view is

accepted, when

possibly prior

hypothesis

(1958-79), Canada

Argentina

rejected and

the

both hypotheses are

the period

in the
superior

US the
to

the

prior to
Monetarist
Keynesian

equations (Rae; Turnovsky and Wohar).

In Germany

[Sauer) the

Keynesian view is a superior

explanation.

In the period with fixed exchange rates, one would

expect that

money is

endogenous and therefore should not be a

good explanation of inflation. However, the Monetarist equation

also fails

during the

indicates that
for Western
as yet

period of flexible exchange rates. This

what is true for the US is not necessarily true

Europe. Although the NCE always fails, we have not

succeeded in

finding an

inflation equation

which

is

valid for all countries, exchange rate regimes and periods.

The coefficient
aggregate in
which the

"a" on

the inflation

rate of

expansion. The

the growth

of the

monetary

equation (4) measures the speed at

inflation converges to the rate of monetary

"half-life" is

the time that it takes for half

of the

full effect

the US

and only 1.7 quarters in Argentina (Domenech). The rate

of inflation
times as
implied by

to occur. The half-life is 4.8 quarters in

converges to the rate of monetary expansion three

fast in Argentina as it does in the US. The reason is

the ASYMPTOTICALLY RATIONAL EXPECTATIONS HYPOTHESIS

to be discussed below. When inflation is very high, people know

that the

cause is

react quickly
where inflation
and people

the high budget deficit; and anticipations

to the

is lower,

place more

analysis than

current fiscal

policy.

In the countries

it is attributable to many factors;

weight upon

their priors in a Bayesian

they do to the current monetary or fiscal

input.

19

In terms

of the chain of causation above, the link between the

current monetary input and the expected trend rate of inflation

is weaker in low inflation countries.
E. Post

1980 Price

Federal

Experience

Reserve

and

the

the

Dilemma

of

the

Monetarist/Keynesian

Hypotheses
In october
operating procedure
the federal

1979, the

and switched

funds rate

the growth

of the

the Federal

summarized his

Reserve

changed

its

from a policy of controlling

to one where the object of control was

money supply. The subsequent experience led

Reserve to

monetary growth.

Federal

de-emphasize the

policy of targetting

In his testimony to Congress, Chairman Volcker

attitude concerning

the use

of HI as a policy

tool.
"Experience over
the

difficulty

monetary policy
two simple,

-I

in current
HI alone

half of 1986 underscored

impossibility-

of

conducting

circumstances according

during this

institutional transition
future price
the broader

is not

pressures ... The

aggregates, as

economy and

to one or

period

today a

of

economic

reliable measure

more restrained

well as

the

and
of

performance of

performance

of

the

prices themselves, point in a different direction"

Nuetzel,

The
concerning the
money as

say

preset criteria ... The weight of evidence strongly

suggests that

(quoted in

the first

would

1987).

difficulty
use of

a policy

that

the

Federal

Reserve

faced

either HI or the growth of high powered

tool is easily seen in the following set of

data.
rate of growth of

1967-82

1980-86

HI

6.4% per annum

9.0% per annum

H2

9.2

9.4

adjusted monetary base

7.4

7.7

GNP price deflator

7.0

4.9

20
The growth
decelerate,

but

of the

there

was

three monetary aggregates did not

a

significant

deceleration

of

inflation. The situation since 1983 : 4 was even more dramatic.

The growth

of the

monetary aggregates increased substantially

declined to 2.2 % per annum. These phenomena led

and inflation

the Federal Reserve to abandon its earlier policy of targetting

the growth of monetary aggregates.
The questions
in the

inflation rate

incorrect?

The way

then become
? Is

: what caused the decline

the monetarist inflation equation

to resolve

these

questions

is

to

re-

estimate equation (7a) and use the growth of high-powered money

(ghpm) as the monetary aggregate in place of M1, because of the
financial deregulation.

This is done in equations (7b) for the

period 1957-80 and (7c) for 1967-86.

In each case, the form of the equation is :
(1-a)pi(t-1) + a ghpm(t-1) + b U(t-1) + c

(7b ; c) pi(t)

RBAR-SQUARE 0.89

OBS = 24
OW 2.27

(7b) 1957-1980

SEE 0.93

variable

coefficient

t-statistic

CONSTANT

1.7

1.98

LAG INFL

0.72

5.98

LAG GHPM

0.37

3.22

-0.40

LAG UNEMP

-2.5

equation

This

is

most

consistent

with

monetarist theory.
lagged growth
unity.

The

.w.e.a.k.

The coefficients of lagged inflation and

of high powered money are significant and sum to
Keynesian effect has the right sign and is

significant.

( 7 c) 1967 -1986

OBS

RBAR-SQUARE 0.77

OW 2.7

20
SEE 1.08

variable

coefficient

t-statistic

CONSTANT

0.57

0.41

LAG INFL

0.75

6.2

LAG GHPM

0.72

3.2

-0.68

-4.3

LAG UNEMP

the

21

The Keynesian

variable, lagged unemployment, becomes

more important

in explaining

the deceleration

inflation. The

lagged growth

of high

of the

powered money

recent
is still

statistically significant. However, the sum of the coefficients

exceed unity, which should not occur.
The conclusion
monetarist hypothesis
and Keynesian

is that

the

-which is

arguments- are

elements

of

the

weak

a synthesis of the monetarist

still valid

in

explaining

the

deceleration of the inflation since 1980. However, the standard

error

of

estimate

unemployment,

increased.

however,

inflation decelerated.
concerned about

has

The

been

effect

of

main

reason

the

the

high

why

the

But the Federal Reserve should still be

the effect

of the high growth of the monetary

base upon future inflation.

F. Real Stock Market Returns
The weak
the

growth

of

Monetarist hypothesis
the

economy

and

explains more

the

variati.ons

about
in

the

unemployment rate than does either of the other hypotheses, and

has fruitful implications for financial markets

in particular

real stock returns.

For example,

Jason Benderly and Burton Zwick (1985a)

used the

Monetarist model

why they

seem to

to explain

real stock returns, and

be negatively related to the current rate of

inflation.
Their argument is that real stock returns in year t-l
denoted by R(t-1) reflect the growth of the economy from t-l to
year

t.

Monetarist

The

growth

monetary aggregate,
1, denot'ed

of

equation

the

(Sa)

economy

above.

Given

G[y(t)]

is

given

the

growth

of

by
the

a rise in the rate of inflation in year t-

by pi(t-1),

lowers the growth of real balances and

hence the growth of output G[y(t)].

Since real

stock returns in year t-1 reflect G[y(t)]

; and a rise in pi(t-1) reduces G[y(t)], a negative relation is

observed between

real stock returns and the rate of inflation.

This explanation is a further indication of the strength of the

Monetarist hypothesis.

22

II. MICRO FOUNDATIONS OF MACROECONOMICS

A crucial variable in all three schools of thought is
the rate of unanticipated price change. Recent theoretical work
on the

micro-economic level investigates how MRE occurs,

occurs at

all. There

is a

if it

growing number of theorists who do

not accept the MRE hypothesis as an axiom like expected utility

maximization. There
to

determine,

in

is almost
an

no way to examine a macro model

objective

way,

what

appropriate equation

for anticipations.

composite hypotheses

in

equation for
of this

macro

price anticipations.

paper

concerning:

examines

evidence

is or is

There

model

to

are

not
too

disentangle

the
many
the

Instead, the remaining part

drawn

from

micro

markets

interest rates, futures prices and foward rates on

foreign exchange.
A. Bayesian Errors. Muth Rationnal and asymptotically
rational expectations
The forecast
p(t+l), and
denoted by
equation

error between the price in t+l, denoted

the subjective

or market

E*p(t+l;t), can

anticipation at time t,

be decomposed into two elements in

(S).

(S) p(t+l) - E*p(t+l;t)

[p(t+ll - Ep(t+ll] +
[Ep(t+l) - E*p(t+l;t)]

(Sa)

u + B(t>

The first

term is

the difference between the actual

price p(t+l)

and the

abbreviated

as Ep, that is based upon the true model. This is

the expectation

objective expectation

that would

be known

Ep(t+l), which is

to those people who knew

the true distribution of p(t+l) abbreviated as p. Difference pEp. denoted

by u,

is an

unavoidable error which results from

positive price variance around the objective expectation.

The

second

term

is

the

difference

between

the

objective expectation Ep and the subjective expectation held by

the market E*p(t+l;t), abbreviated as E*p.

I call difference Ep

23
- E*p

a BAYESIAN

ERROR, denoted

by B(t).

This error will be

related to information that is available at time t.

An easy
follows. Suppose

way to

that there

distributed numbers
the mean
from the

is

the Bayesian
large

urn

error is
with

as

normally

which are prices. The public does not know

price, but

an individual

think of

knows the variance. At regular intervals,

incurs a

given cost

per unit to sample numbers

urn .. On the basis of the sample, he updates his prior

estimates of

the

estimate E*p

of the mean. There is a Bayesian error denoted by

B(t)

mean

and

obtains

subjective

posterior

[Ep - E-p] at each time and there is a variance var B to

these Bayesian
information

errors. The

available

Bayesian error

at

time

will be related to

contained

currently drawn.
Bayesian

The MRE

hypothesis assumes

error.

The

expectation. There

public

is no

that B

always

=0

knows

attention paid

convergence of Bayesian errors,

in

to

the

sample

; there is no
the

the

objective
process

of

in the MRE framework.

Convergence of stochastic variable requires that both

the expectation
zero.

If

E[B(t)] and

the variance

the variance var B(t) converge to

did not

converge, then

it is possible

that there are growing positive errors which are exactly offset
by negative errors.
The

ASYMPTOTICALLY

HYPOTHESIS is

concerned with

convergence of

the Bayesian

(ARE)

EXPECTATIONS

RATIONAL

the determinants of the speed of

errors. The

MRE hypothesis

is a

very special case of ARE.

It can

be shown [Stein, 1986 : ch. 3] that:

Bayesian errors

are serially

(1) the

correlated. There is no way that

the market can exploit this phenomenon to make profits, because

no one

knows that he has made a Bayesian error (because no one

knows the objective expectation Ep). These errors arise because

some weight

is attached

the weight
attached

is on
to

the

to the

initial priors and not all of

the sample mean. As time goes by, the weight

initial

prior

declines

attached to the sample drawn. Moreover,

Bayesian error
prices in

is positively

the urn,

relative

to

that

(2) the variance of the

related to

the variance

of the

and negatively to the cumulative number of

24

items sampled,

denoted by

N(t). The equation for the variance

of the Bayesian error is given by (9a).

(9) var B(t)

var p/N(t)

The variance,
errors converge
N(t) of

and the

asymptoti~ally

items sampled

that the

expectation, of

the Bayesian

to zero as the cumulative number

grows indefinitively

large:

provided

distribution is stationary. If the distribution keeps

shifting, convergence may never occur.

The cost of sampling prevents people from taking very
large samples
rather than

at any time, and thereby produces an asymptotic,

an immediate,

sampling, the
; and

convergence. The

lower the cost of

larger will be the sample size drawn at any time

there will

be a

faster rate of convergence. Similarly,

the greater the variance of the initial distribution, i.e.,.the

greater the

uncertainty, the

slower

will

be

the

speed

of

convergence. Points (1) and (2), and equation (9) describe in a

compact

way

the

ASYMPTOTICALLY

RATIONAL

EXPECTATIONS

HYPOTHESIS.
In the
MRE hypothesis,
zero,

is

with the

subsequent sections it will be shown that the

which claims

that the

Bayesian error B(t) is

inconsistent with the data. but the ARE is consistent

micro-economic data.

nominal interest

The evidence will concern:

rates and anticipated inflation;

[B)

[C) pricing

in futures markets ; and [0] forward foreign exchange rates.

B. Nominal interest rates and anticipated inflation
The

nominal

decomposed as

the sum

which changes

slowly, and the subjectively anticipated rate of

rate

of

interest

i(t)

is

usually

of two parts : the real interest rate r

inflation from t to t+l denoted E*pi(t+1;t) abbreviated E*pi.

(lOa) i(t)

r+ E*pi(t+1;t)

The difference

between the

actual rate of inflation

pi(t+1) and the subjectively anticipated rate is (lOb), the sum

of an unavoidable error u and a Bayesian error B(t).

25

(lOb) pi(t+1) - E-pi(t+1;t)

It follows
of inflation

u + B(t)

that the relation between the actual rate

and the

nominal rate

of

interest

is

equation

(10c).

=-

(10c) pi(t+1)

r + i(t) + B(t) + u

If the
no Bayesian

MRE hypothesis were correct, so that there is

error, then

several relations should be observed.

First:

there should

be a direct relation between the nominal

rate of

interest and

the rate

regression of
of interest
should be

the rate
on an

unity;

and the

included in

Bayesian error

intercept should be minus the real

contemporaneous variables,

a predictor
It

in

"h" period loan, the regression coefficient

: no

rate as

Second

of inflation from t to t+h on the rate

rate. Third

period.

of inflation.

B(t), should

which

add to

would

be

the interest

of inflation from the present to the next

will be

shown that

each of

these

statements

is

contradicted by the data.

1. The Friedman-Schwartz Evidence
Friedman and
first relation,
result was
War II

Schwartz (1982) were concerned with the

over the

that

long

"the data

run

1867-1975.

remarkable

for all sub-periods before World

behave as if prices were expected to be stable and both

inflation and

deflation were unanticipated : nominal yields on

nominal assets

average much

the same in periods of rising and

falling prices ... " [po 10]. The correlation between the rate of
inflation and the nominal rate of interest is not significantly
positive.
The relation
"After World

War II,

began to

change in

the financial

the recent period.

markets began

to

behave

differently. Beginning in the 1960's, there is a gradual shift,

in both

the United

prior pattern...
inflation, so
variable" (p.

interest rates

real
11).

States and
returns

on

the United
start to
nominal

Kingdom, from
parallel
assets

rates

become

the
of
less

26
2. Recent Evidence
Fama and
(10c) and

the

others

FISHER

evidence. Their

addressed

hypothesis,

hypotheses were

themselves
on

the

that the

to

basis

equation
of

recent

coefficient of

the

interest rate is unity, and the coefficients of contemporaneous

variables is zero. That is,

in a regression equation based upon

(10c), there are no Bayesian errors.

For

the

period

1953-71,

Fama

reported

that

the

interest, rate on government issues with a one year maturity had

a

one-to-one

correspondence

inflation of

a subset

with

the

subsequent

rate

of

of the consumer price index. This would

be support for the MRE.

However, an
contradicts Fama's
He obtained
is not

examination of

the data by Barro (1986)

assertions and rejects the second relation.

several results.

(a) After 1971, the interest rate

useful as a predictor of inflation.

there is

a lot

of available

subsequent inflation

(b)

information that

In all periods,
helps

predict

that is not contained in the nominal rate

of interest.
Several equations (11.1)-(11.3), which are subsets of
equation (10c)
hypothesis

above, show

the serious

Standard errors

are shown

weaknesses of the MRE

in parentheses.

(Barro

1986) .
1953-71 (11.1) pitt)

.01

+ .82 i ( t-1 )

( . 15)

( .006)
This

equation

coefficient b

.82

gives

is not

hypothesized value of unity.

the negative

of the

mixed

results.

significantly different

(i)

The

from its

(ii) The constant, which should be

real rate,

is not significantly different

from zero. Result (ii) makes no sense.

But this

equation is

misleading.

If one adds vector

X(t-h) of currently available information, consisting of lagged

inflation pitt-h)
equation (11.2)

and lagged

is obtained.

money growth

mutt-h) to (11.1),

27

1953-71 <11.2) pi(t)

.001

( . 007 )

. 73 P i ( t - 1 )

( . 35 )

-.30 pi(t-2) +
( . 17)
Now the

.14 i(t-l)

( .32)
.26 mu(t-2)

.49 mu(t-l)

( . 18)

( .23)

MRE hypotheses are clearly rejected.

(a) The

coefficient of the interest rate is not significantly different

from zero,

and is

significantly different from unity.

(b) The

vector containing contemporaneous information with the interest

rate is significantly different from zero. That is B(t) adds to
the explanation of subsequent information.
For the

recent period, matters are worse for the MRE

hypothesis.
1972-82 (11.3) pi(t) = .061
( .025)

.19 i(t-l)
( .30)

Even in the crude form that Fama used, the hypothesis

that the coefficient of the interest rate is unity is rejected.
The conclusion

is that,

insofar as

the Fisher

hypothesis is

concerned, the MRE hypothesis is rejected.

C. Evidence from Futures markets
Considerable light

is shed

upon

the

anticipations

hypotheses from studies of futures markets in foreign exchange,

financial

instruments

and

traditional

commodities.

These

much more active and larger than the corresponding

markets are
spot markets.

For

Treasury bond

futures and futures on the S&P-500 index are ten

times that

easier to

the

daily

volume

of

trade

in

of the volume on the New York Stock Exchange. These

markets can
no price

example,

only flourish when there is uncertainty. There are

rigidities or

frictions in these markets.

test anticipations

hypotheses on

a micro

It is much
level in

these markets than from tenuous inferences from macro models.

In futures
concerning, the

markets we

focus

upon,

and

have

data

futures price q(t+h;t) at time t of a contract

28

which calls

for delivery

at subsequent date t+h. For example,

consider the

futures price

Japanese yen

to be

on January

delivered in

9 of

a US

March. Denote

dollar for

this price as

q(3;1) and abbreviated as q.

Continue with
This forecast

the approach

error (p-q)

taken in

equation (8a).

can be decomposed into three parts.

The forecast error has three elements.

(12)

(p-q)

(p-Ep) + (Ep-E*p) + (E*p-q)

The first

term (p-Ep)

The second term (Ep - E*p)

term (E*p-q)
to take

u +

B + R

u is the unavoidable error.

B is the Bayesian error. The third

is a risk premium necessary to induce people

either a long or a short position in the risky futures

contract.
The evidence

from a

study of many futures market is

most relevant in evaluating the validity of the MRE hypothesis.

If the

MRE were correct, the Bayesian errors would be randomly

distributed among groups ; and there would be no Bayesian error

in the

market. Then,

futures markets

would

not

affect

the

relation between the subjective and objective expectations, but

would just determine the magnitude of the risk premium.
1. Futures markets affect the Bayesian errors
It

has

theoretically and
the convergence

been

shown

(Stein

1986

ch.

5),

both

empirically, that futures markets accelerate

of the

Bayesian errors

to zero. Prior to the

development of futures markets, the MRE can always be rejected.

After its

development, the

MRE hypothesis fares better. These

results are consistent with the ARE hypothesis.

First:
in Maine

and Idaho,

development of
year

A(t)

in some

there were

cobweb cycles

prior

to

the

futures markets. The acreage planted in a given

depended

previous year's

commodities, such as potatoes grown

positively

price at

and

significantly

upon

the

the harvest p(t-1). This relation is

inconsistent with rationality of expectations.

Futures markets
in

Idaho

potatoes.

The

developed in Maine potatoes, but not

cobweb

cycle

continued

very

significantly in the case of acreage planted of Idaho potatoes.

29
However it disappeared in the case of Maine potatoes, such that
there was

no

longer

any

significant

relation

between

the

acreage planted A(t) and the previous harvest price p(t-l).

There were
when there
futures

clearly very

significant Bayesian errors

were no futures markets in this commodity. When the

markets

were

developed,

the

Bayesian

errors

were

reduced significantly.
Second : consider the case of "storable" commodities.
A commodity

is purchased

the commodity

at time

t becomes

physical depreciation,
positive

marginal

including risk
the commodity

at price p(t) at time t. One unit of

m<1

m units

and

productivity,

premium is
from t

expected price at t

if

at t+l.

the

m>l.

If there is

commodity

The

has

interest

rate

i(t). The marginal cost of carrying

to t+l

is p(t)(I+i)/m. The subjectively

to prevail at t+l is E*p(t+l;t).

When the commodity is storable the following relation

must be

satisfied if people optimize. Marginal cost,

risk premium,

should equal

the subjectively

including

expected

price.

Th i s i s (13. 1 ) .
(13.1) E*p(t+l;t) = [(I+i)/ml p(t)
Using equation

(8a), which

relates the subjectively

expected price to the realized price, the optimization equation

is (13.2).
(13.2) p(t+l)
If the
there are

[(I+i)/ml p(t) + B(t> + u

MRE hypothesis

no Bayesian

were correct

errors, then

such that (B=O)

the price

at t+l

should

depend upon the current price, but not on earlier prices p(t-h)
where h>O.
This

hypothesis

was

examined

for

many

storable

commodities (see the discussion of the work of C. Cox in Stein,

1986 : ch. 5, with references to the empirical literature). For
each commodity, two periods were examined.
there was

no active

second period,
commodity.

futures market

there was

an

active

In the first period,

in the
futures

commodity.
market

In the
in

the

30

The statistical hypothesis is contained in regression

equation (13.3)
price pet)

where spot price p(t+1) is regressed upon spot

and a

vector of

past spot prices pet-h), h>O. The

prices are daily spot prices at weekly intervals.

(13.3) p(t+1)

a + b pet) + c. pet-h) + u

The MRE

hypothesis is

errors. Therefore,

that there

the contribution

are

no

of c.p(t-h),

Bayesian

h>O to

the

prediction of p(t+1) should not be significantly different from

zero. Compare (13.2) with (13.3).
The F-statistic,
pet-h), was

concerning the

calculated for

contribution of

each commodity

c.

in each of the two

periods: with and without futures trading. The results are the
same for
in the

each commodity.
commodity, the

(1) When there was no futures trading

F statistic

is highly significant. The

vector of past prices (h>O) adds significantly to the explained

sum of

squares. This rejects the MRE hypothesis that there are

no Bayesian

errors.

commodity, then

(2)

When there was futures trading in the

F is not statistically significant. One cannot

reject the MRE hypothesis that there are no Bayesian errors.

The results indicate quite clearly that the existence
of

futures

markets

equilibrium.
seems to

accelerates

the

convergence

to

MRE

In fact, the existence of an active futures market

be a

necessary but

not a sufficient condition for a

rapid convergence to MRE.

2. Differential profits in futures markets

There is
MRE is

another important

not correct

arise from
makes a

in general. Profits in futures markets can

several sources.

smaller Bayesian

superior forecasting
not exist
particular
objective

; and
group

costs of

One is

that the
is

not
Ep

ARE claims

sampling will

that a

error than

ability. The

expectation

expectation. The

set of evidence that the

particular

MRE assumes

that this does

subjective expectation
significantly
which

is

that the

take larger

group

does the market. This is

different
equal

to

group with
sized samples

E-p of

any

from

the

the

market

the smaller
of current

31

information and,

on the

basis of equation (9), they will have

smaller qayesian errors.

The expected profits of a particular group trading in
the futures

by EZ. is given by equation (15).

market, denoted

[See Figlewski ; Stein, 1986 : ch. 51.

(15) EZ.

f(l-b.)

If the

Bayesian error

of a

particular

group

were

regressed upon the Bayesian error of the market, the regression

coefficient in equation (15.1) would be b . This regression can
only be

done theoretically,

unobservable. However,

since

the

the expected

Bayesian

errors

are

(average) profits EZ. are

observable. A group will make positive expected profits,

if its

Bayesian error is smaller than that made by the market.

(15.1) b t
f

= cov(B.,B)/varB

positive function

Bayesian error

= market

; B

of this

group; EZ.

Bayesian error ; B.

the

expected profits made by

the i-th group in the futures market.

The
differences

evidence
in

speculators and

is

that

forecasting
small

there

abilities

(amateur)

are

between

significant
professionnal

speculators

and

between

larger agents (professionnal speculators and large commercials)

and small

agents

profits of

the

(either
larger

superior forecasting
The larger

a whole

Bayesian errors
errors are

than does

a systematic

or

speculators).

generally

derive

from

than from

smaller Bayesian

whereas

not random

differ in

agents

abilities rather

agents make

market as

commercials

The
their

risk premia.

errors than does the

the smaller

agents make

larger

the market as a whole. The Bayesian

variables with
way among

zero

groups as

means,

but

they

predicted by ARE

equation (9). This is another rejection of the MRE.

In the conventional theory of speCUlation, all groups
have MRE

but one

profit from

the risk

from commercials
this risk

group, usually
premium (R

who want

referred to

to hedge

as speculators,

Ep - q) which they receive

away risk.

Keynes called

premium resulting from short hedging by commercials,

32
"normal backwardation".

the MRE/normal backwardation

states that there are no Bayesian errors, but

hypothesis which
the

There is

profit

speculators

from

the

"normal

backwardation".

this view, the commercials should lose on average

According to

in the futures markets.

Hartzmark examined
commercial and
report their
Commission)
1977 to

non-commercial traders
daily positions

in

the futures

December 31,

reporting commercial
the losses

of the

pure speculators.
table 1
markets.

(who

are

the

required

markets during

1981. The

profits

and non-commercial

the period July I,

and

losses

traders are

of

the

equal to

the large commercial firms make profits on

their

contradicts

short

positions

the

in

MRE/normal

the

RETURNS TO REPORTING TRADERS OF FUTURES 1977-1981

($ millions)
Commodity

mean return

commercial
non-commercial

0.35
0.03

commercial
non-commercial
pork bellies
commercial
non-commercial
live cattle
commercial
non-commercial
feeder cattle
commercial
non-commercial
T.bonds
commercial
non-commercial

0.45
0.04
2.80""
0.07"
-1.02"
0.55""
0.49
0.64""
5.95"
-0. 11

futures

backwardation

TABLE 1

wheat

to

to the Commodity Futures Trading

hypothesis.

oats

large

An examination of the evidence summarized by

long and
This

returns to

non-reporting (small) traders who are often

shows that

both their

the daily

33

T-bills

1.20-0.04

commercial
non-commercial

All markets
commercial
non-commercial
Source:

Hartzmark (1987,
table 3)
; his original study is
cited in Stein, 1986 : ch. 5). The -(--) = significant
at 10~ (1~) level

D. Foreign exchange markets

The

foreign

exchange

markets

are

important

and

interesting. Forward markets are highly developed and have been

used for

many years.

exchange market
hypothesis.

The

forward/futures

A remarkable

is that

existence
markets

phenomenon in

they are
of

seems

the foreign

not consistent with the HRE

active

to

be

and

necessary,

developed
but

not

well known

and

sufficient, for a rapid convergence to HRE.

This can
simple way
is a

be shown

to see
to

connection with

ways. A

this is to show that in these markets there

total violation

corresponds

in many

of the

the

FISHER

FISHER
CLOSED

equations (10)

concerning futures

OPEN

hypothesis,

hypothesis

(11) above,

which

discussed

or equation

in
(12)

markets in storable commodities. Focus upon

the latter, repeated here as equation (16), when the spot price

= E-p-

has been subtracted from both sides. The risk premium (R

q) has

not been

failed to

written explicitly

find any

evidence for

because researchers

have

its existence in the foreign

exchange market.
(16)

[p(t+1) - pIt)]
The HRE

[q(t+1);t) - pIt)] + B(t) + u

hypothesis which

claims that

there are

no

Bayesian errors, just unavoidable ones, implies the following.

A regression

of the (logarithm) of the spot price at

time t+1 less the (logarithm) of the spot price at time t, upon
the (logarithm)
time t+1

of the futures price at time t for delivery at

less the

(logarithm) of

the spot

price at

time t.

34

should yield

a regression

coefficient,

denoted

by

"b",

of

unity.
There is
covered interest
premium on

another way
rate parity

the foreign

to look

at equation (16). The

hypothesis is

exchange [q(t+l;t)

that the

forward

- p(t)], where the

variables are logarithms, is equal to the domestic less foreign

interest rates

on securities

of comparable

risk, denoted

by

i(t). Equation (17) holds exactly in the Euro-currency market.

(17) q(t+l;t) - p(t)

The forward

i(t)
foreign exchange will be at a premium if

the domestic interest rate exceeds the foreign rate.

Equations (16)

and (17)

then imply

(18),

equation

which is the FISHER OPEN hypothesis

(18) p{t+l) - p(t)

If the
rate,

i>O,

i(t) + B(t) + u

domestic interest

rate exceeds

the

foreign

then the foreign currency is expected to appreciate

relative to

the domestic currency. The Fisher Open equation is

the crucial

dynamic equation in a large part of the literature

concerning exchange
the analysis

It makes

no difference to

whether equation (16) or (18) is used to test the

MRE hypothesis.
have to

rate dynamics.

It is

search for

easier to use (16) because one does not

interest rates

on securities

with

equal

risk.
Every study
the hypothesis
done by
Table

that b

the Bank
2

"b"

the

that the

hypothesized value

standard error

is so

foreign exchange
The most
and
the
is

standard
one

month

significantly

value is

market rejects

comprehensive study was

(Longworth, Boothe

values

coefficient

asterisk indicate

definite.

1.

concerning

Generally, this
from the

of Canada

presents

coefficient

of the

and Clinton).
errors

of

forward
negative.

significantly

the
rate.
The

different

of unity. In the other cases, the

large that

one cannot conclude anything

35

TABLE 2
EVIDENCE FROH FOREIGN EXCHANGE HARKETS 1970-1981
ONE HONTH FORWARD RATE RELATIVE TO US DOLLAR
COUNTRY

REGRESSION COEFFICIENT

STAND. ERROR

.389--

Canada

.550)

France

- 1.8376-

.8756)

Germany

.5333
.4786-.4803
- 1.5330--

(1.4169)
( .4716)
.4174)
.8434)

Italy
Japan
UK
Source

Longworth, Boothe and Clinton.

rent from unity at .05 level.

Contrary
domestic interest

to

the

rate exceeds

foreign exchange

tends

contradicts the

Fisher

to

significantly diffe-

Open

hypothesis,

the foreign

decline.

The

if

the

rate the price of

empirical

evidence

crucial equation in the exchange rate dynamics

literature.
The HRE hypothesis, or its counterpart in the finance
literature THE
the data.
and can

EFFICIENT HARKET HYPOTHESIS,

The ARE

hypothesis is

explain why

a negative

is contradicted by

consistent with the evidence

"b" is

1986:

found (Stein,

ch.5).
Hy

conclusions

are

as

follows.

First

futures
a rapid

necessary, but not sufficient,

for
markets are
convergence to HRE. Second: at any time, there are systematic
differences

among

commercial firms,

groups
and to

in their
a lesser

Bayesian

errors.

extent large

Large

speculators.

profit at the expense of small speculators. In terms of the ARE

hypothesis. the
errors because
information are
groups.

Third

large commercial
their costs

of sampling

smaller than
even

firms make

when

must be
there

or

smaller Bayesian
obtaining

incurred by
are

well

the

market
other

developed

forward/futures markets, such as in foreign exchange, there are

36

very significant
hypotheses are

Bayesian errors; and the MRE and Fisher Open

not consistent

with the

: It is

incorrect to

use the

macro model,

as is done by the NCE. Fifth: the ARE hypothesis

which implies

MRE hypothesis

data. Fourth

the speed

as an integral part of a

of convergence

(eqn.9) is a superior

anticipations hypothesis. The micro evidence is consistent with

the empirical

results obtained

from the macro tests discussed

in part I above.

REFERENCES
Barro,

~obert
(1986) Futures
Markets and the Fluctuations
Inflation, Journal of Business, 59, April, S21-38

in

Barro, Robert (1978) Unanticipated Money, Output and the Price Level
in the United States, Journal of Political Economy, 86, 549-80.
Benderly, Jason and Burton Zwick (1985a)
Inflation, Real Balances,
Output and Real Stock Returns, American Economic Review, 75,
December, 1115~23
Benderly, Jason and Burton Zwick (1985b) Money,
Unemployment and
Inflation, Review of Economics and Statistics, 67, February,
139-43
Blinder, Alan (1987) Keynes,
Lucas and Scientific Progress, American
Economic Review, Papers and Proceedings, 77, 130-36
Carlson, Keith (1978)
Inflation, Unemployment
Federal Reserve Bank of St. Louis, n09

and

Money,

Review

Domenech, Roberto (1987) Output and Inflation in a Semiindustrialized

Economy: Argentina 1950-80. Ph.D. Thesis, Brown University
Fama, Eugene (1975) Short term Interest Rates as Predictors
Inflation, American Economic Review, June, 269-82

of

Figlewski, Stephen (1982) Information Diversity and Market Beahavior,

Journal of Finance, 37, 87-102
Friedman, Hilton and Anna Schwartz (1963) A Monetary History of the
United States 1867-1960, Princeton University Press

37

Friedman, Hilton and Anna Schwartz (1982) Monetary Trends

United States and the United Kingdom,
University of
Press

in the
Chicago

Hartzmark, Hichael
(1987) Luck
Versus Forecasting
Ability
Determinants of Trader Performance in Futures Markets, Working
Paper, Department of Economics, University of Michigan
Longworth, David, P. Boothe and Kevin Clinton (1983) A Study of the
Efficiency of Foreign Exchange Markets, Ottawa: Bank of Canada
Lucas, Robert E. (1973) Some International Evidence on the OutputInflation Tradeoff, American Economic Review, 63, 3, 326-34
Hodigliani, Franco and L. Papademos (1976) Monetary Policy for the
Coming Quarters, Federal Reserve Bank of Boston, New England
Economic Review, March/April
Neutzel, Philip (1987)
Uncertain Times.
February

The FOMC in 1986

Flexible Policy for
Review. Federal Reserve Bank of St.
Louis,

Rangazas. Peter and Dewan Abdullah (1987). Testing Some Monetarist

Propositions. Working Paper, University of Toledo. Ohio
Rea. John D. (1983) The Explanatory Power of Alternative Theories of
Inflation and Unemployment. Review of Economics and Statistics,
LXV. May. 183-195
Rush. Hark (1986) Unexpected Money and Unemployment. Journal of Money.
Credit and Banking. 18. 259-74
Sargent. Thomas (1976) A Classical Macroeconometric Model for
United States. Journal of Political Economy. 84. 207-37

the

Sauer, Christine (1987) A Test of Alternative Macroeconomic Theories

in an Open Economy : Germany 1960-85. Ph.
D. Thesis. Brown
University. to be published by Springer Verlag
Stein, Jerome L.
(1982) Monetarist, Keynesian
Economics. Oxford: Basil Blackwell
Stein. Jerome L. (1986)
Basil Blackwell
Tobin.

The Economics

of Futures

and

New

Classical

Markets. Oxford:

Ja~es
(1975) Monetary Policy.
Inflation and
Papers on Monetary Policy. Bank of Australia

Unemployment.

Turnovsky, Stephen and Hark Vohar (1984) Monetarism and the Aggregate
Economy:
Some Longer-Run Evidence. Review of Economics and
Statistics. 66. November. 619-29.

A NEO-STRUCTURALIST HODEL OF INFLATION AND UNEHPLOYHEN"I<*)

Jean-Paul AZAH
CERDI

University of Clermont-Ferrand

INTRODUCTION
On the

border between usual macroeconomic theory and

development theory,
which can

one can

be more

approach to
be found

find a

or less

bunched to

inflation. Although

in the

set of

analytical themes

form the structuralist

some unity of inspiration can

latino-american works which

are so labelled,

it does not seem possible to speak of a structuralist theory of

inflation,

in a proper sense.

But some

recurrent propositions seem to playa major

role in this theoretical movement :

1) the

supply of consumption goods is inelastic, and

particularly agricultural supply;

2) wages resist to inflation, thanks to trade unions,
or because of minimum wage policies;
3) the various sectors of the economy do not react to
inflation in the same way, and even less so to disinflation, so
that relative prices are affected;
4) the
be found

in the

basis of

macroeconomic explanation is not to

behaviour of

some aggregates (like the money

supply or

the level of government expenditures), but rather in

the value

of some parameters (like elasticities, propensities,

or rates). For example, the fiscal system matters more than the
level of

taxation. Other

concerned with
following, where
Findlay,

H The

( 1988) .

propositions from

international trade,
we

only

analyse

and are
the

this movement are

neglected in the

closed

economy

(see

1973).

French version

of this paper has been published as Azam

40

In the

present article,

model of

inflation and

flavour,

emphasizing

assumptions

and

characteristics.

we are

unemployment
inelasticity,

deriving

having

the

structuralist

rigidity,

results
we

Consequently,

microeconomic foundations

trying to produce a
and

provide

asymmetry

"structural"

from

some

rigourous

inelasticities, taking into

to some

account the rationing which the assumed price rigidities imply.

For

this

reason,

the

label

"Neo-Structuralist"

is

more

appropriate than "Structuralist", for the following work, which

does not follow strictly the line of that movement.
We

use

two-sector

producing intermediate
in Hicks

(1974). To

been called:
as the

odel,

goods, whose

including

sector

antecedent could be found

some extent, the present paper could have

"The Crisis in Keynesian Economics Reconsidered",

model developed

here

uses

some

of

the

assumptions

appearing in this non-mathematical book of the 1972 Nobel Prize

winner. Hence,

we assume

"flexprice" markets,

that "fixprice" markets exist beside

allowing for

some influence of inflation

on relative

prices. Similarly,

goods plays

a major part in our model, as in Hicks' book. But,

it is

rather reasonable

between this

the

market

for

intermediate

to establish as we do here a parallel

book of Hicks' and the Structuralist approach, as

Basu (1984) has already done.

But the

aim of

the present

model is

not

only

to

formalize more or less faithfully a given theoretical approach.

Its interest lies mainly on the new light which it sheds on the
long-run relationships

between inflation and employment. We do

not find in the following a "vertical long-run Phillips curve",

but we

end up on the contrary with a model where employment is

maximized in

the long

corollary, this
costs

of

model provides

perfectly

employment.

run when
expected

Curiously

then,

prices

are

constant.

As

a precise analysis of the real

inflation,
our

in

terms

Neo-Structuralist

of

lost
model

provides microeconomic foundations to the positive slope of the

long-run Phillips curve, which Friedman (1977) hypothesized.
The assumptions of the model are presented in a first
section, while
firms and

the second

the aggregate

one describes
supply curve

the behaviour of the

which

it

implies.

No

41

effort is

done to

the demand

"close" the model by a detailed analysis of

side,

mechanism to

and

one

only

assumes

prevent equilibrium

curve, whether

in its

dynamic long-run

from

static short-run

version, which

is the

that

lying

there
on

the

version,
topic

is

or

of

no

supply
in

the

its
third

section.
1. THE ASSUHPTIONS OF THE HODEL
1.1 The technologies of the sectors
We

analyse

here

model-economy

including

sectors, producing a quantity Y of the final good,

and a

quantity X of the intermediate good,

goods are

sold respectively

denote L y

and Lx

sectors,

and

at the

two

in sector I,

in sector II. These

money prices

p and v. We

the quantities of labour employed in the two

the

decreasing returns

money

prevail in

wage.

Lastly,

the two

we

assume

sectors, so

that

that
the

production functions satisfy the following assumptions

(2)

FL > 0, Fx > 0

( 3)

FLL < 0, Fxx < 0, FXL > 0

D IE FLLFxx - F 2 LX > 0

(4)
(5 )
(6)

F(L y

X)

(1)

X = H(L x )
H' > 0, H" < 0
In the

the subscripts

expressions of
denote the

assumptions (2),

(3) and (4),

partial derivatives with respect to

the corresponding variables.

1.2 The indexed wage
As far
concerned,

we

as the

working of

select

assumptions

"flexprice" markets
market, we

exist side

assume that

the
so

various
that

markets

"fixprice"

is
and

by side. First, for the labour

full employment

is not

feasible, and

hence that no quantitative constraints can ration firms on that

market.

In

other words,

we restrict our attention to the case

42
of

unemployment

equilibria.

Moreover.

we

assume

that

the

nominal wage w is perfectly indexed on the price p of the final

good. and we denote s the fixed value of the real wage:
(7)

sp

In order
to (at

to justify this assumption. one could refer

least) four

theories. First.

appealing rationale
assume that

trade unions

imposing the

real wage

following Friedman's
has decided

and

this

is

the

more

within the structuralist "vision", one can

are very

powerful, and

in

s. Then, one could as well assume that

advice (Friedman,

to use

succeed

an indexation

1974), the

mechanism of

government

wages on

the

prices of

the consumption goods purchased by wage earners. But

one could

as well

which could
For example,

for a

rigid real wage

to less well organized labour markets.

one could

refer to Lewis's unlimited supplies of

labour hypothesis
constant MRS

find justifications

be applied

(Lewis, 1954),

between leisure

or to Sen's assumptions of a

and consumption

in the

peasant

sector (Sen, 1966).

1.3 Market clearing for the final good
Then, at the other extreme, we assume that the market
for the

final good Y is instantaneously cleared because of the

perfect flexibility

of the nominal price p. This assumption is

akin to

Postulate of

the "First

maintained in
being very
apply this

which Keynes

the General Theory, and can thus be justified as

common, used

well justify

the Classics",

it by

by nearly

every theorist. One can as

empirical arguments, provided one does not

analysis to

the very

short run

(see Azam,

1986,

chap. II).
1.4 Sluggish

adjustment

on

the

intermediate

good

market
Lastly, for
mixed assumption.
quite natural

the intermediate

As Hicks

to assume

good market,

(1974) explicitly

that the

we use a

does, it

seems

price v of the intermediate

43

good reacts to the mismatch of supply and demand less quickly

than the price p of the final consumption good. One can regard
the position of the firms in this sector as being governed by
complex relationships. like those implied by sub-contracting.
There are good reasons to establish "customer's markets" in
this case. following Okun's wording (Okun. 1981). Price changes
are costly and imply the risk of breaking relationships of
trust that the sub-contractors are not prepared to bear
untimely. "Continuity"
plays a major role. which Okun
emphasizes. as well as the sense of "fairness". which is
stressed by Hicks. Similarly. it is plausible that the price
agreed between the firms on this market does not fluctuate in
order to pass on every changes in costs. but is rather related
to "normal costs". as assumed by Hicks. Therefore. it is
reasonable to assume in this Hickso-Structuralist framework
that the price v does not adjust instantaneously. neither for
clearing the market. as does the price p on the final good
market. nor for passing on entirely cost changes. as does the
price w on the market for labour. We assume on the contrary
that these two influences are only effective with a lag. as
time passes.

2. FIRHS BEHAVIOUR AND AGGREGATE SUPPLY

2.1 The two regimes of the model
We assume that firms in the two sectors seek to
maximize profit.
regarding
all
the
prices
as
given
independently of their decisions. The assumptions that we have
already presented imply as well that no quantity constraints
can be rationally perceived by firms on the labour market and
on the market for the final good. On the contrary. the
intermediate good market. the price of which does not adjust
instantaneously. can give rise to quantitative rationing. We
assume that transactions on this market satisfy the principle
of voluntary
exchange and that the rationing scheme is
efficient (see Benassy. 1982). Using the concept of Benassyequilibrium. whereby perceived constraints are equal to actual
constraints. our model possesses two regimes. according to

44
whether supply

XS or

demand Xd

intermediate-good market.
where supply

of that

is

We call

good is

the

short

shortage

side

regime

a bottleneck,

of
the

and we

the
case

call the

reverse case the glut regime.

The behavioural
structure of

functions that

we use

to form

the

the model in each regime are derived from solving

the following programs

max pY - wLy

(8 )

X"

( 10)

max vX - wLx
H (Lx)

s.c. X
( 11 )

Under the
(11) cannot

vX

F (Ly,X)

s.c. Y
(9 )

stated assumptions,

constraints

(9)

and

be simultaneously holding with equality, except at

the equilibrium point of the intermediate-good market, which is

the

borderline

between

problem similar
of quantity
has been

the

two

disequilibrium

to the program (8) -

constraints, giving

previously dealt

regimes.

(9), with a larger number

rise to a four regimes model,

with by

the present

author (Azam,

1982), assuming that the intermediate good is imported.

2.2 The Glut Regime
The notional
corresponding to

behaviour of

the case

the firms

where X

<

XS,

in

sector

I,

is derived from the

following first order conditions:

(12)

pFL

(13)

pFx

Taking the

total differential of these equations, one can find

the notional demand functions for inputs :

d(W/ P \

d(V/PY

45

for labour
and
notional demands
words, the
In other
intermediate inputs in sector I are decreasing functions of the
s) and of the real price of good X (vIp) .
real wage rate (w/p
It follows that the notional supply of final good Y is a
decreasing function of the real wage s and of the nominal price
v, and an increasing function of the price p.
Except in the border-case where the intermediate-good
market is
sector I

cleared, the
is only

notional behaviour

effective in

of

the

the glut regime.

firms

in

In this case,

firms in sector II are rationed on their outlets, with:

(15)

Lx

H- 1 (X d

in self-explaining notation. Consequently, in this regime, the

total demand for labour LE

= Ly +Lx

is a decreasing function

of s and v, and an increasing function of p.

2.3. The shortage regime
In the

shortage regime,

results. First,

sector II

one gets

is not

rather

different

rationed, and its behaviour

results from the first-order condition:

(16)

vH'
One

thus obtains the

notional demand for labour by

this sector Lx, and its notional

differentiating (16), and using (5)

supply of output

H'

( 17)

price p,

(H' ) ..

Verbally, the
sector is

a decreasing
on which

notional demand
function of

for

the real

X-, .by

d(W/P

d(v/p>
labour

by

this

wage and of the

the nominal wage w is indexed. It is on the

46

contrary an increasing function of price v. The notional supply

of intermediate goods follows immediately, and has therefore
the same characteristics, as far as the signs of the impacts on
it are concerned.
Then, the
shortage regime
is characterized by
quantity rationing perceived by firms in sector I on their
purchases of
good
X. We, therefore, get the marginal
conditions
( 12)
( 13' )

pFL = w
pFx > v
R

The effective Ibehaviour of labour demand Ly and

I
final-good supply yP follows by differentiation, combining (12)
and (1)
( 18)

C~:)-hLJ (FL

-FLX)

("(W/P~

FLLFx - FLXFL

dX-

This gives the structural forms of the demand for labour - and
supply of output functions of the firms in sector I in the
shortage regime.
To get the reduced form of these functions, in order
to describe the global behaviour of the economy in the shortage
regime, one must replace dX- in (18) by its expression coming
from (17). One thus gets results which are very different from
the previous case. Let's begin by describing the total demand
R

for labour LP = Ly + Lx. One finds by using (17) and (18)(1)

v
v
v
dLP= (FLLH"--)-l(H"--+FLL-H'FLx)ds+H'(H'FLx-FLL)d(--)
(19)
p
p
p
In other words. in the shortage regime. employment is
an increasing function of the price of the intermediate good.
and a decreasing function of the real wage rate and of the
R

(1) We have dLP = dLy + dL x But. (17) gives dL x as a function

of d(w/p) and d(v/p). and (18) gives dL y R as a function of
d(w/p) and dX-. To get the result in (19). one must sUbstitute
in the equation of dLyR the value of dx- as a function of
d (w / p) and d ( v / p) found in (17).

47

price of the final good.

2.4 The reverse slope of the static supply curve
One could easily show that national income, here
equal to output of the final good in the ~hortage regimedenoted yP- has essentially the same comparative statics as
employment LP, which we have just presented. Thus, one finds
very naturally that income is a decreasing function of the real
wage rate s. The positive effect of the intermediate good-price
v on
national income
may be less intuitively obvious.
Nevertheless, one understands that a price increase for this
good is required to ease the bottleneck, by inducing the firms
in sector II to produce more. But, what is probably a priori
the least obvious fact in this regime, is that national income
is a decreasing function of the nominal price p of the final
good. This is probably a feature of indexation that Friedman
(1974) had not taken into account in his plea: here, inflation
is passed
on as
wage increases, and this reduces the
profitability of sector II, which sells its product on a
customer's market. Consequently, this induces it to tighten the
bottleneck.
We thus get in the shortage regime a final-good
supply curve which is not only inelastic with respect to its
price, as in the structuralist vision, but which has a negative
slope! This result, which is a priori little intuitive, does
not appear in Hicks or in the Structuralists. Nevertheless, it
follows from assumptions which are very close to theirs.
We thus have now a supply curve for final good
supply, and a demand curve for labour demand, whose slope
changes sign when the regime changes. A real price increase for
the intermediate good has a positive impact on output and
employment in the shortage regime, whereas it has a negative
impact in the glut regime. But this result is based on the
assumption of nominal price rigidity for good X, and must
therefore be regarded as a step in the direction of a more
dynamic analysis of the relationships between inflation and
unemployment.

48
3. THE LONG RUN SUPPLY CURVE
3.1 Adjustment of the intermediate-good price
In the

same way

rigidity assumption

as it

is never

for any nominal price.

easy to

warrant a

it is never easy to

describe then

how such a price evolves as time passes. We have

justified the

rigidity of v by reference to the "structuralist

vision".

invoking

intermediate

the precarious

good

sector.

customers' markets.
violent than

where

which
price

elsewhere. But

an instrumental

position

part. to

leads

of

firms

them

to

fluctuations

in

the

establish

must

be

less

this assumed rigidity only played

produce the analytical tools that we

are now going to use.

Obviously. we
rigid price

v. for

inflation and

do not

mean to

the analysis

unemployment. On

carryon

assuming

of the relationships between

the contrary. we are now going

to assume that this price is subjected to both the influence of

indexation.

like

the wage

of Supply and Demand".

the contrary
labour and

rate. and the influence of the "Law

like the price of the final good. But to

of wh_at we have assumed regarding the markets for

final good.

acts instantaneously.
act simultaneously

where only
we assume

as time

one of

the two influences

here that

the two influences

passes. even

if one regards it as

being quite fast.

Moreover. we
neo-structuralist

are not seeking to develop completely a

model

of

the

dynamics

of

inflation

and

unemployment. For. as we announced in the introduction. what we

are really

interested in

of perfectly

here.

expected inflation.

seeking to

analyse in

trend rate

of inflation

fashion of

a "Long-run

short-run dynamics.

our model

is to bring out the real costs

Consequently.

we

a relationship

(expected) . and
Phi 11 ips Curve" .

are

between

employment.
We thus

in

only
the
the

neglect the

and we need not specify the demand side of

the final good market in this model. and we only need to assume
that steady
which is

states. defined

equivalent) exist

by the
and

are

constancy of vIp (or w/p.

stable.

Hence.

we

are

implicitly assuming that inflation or deflation are financed by

the quantity

of money

growing at an appropriate rate. or else

49
that real

cash balances

demand. We

do not

problem set.
v as

do not

need to

affect at

all real effective

go into these details to solve the

We only need to assume that the rate of change of

time passes,

we denote v,

whi~h

is an increasing function

of the trend rate of inflation, denoted p, and of excess demand

for the intermediate good Xd
(20)

= V(p,X

function, and
of this

Vp and

that the

=a

Vx the partial derivatives of this

we assume that they are positive. The properties

function play

below, and

X-

X-), V(O,O)

We denote

we must

a crucial

part to establish the result

present them

dynamic adjustment

of

carefully. One
the

price

could assume
satisfies

the

following inequalities:

(21)

This squares
the

Vp < 1 and

<

above

(21) as

Vx

well with

exercises.

expressed in

<

But

one

a bit

00

the "vision" which underlies

can

regard

too restrictive.

the

conditions

We can

get the

sought result by assuming only that :

(22)

Vx
<

00

- Vp

The marginal

benefit in

terms of generality is that

one can assume nearly perfect indexation (Vp ----> 1), provided
Vx has

the required

hypothesis

aims

at

properties
ruling
p, on

out

for

(22)

perfect

the first

to
and

hold(2).

This

instantaneous

indexation of

v on

excess demand

has actually a positive impact on this price, o.n

hand, and to ensure that

the other hand.

(2) For example, the function :

v = ap + b(l - a)(X d - X-)
satisfies assumption (22) even if a ---> 1, provided b > O.

50

3.2 Unemployment and price fluctuations

the analysis

Then, restricting

(where v

p),

one gets

as a

to the steady states

"long-run Phillips

Curve" the

diagram of figure 1. The set of deflationary steady states (p <

0)

corresponds

to

the

inflationary steady
regime. To

glut

regime,

whereas

the

set

of

states (p > 0) corresponds to the shortage

compute the slopes of these two portions of curves,

one begins by computing the total differential of (20) with v

p. One then gets :

(23)

Vx

dp

- Vp

Then, using again the equations of Xd and XB, as they

are written in (14) and (17), one finds for a fixed s :

(24)

- XB)

d(X d

FL.L.

d(v/p)

P(H'

)2

< 0

H"v

Lastly, one can relate the long-run rate of inflation

given by

(23) to

regime, and
already

LE in

studied.

the level
the glut
Denoting

of employment

LP in the shortage

regime, whose properties we have

p,
E,
I
the
superscript

corresponding to each regime, one can get :

(25)

d(X d

dp

is positive

and finite.

negative. from
described in
of (vIp)

XB)/d(v/p)

dL1/d(v/p)
From assumption

we have

(24).
(25) is

(22), the

first term in parentheses

The numerator

Consequently,

of the

the

sign

second term is
of

the

slope

the opposite of that of dL1/d(v/p). But,

already seen that employment is an increasing function

in the

regime. This

shortage regime.

fact explains

and decreasing

the increasing

in the

function

~lut

relating

inflation and unemployment. when inflation is positive, and the

decreasing relationship

which prevails

when the

reverse case

51

prevails, as described in figure 1.

CONCLUSION
In the model that we have just analysed, steady state
employment is

maximum when

constant, for

a given

concerns the
Phillips

the price

value

of

of the

the

real

"neo-structuralist" equivalent

Curve".

It

assumptions, but
and deflation,
as compared

is

not

it equally

only

good

rate.

of the

non-vertical,

implies real

measuring these

final
wage

is
This

"long-run
under

our

costs for inflation

real costs by the loss of jobs

to the feasible maximum, for a given real wage. We

therefore end

up with

assimilates long-run
constancy.

Then,

an "anti-structuralist" doctrine, which

employment policy

even

"immutable", there

if

the

remains in

real

and
wage

steady-sta~e

is

price

regarded

as

the model some possibilities to

fight unemployment.
To get

to this result, we have assumed three markets

where price .determination follows different rules. Upstream, we

have assumed

a very

labour market.
final good
its

price.

fast indexation

Downstream, we

of the wage rate, on the

have assumed

a market

for the

which is cleared by the instantaneous adjustment of

Between

intermediate good

these

market has

two,

we

a price

have

assumed

reacting to

that

the

these

two

types of influences, but with a lag.

REFERENCES
Azam, J.P.

(1988), "Un modele neo-structuraliste d'inflation et

chomage", Revue d'Economie Politique, 98,pp. 78-89.

Azam. J.P.

(1982), "L'impact

commerciale en
1089-1114.

macroeconomique de

la politique

desequilibre", Revue Economique, 33, pp.

52
Azam.

(1986).

J.P.

Theorie

macro-economique

et

monetaire,

Paris, Nathan.
Basu. K.

(1984), The Less Developed Economy, Oxford, Blackwell.

Benassy. J.P.

(1982), The

Economics of Mark&t Disequilibrium,

New York : Academic Press.

Findlay. R.

(1973),

International Trade and Development Theory,

New York : Columbia University Press.

Friedman. H.
F~iedman.

(1974), Monetary Corrections, London:

lEA.

H. (1977), Inflation and Unemployment, London: lEA.

Hicks. J.R.

(1974), The Crisis in Keynesian Economics, Oxford

Blackwell.
Lewis.

W.A.

(1954),

"Economic

Development

with

Unlimited

Supplies of Labour", Manchester School, 22, pp.139-191.

Okun.

A. (1981), Prices and Quantities, Oxford

Sen. A.K.

(1966) ,

"Peasants

Surplus Labour",
425-450.

and

Journal of

Duallism

Blackwell.

with

or

without

Political Economy, 74, pp.

53

p
;II,

r----+.---L

L(s)

TOWARDS A MONETARY THEORY OF A PROCESS OF CHANGE(*)

Mario AMENDOLA (University of Roma) and Jean-Luc GAFFARD
(University of Nice - CNRS LATAPSES - Avenue Albert Einstein.
Sophia Antipolis 06560 VALBONNE CEDEX)

The divorce
of changes
shows the

in the

between monetary theory and the analysis

productive capacity

failure of

and monetary

disequilibria into

contemporary

crises

technological
full

impacts

call

and

economists seem

technique of

on

transformations
that the

phenomena only

the

while
radical

under

way.

appraisal of the

requires short

term

Schumpeter had already pointed out that

of

Monetary

significant restriction
the short

model. Thus.

attention

on thinking

of monetary

votaries

a proper

the

productive

to go

analytical frameworks.
: "Modern

of the economy clearly

modern economic theory to integrate real

: they

Analysis
assume

the

introduce

most

organization

and

production and the capital equipment as given (in

run}

( ... ).

The reader

should observe (a) that the

restrictive assumption in question excludes the very essence of

capitalist reality.

all the

phenomena and problems of which -

including the short run phenomena and problems - hinge upon the
incessant creation of new
that. because

of this.

and novel capital equipment. and (b)

a model

framed upon

this restrictive

assumption has next to no application to questions of practical

diagnosis. prognosis.

and above

all. economic

policy

unless

reinforced by extraneous consideration" (p. 280).

This paper
problem. restoring
capacity at
both a

is an attempt to take into account such a

the analysis

the heart

of changes

of monetary

different representation

in the productive

theory. This

will require

of the process of production.

with respect to dominant one. and a different interpretation of

the concept of liquidity and hence of what has to be understood
as liquidity

preference. We

shall thus

be able to work out a

(*) This paper stems from: M. Amendola and J.L. Gaffard "The
Innovative choice.
An Economic Analysis of the Dynamics of
Technology" Blackwell. Oxford. 1988.

56

truly sequential
up for

framework which will provide the adequate set

the analysis

of a

process of

change

in

monetary

economy.
The analytical framework

Sequence analysis

requires in

the first

place that

attention be paid to the fact that production takes time.

This is
portrays the
over time

process of

in a Neo-Austrian model, which

production as a scheme for converting

a sequence of labour inputs into a sequence of final

output (Hicks,
and

what happens

the

1973). The

technical

model highlights the time structure

intertemporal

complementarities

of

the

process of

production, thus making it possible to focus on the

"making of

the machine"

of change

as the relevant moment of the process

of productive

capacity.

In this context the capital

goods become the particular expression of each kind of process,

within which
cannot be

they are

identified with
seen as

produced, cannot

transfered;
a stock

a result

process of

classical tradition
capital and
propose to

of physical

(although an

production.

That

kept

essentially

processes of

it and

goods which are actually

intermediate
is

why

result)

going

alive

by

the

shared

by

Schumpeter

conceive capital

available for

exist outside

"capital", therefore, cannot longer be

as a

financing the

Austrian

fund,

carrying them

of

the

to

the

theory

(1934)

made

labour required

production and

back

up

of

of
we

money,

for starting the

out (Wages

Capital, however, does not coincide with money. Money,

Fund).
in fact,

can also be stored.

As was
essential in
where there

pointed out

real sequence
is a

process conceived

periods makes

(1973),

economies, that

sequential learning
~s

money
is,

in

becomes
economies

that implies a decision

a related sequence of choices and not as a

predetermined, although
dates, succession

by Hahn

referring at transactions at different

of decisions.

it possible

environment than

is

becomes relevant

less as

to

In fact, when the sequence of

get

available

the character

of liquidity

"liquidity is

not a

in

more
the

information
first

on

period,

the
money

a contingent store of value than for

attached to

property of

it.

a single

In

this context,

choice;

it is

57
matter of

a sequence

1974. p. 38)

of choices.

a related

sequence" (Hicks

; the reference to liquidity preference is used to

explain choice between (relatively liquid) financial assets and

( relatively illiquid) real assets.
Thus. when
that something
spec i f i ed .

there is

new and

the feeling

or the

different. although

perception

not

yet

clearly

is

going or needs to happen. accrued preference for

liquidity. as

the attribute of assets whose acquisition can be

easily revoked

as opposed to assets that on the contrary imply

commitment to

a given

course of

action. emerges

as the most

relevant consequence of the appearance of uncertainty.

the more

the agents

decrease in

expect

to

the confidence

learn.

the

greater

In fact.
is

the

they have in the existing state of

affairs. because the greater is the likelihood of a substantial

revision of this in the future.
expectations. no
follow.

reflecting

change

it

the old

in the

the appearance

will

considering the

A modification of the long term

longer formed
in

fact

be

one. although

of a
a

advisability of

same

way.

case for

signal

then

qualitative

that

substituting a

will
agents

are

new model for

the new relations on which to rely. and

which will define the new model. have still to be learned. This
will result in a search for flexibility. expressed in the first
instance. as
demand for

we have

already mentioned. by an increase in the

flexible positions - that is positions allowing for

waiting, postponement
liquid assets

or quick revisions of decisions. Holding

(money) appears

then as

the

most

appropriate

the

sources

answer.
However. when
process

df

learning

the nature
involved

and

are

considered.

of

the

the

problem

appears in a different light.

Holding liquid
postponement

of

implies. may

in fact

detailed

concerns existing
is expected

assets, with
a

binding

correct

decisions

response

opportunities, about

of the

that

when

this

learning

which more information

to appear. or the sequential arrival of attractive

new opportunities

as the

Flexibility.

in

to diminish

the options

then a

and

be

the possibility

result of

this case,

choice that

the mere passage of time.

has a (static) significance:

for the future".

does not

reduce the

"not

A flexible choice is
future

alternatives

58
associated

with

the

it

most

liquid

in

choice,

this

perspective, is also the most flexible one.

not

of

But when

learning can

waiting,

holding

only be

liquid

the result of doing,

assets

is

no

longer

the

appropriate response to the search for flexibility, because new

opportunities will

not become

available independently

of the

decision makers' actions. Flexibility, then, acquires an active

(dynamic) character: "to increase the options for the future",
; and a flexible choice will be not so much a choice which does
not diminish
as a

the capacity to respond to oncoming opportunities

choice that

will itself

bring about

new opportunities,

thus enlarging the gamut of future options.

The counterpart
active interpretation
passage from

the

of this

of the

passive

concrete expression

shift from

concept of

posture

of

a passive to an

flexibility
waiting,

is

which

the
finds

in an accrued preference for liquidity, to

the active posture represented by an innovative choice, that is

a choice

to start

and carry out altogether, new and hence not

yet clearly defined processes of production. The most important

problem then

becomes that

of the viability of such a choice:

which, as we shall see, depends mainly on the constraints which

will affect

the innovative

their evolution

in time.

disequilibrium, as
expectations to

processes of
In fact,

a result

which

the

of a

production,

the appearance
modification of

existing

capacity

is

and

on

of a stock
long

term

no. longer

adjusted, brings also automatically about a flow disequilibrium

(an inequality
thus makes

between current

a financial

demand and current supply).

It

constraint appear which (together with

the human resource constraint) is to playa central role in the

articulation and

in the

effective evolution of the process of

chance as a sequential process.

Money plays
framework.
on the

It

breaking of

the previous

an essential

makes it

possible

role

in

this

analytical

in the first place to figure

a sequence without being obliged to assume

existence of

an exogenously

given technological

advance. The increased demand for liquid assets, as a result of

a modification

of long term expectations which reflects a loss

of confidence

in the

existing state of affairs and the search

for something

new and different, acts in fact as a signal that

59

allows the
the

case for

model,

to

disequilibrium

acquire

kind of

change, originating within

analytical

resulting

expectations can
points to

a qualitative
from

take a

relevance.
modification

monetary form,

the qualitative

The

and this

stock

in

the

immediately

nature of the change required. This

disequilibrium, on the other hands carries its effects

down the

sequence in

in which

only real

a different way with respect to the case

stocks are

difference between

considered, thus stressing the

process

interpreted

as

qualitative

change and a process treated as a quantitative adjustment.

But the
making the

presence of

case for

acquire analytical

money is not only essential for

a qualitative
relevance.

It

change to
also plays

appear

and

to

a marked role in

determining the viability of the process of innovation which is

the expression of such a change: as we shall see in particular
when dealing
of a

with the effects of the financial constraint (and

policy aimed

at modifying

it) on

the evolution

of the

economy on an innovative path.

The analysis

of a process of qualitative change will

then appear essentially as a monetary analysis.

A sequential model
The essential feature of a process of production of a
Neo-Austrian type

is its articulation in time. Let the profile

of a unit process be described by

[ao,

where

a' k

and

c'

81, . . . 8

= (ak
= [C~+l"

1,

The a's
required for

.l, ... 8

rn + H

ak2,'" aka) ,

c]

= 0,1, ... m+M

. C~+H]

are

the

(heterogeneous)

labour

inputs

starting the process (at time 0) and for carrying

it out through the m+M periods of its life. and the c's are the
final outputs
the m

appearing from period m+l onward, that is, after

periods which

phase of

are assumed

to make

up the construction

the productive capacity. Each given profile defines a

particular technology.,
In steady-state.
the processes

will have

with an

unchanging technology, all

the same time profile; new processes

60

will be increasing at a constant rate g (which will also be the

growth rate

of the

economy) and

the age

composition of

the

population of processes will remain constant.

The rate of starts of new processes in period t, will
then be,
(l+g)T xo(t-T)

xo(t)

while the

other relevant

magnitudes of the economy, always in

period t, will be :
- the employment of the different types of labour, LD(t), given
by
aoxo(t) + Ax(t)
where A = (a"a 2

- total output,

reckoned as the money value of aggregrate final

a~+M)

and x'(t)

output given by

where PR(t)

is

the

price

(routine) processes,
and

equal

[X~+l(t)

, ...

to

the processes

the

call it R(t),

one

x~+M(t)l

of

in

output

of

the

existing

in terms of money (constant

steady

state),

and

xu(t)

is the row vector whose elements represent

of production

in the

m+l to m+M periods of the

utilization phase respectively.

- the

Wages Fund,

that is

the amount

of financial resources

(money) required by the labour applied to start and to continue

carrying on the processes of production still alive, given by,
WIt)

w'(t)LD(t)

where w'

(Wi'

W2, . . . ws

exogenously determined
correspond to

is the vector of money wages rates,

(and constant in a steady state), which

the different

types of

labour employed

in the

processes of production
- the

money value

of the aggregate demand for final output at

the ruling price given by

61

- the producers' financial resources available for final demand

(consumption out of profits), given by
Q(t)
which

one

is

of

the

components

of

P*(t)

and

which

is

exogenously determined
-' the exogenously determined inflow of money
~M( t)

which is the difference between loans and repayments.

The relations
economy reflect
each period

between the relevant magnitudes of the

the sequential structure of model, both within

and between

successive periods. In a steady state

these relations will be :

- Q(t)

~JHt)

which shows
financing

that the
the

consumption)

processes
have

proceeds of

resources available to the producers for

an

of

production

internal

the sales

source,

(and

their

represented

by

own
the

of the previous period - equal to P*(t-

1), that is to the value of the aggregate demand in that period

- and

an external

source

~M(t),

where

M(t) increases from

period to period at the rate g ;

P*(t)

W(t) + Q(t)

which shows

that current demand depends on the money income

actually devoted to the purchase of final output ; in a steadystate with
the

only routine

income

processes it is equal to t'he whole of

available,

since

there

is

no

search

for

flexibility;
P(t)

= eP*(t) =

(l+g)P*(t-l)

which shows

that the

each period

is determined

value of

money value of total final production in

aggregate final

on the
demand in

basis of the expected money

the same period, and that

62
the latter results from a growth rate equal to the one realized
in the

previous period

(i.e. the constant rate g in a steady-

state) .
Assume now
state sequence.
expressed by

that there

This will

is a

breaking of the steady-

result in a search for flexibility.

a reduction

in the

proportion of money actually

devoted to production and/or consumption.

W(t) = r(t)[p-(t-ll +
P-(t)

<

A flow
Rs(t)

i1M(t) - O(t)]

o(t)[W(t) + O(t)]

where r(t). ott)

P(t) >

in such a way that

disequilibrium will then appear in period t

P-(t) -

which implies

R(t) -R-(t)

assumed that

that stocks

of final

output-

- are involuntarily accumulated. as it is

price changes

do

not

take

place

within

each

period.
This
considered by
formulating

short

(rightly) as
case the

disequilibrium
the producers
terms

the signal

producers give

appearance of
consists in

either
result of

expectations
of a

(i.e.

mistake
eP-(t.

structural change.

a quantitative

(wrongly)

be

in
or

In the first

interpretation to the

an excess supply and consequently their reaction

a simple

investment targets
according to

can
as the

revision of

the

final

production

and

of processes of production still carried on

the established technology. The economy continues

to follow a routine path.

a qualitative

In the second case the producers give

interpretation to

what is

going on,

and their

reaction then consists in a revision of the decisions on how to

invest, by setting off on an innovative path.
When the

economy keeps following a routine path,

relevant magnitudes and the relations between them will be

(1) LD(t)

aoxo(t) + Ax(t)

(2)

P(t)

PR(t) [R(t) + Rs (t-l)]

(3)

W(t)

w'(tlLD(t)

(4) P-(t)

(5) ott)
(6)

W(t)

PR(t)R-(t)

P(t) - W(t)

its

63

where SR(t-l)

[R"(t-l) - R(t-l)]

unvoluntarily accumulated

by the

shifted to

period so

the following

0 are the stocks of money

>

consumers and

automatically

as to increase the demand

for final output in it,

(7)

pet)

o(t)[W(t) + PR(t-l)SR(t-l) + OCt)]

(8)

pet)

eP"(t) = [1 + g(t-l)]P"(t-l) + PR(t-l)SR(t-l)

where KR

is a price reaction coefficient, and which shows that

price changes

in each

period reflect the appearance of excess

demand or of excess supply in the previous period ;

(10)

w'(t)

(under

PR(t)W'

the

hypothesis

of

fixed

real

wages) .
When the economy starts moving on an innovative path,
the profile
change

of the

at

each

expression

of

(innovative) processes
successive

the

step,

ongoing

of production will

this

change

technological

and

being

the

productive

transformation. This is described by

[ba(t), b 1 (t), ... b n + 1 (t), ... bn+N(t)
where b'

(t)

[b k

(t), b k 2 (t),

; d(t)]

... b k

...

(t) ]

0,

1, ... n+N

and d' (t) = [d n + 1 (t), ... dn+N(t)]

80th the

input and

the output coefficients become a

function of time. More precisely, they become a function of the

workforce's acquaintance
which, although
mere passage
number of

with

taking place

of the

new

in time,

latter but is,

innovative processes

the moment

the

an innovative

productive

problems,

does not depend on the

in turn, a function of the

of production carried out from

choice was

first

made

up

to

the

particular period considered.

This process
helping to
processes of
and the

of research

specify on

the way

and experimentation, while

the profile

of the innovative

production, at the same time causes the upgrading

enrichment of the human resources that are involved in

those processes
new skills
possible to

; thus

leading to

and qualifications
devise

and

the appearance of entirely

which will

implement

themselves make

altogether

new

forms

it
of

64
production (and

consumption). A

productive options,
resources,
sets in

is

process of

associated with

creation

a modification

of
of

new
human

fuelled in this way by the learning process that

as the

result of

an innovative

choice

and

of

the

carrying on of innovative processes of production.

Let LS(t)
elements represent
labour resource
structure of

at time

fact the

labour availability

the different

the human

successive period.
is in

be the

skills of

vector

whose

the heterogeneous

t. A different vector will define the

resource

Each element

result of

at

the

beginning

of the vector,

of

each

in each period

demographic and educational factors

and of the above mentioned learning process, that is :

where gh

is the

proportion of

workers of skill h accruing in

period t owing to the demographic and educational factors ; and

fh and f h -

are the subtraction and the addition to the workers

of skill

h due

process. We

upgrading resulting

assume, further,

existing ones
size of

to the
in each

from the learning

that new skills are added to the

successive period, so that not only the

the labour availability vector's elements but its very

dimension is modified.
Given the
the skills
devise, the
on the

the processes of production that

labour availability

type and

be carried

range of

of the existing human resources make it possible to

vector sets a constraint both

the number of the processes that can actually

on in

each given

period: as only those processes

can be started and/or kept alive (and in the amounts) for which
the

req,uired

proportions.

labour
However,

inputs
this

are 'available
constraint

is

in

the

modified

right
by

the

process of learning and is made less stringent as more and more

innovative processes are carried out.
From the

moment T

= 0,

when an innovative choice is

first made, onwards, there will be three distinct phases which,

following Hicks

(1973), we

shall call

the early phase and the late phase.

the preparatory phase,

65
During the
that is.

when

production has
magnitudes of

the

preparatory phase
output

not yet

of

the

appeared on

the economy

(from T

innovative

=m

0 to T

processes

the market)

of

the relevant

and the relations between them will

be :

(la) LD(t)

Ax(t) + bo(t)Yo(t) + B(t)yC(t)

where yo(t)
is the rate of starts of innovative processes and,
yc(t) is the vector [yC , (t),yC n (t)] whose elements represent
the

innovative

processes

of

different

age,

still

in

the

construction phase, carried on in period t,

(2a) P (t )
w' (t) L (t)

( 3a) \.[( t )

(4a) P-(t) = PR(t)R-(t)

(Sa) OCt)

pet) - wet)

(6a) wet)

r(t)[p-(t-1) - PR(t-llS R (t-1) +

(7a) P-(t)

~M(t)

- OCt)]

o(t)[W(t) + PR(t-l)SR(t-1) + OCt)]

(8a) pet) = eP-(t) = [1 + g(t-1)]P-(t-1) + PR(t-I)SR(t-l)

(9a) PR(t)

PR(t-1) + KR (t)PR(t-1)[R-(t-1) - R(t-I)/R(t-I)]

(lOa) w' (t) = PR(t)W'
During the

that is

early phase

(from T

m+1

to T

= m+M

when the output of both the routine and the innovative

processes of production is on the market) we shall have:

(Ib) LD

Ax(t) + bo(t)Yo(t) + B(t)y(t)

where yet)

[Yl (t),

elements represent

Yn(t), ... Yn+N(t)]

all the

is

innovative processes

vector

whose

of different

age carried on in period t

(2b) pet)

where y"(t)

PR(t)[R(t) + Rs (t-1)] + Px(t)[d'(t)y"(tl + Is(t-I)]

= [Yn+l(t), ... Yn+N(t)] is a vector whose elements

represent the innovative processes of different age, yet in the

utilization phase, carried on in period tr
and Is(t)

I(t-I) - I-(t-l)

>

0, where I

d'y" is the output

of the innovative processes and 1- is the amount of the same

output actually absorbed by the market

66
(3b) W(t) = w'(t)LD(t)

(4b) P-(t)

PR(t)R-(t) + Px(t)I-(t)

(Sb) Q(t)

P(t) - W(t)

(6b) W(t)

r(t)[p-(t-l) - PR(t-l)SR(t-l) -Px(t-l)Sx(t-l) +

~M(t) - Q(t)]

where Sx(t-l>

(7b) P-(t)

I-(t-l) - I(t-l) >0

o(t)[W(t) + PR(t-l)SR(t-l) + Px(t-l)Sx(t-l) +

Q(t)]

(8b)

eP-(t)

P( t)

[1 + g(t-l)]P-(t-l) + PR(t-l)SR(t-l) +
Px(t-l)S:r(t-l)

PR(t)
(9b)

px(n+l)

PR(t-l) + KR(t)PR(t-l) [R-(t-l) - R(t-l)/R(t-l)]

= 6(n+l)P(n+l)/CI(n+l)

where CI(n+l) is the productive capacity of the

innovative output, inherited from the past at time n+l
P:r(t)

p:r(t-I) + K:r(t)Px(t-l)[I-(t-l) - I(t-l)/I(t-l]

t > n+l, where Kx(t) is a price reaction coefficient

(lOb) w'(t)

= p(t)w'

where p(t) is a price index

= T(t)P(t)/PR(t)
= 6(t)P(t)/P:r(t)
R-(t) = T-(t)P-(t)/PR(t)
R(t)

I(t)

(lib)

I-(t)
where Toutput

= 6-(t)P-(t)/Px(t)

and 6of

express the

the

routine

respectively, and
T(t) + 6(t) = l,T-(t) + 6-(t)
and 6(t)

= 6-(t-l)

During the
all the
output of

the

the innovative

(from T

have been

= b o ( t ) y ( t) +.B ( t ) y ( t )

= m+M+l

scr~pped

processes keeps

market) we shall have

( 1 c) L D( t)

innovative

for

the

processes

I, 6-(t) > 6(t)

late phase

routine processes

preferences

consumers~

and

onwards when
and

flowing on

only

the

to

the

67

(2c) P(t)
(3c) W(t)

Pz(t)[d'(t)yU(t) + I. (t-1)]
w'(t)LD(t)

= Pz(t)I(t)

(4c) P-(t)
(5c) Q(t)

P(t) - W(t)

(6c) W(t)

r(t)[p-(t-1) - pz(t-1)Sr(t-1) +

~M(t)

- Q(t)],
with r(t) >1

(7c) P-(t)

o(t)[W(t) + Pr(t-1)Sr(t-1) + Q(t)],

with o(t) >1

= eP-(t) = [1
(gc) Pr(t) = Pz(t-1) +
(10c) w'(t) = Pz(t)w'

(8c)P(t)

On the

+ g(t-1)]p-(t-1)

+ Pr(t-1)Sr(t-1)

Kr (t)Pr(t-1)[I-(t-1) - I(t-1)/I(t-1)]

innovative path, the solution of the model in

each period - and hence the pattern of evolution of the economy

- is

given by

vectors x(t),

the values

of the

u(t), y(t),

whose elements

scalar yo(t),

v(t), the

represent the

and

latter being

of

the

the vector

innovative processes scrapped at

time t. The processes of production are scrapped. if necessary,

in an- order reflecting a flexibility criterion that focuses on
expected final
time) as

output (both

growth

the

(first the

a reduction

processes will

nearness

expected

stringent

in

financial

a reduction of demand, and hence of

processes scrapped will be those in the

utilization phase
there is

less

and its

there is

constraint. When
the

of

an index

its amount

be

older, then

of resources,

the

first

to

the younger).

When

the rate of starts of new

be

cut.

followed

by

the

processes in the construction phase (first the younger. the the

older).
The monetary

analysis of an innovative choice versus

a routine choice
The analysis

will be

stochastic simulations
would happen

developed

that will

under alternative

exogenous variables

and the

by

means

of

non-

make it possible to see what

assumptions in

parameters of

terms

the model,

of

the

set at

values corresponding to various scenarios whose consequences we

would like

to evaluate. This has been done in order to explore

the articulation

of the

sequence in

time that represents the

evolution of the economy under the alternative hypotheses made.

68
in order to bring to light the relevant moments and connections
of this sequence.
On a

routine path.

The labour

is already perfectly

adjusted to the prevailing technology, and there is no learning

; the

only existing constraint is therefore represented by the

available financial
activity of
affected

resources that

the economy,

mainly

by

determine

the

levels

of

and whose amount and destination are

the

values

taken

by

the

exogenously

determined inflow of money AM (and by the take out 0).

In a first scenario the growth rate of
adjusted to

the lower

final demand
and/or

existing state

is

level resulting from an initial fall

- which

reflecting,

(and 0)

~M

follows a

reduction in

the value

in

of r

in the model, a loss of confidence in the

of affairs

- and hence forward adjusted period

after period to the actual growth rate of the economy, whatever

this happens

to be.

The growth

of the economy is immediately

stabilized, unless there are further reductions in rand/or o.

This lower
processes of

growth rate

production

scrapped, and

this

in

will

however,

the

go

implies

utilization

on

as

long

that

phase

as

the

some

must

be

inherited

productive capacity is not fully adapted to the expectations of

final demand.
in the

During this period, the funds that were invested

processes scrapped are lost and producers are left with

the corresponding
degree of

debts, there is therefore an increase in the

indebtedness which, sooner or later, will call for a

reduction in

the growth

stronger financial
investments and

rate of

AM. This

constraint, a

a fall

smaller

in current

will result in a

Wages

Fund,

production. The

lower

immediate

consequence is the reappearance of a flow disequilibrium in the

current

period

the

disequilibrium down

carrying

the sequence

of

the

will then

resulting
cause the

stock
growth

rate of the economy to keep falling in each successive period.

In a
the rate

second scenario, A M (and 0)

at which

the sequence
actual growth
situation. Two

the economy

which has

phases can

economy converges

was faring before the break in

brought about

rate, thus

is kept growing at

trying to

the

reduction

return

be considered.

to

the

in

the

original

In the first one the

to the original growth rate through a damped

oscillatory movement.

In a second phase - when the oscillations

69
in the rate of starts that have occurred during the first phase
begin to

effects

produce their

becomes

explosive

enormously:

and

disarray. The

attempt to

maintaining the growth of

into difficulty

level

the

the economy

the
of

indebtedness

gets into

an intolerable

get

to

back

movement

oscillatory

potential

increase
state

of

growth

by

L\.M (and 0) at the original rate runs

because it

induces an

increasing variance in

the age structure of the processes of production, which becomes

a source

of instability

fluctuations in

in the

the productive

sense that

it

capacity that

brings

about

become more and

more pronounced as time goes by.

On an

innovative path,

the effects

of

learning

process associated with the carrying on of innovative processes

of production

must be

human resources
releasing of

taken into

by bringing

account.

into higher

the human constraint.

Learning

affects

skills which imply a

It is therefore on the pace

and the intensity of the learning process that the viability of

an innovative

choice finally

depends. However,

the

learning

process, and hence the pattern of evolution of the economy, are

deeply influenced by the policies followed that, as in the case
of a routine choice, are stylized by the different values given
to the growth rate of

L\.M (and 0).

In a first scenario the growth rate of

made to

follow the

there is

a human

resources

actual growth
constraint at

available

effectively

be

for

used.

The

rate of
work, not

employment

in

resulting

fall

L\.M (and 0)

the economy.
all the

When

financial

production
in

is

final

can
demand

triggers off a succession of excesses in supply and demand that

reflects the

way in

which short-term expectations are revised

in the

model, and

of the

economy that,

preparatory phase
when learning
disappear. From
is stabilized,

that causes fluctuations in the growth rate

although damped,

and even

in production

beyond -

last all

through

the

that is up to the moment

causes the

labour constraint

to

this moment onwards, the growth of the economy

and none of the processes of production must be
if (and When) an increase in the value of r

scrapped. Later on,

and a above unity allows an over - functionning of the economy,

the growth

rate of

L\. M can be reduced below the actual growth

rate of the economy {as the internal resources gradually become

70
available to replace the external ones). thus making possible a
reduction of the indebtedness.
scenario portrays

The second

AM
If

inchanged. that is at the level i t had before

(and 0)

fall due to the reduction in the value of

the initial

o.

to return

to its potential growth by keeping the growth rate

the economy
of

the attempt

the

human

fluctuations in

constraint

is

particularly

and/or
the

strong.

final output may not only lead to fluctuations

in the rate of starts of new processes but. given the scrapping

rules followed,
still in

also a

the phase

serious increase
lenghtening of

scrapping of

of construction.

in

the

the period

degree

of

processes of
such as

production

to result

indebtedness

and

in a
in

of time that will elapse before the

output of the innovative processes will start flowing on to the

market. The path followed then,

is clearly not viable.

Conclusion
While it
of money

one is

133), the

both the

the sequence

undergone by
appears as

"not true that by getting rid

automatically in equilibrium" (Hicks.

analysis carried

for determining
time of

is certainly

which portrays

the economy.
a policy

out shows
nature and
In

aimed at

this

1973 p.

that money is essential

the actual
the

process

context

rendering an

evolution in
of

change

monetary

policy

innovative choice

(i.e. a qualitative change) viable according to given criteria,

and its most relevant aspect is represented by the articulation
in time

of the

intervention rather

is, not so much how much as when.

than its intensity : that

71

BIBLIOGRAPHY
Amendola, M. and J.L. Gaffard (1988), The Innovative Choice. An
Economic Analysis of the Dynamics of Technology, Oxford,
Basil Blackwell.
Hahn, P.H.

(1973), On the Foundations of Monetary Theory,

Parkin

and

A.R.

Economics,

Nobay

London,

(eds.),

Longman

Essays

Group,

in

in M.
Modern

reprinted

in

Equilibrium and Macroeconomics, Oxford, Basil Blackwell.

Kicks, J.R.

(1973), Capital and Time, Oxford, clarendon.

Hicks, J.R.

(1974, The

Crisis in Keynesian Economics, Oxford,

Basil Blackwell ..
Lundberg,

E.

(1937),

Studies

in

the

Theory

of

Economic

Expansion. Reprints of economic classics, New York, A.M.

Kelley 1964.
Schumpeter, J.A.

(1934),The Theory

of Economic

Development,

New York. Harvard University Press.

Schumpeter, J.A.

(1954), History of Economic Analysis, London,

Allen and Unwin.

RULING OUT MULTIPLE EQUILIBRIUM PATHS IN MONETARY OPTIMIZING

MODELS : NECESSARY AND SUFFICIENT CONDITIONS
Didier LAUSSEL and Antoine SOUBEYRAN
Universite d'Aix - Marseille II
CEFI Chateau La Farge
1. Introduction
The monetary optimizing model, where a representative
consumer, facing
an

parametrically a path of money prices, solves

intertemporal

infinite

horizon

problem

and

where

the

equilibrium path of prices is defined by the condition that the

exogeneous path
this problem,
attempt

of money and consumption goods supplies solves

was first

to

provide

assumption

widely

models. Up

to now

original
confusing.

It

used

been contradictory

can't be

[1974]

the

saddle

original

ruled out

path

expectations
from Brock's

and

Rogoff

[1983],

of

Gray

is rather

Obstfeld

[1984],

utility function -zero monetary

ruling

[1983]

an

rational

framework the

and incomplete.

Rogoff

in

which evolved

exception

this simple

possibility of

allowing to

the

the separable

Even in

Brock's [1974]

for

monetary

(Obstfeld

has, with

Obstfeld and

in

Brock

and Rogoff [1986], Obstfeld [1984])

concentrated on
growth case.

by

justification

the literature

contribution

[1984], Obstfeld

been the

introduced

out

proved

conjecture,

results have

The debated

question

divergent
that,

in

price

has

paths.

opposition

hyperinflationary

to

paths

on optimality grounds but gave a condition

rule them out on feasibility grounds. However they

as well as Gray [1984] wrongly asserted that the transversality

condition is

sufficient to

rule out

hyperdeflationary

paths

(for a correction see Obstfeld and Rogoff [1986]).

Obstfeld has shown that,

in the non separable utility

function case, price paths may be stable (though convergence to

the

steady

state

equilibrium paths

may
may not

occur

only

be unique

in

infinite

time)

(Obstfeld [1984]).

and
In a

companion paper (Laussel and Soubeyran [1987]), we demonstrated

that there may also be cyclical equilibrium price paths.
This paper

deals with

the uniqueness problem in the

general framework of a non-separable utility function,

non zero

74
monetary growth
necessity and

model. The

first step we take is to prove the

sufficiency, for

optimizafion problem,
Value Condition

of the

the

consumer's

intertemporal

Euler equation and of a Terminal

(which is the same as Benveniste - Sheinkman's

[19821, but the conditions of their theorems are not met in the
monetary model

studied here). Then we proceed to study the set

of equilibrium

price paths.

We first

show that assuming that

consumption is never a Giffen good is enough to rule out stable

and cyclical

price paths.

conditions and

a set

of

unstable (divergent)
different from

necessary

price paths.

come from

analyses

[19831, Gray

give a

of

which substantially

conditions

sufficient

to

rule

out

are markedly

in the existing literature.

the mistakes which crept into the

hyperdeflations

[19841 and

set of

These results

corresponding ones

These differences
previous

Then we

from the

modifies the

(Obstfeld

and

Rogoff

non separability assumption

study of hyperinflations. We

generalize the existing results on hyperinflations (which first

appeared in

Obstfeld and

importance for

prove entirely

the long-run

demonstrating

the

between real

money and

consumption.

new results showing the importance for

optimality of

decrease of

[19831),

assessing their feasibility of the complements-

substitutes relationship
Then we

Rogoff

hyperdeflations of

the

speed

of

the implicit rate of return on money holdings when

real balances tend to infinity.

2. The model
Let us
an infinitely
The consumer
continuous

consider an

economy composed of two agents :

lived representative
faces a

function

constraint M(t)

price level
of

P(t)

time
(y -

consumer and a Government.

P(t) which

and
c(t

an
+

is a

piecewise

instantaneous

budget

T(t), where M(t), c(t),

T(t) and yare respectively, at time t, the money holdings, the

consumptipn level,
the

consumer's

the net

constant

transfers from
exogeneous

the Government and

income.

The

Government

chooses a rate of growth 9 of the money supply Me and transfers

the money

created to

constraint 9MS(t)

the consumer

according

to

its

budget

T(t).

We assume that the representative consumer wishes to

75

mIt~

MaxI: u(c(t).

e- 6 t dt

s.t.

(1)

M(t)

(2)

( M( t). c ( t

(3)

M(O)

u :

R2+

R+ the

6 E

PIt)

(0)

(y E

cIt~

+ T(t).

R2+.

Mo.
->

R is the instantaneous valuation function.

constant psychological rate of discount and mIt)

M(t)/P(t) are the real money balances.

Let (M-(t). c-(t) be a solution of (0). We assume the
goods and money market always clear
(4 )

c-{t)

y. all t E [0,00]

(S)

WIt)

M-(t). all t E [0.00]

Note

that the

budget constraints

together with the

equilibrium conditions imply that M-(t) = 8M-{t), all t.

Assumptions
(A.1) u

R2+

-> R

is twice-continuously

differentiable and

strictly concave with non-negative first derivatives.

(A.2)

I:

u(c-(t), m-(t e- 6 t dt < 00

(A.3) Consumption and real money are normal goods.

(a)

u~uc~

ucu~~

>

O.

(b)

ucuc~

u~ucc

>

O.

i.e.

(A.4) c-{t) is a strictly decreasing function of PIt) for all t

~

O.

(A.S)

(a)

y > O.

(b) Mo > O.

Assumption (A.4)
which.

despite

the

necessarily satisfied
price level

enters the

states a

normality

very natural
assumption

requirement

(A.3)

is

not

otherwise in a model like this where the

utility function

via the

real

money

balances: consumption must not be a Giffen good. We shall show

that

(A.4)

is

equivalent

to

direct

assumption

on

the

76
properties of
we make

the utility

no assumption

function:

on the

conditions" since

function case

such assumptions

properties of

u=/u c

in the

non-separable utility

would be

are the

which

Note that

limit properties of u= (the so-

called "Inada

o.

+ mu c = >

Uc

the

useless:

only to really matter, do

not follow from those of u=.

3. Optimality conditions
The first
and

sufficient
the

take is now to find the necessary

conditions

consumer's optimal
with

step to

for

{c(t),

M(t)}

to

solve

the

control problem (0). The difficulty is here

transversality

condition

at

infinity.

While

the

sufficiency of Lim [A(t)e- M(t)] = 0 has been proved a long

t -> 00
time ago under suitable concavity-convexity assumptions (Arrow
6t

and Kurz

[1971]), Halkin's

"terminal value
the concave

[1974]

condition" (TVC)

valuation function

example
is not

shows

that

this

always necessary.

In

case, Benveniste and Sheinkman

[1982] showed the necessity of the above TVC. However they used
assumptions which
model:

their "tpchnology

and their
when the

are not

always satisfied
set" depends

"interiority assumption"
monetary growth

level tends

rate is

in

the

monetary

on time in this model

does not

hold at infinity

negative or

when the price

to zero. No simple generalization of their results

is available.

Hence we

prove here

new

theorem

providing

necessary and

sufficient conditions for Problem (0). From this

theorem we shall deduce the necessity and sufficiency of BS'TVC

in our model.
Let h(.)
parameter

and

{c(

be a
t) ,

fixed function

'"
M(t ) }

of

candidate

time,

solution

{c(s) a h (s)} , all s

'"
generates the state variable M(t) + a ~: P(s) h(s) ds.

Obviously the

control path

ERa
to

(0).

E [O,t],

Definitions
(D.1) An

-v

ALhLl~~~h~e

is a

comparison path

{c(t)

ah(t), M(t) + aI: P(s) h(s)ds} corresponding to a function

h(.) and
family) of

a parameter

value a.

all arbitrage

function h(.).

1\ (h(.) is the set (or

schemes corresponding to a given

77

(0.2) n(h(.

is the

corresponding to
(cit), M(t)}

set of

a given

feasible

arbitrage

function h(.),

schemes

i.e. the

set of

I\(h(. such that (cit), M(t E R +, all

t E [0,00].
(0.3)

A(h(.

is the set of

for a given function h(.),

parameter values such that,

(C'(t) - ah(t), M(t) + a

I:

PIs)

h(s)ds} E n(h(. .
Only the
other

families

feasible paths

:f

that n(h(.
interest.

(or.

of

arbitrage

than the candidate

( a)
(b )

of

Problem

A(t)

A (t)

X (t)

if
.G:$ if
.G
if

~ ( t)

are

N(O)
R-

A(h(.C

R+

of

(N(O)

( wit h

J (t) =

at

is

continuous

:f

P ( s) h ( s ) d s )

is a neighbourhood of 0),

,...

- ah(t)}

the criterion

evaluated along the control

depends only

on the

parameter a

and we

write

v (a )

I:

For a path (cIt), M(t)} and a given function

value of

path (cIt)

exists

function h(.) such that A(h(.

J ( t)]

A(h(.C.

Necessity.

h(.), the

there

- un.! P ( t ), all t

any arbitrary

A(h(.

iff

all t

G = 1 i m [A (t) e - <> t
t-;>oo
exists and
.G

(0)

such that

P (t) = u c

(c) for

~.

:f

A(h(.

c(t) > 0. all t E [0,00], M(t) > 0, all t E [0,00[,

a solution
function

including
i.e. such

'"
'"M(t)}
(A.2), a path (c(t),

Theorem 1. Under assumptions (A.I) such that

equivalently,

schemes
solution,

u ( c( t) - ah (t) ,

(C'(t) , M(t)} solves Problem (0)

possible functions h(.).

PIt)

iff

E aArg Max

(V(a for all

EA (h ( . ) )

78
Under assumptions

(A.1) -

(A.2) we use in Appendix a

standard theorem to show that V'(O) exists and

I:

V' (0)

Integrating by

[-U c hIt)

parts the

P( t>

second part of the above equation we

obtain :

100
0
100
h(t)q(t)dt - lim J(t)e- 6t t _____ e- 6
t->oo
PIs)
u~

V' (0)

-u c e- 6t + PIt)

where q(t)

I:

and '"
M(t)

Since cIt)

u~

e-

(s- t1

ds,

6sds ,

PIs)
are strictly

positive for

all

finite values of t, there must exist non trivial functions h(.)

and A(h(.
function

I:

h(.)

such

O.

all t equal or larger than some T > 0

for

Now if {~( t>,

~N(O).

q(t)h(t)dt

q(t)

= o

J(t)

such that

that

To

11 (t)}

is to so I ve (0), every

Hence there

must

res~ect

(since k

constant), we obtain

is a

The necessity

of

the

time and

Euler

must

k E R

exist

kP(t> all t E [O,Tl. Setting )\It>

q(t)/P(t) with

to

P(t>h(t>dt

verify
such

that

uc/P(t>, derivating
result to 0

equating the

A(t) = 6A(t) - u~/P(t).

equation

(T.1

(a)

(b

is

established.
From the
V'(O) above
M(t)}

is

necessity of

we deduce

to solve

(a) and

that V'(O)

problem (0).

(b) and

must equal
The necessity

results

(-G)

if

on

{c(t),

of (c) follows

trivially.
Sufficiency. By concavity of u (A.1), u('C(t> , }l'(t)/P(t u(~(t)

u~

- ah(t), -- (~(t) + aJ(t) ~ U c ah(t) PIt)

PIt)
Integrating from 0 to +~ allows us to write

VIOl - VIa)

PIt)

aJ(t).

- aV' (0)

79
V is obviously a concave function and the sufficiency
of (al -

(bl -

(cl follows directly.

A serious drawback of theorem 1 above is that it does

not provide

us with

used directly.

a Transversality condition which could be

Theorem 2 below shows that T.l.

(cl

implies the

usual TVC (limA(tl e- 6 tM(tl = Ol

t-> ...

Theorem 2.

LimA,(tle- 6 t H(tl

t-> ...

Necessjty Let

fr..Q.Qi.

[O.T]' g(tlP(tl
J(tl

iff T.1.

= H{t)

(cl holds

"-

P(tlg(tl

M(tl +

all t ~ T.

M(t).

i:P{s)g(SldS

MolT. all

Obviously

for all t

larger than T. Now A(g(.

since A(g(.)l always includes the segment [-1.0]. Now

T.1.(cl implies that lim

A ( g ( .)

= [- 1 a 1.

T. 1.

A (tle- M(tl = a (note

I i mA( t) e - Ii( t) :!i a
6t

t->OI
(c) - >

that when

6 t

t->OI

which may only be satisfied with =).

Syfficiency.
From the
for all

non negativity constraints M(t) + aJ(t)

a E A(h(.. Multiplying both sides byA (t)e-

given that

limA (t)e- 6 t M(tl = 0, to conclude that a G ~

all a E A(hL. Now A(hL) N(O) -> G

:!i

and A('h(.C

R+ -> G

Let

simply

us

= 0,

leads.

6t

for

A(hL C R- -> G

O. Q.E.D.
note

to

conclude

here

that

transversalityconditions provide no additional requirements in

,."

the case where lim M(t)

t->OI

=a

4. Equi libria
x(t)

We shall

A (t)M(t).

now introduce a new variable x(t) such that

Equivalently. using

Differentiating with
equation (T.1

(al -

respect to

T.1 (a),

x(t)

uom.

time and using both the Euler

(b)l and the equilibrium conditions (4) we

obtain a law of motion for x(t) = (m(tll.

Definition 1.

A candidate

Equilibrium Path

(CEP) is

path

80
(x(t)} defined by

x (t )

(6 )

u,"

(9 + 6 -

(y,m

x(t),

which is Euler path equilibrium values (y,MS(t)} substituted in

it.
The

Lemma

below

correspondence between
It will

enable us

will

a price

to speak

establish

one-to-one

path (P(t)} and a path (x(t)}.

equivalently of equilibrium price

paths or of equilibrium x-paths .

.l...emm.a-1..

Under assumptions

(A.1)

(A.4), x(t)

decreasing function of P(t) for all t

~.

From T.1

(a)

(P(t) u cc )-l(u c

0,

it is easy to calculate that dc-(t)/dP(t) =

+ mu c ,")'

(u c + mu c ,") > O.
<

is a strictly

Now

From (A.1) and (A.4)

dx(t)/dP(t) = -

it follows that

(P(t-2M(t)

(u c + mu c ,")

O. QED.

Definitiofi-Z : A Full Equilibrium Path (FEP)

(a) a price path (P-(t)},

is equivalently:

with P-(t) > 0

all t

0,

such that (y,MS(t)} solves Problem (Q) where the consumer faces
(P-(t)} parametrically;
(b)

candidate

equilibrium

path

V-

satisfying both a feasibility condition,

~

0, and a transyersality condition,

{x-ttl}

(CEP)
(x-(t ~

all t

lim x-(t)e- 6 t = 0
t -> + ro
(M-(t)/P(t

= ~(P(t,

which by Lemma 1 is a strictly decreasing function.

It would be

Let us

write x(t)

easy to

= uc(y,M-(t)/P(t

show that

the feasibility

x-(t) =

~(P-(t for all t ~ O. Note that

V-

condition

(x(t ~

other feasibility condition P(t)

Let us
and (A.4)

imply that

The separable
Uy

It

write x

(Figure 1.8)

0 corresponds

the

= uc(y,m)m =V(m). Assumptions (A.1)

~'(m)

>

0 (this follows from Lemma 1).

below (Figure

cases where
and

1.A) together

\>(0) =

V(O) > 0 (Figure 1.C).

from the substitutes case (where

with the two

0 but V'(m)

distinguished the complements case (where

uy

to

O.

utility function is a special case where"'(m) =

is pictured

other possible

:f

u'c(y,m)

In Figure 1.8 we have

V' (m)

~'(m)

is larger than

is lower than u y

).

81

v(m)

FIGURE 1.A.

______________________________

~m

v(m)
u (y,O)m
y

v(O)

v(m)

FIGURE 1.B. (I: complements'case,

II: substitutes'case)

FIGURE 1.C.

82
SinceV : R+ -> R+ is a strictly increasing function,
there is a well-defined, strictly

V- 1

'\J

Let us define i\.(x)

=-

u'"

increasing

(y,"y- 1 (x

inverse function

and substitute it for

Uc

u",/u c

in (6). We obtain the following first-order differential

equation in

x,

the

solutions

Equilibrium Paths (CEP)

x(t)

(1)

Lemma 2.

(6

of

-X (x(t

which

are

the

Candidate

Under Assumptions

x(t)

(A.1),

(A.3)

and (A.4)

the

only

possible CEP are the following ones

(a) Steady-state paths such that x(t)

the solution ofi(x)

(b)

e ;

6 +

Divergent

(Hyperdeflationary paths)

X'

or dx(t)/dt

From (7) it follows

(x(t) )x(t).

that

<

dx(t)/dt

(A.2)

from

But

(Hyperinflationary

: we

dx(t)

it

that d(u",/uy)/dm < 0 and from (A.1) and (A.4)

increasing function

dx(t)/dt > 0

paths such that either

p.a.t.h.s) .

~.

x, where x is

it! t)

dx(t)

is

V-

easy
1

to

show

is a strictly

may conclude that)x) >

o.

Lemma 2

follows trivially.
The main
it allows
as have
key to

thing one has to learn from Lemma 2 is that

us to rule out such "convergent speculative bubbles"

been studied

this result.

multiplicity of

by Obstfeld [1984] : Assumption 4 is the

Not surprisingly

equilibrium paths

it turns

in Obstfeld

out that

the

[1984] was the

direct consequence of consumption being a Giffen good, at least

in a neighbourhood of the steady-state. Assumption 4 allows one
to rule

out also more paradoxical behaviours (such as cyclical

paths) which

would become

possible with

consumption being
Giffen good over some ranges of values of m.
We can

now proceed

to

study

the

set

of

Full

83

Equilibrium Paths in order to find out conditions under which

this set has at most a unique element : the steady-state path
when it does exist. This amounts to finding conditions for
ruling out hyperdeflationary and hyperinflationary paths.
Theorems 3
and 4 deal respectively with sufficient and
necessary conditions. Lemma 3 states a result which will be
useful in both theorems' proofs (note that it's simply Obstfeld
and Rogoff'
[1986] "Theorem I"
for a similar result see also
Tirole [1985] page 1098). It is only an equivalent way of
stating the transversality condition and as such gives no
additional information on how to rule out hyperdeflationary
paths.

L~mma

3. Lim

t->oo

x(t)e- 6 t

where f(x) =X(x)

~.

0 ->

i: f(x(sds

r-

f(x)

Xo x(cS-f(x

- .....

- e, x = lim x(t)
t->oo

Dividing both sides of equation (7) and integrating from

to t yields

f(x(sds.

log x(t) = (cS +

e)t -

i: x(x(s) )ds = cSt - i:

It follows that lim [x(t)e- 6 t ] = 0 -> lim(log(x(t

t->oo
t->oo
- cSt) = -00 -> ]0 f(x(sds = +00. Using (7), a simple change of
f(x)dx
variables leads to ]of(x(sds = ]xo
x(cS - f(x
Theorem 3. Under Assumptions (A.1) to (A.5), if

r"

r"

rK'

mUm

(a) lim ( - ) > 0

m -> 0 U c
(b) lim (f(x < 0, or
x -> 00

(c) lim (f(x

x -> 00

and]cx < 1, k > 0: E:(x) S -k[f(x)]"" all x ~ xo,

where t,(x)
dlog(f(x/dlog x.
the equilibrium set n has at most one element, the steady-state

84

path.
hyperinflationary
E.r..wU..
(a)
equivalently written as :

V ' (m) lit = (cS

(8 )

+ e)

V (m)

Assumption (A.4) implies that

sufficient condition for an
infeasible

Equation

CEP.

may

be

- mUm
~ '(m)

>

0,

all

hyperinflationary

(which is also a necessary one)

o.

CEP

to

be

is

lim(m) < O.
m->O
one concludes that

Dividing both sides of (8) by ~'(m),

m
lim (m) < 0 <-> 1 im (-u m) > 0 (lim [ ~ (m) I
m->O Uc
m->O
m->O
m
1 im
=0
m->O 1 + (mucm/u c )
since from (A.4) mucm/u c > -1, all m
shows that 1 im [mum
m->O

since

(7)

'iJ ' (m) ]

0 ; a similar argument

V' (m)] > 0 -> lim [mum/u c '] > 0).

m->O

(b) Hyperdeflationary CEP. Let us first note that,

f'(x)
is negative and x increases along an hyper

deflationary CEP, lim fIx)

then

0)0

X->O)

f(x(sds

<

<

0 ->

f(x(s

from

Lemma 3,

condition does not hold.

If lim fIx) = 0 and a < 1, k > 0

<

0, all s >

S, and

the transversality
.(x) :s -k[f(x)]"',

X->O)

then

fIx)
X

f'(x)

:s - - - -

kf(x)'"

From Lemma 3 it then follows that 1= i: f(x(sds :s

-f'(x)dx
kf(x)"'(cS-f(x

<

i:(Xo)

df(x)

kf(x)"'(cS-f(xo

f(xo)l-",
k(cS-f(xo(l-a)

<

+0)

T.3 (a)
is a sufficient (and necessary) condition for ruling
out hyperinflationary CEP on infeasibility grounds (obviously,
the transversality condition is always satisfied along an

85

hyperinflationary CEP).
One should note that the condition

lim mUm> 0 which

m->O
appears in the literature (Obstfeld and Rogoff [1983], Gray
[1984]) is no longer sufficient nor necessary in the general
non-separable case : the reason is that Uc depends now on m and
that its limit properties near zero do matter. For instance,
the condition lim mUm> 0 may be satisfied while hyperinm->O
flationary CEP remaining feasible if
Uc (y,O) = 0, a case
where real money and consumption are very strong substitutes.
It may be also that lim mUm = 0, while all hyperinflationary
m->O
CEP are infeasible if
ue(y,O) = +00, a case where real money
and
are very strong complements (when real balances
consumption
are zero, the consumer enjoys no additional amount
of
consumption good). Lemma 4 below provides us with a
set of conditions implying that lim [mum/u c ] > 0
m->O
Lemma

~.

if (a) lim mUm> 0 and Uc (y,O)

m->O
(b) lim uc/u m
m->O

< + 00,

o and lim [d(ue/um)/dm]

m->O

or
< + 00,

mUm
then lim --- > 0
m->O Ue
The
reader. If

proof of (a) is obvious and will be left to the

lim ue/um = 0, we may apply l'Hospital's rule to
m->O
lim [d(ue/um)/dm] - 1 . Note that (a)
conclude that lim mum/u e
m->O
m->O
is direct generalization of the well-known condition lim mUm
m->O
> 0 :
one has simply to assume. that the marginal utility of
consumption takes on only finite values.
T.3. (b) - (c) state conditions which are sufficient to rule
out hyperdeflationary CEP. Contrary to results which appeared
Gray
[1984],
the
in Obstfeld
and Rogoff
[1983] and
transversality condition itself is not sufficient to exclude
hyperdeflationary CEP from the equilibrium set: at the root
of this error is the wrong conjecture that lim f(x(t=O is
x->oo

~.

86

I:

sufficient that
f(x(tdt (to show that this is not true,
let f(x(t = lIt . . ). The economic interpretation of T.3. (c)
is the more straightforward in the zero monetary growthseparable utility function case
it then means that the
implicit (or subjective) rate of return on real money balances,
u~/uc, has
an elasticity with respect to real balances m which
is bounded above by increasing function of m
under this
assumption the rate of return on money holdings decreases
quickly enough while real balances increase so as to make its
integral converge along any hyperdeflationary CEP. Note that
T.3. (c) implies Brock's [1974] sufficient condition for ruling
out hyperdeflationary CEP: f(x) S x- k Let us take a = 0 in
T.3. (c), and, now
E. (x) S -k, all x <! xo, imply that log
(f(x S -klogx and. hence. f(x) S x- k
Theorem 4. Under Assumptions (A.1) to (A.5)
mu~

if lim
o
m->O U c
belongs to the equilibrium set 0
(a)

every hyperinflationary CEP

(b) if limf(x) > 0 or if limf(x) = 0 and k > O. a <! 1

x->oo
x->oo
E(x) <! -k[f(x)]a. all x <! Xo. every hyperdeflationary CEP
belongs to the equilibrium set 0
~L.
(a) ~erinflationary CEP. An obvious condition for an
hyperinflationary CEP to be feasible is lim(m)
O. From equam->O
tion (8) it is. as shown in the proof of Theorem 3(a). equivalent to the condition lim [mu~/uc] = O.
m->O

<

0 ->

I:

(b) merdeflationary CEP. if lim f(x(t

> 0, f' (x)

dt infinite along any hyperdeflationary CEP.

f(x)
f'(x)
If lim f(x(t = O. .(x) <! -kf(x) ->
kf(x)a
x->oo
x
f(x(t

It follows that I

I: f(x(sds

<!

I:(X

o )

df

kf(x)a(o - f(x

f(x) l - a
[---------]f(xo) = + 00.
(1-a)k6 o
Now the transversality condition is obviously satisfied.

>

87

Theorem 4.(a) has the immediate consequence that (see

Theorem 3. (a
condition

lim mu~/uc > 0 is a necessary and sufficient

m->O
for ruling out hyperinflationary CEP. Lemma 5 below

states some conditions,each of them implying that lim

m->O

L..~mlluL5..

If (a)

u~

lim
m -> 0

<

OD

Uc

mu~/uc=O

or

(b) lim mu~ = 0 and uc(y,O) > 0 or

m -> 0
(c) lim mu~
m -> 0
) lim
m -> 0

<

and

00

u~

+
Uc

00

Uc

(y,O) = +

00,

or

and lim d(uc/u~)/dm

m -> 0

00

then lim mu~/uc = 0

m -> 0
The p..r_Q.Q.i iss imp I e and we I eave it to the reader.
From T.4 (b) we learn that an hyperdeflationary CEP
is an equilibrium whenever the growth rate of the money supply
is negative, a result which can be traced back to Brock [19751
who was the first to put it forward. The same result obtains
when limX(x) > 9 ~ 0, an occurence which, in the non separable
x -> 00
case,
is not inconsistent with the usual assumption in the
literature that

lim u~ = 0, as illustrated by the example

m->oo

below :
u(c,m)

[cxc +

~m

+ log (W m + JJ

1,

0 <

<

1, cx,

~,W,

JJ > 0,

where
u~
Uc

~ W m + ~JJ +W

aW m + CXJJ

is a decreasing function of m, is bounded

below by (~/cx) and bounded above by (~JJ +

W )/cxJJ.

Note that lim

m->O
we can conclude that

= 0 and lim u~ = + 00. From Theorem 4

m->O
every hyperdeflationary or hyperinflationary CEP
u~

the utility function above is a full equilibrium.

generated by

88
The

second

part

of

T.4 (bl

states that an hyper-

deflationary CEP is an equilibrium, even when lim f(xl = 0, if

x->oo
the elasticity of f(xl with respect to x is bounded below by an
increasing

function of x. The following functions satisfy this

condition:
f(xl

A similar

case is

suggested to
Fernandez.

[ko: 10g<1 +
f(xl

function

previous statements
transversality

in the

condition

(k,o:l E R+x[I,+oo)

[logx)-l,

the

Rogoff (1986)

Obstfeld and
These

xl)-1/~,

are

so

example
by G.

many

enough

to

R.

counterexamples

to

rule

hyperdeflationary CEP.

APPENDIX

Let us write f(a,tl

follows that V(al

I:

f(a,tldt.

We have to complete the proof of T.l to prove that

V'

(a

df

(a,tldt.

We shall use the following standard theorem

was

Calvo and

literature according
is

which

to which the
out

all

89

THEOREM
Let yea}

I:f(a,t}dt, f

If:

(i) f

af

( ii )

~a

(iii)

[- , ]x

wo

-> R+

[- , ]x R+

is continuous,

-> R+ is continuous,

I:f(a,t}dt exists V a E [- ,

( i v) Sup
aE [-, ,, ]

af

T.

1.

---

oa

(a,t}dtl

-> 0 as T,T'-> +

then V' (a) exists V a E [-, ,, ] and V' (a)

Conditions (i),

00

(a,t)dt

(ii) and (iii) are obviously satisfied in our

model. Cond i t i on (i v) is less s imp Ie. l&.LJ.l.fL.d.e.lllQ-fis..t..r..a.t..e.._..th.at.

.t.his last condition is ~.I-ii...Le.d when A.l. (concavity of
holds and G (

u(.

exists.

1 im
t->oo

Proof
a) from concavity of the utility function u(c,m), a E [- , 1
->

f(a,t)

is concave t

since

Jt

+ (_)2 U ............ <

(b)

let 0 < T < T'

it follows from before that

90
is concave

aE[-C,l

c) from concavity of L(a) it follows that

dL(a)

1--1

a E [- /2. /2],

L(-) - L(-f,/2)

f:,/2

da

dL

Max

[I

L() - L( /2)

/2

I.

1]

We can now
under the sign

S kTT'

use

the standard

theorem of

derivation

when T and T' are finite to justify :

(a,t)dt

(a)

da
Since

dL(a)

1--1
da

follows that sup 1

a E

I:'

S kTT'

for all a E [- /2.

/2] it

df (a,t)dt

da
[-12. 1 2 ]

But kTT"
the existence of

tends to 0 as T. T' tend to infinity since

I:

f(a,t)dt implies that lim ILTT(a)

T,T'->oo

1=

0,

Q.E.D,

VaE[-,l

When a is constrained to be non negative the proof is

slightly different, We define
dL( 0)

k TT

= max [1--1,
da

L( ') - L(

/2

12)

11 with

concav i ty, ,
It is not difficult to show that

dL(a)

1--1
da

S k TT

, by

91

da

Now since lim (-AT e- 6t J T ) exists, lim ![-AtJte-6tjT!

T->Q)
T, T' _> Q)
T
and k TT tends to

zero

as

time

tends

to

infinity.

Q.E.D.

REFERENCES
D'Autume, A. and P. Michel: "Transversality Conditions. Budget
Constraints and
Equilibrium

in

of a Perfect-Foresight
Model".
European

the Determinacy
a

Monetary

Growth

Economic Review. 31 (1987).

Benveniste, L.M.
Dynamic

and J.A.

Sheinkman

Continuous Time

"Duality

Models

Optimization

Case", Journal

Theory

Economics

of

of Economic

for
the

Theory. 27

( 1982). 1 - 19 .
Brock, V.A.

: "Money and Growth : the Case of Long-Run Perfect-

Foresight" International

Economic

Review.

15

(1974),

Monetary

Model".

750-77.
Brock, V.A.

"A

simple

Perfect-Foresight

Journal of Monetary Economics. 1 (1975). 133-50.

Calvo, G.A.

: "On

Models

of

Money

and

Perfect-Foresight".

International Economic Review. 20 (1979). 83-103.

Gray, J.A.

: "Dynamic

Models:

Instability

in

Rational

Expectations

An Attempt to Clarify". International Economic

Review. 25 (1984). 93-122.

Laussel, D. and Soubeyran. A.: "Speculative Bubbles in Monetary
Optimizing Models: Towards a Synthesis". mimeo (1987).

92
Obstfeld, M.

"Multiple

Stable

Equilibria

in

Perfect-

Foresight Model", Econometrica, 52 (1984). 223-28.

Obstefeld M.

and Rogoff,

Maximizing Models

K.

: "Speculative Hyperinflations in

: Can We Rule Them Out ?", Journal of

Political Economy, 91,(1983), 675-705.

Obstfeld, M. and Rogoff, K.

: "Ruling Out Divergent Speculative

Bubbles", Journal of Monetary Economics,

17 (1986). 349-

62.
Tirole, J.

"Asset

Bubbles

and

Overlapping

Econometrica. 53 (1985), 1071-1100.

Generations",

THE RANDOH

THE OPTIHAL SHARING HONEY-BOND IN THE PORTFOLIO

CHARACTERISTICS APPROACH
Jean-Harie ROUSSEAU

Universite de Bretagne Occidentale

12, rue de Kergoat

Yhen CLOVER
double constraints
double function

B.P 331 - 29273 BREST CEDEX

(1967) establishes
in the

which is

the existence of the

consumer objective
fulfilled

by

he recalled the

money

as

medium

of

exchange (income constraint) and as store of value (expenditure

constraint).
The

selling

monetary balance,

of

the

in contrast

products
it is

gives

the

agent

necesserary for buying a

product to hold of an equivalent purchasing power.

The relative
of course,

equal

marginal rate

price of the two functions of money is,

to

one.

It

of substitution

follows

that

between the

optimally,

two

the

corresponding

demands is equal to one (1).

It is
money into
following

hence possible

account, by
P.

to take

associating to

DAVIDSON

(1973),

D.

the two
them a

FISHER

and

functions of
characteristic
more

lately

H. BOURGUINAT (1987) regarding the international aspects.

The analytical setting of such an approach as already
been defined

by K.J.

consumer choices

not

LANCASTER who
to

the

goods

has proposed to apply the

themselves

but

to

the

characteristics which are associated.

The extension
monetary and

of the

financial assets

LANCASTER's

approach

to

the

raises two questions: first to

define a measure of the characteristic of these assets (see for

example P.

ARTUS et

J.M. ROUSSEAU

(1987

and

secondly

to

elaborate an accurate theoretical frame-work.

* I am very indebted to the Professor DE BOISSIEU and P. ARTUS

for their encouragements to carryon my investigations in this
subject. Many thanks also to Professor C.
CRAMPES for his
helpful comments.
(1) on this point cf D. FISHER (1978), annex B, pp. 239-245

94
As regards this last point we have already formulated
the main

results in

the case

of certainty

(J.M. ROUSSEAU

1984) .

We deal

now with case of uncertainty. There, the two

characteristics of

the assets

O. and the store of value O2

The aim

of this

are random variables (liquidity

),

paper is to analyse how the sharing

money/bond is done.
In order
of certainty

to allow a comparison between the two cases

and uncertainty,

we will

determine very rapidly

what is the optimal sharing of bond under certain future.

random future,

the agent

expectation drawn

is supposed

from this

Under

to maximize the utility

characteristics

(J.

J.

LAFFONT

1985) .

The sharing
deduce money
other side.
two sorts

money/bond in

demands on

allows

to

the one side and bonds demands on the

The obtention
of assets

the portfolio

of the

can be

optimal sharing between the

done by

applying the LANCASTER's

characteristic approach.
The

holding

of

an

asset

allows

to

obtain

simultaneously two services.

1. A store of value service.

The asset
the wealth

A(l +

which yields

the rate

r allows to obtain

r) which is equivalent to a purchasing power

of A(l + r)/p, where p is the general price level.

2. A liquidity service
As the

agent decides

to convert the wealth A(l + r)

into money he obtains :

A(l + r)(l - c)

where c is the liquidation cost

or
AU +

where I

1)

=r

- c(l + r) or

approximately I
Let

r - c

O. and O2 be the characteristics of liquidity and

95
store

value

of

composition of

the

agent

is' reaching

his portfolio

so as

to

determine

to maximize

his

the

utility

function under the budgetary constraint.

First introduce the result in the certain case.
The wealth, w, of the agent is divided in two parts T
(bonds) and M (money) which can be put as
M

(1 - 9}w

and

The problem

9w where 9 E [0,1]

consists then

to determine

the optimal

value of

9. Each of the assets yields both a liquidity service

(01) and

a store

of

value

service

(02)

according

to

the

following array.
+ r)

(1

TO + 1)

p
M
M

p
Portfolio

02

01

where 0 1 = w(l - 91}

p02

w(l + 9r}

The agent objective is then determined by

Under the

constraint rOl

l.p.02

(r -

l}w which

represents the budget constraint in the characteristic's space.

In the special case where Y
(r

0 1 .02 , it follows that:

(r

1)

Yo

1)

.w

and
-1

96

e-.

that corresponds to the optimal sharing

1 + r

e-

--r.l

for

e-

E (0,1) we have 1 E (-r, 1 + 2.r

which requires 1 to be negative or that c

>

1 + 2r

If c

<

r +

the agent will only holds bonds in Portfolio.

1 + 2r

Money is

dominated by

bonds. Bonds give simultaneously levels

of liquidity and store of value services which are greater than

the ones given by
Portfolio.

money holding. Money is eliminated from the

We will now deal with the uncertain case.

Assume that

yield rate,

r, liquidation

cost c, and

general level of price p are random variables. It follows that

1 is also a random variable.
The target

of the

agent is

to maximize the utility

function expectation.
The agent objective is then

under the constraint r01 - 1.p02

(r

l)w

We shall still assume that

Reformulating the first order condition in this case:

97

Ee:-)
This value
interval (0,1).

can be

As usualy

accepted only if S belongs to the

we

shall

study

the

first

order

condition
2{r.T.w + (a 2

cov(r,c)w + cov(r 2 ,l/p) - cov(rc,l/p)}S +

r.w + T.w + 2cov(r,l/p) - cov(c,l/p)

where r, T, w denote the expectation of r, 1,

We can

now study

the

influence

of

lip.
the

and

expectations and standard deviation on S.

1st. case

cov(X,Y)

=0

v X,Y

All covariances will be assumed 0 or neglected.

In this

case we will find again the quoted condition

since the first order condition is

2(r.I.w)S + r.w + T.w
hence

r.T
The examination
the following
portfolio is

results

first order

of the
if

uniquely composed

>

0 (i.e Sof bonds.

condition gives
1) then the whole

Money is completely

dominated by bonds and hence is taken away.

There can be money in the portfolio only if

T < o.

98
The

condition

expectation of
expectation)

<

the

(i.e

nominal

yield

the bond is weaker than the cost of liquidation

is

a necessary

condition of having some money in

optimal portfolio.

~he

For insuring
come again

to conclusion

case.

If
when

the balance

then 8

< 0,

demand not

identical to
M

to be zero, we

those of

the

certain

will belongs to the interval (0,1)

will take values belonging to (r.2F).

More precisely we have

(there will

<

be some

money in the portfolio)

if c

r + -------.

>

If c

2r 8

will

1 + 2r

always be zero
In a more general way we can then state the following
conclusion :
In the case where the dispersion of the values of the
random variables
be zero

around the mean is weak (the covariances will

or neglected)

the optimal

sharing of

the

portfolio

depends only on the expectations rand c.

The portfolio
small values
portfolio

of

will be

c (c

will

be

intermediate values

F). For high values of

<

composed
of

wholly composed of bonds

exclusively

c (c

of

conclusions

2F) the

money.

For

the portfolio will contain both money

and bonds and the share of bonds will decrease with

The

>

for

are

strictly

similar

c.
to

those

obtained in the certain case.

2nd case.
There,02 r

cov(X,Y)

=0

can not be neglected.

The 8 formula becomes then :

I

+ r

= 1..----------

We have

X,Y

except for X

Y=r

99

69
sign

6r
The share

of bonds

will increase

with the

average

nominal yielding if 12 > 02 r .

69
sign

61
The share

of bond

demands decreases)
liquidation cost)
The two
with those
use of

increases (and

when l{average

hence the balance

nominal yielding

- average

increases if r2 > 02 r .
results should

be compared

and

contrasted

obtained by conventional portfolio theory where the

a quadratic

liquidity in

utility

the optimal

function

portfolio

leads
which

increases provided r2 < o2r (P. AFTALION

to

decreases

share

of

when

& C. VIALLET (1977.

Furthermore

69
sign
6c

The

share

liquidation cost if r2

of
>

bonds

order moments

first case

by

with

the

average

02r (2)

Integrating uncertainty
the first

decreases

through other

modifies the

elements than

results obtained in the

pointing out more precisely the role played by

the risk of rate (mesured by or).

To the

question:

"why does the investor always hold

money in his portfolio 7" the answer is always the same because
money dominates

bonds as

regards liquidity (wich is expressed

(2) In the converse case where r2 < 02 r . the second order

condition for reaching a maximum is no more satisfied and we
have then a corner solution 9* = 1. The whole portfolio is composed of bonds.

100

by

1 =r - c

<

0).

Taking

elements

other

account

into

for

facing

uncertainty will change the answer.

3rd case
In this case, the formula of
r

and coy (r,c)

becomes

The attractivenes of this case lays in that it allows

to read the following conclusion :
It is

possible to

prove that

contains money

despite the

positivenes of

(1

yielding of bond
In other
by the

the optimal portfolio

the

net

expected

0).

>

words, although

money in totally dominated

bond as regards characteristics it is still held in the

portfolio.
As an example let us take the case when c
It is

that

r>

then possible

0 and

r.T

necessary

02r

to find

ar + b

values of a and b such

cov(r,c) < O.

condition

for

having

there

two

inequalities simultaneously verified is that a > 1.

So if the liquidation cost is supposed to be a linear
function of the nominal interest rate it will be necessary that
Oc

> Or since a

It is
when the

is precisely the proportionality coefficient:

hence possible

to obtain the announced result

risk on the liquidation cost is greater than the risk

on the nominal yield.

The meaning
if we

suppose

the bond

T =0

provides the

lays then in the risk.

of this
(i.e

conclusion appears more clearly

r = c).

same service

In terms of expected values

as money.

The difference

If the
risk of

interest rate

risk of

liquidation

cost

composed of the bond C9-

the

Or

whole

is greater than the

portfolio

will

be

1)) but conversely the portfolio may

also include some money.

This conclusion
that is

can be

extended to

the case I > 0,

the case where the expected value of

the one of

c.

Taking
certainty

standard

allows

to

deviation

predict

into

some

is greater than

account

portfolio

under

an

sharing

not

reducible to the certainty case.

As

conclusion

characteristics allows
a portfolio.

It is

and a

demand

bonds

the

approach

by

the

random

to determine the optimal composition of

then possible
apart

from

to determine a money demand

the

conventional

portfolio

theory.
But, furthermore

it is

to' obtain the

also possible

random demand of liquidity service O. and of store of value O2

It

is

hypothesis we

nevertheless
have formely

necessary

assumed

to

about

release

the

some

magnitude

of

covariances.
More
influence of

explicitly,

it

a price variance.

relatively to

is

necessary

to

assess

the

If 9 is homogeneous of degree

the price p and to the initial wealth w, a shift

in distribution of the future price p will modify the numerator

and determinator of the ratio which defines 9.
That will
other elements
process of

take place

I and

through covariances

c involved

the investor

in

the

utility

with

the

maximising

even if the expected purchasing power

is assumed to be constant.
Taking into

consideration the

random

character

of

service yielded

by the financial and monetary assets imply the

optimal sharing

to depend

on other

elements than the central

values of the random variables used in the economic calculus of

the investor.

102

REFERENCES

& Viallet,

Aftalon, F.

C.

(1977)

- Theorie

du portefeuille.

Analyse du risque et de la rentabilite. Paris, P.U.F.

Artus,

& Rousseau,

P.

caracteristiques

(1987)

J.H.

"Valorisation

qualitatives

Communication au

Colloque de

des

des

obligations"

l'A.E.A. sur

les modeles

monetaires et financiers. Geneve. 23 et 24 janvier.

Bourguinat,

(1987)

H.

Les

vertiges

de

la

finance

internationale. Paris, Economica.

Clower, R.

(1967) - "A Reconsideration of the Microfoundations

of Monetary

Theory", Western Economic Journal, December

; reimprime

in CLOWER

Ed.

- Monetary Theory, Hardmonds

worth, Penguin (1969).

Davidson,

P.

Money

(1973)

and

the

Real

World,

London

MacMillan.
Fisher, D.

(1978) -

Monetary Theory and the Demand for Money.

London, Martin Robertson

Laffont, I.J.

(1985) -

Economie de

& Co. Ltd.

Cours de theorie microeconomique : 2 -

l'incertitude et

de l'information.

Paris,

Economica.
Lancaster, K.

(1966) -

"A New

Approach to

Consumer Theory",

Journal of Political Economy. Avril.

Rousseau,

J.H.

(1984)

portefeuille:

-"Services

le cas

de

monetaires

deux

actifs."

et

choix

de

Document

de

travail, Universite de Bordeaux I, multigraphie, 40 p.

A SURVEY

DEBT CONTRACT UNDER IMPERFECT INFORMATION

Xavier FREIXAS
Universite de Toulouse I et GREMAQ

Place Anatole France 31042 TOULOUSE Cedex

The classical
appear irrelevant

analysis in terms of supply and demand

when the

problem

at

hand

is

related

to

lending, and to the credit market. The limits that are imposed
to the level of a firm's debt by its creditors, or the effect
of a

firm's capital

would still

structure on its value are phenomena that

remain unexplained

had the supply-demand analysis

been uniquely employed.

The

credit

characterized by

market

the

( 1 i ke

labour

the

heterogeneity

of

the

market)

agents.

If

is
the

investor could have a perfect information an equilibrium with a

different amount
agent would

lent and

be obtained.

a different interest rate for every

But, in

general, the information is

imperfect, and the borrower quality is unknown to investors

(except, may be, the quality of the largest corporations that
are evaluated
by
rating
agencies).
Consequently.
the
transactions that ideally should be realized in different
markets end

up taking

place in

unique

market

where

the

investors face potential borrowers of different characteristics

that they cannot recognize. The credit market results from all
the contractual

relationships between

potential investors and

potential borrowers that can be established with the investors'

imperfect information.
In this contractual relationship, the borrower has an
advantage for

two reasons

on the one hand. he will have,

in

general, a better information concerning the characteristics of

the project
may be

he is willing to undertake; on the other hand. he

able to

modify the

characteristics of this project in

response to his own objectives, as they jointly result from its

own

charac~eristics

(risk aversion, ... )

and from

the

credit

contract. As a consequence, the framework that allow us to

model the
credit relationship is the one of asymmetric

104

information, with both adverse selection and moral hazard, and

this explains why, to some extent, the analysis may be complex.
The analysis of the credit contract is the starting
point for an understanding of the credit market. The theory of
contracts, as a relationship between a principal and an agent
allows us to provide a justification for the use of the debt
contracts that are commonly used and that we will call
"standard debt contracts". Indeed.
it is interesting to state
out clearly under what conditions an optimal contract is
characterized by 1) a fixed repayment that is independent of
the firm's cash flows and 2) a bankruptcy mechanism with a
switch of control to the defaulting firm's creditors that may
obtain then a maximal repayment. This is done by use of
incentive compatible
contracts under
the
assuption
of
observation or bankruptcy costs.
The use
of standard
debt contracts has as a
consequence the fact that it endows the borrower with a convex
profit function,
so that a risk averse agent may become. in
fact. a risk lover. With the limited liability clause, this
implies that the agents will often prefer to choose more risky
projects. This
effect may
be mitigated in a long run
perspective. since in that case. a firm that defaults bears the
opportunity cost of not having access to the credit market to
finance new projects.
In order to survey this area. we will first start by
describing an
ideal contingent
contract. Then. we will
introduce the different asymmetric information restrictions and
the consequences on the contracts. Namely. after describing
contingent contracts (section 1) we will assume first (section
2) that the asymmetry of information relies on the value of the
realized cash flow, then we will assume that it concerns the
level of risk that the firm chooses (section 3). The control
issue in the bankruptcy will be discussed in section 4. and
finally we will examine the dynamics from a long run point of
view, that softens the incentive problem (section 5).

105

1. Optimal incentive contracts

In order to have a better definition of the borrower
lender relationship, which we take here to be a firm and a
bank, it is interesting to examine beforehand the extreme case
in which the markets are complete and the contracts can be made
contingent. In such a framework, the contracts would take into
account, at each period, the following element~ that depend on
the state of nature that occurs
l(st) the amount of the loan
R(st) the amount to be repaid by the firm
S(St) the revenue of the firm
G(st.) the collateral
I ( St ) the investment choosen by the firm
X(st.) the cash flow obtained by the firm
I n a com pie t e con tin g e n t ma r k e t, .:t..h.e..T.:_e_. __.l.B-..n.Q...--H.!:l..l.l
.d..e..Lin.e..d.......m.!:l..an.i~__....f..Q.r_Hh.a.t__..b..a.n.k.r.. u..p..t..c.,y_.u , sin c e the a b sen ceo f
any repayment in some states of the nature is taken account ex
ante, and there is no need to turn the firm over to its
creditors, since the actions that they would have taken in that
state of the nature are also specified in this contract. It is
precisely because it is impossible, or too costly. to write
contingent contracts that bankruptcy will play an important
part in the credit contract.
It is costumary to impose a limited liability clause
into the credit contract, and we will take this constraint as
given, although
we think that there may be interesting
clause
(an
the justification
of this
developments in
interesting contribution is the one of Sappington (1983)). The
firm's bankruptcy can then be defined as the fact that the
limited liability clause is invoked.
If we disregard contingent contracts as too costly,
the richer class of contracts that may be considered is
certainly the class of contracts that are contingent with
respect to the cash flow. Among this type of contracts, we will
define precisely the standard debt contract that is inspired by
Gale and Hellwig (1985) by the two following properties:

106

if the firm does not go bankrupt, the repayment to the lender

is fixed and independent of the firm's cash flows.

-

if

the firm goes bankrupt, then 1) the firms'

control of

the firm

and/or 2),

the firms'

creditors take

repayment to

its

creditors is the largest amount possible.

Several

types

of

models

have

established

the

optimality of this type of contract thus giving a justification

for its

use.

It

depending on
viewed as

is possible

to distinguish

the characteristic

primordial

the

of

the

revenue

two

approaches,

bankruptcy

stream

or

that

the

is

firm's

control.
2. Optimality of the standard debt contract
Two models,
have established
contract to

Still, they

by Townsend

(1979).

information concerning

will examine

auditing

for the

standard ctebt

These are Diamond (1984) and Gale and

differ on one point:

firm's cash

quite similar,

and they are related to a more general approach

first developed
asymmetry of

features are

sufficient conditions

be optimal.

Hellwig (1985)

that we

.whose main

first,

flows provided

cost),

while

the

the

models
firm's

assume
cash

an

flows.

in Gale and Hellwig's model,

borrower is able to know the

he pays

this

Both

a cost

is

(for instance

impossible

in

an

Diamond's

framework.
Since

this

model

revelation principle
use of

uses

dominant

strategies,

the

allows us to describe these strategies by

direct mechanisms.

The firm

announces a value for its

cash flow, and in response to that value, the borrower (a bank)

will decide
cost, or

whether to

not.

would imply

(Mixed

a very

observe the cash flow, thus paying the

strategies are

here excluded since they

low probability of observation with a very

high penalty).
A debt
of the

contract will

cash flows

cash flow

in Y

Then, for

the cash

incentive for

that are

it is

specify, first of all, a set Y

such that if the firm announces a

not optimal for the bank to observe it.

to be

in the set y, the

announce the

cash flow that is

flows restricted

the firm

is to

107

associated with

the minimal

repayment, whatever the true cash

flow y.
Min R(yl

YE Y
For such

a mechanism

to be incentive compatible, we

need
R(yl = R
in other words, the repayment function has to be constant.
The implication
firm does

not repay

observation cost
Therefore,

it

announce a
R. On

so

will

given the

is the

following:

R, the

creditor will

as

check

to

never

cash flow

the other

of this

be

the

optimal

choose to

realized
for

if the

the

pay the

cash

flow.

borrower

to

corresponding to a repayment superior to

hand, when

existence of

a low level of cash flow obtains,

a limited

liability constraint,

it is

interesting for the borrower to declare that a low value of the

cash flow

has occured, and the creditor will choose to observe

the firm's

cash flow.

the incentive
shown that

We therefore obtain the general form of

compatible credit

contracts.

It

remains to

be

the optimal ones will entail a maximum repayment in

the low cash flows case.

To do
its creditor

so, Gale
are risk

contracts are

and Hellwig assume that the firm and

neutral, so

those that

minimize

that the
the

Pareto efficient

expected

observation

costs. These contracts are then characterized by the fact that,

for a

given expected

will not be observed,

il

if

the cash

repayment, the

set y,

whose cash flows

is maximal. But this implies:

flow is

superior or

equal

to

R,

then

the

repayment is R
iiI

if

the cash

maximum

that

clause. This

flow is
is

inferior to

permitted

defines the

given

R the
the

repayment is the
limited

firm's bankruptcy

liability

as the case in

which the cash flows are observed and become the property of
the creditors.

108

The risk
role, since

plays

neutrality assumption

otherwise

there

could

be

an

important

trade-off

between

observation costs and risk sharing. But clearly if there is not

"too

much"

risk

consequence of
the firm
rate of

aversion

the

result

will

still

hold.

the risk neutrality assumption is the fact that

will invest

all its assets in the project, since the

return on its assets is equal to the rate of return of

the lender plus the average observation cost.

Diamond's (1984)

model considers

a bankruptcy

cost

which is

endogenous and non pecuniary. For instance, that cost

may

the

be

opportunity

cost

of

entrepreneur to

deal with

determines then

simultaneously the

the non

pecuniary cost.

pecuniary,

Since

spent

optimal debt

the

by

procedure.

bankruptcy

the

Diamond

contract and
cost

is

non

efficient contract
with the

will minimize

the bankruptcy

same incentive constraints as before, the

have a constant repayment cost. The optimal contract

firm must
will be

time

it is not an income for another agent, and therefore

every Pareto
cost. Yet,

the

the bankruptcy

here characterized
flow is

the cash

bankruptcy that
cash flow.

by a constant repayment R whenever

superior to

R and

to the

is equal

pecuniary cost of

difference between R and the

firm's point

From the

a non

of view

the

cost

of

the

repayment is always constant, so that the contract is incentive

compatible. The
the fact

difference with respect to Gale and Hellwig is

that R(y)

and

are determined simultaneously in

~(y)

Diamond's model.
One of
that is

the interesting

developed by

financial
a contract

between a

firm and

In both

information related
Financial

that it

view,

justifies

its bank or between a bank and

cases there

to the

a financial

number of

firms and

asymmetry

of

therefore justified by the

intermediary will lend to a much larger

the non

diminish with

the number

To understand

why there

larger number

exists an

observation of the borrower's cash

intermediation is

fact that

to a

the fact

Diamond is

this

intermediation. Clearly, this model describes as well

its depositors.
flows.

implications of

pecuniary

costs

by

firm

will

of firms to which a loan is granted.

are economies to scale in the lending

of firms

we have

to examine the expected

109

profits of the financial intermediary, that

risk neutral.
The expected profits are

E(w>

=E

[i!l gi ] - HN

[P(~gi ~

Hn > +

P(~gi

is assumed to be

<

Hn>]

where N is the number of firms. gi is the partial repayment by

firm i when it goes bankrupt and HN is the amount that the
financial intermediary has to repay to its depositors. which
because of the non pecuniary cost is seen as a constant cost
from the financial intermediary point of view.
When the number of loans increases, the probability
for the financial intermediary to go bankrupt,

[~

gi

<

HN ]

diminishes. and this, in a free entry financial market. implies

a decrease of the repayment H and better loan conditions for
the firms.
Consequently, given these economies of scale, the
model gives a justification for a limited number of financial
intermediaries with a low probability of bankruptcy.
It is worth noticing that the Gale and Hellwig's
model would lead to the same conclusion although the result
would be, then, trivially obtained. Indeed, since there is a
fixed cost of observation the economies of scale of granting
more loans are obvious.
3. Optimal capital structure
The conclusions of the preceeding models bring in
another question
by establishing the optimality of the
standard debt model. (which implies that the Modigliani-Miller
theorem is invalidated in an imperfect information setting>
these models show that it is never optimal for a firm to
increase its capital by an increase in its stock. Besides, in
the framework these models assume, the stockholders would never
be able to check that the cash flow that the firm has announced
is the one that has been realized.
Intuitively, it seems clear that a firm which is

110

almost completely
risky projects,

financed by
since if

limited liability

the firm

view, the

choice of

the project

tend to

take

will retain

moral hazard

investment projects,

very

is not successfull, the

clause will still protect him, while,

project succeeds
point of

debt would

if the

the profits. From this

issue is

that of the firm's

and this may appear as a limit

to the firm's possibility to borrow from the financial market.

To illustrate the firm's behaviour, we have developed
in Freixas

(1987) a

projects are

model in

available to

which two

every firm,

types of

investment

one safe and the other

one risky,

both having the same expected value that depends on

the firm's

characteristics. We have assumed,

in addition, that

the lender

which is

to

quality of

the borrower, so that both moral hazard and adverse

risk averse

is unable

identify

the

selection are present in this model.

Since the

two investment

projects

differ

only

by

their risk, the optimal contract should induce the firm to take
the less

risky decision,

the less

risky returns

while giving
stream. This

the risk averse lender

is possible

only with a

returns stream that is a linear function of the cash flows, and

this is interpreted as a combination of debt and stock.
4. Optimality of the bankrutpcy control switch
The above

models focus

on

the

incentives

in

the

credit contracts in a one period contract between borrowers and

lenders. Still,

by taking

only one period it is impossible to

analyze the change in the firm's property rights that will take
place in the event of bankruptcy. We know that in this case the
creditors will
making and

replace the stockholders in the firms decision-

choose either

to liquidate

or to

reorganize

the

firm.
Aghion and
of the

Bolton (1987) explain this characteristic

credit contract.

contracts are

too costly

They assume
to be

that complete contingent

used, thus

restricting their

analysis to incomplete contracts, where the actions to be taken

by the managers in each state of the nature are not specified.
The model
three possible

considers only

strategies:

two states

continuation of

of nature, and
the same type of

111

management,

innovation

flow that

obtains is

depend on

the state

action that
state of

or closing
a

that is

so that

nature that

it is

the firm. The cash

whose

distribution

realized and

on

the

impossible to infer the

has occurred from the observation of the

flow. Still,

that changes

variable

of nature

is taken,

realized cash

random

down of

the observed cash flow is a signal

the probabilities

of the

two states

of

nature

according to the Bayesian updating rules.

Assume that
the firm's
for the

investors differ

control the

is 9 1

parties cannot

approach the

issue contingent
it

while it is optimal for the investor to

abandon the

contingent

on

~he

If the two
states

of

firm's control depend on 9, then the best

optimal decision

on the

value of

correlated with
is more

make the control

the cash flow y, since this

the state

interesting for

is to
of

nature.

In

other

the firm's stockholders to

firm's control to the firm's creditors rather than

to compensate

them. The

easy to

debt holders
inferior to

make contracts

that the

variable is

it is

a way that it is optimal

firm whenever the realized state is 9 2

nature, so

words,

in such

stockholders to control the firm whenever the state of

nature realized

way to

the expectations of the stockholders and

role of bankruptcy appears then since

establish that
will only

be of

some critical

the passing

over of control to

interest if

the cash flows are

level corresponding

to the nominal

level of debt.
5. Dynamic contracts and reputation
It has
is less

been often argued that the incentives problem

important in the long run since the fact that the game

is repeated

may induce

may clearly

be the

firm's bankruptcy
to the

absence of

the borrower to be more cautious. This

case here,
has an

since in a dynamic context the

opportunity cost that may correspond

future lending.

Consequently, the firm has

some incentive not to default.

This intuition

has been

formalized in

the repeated

games context with imperfect information by the introduction of

the concept

of reputation

: the

reputation is

the effect on

player A's strategy of his taking into account the consequences

112

of his

actions into

player B's beliefs.

If player B's beliefs

are not modified after the observation of A's strategy there is

no reputation
time, and

effect. Therefore, reputation is builded through

like an

periods to

investment, entails

a loss

in

the

first

obtain a gain later on in the game. This gain comes

from the effect on the other player beliefs.

This idea developed by Kreps and Wilson (1982) on the
one hand,

and Milgrom and Roberts (1982) on the other hand has

been applied

by Diamond

(1986) to

study the relevance of the

reputation effect on financial markets.

His model is based on the following structure
- there

exists a

represents the
that the

riskless

asset,

and

the

collective financial market,

investor,

which

is risk neutral so

expected return on the loans that are proposed to the

firms are always equal to the riskless rate. Consequently, when

the probability

of default of a borrower decreases, the facial

interest rate on his loans will also decrease.

- the

investor cannot

observe the

cash flows nor the quality

(or the probability of default) of the different agents.

- there are three types of firms, all of them risk neutral, the
ones that

have a

zero probability of default (non risky), the

ones that

have a

large probability

larger profit

of default

(risky) and a

in case of success, and the ones that may choose

their investment

project, that

is they can choose to be risky

or non risky. This third type of firms we will call strategic.

- the

cash flows

for a

successful 1 firm

is such that if the

horizon of the firms was limited to one period, everyone would

choose the
firm are

risky project.
such that

Also, the

cash flows

of the risky

the lender will make losses on those loans

that go

to risky firms. As a consequence, the lender will only

lend to

the good borrowers, and a firm that has once defaulted

will never obtain credit again.

A

sequential

characterized by
its cash

flow is

repay their

Bayesian

is

then

shown

to

exist,

the following behavioun of the firms:

superior to

debt (ii)

the debt,

if it is inferior

the firms

(i)

if

prefer to

the lender obtains a

zero return to its loan. Simultaneously, the lender updates its

beliefs and

limits its

loans to

the firms that have not gone

113

bankrupt, and figure out the probabilities for a given firm to

be of one of the different types, given that it has repaid
during the first T periods. Clearly, the longer the history of
a solvent firm, and the lower the facial interest rate that it
will have to pay on its loans. This modifies the strategic
agents profit function. Indeed,
in a model with only one
perLod, they will always prefer a risky investment project,
while for a longer horizon they will take into account the
opportunity cost of being deprived of credit in the future.
Thus, for a sufficiently long horizon,
there will be a
reputation

effect,

with

the

strategic

firms

building

reputation of non risky firms by choosing the non risky

investment project. (Notice, though, that for a finite horizon,
the strategic firms will always choose a risky investment in
the last period, since the problem is then the same that in a
one period framework). More precisely, Diamond shows that there
may be two types of behaviour for the strategic firms : ei~her
they choose the non risky project from the first period,
(immediate reputation) or else they gamble by undertaking the
risky project at the beginning and,
if they are successfull,
they switch to the non risky one to benefit from the reputation
the~ have involuntarily built (acquised reputation).
Thus, as it is clearly from Diamond's paper, taking
into account the effect of a long run relationship make two
phenomena appear : on the one hand, the interest rates that a
firm will

obtain on its loans will depend on its history of

payments to the lender ; on the other hand the incentive
problems, which are a consequence of the use of the standard
debt contract, soften when the long run perspective is adopted.

To conclude,
it is important to understand that all
this litterature is at its start. The different characteristics
of the debt contract are now better explained. On the one hand,
auditing or observation costs for the cash flows will make
optimal for the firm to issue standard debt. Yet, if this is
the case, the moral hazard issue related to the firm's choice
of rnvestment projects is worsened. Taking into account this
problem leads to an endogeneous characterization of the capital

114

structure as

being composed of debt and stock. and to a better

understanding of

why the

control of

the

firm

goes

to

its

creditors when the firm goes bankrupt. All these problems still
arise in

a long run perspective. The difference comes from the

fact that

the firm

has an additional bankruptcy cost that is

given by the cost of its beeing denied credit in the future.

It

is clear

will

be

and

in

that in

explored again.

the years

to

come

possibly within

this

dynamic

problem
setting.

connexion with a justification of the limited liability clause.

That will give a much more clearer understanding of the working
of the credit market.

REFERENCES

Ashion, P.

et P.

approach

Bolton (1986).
to

"An

bankruptcy

"Incomplete

and

the

optimal

contracts"
financial

structure of the firm" D.P .. MIT.

Diamond, D.

(1984). "Financial

Monitoring" Review

Intermediation

of Economic

and

Delegated

Studies. July. pp. 393-

414
Diamond, D.

(1986). "Reputation

WP 134.

Graduate

School

Acquisition in Debt Markets".

of

Business.

University

of

Chicago.
Freixas, X.

(1987). "On

Debt and

Stock as

Optimal financial

instruments". Cahier GREMAQ. Universite de Toulouse.

Gale, D.

et H.

Hellwis

Contracts

(1985).

"Incentive

Compatible

Debt

The One Period Problem". Review of Economic

Studies. October. pp. 627-646.

Kreps, D.

and

R.

Wilson

(1982)

"Reputation

and

Imperfect

Information" Journal of Economic Theory. August.

Hilsrom, P.

and J.

Roberts (1982)

"Predation. Reputation and

Entry Deterrence". Journal of Economic Theory. August.

115

Rothschild, M. and J.E. Stiglitz (1970) "Increasing Risk: I, A

definition", Journal of Economic Theory, vol. 2., pp.
225-243.

Sapington, D.
(1983),
"Limited Liability Contracts between
Principal and Agent", Journal of Economic Theory, 29,
pp. 1-21.
Townsend, R.M. (1979), "Optimal Contracts and Competitive
Markets with Costly State Verification", Journal of
Economic Theory, 21, pp. 265-293.

CAUSAL RELATIONS AHONG THE SOURCES OF HONEY SUPPLY

THE PORTUGUESE CASE
Hario AN TAO - University of Lisboa (ISE)

INTRODUCTION

I.

The global
is usually

monetary situation

of a national economy

simply expressed by the following CBS (consolidated

banking system) equation:

( 1)

DLX

where DLX

CLSP

is net

CLEP

HZ + DIV

foreign reserves.

public administrative

sector. CLEP

the

made

private

sector

individuals. HZ

is the

up

of

volume of

CLSP is net credit to the

is net

internal credit to

companies

and

private

monetary assets held by the

private sector and DIV are sundry items.

In the
since the

monetary policy gradually adopted in Portugal

second

half

instrument employed
(total internal

of

the

seventies

major

policy

by the authorities has been control of TIC

credit). However.

monetary planning

the

has in

since the CLSP component in

general been considered an exogenous

variable. particularly due to the frequent budget deficits

financed largely through the creation of money by the central
bank. TIC control has mostly involved regulating the CLEP as a
means for achieving proposed objectives in terms of output.
inflation and balance of payments.
This brief

outline of

the fundamental nature of the

monetary policy implemented raises certain important questions.

some of which are listed below:
a) Have crowding out effects been experienced?
b) In

accordance
balance of

with

the

payments been

the expansionist

monetary

approach.

has

the

fundamentally determined by

or contractionist

nature of

credit

policy and in particular by the importance of CLSP ?

118

A number
type

of

of approaches can be adopted to answer this

question.

and

identification

of

possible

direct

feedback relations is of prime interest.

Empirical analysis
recently

been

possible

autoregressive
adjusted and

of dynamic feedback relations has

(1)

systems

using

modelling

stationary variables

of

new

method-vector

previously

- hereinafter

seasonally
referred

to

simply as VAR.
An alternative
consists

of

method used in classical econometrics

specifying

equations. from

which

models
in

for

general

structural
simplified

simultaneous
structures

in

reduced form can be derived.

The choice
dichotomous

to

nature.

disadvantages

be

made

Weighing

normally

could lead

correct procedure

in an

and

suggestions thereby
dynami~

econometric

necessarily

not

up

expressed

methodologies (2)
specification

is

the

of

advantages

for

each

to situations

of

a
and

these

in which the most

initial phase is to make use of a VAR

subsequently,
obtained, to

using

the

information

and

go on to specify a classical

model. This is one such situation, as will

be seen below.
In summary

the advantages normally attributed to VAR

models are mostly the following:

a) Sound evidence can be obtained for classifying variables
into

endogenous

and

exogenous

and

for

excluding

variables which are not very soundly justified ;

b) A

relatively precise
number of

indication can be obtained on the

lags in each variable to be included in the

model ;
(1) C.
Sims (1980)
produced an important pionnering paper. In
terms of VAR specifications for the Portuguese economy. only
two references are known prior to this - Barbosa, A.
Pinto
(1984) and Teixeira dos Santos, F. (1986). The first case is a
bivariate model on the relation between inflation and output
while the second is a trivariate model adding those variables
to the money supply.
(2) On this issue see Genberg. H. et al ..
(1984), for example.

(1987) and Buiter, W.

119

c) Reasonable precision is possible in visualising the main

transmission mechanisms

through

which

macroeconomic

policies or exogenous shocks on uncontrolled variables

produce effects

on target

action governing

variables and the rules of

administration of

such policies can

be inferred.
There are
discussion

two basic

should

be

reasons why

subject

to

the problem

particularly

under
careful

methodological treatment open to various options.

Firstly, economic

expectations play

a vital role in

the development of the principal variables. Thus it is far from

certain that

future behaviour

will substantially reflect past

development. At the same time Lucas' critique on the effects of

macnoeconomic policy
policy measures
activity could

could be

highly relevant, and if so only

not anticipated
influence the

by those involved in economic

present

and

future

course

of

target variables.
In

this

context

it

dynamic models with rational

general

these

simplification
empirical

models
which

is

therefore

e~pectations

require

which

they

hypotheses

distance

might

that

should be adopted.

strong

considerably

realities

suggested

them

explain

from

In
for
the

predominate.

Furthermore, the estimation of econometric models with rational

expectations generally
exogenous variables

demands that

be forecast.

the future

The

quality

behaviour
of

of

estimation

therefore depends on minimising the forecasting errors for such

variables (and
within the

also on

eliminating possible

structure of

multicolinearity

forecasts) by using ARIMA forecasting

models, transfer functions or autoregressive vectors.

Secondly,
function against

monetary

and

foreign

exchange

policies

a background of considerable uncertainty. and

significant forecasting

errors on

the part of the authorities

are frequent.
Moreover,

in

the Portuguese

between the

seventies and

changes and

financial

probably affected

eighties, there

innovations

the nature

predominant variables

case for

and the

of some

of causal

a long period

were institutional
significance which

relations between the

stability of

the parameters of

120

the models estimated.

To justify the relevance of these observations to the
Portuguese case

we will

begin with

a closer

examination

of

monetary programming.
In a stylised way the programming which has been made
for one

or more

years could be described taking the change in

annual variations in flows of the balance sheet restriction (1)

as a

yardstick. To

explanation (3),
increase in

summarise

the

the increase

in

traditional
demand

central

for

H2

bank

which

the

the money supply should satisfy in the near future

is derived from the basis of forecasts for y (rate of growth of

real output)
future

and u

levels

of

administratively)
variation (4),
residually,

(rate of inflation) and in accordance with

interest

rates

(to

be

; having established an objective for the DLX

recourse to external credit (EC)

in

development of

established

accordance
the CAB

with

the

is determined,

forecast

available

on

(current account balance), and also by

virtue of (1) the maximum variation possible in the CIT is also

obtained;

finally, having

identified public sector financing

needs, the CLSP and residually the CLEP are determined.

It could
characterises the

be maintained
period from

that this outline reasonably

1977 to

1985, despite the fact

that improvements in monetary and foreign exchange programming,

changes in

objectives or

pursued, and

stages of

the emphasis

with which

they

were

expansion and contraction (stop-go)

in

administering the respective policies, normally contrary to the

cycle of

the international

this period.

Furthermore, since

the financial

markets,

investment been
alternative to
which is
by the

provided

not

with

bank credit,

1986 with

only
an

has

the stimulation of

financing

important

and

of

private

attractive

but also budget deficit financing

still excessive, has in large measure been taken over

bond market.

position is
(5))

economy, have been observed during

which,

In this

adopted by
placed

in

context a particularly important

BT issues
the

(new type of Treasury Bills

primary

market

(3) See 1979 Bank of Portugal Report, pp.

with

financial

133.

(4) From mid-1983 the objective became the CAB balance.

121

institutions, were eventually mostly transferred to companies

and individuals
either permanently or through repurchase
aggreements.
Furthermore, during the first half of the seventies
and in previous periods the nature of monetary policy and its
background were considerably different from what has just been
described.
As regards the uncertainties and inaccuracies faced
by the

central bank,

we will

illustrate them

by summarising

their own claims (6) relative to 1985.

The forecast for u for 1985 was 22 %, which turned
out to be excessive since the rate of inflation eventually
observed was 19,3 % mostly due to the unforeseen fall in the
dollar. The target established for y was 3 %, reflected in
nominal terms in a 25 % growth of output, which involved an
external deficit of 4,1 % (CAB/GDP) . There was, however, an
external surplus of 1.8 % which in fact would have been even
greater had
monetary policy not been involuntarily more
expansionist than planned.
This enormous divergence between what was forecast
and what actually happened meant that, when the panned CIT
ceiling was not adjusted in time, there was an increase in the
counterbalances of the creation of money supply greater than
forecast.
turn this unforeseen expansionist shock
In its
unleashed powerful effects amplifying or correcting divergences
in other variables - a fall
in the deceleration rate of u,
increase in nominal y and overheated domestic demand, among
olher things.
The moral to be drawn from these considerations is
therefore substantiated with a note of caution regarding the
quality of econometric results obtainable within the framework
of the different methodologies applicable. This is more than
sufficient reason
out a VAR model.

to adopt

an eclectic methodology and to try

(5) The first issue of these TBs was August 1985.

(6) See 1985 Bank of Portugal report, pp. 159.

122

The outline of this article is therefore as follows:

Section II
the VAR

explains the

for two

periods is

1972.111-1986.IVand
there was

selected

of policy

; in Section III
longer

one

from

to the period when

another corresponding

most homogeneity

1986.IV
of the

methodology adopted

mixes. from

1976.1

to

Finally. Section IV examines the power of explanation

models by

(impulse

means of

response

two types

functions)

forecast error

variance). and

also presented

on possible

of analyses

and

DFEV

called IRF

(decomposition

certain final

of

observations are

extensions for models suggested by

the respective results.

II. VAR HODEL PROCEDURES
VAR models
between variables
series to
be more

with trend-free

or seasonally adjusted time

stringent (7)

we must obtain series that are at least

i.e. series

whose first and second moments

and invariable relative to time. Normally the first

a representation

information contained
variances and
on and

relations

the resulting series are stationary. To

are finite

with

in data

zero
in a

covariances {n t . a }

summarised.

series in

(1969) causal

ensure that

weakly stationary.
moment (

analyse Granger

such a

mean)

is

removed

sequence of

matrices

and
of

{E(Yty'a)} is concentrated

It must be noted that for weakly stationary

sequence the

matrices

depend

on

the

t-s

difference and not on the t and s moments individually.

For better

clarification of what is to follow let us

consider that VAR model construction includes five stages.

The

first

stage

has

just

been

referred

to

transformation of original into stationary series.

In general
first differences.

it is

sufficient to

use the operator of

once or twice. to remove trends existing in

the original

series. Seasonal

generally be

eliminated with

influence. when
the

help

of

the

suspected. can
operator

of

differences or by means of seasonal dummies.

There are various tests for identifying the existence
of stationarity.

or more

correctly. for not rejecting it. The

(7) See for example Aoki. M.

(1986).

123

function of

autocorrelation plays

respect, as

is known,

respective graph
or are
the

and in

makes it

not autoregressive
residual

variables

critical

many cases

role

in

examination of

this
the

possible to infer whether there are

coefficients or
of

moving averages in

supposedly

the

stationary

deseasonalised series.
Two

are

tests

frequently

used

to

test

the

significance of the autocorrelation coefficient.

These are

the t

significance Bartlett test for each

coefficient and the Box-Pierce (0 statistic) test.

It is
spectral

also useful to use partial autocorrelation and

analysis

stationality or
As

regards

to

corroborate

and

deepen

awareness

of

lack of seasonality in the transformed series.

the

latter

study

of

the

Durbin

accumulated

periodgram graph and concomitant application of the KolmogorovSmirnov test are particularly important.
All

these

procedures

have

been

adopted

in

the

Y(t)

of

empirical analysis in the following section (8).

After the

first stage

there is

vector

stationary stochastic variables whose relations, to be put into

models, represent

the functionning

of a given economic system

in dynamic terms.
The second

stage consists

the first,

which we

will call

Sims first

suggested it

of selecting two models :

the VAR (Sims) model, since C.

in 1980,

imposes the restriction that

all variables

of the

second, which

we have called the restricted model, arises from

using an

model have the same number of lags ; the

optimal criteria

in general

suggested by Hsiao (1981,

lags varies

1982) and

in this

model the

number of

between

As regards

the first

model the choice of M could be

variables.
imposed ad
the Hsiao

hoc. Furthermore there is nothing to prevent use of

criterion for

establishing M, the procedure we have

adopted.
We will

now go

on to

the question

of choosing the

(8) The results of these tests have not been included so as not
to overload the text. Figures are available on request, as are
other results quoted but not presented.

124

second mode I .
In

formal

terms

VAR

model

proposes

an

autoregressive representation which can be denoted by

(2)

Y( t)

where D(t)
is an
i

vector of deterministic regressors and e(t)

is the

variable

error

1, ... ,H, as

vector, not

a white

noise

correlated to

type.

The

any Y(t-i),
1, ... ; H)

Ai ( i

represent squared matrices of the lagged variables coefficients

and

the

zero

precisely

restrictions

from

criterion. This
many steps

of

Hsiao criterion

as there

estimation the

which

application

might
the

be

imposed

aoresaid

result

efficiency

consists of minimising,

in as

are endogenous variables, by means of OLS

Akaike statistic

relative to

final prediction

error - which we will call the G statistic, where (9)

N+P

SSR

N-p

N
It must be noted that adoption of the Hsiao criterion

to specify

the polynomial structures of lags in variables also

allows the

concept of

operationally,

and

Granger causality

(1969) to be defined

brief

on

digression

this

matter

is

therefore justified here.

The
variables in

Granger

concept

a simultaneous

defines

equation

causality
model

between

according

to

temporal precedence criterion.

This will be illustrated with a trivariate model.
Let Z(t)
at the

moment t

stationary and

".

'" ...

(P,E,R) be the past information available

on the

P,E and

R variables representing the

deseasonalised series

transformation of the

respectively obtained by

cLsp, CLEP and DLX series.

(9) For a summary explanation on this procedure see Teixeira

dos Santos,
op.cit .. It will be noted that the G statistic
corresponds to the Akaike,
H. (1969)
statistic for measuring
the final
prediction error
(FPE), in which N refers to the
number of observations, P the number of regressors and SSR the
sum of squared residuals.

125

Each pair
different types
identified in

of

identify

In

Granger

a trivariate

causal relations
figures.

of variables

causal

Granger

relations

causal

which

I defines

relations
to the

among

the

be

all these

have been

Portuguese

can

G statistic

relations among

these relations

variables, corresponding

found in one of the

model. Annex

by quantitative

this study

seventies and

will be

applied to

P,

case

and

during

R
the

eighties, and the main conclusions drawn will be

outlined below.
Merely for
formal accuracy,

the

purposes

the group

of

illustration

of possibilities

and

for the

for

P and R

pair are presented below

a) R

directly causes

P iff

(if

and

only

if)

the

prediction improves when R is included in the group of

regressors used.
explained in
causes P

In the terms of the notation used and

Annex I,

it is

said

that

"R causes 2P and causes 3P",

directly

i.e. either in

a bivariate or in a trivariate model.

b) R

indirectly causes P iff "R causes 2P, but less than

E, R does not cause 3P, R causes 2E and causes 3E".

c) Type

I spurious

causality between Rand P exists iff

"R does not cause 2P and causes 3P".

d) Type

II spurious

causality between Rand P exists if

"R causes 2P, but less than E, and R does not cause 3P
(as in

b)

in addition,

opposed to

b) and

reverse of

b)".

E is

used, R

used the

E causes

not cause

2R and

causes

3E
3R

(as
(the

In this case it can be seen that when

does not

cause E. However,

criterion suggests

situation could

R does

occur when

that R

if E is not

causes

P.

This

E directly causes P and R

is a "proxy" of E.
e) There is a direct causal relation ("feedback") between
Rand

observed.

whether

relation

a)

or

the

reverse

is

126

f) Finally.

R does

not cause

conditions are identified.

P if none of the previous

i.e ..

f.l)

"R does not cause 2P or cause 3P".

f.2)

"R does not cause 2P and E does not cause 2P".

It must

be noted that this situation is equivalent to

saying that

the absolute G minimum is obtained at the

end of the first step. which allows P to be defined as

an exogenous variable.
f.3)

"R

does not

cause 3P or 3E and E does not cause

2R or does not cause 3R".

It is

now clear

that at the end of the second stage

there is a restricted VAR model available in the sense referred

to.
It cannot

be stated however. that it contai n.s any of

the causal

relat ions just

problem of

veri fy i ng the

furthermore under! ie
consisting of
the various

outlined. There
nature

all VAR

comparing the

of

these

models. and

is

therefore

relat ions.
there is

the
which

method

G statistic figures associated to

alternatives available. We would like to point out

that as yet we have not found any reference to this question in

the literature.

which seems

rather significant.

and we

will

therefore deal with the issue in the following section.

To conclude

the outline

of

the

second

stage

the

restricted model obtained can be presented as follows

[i] [::1.

A( )
[ D( )
G( )

(3 )

where the

letters A and I refer to polynomials in the backward

shift operator L and where their number can be ascertained with

the aforesaid

procedure

in

steps

for

minimising

the

statistic.
In the third stage the accuracy of the model selected
in

the

previous

stage

is

tested.

comparing

it

to

close

127

variants, with more or less zero restrictions.

section the

most important

explained and

cases in

methodological

which in

In the following

elements

will

be

our opinion the stage should

not be omitted will be characterised.

In

the

fourth

plausible restricted
mean square
for out

stage

the

predictive

of

models is examined, and RMSE (root of the

error) and

U Theil's statistics are normally used

of sample predictions. As a comparison

to analyse

capacity

the predictive

~e

also decided

of the VAR (Sims) reference

capa~ity

model.
After

this

procedure

definitively

selected

restricted model is obtained to allow the fifth and final stage

to be

approached, where

(to the

the system's

ortonormalised responses

studied. This
model to

in the

necessitates changing

error variables)

the form

is

of the selected

express the endogenous variables as a function of.the

innovations

(moving

transformation is
to which

response to innovations

average

representation).

This

based on a recognised Wold theorem according

any stochastic

by the

sum of

of the

moving average

stationary process can be represented

two components, one deterministic and the other

type. Expression

(2) can

therefore be

rewritten as
D(t) + e(t)

(4)

[I - J(L)]

where I

is the identity matrix and J(L) a matrix of polynomial

y(t)

functions in

the L

operator, with

orders between

Denoting the

non-deterministic component

and

M.

of Y(t) as W(t), arid

assuming sufficient conditions for the existenoe of the inverse

matrix, this can be written as
(5)
where

W(t) = [I - J(L)]-l e(t) = M(L).e(t)

M(L)

is

therefore

the

aforesaid

moving

average

representat i on ..
Having reformulated

the model there are two types of

question to consider
a) How

the innovations

affect the future values of each

128

variable (the so-called "impulse response functions");

b) What

is the relative importance of each innovation in

explaining

the

variables

forecast

of

the

(decomposition

error

model

of

in

variance
future

of

periods
variance

error

forecasting

the

analysis).
To conclude

this methodological

explanation certain

problems arising in this last stage must also be dealt with.

Firstly. since
in

different

units

the variables

of

are normally expressed

measurement.

the

impulses

must

be

standarised by equalising unit variation of each error variable

to the respective standard deviation.
Secondly.

the

error

contemporaneously related.

It is

variables
therefore

are
not

generally
possible

to

disturb only one equation in the system each time. with each of
the

possible

diagonalising
residuals. The

n.

unitary
the

innovations.
matrix

of

the

diagonalisation procedure

without

previously

covariances
adopted by

of

the

Sims (in

1980) consists of premultiplying the

e(t) vector by a D inferior triangular matrix. with ones in the
principle
diagonal.
which
transorms
e(t)
into
a
the aforesaid

article of

contemporaneously independent error variables vector.

In

this

way

the

formula

(5)

can

be

rewritten.

simplifying the notation and putting t at a lower index.

(5')

Wt

= C(L)r t =

I
Ctrt-l
i=O

j. With
this error
variables
is the innovation occuring at t and which
at time only effects the j equation of the n equations in the
system. However.
in subsequent periods the effects of an
impulse on rJt spread through the whole system, propagated by
the C 1 matrices.
As Buiter (1984) observes.
the problem with this
procedure is that there are n!
linearly independent error

=
0,
with OrlrJ
representation. rJt

129

orthogonalisation procedures.

The C l

matrices

can

therefore

theoretically be rather sensitive in this respect.

However, as

Doan and

Litterman (1981)

have stated,

adopting Choleski's decomposition for the orthogonalisation, it

is to

be expected that the order of variables in the factoring

prOcess will
provided

not

the

significantly

correlation

influence

between

the

Cl

matrices,

innovations

is

not

very

significant.
It must
applied rather

also be noted that the Choleski procedure is

generally -

the Gram-Schmidt

in fact

procedure (which

it corresponds exactly to
is frequent

in

non-linear

programming algorithms), when the series under study are weakly

stat~onary.

Furthermore,

the

simplified considerably
procedure in

particular,

matrix arising
as

linear

information as

Yl

of

innovation

are

triangular

block

contains

these

the

same

type

of

1, ... ,N vectors with zero mean and

is the group of innovations of a

vectors, established

series, if and only if whatever k,k

standard deviations

and

vectors, or to be more technical, the

i

finite variance
data

vectors,

time. This being the case, the

vectors

the data

group of independent Yl
constant and

prediction

non-singular

invariable in

innovation

group of

the

combinations

of

of

from it allows the data vectors to be expressed

combinations are
group

problems

for such series. By using the Choleski

generated by

in

weakly

stationary

1, ... , N is, the space of

the sub-group YJ

1, ... k

is identical to that generated by the YJ homologous subgroup.

In

the

present

study,

and

to

certain

extent

reflecting the concern outlined, the Choleski decomposition was

used

and

models

were

estimation

procedure,

(seemingly

unrelated

subject
which

to

we

will

regressions),

estimated coefficients

with the

the
so

recognised
abbreviate
as

to

Zellner
as

SUR

correct

the

information contained

in the

contemporaneous correlation between error variables.

III. SELECTION OF THE TRIVARIATE HODEL AND RESULTS
Before the

explanation it should be pointed out that

all results mentioned in this and the following section are set

130

down in Annex II.

A quaterly data base was first constructed from
1965.IV to 1986.IV for the aforementioned original series.
CLSP. CLEP and DLX (10).
The following procedures were carried out to specify
the intended VAR model. Original series were transformed into
stationary series (11). after previously applying to the
logarithms the Box-Jenkins conventional model
(6 )

Y-(t)

where LiY(t) = Y(t-i). and Y(t) is the vector of the logarithms

of the original series.
After the first stage specification of the benchmark
VAR (Sims) model was carried out.
It will be noted that since all equations in the VAR
(Sims) model have the same regressors. OLS estimation from
equation to equation is asymptotically efficient.
To establish M. rather than arbitrarily imposing a
number. the first step of the Hsiao criterion was used. Thus
for each period and for each dependent variable. p. E and R
respectively. the number of lags which minimize the G statistic
were identified. The greatest of these numbers was then taken
for M.
(10) Explanatory advantages cannot a priori be attributed to
the variant in the retained restriction (1) in relation to
other close variants. However. a prelimirary test using VAR
(Sims) models was carried out with M = 8 for the 1976.I-1986.IV
period for the various possible alternatives. Results obtained
show the retained variant to be far superior in terms of
quality of fits and other normal statistical indicators.
(11) Application of model (6) was shown to be sufficient for
ensuring stationarity and seasonal adjustment. Other variants
were tested to confirm the relative good quality of this model.
in particular using the first differences operator a second
time and/or dummies for detecting seasonal influence. Barlett
tests and Q and Q- statistic values. and also examination of
autocorrelation and partial correlation functions does not show
any gains worthy of mention in terms of statiscal quality
relative to the alternative adopted. and the latter are
therefore retained.

131

It was

then observed

(76.I-86.IV) in
were more
previous

which monetary

consistent and

= 8.

establish H

years

was

in

observed that

the interval

was, however,

seen that

prediction error
was considered

for the

sequence

(FPE) very
to be

Table I,

of

to

period
to

to
rise

demanded and

the

it was

exceed 72.III-86.IV.
estimates

It

G statistic

values, the
of

the

final

close to each other, for which it

perfectly sufficient

exaggerated to establish H

a
have

original series,

could not
> 8

sufficient

such

the lags

period

exchange policies

H would

that

for H

shorter

was

extending

view of

data available

and foreign

noted

that in

generally presented

the

coordinated, it

However,
it

considerably. so
limitation of

that for

and

perhaps

even

16 for the longer period (12).

Annex II,

presents

the

main

statistical

indicators of the OLS estimation for the VAR model selected. As

can be

seen from this table the 0- (13) statistic does support

the hypothesis that there is no autocorrelation in the residual

series of
fits are

the three

equations estimated

generally acceptable,

except

in each
for

the

period. The
dependent

variable equation in the short period.

It will
F-test in

be noted

both periods

possibility of

that the significance level of the

suggests crowding

explaining the

out effects and the

DLX fluctuations

by

means

of

variations in the CLEP.

(12) Orthodox application of the first step of the Hsiao
criterion suggests that H
12 and H
19 should have been
established respectively for the short and longer periods.
However, the G statistic values respectively for H = 8 and H =
16 are not significantly greater than those obtained when H
reaches those values, so that according to an "Occam's razor"
logic, the 'option taken is justified. The relevance of this
alternative is
further demonstrated when G is minimised
throughout the three stages of the Hsiao criterion
the
absolute minimums for G with H = 8 and H = 16 do not differ
significantly from such minimums where H = 12 and H = 19.
(13) The statistic

0-

= N(N+2)

I (N-k)-l 2(k)
k=l

is an adjusted and more stringent

test C'eg. Harvey (1981).

version of

the Box-Pierce

132

The Hsiao procedure was then used in three stages for

N+p

minimising the G = ___ .~2 statistic, referred to in Section II.

N

The following restricted VAR models were obtained, reflected in

the different orders of the lagging polynomials of A to I :

(7 )

(8 )

c(O)]

[i1.

[AlB)
D(6)

[it

[A( 16)

8(6)
E(8) F(O)
G(O) H(6) 1(8)

D(6)
G(O)

respectively for

[i] [::1.

8(3) C (0) ]
E(8) F(O)
H(3) I ( 16)

[i]

1976.1-1986.IVand

[::],

for 1972.111-1986.IV.

It

can be seen that the models have very similar patterns.

Models (7)
optimised according

and (8)

represent the VAR specifications

to the

Hsiao criterion

and therefore

in

theory should not be modified, which means that the third stage
referred to

should

application of
minimums for
than the

excluded.

the Hsiao
G for

Furthermore,

criterion shows

each equation

if

that

are not

the

empirical
absolute

significantly less

G figures corresponding to polynomial structures with

fewer lags,
of the

be

the problem

additional lags

other hand,
than what

if M

can be posed of testing the relevance

thereby included

has been

corresponds to

first step

of the

in the model. On the

established ad hoc at a value less

the minimum

Hsiao criterion,

of G

obtained from the

the significance

of

the

omission of these extra lags must also be tested.

It is
carrying out
VAR model
believe

therefore our
or excluding

opinion that the suitability of

the third

stage (fine tune) of the

construction should be evaluated in this context. We

the

clarifies a

heuristic

criterion

methodological problem

just

suggested

which has

suitably

not as yet been

correctly answered.
This problem is important in so far as it might arise
whenever the case under study is characterised by the existence

133

of relatively high lags, which in fact occurs with our

trivariate VAR model. The significance of a certain number of
lags is easy to test. A statistic described in Doan and
Litterman, op. cit., can be used for this purpose, which we.
will call A, ~here A = (N-c)[10gII , I-logII 2 1), which has an f(2
distribution. I,

and I2

covariances obtained

are the residual series variances and

by OLS

estimation applied

to

each

VAR

model pair which only differ from each other because of the lag
interval being
the so

tested. N is the

called multiplication

number of observations, c is

correction factor

IIll is the determinant of I l .

Having carried out the

third step

and as

it was

usual

observed

that certain adjustments should be made either to (7) or (8) so

that the selected restricted VAR models were

(9 )

( 10)

[it

[i1.

[A(8)
0(6)

8(4)
E(O) C(4)]
F(O)
G(O) H(4) 1(8)

[A(0(3)16)
G(O)

[p] ["']
E
R

B(3) C(O) ]
E(8) F(3)
H(3) I (8)

[i]

respectively for the short and long

We

then

e2
e",

[::1.

periods~

identified

the

models

accordance

implied in

these

defined in

Annex I.

in

Granger

causal
with

the

relations
criterion

As has been said, this is an excerc1se we

have not found in recognised VAR literature.

The principal

conclusions of

this analysis (14) are

(14) By way of example the G statistic values throughout the

three-stage minimisation process for model (7) necessary for
establishing this type of conclusion are as follows:
For P var
G(8.6,0) = 8.304
G(8.0.0)
10.82
G(8.7.6)
12.22
G(B.0.6) = 9.735

FPE
For E
G(6.8.0) =
G(0.8.0) =
G(7,8.6)
G(0.B.6) =
G(6.B.7)

X 10 3
var
0.309
0.415
0.293
0.348
0.505

For R var
G(0.6.8) = 336.5
G(6.0.8) = 498.0
G(0.0.8)
546.1
G(7.6.8) = 600.1
G(6.7.B) = 600.2

134

as follows
a)

In

model (7),

there is

direct

feedback

relation

between E and P, while R causes neither P or E. On the

other hand,
occur for

P indirectly

causes R and the same would

E if the indirect causality did not have to

be fulfilled
than E",

by the condition:

this necessarily

if "P causes 2R,

less

excludes E as the indirect

cause of R.
b) The

transformations on (7) which cause (9) provides an

: all

interesting result
direct

relations,

feedback

compatible with
the next

the variables

the logic

which

have
is

paired

perfectly

of the dynamic analysis in

Section, although

it

is

not

necessary

condition.
c) In

(8), E

directly causes

spurious causal

P while

relation in

governs R

in explaining

neither P

nor R

cause E.

R has

relation to

a type
P.

II
thus

P fluctuations. Furthermore,
Finally, E

governs P

and

directly causes R.
Comparison of
the years
(in the

72 to

75 weakens the consistency of monetary policy

long period

referred to

(8) and (7) suggests that inclusion of

E is

above, while

an exogenous

variable in the sense

in the short period its fluctuations

are also dependent on crowding out effects).

d) As

for the

which give

short period,
rise to

the transformations

(10) also

on (8)

produce more desirable

causal relations.

For the

modifications are

not introduced the relations remain

first

equation

in

which

This should be read as follows.

For example,
G(8,7,6),
for
regressions in which P is the dependent variable means that the
minimum G statistic in the third Hsiao step
(after having
established the number of P and R lags as 8 and 6 respectively
in the two previous steps) was achieved when 7 lags were
allowed for E. In terms of notation used in model (7), G(8,7,6)
would be expressed as A(8) 8(7) C(6). Since G(8,7,6) does not
correspond to the absolute minimum which as seen was observed
for G (8,6,0), A(8) 8(6) C(O) was established in model (7).

135

: E governs R and directly causes P. Furthermore, both

P and

R now

directly cause

E. Finally,

despite the

reduction in the I polynomial, the causal relations in

the third

equation are

unchanged

E governs

P and

directly causes R.
We believe
of analysis.

the relations

Identification of

implicit in

VAR models

postulating the

intrinsic dynamics
usefulness of

is very

dominant

for

terms, there

implication relations,

and

system under

information

classical econometric

important,

mechanisms

of the

such

obtained justify this type

the types of Granger causality

in both

particularly
directions

for

in

the

study. As regards the

constructing
are still

models

in

reservations

senses, between Granger causal

and non-invariant relations (IS) do not strictly exist.

Beginning the
VAR model,

it is

estimated with

the Zellner

estimations for

The

SUR procedure

main

the (7)-(8)

Table II.

quality of
and that

in the

construction of a

first seen that the (7) and (10) models were

previous Section.
shown in

fourth step

Thus it

the (9)-(10)
both are

statistical
and (9)-(10)
can be

referred to
indicator

in the

in

these

pairs of models are

seen that the statistical

pair is superior to the (7)-(8) pair,

superior to

the corresponding

VAR

(Sims)

models in Table I.
An additional
(10) pair

should be

performance, using

test on the value of choosing the (g)made in

terms of

comparative predictive

the RMSE (root of average square error) and

U Theil statistics, as has been said.

Results obtained for out of sample predictions as far
as the fourth future quarter are shown in Table III.
It can be seen that (g) and (10) are clearly superior
as forecasting

models to

the others

and also

to the "naive"

model, which presupposes no ohange throughout the period.

Considering the

evidence produced as a whole,

it can

therefore be

concluded that the obvious choice is that of pair

(9)-(10). We

will now

go on

to the dynamic analysis which is

the subject of the next Section.

(15) See Buiter, W., op. cit ..

136

ANALYSIS OF

IV. THE DYNAMICS OF THE MODELS

INNOVATIONS
As has
on results

been stated in Section II we will now comment

obtained with

two types of analysis concerning the

dynamic behaviour of the system :

a) Impulse

response functions

the reactions

of the

which can be used to infer

variables in the system to each

possible impulse throughout the period ;

b) Decomposition

of forecasting error variance, which can

be used to observe the development of the significance

of each

innovation in

the forecasting error variance

for each variable.

Accumulated figures

for up to 20 future quarters for the short

period and 24 for the long period were calculated.

The order of variables adopted in orthogonal ising the
entire covariance matrix of the innovations corresponds to that
of the models retained, i.e., P, E, R,
the standard

used was

in descending order, and

naturally standard

deviation for

each

innovation.
For model
response functions

(9) Figures
of P,

to 3 show standards for the

E and R to each innovation, which we

have referred to as i p , i. and i r

Figures

4 to 6 show the same

information for model (10).

For the 1976.I-1986.IV period and therefore for model
(9) the most important observations are as follows:
a) The

P responses

the expected
for ip and i e
b) The

are long lasting, increasing and have

signs (+ for innovations in P and E, ie.

and - for R innovation,

responses

differentiated:

have

the

when the

expected

ie., for i r
signs

and

).

are

impulse is on the variable

itself. the response is significant and long lasting;

for the
the first

i p impulse the response is significant during

year and

is attenuated

subsequently

(the

137

the response to an impulse on R

crowding out effect)

is weak ;

c) The response of R to innovations on the variable itself

is intense but temporary and is weak relative to i p
and ie, with a minus sign for the first and a plus
sign for the second case. This observation means it is
possible that the R fluctuations depend more on other
variables than on
monetary policy.

the

CIT

ceiling

established

in

For model
(10), corresponding to the 1972.111-1986.IV period,
the basic evidence is as follows:
d) P responses are not long lasting, they are more intense
when the impulse is on the variable itself and they
are irrelevant for i r .
e) E responses are temporary too and the response of ir is
also irrelevant. The crowding out hypothesis is still
suggested by the evidence and is particularly intense
during the first year, as in the previous case.
f) Although there are increased fluctuations the response
pattern for R is similar to that identified for model
(9) .

We will now analyse the evidence on decomposition of the

forecasting error variance for models (9) and (10), which are
reproduced in Tables IV and V.
It will be noted that the patterns identified in the
two periods are very similar, and the following observations
therefore cover both cases :
g) The fluctuation of each variable is particularly due to
innovations in the variable itself.
h)

Impulses on R have
variability of P and

a significant effect on the

E in the 1976.1-1986.IV period,

138

and in

all cases

it is greater than observed for the

extended period, as expected.

i)

The

combined

effect

of

reasonably significant

impulses

on

in explaining

and

is

the variability

of R.
In summary, this analysis confirms the previous one relative to
the impulse
answers to

response functions,
the two

and

both

suggest

that

the

questions asked in the introduction are as

follows
1) Crowding

out effects

the adoption

have existed, particularly after

of more

consistent monetary and foreign

exchange policies (1976.I-1986.IV period).

2)

In

this period,

although it also helps to explain the

DLX fluctuations,

internal credit

does not

seem

to

have had a dominant role.

The

observation

latter

could

be

reconcilied

the

with

possibility of Portuguese experience subsequent to 1976 showing

the existence

of a

significant trade-off between the stock of

foreign currency reserves (DLX) and the effective exchange rate

(TCE), particularly in terms of the alternative between loss of
foreign exchange or devaluation.
There is

an additional

possibility that

there were

pronounced alterations in preferences of the different economic

policy makers

relative to the two trade off alternatives under

consideration throughout
former sometimes
therefore not

the period. with a leaning toward the

and toward

evident that

the latter

other

the trade-off

times.

It

is

relation was stable

throughout the period, or even that it existed systematically.

Furthermore. devaluation
likely to

have a

extent they
situations or

high positive

are complementary
time intervals

and contraction
relation since

of CIT are
to a

instruments. However.
in which

certain
a priori

such correlations have

been relevant or have had the opposite sign cannot be excluded.

Such questions

of great importance in explaining the

139

nature of

monetary and

clarified by

foreign exchange

empirical analysis.

policies can only be

They stimulate

and

suggest

directions for extending and refining this analysis. Given that

Portugal is a small open economy it is also relevant to measure
the relative

importance of

extensions are
which could
Given the
expected

from the

reliability of

confirm them.

Naturally such

an indirect test of the validity of conclusions

be drawn

that

external impulses.

the

present trivariate VAR models.

most results

foresaid

extension

presented,
and

it is to be

refinement

will

140

ANNEX I
Types of Granger causal relations resulting from application of
the Hsiao criterion to a trivariate YAR model.
1. Notation
A

Let Z(t)
and R.

A '" "
(P,E,R)

be all

past information available on P, E

Let G(PIP) be the minimum G value statistic in the group of OLS

regressions in

which the

explanatory variables

endogenous variable

are,

in

is

rising sequence,

and

their

the

lagged

variables P(t-i) from i=O to i=M.

"'''
.J\"
Let G(PIP,E)
and G(PIP,R)
be the minimum G values in the second

stage of

application of

the Hsiao

same way

after separate

addition of

criterion obtained

in the

the E(t-j) and R(t-k),

'"

and k from 1 to M regressors to the corresponding G(PIP) model.

The lower

of these
"

two minimums is taken.

It is accepted that

this will be G(PIP,E).

A

Let G(PIZ)

then be

application of
addition of

the minimum

the Hsiao

the R(t-l),

'"

G value in the third stage of

criterion obtained
1 from

to

"
corresponding G(PIP,E)
model.

as before

regressors

after
to

the

2. Granger causal relations.

a) R

di~ectly
...

causes P if and only if

"

,..

"

"

.A

G(PIP,R) < G(PIP) and G(PIZ) < G(PIP,E)

which leads to "R causes 2P and causes 3P".
It can

be seen

that in this case the P prediction improves

when R is included in the group of regressors.

b) R indirectly causes P i f and only i f
tilt

'"'

G(PIP,E) < G(PIP,R) < G(PIP) and G(PIZ)

....

...

""

"""

"""

G(PIP,E)

~
""

and G(EIE,R) < G(EIE) and G(EIZ) < G(EIE,P)

i.e., "R causes 2P but less than E, R does not cause 3P and
R causes 2E and causes 3E".
c) There

is type

I spurious causality from R to P if and only

141
,.A

""'''''''

if G(PIP.R) ~ G(PIP) and G(PIZ) < G(PIP.E)

and therefore "R does not cause 2P but causes 3P".
d) There is type II spurious causality from R to P if and only
"'A. < G(PIP.R)
"''''' < G(PIP)
"" and G(PIZ)
. . ~ G(PIP.E)
...."
if G(PIP.E)
""
"
;It
"'" '"""
"'"
and G(EIZ)
~ G(EIE.P)
and G(RIR.E)
< G(RIR)
and
....
-:1 .....
G(RIZ) < G(RIH.P)
i.e .. "R causes 2P but less than E.R does not cause 3P or 3E
and E causes 2R and causes 3R".
In this case whenever E is used R does not cause P. However.
if E is not included R causes P. This situation can arise
when E directly causes P and R is a good proxy of E.
e) There is a direct feedback relation between Rand P if and
only if a) and the inverse occur simultaneously. Therefore
....
...
....
... .....
.... ....
G(PIP.R) < G(PIP) and G(PIZ) < G(PIP.E) and G(RIR.P) <
....
.... ....
A
G(RIR) and G(RIZ) < G(RIR.E)

,.

f) R does not cause P if and only if

f.1) "R does not cause 2P and does not cause 3P". i.e ..
.... '"
....
""
,. A
G(PIP.R) ~ G(PIP) and G(PIZ) ~ G(PIP.E)
f.2) "R does not cause 2P and E does not cause 2P". i.e ..
"'.....
......
AA
.....
G(PIP.R) ~ G(PIP) and G(PIP.E) ~ G(PIP).
In this case the absolute minimum of the Hsiao
criterion is reached in the first step. which suggest
that P should be classified as an exogenous variable
in the system concerned.
f.3) "R does not cause 3P or 3E and E does not cause 2R or
does not cause 3R". therefore
....
..... '"
~
..........
G(PIZ) ~ G(PIP.E) and G(EIZ) ~ G(EIE.P) and
A~
.....
A
~A
G(RIR.c) ~ G(RIR) or G(RIZ) ~ G(RIH.P).

142

ANNEX II
TABLE I
Statistical indicators of OLS estimation of VAR (Sims) models
H

=8

Equation

0R- 2
Value l.s.

SSR

P
E

21.7 .24
14.8 .68
9.5 .95

.249
.006
6.41

76.1-86. IV

.95
.39
.12

c. V.
P
E
R

= 16

72. I I 1-86. I V
0*

Value l.s.

R~2

SSR

CV

20.3
23.9
13.8

.84
.62
.63

3.95
.004
4.52

P,E
E,P
E

.48
.30
.88

TABLE II
Statistical indicators of SUR estimation
of restricted VAR models
Hodel

P
E
R

Hodel (8)

(7)

0"
R- 2
Value l.s.

SSR

18.6 .41
15.4 .63
18.5 .42

.336
.033
7.96

.95
.05
.29

Hodel (9)

0*

Value l.s.

20.0 .52
9.2 .99
14.0 .87

R-:Z

SSR

.68
.37
.59

5.94
.026
5.54

Hodel (10)

R- 2
0*
Value 1.s.

SSR

13.1 .79
16.9 .53

.018
8.81

0R- 2
Value 1. s.

SSR

18.7 .60
19.7 .54

.017
6.01

------------------------------------------------18.5 .43
.95 .295
22.8 .35
.75 4.43

P
E
R

.42
.44

.44
.57

Note
l.s. means level of significance and
variables;
i.e., variables whose values
significance.

CV means causal
have F < 0.1

143

ANNEX II
TABLE III
Predictive performance of VAR models selected
VAR (Sims) Hodels
76.1-86. IV

M = 8

No future
quarters
p

RMSE U

RMSE U

M=12

RMSE U

RMSE U

72.1I1-86.IV
E

RMSE

RMSE U

--------------------------------------------------------------1 .158 .241 .015 .761 .406 1. 47 .598 1. 01 .008 .647 .262 2.53

2 .199 .369 .019 .909 .555 1. 14 .660 1. 20 .009

3 .240 .333 .015 .821 .546 1. 39 .335 1.48 .009
4 .144 .487 .012 .742 .528 2.36 .153 1. 19 .012
Hodel
P

RMSE

(7)

E
RMSE

Hodel (8 ) 72. I 11-86. I V

76.1-86. IV
U

R
RMSE

.546 .367 2.19

.388 .366 2.28
.444 .138 1. 16

RMSE

E
RMSE

R
RMSE

--------------------------------------------------------------1 .167 .255 .024 1.20 .216 .788 .517 .874 .014 1.07 .115 1. 12

2 .212 .392 .016 .785 .412 .843 .607 1. 10 .017

3 .258 .358 .009 .538 .349 .877 .304 1. 34 .019
4 .179 .608 .014 .865 .246 1. 10 .148 1. 16 .017

Hodel (9) 76.1-86. IV

P

RMSE

E
RMSE

R
RMSE

1. 01 .218 1.55
.766 .246 1. 53
.572 .208 1. 75

Hodel (10) 72. II 1-86. IV

P

RMSE

E
RMSE

R
RMSE

--------------------------------------------------------------1 .134 .205 .020 .990 .211 .780 .514 .869 .008 .589 .067 .652

2 .176 .325 .018 .889 .388 .794 .619 1. 12 .009 .540

3 .229 .319 .010 .587 .452 1. 14 .246 1.09 .012 .456
4 .166 .563 .013 .820 .153 .688 .081 .633 .013 .438

.130 .929
.095 .594
.128 1.07

144

TABLE IV
~

DECOHPOSITION OF VARIANCE
of prediction error variance in future k quarters explained
by each innovation
Hodel (9) 76.I-86.IV

Forecast
variable

K=l

Future quarters
K=4
K=8
K=12

K=16

K=20

100.0
0.0
0.0

58.0
32.2
9.8

59.7
29.7
10.6

61.8
27.0
11.2

64.0
24.8
11.2

65.6
23.2
11.2

ip
i ..
ir-

5.1
94.9
0.0

8.7
86.8
4.5

9.3
82.2
8.5

9.3
81.3
9.4

9.3
81.1
9.6

9.3
80.9
9.8

ip
i ..
i r-

7.9
1.5
90.6

6.4
22.3
71.3

6.5
23.4
70.1

6.7
23.8
69.5

6.7
24.0
69.3

6.7
24.0
69.3

Impulses

--------------------------------------------------------------

TABLE V
Hodel (10) 72.III-86.IV
Forecast
variable
Impulses
p

Future quarters
K=l

K=4

K=8

K=12

K=16

K=20

K24

100.0
0.0
0.0

72.9

62.4 58.4
37.'1 40.2
0.5 1.4

61.2
37.2
1.6

57.3
41.2
1.5

51.4
47.2
1.4

27.0
0.1

ip
8.6
11.4
11.7 12.6
12.7 12.9
15.3
87.3
85.8 84.3
84.3 84.1
91.4
81.9
i ..
0.0
1.3
2.5 3.1
3.0
3.0
2.8
ir-----------------------------------------------------------R
ip
6.8
6.5
7.8 8.3
8.6
8.6
8.6
0.0
25.9
37.3 38.6
41.7 41.2
41.7
i ..
93.2
67.6
54.9 53.1
49.7 50.2
49.7
i r-

145

Cumulated impulse response functions (IRF) to

Si,i,i~

l p

rS

Model (9)

Model (10)

1976.1 - 1986.1V

1972.ill - 1986.IV

--- .......... , - ,
......

_---------

Fig. 1 De P

Fig. 4 De P

Fig. 2 De E

Fig. 5 De E

Fig. 3 De R

Fig. 6 De R

P, E and R responses to ip
1-++

P, E and R responses to ie
P, E and R responses to ir

....

..................

..... _ - /

"

,/

146

BIBLIOGRAPHY
AKAIKE,H.
(1969)
"Statistical predictor
identification",
Anals of the Institute of Statistical Mathematics, vol.
21, 1969.
AOKI, H.
(1986), Notes on Economic Time
System theoretical
perspectives,
Berlin, 1986.

Series Analysis
Springer
Verlag,

BARBOSA, A.
PINTO (1985)
- "Infla9ao e produ9ao en Portugal
vol. IX, n l, Jan/85.
1953-1980, Economia.
O

BRANSON, W.
(1984a) "Exchange rate policy after a decade of
floating" in Exchange Rate Theory and Policy ed. BILSON,
J.E. MARSTON, R., Univ. of Chicago Press, USA, 1984.
BRANSON, W.
(1984b) "A model of exchange rate determination
with policy reaction: evidence from monthly data", Open
Economy Macroeconomics Conference, Lisbon, June/84.
A.
(1984)
"Testing for the
BURBRIDGE, J.
and HARRISON,
International
effects of oil price rises using VAR",
Economic Review, vol. 25 N2, June/84.
(1986) "Monetary policy, fiscal policy
CHOWDHURY, A.
et al
Southern Economic Journal,
and investment spending",
Feb/86.
DOAN, R. and LITTERHAN, R. - "User's manual rats version 4.30",
VAR Econometrics, Minneapolis, USA, 1981.
FACKLER, J.
(1985) "An empirical analysis of the markets for
goods, money and credit",
Journal of Money Credit and
Banking, vol. 17, n l, Feb/85.
O

GENBERG, H.
et al (1987) - "The relative importance of foreign
and domestic disturbances for aggegate fluctuations in
the open economy", Journal of Monetary Economics, vol.
19, 1987.
- "Deficit budg'taire et
GIRARDIN, E.
and HAROIS, W. (1984)
Universit' Paris I,
d'ficit externe" , Document n091,
Sept/84.
GRANGER,
C.
(1969)
"Investigating causal
relations by
econometric models
and
cross
spectral
methods",
Econometrica, vol. 37, n03, July/69.
HSIAO, C.
(1981) "Autoregressive modelling and money income
causality detection",
Journal of Monetary Economics,
vol. 7, n l, Jan/81.
O

HSIAO,

C.
(1982)
"Autoregressive modelling and causal
ordering of economic variables",
Journal of Economics
and Dynamic Control, vol. 4, Aug/82.

147

NELSON, C.

(1973) - "Applied time series analysis", Holden Day

Inc. San Fransisco. 1973.

OCDE (1982)

- "Financement du d'ficit budg'taire et contr6le

mon'taire". S'rie Etudes Mon'taires. Paris. 1982.

RIBEIRO, C.

SILVA (1987) - "Consequ3ncias das

expectativas
racionais na especifica910 de modelos econometricos"
Doc. trab. n036 CEMAPRE-ISE. Lisbon. Aug/87.

SAMUELSON, P.

(1983)- "Foundations of Economic Analysis".

Enlarged Edition.
Harvard Economic Studies.
vol.
80,
1983.

SANTOS. F.

TEIXEIRA (1986)
"Money prices and output
in
Portugal:a vector autoregression analysis". Association
Sud Europea
de Economia
Teorica.
Conf.
Papers,
Marseille. May/86.

Part 2 : MONETARY POLICY GAMES

MONETARY POLICY CREDIBILITY AND COORDINATION

Kevin CLINTON and Jean-Claude CHOURAQUI
OECD Department of Economics and Statistics
2, rue Andre Pascal 75775 PARIS CEDE X 16

In

assessing

the

appropriate

stance

of

monetary

policy. one cannot just rely on the simulated effects of policy

in econometric
policy change
and the

models.

In the real world the impact of a given

will depend heavily on the state of expectations

degree of

confidence in

the monetary authorities.

In

this regard. Section I of this paper considers some fundamental

issues related
that have

to the credibility of non-inflationary policies

received much

attention in

the

recent

analytical

literature on economic policy. Section II discusses briefly the

extent to

which the

enhanced through

efficacy

of

monetary

international cooperation.

policy
More

could

be

details

on

these topics can be found in the literature review presented in

Section III.
I. TIME CONSISTENCY AND CENTRAL BANK REPUTATION
A number

of economists

discretionary monetary
stability. This
concept of

policy may

conclusion is

classical assumptions,

it takes

A policy

fully into

to be

future.

a monetary

this sense,

zero inflation

argued

be incompatible

derived from

discretionary decisions
In

recently

that

with price

some strong

neo-

including rational expectations, and the

time consistenCY.

consistent" when

have

made by

is said

to

be

"time

account the anticipated

the authorities in the

policy aimed constantly at

can be said "time inconsistent", because if the

This paper is an extract of the OECD Department of Economics

and Statistics Working Paper N 39, entitled "Monetary Policy
in the Second Half of the 1980s : How much room for Manoeuvre
?". The views expressed therein are those of the authors and
do not necessarily represent those of the OECD or of the
Governments of its Member countries.

152

public

believe

does

that

wi 11

prices

stable.

remain

policymakers have an incentive to allow a temporary increase in

monetary

to

expansion

inflationary monetary
problem.

If

renege on

output

boost

policy may

the public

(1).

Thus.

non-

have an inherent credibility

suspects

that

the

authorities

will

their commitment of price stability. expectations of

inflation will

cause nominal

wages and interest rates to rise

immediately. Assuming. however. that the authorities maintain a

non-accomodating stance.
than expected.

inflation will

turn out

to be lower

so that realised wage rates will prove too high

for achievement of the full employment level of output and. for

the same

reasons. ex

surprisingly high.
get output

Consequently.

closer to

inflation to
words. a

post real

commit all

only by

expectations of

time-consistent policy
unless the

their future

rates

will

appear

monetary authorities can

th~

full employment

match the

some inflation.

interest

allowing

the public.

(2) will

in general

monetary authorities

policy decisions

some

In other
inv.olve

can credibly

to the objective of

price stability.
This line

of reasoning

has led

some economists

to

recommend basic reforms. that would put tight legal constraints

on the

discretionary authority of central banks. to allow only

non-inflationary

policies.

interest for

their audacity

difficult to

envisage how

important

are

the

Such

suggestions

than their
they

implications

discretionary monetary

policy as

These implications

are

consistency throws

them into

not

novel

it

implemented.

time

of
is
but

of

realism. since
be

might

are

consistency

actually
the

sharp relief.

notion
First.

more
it is
More
for

practised.
of

time

arguments

(1) See Kydland and Prescott (1977)

(2) Therefore.
a "time-consistent policy" has a different
meaning than that generally attributed to "consistent policy".
Consistency is usually regarded as a virtue. implying constancy
and. in the context of monetary policy.
a willingness to
persist with a non-inflationary stance so as to enhance its
credibility. This type of policy is defined as ~~O~~~ in
the technical
literature.
A
precommitment solution thus
represents a situation in which the public expects.
and the
monetary authorities allow. no inflation. But while this would
be the best long-run solution. it may not be compatible with
the incentive offered to policymakers.

153

that monetary authorities must take a long view and be

concerned to build a strong reputation for resisting inflation
are enhanced.

A central

creates money

rapidly,

increasing the

bank that takes a short view and

only temporarily, risks permanently
rate

inflation

output-inflation
stably linked

and

Second,

tradeoff.

monetary targets,
The central

if

in circumstances

to ultimate

worsening

the

short-run

arguments

for

announced

where money aggregates are

nominal objectives, are reinforced.

bank is then able to prove its commitment by means

of an important intermediate variable that is easily monitored.

The view that inflation could be quickly reduced with
minimal output
approaches,

requires

authorities.

But

policies is
that such

losses, as implied by some rational expectation

absolute

once

the

taken into

authorities to
has rested

incentive

for

in

the

monetary

time-inconsistent

account, there is no reason to believe

confidence can

considerable time,

confidence

exist. In

or some

practice,

very sharp

establish credibility.

on demonstrated

it

shocks,

will

for

take

monetary

In the 1980s credibility

resistance to inflation pressures.

Monetary targets, where they have been met, have helped in this
respect by

providing a yardstick for proving the commitment of

the authorities.
larger

in

the

discretion is

While the
absence

likely to

for expectations
in

the

1980s

room

of

for

targets,

be more

discretion
the

may

exercise

appear
of

such

hazardous because the anchor

is less firm. Reducing inflation expectations

has

involved

reduced

output

and

increased

unemployment, which might be regarded as part of the costs of

inflation, and
of the
associated loss of central bank
credibility of the 1970s.

II. POLICY COORDINATION

Several empirical
evaluate the
agree on

studies

potential gains

a number

that

have

attempted

to

from policy coordination broadly

of important points, despite differences in

models used. These points may be briefly stated as follows:

i)

Small gains

can be

rational insular

derived from

policies. In

coordinated

vis-a-vis

the conditions of the mid-

154

1980s, the

gains stem

monetary policies
in increased
imbalances.

in large

part from

in cooperative

output and.
Inflation

more

relaxed

solutions. which result

in some models.

is higher.

in reduced trade

but not

by so

much to

negate the net welfare benefit (3)

ii) Gains

in dynamic

models may

arise from improved timing.

Disinflationary policies would be applied less abruptly in

a cooperative
welfare loss
over time

regime so
is less,

is the

that the

present value

although the

same as

of

the

sum of output losses

in a non-cooperative regime to

achieve a given degree of desinflation.

iii) A

potentially

important

developing countries.
to the

side

benefit

accrues

to

the

which are not assumed to be parties

policy coordination. but which gain from increased

demand and

lower interest rates in the OECD area and from

improved terms of trade.

There are

two interacting

reasons for the inference

that the uncoordinated monetary policies of the 1980s have been

too tight.

First. policy

price stability

has revealed a strong preference for

relative to

output growth. Second. the models

typically show negative short-run spillovers on the price level

from monetary

policy --

i.e. an

disinflationary (inflationary)
depreciat~on

(appreciation)

preference for

price

easing

(tightening)

effect abroad
of

stability

the
of

because

domestic
itself

has

of

the

currency.

The

would

of

course

justify a degree of monetary restraint, but empirical estimates

suggest that

in conjunction

with the

second

factor

it

has

produced a systemic bias towards unduly tight policies.

However,
potential gains

important

qualifications

from coordination

attach

to

the

that have been found by the

empirical studies :

(3) Welfare is assumed in these studies to be a function of

variables such
as output,
inflation and current account
balances. The precise form of the function. in particular the
weights put on the different goal variables, can affect the mix
of policy recommendations that is derived.

155

i)

The smallness of the estimated gains and their uneven

distribution is a warning there may not be sufficient
in particular
incentive for all countries to cooperate,
for the United States.

ii)

The models differ and it is uncertain if any is an

adequate representation of the real world. Cooperation
based on incorrect models could easily be worse than noncooperation.

iii) Recent theoretical analyses have shown that coordination

may result in welfare losses, because of uncertainties and
expectational factors that have not been taken into
account in the empirical work.
Most important

in this

last

respect

is

that

the

estimated gains in dynamic models often rely implicity on timeinconsistent policies. Such policies are not necessarily more
credible just because an international agreement is made, and
they may be less credible. Because fear of exchange rate
depreciation is reduced, international coordination could yield
solutions that are too inflationary and on balance worse for
welfare over time than non-cooperative solutions. Some authors
conclude that unless binding constraints or the authorities'
strong reputations can make non-inflationary policies credible,
cooperation is futile.
III. A LITERATURE REVIEW
This Section
provides a
brief explanation
of
analytical concepts used above, and highlights some of the more
important findings in the recent literature on the issues of
time inconsistency, credibility and international coordination
(4). Although the focus is on monetary policy, the issue of

(4) The latter is defined as

i.e. one in which no country
other being worse off.

a "Pareto-efficient" situation,
can be better off without some

156

cooperation

involves

necessarily

monetary-fiscal

mix.

empirical setting

The

analysis

of flexible

discussion

some
is

limited

of

to

the

today's

exchange rates and high capital

mobility and substituability.

A. Policy cooperation as a strategic game
1. Game theory approach
In

independently,
reactions of

non-cooperative
adapting

decisions

the others.

cooperative, or
other parties'

game,

The most

to

countries
actual

or

act
expected

common assumption

of non-

competitive, behaviour among countries is that

behaviour is

taken as

given. This is known as

the "Nash assumption". An alternative concept sometimes applied

is the

"Stackelberg assumption"

leader, setting
as best

as

since it is a fairly robust finding in the

multicountry models

that

countries

assumption, with
be a

country acts

its strategy assuming that others will respond

they can

other OECD

that one

U.S.

but

policy

not

vice

significantly
versa,

the

affects

Stackelberg

the United States behaving as a leader, might

more relevant

assumption than the Nash. Non-cooperative

games yield

outcomes in

better off,

with no

which one country can usually be made

other being

worse off,

by a cooperative

rearrangement of strategies. The gain attributed to cooperative

policies can

thus be

calculated as

non-cooperative solution
(5). For

any gain
(i)

must hold

on ultimate
domestic
countries

the difference

between a

optimal cooperative

solution

from cooperation to exist, three conditions

policies in

negligible spillover
countries' policy

and an

any given country must have non-

effects on other countries

instruments must

(ii) foreign

have an independent effect

objectives distinct from that obtainable by mixing

instruments
must

not

appropriately
have

enough

and

(iii)

instruments

individual
to

achieve

independently all objectives.

Niehans

(1968)

anticipated

much

of

the

recent

(5) These refer to the large-scale international macroeconomic

models developed,
respectively,
by the Japenese Economic
Planning Agency and the U.S. Federal Reserve Board.

157

research in
reserve

a theoretical

currency

cooperation. tax
too

tight.

system.

He

policy might

since

maintaining

study of

the

external

employment [cf.

fixed

exchange

concluded

that.

rate.
without

be too easy and monetary policy

latter

balance

is

mainly

and

the

directed

former

Mundell (1962)]. Hamada (1974.

towards

towards

full

1976). focusing

just on monetary policy. showed that non-cooperative strategies

under

fixed

deflation

exchange

or

and

the

individual objectives

the

exact

bias

of individual
growth

of

in his
exceeds

be overly

an example

be

the sum

(exogenously determined
policy will

might

inflation.

relation between
objectives

rates

biased

towards

depending

balance

on

of

If

supply

the
of

the

payments

international

model).
the

too

reserves

sum

of

reserves

the
then

contractionary. and vice versa. This is

of conflicting

country objectives.

Canzoneri

and

Gray (1983) show that undesired non-cooperative biases can also

emanate from

the structure

of spillovers. They consider three

configurations of spillovers: symmetric-negative ("beggar-thyneighbour"

situations).

situations) and
policies are

asymmetric.

the

be a

In the

are deflation-biased

symmetric-negative

countries attempt

case.

to offset

round of

competitive exchange

symmetric-positive case.

since no

country gives

policies

enough weight to

impact abroad of expansionary measures at home.

the beneficial
Biases from

("locomotive"

spillovers by expansionary measures at home.

example would

rate depreciations.

of the

In

inflation-biased as

mutually negative
A concrete

symmetric-positive

asymmetric games will depend on the precise nature

asymmetries. but

mixes across

they can

lead to

conflicting policy

countries. For example. a short-run payoff to the

United States

can be

derived in certain circumstances from an

expansionary fiscal/tight money mix [see. e.g .. Sachs (1985)].

In an
in the

of nominal wage rigidity

United States and real wage rigidity in Europe analysed

by Asikoglu

(1986). the United States essentially has only one

instrument to
instrument that
another to
does not

asymmetric situation

influence aggregate demand. while Europe has one

can affect

affect the

real output

price level

(fiscal

policy)

(monetary pol icy).

and

Europe

need to cooperate when the two instruments are up for

negotiation. while the United States does not want to cooperate

158

if only monetary policy is on the bargaining table. This nicely

illustrates the

point that cooperation can be expected only if

potential gains are available to all participants.

While theory
cooperation.

it

tells us

where to

look for gains from

does not say how large the gains might be. The

next section discusses some empirical findings on this score.

2. Empirical results
The

landmark

coordination is
solution with
and

study

Oudiz and

of

potential

gains

from

Sachs (1984). which compares a Nash

an optimal

cooperative solution.

using the EPA

MeM models (6). One novelty of this study is that it infers

the characteristics

of governments'

objective functions

from

the multipliers of the models and from the assumption that each
country (the United States. Germany and Japan) does the best it
can without

cooperating. Then.

from synthetic

values for the

policy instrument settings and for ultimate objectives over the

period 1984

to 1986.

the preferences of the three governments

are estimated. Preferences are "revealed" to be highly weighted

against inflation
favour of
for the

in the

United States

current account
three countries

and Germany.

surpluses in
and the

and

in

Japan. The output gap

trade balance for the United

States are revealed . on the other hand. to have lower weights.

These inferences.
path for

it must be emphasized. depend on the baseline

the 1984-86

period as

well as

the structure of the

models.
The gains
Sachs stem

from coordination

mainly from

rates. With

suggest more

a coordinated

by

reduction

Oudiz
in

and

interest

both models cooperation implies increased monetary

expansion everywhere.
more fiscal

derived

but while the EPA results also recommend

contraction in

all three

fiscal expansion

in the

countries. those of
United States

MeM

and more

fiscal restraint in Germany and Japan. This odd result from the

MeM --

that countries

already doing

on fiscal

should have done more of what they were

policy --

is a

consequence

of

the

(6) It complicates the argument. but does not change its

essence. to recognize that some inflation might be optimal in a
world with distorsions caused by non-neutral taxes. monopoly.
etc. [Barro and Gordon (1983) 1.

159

revealed

preferences

approach

and

of

the

fact

that

no

consideration was given to the longer-run sustainability of the

policies.
to U.S.

If

instead higher weights are attached to output and

fiscal and/or

increased U.S.
(1985).

trade deficits.

the recommendation for

fiscal expansion does not survive [Ishii et al.

Sachs

and

McKibbin

expansionary monetary

(1985)1.

policies does.

but

that

for

more

In all cases considered.

the derived benefits. relative to the non-cooperative solution.

of increased
some

output and

increase

in

employment more

inflation.

given

than compensate

policymakers'

for

apparent

preferences.
A common
assessed from

finding is

the

that

objective

countries (invariably

a group
1 per

Taylor (1985)
neo-classical

to

welfare
the

gains.

cooperating

of OECD countries) are small -cent or

so of

GNP. Carlozzi and

contend that the gains from coordinated policies

are empirically
that tend

net

functions.

usually no

more than

the

negligible. Since
assumptions

their analysis

(including

rational

is based on
expectations)

to reduce policy effectiveness in the short run. and

eliminate it

entirely in

the long run. the question is raised

as to whether their conclusion merely reflects an underestimate

of the

effects of

answer seems
in a

to be

wide range

Canzeroni and
has some

policy -of models.

For example.

Minford (1986)

cases coordination

of

spillovers.

The

that the small estimated gains are derived

very large

(1986) allows

especially

from the

results derived

Liverpool model. which

monetary spillovers.
yields only

for uncertainty

by

agree that

in many

second-order benefits. Frankel

as to

model specification. and

further subverts the positive findings. Frankel finds that even

where goals

are the same. use of different models by different

parties. neither
real world.

of which

is an

exact representation

of the

would be likely to cause welfare to be lower under

coordination.
Two factors.
not be

overlooked.

timing of

more favourable

coordination can

Eir~.

policies [e.g.

to cooperation. should

Sachs (1983)1.

result in

better

If the starting point

is one of high inflation. the optimal selfish policy would be a

sharp tightening
exchange rate

of monetary

to appreciate

policy. which
and

dampens

causes
domestic

the

real

inflation

160

rapidly.

Competitive

selfish

international deflation.
participant attempts
advantage, each
loss is

policies

then

With cooperative

to exploit

imply

sharp

policies, since

the exchange

no

rate to its own

disinflates more slowly. The cumulative output

the same, for a given total degree of disinflation, as

with competitive
so, with

policies, but

normal rates

outcome yields

of

developed countries
higher demand
might have

time

higher welfare.

industrialized countries

is spread

preference,
~e~Qnd,

yields

and improved

participating countries

lower

LDC terms

gain from

the

cooperative

cooperation between the

side-benefits

(LDCs), favouring

more to

into the future and

to

the

interest

of trade.

lessrates,

Indeed

LDCs

increased coordination than the

themselves [e.g.

Sachs

and

McKibbin

(1985)].
3. Other approaches
i) Exchange rate and world money growth rules
McKinnon (1984) proposes a monetary agreement between
the United

States, Germany

and Japan

which would incorporate

exchange rate target zones and a constant rate of growth of the

combined money
in the

supply. The

global policy

Although currency

idea is to avoid unintended biases

stance cause

by currency

substitution is

not

important

[see e.g. Dornbusch (1983), Boothe et al.

Sachs (1986)
some merit
This

is

show that
in the

his

empirically

(1985)], McKibbin and

McKinnon's proposal

context of

because

substitution.

nevertheless

has

a worldwide inflationary shock.

scheme,

like

the

full

cooperative

arrangements, bans attempts at competitive appreciations, which

might otherwise cause the world interest rate to rise too much.
More simple
degree of

proposals have

exchange rate

encouraging

code

substitute for
suggested the

from underlying

of

behaviour

that

the
could

to limit

the

intention
serve

as

of
a

explicit policy coordination. Some authors have

formation of

[e.g. Williamson
exchange rates

been made

flexibility, with

target zones

(1985)]. Critics
do little

good if

macro policies

for major

countries

argue that target zones for

they direct

[e.g. Dornbusch

attention away
(1983)].

The

161

majority view

of the

system (1985)

was that such a proposal did not offer prospects

of improvement
are only

G-10 study of the international monetary

of the

present situation.

Exchange rate rules

a good substitute for explicitly coordinated policies

in special

cases. Although

manipulation

of

distortions in

exchange

they might
rates,

policy mixes,

help avoid competitive

they

also

migh~

lead

to

and to systemic biases in policy

stance [Hamada (1974), Johansen (1982)].

The European

Monetary System (EMS) has been assessed

from the viewpoint of the theory of policy coordination in some

recent articles [e.g. Melitz (1985), Oudiz (1985), Giavazzi and
Giovannini

(1986)].

The system

attempts to

encourage more or

less symmetric policy adjustments among it.s members and accepts

exchange rate
e.g., the

realignments at

more frequent

intervals

than,

Bretton Woods system in practice did. Studies of the

operation of

the system

implications

for

symmetry of

EMS

have explicitly
countries

adjustment between

concentrated on

themselves

members

--e.g.

on

the
the

rather than on the

broader internatIonal

context. No clear consensus emerges from

these studies

the rules

on how

of the

system might

be best

modelled, or on its benefits to members.

ii) Judgemental approaches
A large

number of

authors have

described what they

see as desirable policies, arguing from their own judgement and

a varying

amount of

Buiter (1985)
the world

explicit theory

and empirical

evidence.

derives optimal policy responses for the rest of

in response to a

U.S. budget cutback (of the Gramm-

Rudman-Hollings variety). One such response is fiscal expansion

outside the

United States

remains unchanged.
would then

Within the

be offset

(improved U.S.

such that

by a

the world

interest rate

United States, fiscal restraint

real depreciation

competitiveness). Another

of

the

response would

dollar
be a

one-shot increase in the world money supply, which would reduce

real interest rates and cause a temporary increase in inflation
rates. A

permanent increase

also offset
cost of

in the growth rate of money would

output effects of the fiscal restraint, but at the

permanently higher

inflation.

Some

questions

about

162

confidence obviously
similar package

arise from

of measures.

United States.

these

proposals.

However

including fiscal restraint in the

fiscal expansion

in the rest of the world. and

some monetary expansion. has been widely advocated [e.g. Marris

(1985)]. As
and in

before. the

conclusion is that the gain in output

reduced current

account

imbalances

seems

worthwhile

relative to the increase in inflation.

B. Time consistent policy and central bank reputation
1. Conceptual issues
In models
monetary

with the

neutrality.

necessarily produces

classical property
inflationary

an

of

monetary

additional

inflation

accompanied by any gain in output. Therefore the best

the long

run is

policy

worse results in the long run than a non(7). since

inflationary policy

firmly believes

long-run

one that
that the

allows no

inflation.

monetary authority

is

not

polic~

in

If the public

is committed

to

such a policy, and if in addition the policymakers hold to that

commitment. a

favourable outcome

is likely.

i.e. no inflation

and continuous full employment of ressources (i.e. unemployment

at the

"natural" rate).

precommitment, the

But

given

policymakers can

public
achieve

belief
an

in

even

their
better

outcome by reneging temporarily. This will be the case if there

is a short-run tradeoff between unemployment and inflation, and
if both
higher

society and
output,

central bank

even

beyond

can then

increasing the
if the

the central
the

bank have

a preference

full-employment

improve welfare

in

the

rate.

short-run

for
The
by

money stock. However this option is viable only

public is convinced that in future periods the monetary

authorities will revert permanently to the no-inflation policy.

Otherwise inflation
and prices.

and the

premiums will be built into wage contracts

price level will rise at once without any

increase in output.
A "time-consistent"

policy can

be

defined

as

one

which takes fully into account the discretionary actions by the

authorities in

the future

[Kydland and

(7) Barro (1986) uses a similar concept.

Prescott (1977)].

In

163

this sense,
monetary

it can

policy

be said
is

can in

is credibly

general achieve

commitment of
are clear

to inflate,

outcome by
For this

reneging on

the

reason, unless there

guarantee such

this commitment,

commitment

the

If the monetary authorities

even though they intend to do so

the public, which is aware of the temptation

will revise

inflation. When

involves

established, the central bank

public might be sceptical about it.

only temporarily,

it

implement them. Once a non-

a better

price stability.

constraints that

divert from

since

to actions in the future that might not

when the time comes to

inflationary policy

constant non-inflationary

"time-inconsistent"

committing authorities
be optimal

that a

its attitude and start to expect some

no constraints

are placed on the authorities,

the situation is likely to slip over time into one in which the
inflation rate

is just high enough that policymakers will find

any further increase unacceptable.

rationally

expects

just

the

In this situation the public

rate

of

inflation

that

is

delivered. Therefore, with no binding commitment on the part of

the authorities,

a time-consistent

policy in

general

allows

some inflation.
These
treatment in

ideas
recent

have

been

literature

given
on

the

rigorous

theory

policy, which can be illustrated as follows:

of

formal
economic

164

EXPECTATIONS OF THE PUBLIC

Some inflation
No inflation
{policymakers not
{policymakers
credible}
credible}

ACTUAL POLICY

No inflation
{Precomm'i tment}

***

Some inflation
{Precommitment
ignored}

**

****
Time inconsistent
policy solution

There are

Time consistent
solution

two possible expectations and two possible

outcomes illustrated
(i.e. one

Time inconsistent
expectations
solution

Precommitment
solution

here, yielding

outcome for

system). Policymakers

four

possible

solutions

each of the four possible states of the

may or

may not

stick to their declared

target of no inflation
the public mayor may not believe
them.
The asterisks (stars)
indicate the ranking of the
solutions in terms of social welfare ; the ranking rises as the
stars increase. The only two possibilities that are sustainable
in the long run are the precommitment solution and the timeconsistent solution,

because only

in those

solutions are

eL

plans realized. However in any decision period the highest

ranking (four stars)
is awarded to the time-inconsistent

policy, if

it is

feasible. The precommitment outcome, best in

the long run, gets only three stars. At the other end of the
spectrum, the
{one star}
worst solution
is when
the
policymakers are committed but lack credibility -- an output
loss is then caused by the central bank's refusal to accomodate
the higher

wages and

consistent policy,
by the
stars.

prices built

which just

public, avoids

into

contracts.

time-

ratifies the inflation expected

this output

loss and

so receives

two

The illustration highlights two implications. First,

whatever the true intentions of the authorities,
the best
outcomes can

be acheived

only when

the public

believes that

they will allow no inflation. Therefore the authorities can

always be
expected to
announce that
they will follow
disinflationary policies regardless of whether they have the
will or

the means

to do so. Second, the authorities avoid the

165

worst, and

might achieve

in the short run.

the best, by allowing some inflation

In theory it is easy to assign weights to the

preferences under

each solution

that will

central bank will opt for some inflation.

loss of

output of

guarantee that the

If society values the

the precommitment/low credibility situation

highly enough, then the monetary authorities may be forced into

an accommodative inflationary stance against their will.
These are
they demonstrate
bank to

conclusions with

establish

policy without
manoeuvre.
the true

the

credibility

some external

for

of

an

constraint

anti-inflationary

on

its

freedom

of

In a situation where the central bank has discretion

nature of

observing its
countries the
resources to

its intentions

actions over
private sector
"central bank

recognize that

an announced

In

their

model

announced targets

of time

inferred
and

in

by
some

does in fact devote considerable

the speed
change
of

policy
is

monetary

increases. Backus

and

Meltzer

with which the public

in

reputation

the precision

be

watching". Cukierman

credibility as

increases as

can only

a period

(1986) define
occurred.

wide ramifications,

that it might be very difficult for a central

has

actually

parameter

control

and Driffill

in

which
hitting

(1985) pose

the dilemma more sharply, by defining two types of policymakers

--"strong"

(inflation

resisting) and "weak" (inflation prone).

Moreover the

authors define

lost, cannot

be regained.

credibility as
As long

as the

a state that once

authorities do not

inflate, the public has some confidence that the policymaker is

strong. However,
cheat on

since the public is aware of the incentive to

the announced

policy, once

the policymaker

reveals

himself to be weak he is forever perceived so.

Considerations of

this kind

have radically

the arguments

for monetary

rules.

The

recently been

argued, must

visibly and

shifted

required rule,

it has

permanently bind

the

monetary authority to a no-inflation objective, which is a much

stronger thing
then centres
for the

on the

central bank,

the stability
(1983)

than a

constant-money-growth rule.

proper legal and constitutional framework

rather than on empirical arguments about

of demand

describe

as

permanently constrain

The debate

for money.

discretionary

Thus,
a

Barro

system

the central bank to

that

and

Gordon

does

price stability.

not
In

166

their model,

which has

discretionary system

strong classical

properties.

such

inevitably produces inferior results to a

system bound by law to no-inflation rule. Since it is difficult

to imagine what concrete form the binding laws could take, this
argument is not very relevant to the real world.
Instead, one
the practical
If

the

central

credibility

is led to focus on

and

bank

establishes

price

stability

attractive short-run

opportunities

generally recognised

that this

this case

to

is a

constraints

external

high

reputation

foregoing
inflate,

for

apparently
it

very valuable

will

be

asset.

In

can be achieved through an

-- the concern not to worsen the tradeoff

endogenous incentive
unnecessary.

a
by

pre-commitment solutions

and

which is

repu~ation.

alternative to a rigid set of external controls.

and

Barro

policymakers' concern

the

on

central

(1983)

Gordon

for reputation

bank

show

become

that

the

is related to the length

of their time horizons. A central bank that takes a long enough

view (i.e.

has a

low enough

rate of

time discount)

will be

deterred from short-run inflationary policies by the inevitable

loss of

reputation.

It

weights highly

the prospect

that the

public can "punish" the policymakers by revising its opinion of

their credibility,

and hence

permanently worsening the policy

tradeoff.
2.

Implications for policy coordination

Dynamic models
issue implicitly
raise no

that do

assume that

credibility problem.

not explicitly

the

address

time-inconsistent

the

policies

If precommitment by policymakers

is both possible and credible. then indeed it can be presumed a

priori that
van der

cooperation is preferable to non-cooperation [e.g.

Ploeg (1986)].

cooperation is
are admissible.
more

inflation

considers that

existence of

.if only

since cooperative
than

non-cooperative

This is because,

inflate the

a net gain from

time-consistent solutions
solutions of that kind have

cooperation between

worsen welfare.
incentive to

But the

not assured

ones.

Rogoff

(1985)

central banks could easily

in a cooperative setting. the

money stock is increased by reducing

the fear of exchange rate overshooting in individual countries.

In Rogoff's

model this

raises the

rate of

inflation without

167

yielding a gain in output. Oudiz and Sachs (1985) however argue

that cooperation
higher rate

might improve

of inflation

social

if it

w&lfare,

increases

the

despite

the

stability

of

exchange market speculation.

Finally, policy coordination does offer the chance of
an optimistic
dilemma.

resolution of the credibility-time inconsistency

If the perceived gains from cooperation are very high,

agreement between
that time

inconsistent policies are credible, because the cost

of reneging
very big

countries might persuade the private sectors

would be prohibitive. However, the proviso here is

in the absence of any evidence showing large gains of

this kind.

REFERENCES
Asikoglu, Y.

(1986), "Macroeconomic Interdependence and Policy

Coordination Between
Theoretical Approach",

the U.S.

and Europe

Discussion paper

Game-

n0631, Queen's

University, March.
Backus, D.

and J. Driffill (1985), "Inflation and Reputation",

American Economic Review, June pp. 530-38.

Barro, R.J.
with

(1986), "Reputation

in a Model of Monetary Policy

Incomplete

Information",
Journal
Economics, vol. 17, n 0 1, pp. 3-20.

Barro, R.J.and

D.D. Gordon

Monetary Policy

in a

(1983),

"A

Political

Natural Rate

of

Monetary

Theory

of

Model", Journal

of

Political Economy, August, pp. 589-610.

Dooth&, P.,

K. Klinton,

International

A.

Asset

Cote

and

D.

Longworth

Substitutability

Evidence for Canada, Bank of Canada, Ottawa.

(1985),

Theory

and

168

Buiter,

Y.H.

(1985),

"Macroeconomic

Interdependent World

Policy

Economy:

An

Design

Analysis

in

of

an

Three

Contingencies", NBER Working Paper n01746, November.

Canzoneri, M.B.,
Policy

and J.A. Gray (1983),

in

an

"Two Essays on Monetary

Interdependent

Finance Discussion

World",

International

Papers, n0219, Board of Governors of

the Federal Reserve System, February.

Canzoneri, M.B.

and P.

Minford

Policy Coordination
Discussion Paper

(1986),

Matters:

Series,

"When

International

An Empirical

n0119,

Centre

Analysis",

for

Economic

Policy research, London.

Carlozzi, N.

and J.B.

Mobility and
J.S. Bhandari

Taylor (1985),

the Coordination
(ed.),

Exchange

"International
of Monetary
Rate

Capital

rules",

Management

in

under

uncertainty, M.I.T. Press.

Cukierman, A.

and A.H. Meltzer (1986),

Credibility

and

Inflation

"A Theory of Ambiguity,

Discretion

Under

and

Asymmetric Information", Econometrica, vol. 54, n05,

pp.

1099-1128.
Dornbush,

R.

(1983),

Interdependence",

"Flexible

Exchange

rates

and

IMF Staff Papers, vol. 30, Nl, pp. 3-

30.
Frankel,

J.A.

(1986),

"The

International Macro

Sources

Models and

of

Disagreement

among

Implications for Policy

Coordination", mimeo, University of California, March.

Giavazzi,

F.

and

A.

Giovanni

(1986),

"European

Currency

Experience", Economic Policy, January.

Group of

Ten (1985),

"The Functioning

Monetary System",

IMF

of

the

International

169

(1974),

Hamada, K.

"Alternative

Interdependence
Monetary
System,

of

Exchange Rate Systems and the

Monetary

Policies

and

Policies,

the

in

National

International

Financial

(R.Z. Aliber ed.), University of Chicago Press:

Chicago, pp.13-33.
Hamada,

K.

(1976),

"A

Strategic

Analysis

of

Monetary

Interdependence", Journal of Political Economy, vol. 84,

n04, pt.
Ishii, N.,

1, August. pp. 677-700.

W. McKibbin

Policy Mix,
Aspects

and J.

Sachs

(1985),

"The

Economic

Policy Cooperation and Protectionism; Some

of

Macroeconomic

United States,

Interdependence

Among

the

Japan and Other OECD Countries", Journal

of Policy Modeling, vol. 7, N4, pp. 533-72.

Johansen,

L.

(1982),

"A

Note

International Equilibrium
Journal

of

Kydland, F.W.

the

Possibility

of

an

with Low Levels of Activity",

International

February, pp.

on

Economics,

vol.

98,

nOl,

Rather

than

1-23.

and E.C.

Discretion

Prescott (1977),
The

Inconsistency

"Rules
of

Optimal

Plans",

Journal of Political Economy 85, n03, June, pp. 473-491.

Marris, S.

(1985),

a Risk",
n014,

"Deficits and the Dollar: The World Economy

Policy Analyses

in International

Economics,

Institute for International Economics, Washington,

December.
Melitz, J.

(1985),

"The

Welfare Case for the EMS", Journal of

International Money and Finance, December.

McKibbin, W.J.

and J.

and Fiscal

Sachs (1986),

Policies in

"Coordination of Monetary

the OECD" ,

NBER Working Paper,

n01800, January.
McKinnon, R.I.

(1984),

Stabilization",

"An International Standard for Monetary

Policy

Analyses

in

Internattonal

170

Institute for

Economics, n08,

International Economics,

Washington, March.
(1963), "Capital Mobility and Stabilization Policy

Mundell. R.
Under

Fixed

Journal of

and

Flexible

Economics and

Exchange
Political

Rates",

Canadian

Science

29,

n04,

in

Open

November, pp. 475-85.

(1968),

Niehans. J.

"Monetary

Economies Under

and

Fiscal

Fixed Exchange

Approach", Journal

of Political

Policies

Rates:

An

Optimizing
n04, Part

Economy 76,

II, July/Agust. 893-920.

Oudiz.

( 1985) ,

G.

"Strategies

Economiques

Confrontation 7,

Coordination ou

Europ~ennes

INSEE Working

Paper.

n08506, June.
Oudiz.

G.

and

J.

Sachs

Coordination among

Papers on Economi cActi v i ty,

Oudiz,

G.

J.

and

"Macroeconomic

(1984).

Policy

the Industriial Economies", Brooking

Sachs

1, pp.

( 1985) ,

1-75.

"International

Policy

Coordination in Dynamic Macroeconomic Models", Chapter 7

in

International

Buiter and

R.C.

Economic
Marston

Policy
eds.),

Coordination
Cambridge

(W.H.

University

Press.

K.

Rogoff.

( 1985) .

Cooperation

be

"Can

Monetary

International

Counterproductive

International Economics.

7"

vol.18. n03/4,

Policy

Journal
May. pp.

of
199-

217.
Sachs, J.

(1983),

Dynamic

"International

Macroeconomic

Policy

Model",

Coordination

NBER

Working

in

Paper.

n01166, July.
Sachs, J.

(1985),

"The

Dollar and

the Policy

Mix

Brooking Papers on Economic Activity 1, pp.

1985" ,

117-97.

171

Sachs. J. and C. Vyplosz (1984), "Real Exchange Rate Effects of

Fiscal Policy", NBER Working Paper, nOl16, January.
Sachs. J.

and V.

the

OECD

HcK i bb i n (1985),
and

LDC

"Macroeconomic Policies in

External

Adjustment",

Centre

for

Economic Policy Research Discussion Paper 56, March.

Van der

Ploeg. F.
Long-run

(1986),

"Capital Accumulation,

Conflict

Discussion Paper

in

International

Inflation and
Objecti ves" ,

Series lIS, Centre for Economic Policy

Research, London.
Villiamson. J.

(1985), The Exchange Rate System,

International

Economics

International Economics
1983) .

Policy

Institute for
Analyses

in

nOS (original version published

A FOLK THEOREM OF MONETARY POLICY

Carlo CARRARO
University of Venice and CEPR

1.

Introduction

The

analysis

economic policy

of

by using

effectiveness
game theory

contemporary macroeconomics.
and the

and

is a

optimality

of

crucial issue

of

International policy coordination

role of monetary policy are the topics more frequently

analysed.
Most papers
the following

have been concerned with the solution of

dilemma, proposed

by Kydland-Prescott

(1977)

the optimal policy is in general time-inconsistent, whereasthe

time-consistent policy
can be

re-phrased in

is in general sub-optimal. This dilemma

the following

way:

the strategy which

maximizes the policymaker's (expected) utility is in general

domi nated by a different (hence sub-opt imal) strategy. (1)
This dilemma was solved in several ways: KydlandPrescott (1977)
propose to force the policymaker to be
committed to
mechanism

the announced
Barro-Gordon

trigger strategy

policy rule by an institutional

(1983),
instead, use Friedman's

equilibrium concept

(see Friedman, 1971), to

show that a reputational mechanism can force the policymaker to

be committed
infinite. The

to his announcement,
same result

is

if the deterministic game is

shown

for

incomplete information

by Backus-Driffill

use of the sequential

Wi 1 son (1982).

equilibrium concept

This paper

aims at

providing more

above dilemma by showing that

(i) there
exists a

strategy,

finite

games

with

(1985a,b), who make

proposed by Krepsinsight
named

on

the

Stackelberg

(1) Kydland-Prescott's (1977) dilemma is proposed within a

deterministic (perfect foresight) framework.
If we
introduce
uncertainty, the dilemma coincides with the Newcombe's problem
discussed by Gardner (1973).

174

trigger strategy, which maximizes the policymaker's utility and

which cannot

be dominated

by any

other strategy; conditions

for this strategy to be effective and credible are provided ;

(ii) Folk
strategies, the

theorem

if

policymaker can

which Pareto-dominate

we

allow

achieve any

the time-consistent

for

trigger

of the

outcomes

Nash outcome

(more

generally, any individually rational outcome can be achieved)

(iii) Folk
strategies into

theorem

if

a hierarchical,

we

introduce

trigger

sequential, framework,

where

the policymaker is the dominant player of the game, then he can

achieve any

of the

outcomes he prefers to the time-consistent

Nash outcome
(iv) either

a coordination

problem characterizes

a general

problem or a credibility

policy problem and leaves its

outcome indeterminate.
In order to simplify the analysis, we will consider a
specific

economic

example,

the

Monetary

Policy

Game

(see

Cukierman, 1986), even if we want to emphasize that our results

have general validity (see Carraro, 1987).
The Monetary

Policy Game

and conflict

between the

sector. The

crucial point

monetary policymaker

analyses the

interactions

monetary policymaker and the private

is the following:

has a

zero inflation

suppose that the

target and that he

values output growth over its natural rate; the "expectational

Phillips curve"
output target
fooling the

implies that
only

by

the policymaker

inducing

private sector.

can achieve his

inflation

shocks,

i.e.

However, being

private

sector's

expectations

rational,

behaviour is

perfectly anticipated and the outcome of the game

is sub-optimal,

the

policymaker's

by

time-in'consistent

i.e. output growth is equal to its natural rate

and inflation is positive. Sub-optimality derives from the fact

that

the

monetary

inflation target
monetarist rule
to output

policymaker

by committing
that money

natural rate

can

achieve

himself to

at

the

least

his

old-fashioned

stock rate of growth must be equal

of growth. This is therefore a typical

example of the policy dilemma previously presented.

In this paper we want to show that Kydland-Prescott's
dilemma constitutes a partial view of the policy problem, since
there may

exists a

time-consistent monetary

policy such that

175

the policymaker

can achieve

both

his

inflation

and

output

target, under suitable assumptions on the information available

to the players of the Monetary Policy Game.
Furthermore, we

show that

many more outcomes can be

achieved by the monetary policymaker, thus proving an analogous

of the

Folk theorem in the industrial organization literature.

The novel

element of the paper, with respect to the industrial

organization literature,

is that

hierarchical, sequential,

a Stackelberg

framework

(a

game) will be considered. The nature

of the game between the policy authority and the private sector
is

indeed

hierarchical

application of

and

the results

organization literature,

makes

it

proposed in

being

those

unsatisfactory

the

the recent industrial

results

conceived

for

symmetric (Nash) players.

In the
analysed in

next section,

we

consider

the

same

model

Barro-Gordon (1983a,b), Backus-Driffill (1985a,b),

Barro (1986),

Cukierman (1986), and we characterise the normal

form of the repeated Monetary Policy Game.

Then,

in

section 3,

perfect equilibria
the monetary

we show

subgame

exist, when the private sector can threaten

policymaker. This

infinite policy

that infinite

extends

games. Furthermore,

previous

results

on

a coordination problem of

the type emphasized in Schotter (1980)

is shown to arise.

In section 4, we consider the case where the monetary

policymaker can

also threaten

the private

represented

player),

as

monolithic

sector

and

we

(initially
analyse

the

sequential equilibria that can be achieved under the incomplete

information
Strategy
1986a). (2)

assumption.

equilibria
It

will be

In

are

particular,
considered

shown that

Stackelberg
(see

Carraro,

Trigger
1985,

a larger set of equilibrium

(2) Stackelberg Trigger Strategy equilibria are sequential

equilibria such that the policymaker's desired outcome is the
actual outcome at all
stages of the game. The equilibrium is
Stackelberg because the players of the game are not treated
symmetrically and only the dominant player has the power to
raise threats.
The equilibrium
is Trigger
because the
policymaker threatens to punish the other players of the game
whenever his desired outcome
is not achieved.
A similar
equilibrium concept for Nash games has been proposed by
Friedman (1971, 1985) and applied, for example, by Green-Porter
(1984), Rothemberg-Saloner
(1986).
The relationship between
Friedman's Trigger Strategy equilibrium and the Stackelberg

176

outcomes can
bliss

be achieved

point

whenever

coordination

his

problem

credibility problem.
exists a

and that the policymaker attains his

In particular,

is greater

is

believed.

The

into

transformed

therefore

is

monetary policy

and output

announcement

it will be shown that there

strategy such that inflation is zero

than its

natural rate at all stages of

the game.
Section 5
Policy Game

discusses the

when the

monolithic player,

solution

private sector

but is

of

is not

the

Monetary

represented as a

characterized as an infinite number

of atomistic players.
Finally, section
previous results
of

the implications of the

and their robustness with respect to possible

respecifications of
functions,

6 analyses

the monetary

model, of

the players'

loss

the players' strategic behaviour and information

set.
2. A Model of Monetary Policy
A standard,
discuss the

role of

models (see,
Gordon,

ingredients of

the following

real

negatively
targets

may

i.e.
the

economic

inflation. Fourth,
over optimal

increases in

Third.

of

the
not

Barro-

a discussion)

1986).

the inflation
there

is

policymaker

activity

by

first.

general price
an

rate

on

policymaker.
coincide

with

cash

aims

at

inducing

balances

Finally.
the

the

public's

level.
expand

"expectational
achieving
unexpected

expectations are rational. Fifth.

inflation tax
by

1977)

1985a,b ; Cukierman,

affecting the

activity.

Phillips Curve".
expansions

to

in rational expectations

1986. for

works by

economic

used

the model (the constituent game) are

(see Barro.

Secondly. unexpected

been

Kydland-Prescott,

; Backus-Driffill.

monetary policy

has often

monetary policy

for example,

1983a,b

The basic

simple, model

is

inflation
considered

policymaker's
targets.

For

example. two different rates of growth of output (or inflation)

may be

the objectives

of

the

policymaker

and

the

public.

Trigger Strategy equilibrium is discussed in Carraro (1985). A

brief presentation
of the
Stackelberg Trigger
Strategy
equilibrium concept is provided in the Appendix.

177

Therefore, the goal of our analysis of the standard Monetary

Policy Game is to study equilibria resulting from the conflict
between agents who aim at achieving incompatible targets.
The first equation of the standard model is a rate of
growth version of the "expectational Phillips Curve"
(2.1) Yt
where Yt and y~t are the rates of growth of,
respectively, real
output and
potential
output
(which
corresponds to the natural rate of unemployme~t) ; Pt and pet
are, respectively,
inflation and expected inflation. Several
plausible reasons to justify the introduction of eq. (2.1) are
provided in Lucas (1973), Barro-Gordon (1983a). Output rates of
growth are introduced for simplicity's sake. Eq. (2.1) can be
interpreted as private agents' reaction function, who decide
output as a function of actual and anticipated prices (both as
a function of the monetary strategy). Notice that introducing
(2.1) implies that the outcome of the game must belong to
private agents' best reply function. This assumption is fairly
unnatural in game theory and will be relaxed later on.
The second equation, derived directly from money
quantity theory, describes aggregate demand (see Barro-Gordon,
1983a) and defines the role of monetary policy
Yt - Pt
where m+t is the money stock rate of growth.
The policymaker's loss function at time t, VMt .
defines two policy objectives: output rate of growth should be
greater than its potential level and the price level should be
stable. Following Barro-Gordon (1983b), we assume that the
policymaker values inflation shocks, that is Pt > pet, only
over some range and that the optimal inflation tax on cash
balances is normalized to zero.
Therefore, we have:
(2.3) VMt

178

The first
inflation shocks,
inflation. The
rate),

term of

the equation

whereas the

parameters

second
(which

is the benefit from

term

is

depends

the

on

cost

the

of

natural

and b o are assumed to be positive.

Using (2.1),

eq.

(2.3)

can

be

re-written

in

the

output

and

following way:

Defining X t
potential output
e~/o

by

and

as the

rates of

difference

between

growth (i.e.

then re-normalizing

Xt

Yt - Y*t), X

the loss function by multiplying

we obtain:

(~/O)2,

(2.4)

The policymaker's

control variables

bo(~/o)2

are

the

money

stock mt and the relative monetary announcement mat.

Given the
( 2. 1)

above definitions,

we can

re-write

eqs.

( 2 . 2) as

(2.5.1)

and
(2.5.2)

mt

Finally, we assume that expectations are rational,

that eqs.

(2.5)

imply:
met

(2.6)

where met
stock rate
agents'

so

is the

expected deviation

of growth

from

its

natural

expectations

may

not

coincide

announcement mat.
Solving equations (2.5), we obtain

level
with

of

the
Y*t.
the

money
Private
monetary

179

(2.7)

1 +

(2.8)

pet. +

pt.

1 +

From eqs.
of output

(2.6) (2.7), i t is evident that deviations

from its

natural

rate

can

be

obtained

only

by

unexpected increases of the money stock rate of growth over its

natural rate.
Xt.

Therefore, the

x only by cheating,

which,

if

is the

i.e. by announcing a policy decision

believed, will differ from the actual decision. This

basis of Barro-Gordon's (1983a) results which show that

the monetary

authority has always the incentive to depart from

the announced
that,

policy in order to achieve a lower loss. Knowing

the private

monetary policy

sector anticipates

and the

(i.e. Xt.

only if

the policy

outcome of

the

time-inconsistent

the game

is

sub-optimal

0 and Pt. is positive). As shown in the next section,

information is

time horizon

assumed, can

monetary policymaker
that Xt.

monetary policymaker can achieve

is still

to be

is

infinite

a reputation
committed to

or

incomplete

mechanism force the

his announcement, so

zero but no inflation is induced by monetary

policy.
The above

equations

define

the

standard

monetary

game, but do not define yet all the elements of the constituent
game (furthermore,

eq.

(2.1),

being a

reaction function,

is

redundant).

In particular, the private sector loss function and

the "rules

of the

will completely

game" must

define the

be specified.

The next section

constituent game

and explores its

equilibrium outcomes.
3. Friedman's Trigger Strategies and the Coordination
Problem
This section
Gordon's (1983a,b)
depart from

analyses

the

their analysis

in that

private sector's loss function and

of the

implications

of

Barro-

and Backus-Driffill's (1985a,b) results. We

we explicitly

expli~itly

write

the

define the "rules

game". We use a loss function consistent with (2.1), so

180

that

and

Barro-Gordon's

derived by

Backus-Driffill's

assuming that

results

can

be

all outcomes of the game must belong

to private agents' best reply function. This assumption will be

relaxed in

the

next

section.

Furthermore,

we

assume,

for

simplicity, the private sector as a monolithic player that aims

at achieving

a stable output growth (x t = 0).

In contrast, the

monetary authority tries to achieve Xt = x* only by fooling the

private sector.

We emphasize

that the

results below

can

be

shown even without the monolithic player assumption (see Barro,

1986) .

The private

sector's

loss function

can be written

as
Vpt

(3. 1 )

(Yt - Y*t)2 +

so that

- pe t )2

~(Pt

X2t +

- pe t )2

~(Pt

private sector's

rational reaction function

is given by (2.1).
Private sector's decision variables are output Xt and
price expectations pet.
Define the players' strategies as Spt
SMt

(mat,m t ), where

SMt. The

to Spt

strategy space is therefore St

define the
players

Spt belongs

loss vector

of

the

constituent game

Vt

game

(X t , pet) and

= SMtXSPt.

Furthermore

(VMt,V pt ). Let N be the number of

(two

at time

and SMt belongs to

is

in

this

section).

denoted by

Then,

the

r(N,St,V t ). Notice

that no dynamic equation links different stages of the game, so

that the subscript t can be omitted.
Three outcomes
(remember that,
on the

of the game are particularly relevant

in accordance with the macroeconomic literature

Monetary Game,

we assume that all outcomes must belong

to the private sector's reaction function (2.8

(i) Control.
cheat so

that mt

(2.8. Consequently,
becomes mt
players'

0,

The monetary

= mat
the

met

implies Xt

0 (see eq.

money stock optimal rate of growth

which implies

losses are:

authority does not try to

pet

Pt

0 (see

eq.

(3.1.

The

181

X*2

(3.2)

( i i) .Ch.e..a.t..ins. The monetary author i ty announces m" t

and,

if the announcement is believed (pet = 0), minimizes his

loss function

by setting

mt

private sector believes the announcement (pet

outcome of
The players'

the game

is X t

m*t.

~(1+~)x-/(~2+b)

~2x-/(~2+b)

The

0), so that the

and Pt

~x-/(~2+b).

losses are:

(3.3)

bx-2/(~2+b)

~3(1+~)x-2/(~2+b)2

(iii).D.i.l;tc. r..e...tLo...n.
anticipates that

The

private

monetary authority's

time inconsistent

sector

correctly

announcement mat = 0 is

and that the optimal reaction to the private

sector's expectations is

(3.4)
~2

+ b

that can be obtained by minimizing (2.4) with respect

to mt , subject to (2.7)(2.8). Moreover,

(3.5)

and (3.4)

imply:

~x*/b

which substituted
the outcome
players'

(2.6)

of the

into (3.4)

gives mt

game is X t

0 and Pt

~x*/b.

~x-/b,

Then

so that the

losses are

(3.6)

The

relevant

outcomes

summarized by the following table

of

the

game

can

also

be

182

Table 1.

Private Sector

Monetary
Authority

mat=O, mt =(3.4)

where the first column gives the outcomes of the game

when the

announcement is

the outcomes

when the

believed and the second column gives

announcement is

not

believed

by

the

private-sector.
Table

defines the standard Monetary Policy Game as

described, for example.

Notice
authority

has
r
strategy m t

in Cukierman (1986).

that

YC Mt

always

an

yr Mt .

so

that

incentive

to

cheat.

<

is therefore

the

monetary

The

optimal

dominated. Furthermore, yd Mt >

yrMt, so

that the stable Nash equilibrium of the monetary game

under complete
information (pdt
~xM/b.
x d t = 0)
is suboptimal

(the

domi nant

strategy

is

sub-optimal).

Th i s i s

Kydland-Prescott's dilemma.
Consider now
the repetition
defined.

'vi i =

T
L

t=O
players'

(T times)

Each player
at i Y it.

(T

of the

now

= M, P.

discount

repeated game

the game

r(N.St.Yt.a.T) which
constituent game

minimizes

the

The vector

is just
previously

intertemporal

a = (aM, a p

loss

def i nes

the

factor, whereas T defines the stages of the

00).

The link among different stages of the

game is constituted by players' memory.

results

If complete

information is

imply

the

r(N,St,Yt,a,T)
game is

that
is

equilibrium

(pdt,X d t ) at

assumed, Selten's (1978)

of

the

monetary

all stages of the game,

game
if the

repeated a finite number of times. Only if the game is

repeated an

infinite number

of times

and a

is

sufficiently

183

close to

1971) or if incomplete information is

one (Friedman,

assumed (Kreps-Wilson,
(at least

in the

Backus-Driffill,

first stages)

of

the

is based

on

three

outcome

~he

repeated

1985a,b ; Fudenberg-Maskin,

This result
First, the

1982), can (prt=o, xrt=O) be

game

(see

1986).

basic

assumptions.

private sector is supposed to form his expectations

in the following way:

if mT =0 at all

(3.7)

<

otherwise
This implies
sector to

an implicit

threat

from

the

private

the monetary authority, which is punished forever by

the private

sector (who

anymore), whenever

does not

a money

believe the

stock rate

monetary

rule

of growth greater than

zero is chosen.
Suppose the
(see Frtedman,
adopts the

game is

1971) close

infinite and aM is sufficiently

to one.

trigger strategy (3.7)

thus forcing

the monetary

Then, tfie

private sector

in forming his expectations,

authority to

choose mt = 0 (at all

stages of the game), in order to avoid the sub-optimal outcome

(x d t , pdt)
at all
future stages of the game (see Friedman,
1971,

1986).

By using

same conclusion
punishment lasts
1983a,b the

Green and

can be

Porter's (1984) results, the

obtained even

a finite

if the

private

s~ctor

number of periods (in Barro-Gordon,

punishment lasts

only one

period,

so

that

the

punishment is not effective).

Furthermore, a
behind the

equilibrium

sequential

Driffill (1985a,b)
mt =0

similar reasoning can be shown to lie

to show

concept

used

by

Backus-

that the monetary authority adopts

(at almost all stages of the game) even when the game is

finite and

incomplete

information

incomplete information

is necessary

is

assumed.

only

if

equilibrium of the monetary game exists. Otherwise,

Nash equilibria
results can

be used

sector's trigger
forced to

exist, Friedman's
to

show

strategy such

that

(1985) and
there

Notice

that

unique

Nash

if multiple

Abreu's

exists

(1986)
private

that the monetary authority is

choose the optimal rule mt = 0 even in deterministic

184

finite games.
The second
raised

only

by

crucial assumption

the

private

authority is

simply allowed

being Pareto

superior to

sustained

by

sections, we
Backus and

the

sector's

that Barro

Driffill's (1985a,b)

allowing the

threats

whereas

the

are

monetary

to pick one of the outcomes that,

the Nash outcome (x d t , pdt), can be

private

will see

sector,

is that

threats.

In

and Gordon's

the

next

(1983a,b) and

results can be generalized by

monetary authority to threaten the private sector

(even when an infinite number of atomistic agents is assumed).

The third
outcomes must
By

relaxing

obtained.

In

basic assumption

belong to
this

is that

all equilibrium

private sector's best reply function.

assumption

more

general

results

can

be

particular, it can be shown that all individually

rational outcomes

are equilibrium

outcomes of

the game

(see

Fudenberg-Maskin, 1986).
For the moment, we want to emphasize that even in the
simpler

strategic

setting

Backus-Driffill (1985a,b)

used
and

by

many

Barro-Gordon
others,

(1983a,b),

the

problem

of

multiple solutions arises.

Suppose

that

the

private

sector

forms

his

expectations in the following way:

(3.8)

otherwise

where mat is the announced monetary policy.

If mt
private sector

= mat,

we have

does not

yrpt =

have any

Yd pt = 0, so that the

incentive to

cheat.

If the

game is

infinite, this implies that the monetary authority can

choose any money stock rate of growth mat = mr t such that :
(3.9)
The proof

is an

easy application

of Theorem 3.3 in

Friedman (1986).
Hence the

time-consistent outcomes

of the

infinite

185

monetary game belong to the following set

{(xrt,pr t ) : 1 > aM > (yrMt-yc Mt )/
(ydMt-yc Mt ) ; t = 1. .. T}

(3.10 )

The same

conclusion can

be achieved in finite games

if
(i )

( incomplete

information) the private sector

assigns a positive probability to mt = mr t ; alternatively, the
private sector does not know with certainty the policymaker's
loss function. Then, the sequential equilibria of the game
belong to the set IR N. The proof follows Backus-Driffill
( 1985a, b) .
(ii) (bounded rationality) each player's strategy is
within E in loss of the minimum possible against the other
player's strategy. Then, the E-equilibria of the game belong to
the set IR N. the proof follows from Fudenberg-Levine's (1983)
results.
Two problems arise
(a) If the monetary authority minimizes WM with
respect to {mt
t=1 ... N} over the set IRN and (x r t = 0, prt
=0) belongs to IR N, then the solution is mt = 0 at all stages
of the game. However, being (3.8) enough to assure Pt = pet and
Xt
0, the private sector is indifferent among any value of
mat and may not believe monetary authority's announcement if it
is not backed by concrete measures which support it. This
objection was raised by Rogoff (1986) who emphasized that a
coordination problem therefore arises.
In other words, the
private sector
should impose
to
the
policymaker
the
policymaker's desired solution. Why should two conflicting
players coordinate in such a perfect way? Furthermore, it is
much more plausible to think that the policymaker tries to
impose his desired solution to the private sector. This idea
will be explored in the next section.
(b) At least from a descriptive viewpoint, the
existence of multiple solutions is rather troublesome. Is there
any way to reduce the multiplicity of possible outcomes
previously emphasized? Rogoff (1986) remarks that multiplicity
arises from the introduction of discontinuous state-dependent

186

trigger strategies.
function of
unique

If players' strategies must be a continuous

state variables

time-consistent

Monetary Policy
stages of

(subgame

Game is

the game

(the inflation

rate),

perfect)

then

the

of

the

solution

the Nash equilibrium (Xdt,pd t ) at all

(see Stanford,

1986a ; Robson,

1986). This

conclusion is not true and multiplicity re-appears if :

accelerated

E-equilibria

of

the

discounting are considered (see Kalai-Stanford,

- average

instead of

considered (see Stanford,

- continuous
each player'
decision

discounted loss

functions are

1986a)

state dependent threat strategies where

strategy

are

with

game
1985)

also

considered

depends
(see

on

the

Carraro,

other

1986b).

player's
This

last

possibility will be examined in the next section.

4. Stackelberg

Trigger Strategy

and the Credibility

Problem
Let us now consider the Monetary Policy repeated game
r(N,St,Vt,a,T) and
without imposing

let us

determine its

the restriction

equilibrium outcomes

that they must belong to the

private sector's best reply function. Furthermore, we introduce

a

hierarchical

structure

monetary policymaker
sector.

In

other

authority adopts
impose his

to

raise

words,
a more

the game,

we

the

game,

threats
assume

by

allowing

against

now

that

active monetary

own desired

we consider

into

the
the

the

private
monetary

policy, by trying to

solution to the private sector. Hence,

the monetary policymaker as the dominant player of

where dominant

desired solution

to the

is the player who can impose his own

other players.

Finally we

assume no

discounting.
The basic
attempt by
produce

feature

the policymaker

more

than

policymaker's effort
only by

Appendix,

Y*t.

is vain

game

if he

As

is

therefore

the private
previously
tries to

becomes effective

appropriately threatens
Trigger

the

to induce

Yt

cheating, but

Stackelberg

of

if

sector
shown,

achieve X t
the

the
to
the

policymaker

the private sector. We want to use the

Strategy

equilibrium

described

in order to derive the policymaker's optimal

in

the

str~tegy

187

and to show under what conditions his threats are effective and
credible. For
a monolithic
and many

the moment, we assume that the private sector is

single player.

The game

atomistic producers

between the policymaker

will be

discussed in

the

next

section.
Suppose

the

monetary

following continuous

authority

announces

(discontinuous strategies

are considered

in the Appendix) trigger strategy (linear for simplicity)

- x)

(4.1)

In other

words, the

the

if X t

S x

monetary authority threatens to

increase the money rate of growth (thus inducing inflation) any

time the

private sector

authority's desired

does not

comply

output growth x.

with

the

In contrast,

monetary

if X t

x,

the policymaker adopts his rational reaction function (3.4).

Notice that
optimal strategy
This is

mt

desired output

when X t

an important

is
x

In other

policymaker's desired

monetary

and that VMt

property of

growth. which

desired outcome.

the

authority's

=0

in this case.

the monetary

implies the

words. we

credibility of

can also

outcomes must

policymaker's

belong to

say

that

the
all

his best reply

function.
Is strategy
private

sector

(4.1) effective

accepts

to

and credible? Will the

produce

x?

Xt

Will

the

policymaker actually carry out the announced threat whenever X t

does not

coincide with

x? What outcome is likely to prevail

at all stages of the repeated game?

To answer
outcomes of
form under

the constituent
the "rules

Then credibility
by using

the above
of t~e

questions we
game which

derived in

the

summarizes its normal

game" assumed

and effectiveness

the results

first derive

in this section.

conditions will be studied

Carraro

(1985,

1986a)

and

briefly presented in the Appendix.

(i)~.

strategy (4.1)

Let us

is effective

first assume
and credible.

that the
In this

trigger
case,

as

188

shown in

the Appendix, strategy (4.1)

is also time-consistent,

since it aims at achieving the policymaker's lowest loss. Then,

the private

sector finds

implies mt
players'

by (4.1)

(ii)
but not

credible and

x-.

In

the policymaker

whenever the

other words,

punishment but
pet.

=0

by (2.6.2). The

V p t = x- 2

The trigger

DiB~~tiQn.

the policymaker's

pet

V- Mt

announced threat

and Pt

losses are
(4.2)

Xt

to produce X t = x-. This

it optimal

strategy is effective,
does not

carry out the

private sector does not produce

the private sector anticipates that

reaction to

his

choice

will

not

be

the rational strategy (3.4) so that met =

This implies

mt

Pt

minimizes his loss function,

~x-/b.

the
~x-/b

Then the private sector

thus choosing X t

= O.

The players'

losses are :
(4.3)
This equilibrium
of the

xd

standard Monetary

coincides with the Nash equilibrium

Policy Game

(i.e. pOt

( iii) .
credible,

but

punishment.
the

The

policymaker

trigger
actually

strategy

is

carries

out

not
the

announced threat in order to increase his reputation and induce

the private sector to adopt X t
game. The
~x-/b,

private sector's

but

now the

x- in the future stages of the

expectation is

threat is

still met

carried out and mt

kt)x. The outcome of the game is :

(4.4.1)
and
(4.4.2)

that implies the following players' losses

p8 t =

ktx t + (1-

189

(4.5.1)

and

(4.5.2)

Therefore, the

relevant outcomes

of the game can be

summarized by the following table

I.a.b.Le__.2...
Private Sector
Xt=X,

m- t =(4.1),m t =(3.4)

p8 t =mol3l. t

Monetary
Authority
m- t =m~_ = ( 4. 1 )

As stated
only equilibrium
"rules of

in the

if

Appendix (see

of

this

repeated a

finite number
standard

section,

is

{pOt=~x*/b,xOt=O}

complete information

obtained the

Proposition 2), the

of the repeated game r(N,St,Yt,a,T) under the

the game"

(discretion) equilibrium
game,

YOM,Y Op

Y*M'Y*P

of

result

is

assumed

times.

We

previously

the

stable

Nash

at all stages of the

and

the

have

game

therefore

described

is
re-

for

the

Monetary Policy Game.

However, by
assuming that
whether the

assuming incomplete information (i.e. by

the private

sector does not know with certainty

policymaker finds

punishment), i t

is possible

it optimal

to

carry

out

the

to show (see Proposition 3 in the

Appendix) that the policymaker's best outcome (Pt

0,

Xt

x)

190

is the

sequential

equilibrium

of

the

game

under

suitable

(effectiveness and credibility) conditions. (3)

Let us

first check

authority's strategy.
(Pt =

0,

Xt

effective)

the

effectiveness

of

monetary

In the Appendix, we show that the outcome

x*)

is

inducible

lie

monetary

policy

is

if VP Mt > V* pt, that is :

(4.6)

Furthermore, the

trigger strategy

(4.1)

is shown to

be credible if :
(4.7)

when the game is finitely repeated (see Proposition 3

in the Appendix)

; that is i f :

(4.8)

where 6
reputation,

can

i.e.

>

----------------------

be

interpreted

as

the

policymaker's

6 is the prior probability that the threat is

actually carried

out by

the policymaker

whenever Xt

is

not

equal to x*.
Furthermore, the
the

game

is

repeated

credibility condition
an

infinite

number

is 6
of

>

times

0 if
(see

Proposition 4 in the Appendix).

Finally, the
VO pt , that

is if

trigger strategy

trigger strategy is necessary if V*pt >

bx* >

O. Assuming

therefore x*

>

0,

any

with punishment component kt{xt-x*) such that

(3) Therefore, whereas Backus-Driffill (1985a) showed that (pr t

= O,x r t
can be the outcome of the repeated game
if
= 0)
incomplete information is assumed, we show that, under the same
assumption, (Pt = 0,
Xt = x*) can be the outcome of the
repeated game
if the Stackelberg trigger strategy
(4.1)
is
effective and credible. Notice that the monetary authority
prefers (Pt
0,
Xt =x*)
to (prt,X r t ),
so that, if (4.1) is
effective and credible, the relevant outcome of the repeated
game will
be characterized by zero inflation and output growth
over its natural rate.

191

(4.6) and
of the

(4.8) are satisfied,

induces (Pt=O,xt=x") as outcome

finitely repeated Monetary Policy Game (see Proposition

3) .

In particular,
section,

using the

in this

it is easy to show the following proposition

.Er:.Q.,p..9...s...i..t..i.Q1LA
where p

results presented

: If k t <
and the

[~(1+~)11/2,

-[p-~l-l[(1+~)b+p(~-b)l

credibility

0,

<

condition

>

is satisfied, then {xt=x",

[b2(1+(1-kt)~)1/[~(1+~)1-kt)b-~)21

Pt=O ; t=1, .... T}, can be sustained as a sequential equilibrium

of the Monetary Policy repeated Game.
Notice that

a large negative punishment parameter k t

is more likely to satisfy conditions (4.6)

(4.8). By (4.1) this

implies that

is threatened by the

a large

monetary policymaker
produce X t

to induce the private sector to

If the threat is carried out, the consequent

= x

inflation burst
sector's

monetary expansion
in order

and output

loss,

so

expansion largely increase private

that

he

prefers

to

comply

with

the

policymaker's desired production decision.

Proposition
announced in

proves

definition of

the policymaker

output grows

the

the Introduction,

consistent (by
such that

first

of

the

results

i.e. that there exists a timesequential equilibrium) strategy

achieves

his

bliss

point,

where

more than his natural rate and inflation is zero.

Hence, this strategy cannot be dominated by any other strategy.

This

changes

the

nature

of

Kydland-Prescott's di lemma

and

increases monetary policy effectiveness.

The third
previous section)

result
can be

(the

second

derived from

was

proved

in

the

Proposition 1 and 3 of

the Appendix, which define the inducible region of the game and
the credibility

condition. The

monetary authority

can indeed

achieve all outcomes that belong to the following set:

(4.9)

IRs

{(x*t,P*t)

>

>

(V*pt-VOpt)/(VPpt-VOpt)
t

where V*Pt
loss when
(x* t, p* t) .

= VPt(x*t,P*t)

the monetary

is the

private

= 1 ... n

sector's

authority achieves his desired outcome

192

Again, the
However,

in

problem

th i s case,

policymaker can

of

multiple

solutions

arises.

it is more plausible to think that the

minimize his

loss function

over the set IRs,

and achieve his minimum inducible loss :

(4. 10)

= {x" t ; t = 1 ... T}, p" = {p"t ; t

Stackelberg Trigger Monetary Strategy.

where x"
and mt

is a

l ... T}

In

this

case, indeed, the policymaker, by appropriately threatening the

private sector,
can achieve his most preferred attainable
outcome.
Hence, given
it exists,

is unique.

6, the outcome of the repeated game,

However,

reputation,

is

be assumed

to be

a subjective

Hence, the

policymaker may

6,

the

policymaker's

if

prior

parameter, which cannot plausibly

known by the policymaker (or by economists).

try to

induce a

desired solution

which does not belong to IRs, so that either he carries out the
announced punishment or he acquiesces. In both cases, multiple
(sub-optimal) equilibrium paths are possible (see CarraroSiniscalco, 1987)
be avoided.

and the problem of multiple solutions cannot

Furthermore,
understood from

the

fourth

result

can

easily

be

the above discussion. The coordination problem

which arises

when the

authority in

order to force monetary policy to be committed to

the announced

optimal

private sector
rule,

has

now

threatens the
become

monetary

credibility

problem, being the outcome of the game strictly dependent on

the credibility parameter 6.
In both cases, the equilibrium
path of the game depends on parameters which cannot be uniquely
determined, thus leaving the solution of the Monetary Policy
Game indeterminate.
In the previous section. the multiplicity
of equilibria was parametrized by the coordination parameter
m~t =
m- t . whereas in this section,
the multiplicity is
parametrized by the credibility parameter 6.
5. Atomistic Players
In

the

previous

section

the

Stackelberg

Trigger

193

Strategy equilibrium was computed and analysed under the

assumption that the private sector can be represented as a
monolithic player. We now relax that hypothesis by assuming
that the private sector is composed by an infinite number of
atomistic players.
Let us further assume that all players are identical
so that, player i's loss function is
(5.1)

Vlt = (Ylt - Y*lt)2 +

i

~(Pt

- p8 1t )2

1,2 ... (I)

where Yl is player i's production decision and pel is

his price expectation. Then, by appropriately defining X*l'
each player's reaction function is :
(5.2)

so that eq. (2.1) can be derived by aggregation.

The rest
of the model is not changed by the
introduction of the atomistic players hypothesis.
The crucial assumption of this section is that each
player is so small that aggregate output is not affected by
changes of the production decision of one player only. We then
study credibility and effectiveness of the trigger strategy
(4.1) .

Suppose each player is uncertain whether the threat

is actually carried out whenever Xt does not coincide with x*
and assume that the prior probability that each player assigns
to the possibility that the threat is carried out is equal to
61
6, for all players. We need to study two games : the
Monetary Policy Game between the policymaker and the atomistic
producers and the game between player i and all the other
producers.
We first study the latter game in two cases : when
the policymaker actually carries out the threat if Xt does not
coincide with x and when he prefers to adopt his rational
strategy (3.4). In the latter case, the normal form of the game
between player
and all the other atomistic players can be
summarized by the following table:

194

Table 3.
Player i

-2

,x - 2

Player j

for all j=t=i

X.,t = R.,(m t )

YOpt,Y+it

is
defined by
i's los&

(5.2),

x, pe. t

Xit

YO p t

when he

the announcement

player

YOpt,YOpt

i's

best

reply

is defined by (4.3) and

believes the

function

Y+. t

is player

monetary announcement and sets

0, but all the other players does not believe

and follow

their rational

reaction strategy

(5.2). We have:

(5.3)
Notice
XJt
Pt

that

= x, aggregate
= O. Hence player

when

all players but player

outcome is

Xt

so that mt

i finds it optimal to choose

Xit

adopt
x and

0, so

that his loss function achieves its lowest value.

It
X.

is easy

RF.(m t

to

prove

that

Y+. t

>

YO p t ,

so that

is player i's dominant strategy at all stages of

the game.
An analogous
case, that
threat if

is when
Xt

result can

be obtained

the policymaker

in

the

actually carries

other
out the

does not coincide with x*. The normal form of the

game is summarized by :

195

Table 4.
Player i

*2

,X

*2

Player j
for all j:fi
x.Jt = Rp i

(m t

YPPt,Y-

it

for all j:f i and

o if x.Jt = X
XP t otherwise, where x P t
is defined by (4.4.1).
Moreover, YP pt is defined by (4.5.2) and V-it is player i's
loss when the punishment is carried out even i~ he adopts the
policymaker's desired strategy. We have:
(5.4)
where pP t i s def i ned by eq.

(4.4.2).

Even in this case,

it is easy to prove that V-it >
YP pt , so
that Rpi(m t ) is player i's dominant strategy at all
stages of the game.
Combining the results of Table 3 and 4, we can
therefore conclude that there is an incentive for player i to
adopt Xit
Rpi(mt ) whatever the other players' decision and
whatever the policymaker's strategy.
Furthermore, being all players identical, Xit
Rpi(mt ) is the dominant strategy for all players, so that no
producer will comply with the policymaker's desired production
decision. Hence, th~ trigger strategy (4.1) seems to be
ineffective, whatever its credibility.
However, each
profitable to

player knows

choose Xit

that when

different from

x,

he
all

finds
the

it
other

players will find it profitable to choose xJt different from

x*, j:fi, so that Xit = 0 for all i, when the punishment is not
carried out, but Xit = x P t and Yit
VP pt for all i, when the
monetary authority actually punishes the producers. A typical

196

Prisoners' Dilemma

therefore arises.

Suppose 6

is such

that

is satisfied. so that the optimal monetary strategy (4.1)

(4.8)

is credible. Then each producer is better off by choosing Xit

x.

but

has

always

an

strategy) to choose Xit

incentive

(whatever

the

monetary

Rpi(m t ). thus achieving a sub-optimal

loss.
The

above

discussion

~x/b)

that.

being

the

the outcomes of the game are st i 11 (x t

producers identical.
x. Pt = 0)

implies

if the monetary strategy is believed.

(x t

= O.

Pt

if the policymaker is not believed but the punishment is

x P t . Pt = pP t ) if the punishment is
out. and (x t

not carried

carried out. Therefore. the aggregate game between the monetary

policymaker and

the atomistic

players is

still described

by

Table 2.
As shown
the incentive

not to

strategy (4.8)
Dilemma can

1986;

above. there

is however

choose Xit

is credible.

be found

x. even when the monetary

The solution

in several

Benoit-Krishna. 1985).

for each producer

to this

ways (see

Suppose. for example. that each

atomistic player adopts a Tit for Tat strategy.

[~:.21

(5.5)

Prisoners'

Fudenberg-Maskin.
i.e.

otherwise

where X11 = x for all i.

As

shown

assumption all

by

Kreps

and

players choose

Xit

al.

(1982).

under

this

x when they believe the

monetary announcement. until they observe xJt different from x

for some

and

announcement is

some t.

As a

consequence.

credible. the

if

the

monetary

sequential equilibrium

.
x

of

the

Pt = 0) at least in the early stages of the

game (see Kreps and a!.. 1982)

game is

(Xt

We
uncertain not

only

need

to

assume

only whether

punishment whenever
producers actually
desired strategy.

each

the policymaker
x.

accept to
In other

that

but
comply

also

carries
whether

with

words. each

producer

the

out
the

is
the
other

policymaker's

player does

not know

197

with certainty

the loss

function of

all the other players of

the game.
Summing

up,

under

the

above

general

incomplete

information assumption, (x t = x*, Pt = 0) is the outcome of the

monetary policy finite repeated game at least in its first
stages,
if
the effectiveness
and credibility conditions
provided in the previous section are satisfied and each
atomistic producer plays Tit for Tat. The credibility condition
is of

course weaker (see Proposition 4 in the Appendix) if the

repeated game time horizon is infinite.

Under this latter assumption it is also possible to
use Friedman's (1971) results to show that (x it = x* for all i)
is the equilibrium point of the repeated game among the
atomistic

producers

when

the

monetary

strategy

(4.1)

is

credible.
Finally, if
multiple Nash

equilibria of

atomistic producers
possible to

RFi(m t ) is

not a

contraction (so

the single-play

exist), and

game

that

among

the

(4.1) is credible, then it is

show (see Friedman, 1985 ; Abreu, 1986) that there

exist a credible production strategy for each producer such

that (x it = x*)
is the equilibrium point of the repeated game
even when the time horizon
information is assumed.
Alternatively, it

is

finite

is possible

and

to

use

no

incomplete

the

bounded

rationality assumption and the E-equilibrium concept to provide

conditions for

(x it

finite repeated

x*) to

game among

be the equilibrium point of the

the atomistic

producers when

the

monetary announcement is credible.

Hence, the
not modified

conclusions of

by introducing

Kydland-Prescott's dilemma
policy problem
eliminated.

and the

an

the previous

atomistic

constitutes a

multiplicity of

section are

private

sector

partial view
solutions

of the

cannot

be

6 Conclusion
This paper
policy. The

has shown

Folk

Theorem

of

monetary

first part of the theorem uses standard results in

the recent industrial organization literature to prove that

198

p r i y ate sec tor ' s---.t.hr:.e..a..t..lLC..an._. .,S..u.s..t..a.in.._. .anY-_......Q.JJ..t. c. Q.me

. ...___w..hJ...c;..h..---.i..s.
Eareto supeiQ~--t~~ash outcome as the equilibr~QY.t~~
of the
s.ame. The second part of the theorem uses the
Stackelberg Trigger Strategy equilibrium concept to show that
the mo ne tar y

pol ic~x::...:_.s........t.hr..e..a.t..s_. ._c_an..... .s...u,S_t"a.i.n._~_.Q..u.t..c..Q.IruL.h..e

p..r:.e.r..e.rJL.t..~t.h.e Nas~Q..U.1...c. Q.lWLAS the equ iii br (urn outcome Qf the

&.aJD,e

These results have been shown both for infinite

deterministic games and for finite games with incomplete
information or bounded rationality
both for monolithic and
atomistic agents private sector.
Furthermore, it has been shown than the multiplicity
of solutions is parametrized either by a coordination parameter
or by a credibility parameter and that the multiple solutions
problem cannot be solved by using continuous state dependent
strategy.
concepts

It is therefore crucial to look for equilibrium

which can restrict the number of equilibria by

defining more
precisely beliefs
and behaviours
out of
equilibrium (see, for a first attempt, Kohlberg-Mertens, 1986).
Finally, we proved that there may exist a credible
strategy which minimizes the monetary policy's loss function
and which is not dominated by any other strategy (i.e. the
po1icymaker achieves his most desired outcome).
The economic fact which explains why the private
sector can be induced to comply with the po1icymaker's desired
outcome is that inflation is costly.
In the simple monetary
model previously considered, only unexpected inflation can be
used by the monetary authority to threaten the private sector.
In this case, the private sector evaluates the probabilistic
cost of
being punished
when he
does not
choose the
po1icymaker's desired strategy, i.e.
the cost of having an
inflation rate higher than expected, and then he decides
whether to comply with the policymaker's decision.
However, if both expected and unexpected inflation
are costly, then the monetary authority's strategy is more
effective. In particular, the announced monetary strategy, i.e.
a monetary expansion whenever output growth is not greater than
its natural rate of growth, is more likely to be credible when

199

both expected

and unexpected

inflation are

included

in

the

private sector's loss function (this is proved in Carraro, 1986

b). The
into

economic rationale

the

private

variability is

for including

sector's

costly and

loss

expected

function

that output

is

inflation

that

output

variability depends on

the inflation level.

Finally, the
the paper

is fairly

indeed realistic
out a

trigger monetary
plausible and

strategy proposed

probably "deja

vu".

in
It is

to expect that the monetary authority carries

further monetary expansion whenever his output target is

not achieved.

APPENDIX
Let VO(SO,SF)
policymaker's loss
function, which

and VF(SO,SF)

function

and

are assumed

preferences over

be,

the

to

respectively,

private

represent

the respective

sector's

the

targets. The

the

two

loss

players'

subscript D

is

used for the policymaker since he is assumed to be the dominant

player of

the game.

the subscript F,

Similarly, the private sector, denoted by

is assumed to be the follower.

The loss

functions Vi'

defined, continuous
assumed to
D,F,

is

i =

and bounded

be concave

the strategy

on S

with respect
of each

D,F, are
= SOXSF

assumed to be
and euch Vi is

to Si' where Si E Si'

player and

Si is

i =

the strategy

space.
The constituent game is therefore r(N,S,V) where N is
the number

of players (two in our case) and V = (VO,V F ).

consider T

repetitions of

the players
we obtain
ai =

the constituent

If we

game and we assume

to discount future losses at the rate ri'

i = D,F,

the repeated game r(N,S,V,a,T) where a = (aD,aF) and

1/(1+rl)'

=D,F.

Both

finite

and

infinite

horizon

repeated games will be considered.

Following

Carraro

Strategy equilibrium
dominant player's
stage of the game.

(1985),

Stackelberg

Trigger

is a sequential equilibrium such that the

desired outcome (SO,SF) is achieved at any

200
Let us consider the constituent game r(N,S,V).
Suppose

that

optimum outcome,

(SO,SF)

is

the

dominant

i.e.

(A. 1 )

arg

min VO(SO,SF)
SO,SF

and let VO,V F be the relative players'

Then,

player's

the

dominant

player's

losses.

Stackelberg

Trigger

Strategy is defined as :

if SF = s F

(A.2)

otherwise

where sPo
player threatens
comply with

defines the
to carry

the dominant

punishment that

out any

the dominant

time the follower does not

player's desired

strategy SF'

The

Stackelberg Trigger Strategy must be such that :

(A. 3. 1 )

arg

min
SF

VF(stBo,SF)

and

(A.3.2)
where Ri(sj)
(4)

defines player

i's best reply mapping,

i.e.
(A.4)

arg

D,F j

In other
the follower

D,F

words, the

is induced

dominant player's

=1=

min
Si

Vi(Si,Sj)

strategy stBo must be such that

to choose,

desired strategy

in his
s F'

own

Then,

interest,the
the

dominant

player finds it optimal to adopt so.

Two properties

must be

satisfied by the Stackelberg

(4) For the sake of simplicity,

the mappings R i { , ) , i = L,F,
are assumed to be contractions, so that the Nash equilibrium of
the single-play game is unique.

201

Trigger Strategy

for the

policymaker to

any stage of the repeated game: first,

i.e., assuming
out the

that the

achieve (S*O,S*F) at

stso must be

~~,

dominant player is committed to carry

punishment, then the follower must prefer to adopt S*F

rather than be punished; secondly, stso must be .c..Le..d.l. b. l.e..,

the follower
that the

must assign

a positive

dominant player

threat whenever

i.e.

probability to the event

actually carries

out the

announced

differs from S*F' and this probability must

SF

be such that S-F is actually adopted by the follower (5)

A

detailed

credibility problem
found in

analysis

of

the

for Stackelberg

Carraro (1985)

effectiveness

and

Trigger Strategies can be

where the

following propositions are

shown
F'Lo"p"Q..s..i..t..i. Q.ll-l

If the dominant player's most desired

outcome (S-O,S-F) cannot be induced by the strategy (A.2) where

sP o is defined by
(A.S)

arg max VF(SO,SF)

So

then it

cannot be

induced by

any other Stackelberg

trigger strategy.
Hence,

Proposition

Stackelberg trigger

defines

strategy and,

the

most

powerful

consequently, the inducible

region of the game :

IR

(A.6)

are

where
follower's loss

when SF

player actually

carries out

eq.

(A.S)

the

differs from

dominant
S-F

and

player's
the

and

dominant

the announced threat. Notice that

implies:

(5) If the follower

is certain that the punishment is carried
out whenever he does not comply with the dominant player's
desired strategy,
then effectiveness of the STS strategy
is
sufficient for (S-L,S-F) to be the outcome of the game.

202

VP F

min
SF

max
So

VF(SO,SF)

Therefore, the inducible region IR defines the set of

outcomes that can be induced by the strategy (A.2).
Let us

consider now

where we assume at = I,

the repeated game r(N,S,V,a,T),

D,F.

If
> VOo ,

and VP o

the only

finitely repeated

complete information

sub-game

perfect

is assumed

solution

of

the

game is (SOO,SOF) at all stages of the game,

where (SOO,SOF)

is the Reverse Stackelberg equilibrium of the

single play game and VOO,V OF are the relative players' losses.
Therefore, under

complete information,

player never

finds it

profitable

whenever the

follower does

strategy and chooses SF

A different
as in

to

carry

not believe

the dominant

out

his

threat

the announced trigger

SOF'

conclusion can be achieved if we assume,

Kreps-Wilson (1982), that the follower does not know the

dominant player's

loss

function,

so

that

he

is

uncertain

whether the threat is actually carried out whenever he does not

comply with
define as

the dominant
6 the

assigns to
the first

player's desired

prior probability

the possibility
stage of

that

that the

the game

strategy.
the

Let

private

us

sector

threat is carried out at

and let

us

suppose

that

this

probability is revised at any stage of the game by using Bayes'

rule (see Kreps-Wilson,

1982). Notice that 6 can be interpreted

as the policymaker's reputation.

Then we can prove (see Carraro,
Proposition

If

in

1985)

complete

information

is

assumed, the strategy stBo can induce the follower to adopt S*F
at any stage of the finitely repeated game if and only if :
(A.7.1)

(S*O,S*F) E IR

(A.7.2)

6 > (V*F - VOF)/(VP F - VO F )

Then the

sequence {SOt

1,2 ... T} can

game.

be sustained

as a

sequential equilibrium of the

203

Proposition 4 : If incomplete information is assumed,

= 00,

(A.7.1) holds and T

1,2, ... }

the sequence {SOt

s"o, SFt

S"F ; t

can be sustained as a sequential equilibrium of the

game for any 6 > O.

As a
global

consequence, the

minimum

conditions are

of

his

loss

dominant player
function

satisfied. Hence,

at no

any

achieves the

time

the

above

stage of the repeated

game the dominant player finds it profitable to depart from the

announced trigger

strategy

st- d

which

is

therefore

time-

consistent.
The above

results define

the main properties of the

Stackelberg Trigger Strategy and the implied equilibrium of the

game.

REFERENCES
Abreu,

D.

(1986),

"Extremal

Equilibria

of

Oligopolistic

Supergames" , Journal of Economic Theory, 39, 191-225.

Backus, D. and Driffill, J.(1985a),

"Inflation and Reputation",

American Economic Review, June.

Backus, D. and Driffill, J.

(1985b), "Rational Expectations and

Policy Credibility following a Change in Regime", Review

of Economic Studies, April.
Barro, R.J.

(1976), "Rational

Monetary Policy",

Expectations and

Journal of

the

Role

Monetary Economics, 2,

of
1-

32.
Barro, R.J.

(1986), "Recent Developments in the Theory of Rules

versus Discretion",

Economic Journal,

Supplement,

96,

Discretion

and

23-37.
Barro, R.J.

and Cordon,

Reputation in

D.

a Model

Monetary Economics,

(1983a),

"Rules,

of Monetary

12, 101-121.

Policy", Journal of

204
Barro, R.J.

and Gordon,

Monetary Policy

D.

(1983bl,

in a

"A

Positive

Natural Rate

Theory

of

Model", Journal

of

Political Economy, 91, 589-610.

Benoit, J.P. and Krishna, V.

(19851,

"Finitely Repeated Games",

Econometrica, 53, 905-922.

Carraro,

C.

(19851,

Policy

"Hierarchical

Analysis",

Games

chapter

Macroeconomic Policy

Analysis",

Princeton University.

Also in

D.

(19871,

eds., Developments

for

Macroeconomic

of

"New

Ph.

D.

Methods

for

Dissertation,

Carraro, C. and Sartore,

of

Control

Theory

for

Economic Analysis, Martinus Nijhoff, Amsterdam.

Carraro, C.

(1986al,

Martos,

B.

"The Theory of Optimal Announcements",

and

Modelling and

Ziermann,

Control of

C.

(19871,

eds.,

National Economies,

in

Dynamic
Pergamon

Press, Oxford.
Carraro, C.

(1986bl,

"Stackelberg Trigger Strategy and Monetary

Policy", Nota di Lavoro 8612, University of Venice.

Carraro,

(19871,

C.

"Credibility,

Indeterminacy of Macroeconomics",
Y. and McKenzie, G.

Reputation

and

the

in Artus, P., Barroux,

(19881, eds., title to be announced,

Kluwer Academic Pub., Dordrecht.

Carraro, C.

and Siniscalco,

D.

(19871,

"Producers' Strategies

in the Oil Market: a Game-Theoretic Framework", mimeo.

Cukierman, A.

(19861,

Some Recent

"Central Bank Behaviour and Credibility:

Theoretical Developments",

Federal Reserve

Bank of St. Louis Review, 5-17.

Friedman,

J.\I.

(19711,

"A

Noncooperative

Equilibrium

Supergames" , Review of Economic Studies, 38,

1-12.

for

205
Friedman.

(1985),

J.V.

"Cooperative

Horizon Noncooperative

Equilibria

in

Finite

Supergames" , Journal of Economic

Theory, 35, 390-398.

Friedman.

J.V.

(1986),

Game

Theory

with

Applications

to

Economics, Oxford University Press, Oxford.

Fudenberg.

and

D.

Equilibria

D.

Levine,

of

Finite

(1983),

and

"Subgame

Perfect

Horizon

Games",

Infinite

Journal of Economic Theory, 31, 251-268.

Fudenberg. D.

and Haskin.

Repeated

Games

E.

with

(1986),

"The

Discounting

Folk

or

Theorem

with

in

Incomplete

Information", Econometrica, 54, 533-554.

Gardner. P. (1973), "Mathematical Games", Scientific American.
Green. E.J.

and Porter. R.H.

under Imperfect

(1984), "Noncooperative Collusion

Price Information",

Econometrica,

52,

87-100.
Kalai, E.

and Stanford,

Strategies in

V.

(1985),

"Conjectural

Variations

Accelerated Cournot Games", International

Journal of Industrial Organization, 3, 133-152.

Kohlberg, E.

and

Hertens,

J.F.

(1986),

"On

the

Strategic

Stability of Equilibria", Econometrica 54, 1003-1037.

Kreps. D.H.

and Vilson,

R.

(1982),

"Reputation and Imperfect

Information", Journal of Economic Theory, 27, 253-279.

Kreps.

D.H.

et

Finitely

al.

(1982),

Repeated

"Rational

Prisoners'

Cooperation

Dilemma",

in

Journal

the
of

Economic Theory, 27, 245-252.

Kydland. F.

and Prescott.

Discretion

the

E.C.

(1977),

Inconsistency

"Rules
of

Journal of Political Economy, 473-491.

rather

Optimal

than

Plans",

206
Lucas. R.E.

(1973), "Some

International

Inflation Trade-Offs",

Evidence

American Economic

of

Output

Review,

326-

334.
Robson. A.J.

(1986), "The

Existence

Reaction Functions

of

for Dynamic

Nash

Equilibria

Models

of

in

Oligopoly",

International Economic Review, 27, 539-544.

Rogoff.

(1986),

K.

"Reputational

Constraints

on

Monetary

Policy", mimeo, University of Wisconsin.

Rothemberg. J.J. and Saloner. G.
Model of

Price Wars

(1986), "A Supergame-Theoretic

during Booms",

American

Economic

Review, 76, 390-407.

A.

Schotter.

(1980) ,

The

Theory

Economic

of

Social

Institutions, Cambridge University Press.

Selten.

R.

(1978),

"The

Chain-Store

Paradox",

Theory

and

Decision, 9, 127-159.
( 1986a) ,
"On
Continuous Reaction
Stanford. V.
Equilibria in Duopoly Supergames with Mean

Function
Payoffs".

Journal of Economic Theory, 39, 233-250.

Stanford.

V.

(1986b).

Equilibria

in

"Subgame
Discounted

Perfect
Duopoly

Reaction

Function

Supergames

Trivial", Journal of Economic Theory. 39, 226-232.

are

ON THE CONVERGENCE OF BELIEFS AND POLICY TO A RATIONAL

EXPECTATIONS EQUILIBRIUM IN A DUAL POLICY PROBLEM
Tamer BASAR and Mark SALMON
Coordinated Science Laboratory
University of Illinois at Urbana-Champaign, USA
and Department of Economics, University of Warwick, England,

1. Introduction
This

paper considers the question of the convergence of

expectations
asymmetric
the

and policy

in

model

of

monetary

policy

with

and imperfect information between the policy maker and

private sector.

the

policy maker

the

optimal policy

Certainty

In this

is

model the objective function facing

n2n~quadra~

because of the manner by which

influences the private sector's expectations.

equivalence does

not apply to the optimisation problem

and

the optimal policy reflects a ~ control structure in which

the

policy maker

policy
and

action on

also on

when

for the

into account

the information

his ability

the private

capacity

must take

to affect

sector forms
policy

maker

both the effect of his

set facing

its
to

the private sector

how this information is used

expectations.
actively

Despite

intervene

this

in

the

expectation

formation process

there

unique rational expectations equilibrium in the model

to

is a

which both

the expectations

optimal

policy converge.

problem

which is

we
the

describe a

of the private sector we show that

of the

In section

solved analytically

number of

factors affecting

private sector

2 we

introduce the

in section 3.

and

the

policy

In section 4

numerical simulations designed to study

the convergence

to the

unique maximizing

equilibrium of this model.

2. The model of monetary policy and expectations
We
Cukierman

consider the

model of monetary policy introduced by

and Meltzer (1986) and previously analysed in Basar and

208

Salmon

(1987). The

monetary authority

faces

finite

horizon

optimisation problem with an objective function of the form:

(1)

N
(~

- 1/2(m P )2)}

~l(eixi

i=O

where the policy

growth, mP .

instrument is

the

planned

rate

of

monetary

The

private sector's

information set

is imperfect

since it

is

assumed that it is only able to observe mi' the actual rate of

monetary growth rather than the policy maker's planned rate of
growth,
the

mPi' Actual
which has

the effect of adding a random disturbance,

- E [mi I Ii 1
i

ei

mi

mi

mP + III i ,

where
(3)

that

to deviate from

to the planned rate. The monetary surprise, e

(2)

The

is assumed

planned rate as a result of imprecise monetary control by the

authorities
III i ,
by

monetary growth

llIi

is then given

mP - E [mpi L 1 + llIi

0,1, ...

N(O,02)

'"

problem also contains an element of asymmetric information in

the

private

sector

is

assumed

not

to

know

the

basic

preference parameter, Xi, of the monetary authority. The value

taken by Xi reflects the trade off as seen by the policy maker
between the benefit from monetary expansion and the loss from
increased
This

inflation in

it's objective

preference parameter

allowed

to change

components

is considered

over time

leading to

with both

function in
to

be

each period.

stochastic

what is effectively the state equation for

the optimisation problem,

(4)

Xl

and

permanent and transitory

9Xi_l + A( 1-9) + Vi'

1,2, ...

209
The
then

information set

given by

{mi-l,

the set of observations on past money growth,

mi-2,""

structure

mol,

together with

and parameters

constructs

available to the private sector is

(except

a knowledge

Xi)'

The

of

the

monetary

Ii

model

authority

its optimal policy based on a knowledge of its current

preferences

Xi as

particular

well

as

Ii,

nested information

hence

the

asymmetric

and

in

structure. So in general we seek a

policy rule of the form

( Xi' Xi -

(5)

Under

these informational

able

to solve

1 ,

xo,

Ii)

assumptions the

the private

monetary authority is

sector's prediction

problem and will

naturally take this into account when designing policy.

Notice
apparently
form
the

that

quadratic the

due to

the presence

will affect

when

it forms

that

it

may

mechanism

also

choice

of monetary

linear
reflect

this

message

the

monetary

it is not so readily recognised

the

It is this

the effect

of the

optimal

the formation of expectations that

linear

quadratic.

While

conditional

the conditioning variables. So in general the

component
function

of
of

to

the
past

be

the problem

objective

function

monetary
linear

to a

growth

function

may

rates.
of

the

linear quadratic form

take into

account that the monetary authority may

manner by

which the private sector forecasts we

be forced to solve a non-certainty equivalent, dual control

in order

authority.
period

that

facing the private sector

which certainty equivalence applies. As we shall see

affect the

problem

clear

the way this information is used.

expectation

when we

shall

qu?dratic

formation

set returns

one for

also

a linear

is

expectation

non

inflation

information
below

function

operator is a linear operator it is not necessarily a

a nonlinear

Forcing

is

affect

growth on

function of

surprise

While it

influence through

the problem

expectation

of

the information

directly

and hence

channel of

objective

of the private sector's expectation in

it's expectations

second
makes

the

problem is

objective function.

authority

and

although

to find

The discussion

case for

reasons of

without introducing

the optimal
here will

policy for

the

be restricted

tractability and

to

monetary
the

two

conveying the main

excessive notation.

The

reader

is

210

referred

to Basar

complete

discussion of

papers,

and Salmon (1987) and Basar (1987a) for a more

the general case. As opposed to these two

our intention

in the

present paper

is to

consider the

convergence

of policy and expectations to a rational expectations

equilibrium

in the

of

the general

dual problem which arises in the first period

two period problem. By focussing on a single time

interval

we are

economic

parameters on

able to

consider

the

the iterative

effect

of

process and

the

various

also

observe

global convergence confirming our theoretical results.

3. Solution for the optimal policy in a two period
problem
For

the

two

period

problem

(with

1) ,

the

optimisation problem faced by the policy maker is

E{

maxJ

(6 )

i=O

9xo + A ( 1-9) +

x,

subject to

l3 i c(xi,Ii,m P )}

Vi,

Vi

fIl i

c(x,I,m P )

(7)

where

The

change in

discussed

'"

N (0,0 2

) ,

N (0,0 2

'"

i =0, 1

mP(x-E[xIIl) - 1/2(m P )2

the objective

fully in

function from

Basar and

that given in (1)

is

Salmon (1987) and follows from the

equivalence
E{mPxi - E{mPlli}x i

1I2(m P )2}

1/2(mP)2}

E{mPxi - mPE{xill

which

holds, given the nestednes or tower property of conditional

211

expectations, because the following identity is valid,

E[E{mP!Il}xl) = E[E{E{mPlli}xllll})

(9)

E[E{mPII1}E{xlll.})
i

= E[mPE{x.II.})
1

Notice that it is the unconditional expectation that is

relevant given the objective function in (6). Hence the two
optimisation problems are mathematically equivalent despite the
apparent change in preferences. Essentially placing the private
sector's interest on the policy variable mP itself when forming
1

expectations in the surprise component is simply masking what

is
the more
fundamental problem
of the private sector's
expectation
(and hence
surprise) in
the basic preference
parameter Xl'
Now for the last period, i=I, there will be no issue of
information transmission and so the optimal policy can be simply
seen from (7) to be given by

The optimal cost function for the two period problem can then be
written. having substituted the optimal policy rule for the final
period. as
max J a = max

E{1/2~(x.

- E[x.II.)2 + (xa - E[xa - E[xalla)m P

a

- 1/2(m P )2}
a

( 11)

But, given that

I. = ma = mP + ~a,
a

we may rewrite the innovation in period

as

(12)

~a)

x. - E[x.1 I.) = 9(xa - E[xalmP +

so the optimal cost function becomes

v.

212

( 13)

1/2~a2

max

E{1/282~(xo

- E[xo\m P + O])2
o

which

(x o - E[xo\Io])m P - 1/2(mP)2)
o

may be rewritten in terms of the yet unknown policy rule a o

as,
( 14)

1/2~a2

+ max F(a o )

ao

where

F(a o

difficulty
policy

is

in solving

this problem
o

growth

when the private

(or the

lies

is part

in

that

the

optimal

of the conditioning

sector's expectation of monetary

preference parameter)

is taken

in period 1. The

we solve the problem below is to simultaneously solve for the

optimal
1,

predictor (expectation)

for the private sector in period

6, and the optimal policy rule for the government in period 0,

say

a. Since

will

we know that the private sector's forecast function

depend on

government's
sector's
of
a

on the right hand side of (13). The

in the initial period, mP

information
way

the maximand

the government's

optimal policy

forecast function,

policy rule,

rule

will

we need

depend

and in
on

turn

the

the

private

to examine the fixed points

the response mappings 6(a) and a(6). We now show that there is
unique fixed

point

corresponding

to

the

unique

maximising

rational expectations equilibrium in this model.

Using

the

unknown

predictor

function,

6,

in

the

objective function, J, we define

(15)

G(6,a)

E{1/282~(6(I,)-xo)2

( 16)

mP
o

noting

that the

problem"
error

= a ( x o , 10

through its
this,

the prediction

1/2(m P )2)
0

1, = m P + o
o

information set

facing the

recognising

),

+ (x o - E[xo\Io]lm P

10 will

be empty.

The "policy

private sector is to m. i..n..i.m..i..s. e.. its prediction

choice of

6.

will determine

The

monetary

its optimal

authority,
0:,

error as in the following problem;

by

upon

maximising

213

max min G(6,o:)

max F(o:)

( 17)

0:

0:

6(0:)

We next show that the function G in fact admits a unique saddle

point,
in other words there exists a unique pair of policies
(6 M,o:M) such that
G(6 M,o:M)

(18)

max min G(6,ex)

ex
c5

min max G( 6 , ex) ,

c5
ex

or alternatively,

Clearly given any

(17) that

such saddle-point

pair (6 M,ex M)

we have

from

max F(ex)

(20)

0:

and furthermore ex
is the unique maximising
M
M
(6 ,ex ) is unique as a saddle-point solution.
Before
some

presenting the

notation. Let

L =

La be

main result

solution above
we first

introduce

a real solution to the polynomial

equation
La 2

( 21>

(L a
2

'"

92~

+ 02

if

= geL)
)2

a
'"
that provides the largest value for F(ex) and let Ka be given by
X

(22)

Furthermore introduce the function

214
(1-Ke2~)

r (K)

(23)

=
(1-K2e2~)

and the condition L o (1-L o

(24)

< 0

)02
x

..

THEOREM

r(
o

<

(i)
~

equation (21)

The polynomial

(L,

is identical with

maximal F(a), real solution Lo

and admits a

L
with

Lo < 1.
If Lo satisfies (24), the game G admits the unique

(ii)

saddle-point solution

Xo + KoIt
Lo(xo-i o )

(25)

;5
;5

xo + ~ (Lo)I.
r(Ko)(xo - E[xoIIo])

where a- is also the unique solution of (20).

(iii) Condition (24) can equivalently be written as
(26)

1 -

(i)

Existence and

(Lo)2e2~

>

<=>

1 - K2
o

e2~

>

Proof
solution,

Lo

solution

lies

of gel)

fact that

(21)

in

the

changes sign
is

open

right

Here we

interval

maximising

at most

any a

a-,

this conditional

and

(0,1),

the

once in that interval.

r(

(L

verify the pair of inequalities (19). The

is minimised

by the conditional mean of Xo (given I.), and when a

side inequality,

into (23).

hand side follows since G(6,a), given by (15),

for
hand

F(a)

identical with the equation

follows from the substitution of (22)

(ii)

the

follows from the simple observation that every real

of (21)

derivative
The

uniqueness of

mean is linear in I. as given. For the left

note that

G(6-,a)

is a quadratic function

of a, with the coefficient of the quadratic term being

215

(27)

[1 -

L2(02
0
"0

(L 2 02 + 0 2 )2
o

The

condition T

concave

)282~

<

function of

"

= T

directly implies that

a, and

G(6~,a)

being quadratic,

is a strictly

it admits a unique

solution which is
ex(x a

and by (i)
ex(x o )

This

verifies the

the

Lo(x - xo)
left hand

o.

condition T <

side of

the inequality (19), under

Using the fact that Lo satisfies (21), T can

be simplified to
T = 1/2(Lo (I-L o )(02
Xo

and
Note
and

hence the

point
that

this condition,

is

is indeed equivalent to (24).

G(6~,a)

admits a unique maximum,

interchangeability property

equilibria [Basar
(25)

1)

concavity condition

that under
using the

/0 2 ) -

and Olsder

indeed the

of

(1982)]

multiple

saddle-

readily

follows

it

unique saddle-point solution of Gunder

(24) .
(iii) This follows readily by noting that

and hence the condition T < 0 is equivalent to (26).

The

condition (24)

solution
problem

of (21),

of the

Theorem is

and this

depends on

given
the

o
terms

of

the

parameters

of

the

in

only implicitly. A more explicit dependence purely on the

parameters

0 2
K

,0 2 ,

...

8 and

can be seen in the condition

216

(28)

< 40 2 /0 2

82~

which

xa

implies (24). To see this implication, note that in view of

(26), condition (24) is equivalent to

L 2 02
a x

(29)

.0 2
x

(L 2 02 +
a x
a

82~

a
<

0 2 )2

but since
L 2 02
a x

Max

.0 2
x

(L 2 02 +
a x
a

La

82~
x

0 2 )2

the preceding inequality is always satisfied under (28).

Condition
sufficient

(28), or

for the

the less

restrictive one (24), are

linear solution a* given in the Theorem to be

overall maximising, but there is no indication that it is also

necessary.
In fact,
it is quite plausible that the result is
valid

for all values of the parameters defining the problem. Non-

satisfaction

of (24)

not

saddle point

admit a

larger

simply means
(that is

that the auxilary game G does

the upper

value is strictly

than the lower value) ; however this does not rule out the

possibility

that

the

maximising

solution

for

F(a)

is

still

linear.
In fact
if we restrict the monetary authority to linear
(or affine) policies at the outset, say
(30)

then

we can

show through

routine manipulations

that a solution

always exists in this class, and the optimal L is the

maximising solution of (21) (see Basar and Salmon (1987)).

unique

217

4. The convergence to a rational expectations

equilibrium
The game-theoretic analogy used to prove the basic
result of the theorem also suggests a possible iterative process
by which the optimal policy and prediction may be generated.
Since the private sector is maximising G, by a proper choice of 6
while the government is minimising the same by choosing a, a
crude saddle-point iteration on the coefficients of the two
response functions, one the policy rule and the other the
expectation mechanism, would be
(L'n' )

(31 )

r (K' n')
and r are defined by (22) and (23), respectively. Note
where
that if the sequences {K'n'} and {L'n,} generated by (31), for
any initial conditions, converge, then the limits have to be the
optimal solutions Ka and La respectively, defined earlier. The
fact that the pair
Xa + KaI,

constitutes a saddle-point does not imply that the sequences

generated by (31) necessarily converge ; however, as tabulated
below,
in all the runs that comply with the concavity condition
of the second order, that we have tried, convergence has been
achieved after a reasonable number of steps with some initial
conditions
and parametric values requiring more steps than
others. This iterative scheme can be viewed, much as in the
standard cobweb analysis, as the movement in real time towards
equilibrium with a Cournot behavioural interpretation. In other
words, the response by one party to the other's most recent
adjustment,
in policy or prediction respectively, is optimal in
terms of the criterion function G.
The behavioural explanation of this iterative process
may be extended by noting that if adjustment is costly then on
standard arguments the most rapid convergence towards equilibrium

218

would
as

be preferred. There are a number of different possibilities

to the way by which convergence may be accelerated, one of the

simplest

approaches is

factors in (31)

through the

llsL(n> +

This
Over

relaxation

or Jacobi

(31). The

into

the

relaxation

process.

one

step

levels of

This

memory

simple Gauss-

provides

than the

to

another.

more

basic iteration,

be

Different

seen

as

confidence that

most recent

corresponding

than the

Successive

abrupt changes in the response of the other

parameters could

different

of

essentially introduce some memory

behavioural interpretation

from

other's

the class

methods that would describe the basic iteration

relaxation terms

buffering against

party

is in

algorithms rather

adjustment

reasonable

(l-~s)r{K(n

iterative scheme

Relaxation (SOR)

Seidel

by

of

leading to ;

llpK(n> + (l-J.lp) Ll (L(n

(32)

in

introduction

(see for instance, Basar (1987b

response. The

for

corresponding

the parties
level of

relaxation parameter

values

moves

to

the

place in

confidence

away

the

from

each

as

the

the
value

zero, and it attains its maximum at unity.

Figure
representing

shows

the policy

the

graphs

of

the

two

functions

parameter K and the forecast parameter L

given by (22) and (23),

The

and L

exact shape

economic

of these

e, f>,

parameters

a unique

equilibrium

Starting

from any

the

unique fixed

can

(1-Ke 2 f
(1-K 2 e 2 f
functions depends on the values of the
0

and

2
H

be seen

but the

clearly

initial position,
point proceeds

existence of

..

from the

say <Ko,Lol

by iterating

figure.

convergence to

between

the

two

functions.
Table
been
the

1 summarises several of the simulations that have

carried out
rate with

using different values for the discount rate f>,

which the

policy maker's preferences change

and

219
02

which represents the ratio of the variance in monetary

noise

policy

to that in the initial uncertainty in

02

the policy maker's preferences

02
K

number of

observations

simulations.

The first

expectations

equilibrium has

unable

to cause

through

values

of K

and L

all

of F(a)

results
when

the

the weaker

been

instance run

and always

to

the

7)
same

and 8

factors,

discussed

It is

from 110

and

for

is also obtained

to 14.

8 were

reduces

~8'

In general

relaxation factors

required to

in run
Aside
the value

there will exist optimal

the number

of

what these values will be. The values

sensitivity of

it can

and when

it is

closer

to the

(L=l).

Similarly when

or

and

~8

exist.

the rate of convergence

the optimal value of L becomes

certainty

the product of

of the

~p

also be seen that as it becomes both

very small

myopic

small (eg.

of

however simply determined by grid search and

large

the square

number

a minimum and in some case it is possible

from the
of

the

that reduce

it is entirely possible that better values for

becomes

above

noticable from the

that convergence

~p

calculate theoretically

used

to a.

2 ,

a comparison of runs 4 and 8 where the introduction

for these

iterations

conditions

sufficiency conditions fail to be satisfied. The

relaxation

iterations

4, 5

in particular the choice of

of the relaxed iteration procedure discussed above can

seen from

values

sufficient
with respect

for runs

importance

and

we have

to any other point

(see for

been rapid

the parameter values chosen,

violated

to

rational

of initial conditions. Even for widely

has invariably

concavity

to

the

in that

to converge

numerical

where convergence was not obtained, such as in run 9, arose

when

of

the

of K and L for common settings of the parameters. The only

cases

be

from

convergence to

been global

the simulations

initial values

convergence

follow

being that

differing choices

divergent

rate of

equivalent
~

and 9

2 ,

solution

value

the discount rate

change of government preferences

runs 2,3 and 12) the optimal solution returns

to

the myopic solution which obviously requires no iteration. Run

10

shows the

walk

and there

greatest
is

again

case when
is no

deviation from
a

rapid

government preferences
discounting. This

follow a

solution

random

provides

the

the myopic case and as can be seen there

convergence

to

the

rational

expectations

220
equilibrium.
more

The welfare implications of these separate cases are

fully explored

should

also be

matrix

with the

equations

in Basar

noted that

but in

SOR method

and

Salmon

the spectral
is affected

various experiments

(1987).

Finally.

it

radius of the iteration

by the

we found

ordering of the
no effect on the

convergence to equilibrium.

Figure 1

L
K=f(L)

.........

_-/

5. Conclusions

In
and

empirically

equilibrium
the

this paper
the

we have

demonstrated both theoretically

convergence

to

rational

expectations

of the optimal monetary policy for the government and

expectations generating

mechanism of the private sector in a

221

non

conditions
found

unde~

to be

behaviou~al

to the
the

~educing

va~iance

in

Salmon (1987),
a

"Stackelbe~g"

Cukie~man

and

Meltze~

mo~e

the optimal
solution

fo~

(1986) has

Table 1 -

gene~al
~elaxed

and

ite~ations

In

the

pa~amete~

~atio

of

the

unce~tainty

in

the

is

small

ve~y

difficult. As discussed in
de~ived

the same
de~ived

Nume~ical

its

it seems,

gene~al

dominant

2 ,

~atio

policy

we~e

of the

noting.

this

demonst~ated

impo~tance

of

of

theo~etical

simulations and in

the initial

When

is potentially

conve~gence

and

be

that the

the value

p~efe~ences.

gove~nments

wo~th

noise to

moneta~y

The

at least,

is

conve~gence

could

numbe~

is

inte~p~etation

The

p~oblem.

nume~ical

and global.

these calculations

f~om

affecting

cont~ol

conve~gence

~apid

in

ite~ation

which

~elevant

was

conve~gence

to

equivalent

ce~tainty

above

policy

Basa~

co~~esponds

p~oblem

that

a "Nash" solution.

simulations

Initial

Cond r " "

1
2
3
4
5
6
7
8
9
10
11
12

0.95
0.25
0.95
0.95
0.95
0.95
0.95
0.95
0.95
1.0
1.0
0.1

Relaxation
~un

8 when

0.8 1 . 0
0.8 1.0
O. 1 1.0
0.8 0.1
0.8 0.5
0.8 10.0
0.8 1.0
0.8 O. 1
0.8 0.01
1.0 1.0
1.0 10.0
0.1
1.0

pa~amete~s
~p

6
4
3
110**
6*
7
7
14**
FAILED
9
8
3
and

~p

0.9 and

..

(b) Numbe~ of ite~ations to

*Violates 0 2 /02 ~ 4
xo

** Violates

..

(0,0)
0.4905 0.8220
(0,0)
0.4995 0.9593
(0,0)
0.4999 0.9976
(0,0)
0.9548 0.9411
(0,0)
0.6997 0.8181
(0,0)
0.0872 0.9514
0.4905 0.8220 ( 1000 , 1000)
(0,0)
0.9548 0.9411
TO CONVERGE***
(0,0)
(0,0)
0.4656 0.6823
(0,0)
0.0849 0.9217
(0,0)
0.5
1.0

set to unity

0.45

conve~gence

fo~

all

~uns

except

222
02xo/02~

4 and

e2~o2x%2~

4 but not L(1-L)o2

xo/02~

<

--- Violates all three concavity conditions on F(a)

REFERENCES
Basar

T.(1987a), "Solutions
control

problems

to a class of nonstandard stochastic

with

active

learning",

submitted

for

publication

T.

Basar

(1987b),

algorithms

"Relaxation
for

equilibria",

on-line

techniques
computation

Proceedings of

the IEEE.

and
of

asynchronous
noncooperative

Conf. on

Decision

and Control, Los Angeles, California.

Basar

T. and

G.J. Olsder

(1982), Dynamic

Non-Cooperative

Game

Theory, Academic Press.

Basar

T. and

H. Salmon

information

(1987), "Credibility

and the

value

of

transmission in a model of monetary policy and

inflation", submitted for publication.

Cohen

D. and P. Hichel (1984),

by

a time

"How should control theory be used

consistent government

?", CEPREMAP

discussion

paper.
Cukierman

A. and

A. Heltzer

credibility
information",

(1986),

and inflation

"A

theory

of

ambiguity,

under discretion and asymmetric

Econometrica, vol. 54, n05, September,

1009-

1128.
Kydland

F. and E. Prescott (1977),

the

inconsistency of

"Rules rather than discretion,

optimal plans", Journal of Political

Economy, n03.
Hiller

H. and H. Salmon (1985),

inconsistency

of

optimal

"Policy coordination and the time

policy

in

Economic Journal, supplement, 124-135.

an

open

economy",

223
Salmon

H.
(19861,
"Rational Consumption Behaviour", Working
Paper, Department of Economics, University of Warwick.

Whitman

C.
(19861,
"Analytical policy design under
expectations", Econometrica, vol. 54, n"6.

rational

PUBLIC DEBT, INFLATION AND THE COORDINATION

OF FISCAL AND MONETARY POLICIES
Anne LAVIGNE - University of Paris X and C.E.R.E.P.I.
Philippe WAECHTER - University of Paris I (M.A.D. }(*)
and Banque BRED

INTRODUCTION
In

industrial

most

institutions, the
implement

countries

central bank

fiscal

and

distinct

two

and the government, design and

monetary

policies.

The

degree

of

independence between the central bank and the government varies

among

countries

depending

considerations. For
Federal Reserve

historical

instance the

Bank have

independence compared
France. But

on

institutional

Deutsche Bundesbank

a well

to the

and

establish~d

Bank of

and

~he

reputation

of

England and the Bank of

even these latter retain some autonomy in choosing

their operating procedures and instruments. Thus the fiction of

a single

policymaker underlying

the optimal

the traditional approaches to

coordination between

fiscal and monetary policies

should be abandoned.
The design
question

of

instruments

of monetary and fiscal policies is

coordination
than

of

alternative
of

question

l~ss

objectives

cooperation

between

a
and
two

authorities having their own objectives and preferences.

In both

monetary and

fiscal fields the financing of

the budget deficit can raise two kinds of conflicts between the
central bank and the Treasury (I)

(*) We wish to thank Marie-Claude ADAM. Antoine d'AUTUME and

Carlo CARRARO for helpful comments and suggestions on an
earlier draft of this paper. Errors are ours.

(I) Treasury,
fiscal authorities and government are used ~o
indicate the authority which implements fiscal policy. Central
bank and monetary authorities indicate the authority which
implements monetary policy.

2~

il Both

authorities recognize the merits of budget deficits in

stimulating economic activity. Nevertheless the central bank in

charge of

reducing inflation. refuses to monetize the deficit.

The Treasury

in charge of the public debt management wishes to

slow down the growth of public debt.

some creation

of monetary

minimizes the

nominal growth

thus alleviates
this view

base

If the Treasury can obtain

against

of the

its

liabilities

it

stock of public debt and

the real burden of public debt outstanding.

the method

In

used to finance the budget deficit gives

rise to conflicting objectives between the central bank and the

government. Until

the

situation prevailed

end

of

in some

exhibited a

despite the

potential crowding

in

favor

strong

Therefore the

of

the

authorities led

central

to a

this

aversion

conflicting

since

to

central

inflation.

And

out effects of fiscal deficits

were reluctant

bond issuance.

1970's

Western countries

banks there

the governments

the

to reduce

significantly their

conflict was eventually resoJved

bank

the

decreasing

preference

monetary

of

financing

both

of

the

fiscal deficit and to a correlative increase of bond financing.

iil Both the central bank and the government wish to reduce the
public debt

outstanding. As far as the government is concerned

the reduction

of the

public

debt

should

occur

through

an

ongoing monetization

of the public deficit. From the viewpoint

of the

only a decrease in the budget deficit can

central bank

achieve both

the objective

reducing the

public debt

of reducing

inflation and that of

outstanding. Since

the beginning of

the 1980's this kind of conflict has dominated the relationship

between central

banks and

United States.
analyse the

governments.

particularly

in

the

After Tabellini (1986bl. our purpose here is to

strategic interactions between monetary and fiscal

authorities in their attempts to reduce the public debt stock.

Using a

linear quadratic

dynamic

game

this

paper

proposes an analysis of the influence of the strategic behavior

of policymakers on the time path of public debt.
In a

first part.

we assume that the behavior of the

does not

interfere in the conflict between the

private sector
two authorities.

The evolution

of the public debt is compared

under

institutional

settings.

different

We

show

that

the

227

cooperation between
reduction of

the two

public debt

behavior of

authorities leads

outstanding.

the private

sector

is

In

to the fastest

a second

explicitly

part, the

modelled.

We

assume that the private sector anticipates inflation through an

adaptive mechanism.
authorities have

Taking this

behavior into

account,

both

to modify their strategies as a result of the

trade-off between inflation and public debt.

I. Monetary Authorities vs Fiscal Authorities

In this

neutral vis

scenario the private sector is assumed to be

vis

the strategic

interaction between monetary

and fiscal authorities. The central bank and the Treasury share
a common

goal:

outstanding.

the reduction

But

the

creation whereas
through a

the

budget

central

on the

the

of

nominal

wishes

wants

Thus
way

face the

the

bank

Treasury

deficit.

Treasury disagree
outstanding and

of

to

public

to

limit

boost

central

reducing

debt

the
bank

the

following dilemma

money
economy

and

the

public

debt

either

they

cooperate to slow down the growth of the public debt but forego
their own objectives; or they forswear cooperation in order to
achieve their

own objectives,

policymaker will
In doing

bear the

in the

hopes

that

the

other

burden of reducing the public debt.

so they may fail to reduce rapidly the growth rate of

the public

debt. Thus

monetary and
government

the time path of public debt depends on

fiscal objectives
and

on

the

of the

institutional

central bank
framework

and

the

determining

relation between the two.

Within

linear

quadratic

dynamic

game

la

Tabellini, we derive the sustainable equilibria associated with

various institutional
with

respect

to

the

settings. We then compare the equilibria

different

weights

assigned

to

the

objectives of both policymakers.

1.1. The Model
The government

budget constraint

motion of the public debt:

gives the

law

of

228

All variables
denotes the

interest payment
liabilities of
r- is

scaled

of

creation of

to

is the

is the

nominal
budget

income

deficit

d
net

of

monetary base issued against the

the Treasury; r is equal to (l+r-)/(l+g) where

the real

of growth

are

public debt

rate of interest after taxes and g is the rate

real

income.

monetary base

The

central

whereas the

bank

controls

budget deficit

is

the
in

control of the Treasury.

The objectives

of central

bank and

of the Treasury

are described by the following quadratic loss functions:

(2) L"

According to

(2) and

(3) both

policymakers wish to

minimize deviations of the public debt from zero. This behavior

is consistent
public debt
on the

with the

absence of

lump sum

taxes. A

larger

induces larger tax levies in order to pay interest

public debt.

In the absence of lump sum taxes taxation

introduces distorting
debt target

effects on

is normalized

the labor market. The public

to zero in the objective functions.

Assigning two different targets d F and d" leads to more inertia

in the

evolution of

the

public

debt

without

changing

the

results of the conflict between the two authorities.

In addition
growth of
zero is

the central bank wishes to stabilize the

the monetary
consistent with

inflation or

base. The

chosen target normalized to

objectives such

as

the

control

of

external balance target. The parameter T reflects

the structure of the central bank preferences and its degree of

independence vis

vis the government. When

central bank

is definitely

independent:

its monetary

target. Conversely

central bank

merely finances

when

tends to zero the

it cares only about

tends to infinity the

the budget deficit chosen by' the

229

fiscal

authorities.

The

parameter

represents

time-

preference for the present factor as regards the central bank.

The fiscal

authorities on

the other

hand

wish

to

deficits net of interest

minimize deviations
of budget
payments. The budget target reflects the macroeconomic stimulus
desired by the government. possibly dictated by some electoral
considerations. Here again the budgetary target is normalized
to zero

(2). The

assigned to

parameter r

indicates the

relative

weight

the debt target relative to the budget target. The

parameter ~ represents a time-preference-for-the-present factor

as regards the Treasury.
The formulation
seems rather
are usually

peculiar. since
expressed

economic policy.
relationship

of these
in

Thus we

between

loss functions

the preferences

terms
make

final

of
a

the

LF

of policymakers

ultimate

crucial

LM and
goals

assumption

macroeconomic

of
the

objectives

and

monetary and fiscal instruments is invariant through time.

1.2. Monetary and fiscal strategies without
precommitment
Let us assume a game with complete information. The
central bank knows the loss function of the fiscal authorities
and vice
versa.
In this hypothesis neither of the two
authorities can

commit itself

to a given economic policy. The

two authorities simultaneously choose the sequence of their

instruments (mt}t~T and (ft}t~T. This choice of instruments in
t=O

t=O

period t determines the level of the public debt in period t+1.

But this level in period t+l will influence the choice of
policy instruments in period t+l. Thus when determining mt . the
central

bank

takes

into

account

its

influence

on

ft

+ 1

Conversely fiscal authorities will choose f t knowing its impact

(2) With target not normalized to zero the time path of public
debt would be for the benchmark simulation
dt + 1
X d t + y(f - ffi) - z(d F - d M) where x is the solution of
the game in a closed loop Nash framework and y and z are
positive constant. Therefore x is independent of the targets f.
m. d F and d M.

230

on mt

+1'

optimal

response

strategy). Thus,
order to

fiscaf strategy)

strategy (resp

The monetary

the

to
we get

fiscal

a closed

strategy

(resp

is the
monetary

Equilibrium.

loop Nash

In

ease the interpretation we have reduced the game to a

two period game.

From

and

(2)

(3)

we

obtain

the

indirect

loss

functions of each authority:

We now set initial and terminal conditions

(6a) d 1

is given

(6b) oV M (d 3

6d 3

)/

Td 3

(6c) 6V F

(d 3

6d 3

)/

rd 3

By backward
strategies for
the append i x)

recursion we

both authorities

get the

Nash

02 + ar 2 (1+aT)

~ raT (---------------) d
x

(8)

closed loop

(see analytical resolution in

02 +

~r2(1+~r)

~rr(

{me,f e }
2

where

the

superscript

"c"

stands

for

closed

loop

Nash

equi 1 ibrium.
with 0

1 + aT +

~r

From these
public debt:

strategies we

get the

evolution of

the

231

The analysis

of the

closed

loop

Nash

equilibrium

yields several conclusions :

i)

The

existence

associated with
means that
rate of

of

stationary

the condition

equilibrium

that r

is less

is

usually

than one. This

the real rate of interest is inferior to the growth

real income. When monetary and fiscal policies are set

by two

independent authorities, the condition for a stationary

equilibrium is assured if :
(10)

1 +

<

ii) When

condition (10)

equilibrium
absence of
fiscal

+ aT

~r

holds,

strategies

are

precommitment on

authorities,

these

then

time

the

closed

consistent.

the part

of

Thus,

Nash

in

the

monetary

and

strategies

are

both

non-cooperative

loop

credible.
iii) When

the central bank is independent of the Treasury, the

burden of

the reduction

the Treasury.

The

of public debt is mainly sustained by

lower

the

T,

slower

the

rate

of

debt

reduction and the bigger the burden of adjustment placed on the

when the

Treasury. Thus
(T

0), the

the debt

central bank is completly independent

monetary target is fulfilled to the detriment of

reduction. Conversely,

give priority

to the

when the

deficit, the

fiscal

authorities

adjustment of

the debt is

sustained by the monetary authorities.

iv) When
for the
on the

the monetary
present (a
Treasury. As

reduction (3)
bank and

authorities have a high time preference

-+ 0

is higher

) the burden of debt reduction weighs

matter

of

today than

fact,

the

cost

tomorrow for

of

debt

the central

it lets Treasury ensure the debt reduction. When both

authorities have

high time

preferences for

(3) Which is a cost of adjustment.

the

present

the

232
reduction of
rate of

the public

interest is

case, both

debt will take place only if the real

lower than

authorities will

the real growth rate.

In this

achieve their own policy goals to

the detriment of the reduction of the public debt.

v) When

the initial stock of debt is nil, both authorities are

at their bliss point.

Thus when

condition (10)

cooperative behavior,
debt. We

will see

there is

now if

holds, and

even with non-

a progressive reduction of the

other institutional settings modify

the rate of debt reduction.

1.3. Cooperative Equilibrium
Monetary and

fiscal policies are now under the aegis

of one

authority only.

as the

parliament or

This single policymaker can be thought

some council whose members come from the

central bank and from the Treasury.

It will merge the loss functions of the two separated
authorities into a single function which yields:

1/2

with 6 > 0, p > 0 and superscript "s" stands for single.

The original
those of
for the

central

the Treasury
present for

bank

objectives

relative

to

are balanced by p. 6 is time preference

the single

authority

which

solves

the

following optimal control problem

T
(12) Min 1/2 [ ~ 6 t (ft + pm t
"'t,r t
t=l
s.t. d t

+ 1

= rd t

+ ft

)2

+ (r +pT)d 2
t

+ 6T

+ 1

]
(r + pT)d 2
T+1

- mt

Initial and terminal conditions are given by

(13a) d 1

is given as the initial stock of public debt

233
( 13b) P S = (r + Il T) d:3
:3
where p~ is the shadow cost of d:3.
:3
The solutions
the values

of this

of monetary

optimal control

and fiscal

problem give

policies run by the single

authority

~
where

I[-( ~-:_~_:_T-J
r_(

In the

[ - ::-:-:-:-)- ]

,6 d ,

+ (1/1l)]6

[1

[r + IlT]

r6 d 1 ,

cooperative equilibrium

the evolution of the

public debt is then:

( 16) (d S

d S ,d S )
:z
:3
Let us

now compare

the monetary and fiscal policies

run by

two decentralized authorities to those implemented by a

single

controller.

According

equation

to

(9)

and

(16),

coordination increases the rate of public debt reduction if the

following condition holds:

( 17)

(3r

+ eXT)1l

(1+1l) (r+IlT)

< 6

This condition
reduction is
with two

not always

authorities,

This result

implies that
higher with

i.e.

depends on

the rate of public debt

a single controller than

coordination does

the value

of Il

not always pay.

(the

weight

of

the

objectives of the central bank in the overall objectives of the

single policymaker)
present

of

the

aRd of
single

the

time

controller.

preference

When

it

for

the

neglects

the

234
objectives of

the central

always fulfilled
point.

(17) still

nil),

(~

(17)

condition

is

and the fiscal authorities are at their bliss

Conversely

objectives of

bank

when

the

the Treasury

holds and

controller

single

the monetary

ignores

to infinity),

(~tends

the

condition

authorities are

at

their

bliss point.
Figure
and fiscal

represents the static game between monetary

authorities. The

line mm (resp ff)

is the reaction

function of the monetary authorities (resp fiscal authorities).

These two

lines intersect at the Nash equilibrium point. Point

M (resp F)

is the bliss point of the monetary authorities (resp

fiscal authorities).
equilibria with

The dotted

curve represents

the value

respect to

neighborhood of

zero, the

neighborhood of

F. Conversely

of

cooperative

When

~.

cooperative equilibrium
when

is in

the

infinity, the

tends to

is in the

cooperative equilibrium is in the neighborhood of M.

Lastly when
the single

central bank, the Treasury and

both the

authority have

the same

time preference

for

the

present, condition

(17) is

without ambiguity,

the rate of public debt reduction is always

always verified.

Consequently and

higher when policies are implemented by a single controller.

1.4.

Monetary

and

fiscal

strategies

with

precommitment
Cooperative institutional
feasible. However
decentralized

even if

between

arrangments are not always

monetary and fiscal policies remain

two

distinct

authorities,

equilibrium than the closed loop Nash can be reached.

case when
moves. The

each authority
central bank

precommits itself
thus determines

better

It is the

on a

sequence of

its own

sequence of

moves with the knowledge of the moves done by the Treasury. The
sequence of

moves to

response to

the sequence

same conditions
obtained is

which it

commits itself

is the optimal

of moves chosen by the Treasury. The

are required for the Treasury. The equilibrium

thus an

open loop Nash equilibrium derived from a

two-period model. The Hamiltonians for each authorities are the

followings

235

figure.

236
H

( 18) H = a t /2

( 19) H r

H
[rd t
[m'" + Td"') + a t + 1 p
t+1
t
t

+ ft

- mt - d t + 1 )

V/2 [f'" + rd"') + j)t+1p t+1[rd t + f t - mt

t
t

where Pt (resp Pt)

- d +1)
t

is the shadow cost of the state variable for

the monetary authorities (resp fiscal authorities).

Initial and terminal conditions are given by
(20a) d 1 is given
(20b) pH
Td:3
:3
(20c)

pF
:3

rd:3

First order conditions yield the following strategies

(4 )

0 + r"'a

~aTr02

0 + r

r2(a2T+~2r)d1' -~rr

The superscript

"0"

j)

02 + rZ(a 2 T+j)zr)

in equations (21).

(22) and (23)

stands for open loop.

To

these

open

loop

equilibrium

strategies

is

associated the following time path of the public debt:

1'0

-------------------d 1

(23)

0 2 + r 2 (a Z T+j)"'r)

When monetary

and fiscal

-0-z--+--r-2--(-a-:-:-+-~-Z--r-)d)

authorities act in an open

(4) See appendix for the complete analytical solutions.

237

loop framework.

the rate of debt reduction is higher than

they act

closed loop

in a

framework. This

can be

when

seen from

equations (9) and (23). Without ambiguity we can conclude that

dO < de.
3

The

ambiguous and

the superiority

of one

be shown

in the

in order

to ease

equilibria.
ambiguity

with

comparison

strategies is

cooperative

equilibrium

equilibrium over the other. As it will

next part. we have made numerical simulations

the comparison

Table
the

the

we cannot conclude a priori as to

rate

allows
of

of the

us

debt

to

different

conclude

reduction

is

kinds

that

of

without

higher

in

the

cooperative framework than in the open loop.

1.5. Conclusion
When monetary and fiscal authorities are not bound by
a precommitment. the rate of public debt reduction is moderate.
Under

different

increased. This

institutional
is the

settings

this

both

monetary

case when

rate

can

and

be

fiscal

authorities have the same time preference for the present. When
a single

controller is

policies. the
draw

rate of

similar

when

regards its

vis

of both monetary and fiscal

debt reduction

conclusions

precommitment as
accelerated vis

in charge

is accelerated.

each

authority

own strategy.

the closed

loop but

We can

makes

The reduction

is

not vis

vis the

cooperative equilibrium.

II. Fiscal Authorities, Monetary Authorities and the

Private Sector
The strategic

behavior of

both authorities

and the

link between targets and instruments of the economic policy are

modified as

soon as

model. Private
money and

we introduce

sector is

by its

the private

characterized

anticipation behavior

model of

the economy.

It has

will not

directly interact

no

in the

sector in

the

demand

for

by its
in

one

simplified

strategic behavior. so
game

between

fiscal

it
and

238
monetary

Nevertheless

authorities.

behavior. pri.vate

sector put

strategic choices

of both

by

its

an additional

anticipation

constraint on the

authorities. This

will change

the

value of the equilibria we obtained in the first part.

2.1. The new strategic model
The economy is characterized by three equations
(24) Yt

with Xt

Pt - pt-l

(25) Yt

with x"

p" - pt-l

(26) x" - x"

t
t-l

(1

d)(Xt-l

All variables
level. y

denotes the

output at
current

its natural
rate

inflation. I

of

t-l

are in deviation from their stationary

difference between

real output and real

level. p the general price level. x the

inflation.

the nominal

x"

stock of

the

anticipated

rate

of

money. and g is the budget

deficit. h.d. and e are positive parameters.

Equation (24)
(25) an

is an

aggregate demand

Equation (26)

aggregate supply

function

function with a real balance effect.

describes an

adaptative expectations

behavior.

From (24) and (25) we have :

(27)

y~

= h [(x t
t-l = e( at

- Yt-l

(28) Yt - y

- Xt-l) - (x" t - x .. t - xt )

1 )]

where a is the rate of monetary growth.

From (24) and (26) we get
( 29) x" t

x" t

[( I-d) /h]Yt-l

With (28) and (29) in (27) we get :

(30)

Xt

and

[e/(h+e)]a t + [h/(h+e)]xt-l + [(l-d)/(h+e)]Yt_l

239

By assumption the creation of monetary base against

liabilities of the Treasury is equal to the growth of the stock
of money.
Since
aggregate
demand
depends
on
public
expenditures. then by (30) inflation in period t is positively
related to inflation in period t-1. to money creation between
t-1 and t. and to the level of budget deficit in period t-1.
Using the above notations (30) can be rewritten as :
(31) nt+l = ant + bf t + cmt
where n is the rate of inflation per output unit.
According to equation (31) the current choice of
monetary and fiscal instruments influence the motion of future
inflation. For each policymaker the relationship between its
instruments and its ultimate macroeconomic target is not
invariant to
changes in
the instrument
of the
other
policymaker. Thus the inflation rate is explicitely introduced
in the loss function of both monetary and fiscal authorities :

(32) LM = 1/2

[~ at (m 2

(33) LF = 1/2

[~ ~t(f2

t=l

t=l

+ an 2 + Td 2 ) + a T+ 1 (an 2

~n2
t

+ rd 2 ) +
t

T+l

~T+l(~n2

T+l

+ Td 2

T+l

)]

+ rd 2 )]

T+l

where a and ~ are the weight of inflation for the central bank
and the Treasury respectively.
We make the further assumption that the target
inflation of both authorities is zero. Including two different
targets n M and n F would have simply added inertia in the
solution of each equilibrium.
The loss functions (32) and (33) with the constraints
(1) and (31) define the new strategic model (5).
As above. the various equilibria of the dynamic game
are derived from a two-period model. However the analytical

(5) If the targets f. m. d F d M n F and n M are equal to zero at

their stationary level. the budget constraint is not modified
if it is measured in deviation from its stationary level.

240
solutions

are

more

comparison between

cumbersome

the

impeding

different

equilibria.

proceeded to a numerical simulation (6)

2.2. Closed

Loop Nash,

any

superficial
We

have

thus

in order to rank them.

Open Loop

Nash and

Cooperative

equilibria
The central bank chooses a closed loop rule for money
creation in

which each

inflation and

move depends

on the current states of

public debt taking as given the closed loop rule

for budget deficit. Therefore the central bank assumes that the
government will choose the optimal current state feedback rule.
The government

plays

equilibrium strategies
methods since
the closed

the

way.

are computed

The

closed

with dynamic

loop

Nash

programming

they are based on current state variables. Hence

loop Nash

(Analytical resolution
(AlB) and (Al9)

equilibrium
is given

is
by

dynamically
equations

consistent

(AlO),

(All),

in the appendix).

Consider now

the open

decentralized authorities
moves. But

same

the central

loop Nash

equilibrium.

Both

choose their respective sequences of

bank and

the government now precommit

themselves to make their moves without attempting to coordinate

their

strategies.

Hence

each

authority

precommits

to

the

sequence of moves which is the optimal response to the sequence

of moves

announced by

strategies are

its

computed by

(see analytical

opponent.
means of

resolution in

These

open

loop

optimal control

the appendix).

Nash

methods

With no binding

commitment the open loop strategies are generally not credible.

Finally when
monetary and

a single

fiscal policies

policymaker
it weights

implements

both

the original central

bank objective relative to those of the government. The dynamic

game between

two decentralized authorities degenerates into an

optimal control problem. The optimization problem solved by the

single controller then leads to a cooperative equilibrium.
For all
simulation.
~,a,

6,

the equilibria, we have computed a benchmark

In this benchmark simulation parameters

0,

~,

r,

T,

and r are equal to unity and parameters a, b, care

(6) The analytical solutions are described in the appendix.

241

0.6. 0.1

and 0.3

respectively. Each

strategy is

given as

function of state variables w. and d . We have change the value

of each

parameter in

within the

order

compare the various strategies

t~

same equilibrium concept. These simulations are set

out in Table I.
2.2.1. General properties
i)

The

relationship

monetary and

between

inherited

inflation

and

both

fiscal strategies is negative whatever the values

of the parameters are and for each kind of equilibrium. For the
monetary authority.

high inflation

involves a

contractionary

policy and for fiscal authority. high inflation involves a weak

monetization of

the public

debt. Hence the

fis~al

authorities

will have to come up with a sizeable budget surplus.

ii)

Whatever

the

value

equilibrium concept

of

the

is used.

parameters

the inherited

and

whichever

public debt has a

positive impact on monetary strategies and a negative impact on

fiscal strategies.
debt. the

Ceteris paribus.

the higher

the inherited

greater its monetization by the central bank and the

higher the budget surplus to be achieved.

iii) In

period

1.

authorities to
closed loop

generate a

is higher

public

debt

leads

than in the open loop Nash or in

In

contrast.

in the

in

period

the monetary
The

authorities plan

the

closed loop Nash equilibrium

authority in

reason
the

is

reduction

that
of

their strategy
in
the

closed
debt

of

debt

loop.

both

over

periods. The fiscal strategy (resp monetary strategy)

into

strategy) for
debt is

2.

the two other equilibria. Analogous conclusions can be

monetization.

takes

fiscal

budget surplus that is lower in the

equilibrium.

budget surplus
than in

existing

Nash equilibrium

the cooperative

drawn for

the

account

the

period 1

monetary

and 2.

clearly assessed

strategy

the

two

in period

(resp

fiscal

so the reduction of the public

all along

the two

periods.

In open

loop equilibrium each authority care only about the strategy of

the other

within the

same period.

So the

behavior

authority is less smooth than in the closed loop Nash.

of

each

242
Table ,
Comparative strategies in closed loop Nash equilibrium, in open loop Nash equilibrium and in cooperative equilibrium

benchmark

CLH
OLH
cooperation

CLH
OLH
cooperation

CLH
OLH
cooperation

I
I

lid,

1,1"'-,

l itr,

m,ld,

mid,

-0,372
-0,381
-0,505

-0,106

-0,17 2

-0,072

-0,100

-0,129

-0,059

0,332
0,342
0,307

0,085
0,078
0,059

-0,025
-0,025
-0,507
-0,534
-0,533

-0,504

cooperation

-0,580
-0,580
-0,507

ClH
OLH

-0,277
-0,290

cooperation

-0,5"04

CLH
OLH

l,Id,

ns

-0,187

I
I
I

-0,015

-0,015
-0,111

-0,110
-0,109

-0,072
-0,189
-0,188
-0,111

I
I
I

-0,071
-0,067
-0,072

-0,104

changing the weights of public debt

r = 0
-0,072
-0,030
0,559
0,171
-0,074
-0,030
0,559
0,177
ns
-0,095
0,176
0,010
1'= 2
-0,220
-0,080
0,242
0,051
-0,151
-0,008
0,255
0,048
-0,010
-0,108
0,361
0,040
T= 0
-0.240
-0,100
0,023
0,013
-0,162
-0,071
0,023
0,013
ns
-0,095
0,176
0,010
T= 2
-0,141
-0,059
0,490
0,090
-0,115
-0,050
0,444
0,089

I
I

-0,010

I
I
I

I
I

-0,108

In the Benchmark simulation, the values of the parameters f',

a = 0,6, b = 0,1 and c ;; 0,3.

x/Y j = .x/"Y j (see Appendix)

0,361

;,"',"t'

m ,,.,,.,

mi",

-0,248
-0,158
ns

-0,04'
-0,060
-0,142

-0,173

-0,012
-0,040
-0,151

I
I

-0,270

I
I

-0,053
-0,068

0,040

-0,121

0,015

-0,170

ns
-0,344
-0,200
0,015

-0,137

-0,075
-0,083
-0,151

-0,190

-0,031

-0,136

-0,051

ns

-0,137

J, -., 6, ~ and r are equal to 1. Moreover

Table'
ComparatiYe strategies in closed loop Nash equilibrium, In open loop Nash equilibrium and In cooperative equilibrium

, ,Jd,

lid,

1,1"-,

CLH
OLH

-0,372

cooperation

-0,106
-0,100
-O,1J87

-0,172

-0,505

CLH
OLH

-0,361
-0,369
-0,458

-0,100
-0,094
-0,080

'I',

m,Jd,

mid,

m,J....,

ml1fl

0,085
0,078
0,059

-0,248
-0,158
ns

-0,041
-0,060
-0,142

0,094
0,088
0,061

-0,033
-0,032
ns

-0,017
-0,016
-0,076

benchmark

cooperation

CLH
OLH
cooperation

CLH
OLH
cooperation

ClH
OLH
cooperation

-0,381

I
I

-0,382
-0,390
-0,554
-0,367
-0,376
-0,458

-0,377
-0,386

-0,554

I
I
I

-0,112
-0,106
-0,090
-0,104

-0,098
-0,080
-0,109
-0,103
-0,090

-0,072
0,332
-0,059
0,342
-0,104
0,307
ns
changing the weights 01 inflation
<if = 0
-0,050
-0,017
0,350
-0,047
-0,017
0,359
-0,056
0,361
ns
IS = 2
-0,274
-0,115
0,31 B
-0,092
-0,197
0,329
-0,145
ns
0,248
<p = 0
-0,132
-0,060
0,336
-0,089
-0,045
0,346
-0,056
ns
0,361
9= 2
-0,210
-0,084
0,329
-0,168
-0,071
0,339
ns
-0,145
0,248
-0,129

I
I
I

I
I
I

I
I

I
I
I

0,079

0,073
0,064

0,088
0,081
0,061
0,083
0,077
0,061

I
I
I

-0,427
-0,263
0,017
-0,223
-0,132
ns
-0,271
-0,182
ns

I
I
I

-0,052
-0,092
-0,200
-0,029
-0,048
-0,076
-0,052
-0,072
-0,200

243
Table ,

Comparative strategies in closed loop Nash equilibrium, in open loop Nash equilibrium and in cooperative equilibrium

l,td,

lid,

-0,372
-0,381
-0,505

-0,106
-0,100

'i"11',

1,1,1",

m,td,

mid,

m,;....,

ml....,

0,332
-0, 7 2
-0,059
0,342
-0,104
0,307
time preferences
fJ = 0,9
-0,069
0,349
-0,055
0,359

0,085
0,078
0,059

-0,248
-0,158
ns

-0,041

benchmark

CLN
OLN
cooperation

CLN
OLN

CLN
OLN

-0,087

-0,104
-0,098

-0,387
-0,396

-0,112
-0,106

cooperation

-0,492

-0,091

CLN
OLN

0 '590
1--0,590

-0,198
-0,198
-0,120

I
I

-0,158
-0,125

I
I
""=
I
I
U,9

-0,163
-0,123

-0,069
-0,056

0,308
0,317

I
I

0,092
0,085

0,084
0,078

I
I

-0,060
-0,142

-0,239
-0,155

-0,231
-0,147

-0,037
-0,058

-0,041

-0, 5 8

.)= 0,9

----0,470

cooperation

CLN
OLN

-0,344
-0,352

-0,172
-0,129
ns
changing

0 '687
1--0,688

-0,160
-0,159

ns

-0,025 0
-0,025
ns

~=

-0,094
T= 0
ns
ns
-0,050

0,303

0,313

the specialization of authorities

<S"=2,6=OT=0 r = 2
-0,451
-0,285

-0, 16 71
-0,138

0,066

0,102

0,047
0,047

0,024
0,024
1

ns

ns

-0,595
-0,358

-0,130

-O,OBl

-0,106
-0,134

Table'
Comparative strategies in closed loop Nash equilibrium, in open loop Nash equilibrium and in cooperative equilibrium

I,td,

lid,

I,"'",

-0,372
-0,381
-0,505

-0,106
-0,100
-0,087

-0,331
-0,33 8
-0,454

-0,089
-0,084
-0,073

-0,172
-0,072
0,332
-0,129
-0,059
0,342
ns
-0 104
307
changing the other parameters
r = 0,9
-0,154
-0,068
0,296
-0,128
-0,059
0,304
n.
-0,105
0,271

m,td,

Ii",

mid,

m,t"lf,

ml",

-0,248

-0,041
-0,060
-0 142

benchmark

CLN
OLN
cooperation

CLN
OLN
cooperation

l'

CLN
OLN
cooperation

CLN
OLN
cooperation

-0,414

-0,425
-0,555

-0,373
-0,382
-0,482

-0,125
-0,117
-0,102

-0,1 ()8
-0,101
-0,095

0,085
0,078

-0,185
-0,130
-0,010

-0,135
-0,090
-0,0'4

059

-0,158
ns

0,070
0,060
0,049

-0,227
-0,150
ns

-0,046
-0,060
-0,141

1,1

-O,a77
-0,059
-0,'02

b =
-0,061
-0,040
-0,067

0,370
0,382
0,344

0,326
0,337
0,324

O,! 01
0,093
0,070

0,083
0,076
0,051

-0,270
-0,157
ns

-0,227
-0,134
0,011

-0,035
-0,060
-0,143

-0,031
-0,049
-0,127

244
The cooperative

equilibrium is

reduction of

the stock

of public

reduction of

inflation in

characterized

debt in

period 2.

period 1

by

and by a

Thus the strategy of the

policymaker is to provide a high budget surplus in period 1 and

a tight monetary policy in period 2.
Let us
weight of

now compare the strategies when we modify the

public debt

and of

inflation in

the objectives of

both authorities.
2.2.2.

Changing

the

weights

of

public

debt

and

inflation
- Changing the weights of public debt (r,T)
A variation in the value of the debt parameters (r,T)
has a

more significant effect on the closed loop and open loop

Nash equilibrium strategies than on the cooperative equilibrium

strategies (7).
Strategies of
a variation

in its

own parameter

weight

of

the

public

fUnction, their

own strategies

strategies. For

fiscal

the debt,

lower money

creation. Conversely
the weight

debt

a variation in the
in

their

objective

change much more than monetary

authorities,

assigned to
the higher

than to

the other authority. Thus when fiscal authorities

parameter of
modify the

a given authority are more reactive to

the higher

the

higher

the budget

the

weight

surplus and the

for the monetary authorities

assigned to

the debt,

the greater the

monetary creation and the lower the budget surplus.

When

or T

are nil,

the closed loop Nash and open

loop Nash strategies are the same.

- Changing the weights of inflation

(0,

~)

(7) The reason is that in cooperative equilibrium the weight

assigned to the debt is (r + ~T). Thus a variation of r or of T
has only small effects on the value of that weight hence on the
strategy.

245

A variation
no significant
closed loop
is.

effects on

weight assigned to inflation has

the strategy

of debt

reduction in

Nash and in open loop Nash. whatever the authority

Nonetheless

strategies is
matter of

of the

the

variability

a varies

greater when

fact. a

of

monetary

than when

or

fiscal

varies. As a

a (weight of inflation in the

variation of

central bank objective function) has an immediate effect on the

monetization of
monetization.
be more
when the

the debt

In

weight of

has no

a cooperative

sensitive to

above, this

when

equilibrium the situation will

changing the
the debt

direct effect on the

weights of

was modified.

inflation than

As was

explained

is a result of the general strategy in cooperative

equilibrium, characterized

by a

tight monetary

policy in the

second period in order to fight inflation.

2.2.3. Changing the time preference
When one
preference for

of the

two authorities

has a

higher time

the present the cost of adjustment is sustained

by the other one. When the fiscal authorities has a higher time
preference for

the present

lower in

periods.

both

creation rises,
adjustment.

return

the central

Conversely

lower) the budget surplus is

(~is

In

the

bank bearing

when

amount

of

the burden

decreases.

money

monetary
of

the

creation

decreases and the adjustment occurs through the budget surplus.

In cooperative

equilibrium a greater time preference implies a

smoother evolution.
monetary and
Compared to
the money

When a

single policymaker implements both

fiscal policies
the benchmark

creation are

the adjustment occurs over time.

simulation. the

fiscal surplus and

lower in period 1 and higher in period

2.
2.3. Debt

evolution in

Open Loop

Closed Loop Nash Equilibrium

Nash Equilibrium and in the Cooperative

Equilibrium
The equilibrium
time of

the public

made simulations

strategies shape

debt. As

in order

the evolution over

in the preceding section we have

to rank

them by effectiveness. All

246

these simulations are compared with the benchmark simulation

{see
with the same parameters as in the preceding section.
Table 2}
2.3.1. General properties
i} For

all values of the parameters, the public debt reduction

is always
open loop

fastest in

the cooperative equilibrium, then in the

Nash equilibrium

and, finally,

in the

closed loop

Nash equilibrium.
ii} For

all the simulations the inherited public debt has more

influence on the time path of the debt than the initial

inflation. Conversely inherited inflation has a greater impact
on the evolution of inflation than the initial public debt.
2.3.2.

Changing

the

weights

of

public

debt

and

inflation
- Changing the weights of public debt {r,T}
The more
reduction the

both authorities

faster

the

stock

care
of

about

public

public

debt

debt

decreases.

However the time path of public debt is more sensitive to

variations in r {Treasury} than to variations in T {central
bank}. {see Figure 2a and 2b}
Whatever the equilibrium concept is :
i} When T is less than one,
the central bank pays more
attention to its target of inflation. So the monetization of
the

debt

is

low

authorities to
the Treasury

and

this

in

turn

contrains

have a tighter policy. When

neglects its

itself primarily

with its

substantial reduction

public

debt

debt

fiscal

occurs

the

fiscal

is less than one,

target

objectives of

of the

only

and

concerns

stimulus.
through

A
the

monetization behavior of the central bank.

Thus, the
low

stock of

than with a low r.

the public

debt is lower with a

247
Table 2
The comparative evolution of public debt in the open loop Nash
equilibrium, in the closed loop Nash equilibrium and in the
cooperative equilibrium
(

d 2 /d1

d 3!d1

d 2!W1

d 3!W1

(-----------:--------:---------:----------:----------)
(
benchmark
)
(
(
(

CLN
OLN

(cooperation

0,294
0,276
0,187

0,101
0,045
0,040

0,288
0,270
0,179

0,094
0,081
0,037

0,075
0,028
-0,016

0,044)
0,029)
0,021
)

(---------------:----------:----------:----------:----------)
(
changing the weights of inflation
)
(

cr=O)

(
(

CLN
OLN

(cooperation
(
(

-0,016
-0,015

If" ..

(
(

CLN
OLN

(cooperation

0,299
0,280
0,197

0,107.
0,101
0,042

(
(

4I

(
(

CLN
OLN

(cooperation
(
(
(
(

0,296
0,277
0,179

(cooperation

0,292
0,274
0,197

0,153
0,065
-0,024

)
0,091)
0,066)
0,029
)

0,090
0,043

ns

0,099
0,093
0,042

I )
s O )

0,106
0,098
0,037
41

CLN
OLN

ns

-0,017)
-0,016)
0,011)

=2

0,060)
0,045)
0,011)

)
0,060
0,013
-0,024

)
0,029)
0,014)
0,029
)
)

(---------------:---------:----------:----------:----------)
(
changing the weights of public debt
)
(
(
(

CLN
OLN

(cooperation
(
(
(
(

CLN
OLN

(cooperation

0,415
0,414
0,316

0,101
0,047
-0,023

r
0,223
0,211
0,134

0,056
0,053
0,021

(
(

CLN
OLN

(cooperation
(
(

(
(

CLN
OLN

(cooperation
(

0,396
0,395
0,316
0,228
0,214
0,134

0,193
0,193
0,108
0,061
0,058
0,021

T ..

0,103
0,043
-0,023

T -

)
0,080)
0,050)
0,033
)

)
2

0,057
0,020
-0,010

(
(

0,222
0,222
0,108

0,056
0,021
-0,010

0,027)
0,019)
0,015
)

)
0,076)
0,049)
0,033
)

)
)

0,029)
0,020)
0,015
)

248

'ftble 2

The comparative evolution of public debt in closed loop Nash

equilibrium, in open loop Nash equilibrium, and in cooperative
equilibrium

d 2 /d l

d 3/d l

d 2 /lTl

d 3/lTl

(-----------------:----------:----------:----------:----------)
(

benchmark

(
CLN
(
OLN
(cooperation

0,294
0,276
0,187

changing the time preferences

)
(3
0,9
)
0,306
0,109
0,080
0,048)
0,288
0,103
0,029
0,032)

0,101
0,045
0,040

0,075
0,028
-0,016

0,044)
0,029)
0,021)

(-----------------:----------:----------:----------:----------)
(

(
(

CLN
OLN

(
(

)
cc=O,9)

(
(

CLN
OLN

0,303
0,286

0,107
0,101

0,067
0,024

0,040)
0,026)
)

(
(

0,9
-0,014

(cooperation

0,204

0,047

0,402
0,402
0,308

0,204
0,204
0,116

0,021

)
)

(-----------------:----------:----------:----------:----------)
(
)
(

IY=

(
CLN
(
OLN
(cooperation

0,

c O )

-0,025
-0,025
-0,012

-0,030)
-0,030)
0,170)

specialization of the authorities

tf .. 2, 4> = 0,

(
(
(

CLN
OLN

0,265
0,264

0,080
0,079

T ..

0,

0,144
0,073

)
)

0,083)
0,069)

-----------------: --------: ----------: ----------: ---------- )

(
(
CLN
(
OLN
(cooperation
(
(
(
CLN
(
OLN
(cooperation
(

changing the other parameters

0,270
0,257
0,174
0,315
0,290
0,199

0,084
0,080
0,033

O,9)

0,039)
0,029)
0,020
)
)
r=1,1)
0,119
0,082
0,049)
0,111
0,026
0,029)
0,047
-0,017
0,020)

0,067
0,030
-0,016

(
CLN
(
OLN
(cooperation
(

0,299
0,280
0,193

0,108
0,101
0,017

0,092
0,044
-0,026

0,061)
0,046)
0,033
)
)

249

,.,
.'.,,

\,

0,35

.,

0,25

.'.,.,

.'., .,

.,

.'.,

0,2

.'..... ...

....

-'., .....

...... .....

......

.... .........

CL
OL

Figure 2a

0,35

0,25

CL
OL

0,2

Figure 2b

250

ii) When T is greater than unity the monetary authority gives

more weights to its debt target as compared to its inflation
target. The central bank forsakes its independence vis a vis
the government. The monetization of the debt will be higher
than in the benchmark simulation in order to accelerate the
rate of reduction of the debt. Thus this behavior will raise
inflation and lessen the process of debt reduction. When r is
greater than unity fiscal authorities achieve a larger budget
surplus, a proof of their willingness to bear alone the burden
of the reduction of the debt. Thus, the stock of debt is lower
and its rate of reduction is higher when r is high than when T
is high
Finally, cooperative strategies always lead to a
faster decrease in the public debt than in the Nash equilibria.
However the rate of debt reduction is a function of the degree
of cooperation and figure 3 shows that it is not monotonous.
When the weight (~) assigned to the original central bank
increases, the rate of debt reduction increases up to a maximum
where each authority has the same weight in the objective
function of each policymaker, and decreases thereafter.
- Changing the weight of inflation

(o,~)

When the weights of inflation vary in the objective

function of both authorities this does not affect the evolution
of the debt.
However, changes in the weights of inflation a and ~
do lead to contrasting ti~e paths as regards the public debt
\
(see Figure 4). When the central bank assigns an increasing
weight to
its inflation
target the monetary policy is
increasingly restrictive. The monetization of the public debt
decreases, and the public debt is reduced much less rapidly.
Conversely, when the Treasury assigns a higher weight to
inflation, the rate of debt reduction is higher because of two
effects: the budget surplus, on the one hand, and monetary
creation on the other hand both increase.
These conclusions are not significantly modified with
a change in private sector behavior. As a matter of fact a
variation in the value of the parameters in the equation of

251

~
c:

"
l-'

~ t.

I
I
.f II
t

.,.

I
I
.,.

<t

I
t. I

.f
.f

I
I

t, I

'"

i,

+ I

I
.,.

t I

. I

fI

f I
,I

!,

!I

\
~

.+

u:

'"

.f
.f

4-

bO

.j I

::s

"

"

""v....

:l-

,"
~

""- lC.,
.,.""",-

"'""""'-.

.,.

--

.,. -..,. -

~,

\.

,
" "

'''foo..

252

'0
\

I
I

,
I

I
I
I

-\-

..

-\ 0 ,
I
I
'I

..
\

-t

I'

\
.
\..

,oT

1 -\

I
I

\
-\

'

\.

-\-

~~~
~c

"0

r:

"

'V

__~~__--__---- ~----~----~O
\

00
N

'"
N

0,28

0,29

0,30

0,31

.-

-+- -+- .... -+- + -+ .... +

-+-

- .-- . 5

0,1

0,2
0,]
0,'1

+ -+- + -+- ....

0.5

0.6

0,7

0,8

b=c=O .. S

0.9

b=c=O .. 5
_____,.. a

(eLl

\OL

(OL)

(eL)

-f- -+- Ben.chmark

++ ..... + ~ .... + + ...

(OL)

(eLl

Ben.chmark

c=0 .. 9

_ c = 0 .. 9

. - ---_.-.-..---

-.

....-.-.~.

figure

. + . +. -+-. -+-. -+-.-+-. +.-+-. +. -+- .-+-. +.+ . -+-.+ .+.+ .-f-

.... +

. -.- ----.--- . .....- .

--

0,]2

0.33

d 2 /d 1

254
inflation (31)
will change
(31) the

will not

the level

stock of

of the

public debt

(31) -

higher c i s , the

higher the

be higher

inertia

of

see Figure

increase with

will always

evolution of

will always

whatever the

in equation

it

debt. When band c are equal in

public debt

benchmark simulation,
(parameter a

modify the rate of debt reduction,

in the

inflation

is

5). The stock of

the value

of

c.

The

weight of monetary policy in the

inflation, and

the lower

is the monetization of

the debt.
- The

Specialization of

= 2,

The specialization
fiscal authorities
the public

authorities (0

2,

0,

when

the

T = 0)

of authorities

are mainly

debt ( nil and

occurs

concerned with the reduction of

above one) while the central bank

is mainly responsible for the reduction of inflation

(T

nil and

o above one). This specialisation induces a higher rate of debt

reduction than

in the

benchmark simulation

but this

rate is

lower than when both authorities cooperate.

Lastly with

both

and

nil,

the

central

bank

reaches its bliss point. The fiscal authorities are thus forced
to achieve

a high

constraint. These
those described
since the

budget surplus

in order to meet the budget

preferences of the central bank are close to

by Sargent

and Wallace

monetary policy is constant.

the Treasury

was

possible for

it

necessary for

exogenous
to

achieve

and

in their

If the fiscal target of

positive

such

1981 paper,

it

surplus.

would
It

not

be

would

be

the central bank to monetize a part of the debt,

the consequences of which would lead to increasing inflation.

2.3.3. Variation in the other parameters
When monetary
time preference
the debt

and fiscal

for the present

reduction is

(~

authorities have a greater

and a decrease) the rate of

lower than in the benchmark simulation.

This is because the cost of debt reduction is higher today than

tomorrow.

255

2.4. Comparing

the evolution of the public debt with

or without the private sector

For the Nash equilibria. the reduction of the debt is
faster when
(see Table

the private
3 and

constraint that

sector is

Figure 6).

excluded from the analysis

This comes from the supplementary

the behavior of private sector introduces into

the model and into the objective function of both authorities.

These

conclusions

parameters but

are wrong

private sector

(8) the

policy

is

which

authorities. This

hold
when

for

value T

correctly

equals
0

large
zero.

range
Without

of
the

implies a tight monetary

anticipated

by

the

fiscal

necessitates a high budget surplus. With the

private sector. monetary policy has to fight inflation, so that

the rate

of debt

reduction increases for two reasons : on the

one hand.

because of

same size

with or

the high

budget surplus

without private

(which has the

sector) and.

on the other

hand because of the restrictive monetary policy.

In the
not so

cooperative equilibrium.

clear. When

authorities is

the balance

strongly marked

the conclusions are

between fiscal

and

monetary

in favour of the central bank.

then the presence of the private sector accelerates the rate of

debt reduction.

When

policymaker's behavior
and to

or
is to

reduce inflation

sector the

reduction of

are

superior

reduce debt

in the

to

the

in the first period

second. Without

the debt

unity,
the

private

is carried out over the two

periods.
Thus the behavior of the private sector significantly
modifies both

monetary

prevailing institutionnal

and

fiscal

strategies

settings. In

whatever

general this

the

behavior

slows down the reduction of the public debt stock.

This reduction is slower if :
i) both authorities act in a closed loop strategy
ii) both authorities neglect their debt objective
iii) monetary authorities have inflation as their main target;
iv) both

authorities exhibit

a strong time preference for the

(8) Without private sector means the behavior of the private

sector has not been explicitly modelled.

256
'!'able 3
The comparative evolution of public debt with and without
private sector

with)
private sector
)

without
private sector

--------------------: -----------------)
d 2 /d 1

d 3 /d 1

d 2 /d 1

d 3 /d 1

(-----------:----------:---------:----------:---------)
(
)
~

( benchmark

0,290

0,096

0,294

0,101)

0,400
0,222
0,400
0,222
0,301
0,301
0,267

0,200
0,055
0,200
0,055
0,103
0,103
0,080

0,415
0,223
0,396
0,228
0,306
0,303
0,271

0,222)
0,056)
0,193)
0,061)
0,109)
0,107)
0,084)

(r

0
( r .., 2
(T
0
(T
2
0,9
( ~ c
0,9
( r '" 0,9

(3

(------------:----------:---------:-------:-------)
(
)
~

0,272

0,091

0,276

)
0,095)

0,400
0,210
0,400
0,210
0,283
0,283
0,254

0,200
0,052
0,200
0,052
0,097
0,097
0,070

0,414
0,211
0,395
0,214
0,288
0,286
0,257

0,222)
0,053)
0,193)
0,058)
0,103)
0,101)
0,080)

0,172

0,034

0,187

0,040)

0,070
0,110
0,041
0,272
0,200
0,272
0,200
0,188
0,159

ns
0,013
ns
0,090
0,018
0,090
0,018
0,041
0,028

0,072
0,106
0,034
0,316
0,134
0,316
0,134
0,204
0,174

0,036)
0,013)
ns
)
0,108)
0,021)
0,108)
0,021)
0,047)
0,033)

( benchmark

( reO
(r
2
( TeO
(T
2

(3
(

0,9
= 0,9

( r '" 0,9

(------------:---------:----------:----------:----------)
(
S
)
)

( benchmark

( I" :.

0,1
( ~ '" 5
("
20
( r '" 0
(r
2
T = 0
T = 2
o
0,9
r '" 0,9

---------------------------------------------)

257

...

-'"

o '"
II ~

II

I
I

I:Q

-t

,:
t
:
..... ,,
+
'
.... :''
+

./.

,il

N
il

./

...

!'

I
I

'!'

t :

.I

..

...

...

..

. I

~ I

.. I
. I

'!I
.1

I!.
It
,

l(

...

\\

:.-

,,

/"
VI

.If\ 1/
/

"'!.
0

VI

"!.
0

Y.\

-\'

I
I

....

"'!.

,
,,
,

/'

,
,
t ,
....... ,,,
+ ,,
,
.... ,,
,
,,,
,
t

I-'

II

II

t-~J.o;.

I-'

.-l

'".

'.

"

'+

'"

.............

...

'+
'
' ............ + ..

VI

VI

.....

.....

c:::

258
present.
Conversely. the
policies by
the public

coordination of

monetary and fiscal

a single

policymaker accelerates the reduction of

debt. The

coordination between fiscal and monetary

authorities will
the central

be much

bank weigh

more effective

if the objectives of

most heavily in the overall objectives

of the single decisionmaker.

CONCLUSION
During the

1970s. most

combined expansionary
policy. While
limited the

fiscal policy

central banks
monetization of

adopted increasingly
the economy.

of the

developed countries

with restrictive monetary

whishing to reduce inflation have

the debt.

fiscal authorities have

growing budget deficits in order to boost

The perverse effects of this choice of policy mix

are visible today. For most of Western countries. the growth of

public debt

stock exceeds

reduction of
orthodoxy.

the public
However,

implemented by
common goal,

These
the

two

fiscal

the

and fiscal

policies

are

authorities. the central bank and

of

the

authorities

public

debt.

share
yet

they

to achieve it. The central bank wishes to

goal through

reach this

GNP. Hence

a monetary

and

decentralized

reduction

the way

of the

debt requires

monetary

two distinct

the Treasury.
disagree on

the growth

a reduction of budget deficits, while

the fiscal authorities seek it the same goal through increasing

debt monetization.
This strategic
authorities can

easily

theoretic framework.

conflict beween
be

analyzed

monetary and

within

fiscal

dynamic

game

We show with numerical simulations in this

paper that the cooperation and the coordination of monetary and

fiscal policies can improve their well-being and can accelerate
the rate of debt reduction.
When the
it

adds

macroeconomic

objective function
strategies

private sector

are

of the

is included in the analysis

constraint

and

target

policymakers. Monetary

significantly

modified

and

the

and

in

the

fiscal
adaptive

259

anticipatory behavior

of the

private sector slows the rate of

debt reduction.

APPENDIX

We present
closed loop
game in

Nash and

appendix the

The

analogous to

derivation of

the

open loop Nash equilibria of the dynamic

complete information

Government.
with a

in this

derivation

the derivation

single controller.

between the Central Bank and the

of

cooperative

equilibrium

is

of the open loop Nash equilibrium

When both

authorities neglect

the

influence of

private sector

behavior on their own strategies,

the relevant

equilibria are

computed with

o.~.a,b

and c equal

to zero.
1. The closed loop Nash equilibrium
The indirect

loss functions

of monetary

authorities are respectively:

(AI) yM(dt.u t ) = 1/2 Min(m 2 + Td 2 + ou 2
mt
t
t
t

and fiscal

1/2/ Min(f2 + rd 2 +
f

~U2

) +

the constraints

(1) and

1.2

~yF(dt+l.Ut+l).

t
subject to

1.2

(31), and the initial and

terminal conditions:
(A3a) d 1 U1 are given
(A3b) 6yM(d 3 .u 3 )/6d 3

Td 3

(A3c) 6YF(d 3 .u3)/6d 3

rd 3

(A3d) 6yM(d 3 .u 3 )/6u 3

ou 3

(A3e) 6YF(d 3 .u 3 )/6u 3

~U3

The closed
recursion.

loop

Nash

equilibrium

is

computed

by

backward

In period t, the first order conditions (FOCs) for a

260

minimum are
0

(A4) 6V M (d t ,rr t )/6m t

mt +a[(6V M (d t + i , rrt+i)/6dt+i,)(6dt+i/6mt)]
+ a[ (6V M (d t + i , rr t + i ) /6rr t + i ) (6rr t +i /6m t )]
(AS) 6V F (d t ,rr t )/6f t

ft +
+

~[(6VF(dt+i,rrt+i)/6dt+i)(6dt+i/6ft)]

[ (6V F (d t + 1, rr t+ ,) / 6rr t+ ,) (6rr t+ 1/ 6f t) ]

1.1. The solution in period 2

Using the FOes, we obtain
(A6) m",

a(Td 3

(A7) f",

acrr 3 )
+ brr3)

~(rd3

Given

the

inflation, we have

of

motion

of

public

f '"

= -

and

- aaac, M3
Fo

debt

(Mod", + M,f", + M",rr",)/M 3

(Fod", + F,f", + F",rr",)/F 3

(AS) m",
(A9)

laws

~rr,

F,

~(r-cb),

F",

= -

~ab,

F3

l+a(T+ac"'),
+

~(r+b"')

We express the closed loop Nash equilibrium strategies in terms

of state variables in period 2 :
(A10)

mC

. '"

(All) fC

'"

with
Ho

(FoM3+F,Mo)/(F3M3-F,M,)

H",

(F,M",+F",M3)/(F3M3-F,M,)

Go

(MoF3+M,Fo)/(F3M3-F,M,)

G",

(M,F",+M",F3)/(F3M3-F,M,)

261

Using these

strategies,

we

derive

the

motion

of

public debt and inflation:

(AI2) d

1.2. The solution in period 1.

Knowing VM(dz,rr z )

the FOGs

for a

minimum can be written:

(AI4) m1

Xod z + Xzrrz

(AIS) f1

Yod z + Yzrr z

with
Xo

a(T+aT(r+Ho)(r+Ho-G o ) - aoc( a+bHz)(bHo+cGo

Yo

-~(r+~r(r-Go)(r+Ho-Go)

Xz

a(aT(r+Ho)(Hz-G z ) - co - aoc(a+bHz)(a+bHz+cG z

Yz

-~(~r(r-Go)(Hz-Gz)

c~

~~b(a+cGz)(bHo+cGo

~~b(a+cGz)(a+bHz+cGz

Substituting (I) and (31)

into (AI4) and (A1S) yields

the strategies :

(AI7)f 1

= (P o d 1 + P 1 f 1 + P Z rr1)/P 3
= (Ood 1 + 01f1 + OZrr 1 )/03

with Po

(AI6) m1

01

rX O ,P 1

Yo-cYz,Oz

Xo+bXz,P z

aY Z ,03

We express

aX Z ,P3

I-Yo-bY z

the closed

of state variables in period I :

(AlB) mC
1

loop Nash strategies in terms

262
fe

(A19)

with

Ro
R2
So
S2

(OoP3+O.PO)/(03P3-0.P.)
(O.P2+02P3)/(03P3-0.P.)
(P003+p.OO)/(03P3-0.P.)
(P.02+P203)/(03P3-0.P.)
Using theses

strategies we finally get the motion of

public debt and inflation:

(A20) de
2

(A21)

Ue
2

2. The open loop Nash equilibrium

When both authorities enter into binding commitments,
the

open

loop

equilibrium

Nash

of two

simultaneous solution

strategies

optimal

control

derive
problems.

from
The

appropriate Hamiltonan for each authority is :

t,

t,

t,

+ at,+'[pH

t,
+ qH

(rdt,+ft,~mt,-dt,+l)

t+l

(A23) HF
t

~t,/2(f2

~u2+rd2}
t

(rdt,+ft,-mt,-dt,+,) + qF

~t,+l[pF
t+l

where pHt,

(aut,+bf t +cm t -ut,+l)]

t+l

pFt,)

(resp

variables which

(aut,+bft,+cmt,-u t +.)]

t+l

et

qHt,

the central

(resp

qFt,)

are

the

costate

bank (resp. the fiscal authority)

associates with public debt and inflation.

Hence the

FOGs for

a minimum

in period

written
mt

a(pH

t .... t

- cqH

t .... 1

may

be

263
(A2S) 6H F /6ft. = ft. +
t.

~(pF

t. ... 1

+ bqF
t. ... 1

(A26) 6H M/ 6d t.
t.

Tdt. + ocrpM
t. ... 1

pM
t.

(A27) 6H F /6dt.
t.

rdt. +

pF
t.

(A28) 6H M/ 6lT t.
t.

OlTt. + ocaqM

(A29) 6H F / 6lT t.
t.

lTt. +

~rpF

t. ... 1

=qM
t"'1

~aqF

t. ... 1

t.

qF
t.

with initial and terminal conditions

(A30a) d 1 , lT 1 are given
(A30b) pM
Td oo
3

(A30c) pF
3

(A30d) qM
3

(A30e) qF
3

rd oo
OlToo
lT oo

2.1. The solution in period 2

In period 2.

it is easy checked that the FOCs lead to

the same strategies as above

O
(A31) m,..

(A32) f O
,..
Hence
(A33) dO = (r+Ho_Go)d,.. + (H,..-G,..)lT,..
3

264

Knowing

mz

and

fz,

we

determine

the

costate

variables :
(A3S) pM
z

Uod z + Uz 1l z

(A36) pF
z

Wod z + Wz 1l z

(A37) qM
z

Vod z + Vz 1T z

(A38) qF
z

zod z + Zz1lz

with
Uo

T(l+ar(r+H o -G o

Wo

r(l+~r(r+Ho-Go

Vo

o(aa(bHo+cG o

Zo

(~a(bHo+cGo

Uz

T(ar(Hz-G z ) )

Wz

r(~r(H2-GZ

Vz

o(l+aa(a+bH z +cG z

Zz

(l+~a(a+bHz+cGz

2.2. The solution in period 1

Using the
terms of d. and
(A39)

mt

(A40) f

FOGs (A24)

and (A2S), we get m. and f.

11.

(Kodt+Ktft+Kz1T.)/K3
(Lodt+Ltmt+Lz1Tt)/L3

with Ko

ar(Uo-cV o ) ,K t

a(Uo+bUz-c(Vo+bV z ,

Kz

aa (Uz-cV z ) . K3

l+a(U o -cV o -c(U z -cV z ,

Lo

-~r(Wo+bZo)

Lz

-~a

,L t

~(Wo-cWz+b(Zo-cZz'

(Wz+bZ z ) ,L3

l+~(Wo+bZo+b(Wz+bZz

We finally get

in

265

(A42) f O

and
(A43) dO
2

REFERENCES
Alesina, A. and G. Tabellini (1986), "Rules and Discretion with
non Coordinated Monetary and Fiscal Policies" , Mimeo,
University of California
Anderson, T.M. and F. Schneider (1986), "Coordination of Fiscal
and

Monetary

Policy

under

Different

Institutional

Arrangements", Arbeitspapier Nr. 8602, Mimeo.

d'Autume,

A.

(1986),

"Les

anticipations

rationnelles

dans

l'analyse macroeconomique" , Revue Economique, n.2, mars,

pp. 243-283.
Blanchard,O.J., R. Dornbusch and R. Layard (1986), Restoring
Europes Prosperity, M.I.T. Press.
Blinder, A.

(1982), "Issues in the Coordination of Monetary and

Fiscal Policy", NBER Working Paper, N. 982.

Christensen, H.
- Some

(1986), "Monetary Policy and Policy Credibility

Recent Developments",

Working Paper n.

1986-12,

Aarhus Universitet, Danmark.

Cohen D.

and P.

Hichel(1984),

Precommitment. I

: An

"Toward a
Analysis of

Theory

of

Optimal

the Time-Consistent

Solutions in a Discrete Time Economy", CEPREMAR

266
Fisher,

S.

(1986),

"Time

Consistent

Monetary

and

Fiscal

Policies: a Survey", Mimeo, M.I.T.

(1976), "Decentralized

Kydland, F.

Optimization and
Economic and

the

Stabilization

Assignment

Policies

Problem",

Social Measurement,

Annals

of

vol. 5, n.2, pp. 249-

261
Kydland, F and E. Prescott (1977), "Rules rather than Di scret ion
the

Inconsistency

Pol itical Economy

of

Optimal

,vol.

Lucas. R.E. and N. Stockey (1983),

Policy

in

an

Economy

Oudiz.

G.

and

J.

Buiter et

of

Capital",

Journal

of

12, n.l, pp. 55-94

(1984) ,

Sachs

Coordination in

Journal

"Optimal Fiscal and Monetary

without

Monetary Economics, vol.

Plans",

12, n.1, pp. 473-491

"International

R.C. Marston,

Policy

Models",

in W.H.

International Economic

Policy

Dynamic Macroeconomic

Coordination, Cambridge University Press.

Pindyck. R.

(1976),

Policy

"The

Cost of

Formulation",

Conflicting

Annals

of

Objectives

Economic

and

in

Social

Measurement, vol. 5, n.2, pp. 239-248

Sargent. T.J.and N. Wallace (1981), "Some Unpleasant Monetarist
Arithmetic",

Federal

Reserve

Quaterly Review, Fall, pp.

Tabellini.

G.

( 1986a) ,

Monetization of

"Central

Deficits;

Bank

of

Minneapolis

1-17
Bank

Reputation

the 1981

Italian

and

the

Monetary

Reform", Mimeo, University of California

Tabellini. G.

(1986b),

"Money

Debt and

Deficits in a Dynamic

Game", Journal of Economic Dynamics and Control, vol.10,

pp. 427-442
Turnovsky, S.

(1977).

"Macroeconomic Analysi;: CUP.

267
Turnovsky, S. and Y.A.

Brock

Optimal
Foresight

(1980),

Government

"Time

Consistency

Policies

Equilibrium",

Journal

Economics, vol. 13, pp. 183-212.

in

and

Perfect
of

Public

DECENTRALIZED MONETARY RULES IN A THREE-COUNTRY MODEL

AND TIME SERIES EVIDENCE OF STRUCTURAL DEPENDENCE
Masanao AOKI*
University of California, 4731 Boelter Hall - Los Angeles
California 90024 - U.S.A.

1. Introduction
Coordination of actions of several policy makers is
fraught with difficulties of theoretical and practical nature.
To understand the difficulties,
it helps to note that policy
coordination problems are in essence control problems with
multi decision-makers
under imperfect
and
non-identical
information patterns. Since policy makers in several countries
must act
on information
sets which are not identical,
incomplete and imperfect, pooling or exchange of information
may be desirable, were it not for the cost and delays in
gathering, exchanging and processing information, or incentives
for transmitting false information. Decentralized controls can
sidestep some of the costs and delays involved in information
exchange since they are control actions based mostly on locally
available information, possibly supplemented by information
supplied from outside on a few items. (1)
Decentralized coordination
schemes for
a small
country are easier to implement since policy decisions of small
open economies can be based on information set containing
mostly locally available variables and some "key" international
variables which define the international economic environment
which the small open economy faces, and thus they avoid many of
* I thank the participants in the International Macroeconomics
Workshop at the Center for Economic Policy Research, London,
and Marcus Miller in particular, for their discussions and
comments on an earlier version.
(1) Seminal
attempts at
the coordination
question
in
international economiCS via decentralized control are Aoki
(1976) based on Aoki (1972). Alternative approaches to policy
coordination via non-interacting control are discussed in Aoki
(1974). A connection with Mundell's assigment problem and its
dynamic generalization is discussed in Aoki (1976) as well.

270

the difficulties

associated

with

implementation

aspects

of

policy coordination.

Each small open economy is concerned only

with its

objectives and control questions, because

it can

own policy

take the global economic environment facing it as given

and no other spill-over effects of other countrys' actions need

be taken

into account.

small open

economies policy

economies but
effects of
In such

In economics which can't be treated as

also affect

actions not only affect their own

other economies.Spill-over or cross

policy actions on other countries can't be ignored.

"large" economies, decentralized schemes are feasible,

however, only

if decomposition

interactions, and

of policy

external shocks

objectives, dynamic

are realized

in the

sense

described in this paper.

This paper

presents a framework for analyzing policy

rules for large economies in which requisite decompositions are

possible.

To

demonstrate

decentralized policy

simply

schemes, the

the

possibility.

paper adopts

the

of

average-

difference coordinate system of Aoki (1981) in analyzing spillover policy

effects in

a benchmark

model composed

of

three

countries with similar economic characteristics.

Decomposition of
Section 2,

followed by

section lays

out the

policy objectives
that of

is

dynamics in

framework for

examined

Section 3.

analyzing

deviations

in
This
of

economic variables from some "reference" time paths in terms of

the weighted average deviation and deviations from the average.
Section 4

is a

three real

brief account

Theoretically, when

three

elasticities,

when three

for the

of time

series evidence on the

GNP time series of the USA, West Germany and Japan.

i.e.,

size of

economy with

the real

countries
~NP

have

identical

set

of

countries are identical except

(they

different scales),

are copies of a standard

then the

average

deviations

affect the differences but not conversely. We see that although

three

economic

decomposition of
period of

characteristics

are

dynamic interactions

1965 through

not

the

same,

roughly holds

for

this
the

1986. The paper concludes with Section

5.
2. Decomposition of the Performance Index
Let Yi

be the

real output of country i. To simplify

271

the presentation.

out the arguments for a world

composed of three countries. Extensions to an n-country world
is defined as Y1 +
is straightforward.
The world output Yw.
r 12 Y2 +
each Y

we carry

r13Y3 where
in a

constant

expressed at

the real exchange rate to express

common

currency.

(1982) constant

Deviation of
denoted by

rlJ is

output from

Y1

is

dollar. called

some

reference

the

US

GNP

the US real GNP.

or

trend

path

is

6Y , . Thus. deviation of the global output from some

trend (reference)

path is

given by

6Y 1 +

r 12 6Y 2 +

r 13 6Y 3 +

6r 12 Y2 + 6r 13 Y3 .
Define a lower case letter Y. as the relative change
of real GNP. 6Y , /Y , i = w. 1.2.3. Then
(1)
where

Yw = alYl+a2Y2+a3Y3+(6rI2/rI2)a2+(6r13/rI3)a3

are positive

shares of

the real

GNP. This

relation

can

be

rewritten as

where we assume that contribution from changes in real exchange

rates
3
yw=~

'-1

are

smaller

a.Yl'

In

than

the

other

terms.

We

assume

that

examining time periods in which changes in the

real exchange rates are not negligibly small compared with

the

sum of the first three terms in (1). we must carry these in our
analysis. For
Later in
GNP of

simplicity. we

Section 4.

assume that they are negligible.

deviations of

the logarithms of the real

the US. West Germany and Japan. denoted respectively by

InYue.lnY g and InY J . from a common trend time path are regarded
as Yue.Ys and YJ. Note that 61nY.
3
Ya=~

i=l

( 3)

= 6Y,/Y,.

g.1.2.3. Define

aiY,' Define the difference or d-outputs by

yt =Yi - Yw '

=1,2.3.

272

Other d-vGriables

are analogously

defined. Then assuming that

y's are mean zero random variables. and noting that

y?=(yj+Yw)2=(Yw)2+2yw yj+(yj)2,
we see that
Thus. the

var Yi

=var yj + var Yw + 2E (yjyw),

expected value

country's output

of a

= 1,2,3.

weighted sum of the individual

variances is decomposed into a variance of

and the weighted sum of the variances of yd'

y~

S ,

(4 )
L?=l ai var Yi =var Yw + L:'l ai var yt
because the cross product term vanishes because of (2).

Eq.(4) shows that if minimization of the weighted sum

of

individual

objectives of

country's
the three

variances

output

is

the

policy

countries, then it can be achieved by

minimizing the

right side of (4) . The task of minimizing the

of (4) is decomposed into three individual country's

left side

objectives. minimization
of minimizing
(2). Eq.

as "one

of var yd i

the average

(4) may

plus the global objective

deviation subject to the constraint

be used to rephase the coordination question

country optimized var y. while the other two countries

optimize the
natural to

second term
assign the

in (4)

role of

subject to (2)(2).
controlling y.

It may be

to the country

with the largest weight, country 1, say. The reason for this is
explained below.
nature of

assignment

dynamic interactions.

country problem,
rewrite

Such an

Eq.

~31_1alydi=O,

well-known in

(4)

by

using

is

independent

of

the

To see the nature of this nth

the exchange
the

rate literature,

definitional

relation

as

L?=l ai var Yi =var Yw

+ a2[ 1 + a2/ a 2J var y~

+ a 3 [ 1 + a 3 / a 1J var Y ~ + 2 a 2a 3 / a 2 E (y ~ ~) .
The cross-product
reasons

a2a3/a1

term may

be

negligible

for

two

is small. or Yd 2 and yd 3 may be uncorrelated

(2) A special case of this

is where the weights are equal
across countries discussed by Fukuda and Hamada. (1986). The
author thanks them for making available a preliminary draft of
their paper.

273
or nearly

so,

countries 2

unless

there

(As we

and 3.

is

global

will show

shock

common

to

later, country specific

shocks often produce yd 2 and yd 3 which are uncorrelated). As an

then a 2 a 3 /al

Even when the last term is retained, the

above suggests
the largest
and

.046.
a

decompositi~n

country in

countries

var yd 3

a3

,2 and a

that a2

.15.

illustration, suppose

.65.

of policy coordination in which

the world, country I, controls var y_,

and 3

individually

minimize var yd 2 and

possibly separately, or taking E(yd 2y d 3 } into account

and coordinate the policies between countries 2 and 3.

If the

dynamics are

a-variables affect
shall show

recursive,

i.e., if some of the

some d-variables, but not conversely, as we

later, then one can reformulate the minimization of

the weighted sum of the d-varia bles as f<..Ilows ;

With the

above assignment,

the minimization problem

is stated as

Min E (~ajY?)

dj
where

'

i=1,2,3

=Min
dw

The conditional
if the

[varyw

+ Min

J]

dj
i=1,2

variance becomes the same as var yd i

a- and d-dynamics are decoupled

d-disturbances are

uncorrelated.

coordination assumes
countries 2

and 3

the form
take Yw

In

and a-distu:rbances and

this

form,

of Stackelberg

as given

the

game,

and jointly

in

policy
which

minimizes J

while country 1 set Yw to minimize the expression in the square

bracket.
More generally in an n-country world model, we have

f5,

II

L aj

j=1

One of
control yd n
conflicts in

yl= o.
the n-countries,

but take

anyd n

coordination of

11-1

country

-.L ajyf
,=1

say,

must

not

as given,

if

policy actions is to be avoided.

Alternatively, country n may assume the role of controlling yB.

274

3. Decomposition of dynamics
The decomposition

or assignment of policy objectives

into countries suggested above is feasible because the dynamics

for the

a-variables and those for the d-variables of the three

countries turn out to be recursive under certain conditions. We

examine the

benchmark cases

and d-variables
in the

sense

the dynamics for the a-

are completely separated, or at most recursive

that

In

conversely.

in which

a-variables

these

benchmark

affect

d-variables

cases,

but

decentralization

not
of

policy rules is entirely feasible. Dynamics for the a-variables

are shown

to be

several models
have the

decoupled from
by Aoki

same set

those for the d-variables for

(1981, 1982, 1983) if all the countries

of elasticities and of equal size. When the

three countries possess the same set of elasticities but are of

unequal

size,

recursive:

the

the

paper

shows

a-dynamics

that

affect

dynamics

the

the

d-dynamics

become
but

not

conversely (3).
Here to
of three

exhibit the dynamic interdependent structure

simply, we abstract from the current account

coun~ries

dynamics and exclude bonds as assets. The dynamics of the model

of this

paper are due to inflationary expectations that appear

in real

interest rates

and"due

to

sluggish

adjustments

of

output prices. The model is related to that Obstfeld and Rogoff

(1984) and

is a

complicate the
as wealth

(1986a, b). We can

models by adding other sources of dynamics such

changes and

decoupled a-

current accounts without destroying the

and d-dynamics.

regarded merely
the

simplified version of Aoki

as a

(recursive)

decentralization of

The models

used here

should be

vehicle of illustration in demonstrating

dynamic

structure

in

supporting

policy coordination task. Each equation in

the model can be improved in various ways or replaced by better

specified equations.

(3) Models
in which countries are alike in the sense of
possessing the same values for the structural parameters
(elasticities) but of unequal size are analyzed in Laffargue
(1985).

275
Model Dynamics
The model

of this

paper is

specified in log-linear

form by the next 5 equations

(7)

sj/c (t

+ lIt)-sjk

(t)=ij (t)-ik (t)

and
( 10)

Pj (t+l)

=Pj (t) + OYjl + Vjl'

Time index
which is

omitted if

standard LM
Its price
in (9)

is PJ.

may be
from

although the
this

case

under the
the j

of each

Eq.

country price

the expected

Sii

is done in Aoki

country consume

procedure for

since it
the a-d

(7)

(6)

is the

goods. We

complicates

algebras,

interest parity

unit of

0 for

1986a. b)

all three

capital mobility

of the

(1985.

framework still works for

is the

expected value
is a

argument

Eq.

Using a price index rather than PJ in (6) and

assumed perfect

(8)

or

is likely.

j specialized in producing good j.

this refinement

logarithm. Thus

as subscript

no ambiguity

preferable as

as well.

denotes the
value.

is denoted

curve. Country

since residents
refrain

j, k == 1,2,3.

where SJk denotes

country k's

1.2.3. The

formed at

of

relation

currency in

argument
the

next

t+ilt
period

demand equation for good j. Demand depends on

i J - TI j and the real exchange

(9
)
+
defines the expected inflation in
rate Pk - PJ
SJk' Eq.
country j. (For simplicity we use p rather than a price index).
Demand

real interest rate.

shocks are denoted by

output good

price adjusts

11j .

slowly.

Finally.
i. e ..

( 10)

indicates that

6 is sma 11.

Here the

term mJ+llt - mJ could be added to the right side. analogous to

the way done in Miller (1982) or Buiter (1986).

276
3.1. a-Hodel
imply that the a-variables are governed
by a

set of

appears.

In

equations in
taking the

variables average
details.

Note

The a-model

out

no

exchange

weighted average
to

also that

become

out in

model
Aoki

rate

variable

of (8) exchange rate

zero.

the a-model

is therefore

originally pointed

which

Appendix

gives

the

contains no d-variables.
of

closed

(1981). The model

economy

as

is specified

by

or

where

0= <r 1(1- 21;)

weighted average of

and where ~ is a
~ij

as defined in the Appendix.

and

Reduced Form
Substitute (12)

into (11) to obtain the first of the

two equations in is and rrB as

where

Take the expectation of (14) to see that

277

where
are not

V~t=O

is assumed,

observed at

i.e., the average price disturbances

time t. Substitute the above into (13) to

obtain the second equation

We drop

a,

and

to

lighten

notation

in

this

section. Solving this together with (15) we obtain the expected

inflation rate as
( 19)

1t =

y (m - e) - 'YP + lCT\

where

and

( 20 )

=[ - (1 -

05)(m - e - P ) + 11211 ] I D

a-dynamics
Eq.

(14),

(17) and

(19)

yield

the

price

dynamic

equation as
(21)

PHI =Pt +1t+Vt

= y(m - e)t + (1- y) Pt -lCT\ + V"

and the

dynamic equation for the average deviation of the real

GNP from a reference path Yt =o(m-e-p)t+J.1111t]ID .. Eliminating

p from it we obtain the dynamics as

+ ("(lC I ~)11t~1 - [(1 - 2y)1C IS] llta

278

Monetary Rules
This section

examines how

the dynamics are modified

by feedback rules on m.
m =Ki

Solve (20) for i with m = Ki

i
where DK

'"

K(l-crcS)/D~.

for the

=[ (1 -

as) (p + e) + J..Li'lll DK

D + K( l-acS). i.e .. m = t(p +E) + K(jj:zID K )

and t=
Substitute this into (19) to obtain the expression

expected inflation rate under this monetary rule to be

equal to (dropping superscript a)

1t

=[ - ex (p + E) + ~ T\ ]B

where

From (17) and above. we deduce that

( 23 )

=- ex (p + e) + ~T\ .

The price dynamics are given by substituting this into (14) as

(24)

Pt+l

where

p=l-cxa

Eq.

(24)

= PPt + nt

and (23)

=( - a e + ~T\) B+ v.
imply that

under this

feedback rule.

the

dynamics (22) becomes

i.e .. y,a
Note that

is

ARMA (1.1).

Note that

<

<

1. unless crcS = O.

monetary and real demand shocks enter as innovations

279
and with opposite signs. The condition -l<p<l

is equivalent

to

0<

Os

DK

After we
a- and

=(cr~-2D)/2(1-a~).

<2, or K >Kc. Kc

develop the

d-variables, we

the "effective"

joint dynamic equation for the

show that

exchange rate

under this rule, ~~ affects

of country

j and

pf

through a

vector proportional to

The sign

of the

complicated function

, and cr~ > 0

in the

it is easy to see that the

In the other extreme case of cr;>O

derivative is positive.
the

is thus

of the system parameters. However,

external case of crf = 0

cr~=O

avary/aK

derivative

derivative

is

negative.

Thus,

by

but

continuity

arguments, we see that when the LM disturbances dominate the IS

disturbances, the
world interest

monetary policy should stabilize the average

rate with

disturbances dominate
feedback
are in

O.

the LM

not be

~hould

interest rate,

K >

the case

disturbances

used to

i.e., K

In

the

stabilize the

where

the

IS

interest rate
world

average

should be used. These conclusions

agreement with conclusions reached by Fukuda and Hamada

(1986) who trace it back to Poole (1970) for a closed economy.

m

It is

interesting to

random walk

for the

completely from
unequal in

observe that

{P:J

the rule

produces

series and decouples the a-dynamics

the d-dynamics

even when

the

countries

are

size. Under this rule we can see from (22) that the

dynamics for yB become

the effect

of

Comparing this

vIa

now

~~+1

with the

corresponding

appears from Yi4+1'

expression

under

.. -

the

280

cr;

interest rate feedback rule, we derive the inequality on

and

cr~ under which the latter is better.

This inequality
than cr~

confirms our

statement that when

cri

is larger

and crJ then the rule m = Ki is better than m

p rule.

For large K, the inequality somewhat simplifies to

because the effect of 0v become negligible as a 2 becomes small.

3.2. d-Hodel
Define the

d-variables as

corresponding a-variables, e.g. ,yd j

etc. From
are all

the deviations

Yj - ya and pd J

from

the

pj _ pB

the definition, the weighted sums of the d-variables

zero, e.g.,

~jajydj

O. Substracting (11) from (6) we

obtain

Define the exchange rate of country j by averaging over Sjk as

Then (7) becomes

The difference
The result is

of (8)

and (12)

is worked out in the Appendix.

281

d +s ) - -I;yd. + -9 .y a +".d
,. -1t.d) + d (
_po
Y)d =-(}" (.d
))
)).))
'I)

or
(28)
where the

9j = ~ ( 1-3aj) / aj ,

Append i x der i ves

division by

and -

denotes

The rest of the model equations is straight

(l+~).

forwardly obtained:

(9) and (13) yield

1tf, =pf (t + lit) - pf (t),

(29)

and (10) and (14) produce

Pj(t+l)=Pj(t)+8yj(t)+vl-

(30)

Note that
are decoupled

of country
k,k

the d-model

for the

individual countries

from each other. The dynamics of the d-variables

j do not involve the d-variables of other countries

j. The only dynamic interactions come through correlated

disturbances. When
the a-dynamics

3 countries

do not

are about

affect the

equal in size, then

9j

d-dynamics, because

is

nearly zero for each j in (28). To give some idea, suppose that

~.

O. 1, a z

a3

.3

and a

.4,

then

9z

93

.03 and

.25, then the coefficients become 9 z =

If az
.1 and 9 1 = -.1. The d-variable model of country j may be

9 1 = - .05.

93

regarded then
as a

as a small open economy taking y& as given,

small open

interest

rate,

economy can
etc.,

as

be analyzed

exogenous.

The

just

by taking the world

role

of

monetary

reaction functions in a small open economy has been analyzed in

detai I in Aoki (1985).

Reduced Form
Substitute (28)

into (26), and next equate Udjt to 6

times (28) to obtain two equations in i d jt and Ud jt . Solve them
simultaneously to obtain

282

1 =(1 - 08)! D,

where

g =1

=J.l2! D,

-nd,

from (31) and (28)

d-dynamics
Let

From (27),

(30),

(31)

and (32)

the dynamic equations for the

exchange rates and the expected inflation rate of country j are

obtained as

where

=g ,
a22 = 1- <X;.

cp = (8Ill! J.l2)n

a 12

=BIll' D

Two facts are noteworthy about (33). The dynamics for country j
do not

depend on

the a-dynamics.

Zkt,k

j. and the d-dynamics do not affect

283
Monetary Rule
Eq.

(34)

makes

clear

that

so

long
of

md J

country

is

determined as

a function

of the

policy makers

of country

j need no information on d-variables

of the

d-variables

as

j,

other countries. Dynamic behavior of the d-variables of

country j

are jointly

affected by

mdJ and

rna through ye.

examine two decentralized rules.

m j d =K.dR
Ij
ule
From (31), the md

Kid determines i d to be

i d = (gp + nds) t (1 +Kl) + [n(eya +11) + Ie] t (1 + KI).

From (33) the dynamics are modified to become

- Yet (1 +KI)+ [c +KYn /(1 +K/)] (eya +11)

because

After some algebra we obtain

and

c + KYn t (1 + KI) =[

The dynamics are

~:1'

We

284

(35)

where

where

DK =D (1 +Kl)

and where
D 1 =<5(J..ll+K)dID K

D2 = 1- <5

[cr + (J..ll + K) ~ 15K

and

3.3. Dynamic Interaction

We can

now give

interactions between

P~t

the complete description of dynamic

and

Zjt.

Eq.

(21).

be written jointly as

(34)

where

[1-1
=

-X~2 A

-x83 0
va
I
nIa --

0
0

HG=H

0
Y

x83 0

+[ ;'c]
83 c

J..llTi,a / D

~]

(22) and (33) can

285

Eg.

(34)

shows

variables.

In

that

can

affect

only

the

j-country

this sense mdj and mdk,j :j= k are non-interacting

The monetary variable rna can, however, affect all the variables
if 6:j= 0.

(If

6 = 0, then PJ's are not controllable).

The monetary
sense that

rule for

it affects

the model

rna is

more important

in the

globally. Monetary rules for

md J , however,

only affect the j-country variables. With

feedback rule rn a t -- Ki a t , the dynamic matrix changes into

the

where

and the
vector

since x

disturbance

~c/D,

~~

impinge

and ZZt through the

where ~ is introduced in (23), and c in (33).

Effects of disturbance

Under the

p-t,F

rule mat

complete dynamic

on pa,Z't

is distributed through the vector

is modified

decoupl ing obtains ~

through the distribution vector

to

diag(l,A,A}

and

and TJtG affect the model

286

respectively
Dynamic Structure
Dynamic multiplier matrices summarize all information
regarding how
exogenous

exogenous changes in the monetary instruments or

disturbances

informative. Eq.

(22),

transfer function
outputs to

~Q

y,;i

and monetary

can be

1,2,3

and

are

more

and (34) are used to computed the

(dynamic multiplier)

the real

response to

affect
(32)

matrix connecting

shocks.

obtained from

For

example,

the
the

"

(22) and (34). Using to

denote transform we see that

yQ =

rl-

(0"1 D)(1

where

The impact
shift in

multiplier is
T1Q

(~l/D).

Effects

decay to zero at the rate

of

-y(1-y)'

once-for-all
, t

0,1 .... ;

when 11-1 1 <1. Similarly, using (32) we obti!in

y' = [ (0,1 v1(1J -n' [ :::]

~ll

D + OLld I

Dl9] ~.

where

This becomes

where

The transfer

functions in response tOTlf and T1f<:an be similarly

287
obtained. Putting these expressions together. we obtain

Responses to monetary shocks can be similarly summarized by the

transfer unctions as
(37)

y"a

={[

00]
[o][lOOl}
g4 + g:zg3 e

g3

g4

where

and

In both

cases we

variables while
shows that

note that

11- and

m-

e"

~d and

md,

affects yB

affects only
and yd' S

Eq.

yd
(37)

off-diagonal terms in the transfer function arise

only when md j responds to some of the d-variables of the other
country i,j
or pR.
More specifically only when md 2 has a

component responding
transfer function

to

are relevant

the (2,3) element in the

either to pB or

transfer function

transfer function

or to pB

has a non-zero entry. Similarly md 3 must have

components responding
of the

Z3,

to

have

Z2

for the (3,2) element

non-zero

element

of

the

to have non-zero entries. These observations

to our

empirical findings to be discussed in the

next section.
These transfer

functions can

in the original coordination system.

be equivalently stated

In the original variables,

288
dynamic structure
lost. This
With e

so clearly

is one
0

of the

exhibited in

of equal size), the dynamics are

the transfer functions along the main diagonals.

Unequal country
shocks and

(37) are

advantages using the a-d framework.

(three countries

summarized by

(36) and

sizes add

glgze or

monetary shocks,

gzg39 in the case of real

respectively. When

6 is

zero in

(10), the functions g's become

and

have unstable
function
g4 thus
gz and
characteristic of rational expectation models.

The

roots,

4. Empirical Evidence of Interdependence

We used a total of 84 data points covering the period
of 1965.2

- 1986.2

dollars), West

Germany and

constant deutsche
models for

of the

mark and

random trends

real GNP of the USA (1982 constant

Japan
yen

(the

latter

respectively),

two
to

at

1980

construct

and deviations from the trend of the

three real GNP series.

Let Yu . ,
Yg and YJ be the level values of the three
real GNP series. Denoting the real exchange rates of the German
and Japanese

currency with

respect to the US dollar by rg and

rJ' the world real GNP is defined by Yw = Yu s + rgYg + rJY J .

Denoting deviation from "trend" by 6,
e.g., 6Y w = Yw - yOw
where "0" indicates (random) trend

Percentage or relative deviation is

289
where Yw

6Y w/Y"",yu

analogously defined.

= 6Y u ./Y u ... The :variables

Ys and Yj are

The parameters su, Ss and Sj indicate the

shares, Su = Yu/Yw,ss = rsYs/Y w and Sj = rjYj/Y w. To the first

order of approximation the shares are evaluated along the trend
path and

regarded as

deviations

of

the

fixed. One
logarithms

can
of

interpret
the

Yu

etc.

corresponding

as

level

variables, Ys = 6lnY s ' etc. The above implies that

+ as

since au
dropped as

+ aj

1 by definition. The last two terms are

being small

construct models

in two

compared

with

steps. First,

the

first

we build

three.
a model

We
for

"trend" component

y,,]

In
[ In Yg
In y.J
where

=C't,j + Y,
,

represent

trend dynamics

a "trend"

reference time

is of the form

path. The random

't'+1 =A't,

+ short-run

deviations where Ais about .94-.95. The residual vector further

modeled by a 2-dimensional dynamics

Z'+1 FZt + Ge,

Yt =Hzt +et .
The residual vector n t evolves with time by

Yt+l

=HZt+l + et+l
=H(Fzt +Get)+et+i
= HFH+(yt - e,) + HGet + et+i

i.e. ,the dynamic transition matrix is given by HFH+=HF{H'H)-lH'

The vector Ytis transformed into the a-u framework by

290

where

In the

then, SHFH+S- l

a-d framework

is the dynamic matrix

which corresponds to those in Section 3. With au

and aJ
( 38 )

.64,a s

.14

.23 we obtained

SHFH+S-l =

With the

.788 - .077 - .060J

.191 -.261
-.289 - .213
.705

- .796

.63,

shares au

.12 and

as

.25 the matrix

aJ

becomes
(39) [

- .063l

.792 -.071

-.800
.189 -.265
-.294 -.217
.702

The (1,2)

and (1,3)

effects of

elements

yd's ori

compared with

y- and

the other

in

both

not equal

cases
to

represent

zero.

but

the

"small"

off-diagonal entries representing the

effects of a-variables on d-variables. The non-zero entries are

due to m- responding to d-variables or due to the non-identical
elasticities of the three countries.
As discussed

in connection

with (37).

the non-zero

entries in

the (2,3) and (3,2) elements in

(38) and
(39)
d
indicate that m s and
mdJ both responded not only to non d-

variables but

to the

countries. For
modeling

average and the d-variables of the other

further details

method

used

in

on

this

state
paper

space
are

time

found

series
in

Aoki

(1987a,b,c).
There are
weighted average
not be

several reasons for dynamic models for the

of the

exactly decoupled

real GNP

of the three countries will

from the difference dynamics. First.

291

the relative

sizes of

Second, elasticities
Germany and

Japan are

the three
of the

countries are

not the same.

model consisting of the USA, West

not expected to be the same. Third. the

weights used to form the a-variables are actually changing with

time because

the GNP

rates. However.
the idealized

it is

of these

countries

grew

at

different

interesting to see how close or how far

theoretical model

is rather

close to empirical

models in having a small off-diagonal submatrix.

5. Concluding Remarks
This

paper

provides

informationally decentralized
model by
economy

applying the
macroeconomic

demonstrates a
well as

framework

policy rules

for

examining

in a multi-country

average-difference formulation
models.

In

particular.

the

to open
paper

natural decomposition of objective functions as

interaction dynamics. when the economies have the same

set of elasticities.
The quarterly
three

country

real GNP

empirical

decomposition (decoupling)
dynamics is

data are

models.

then used to build

Although

no

perfect

into the average and the difference

observed empirically. empirically estimated models

are approximately so decomposed. and hence such a decomposition

retains broad

features of dynamic interactions among the three

countries.

REFERENCES

Aoki. H.

(1972) -

"On Feedback Stabilizability of Decentralized

Dynamic System". Automatica. 2. p.

163-173.

292

Aoki. H.

(1974) - "Noninteracting Control of Macroeconomic

Variables
Implications
on
Policy
Mix
Considerations", J. Econometrics, 2, p. 261-281.

Aoki. H.

(1976) - "On Decentralized Stabilization Policies and

Dynamic Assignment Problems", J.I.E., 6, p. 143-171.

Aoki. H.

(1981) - Dynamic Analysis of Open Economics, Academic

Press, New York.

Aoki. H.

(1982) - "Dynamic Distributive Effects of Money Stock

Changes in a Three-Country World Model under Flexible
Exchange Rate Regimes", Discussion Paper nOl13, Inst.
Soc. Econ. Research, Osalca University, October.

Aoki. H.

(1983) - "Dynamic Assessment of Financial, Monetary

and Real Disturbances in a Three-country World Under
Alternative Wage Policies", presented at the NBER
Conference on Open Economy Macroeconomics, Cambridge,
MA, August.

Aoki. H. (1985) - " Monetary reaction Functions in a Small Open

Economy" fIJ.E.D.C., 9, p. 1-24.

Aoki. H. (1986a) - " Dynamic Adjustment Behavior to Anticipated

Supply Shocks in a Two-country Model", Econ. J., 96,
p. 80-100.

293
Aoki, H.

(1986b) -

"Effects of Anticipated Real Supply Shocks

and Coordinated

Monetary Accomodation",

Econ.

Study

Quarterly, 37, p. 134-150.

Aoki, H.

(1987a)

State

Space

Modeling

of

Time

Series,

Springer Verlag.

Aoki, H.

(1987b) -

"How

to

Build

State

Space

Models

for

Nonstationary Time Series", Working Paper n0438, Dept.

of Economics, UCLA, March. Also to be presen:ed at the
6th International Conference on Mathematical Modeling,
St. Louis, Aug., 1987.

Aoki,

(1987c)

H.

"Time

Interdependence of
presented at

Series

Evidence

the USA,

the 1987

Far

of

Real

GNP

West Germany and Japan",

Eastern

Meeting

of

the

Econometric Society, Tokyo, October, 1987.

Buiter,

V.

(1986)

"Macroeconomic

Interdependent World

Policy

Economy:

Design

An Analysis

in

an

of Three

Contingencies", IMF Staff Paper.

Buiter, V.

and

H.H.

Hiller

Overshooting and

(1982)

the Output

"Real
Lost

of

Exchange

Rate

Bringing

Down

Inflation", Europ. Econ. Rev., 18, p. 85-123.

Fukuda, S.
of

and K. Hamada (1986) - " Towards the Implementation

Desirable

Rules

of

Monetary

Intervention", Mimeographed.

Coordination

and

294

Kitagawa, C. and W. Carsch (1984)

Space

Modeling

of

" A Smoothness Priors State

Time

Series

with

Trends

and

Seasonalities", JAJS. 79. p. 378-389.

APPENDIX

Derivation of the a-model

Let aJ

>

countries. Define

0 and IJa J = 1 be the relative size of three

the a-variables

as the weighted sums of the

individual variables. For example

where the

superscript a

denote the

a-variable. Obtaining the

weighted averages of (6).

(9) and (10) are straightforward. The

weighted average

left hand side of (7)

of the

is zero because

sJ
0 and IJaJs Jk ~ IJtkajSjk = 0 because
Similarly the right hand side is iA - in = O.
In (8),

we assume

that d Jk

second term of (8) becomes

r.j

aj

r.bj

djk (p" - Pj + Sj")

=d

r.j

aj r."~j a"(p,, -Pj +Sjk)

=d

r.j

aj r.bj a"(p,, -Pj)

=d

r.j

aj(r." a"p" - aj Pj) -

=dpa -a

=0.

r..J

-a

d r.j aj(r."~j a,,) Pj

-

aJ p.J -d P +d r.aJ p.J

=d

ak

SJk

for all

=-

SkJ.

j.

The

295

Since the

elasticity

S'J represents

the

effects

of

output

changes in country j

on that of country i.

(3) we

it is proportional to the relative country

assume that

size aJ/a"

in the third term of

i. e.,

Then
I A. 1 )

Ljaj L"~j Sj"Y" =SL.;aj L"~j (a"laj)Yk

=SLj Lbj a" y"

=SLj (yz -ajYj)
=~

(3yD _y D)

=2SyD

Noting IA.l)

take the

(8) and

(12).

d j k = dak for all

j. the

difference of

Then

+ Lbj Sj" y" + Tlj + <JW -it") -

=-

2S yD

- TlD

<J(if -Ttf) + Tl/ + L"~j dj"(P,, - Pj + Sjk)

+ L"~j Sj" y"

2~ yD.

Using the

posited relation

third term above is rewritten as

Lbj dj"(P,, -Pj +sj,,)=d L"~j a"(p,, -Pj +Sjk)

=d

IA.2)

[L"~j a"(p,, -Pj)+Sj]

=d(-Pj+s),
The

fourth

posited relation

term

~aj /aj for~jj.

becomes,

upon

substituting

the

296
E"j ~j" y" - 2~ ya

=E,,# ~j" (y" -

ya) + (Ebj ~j" - 2~) ya

=Ebj ~j" yt + (E"j

~j" -~) ya

=-~Yf+ej ya
where

= 0 is used. Note that 9 J is zero when aJ

i.e . when the three countries are of equal size.

where Iakyd k

1/3.

FISCAL EXPECTATIONS AND CURRENT ACCOUNT SURPLUS

OF THE MAIN OECD COUNTRIES
Eric GIRARDIN
University of Bordeaux I
Avenue Leon Duguit - 33604 PESSAC

For the
payments

of

OECD

disequilibria

and

countries. the
to 2

GDP in

of

last fifteen

years. the

countries
large

has

scale

current balance of

been

marked

variations.

current account.

Thus.

by

deep

for

most

the surplus of which amounted

GOP in 1973. went into deficit and reached 0.8

1985. Furthermore.

the divergence

of

in the direction of

disequilibria according to the country considered is of course,

the most striking phenomenon of the mid-eighties.
is a

striking contrast

States and

on the

between on

other hand,

the one

Indeed, there

hand, the

Germany and

United

Japan. The former

have a deficit in excess of 3

of their GNP whereas the latter

have a surplus in excess of 4

of their GNP.

These developments

have been

different ways.

Price-effects have

the

terms

guise

of

specifically to

the two

the dollar-shock(s).
approach to

as the

account developments.
between fiscal

trade

fluctuations.

referred to in
This

refers

oil shocks, the counter-oil shock and

A second

the balance

monetary impulses

the other

of

accounted for in three

often been

approach. based on the monetary

of payments.

lays

main explanatory

However, in

the

emphasis

factor of

recent years,

on

current

the contrast

expansion in the U.S. and fiscal restriction in

OECD countries (Chart 1) has been used more and more

often to explain external disequilibria. The "twin deficits" in

the

U.S.

(BOURGUINAT,

1987

MARRIS,

1987)

are

good

illustration of this analysis.

In this
equation over
seven main
France. the

paper we

shall estimate

the floating-dollar

a current

period (1973-1986)

account
for the

OECD countries: the United-States, Japan, Germany,

United Kingdom,

Italy and Canada. We shall try to

298
Graph n' I : Structural budget surplus as a deviation from average

o United States

.,

.J

-Japan

2.5
2
15
1
Po

..

....

.5
0
- .5

JI

-1

-1.5
-2
-2.5

-3

-3_~1
70

74

72

.,

-'

76

\~
78

80

84

82

86

88

OFrance

Germany

Q.

..

8....
0

0
-I

-2
-3

-4l70
6

7'2

76

74

78

80

D \taty

() United Kingdom

2
/

0
Q.

.
Q

....0

60

/I..

'~

-2

'0

-4J

-6

-10

~, "

-12
70

7'2

-8

86

o Canada

84

82

74

i-----'

'\.--

76

78

80

82

'a........ ~
84

Source; computed from PRICE and MULLER (1984) and OI.C.D. (1987a)

86

86

299
isolate

contribution

the

determination

of

contribution of

these

fiscal

of

external

the other

impulses

accounts

the

in

compared

to

the

two explanatory factors usually put

forward.
On a
shown that

theoretical level.

the relevant

budget deficit
VINALS,

impulse

instance:

of
is

empirical point

measured by

with conjonctural

using the

studies
the

have

anticipated

FRENKEL and RAZIN.

1986 ; and. for a synthesis, GIRARDIN,

From an

can be

fiscal

(see for

a number

1985 ;

1987).

of view, the fiscal

impulse

public sector surplus corrected

influences. Two methods are available to get

an anticipated

structural public surplus. First, we can assume

(see FELSTEIN,

1986} that the structural deficit for year t

the best

is

estimate of the deficit anticipated by agents in t-1.

However, such
to accept

an assumption is rather strong. since it amounts

the existence

acceptable predictor
that agents'

foresight.

To

get a more

of future public surpluses, we can assume

forecast use every information available today on

the determinants
a reaction

of perfect

of this future surplus. Then we must estimate

function of

fiscal policy

based on the objective-

variable of this policy.

To test

the influence

of future

anticipated fiscal

impulses on the current account of OECD countries, we shall use

both methods successively. The results of the first method will
be presented
for fiscal

first. We shall then estimate a reaction function

policy. The

afterwards be

budget surplus

used as

account equation

an explanatory

alongside

the

thus

forecasted

will

variable in the current

price

and

monetary

impulse

variables.
Contrary to
confine our

many works

study to

variables. Thus,

we

deviation between
these impulses

the only
will

in this

influence of

take

into

these domestic

for the

main

influence of foreign fiscal

the different countries.

field, we shall not

OECD

account

impulses and
countries,

domestic
not

fiscal

only

the

the average of
but

also

the

impulses on the current balances of

300
1. FUTURE

STRUCTURAL

BUDGET

SURPLUSES

AND

CURRENT

BALANCE
First. we
surplus (SBn t
for the

shall assume
as a

+ 1

year t+l

this same

structural

budget

of GOP) anticipated by private agents

is equal to the surplus measured ex post for

year. Along

current account

that the

with this variable we shall explain the

(CA. as a

of GOP) by price variables such as

the rate of change of the terms of trade (TOE n ). or of the unit

labour costs

(CUT). The

the increase

in the

For all

monetary impulse

monetary base.

these variables.

compared to

the average

compared to

one of

we will
of the

will be measured by

in nominal terms (HGRon).

also

use

seven main

these countries.

their

deviation

OECO countries or

The regressions obtained

for the seven main OECO countries. using ordinary least squares
are given

in Table

those estimates

1. The

will be

main conclusions

presented first.

to be drawn from

We will

then

give

further details for the U.S.

1.1. The relative importance of fiscal expectations
During the
budget surplus

influences the national current account in four

countries: the U.S.

France and

(equation 2) and the U.K. on the one hand.

Italy on

direction of
country

1973-1986 period. the domestic structural

the other.

this influence

considered.

deficit induces

Thus

a current

proportion in the U.K.

is influenced
surpl~s

varies a
an

the

lot

anticipated

account deficit

size

according
structural
in a

one

and

the

to

the

budget
to half

In France and Italy. the current account

by the deviation of the future structural budget

compared to the average of OECO countries. 80

a deviation

spillover

France. Tnis
negative. This
that the

However.

positively on

spill-over amounts
sign difference

the current

to 44
can be

of such

account in

in Italy. where it is
explained by

the fact

deviation of the domestic structural surplus compared

to the average is almost always positive in France and negative

in Italy (see figure 1 above).
The American

future fiscal

impulse exerts a general

301

Table 1 : current accounl balance and fulure sruclural budgel

surplus

United States
CA = 0,:>3:>. 0,727 SBuS + 1 0,162 TDEus - 0,081 TDEj - 0,03 ClITusj - 0,138 TEMPS
(2,67)
<3,63)
(-,B)
H,50)
(-2,38)
R2 .. 0,928
DW ,,2.15
F = 20,S
SCR = 1.89
Japan

CA =0,006 - 0,077 TDEus 0,087 TDEj - 0,326 SBusall.l 0,04 MGROusj 0.163 MGROus
(-1.8'0
(<(,32)
(<(,5:
(2,31)
(2.10)
R2 =0,956
DW .. 2,23
F .. 3,9
SCR .. 1.62
Germany
CA =2,989 - 0,609 SBus71 - 0,080{ ClITital- 0,363 SBukl 0,09 TDEa11- 0,06 MGROjall
(-2.18)
(-2,20{)
H,63)
(2,58)
0,83)

France

- O,075TDEuk
0,27)
R2 =0,93

DW =2,07

F =15,6

SCR .. 2,02

CA .. 1.39 - 0,235 ClITfall- 0.17 MGROfa11 + 0,60{3 SBus+ 1 +0.185 SBf71 - 0.133 MGROusf
(-6,37)
(-.12)
(5,25)
(<(.15)
(-3.00()
R2 .. 0.92
DW -1.76
F -18.5
SCR - 1.16
United l:insdom
CA .. - 0,555 1,009 SBus71 O.99SBukl 0.073 ClIT usuk 0,256 SBa.1171
(5,0)
(5,75)
0,86)
0.87)
R2 .. 0.910{
DW =2.68
F .. 23,8
SCR "3.63
Italy

CA .. - 7.20{ - 1.33 SBpl 0,961 SBa1!71 1.52 SBf7 1 - 0.57 SBuk71 - 0.5 SBit7-l
(-H2)
(2.63)
(3.58)
(-2.52)
H.7)
0,076 TDEit 0.152 TDEal!
(2.04)
(2,81)
R2 =0,927 DW" 2.56 F" 10,87

SCR" 3.39

Canada
CA =- 1.778 - 1.38 SBus2 - 0,13 TDus O,O9TDEusca-l - 0.371 SBus7.1
(-5.77)
(-3,9)
(2,68)
0,81)
R2 .. 0,865 DW .. 1,87
F IH
SCR .. 2.33

302

the current account of all the main OECD

influence on
countries. except for Italy (1). The level of the American
impulse matters for the French and Canadian current accounts.
while the deviation of this impulse compared to the OECD
average is the key variable in Canada. Germany and the U.K.
A one percent (of GOP) deviation of the American
structural surplus compared to the average budget surplus of
the main OECD countries induces a current account surplus in
Germany (0.6 % of GOP). a current account deficit in the U.K.
and Canada (1 % and 0.4 % of GOP).
However an

American structural budget deficit in the

year t+2.
(1 % of GOP)
influences positively the Canadian
current account (for 1.4 % of GOP). Thus a permanent American
budget deficit influences positively the Canadian current
account surplus,
An American budget deficit amounting to 1 % of GOP is
associated with a French current account deficit representing
0.6 %

of GOP.

If this American budget deficit is accompagnied

by a balanced budget in Germany. it induces a Japanese external

surplus (0.3 % of GOP). Thus. the contrasted influence of an
American budget deficit is quite remarkable. This influence is
positiv..e for
the Canadian.
German and Japanese current
accounts. but nea~~e for the British and French ones.
Non-American budget
surpluses do
not influence
except for
two
significantly foreign
current accounts.
countries. Thus.

the deviation

of the British. French. German

and Japanese budget surpluses relative to the average for OECD

countries is linked positively with the Italian current account
for the first two countries and negatively for the last two
countries. A deviation of the German budget surplus relative to
average amounting to 1 % of GOP induces a British current
account surplus in the vicinity of 0.25 % of GOP.
Relative .c..Q..S..t_...ancL. P-C-i.c..e. variables influence rather
account surpluses.
However.
systematically current
this
(1) This
specific result can be explained by the high
correlation between on the one hand the deviation of the
American fiscal impulse compared to the OECD average and on the
other by the deviation of the Japanese impulse compared to this
same reference,

303

influence

differs

rather

widely

between

of the

American and

one

country

and

another.
The growth

Japanese

terms

of

trade influences significantly the current balance of the first

country. Thus,

a 10

increase

in American

terms

of

trade

improves this

same country's surplus by 1.8

worsened by

contrast, the

global influence of terms of trade variations is

~.

This surplus is

when Japanese terms of trade rise by 10

~.

By

rather small for Japan.

A

10

rise

associated with

a 1

country, but

with a

Moreover,

Italy, a

in

in

the

German

current

1.5

In Germany,

this influence

an impact

trade

is

for this same

account

deficit.

improvement in the terms of trade

induces a current account surplus near 0.8

coexists with

of

account surplus

Italian current

10

terms

of the

of GOP.

of the

terms

of

trade

differential increase in unit

labor costs.

The latter

matters only

Indeed, when

the growth

in unit

in France

and Britain.

labor costs is 1

higher in

France than in Germany, the French current account decreases by

0.24

of

GOP.

In

differential with
only by 0.07

the U.S.,
impulaeB.

similar

deteriorates the

unit

labor

cost

current account

contribute in a significant way to

current account

Japanese current
rate of

of GOP.

~~ry

determine the

Britain,

balance is

growth of

the U.S.

of three

countries. Thus,

the

positively influenced both by the

monetary base and by

~he

monetary

growth differential

between the U.S. and Japan. A 1

growth in

accompanied by zero growth of the Japanese

the U.S.

monetary

monetary base improves the Japanese current account by some 0.2

~

of GOP.
By contrast a 1

and the

growth differential between the U.S

German monetary base induces a current account deficit

in Germany. However, the latter is very small : about 0.06

of

GOP.
For France
differential is
higher in

the influence of the monetary base growth

very contrasted

France than

in Germany.

when

monetary

growth

is

the French current account

deteriorates, while the latter improves when monetary growth is

higher in France than in the U.S.

304
1.2. The U.S. case
To explain completely the U.S. current account we had
to introduce

a time

variable. The

latter's

contribution

is

clearly negative (2). The coefficient of this time trend (Table

2. equation

n02) means that. ceteris paribus. the U.S. current

account balance

% of GNP.

deteriorates every year by something like 0.14

It is not possible to know what the specific cause of

this phenomenon

is. To make a comparison possible we also give

(Table 2. equation n01) the regression without this time trend.

As far

as the

quality of

the regression is concerned (figure

2). the inclusion of the trend suppresses almost completely the

errors for

1976 and

1981 and

improves the

precision of

the

adjusted series in the eighties.

However. even with this correction. the simulation is
not very

satisfactory.

Omissions" (N.E.O.)
explain this.

importance

in the

of

"Net

Errors

and

American balance of payments might

Indeed. N.E.O.

amounted on average to 0.19 % of

from 1973 to 1978. and 0.75 % from 1979 to 1986. Thus

U.S. GNP
the large
in 1981

The

gap between the actual surplus and the simulated one

and 1982

(when a

deficit

is

forecasted

while

the

current account is in fact in surplus) may simply be the result

of the

fact that

when account is taken of N.E.O .. the current

account balance becomes negative (figure 3.c).

When we

estimate a

current account

thus

(CA-N.E.O.). the

quality of

the regression

improves.

shrink considerably

from 1981

to 1983

corrected
Errors

and for 1985. The only

important gap which remains is for 1978-1979.

Three
explanatory

factors

coefficient of
value)

main

differences
(Table

the time

2.

trend is

are

noticeable

equation
much

n03).

higher

in
First.

(in

the
the

absolute

: the corrected current account deteriorates by 0.2 % of

GNP. compared to 0.14 % for the uncorrected one. Then.

in place

(2) When this variable

is replaced by an autoregressive term
(CA_ 1 ) . the latter is not significant.

305

Table 2 : Estimates of the American current account.

CA = 0,534

CA

1.189 SBuS + 1 + 0.176TDEus - 0.0% TDEj - 0.068 CUTusj

(5.08)
<3.25)
(-4.17)
(-3.07)
R2 = 0.876
DW = 2.41
F = 15.92

(1)

SCR

3,22

0.535 + 0.727 SBus. 1 + 0.162 TDEus - 0.081 TDEj - 0,034 CUTusj - 0.138 TEMPS (2)
(2.67)
<3.63)
(-.(.13)
(-1.50)
(-2.38)
R2 = 0.928
DW = 2.15
F = 20.5
SCR = 1.89

(CA-E.D.N.) = - 0.033 + 0.2.(5 SBusj.l + 0,14 TDEus - 0,052 TDEj - 0.04CUTusj

(3)
(2.18)
(2.93)
(-2,~2)
(-1.79)
- 0.041 MGRDusj -0.203 TEMPS
(1.15)
(-2,58)
R2 &0.944
DW=1.75
F s 19,5 SCR=1.75

306

Graph n 2 : The quality of the regressions for the American current

balance of payments.
United States
o U.S. observed
T U.S.adjusted
1.5

~: ---~~\---------

Po.

-1

(!)

......
0

...c::

'-'

a)

;::)

b\~

-1.5

u
u

...c::
..s

-2

4)

-2.5

;::)

-3

''""''

-3.5+-~-""----"--r---'--r--~-r-~-""--..---.-~--..-~-..

72

74

1.5

;;::;
z
(!)
b)

76

78

80

82

84

86

as

o U.S. observed

U.S.adjusted

.5

......
0

-.5

...c::

;::)

-)

u
u

-1.5

4)

-2
-2.5

......sc::

'"'

8'"'

-3

.~

-3.5

72

;;::;
z
(!)

74

76

78

80

82

84

86

88

o U.S. observed

U.S.adjusted

c.:..
0
~

....CI

'-'

;::)

c)

u
u

..s

....CI
4)

''""''
u
;::)

-1

-2
-3

"0

Su

-4

''""''

-s

4)

72

74

76

78

80

82

84

86

88

307
of the

level of

the American

fiscal impulse, we now find the

deviation of the latter compared to the Japanese impulse. A 1 %

gap between
surplus

these

two

amounting

to

coefficient of

impulses

induces

0.25

of

GNP.

current

account

Finally,

now

the

the monetary. growth differential between these

two countries

is negative.

differential

induces

an

American current account deficit of nearly 0.04 % of GNP.

For all
regression is

non-American countries,

the quality

of the

quite good.

This is

mostly the case for Japan,

Germany (except

in 1982),

and the

U.K.

The regressions

are also quite satisfactory for France, except

(except in 1985)

(3).

in 1976 and 1978, and for Italy, except in 1976, 1981 and 1984.
The regression is bad for Canada from 1976 to 1979, and in 1982
and 1986 (Chart 3).
The first

method used

future structural
nature

of

the

to

budget surplus
assumption

of

measure

the

is limited
perfect

anticipated

by

the

ad

foresight.

hoc

It

is

undoubtedly more satisfactory to assume that in their forecasts

of

future

fiscal

deficits

informations available
deficits. One

today

must then

economic
on

the

estimate

agents

use

determinants
fiscal

policy

all
of

the
these

reaction

function.
2. FISCAL POLICY REACTION FUNCTIONS
We shall
structural budget

estimate a
deficit for

objective-variables of
the seven

main OECD

thus regress

reaction function
year

t+1

in which the

is a function of the

economic policy for year t. For each of

countries over

the 1973-1986

period, we

the structural budget surplus for t+1,

(from 1974

to 1987), on the following variables for year t

(3) A regression in which we introduced a dummy variable for
1985 for the United Kingdom gave the following result
CA

=-

0,471 + 1,026 SB US7 +

(6,37)

+ 0,533 SB UK +
(7,31)

+ 0,1 CUT usuK + 1,682 DUM85

(3,27)
(2,72)
0,934

OW

2,4

31,9

SCR

2,77

308
Graph n 3 : The quality of the regressions for the non-American
current account balances.

a::

Japan

'"~5

a::

o Jap observed

Jap adjusted

'Q

~4

..

~
~

..

c:

;;;

"'c:"
~

;;;

"l

"'c:"

u
u

c:f -t
...

a -2n

74

76

78

eo

c:

'"'Q<.:>

F adjusted

1.5

..

i......

ee

116

lI'I

82

Geeaany
Ger adjusted

4
3.5
3
2.5
2
1.5
1
.5
0
-.5
-1
-1.5

-2

Feance

eo

o Ger observed

82

OF observed

.5

'"~

;;;

-.5

-3.5

"'c:" -1
::>
e -1.5-2
li
c: -2.5
~ -3
...

a::
8

74

76

o Ita observed

'"uli

;;;
&>

;;;

c:

C
~
8u

~ -3

i...

..

"'" -1

S-2

...::>

...::>~ -~

'-'

86

Ita adjusted

'Q

.~

lI'I

Italy

OUK observed

Ci

..

82

a::

United K.inldoa
UK adjusted

eo

78

-1

-2

..
-3

ee

74

a::

Can adjusted

c;

Canada
o Can observed

.5

1;,"::

tc:
~

"'c:"

-.5
-1

-1.5

e -2
-2.5
c: -3
~

li

::>
u

88

309
- the

rate of

growth of

the consumer

price level

( INF) ,
- the unemployment rate (TCHO),
- the share of public debt in GDP,
- the rate of growth of real GDP,

(DET),
(TCR),

- the exchange rate of the domestic currency against

the U.S.

dollar,

(the

$/DM or $/Yen rate for the

U. S. )

- and the structural budget surplus.

Each time,
deviation of

we also

consider the

these variables

influence

compared to

of

the average

the
of the

seven main OECD countries, and the influence of similar foreign

variables. We always keep only significant variables.
2.1. The main explanatory factors
As the
dollar is

estimation covers

floating,

respect to

Italy, Canada

over which the

dollar exchange rate with

influences significantly the

budget surplus

and Japan.

the period

3), the

the national currency,

future structural
regime of

(Table

in all countries, except for

Indeed, the latter either witnessed a

controlled floating,

or foreign exchange regulations

over the whole or part of this period.

The way
differs in
hand,

fiscal policy

countries such

and France

structural budget

and the

In

against the
German and
the

Franc

U.K. on

the exchange

rate

and the U.S. on the one

the

other.

Indeed,

the

surplus is linked negatively to the domestic

currency's exchange
latter.

reacts to

as Germany

rate in

other words
Dollar is

the former, and positively in the

an appreciation
associated with

U.S. structural
(Pound)

of the

Deutsche Mark

an improvement

of the

surpluses, while a depreciation of

against

the

Dollar

is

linked

to

deterioration of the French (British) structural surplus.

The
compared to

deviation

of

the foreign

the

domestic

one influences

and Canadian

structural budget

British one.

Moreover,

in

unemployment

negatively the French

surpluses, but

Canada, the

rate

level of

positively the
the

dom~stic

310

Table 3 : fiscal policy reaction function

Dependent variable: domestic structural budget surplus

United States

SB =5.713 - 2.094 ($/DM)-1 + 0.146 INFUS7-1 + 0.151 DETUS7-1- 0.956 DUM80

(-12.6)
(7.18)
(10.3)
(-3.90)
RL 0.968
DW =1.62
F : 68.5
SCR =0.38
Japan

SB = - 0.961 + 1.12 SBj7-1 - 0.169 INFj7-1 - 1.571 DUM78 + 1.362 DUM74

(10.9)
(-4.M)
(-3.43)
(2.84)
R2 - 0.96
DW -1.53
F - 53.5
SCR - 1.62
Germany

SB -- 4.40 + 1.03 ($/DM)-1 + 0.081 DETall-1 + 0.465 INFal17-1

(2.84)
(7.17)
(7 ..m
R2 =0.967 DW =2.33
f =97.2
France

SB = - 26.44 + o.n INFf-l - 1.09 (SIFf)-1 + 1.00 DETf-l - 0.56 INFfall-l

(5.14)
(-4.90)
(5.76)
(-5.70)
0.408TCRall-l -0.311 TCHOfall-l
(4.40)
(-2.27)
R2 =0.874 DW =2.4
f ~ 8.07
SCR =0.723
+

United l:in8dom

SB =18.97 -17.9 (S/)-1 - 0.205 DETuk-l + 1.217 TCHOuk7-1 + 2.08 DUM77 -1.7 DUM79
(-5.87)
(-5.67)
(6 ..{9)
(2.85)
(-2.50
R2 =0.961 DW =2.19
F =39.6
SCR = 3.19
Italy

SB =-0.988+ 0.335 DETus7-1-0.626 SBf-l + 1.69 DUMn. 1.92 DUM80 + 1.68 DUM83
(8.62)
(-2.46)
(2.80)
(3.13)
(2.73)
R2 =0.929 DW=2.50
f=21.0
SCR.2.51
Canada

SB = 1.194 0.188 DETca7-1 - 0.417 TCHOca-l -0.5 TCRca7-1 -0.097 INFcaus-l

(3.01)
(-2.95)
(-3.'53)
H}1)
-0.514 TCHOcaus-l + 1.85 DUM74
(-2.30)
(2.80
R2 =0.903 DW = 1.61
f =10.82

SCR .2.20

311

unemployment rate intervenes negatively.

The deviation of the domestic rate of .i.n!..l.a.t...i.a..n
compared to OECO average has a positive influence in the U.S.
and Germany. A domestic price increase greater than average is
thus combined in both countries with an improvement in the
structural
budget
surplus.
In
France,
the
inflation
differential with Germany has a negative influence. However, it
compensates for a positive influence of French inflation on
this country's structural budget surplus.
For instance, when French inflation is 5 ~ while
prices do not increase in Germany, the French budget impulse
will show a surplus in the vicinity of 1 ~ of GOP.
The rate of growth of real GOP plays a significant
(negative) role in Canada,
(as a deviation from average), and
the German rate of growth influences positively the French
budget surplus.
In all countries except Japan, the share of public
debt in GOP has a significant role to play. The deviation of
domestic public debt compared to the average of major OECO
countries has a positive influence in Canada and the U.S. (with
respectively 0,19 and 0,15).
In France, Germany and the U.K.
the structural budget surplus is influenced by the share of
domest i c publ i c debt in GOP. However, th i s i nfl uence is
positive (and very high) in France and Germany (very low), but
negative in Great Britain.
It seems very difficult to explain the evolution of
the Italian structural budget surplus. Thus, we had to use
three dummy variables (DUM) for 1977, 1980 and 1983. The main
explanatory variable is then the deviation of the American
public debt compared to average. When this deviation is
positive, the Italian surplus improves.
Past domestic structural budget surpluses play
a
very minor role. In Japan, the domestic budget surplus lagged
one year as a deviation from the average. has a positive
coefficient slightly greater than one. Thus it seems that Japan
seeks to
maintain a
surplus in
its structural budget
corresponding to this past deviation. In Italy, the level of
the French structural budget surplus lagged one year has a
negative influence. A French surplus amounting to 1 ~ of GOP in

312

year t-l would thus call for an Italian deficit of about 0.63 %
of GOP in year t.
2.2. The contribution of different variables
The graphical
the

different

representation of

variables

surpluses allows

to

the

the contribution of

explanation

of

structural

us to isolate the key factors responsible for

the major fluctuations (figure 4).

In the
~_~otial

to explain
is true

three

can be

"justified" in
one

countries.

the

inflatiQn

which

structural budget
can

measure.

seems to
this price

the German

in the inflation

and the average for 1975. which

In

surplus greater than

Japan.

have played

especially during

fluctuations in

The forecast error for

the increase

the U.S.

we

differential also

influence on

for 1976.

explained by

1976 a

fiscal policy.

of the domestic fiscal impulse. This

U.S. except

differential between
the

OECO

with the average seems to have played a major role

the evolution

in the

that year

main

the

inflation

a major role to guide

the years

1973 to 1977. The

variable have also exerted a major

surplus. except for 1981 and the more

recent years (1984 to 1986) when the inflation differential and

the structural surplus have been moving in opposite directions.
In France.
minor role

the inflation

during some

differential played only a

years (1973

to 1976,

then

1978

and

1981). Indeed, the main turning points in the French structural

budget surplus can mainly be explained both by the deviation of
the French

unemployment rate

compared to

Germany and

by the

German real rate of growth. The first factor explains the large
French deficit

in 1978,

while the

increase in the deficit in

1981-83, and its reduction in 1984-85 can be explained first by

the slowdown

of the

German growth, and then by a new spurt of

growth in Germany.
In the United Kingdom, the evolution of the deviation
of the unemployment rate "justified" a deficit greater than the
one measured

in 1977,

while in

1979. the

evolution

of

exchange rate could enable us to forecast a smaller deficit.

the

313
Unlled Slales
AU_S_adjusled
0 U_S_ obsened.

;;:

""

<.!>
~

Japan
A jap

'0

'"::>

::>
'"

-I

...

Q.

...

Q.

.5

f;;:
il
co

::>

-2

'"

il
co

-3

'"::> ....
"'-;;;..."

-5

'"::>

o jap observed

adjusted

!!

1.5

!!

;;:

"'-;;;..." -I
e...::> -1.5
n
V;

E
<>

-5

-{,

::>

_0

Contribat.ion or the ....riable.

A(lIDMX-1l

O(lIUS7X-1l

.lDTUS7H)

7.

76

7e

80

82

Clinlribalioa or Ute variab les

Ass,m-Il

12

OINfJ7(-1l

10

...

-I

<.!>

-2

""
'0

...
'0
...

... -.
-3

-2

-5
-{,

-7

-4

-{,

7.

76

76

80

82

6.

86

80

62

Graph n 4a : the quality of the regressions for the structural budget

surplus.
Q:.

Ger.any

""<.!>

'0

o Ger observed

Ger adjusted

!!
::>

'"

.5

'"
~

-.5

...::>

;;:

United IUnado.
A

'0

o UK observed

UK adjusted

!!

I~

'"

::>

.
...::>

Q.

Q.

..,il::>co

-I

'"::>

-1.5

-2

... -2.5
E
-3
...::><>
V;

-;;;

-;;;

...

-I

-2
-3

E ....

74

7G

7f:.

60

82

64

66

~
"V;

Coatribatioa or lhe variable.

-5

7.

76

18

80

"(SID.M X-I)

7~,

25

.,.

..

...
'0
...

_I

-2

-2.5
-5
-7.5
-10
125
-IS

-17.5
86

8L

8-4

6'

Coatribalioa of the ....riables

76

80

'<t

<:>

Q-

os

!!

.
.....''""

Q.

.Q

....
g...

Vi

"'-

...

8"'-

'IS

-.5
-I

-1.5

-I

-2

-3
-~

-5

35

30
25
20

15
10
5

-5

-10
-15

74

78

F adjusted

76

78

ft"l

80
DrCA~K-1I

80
.INFn-1I

80

82

82

82

/
86

86

Ii:)

s...
!!
~

e
II>

'"

'a

...'"
"2

g
...

-7
-8

-9
-10
-II

-12

72

-I

-.5

.5'

1.5J

Vi -13

"'-

S
'S

..

-2

-1.5

-2.5

74

78

IJO

Ita adjusted Italy

76

78

80

62

&2

o Ita observed

86

'0

86

~
84

&4

OS8f71-11

Ctrill.tir tbe .... i le

AOE"ruslHI

76

~,

74

"S

.e

!!

-~

-I

1.5

-.5

2.5
-3

'S

.8

-4
72

25
1.5

-6

-5

-4

-3

-2

-I

~sl

1.5

-I

"'- s
'S

Vi

~ -35

e'"

c;

...1
...

.~

15

Graph
quality of the regressions for the structural budget surplus
. n .fb : the
(cont.)

86

of observed

84

.(l/ffX-1I

UCHOI,U(-II

eo.triti. . .r til.....1.....

76

76'

DINFrolll-l1

7~

DOEFn-1I

7~

74

&

76

78

CDd.

80

Can adjUSted

78

Can observed

/104

B6

82

OlhfcM(-1)

80

80

84

OTCAc..71-11

/104

B6

"P

86

/"

arCHOt-.,,(-1)

C tri ti r tbe vari le.

76

76
78
erCHCo:InI-1J

.01(17(-1)

74

315

3. EXPECTED BUDGET SURPLUS AND CURRENT BALANCE OF

PAYMENTS
We can
t+ 1.

for year

use the

structural budget surpluses expected

is over the per i od 1974-1987) - generated

(that

by the reaction functions introduced in the preceding section together with relative price and monetary impulse variables. to
explain the

current balance of payments of the seven main OECD

countries over the period 1973-1986.

3.1. The role of the expected budget surplus
Three countries
domestic structural

U.S .. France

first two

countries.

and the U.K.

it

than the

surplus compared
against 0.78)
higher than

on the

expected

current account

(Table 4). However,

in the

the deviation of the domestic

in

France, the deviation compared

has a positive coefficient. The latter

one for

to the

the deviation

OECD average

because the

in

French-Italian

of

the

domestic

(0.17

table
deviation

the French-OECD average deviation.

coefficient of

the

the foreign one which is the major

Thus,

Italian surplus

is smaller

is now

compared to

explanatory variable.
to the

influence of

budget surplus

i.e. the

fiscal impulse

show an

is

as
much

In the U.S. the

the US-Japan deviation is almost unchanged: a

1 % of GDP deviation between the surpluses of the two countries

generates an

American corrected (4) current account surplus of

% of GDP as against 0.24

presented in section 1 above.

nearly 0.27

In Italy.
surplus both

we now

have an

in

influence

the
~f

regression
the

budget

as a deviation with the OECD average and with the

(4) For the non corrected current account balance (equation 4),
the coefficient of the time trend increases at the exp.ense of
the coefficient of the American budget surplus. In the absence
of the trend. the regression becomes:
CA

= - 0,686

+ 1.023 SBP us + 0.19 TDEus

(3.70)
(2.76)

- 0.129 TDEJ - 0.053 CUTuSJ

(-4.55)
(-2.03)
R2

0.81

DW

= 1.65

9.61

SCR

4.93

316

Table" : Current account balance and expected ruture budget surpluses

Dependent variable (CA) : CUl'rent account surplus as % of GDP

United States
CA =0,073 + 0,528 SBPus+ 1 + 0.165 TDEus - 0,09.( TDEj - -0,018CUTusj - 0,173 TEMPS
(1,99)
(3,21)
(-3,9.()
(-0,85)
(Z,9Z)
RZ .0,908
DW 1.53
F -15,8
SCR 2,<{

(Z')

(CA-E.O.N'> =- 0,258+ O,Z72 SBusj+ J +0.1<{ TDEus - 0,059TDEj - 0,038CUTusj - 0,191 TEMPS
(Z,65)
(3,25)
(-3,09)
(-Z,O<{)
(-Z,58)
- 0,05MGROusj
(-1,50)
R2 =0,953

Japan

(3' )

DW = l.n

F =23,6

SCR = U6

CA =0,.457 + 0,062 MGOusj+ 0,157 MGROus - 0,092 TDEus +0.105 TDEj - 0,303 SBPusall+ 1
(3,07)
(2,47)
(-2,05)
(5,21)
(-<{,24)
R2 =0,952
DW = 1.9Z
F =31,4
SCR = 1.79
Germany.

CA = 3,186 - 0.102 DETita11 - 0.089 MGROjaU- 0.104 TDEall - 0,135 TDEuk -0.357 SBPuk-l
(Z,89)
(2.76)
(-2.61)
(-4,62)
(3.16)
- 0,166 SBPusj.l
( -1.66)

F = 15.1

SCR =2,07

France

CA = - 1.1 3 + 0.662 SBPus1 - 0.07 MGROfall- 0,118 CUffall- 0,078 TDEall-l 0.17 SBPfit.l
(Z.33)
(UO)
U.56)
(-4,28)
(-3.10)
R2 =0,837
DW =Z,OO
F ~ 8,Z4
SCR = 2.36
United Kingdom

CA = -2,756 0,6Z5 SBPukl .. 0,08<{ CUfusuk .. 1.1<{7 SBPus7-1- 0,554 SBPa1l7.1

(6,n)
(2,11)
(5,30)
(2.75)
R2 0,915
DW "Z,20
F 2<{.37
SCR " 3,56
&

Italy

CA = -1 1.1 7 - 0,742 SBput7. J - 1,623 SBPj7. I 1.028 SBPaU7. I - 1.334 SBPit7. I

(-2.50)
(-3.32)
(2.06)
(1,50)
.. 2,09 SBPfit.l + 0,091 TDEit
(2,33)
(1.70)
RZ =0,815

DW=2,01

F5.13

SCR c 8,58

CA" -0,232 - 0,852 SBP us .2 -0.083 TDEus- 0,403SBPus7.J

(-3,88)
(-2,<{4)
(-2,29)
R2 =0,791
DW =2,02
F = lZ.6

SCR = 3.62

Canada

317

French

surplus.

The

explanatory variable
Its coefficient

the

only

level of

is the

presented above.

more than

country

where

the

the structural surplus.

higher than the one of the actual

is slightly

structural surplus
spillover of

is

U.K.

half of

However, this

a British

confirms a

structural

budget

deficit on the current account of this same country.

The American

fiscal

impulse still plays a major role

foreign current accounts. A deviation of 1 % of GOP

to explain
between the

American

and

German

structural

surpluses

thus

generates a Japanese current account deficit in the vicinity of

0.3 %

of GOP.

If this same deviation exists with the Japanese

budget surplus,
paribus, a

the

German

deficit of

deviation is

current

account

shows,

ceteris

% of GOP. Finally, if this

nearly 0.17

taken relatively to the OECO average, the British

current account

has a

surplus in

excess of

1 %

of GOP.

In

France, we still find an influence of the level of the American

surplus, with

a remarkable

stability of its coefficient (0.66

as against 0.64 in Table 1).

As far
of the

as Canada is concerned, the overall influence

American surplus is hardly changed.

In fact an American

structural budget deficit amounting to 1 % of GOP maintained in

year t+l

and year

t+2 now

generates (if it is coupled with a

OECO

balanced budget

for

the

account surplus

of nearly

average)

1.25 %

of GOP

Canadian

current

(as against

1.03 %

above) .
The influence
similar to
find the

the one
level of

account equation
as against

of non

American

explained in
the British

with an

-0.363).

The

fiscal

the first
surplus in

impulses

is

section. We first
the German current

almost unchanged coefficient (-0.357

Italian

current

account

is

still

influenced by the deviation of the Japanese, German and British

surpluses compared to the OECO average. The higher contribution
of the

Japanese surplus

British one,

while the

compensates for
influence of

the

the reduction in the

German

surplus

is

hardly modified. The deviation of the German structural surplus

compared to

the OECO

average now contributes to determine the

British current account. When this deviation equals 1 % of GOP,

the U.K.

has a current account surplus of about 0.5 % of GOP.

Relative price

variables still

playa major role in

318

nearly all

the current

channel of

the terms

of unit

account equations.
of trade in Japan,

relative costs

in the

They still use the

Italy and Canada, and

United Kingdom.

Both channels

appear in Germany and the United States. However, the influence

of German

terms

vanishes. By

of

trade

on

the

contrast, German

Italian

terms of

current

account

trade lagged one year

now exert a negative influence on the French current account.

The

role

negligible way

of

in France.

growth differential
differential with
before. By

monetary

impulses

The influence

with the

in

non

of the monetary base

U.S. disappears

Germany still

contrast,

varies

completely. The

plays a role but smaller than

in Japan and Germany, the contribution of

this monetary impulse is almost unchanged.

3.2. The

contribution of

the different

variables and

the quality of the regressions

For

three

countries,

"rational predictor"

the

regressions

using

the

for fiscal surpluses are almost identical

to the ones using structural budget surpluses measured ex post.

These are: Japan, Germany and the United Kingdom. The study of
the contribution
us

to

of the

understand

different variables (graph 5) enables

the

origin

of

the

current

account

develooments.
Thus,

in

Japan the important residual in 1982 can be

explained by

the positive

the American

monetary

American and

German budget

deterioration of
is of

base

and

trade, but

the

both the growth of

deviation

surpluses expected

the Japanese

course caused

terms of

contribution of

by the
also by

between

the

for 1983.

The

current account balance in 1979

adverse movement
the reduction

of the Japanese
in

the

American

monetary impulse. The reversal of the evolution of the Japanese

terms of trade in 1981, and of the American monetary impulse in
1982 explain
surplus in
noted for

partly the growth of the Japanese current account

the early eighties. However, the strong improvement

the years

between the
recent years,
we must

1981 to

American and

German fiscal

the explanatory

emphasize the

1983 is

mainly due
policies.

to the

gap

In the most

factors are more diverse. Thus.

evolution of the U.S. terms of trade in

0)

(,!)

""

0:
~

'-

'";

-;

c:

,s:,

"ouv -,

0$

-3

-2

c:
72

1.2

IA

.6

.6
.4

~l

-:;
-.4

-.61
n

2.5

1.5

.5

74

76

76

eo

62

eo
+ TOE ...

eo

oJ

.MGR<Mj.

66

.TOEj.

y-

L.~

Contribution of the Yariables

ocurusj.

~-4,

,
M

Q:

'0

"c

v
~

c
'-

"

a.

'0

...

Q.,

-1

-2
72

1.5

-5

-,

-,

-2
72

-1.5

@
c

~.

-2

72

-3

Japan
Jap ad jus ted

76

78

DMGROJs

82

;.,

66

o Jap observed
.at

pi

84

66

O'SBPusall + I

,._.. ~_~--<r-1,

64

,-~

62

6 TD[usa

~--~~-

ao

/\
/
\;1

~v
74

eo

Contribution of the uri.bles

76

aTOI, ..

76

60

~~~-~-.,----~--

eMGPrAlSJ

74

(,!)

""

'o

.,

c:

,s:,

"ov

4 .

-I

-,

.1.5

-1

-.5

.5

-1.51
72

-5

~ -21
t 72
U

"

'0

....

a.
o

;:
$;-10

-2
-251
72

Ger adjusted

Gerllany

v-'.
eo
~

OTDEry

eo

eo

aS6Pru + I

o Ger observed

o SBPusj+ I

Contribution of the Yariables

.MGAO,.,

erOhtl

Graph nO') : The contribution of variables to the explanation of current

jOt.,
accounts.

64

united States
U.S.adjusted
01 U.S. observed

OS8Pusj.

-51-I-----~-~--~~~-_:r:_'

-4

t...
"

'"'

""3u
''c""'

c..

13

(;

...

p..

'0

13

.~

-.5

-II

C"')

'.5
1
15
2

25

3 S.
72

lj
15

51

7<

&

fruce

60
62

66

o F observed

64

GMGROtIIi

66

&SBflIPhl)

\ /'J'Q
60

OTO,1111)

60

66

Ii:

'0

It>

fa

!:!
.,

..

c:
.6

:a

c:

'0

80-

..

0.

-I

-2
-3
-4

-I

1.5

.~

0
.5

5
....

74

74

74

CUTus"

76

OSBPoI17f+1)

76

78

78

78

80

eo

60

82

82

82

64

OS6Pus7('1)

84

.5BPru(+1)

84

66

86

66

CoatributioD of the .ariallies

76

UDite. ~iDl.o.
o OK observed
.. UK adjusted

Graph n"5 : The contribution of variables to the eJ:plaaation of current

accounts.(Cont.)

76

F adjusted

76

OS8USPlt 1)

'/', / \
'd

eoDnilnatioD of the nriables

I \

fJ-.-.. ....-D

.curr,11

j ~
-1.5,

72

25.

ISl~~

7<

;il/~~ =:;~
72

Ii:
'0

1:5

!:!

c:

8
ii

~
-

-I

-2

-I

7<

&

Ita adjusted

Italy

o If a observed

oS6pr,UflJ

66

./

io

-\

OS6PJ7C+1I

Cootrillutioo of tbe "ariab Ie.

.S8P11I71+ II

'S6P,l7{tl)

eTDflll

if

O~6Put.7(.I)

"

~.,~----../...

~~'b

<I __~~__~~__~____~__~~__~~~e6
~ -4L
U

0Q

2
3
30
25
IS

20
10

-3

-I

-s

-10

'0

0-

'0
N

0.

'0
...

41

321

1984,

the

monetary

growth

in

differential

1985,

and

the

Japanese terms of trade in 1986.

For Germany, the residual almost vanishes in 1976 and
diminishes strongly
different in

in 1979.

The errors for 1981-82. not much

the two regressions. can be explained by a strong

negative contribution

of the British budget surplus (1981) and

by a very strong positive contribution of German terms of trade

( 1982) .
In the

United Kingdom.

account deficit

in 1985

the forecast

of

current

is motivated by the large decrease of

the deviation

of the

American fiscal

OECD average.

The inverted

V shaped

impulse compared to the

movement of

the current

account balance between 1979 and 1984 is very well explained by

a

similar

evolution

expected structural
astonishing that

of

the

contribution

of

is

the

domestic

budget

surplus.

It

of

course

most

the terms

of trade

should not

appear

here

directly. but only through the relative labor costs compared to

the United

States. which

playa relatively minor role (except

for 1981).
In

two

countries

the

quality

of

the

regression

improves. This is true for the United States with the corrected
current account balance (which includes N.E.O.)

(5).

The improvement in the quality of the regression when

we use the rational predictor can be explained by the fact that
the evaluation of the deviation of fiscal

impulses (between the

United States and Japan) explains almost alone the evolution of

the (corrected)
The other
terms of

current account balance between 1977 and 1984.

factors play
trade from

only a very occasional role : American

1973 to

1976. Japanese terms of trade in

1986 and monetary impulses in 1985.

Thus we
American current

can better

understand the

evolution of the

account balance in recent years when we use a

rational predictor to evaluate fiscal expectations than when we

restrict ourselves to the assumption of perfect foresight.
In France.

the quality

of the regression is clearly

(5) For the non corrected balance.

when we
include the time
trend. the regression is almost identical to the one obtained
in section 1 above. The contribution of the different variables
is given in the Appendix.

322
improved for

1977.

the picture
to 1977

1979-80

and 1983-84. For the other periods

is hardly changed. The W shaped movement from 1973

can be

explained by

the monetary growth differential

between France and Germany, and by the American fiscal

The error

for 1976

trade in

1975. The

evolution of

is due

to the movement in German terms of

1980 error

the expected

between the

impulse.

can be

explained both

American fiscal

by the

impulse, by the gap

French and Italian structural surpluses and by the

relative unit labor costs (France compared to Germany).

The latter

factor, combined with the monetary growth

differential with Germany, explains the 1981 residual.

In 1985-

86, the rebalancing of the French current account can be linked

to the

American fiscal

expectations, the

relative unit costs

and, slightly, to German terms of trade.

In Italy
presented is
given in

and Canada,

much worse

than before.

the Appendix).

overall movement
followed.

Thus,

explained by

in

(for Italy,
the first

current account
curve effect

graphs are

country,

is still

the

quite well

from 1979 to 1983 is well

the deviation compared to the OECD average of the

Japanese, German
as Canada

However,

of the
the J

the quality of the regressions

is

and British

concerned,

it

expected fiscal
seems

that

corrected measure of the current account,

impulses. As far

one

should

use

just as for the U.S.,

since NEO are quite large in the Canadian case too.

Indeed, the

deterioration in the quality of the regression is also apparent

(see Appendix)
balance. The

in the US for the non corrected current account

important residuals

explained both

by the

negative

for 1981

and

contribution

1982
of

the

can

be

fiscal

impulse and by the evolution of the Japanese and American terms

of trade.

In 1985. the reversal of these movements explains the

underestimation of the current account deficit.

CONCLUSION
The estimation

of a current account equation for the

seven main

OECD countries

(1973-1986)

led

us to

over

the

isolate the

floating

dollar

period

relative importance of the

323
contribution of

fiscal expectations

compared to price effects

and to the influence of monetary impulses.

Evaluating the
under

the

expected

assumption

of

structural

perfect

budget

foresight

surplus

enables

us

to

emphasize four main results.

The future
general influence

American fiscal
on the

impulses seem to exert a

current account balances of the main

OECD countries.

However, the polarisation of the effects of an

American budget

deficit is

for

p~Lt~

the German

most remarkable. These effects are

and

Japanese

current

accounts

but

nesatjve for the French and British ones.

The influence

of

the

domestic

fiscal

impulse

restricted to four countries: France. Great-Britain.

is

Italy and

the United States.

While playing
role, price

a non-negligible and quite gen.ralised

effects are

quite diverse.

Indeed. they take the

form either of terms of trade effects. or work through relative

unit labor

costs. Finally, the role of monetary impulses seems

relatively minor.
This measure

of the

expected budget surplus enabled

us to give a quite satisfactory account of the evolution of the

current account
the United

balance of all main OECD countries, except for

States in

account the

the eighties. However, when we took into

importance of

American balance

Net Errors

of payments,

and

we were

Omissions

able to

in

the

show that the

different explanatory variables used explain better the current

account balance corrected from these errors than the usual one.
It remains
is very

that the

specific.

evolution of the American current account

In

both cases

(corrected and non corrected

surplus), to explain completely this balance one has to include

a time

trend among

towards a
0.2 %

the

exogenous

deterioration of

of GOP

might reflect

every year
the erosion

variables.

the external
has not
of the

This

tendency

surplus at a rate of

been explained here (6).

dominating position

It

of the

United States in world trade. Structural variables representing

(6) A similar finding

BALDWIN (1987).

has

been

emphasized

by

KRUGMAN

and

324
APPENDIX 1
Definition and sources of variables
CA

Current account balance (% of GOP).

OECO (1987a), Table 55 and R5.

CUT

Rate of growth of unit labor costs.

OECO (1980) and (1987b).

DET

Net public debt as % of GOP.

OECO (1987a), Table 15.
Dummy variable which has the value 1
for year t and 0 otherwise.

INF

Rate of increase of consumer prices.

OECD (1987a), Table RIO.

MGRO

Rate of growth of the monetary base.

Computed from IMF (1987) row 14.

SB

Structural budget surplus as % of potential GOP.

PRICE and MULLER (1985), Table 2 and OECD (1987a),
Table 14.

TCHO

Standardised unemployment rate.

OECD (1987a), Table R12.

$/N

Number of units of currency N per U.S. dollar.

OECD (1987a), Table R15.

TCR

Rate of growth of real GOP.

OECD (1987a), Table Rl.

TOE

Rate of variation of the terms of trade. (Unit value

of exports in $/Unit value of imports in $).
Computed from IMF (1987),

XX. - yy,

row 74 and 75.

Deviation of variable YY between country

and the average of the seven main OECD countries.

325
the quality of the international specialisation of this country
could thus

usefully supplement the present study which focused

exclusively on macroeconomic factors.

To test

the robustness

of the preceding results and

to progress in the modelling of fiscal expectations, we assumed

that agents
surplus

"rationally" forecast the future structural budget

by

using

all

the

available

information

on

the

determinants of this surplus.

The study
budget surplus

of

the

of the

reaction functions,
United States,

of

the

structural

main OECD countries using fiscal policy

to oppose .l.a.r:.&.e..._e..c...Q..U.Q.ID.i.e..s. like the

led us

Japan and

main explanatory

determinants

Germany where

variable,

to

medium

past inflation is the

sized

economies

like

France, the United Kingdom and Canada where this role is played
by

the

past

unemployment

rate.

Moreover,

States) or a negative contribution. The share

GOP

plays

major

either
role

in

is

(Germany, United
debt in

have

it

exchange rate

of public

can

when

significant, the

positive

almost

all

countries. By contrast the past structural budget surplus has a

very limited influence.
The estimation of the relative contribution of fiscal
expectations, generated
to show

that the

Germany, Japan
main results.

impulse on

impulse influences
United States,

the fact
as in

using a

the

contribution of
is fairly stable.

first

that the

domestic

current account

France, the

for

the French
half

of

relative price

fiscal
in

United Kingdom

rational predictor

perfect foresight

evolution of

balance in

robust

the current account balance of the

significantly the

as well

Italy. Moreover,
of the

are

We could confirm both the major influence of the

countries, and

assumption of

equations

and the United Kingdom. Those tests led to four

American fiscal
main OECD

by such reaction functions, enabled us

current account

rather than

the
or
the

enabled us to give an account

and American
the

eighties.

current account
Finally,

the

and monetary impulse variables

P.:
w

-;
~

...

I.S

I
,5

IS

2-

-2.5

-I

~-:I

ii

~ --,

~ -,-~

~
'1

-2

"
74

United States if"

78

Append ix 2

0us observed

.S8Pus(+ fl.

84
eo"82"
86

~ ~
OCUTusj.

78

78

eo

,,/\

eo

~
84

Ol1lEj,

84

86

86

72

IS,

78

76

.TDEtlS

60

62

0 ",

64

64

.b"~"

&6

66

eS6Pu57(+ll

Contrillution of the uri.llie s

80

e....

., "'-"'," '"

16

,--.----

7.4

76

OS8Pusl+21

74

82

merican
the contrib ution of variabl es to the explana tion of North-A
- current acount balance s

p;

'-

.S

-.S
-I

;; 'J

~
-;

J:l

~
-IS
~
.
:::>
-2

~ ~S

L:
~

::

~ -_~
-272

~ ~I.S

A l1lEus.

.......

Conuillu tien of tile uriallie s

76

A US adjusted

J\

I~~ f

~
-4

-2
-6

25

-I

-15

~-IO
~ -~

...

~L

Rearessi on inc:tudinl a ti.e trend' .

327
A SELECTED BIBLIOGRAPHY
Bourguinat,

(1987)

H.

Les

de

vertiges

finance

la

internationale. Economica, Paris

Feldstein, H.

(1986) -

"The Budget

Deficit and

Macroeconomics Annual N.B.E.R.

Frenkel, J.

et Razin,

A.

and Internationl

(1985)

the Dollar".

1986.

- "Government Spending, Debt

Economic Interdependence"?

Economic

Journal, vol. 95, September, pp. 619-636.

Girardin, E.

(1987) - Les effets de la politique budgetaire sur

l'equilibre

externe

dans

economies

des

interdependantes. These de Doctorat d'Etat. University

of Rennes I, June, .623 p.
I.H.F.

International

(1987)

Financial

Statistics.

Annual

Edition.
Harris. R.

(1987) -

Les deficits

et Ie

dollar

l'economie

mondiale en peril. Economica. Paris.

O.E.C.D.

(1980) - Main Economic Indicators (1960-1979)

O.E.C.D.

(1987a) - O.E.C.D. Economic Outlook, June

O.E.C.D.

(1987b) - Main Economic Indicators, July.

Price, R.W.R.

et Hijller

structurels et

P.

(1985)

- "Indicateurs budgetaires

interpretation de

politique budgetaire

des pays

l'orientation de la

de l'O.C.D.E.".

Revue

Economique de l'O.C.D.E. pp. 29-76.

Vinals, J.H.

(1986) -

"Fiscal Policy and the Current Account".

Economic Policy n03, pp. 712-744.

THE POLITICAL ECONOMY OF DEBT REPUDIATION

AND EXPROPRIATION IN LDCs

Hartmut R. PICHTThe Kiel Institute of

(Institut

fu~

Weltwi~schaft

The LDC

only a
~isk

debt crisis
of

te~ms

~elatively
add~ess

of the

c~isis

d~amatical

early 1980s

~isk

In

~ole.

debt

1980s has

been

this analysis an attempt is

of

inc~ease

the

f~om

sove~eign

solvency

p~oblems

diagnosed and explained as a

counter-intentional consequence of the implementation of

default clauses
and made

them

concept of

in LDC

banking which

manageable.

mo~e

sove~eign

~isk

is

majo~

played

conside~ations

p~ima~ily

is

onwa~ds

mainly

capacity of the

debt-se~vicing

sove~eign

mino~

the

pe~spective.

from the

the

while

count~ies.

made to

R.F.A.)

Introduction

analyzed in
debtor

Weg 120

Dusternb~ooke~

PO Box 4309 - 2300 KIEL

I.

Economics

Wo~ld

In

~educed

what

desc~ibed

sove~eign

follows

befo~e

c~oss

~isks

the

basic

the main

th~ust

of the analysis is outlined at some length.

Inte~national

as

p~ope~ty~ights.

I'\ational

domain.

c~edit

but

That

contractual

obligations

contracts depend
hono~ing

of

intertempo~al

The
~esea~ch

cont~a~y

debt

enfo~ceable.

claims

c~edit

is

(1).

As
of

obligations

cost-benefit

histo~ically
enfo~cement

cont~acts

to

becomes a
the

c~edit

whe~e

matte~

pa~t

the

enfo~ce

the

self-enfo~cement

calculus on

in the

exogeneously

exist

cor.~equence

autho~
g~atefully
acknowledges
assistance by Angela Husfeld.

(1) Note that

as a means of

acknowledged

not

inst~umentalities

on mechanisms

cont~actual

to

se~vicing

is. no

usually

a~e

of

the
the

of the debtor

ve~y

efficient

gunboat policy was occasionally used

1986 : 14).

(Bo~n

330

only.
between

The relationship
basically a
it

relationship between

involves

an

concluded and

asymmetry

the capital

substantial range
308-9). The

in

of discretion

debtors is

principals and agents,

that

after

the

is transferred,

relationship may

matrix for

creditors and

the debtor

(Jensen and

be described

i.e.,

contract

is

has

Meckling 1976
by a simple payoff

two players (Figure 1) (2). The agent (debtor) A is

assumed to have the choice between the cooperative (A1) and the
noncooperative strategy
contract. Similarly,
choose to

(A2)

with

respect

the principal

provide capital

to

honoring

(creditor)

(cooperative strategy

may
B1),

foreclose access to capital (noncooperative strategy B2)

first place,

i.e., before

the

either
or

to

in the

the contract is actually concluded.

The variables in Figure 1 denote the payoffs to the parties for

all four

combinations of

strategy of

agent A

choices.

It

is

the

noncooperative

which is at stake if one talks about debt

repudiation or willful default.

Figure 1 - The Gains of the Game

Agent's
A1
Principal's
Strategy

Strategy
A2

B1
B2

However, the

risk that debt repudiation is chosen as

the rational strategy on the part of the agent is only one form
of sovereign

risk that

for discussion.
LOC debt
basic

In order

crisis,

it

substitute

applies to

(2) For
( 1988) .

to understand the major cause of the

foreign

One has

and noncooperative

a related,

credit,

i.e.,

to distinguish

behavior on

the principal

this set

interest for the subject chosen

is argued that a look is necessary at the

of

investment (FOI).
government) and

is of

of players

formulation

the part

foreign

between cooperative
of the

agent

(external investor).
as well.

see

direct

The

(host

Figure

noncooperative

Chakravarty

and

Holler

strategy of

the agent

expropriations. The
expropriates is,

now stands

danger that

for large
a

host

scale or

government

macrofor

FOI

thus, the second form of sovereign risk taken

up in this analysis.
What matters
either case

is to

from the

principal's point

provide capital

of view in

only if the payoff for the

agent is expected to be larger in the cooperative than the noncooperative case

capital to

(d l

an LOC

some reasonable

>

d2

Further,

).

agent must

the gain from providing

be high enough in comparison to

alternative where sovereign risk is absent (Cl

> C3)'

In the

present picture

the simplifying assumption, of

course,

is that the principal is risk-neutral in the sense that

variations around expected payoffs do not matter.

The

game-theoretic

framework

just

presented

is

applied and specified in the subsequent analysis to explain why

the debt

crisis has

occurred.

It

is not claimed, however, to

present a complete picture.

It is.

first.

argued

that

with

respect

to

debt

finance the emergence of cross-default clauses in the 1970s had

reduced the
for the

sovereign risks

principals. while

with respect

to

and had made them more manageable

there was no comparable development

noncooperative

behavior

concerning

foreign

direct investments. Consequently. the relative price of the two

substitutes of

external LOC

finance which emerged as a result

of this innovation favored debt rather than equity finance (3).

Second,

the

relationship exists
and the
It is
equity

hypothesis

between the

is

advanced

that

structure of external finance

economic performance of the capital importing country.

argued that
finance

performance.

in

the shift

had

towards debt finance relative to

negative

influence

on

the

economic

particular on the aggregate real marginal rate

of return on capital. which is used as a proxy for the solvency

of a country.
Third, and
problems,

it

behavior with

(3) For
325).

as

is argued

result

that the

respect to

a compatible

new

of

the

induced

move towards

capital

statement see

(Figure

solvency

noncooperative
1)

Folkerts-Landau

seriously

(1985

332

threatens

the

international

financial

Substantial

system.

efforts are required to stabilize the system.

The presentation below is organized in four sections.
In Section II the agent's rationale to either cooperate or noncooperate

is

expressed

principal's perspective
behavior are

formal

the

choices were

sovereign

predictability

In

debt

of

From

the

noncooperative

the different

Section III

risk

of

first.

for the case of debt and FOI. Ten

determinants of

considered.

terms

probabilities

estimated both

potentially relevant
reduced

in

it is shown that the

exposure

and

repudiation

strategic

the

changed

increased

the

relative

magnitudes of the sovereign loss premiums. A Bayesian decisionrule calculus

this effect

is applied.
on the

structure of

adverse implications
It is
made

then shown

Section IV deals with the impact of

external LOC finance and the

for the economic performance of the LOCs.

why the

noncooperative

increased solvency

strategies

more

risks,

attractive

in turn,

again.

The

conclusions are presented in Section V.

II. The Choice between Cooperative and Noncooperative
Behavior
A. The Formal Condition of Noncooperative Behavior by
the Agent
Suppose that

A(t)

denotes

the

stock

of

external

claims <either debt or accumulated FOI) at time t, dA(t)/dt the

future capital inflow (outflow if negative) per unit of time,
the rate

of return

preference rate.

to

be

paid

Let dS(t)/dt

on

A(t),

be the

and

the

time

pecuniary equivalent of

potential future sanctions per unit of time other than the loss
of future

business that might be imposed upon the agent in the

case of noncooperative behavior.

Further,

suppose

that

captures

foreign capital A(t)

perceived

disutilities per

unit of

country. Thereby

iA(t) and eA(t) denote costs, while the other

invested in the

two components stand for additional future resources dA(t)/d(t)

(if positive)
time horizon

and avoided

additional losses dS(t)/d(t). For a

from the present P to the future F, the condition

of noncooperative

behavior on the part of an agent is given by

(1). Whenever the present net value is

(1)i:(dA(t)/dt + dS(t)/dt - iA(t) - eA(t

negative. the
agent. and

noncooperative strategy

the opposite

positive (4).
it is

holds.

if

exp - r(t-F

dt < 0

is advantageous for the

the present

net value

is

As future events cannot be known with certainty.

actually more appropriate to talk about the expected net

present value

being positive or negative. The calculus in this

general form

applies to

direct investments.

foreign credit

In what

as well as to foreign

follows 1(.)

denotes the partial

integral over the chosen time horizon (5).

For practical
is it

to whom

to this

this calculus applies. What are the constraints

decision-making

horizon?

And what

best case

relevant determinants

behind the

included in the formal calculus? In the

decision-making

constraints.

In

formulation is
drawn from

the

present

advanced. Ten

repUdiation and

the

paper

options

much

less

partial hypotheses
issue.

They are

and

its

ambitious

on the

debt

respectively.

put together to

are

explain

choice between cooperation and noncooperation on

of the

net present

its

unit.

expropriation

the literature.

the strategic

agent. The various determinants may drive the

value in

the negative.
positive

What is the appropriate time

one may be able to present a well-specified model of

relevant

the part

unit?

are the

fundamental variables
the

purposes the question arises as to who

sign.

which is

equation (1)
expressed in

respectively.

noncooperative behavior

more to the positive or to

since

decreases if

a negative
the
the

sign

and

probability
net

present

of
value

turns more to the positive. and vice versa.

B. The Determinants of Choice

(4) See Niehans (1986

formulation.

163)

for a

related.

but

different

(5) The time index of the variables is omitted for convenience

in the following presentation.

334

In what

follows,

noncooperation are
for the

in Section

hypotheses on

first presented

FOI case.

procedure (logit

the

In Section

cooperation and

for the debt case and then

C further

below the estimation

analysis) and the data basis is outlined, and

0 the

logit regression

results for

both debt and

equity finance are interpreted.

1. The Case of External Debt
Various authors
default issue,

such as

Bates(1984), to
however,

have presented
Feder and

note only

focussed

countries, and

more
If

les~

repudiation issue

Just (1977)

a few
on

models on

of them.

the

the

debt

and Saini

The studies

debt-servicing

and
have,

capacity

of

at all, on willful default. The debt-

was taken up by Eaton and Gersovitz (1981a ;

1981b), Sachs (1983), and Sachs and Cohen (1984). Complementary

and in

part competing

listed below

as H1

hypotheses

to H1D.

were

When

presented.

rival

They

are

interpretations

are

offered, an additional letter is added.

(H1) The
long-run income

first hypothesis

expectations on

strategies. Eaton

and Gersovitz

income prospects

allow that

relates to

the choice
(1981b;

the impact of

between

the

two

8) argued that high

current consumption

is

divorced

from current income,

i.e., they allow for higher consumption in

the present

expense of future consumption. Governments

at the

may be induced to borrow from abroad, and pay back later on the
assumption that
its

the marginal

constituency

is

high

utility of income on the part of

when

income

is

low.

If

debt

repudiation in the presence of cross-default clauses means that

the probability
obtained from

is high
the

that no future credit I(dA/dt) can be

banking

intertemporarily, then

community

to

smooth

high-income prospects

risk of

noncooperative behavior

is that

"rapidly growing

consumption

will

lower

the

(H1a). A competing view (H1b)

countries may have less incentive to

repay loans, since they do not expect to enter the market again
after the
(Eaton and

period in which a net payment of loans is necessary"

Gersovitz 1981b

future external
net present

: 16).

credit I(dA/dt)

value of

In this case the demand for

is assumed to be low, and the

future capital

flows may

move

to

the

335

negative (equation (1.

(H2)

The

second

hypothesis

refers

to

another

conceivable circumstance where the threat of foreclosing future

access to

the international

community in

the case

credit

markets

by

the

banking

of debt repudiation may be matched by a

counterthreat. A

single borrowing

his odds

away with no or only modest costs are higher.

to get

agent may

well reason that

if he decides to repUdiate debt when other governments do so as

well. Parallel behavior even of a number of smaller debtors may
create a

fairly large

problem for

the banks.

although

each

single one of them would not have seriously affected the banks.
Under these

circumstances banks

might prefer to compromise on

sanctions and even accept substantial lossess.

the

future

credit

forgone

I(dA/dt)

may

relatively small. The hypothesis. thus.

In formal terms.

then

actually

be

is that the probability

of willful default by the agent is higher when others decide to

repudiate as well (bandwagon). For Eaton and Gersovitz (1981b :
14) it

is the

high economic

simultaneous defaults.
here in

that the

for parallel

interdependence

Their hypothesis

which

explains

differs from

the one

interrelatedness of LOC economies is crucial

debt repudiation.

while it is not in the present

formulation.
(H3) Another hypothesis pertains to the international
importance of

a country.

The standing

of some

countries

in

relation to potential sanctions may be much better than that of

others. Various

ways of

are conceivable.
position in

measuring the importance of a country

of course.

one

of

which

is

the

relative

world trade. Population size may be looked at as a

leading indicator

with respect

to future

world

trade.

Hard

sanctions at the occasion of debt repudiation may damage future

relationships and.

hence. are

important countries.
be relatively
positively with

In formal terms.

small for

probability that

less likely

to be imposed upon

I(dA/dt) and I(dS/dt) may

"more important" countries. Thus. the

the noncooperative
the political

strategy is chosen varies

weight of

a country (Eaton and

Gersovitz 1981b : 19).

(H4)
variations

in

investment.

in

It
real

has.

fourth.

absorption.

response to

been

argued

i. e .

external shocks

that

consumption

large
and

are rather costly

336

(Eaton

and

Gersovitz

adjustment costs,

8-9) .

I98Ib

governments may

In

have a

order

to

reduce

strong incentive to

borrow from

abroad to facilitate transition if their economies

are subject

to severe

external shocks.

It follows,

in turn,

that the

probability of willful default is lower in this case,

since in

the light

foregone future

of effective

credit I(dA/dt)

cross-default provisions the

is potentially

large. Notice

that large variations in the export performance as an indicator

of the

vulnerability may

also mean

that

the

country

faces

liquidity problems.
(HS) A further explanation relates to fluctuations in
the domestic
one

economy. The

forward

put

as

consumption patterns

argument parallels

HI.

Governments

strongly to the

may

wish

to

smooth

in the short- or middle-run on the ground

that the marginal utility of income is high when income is low,

and vice

versa.

If

markets I(dA/dt)

future access

would be

to the international credit

severely hampered

in the

case

of

noncooperative behavior, again in view of an effective sanction

scheme, one would expect that the probability of this option is
lower in the case of larger income variations.
(H6) One

may also expect that the gain to be made by

debt repudiation depends on the size of the credit outstanding.

The larger

the value of the resources not to be paid back, the

higher is

the probability

chosen (H6a).
I(iA) and

Th i s app 1 i es,

of course,

to interest

payments

outflow of principal payments (included in I(dA/dt))

(Eaton and

Gersovitz I98Ib

amounts of
agent in

that the noncooperative strategy is

credit may

also reflect

the international

conclusion,

i.e.,

: 302).

On the

other hand,

large

the credit standing of an

market. This leads to the opposite

governments could

speculate to

get hold of

even more external ressources (I(dA/dt)), before noncooperative

behavior becomes an advantageous move (H6b).
(H7)A further hypothesis relates to the dependence of
a country

on imports and, hence, credits from exporters (Eaton

and Gersovitz

I98Ib:

9).

In

the case

exporters may

wish to

continue trade

of

willful

only on

a cash

default,
basis,

because they fear that export credits may be subject to default

as well,
by

i.e., external

default

decision.

transactions may severely be affected

The

probability

of

noncooperative

337

behavior is,
country on

thus, negatively
imports. An

related to

the dependence of a

alternative interpretation advanced by

Sachs (1983

: 20)

embargo.

formal terms I(dA/Dt) and I(dS/dt) in equation

In

is that

of

the

of

p~ssibility

trade
(1)

are affected positively.

(H8) A

further argument

(1985). He

argued that

higher when

income is

rationale is
assumed to

this:

- would

The hypothesis,
low. This
and the

Lachler

expected.

higher when

The
it is

the actual income of the

in equation (1)

of an unexpected income drop.

that

the

probability

the actual

of

debt

income is unexpectedly

makes the more sense, the shorter the time horizon F

larger the

time preference rate r is. This hypothesis

and contrasts

Lachler's
includes

changes. H5
that it

prev i ous 1 y

related to

in situations

H5. The

reasoning

variations, whereas
(H5)

by

Whatever form the penalty may take,

thus, states

both parallels
that

than

The sanction costs - I(dS/dt)

be lower

repudiation is

presented

the probability of debt repudiation is

lower

be positively

agent's country.

was

refers

the argument

anticipated

as

contrasts with

would state

critical difference
to

unexpected

income

of intertemporal consumption

well

as

unanticipated

the hypothesis

a negative

is

income

presented here

relationship

also

when

in
the

income is unexpectedly high.

(H9) Another
"other" sanctions
debt repudiation

hypothesis addresses

on noncooperative
it may

rand/or private

charity

the magnitude

behavior.

In

of

the case of

well be the case that the governments

organizations)

of

affected

lending

countries may not be willing to grant further aid payments, and

they may also use their influence to reduce future aid payments
by international
vary with
the

organizations.

In

other words,

I(dS/dt) may

the dependence on aid. The hypothesis, thus,

probability

of

noncooperative

behavior

is

is that

negatively

correlated with high,aid levels.

(H10) The

final hypothesis

relates to

the level of

the marginal rate of return on capital in the debtor country in

comparison to the costs of external funds.
positive, then
is, the
or

If the difference is

a profit can be made. The higher the difference

higher is the amount of resources that can be consumed

invested

on

the

part

of

the

borrowing

country.

The

338

difference between
of the

the two rates is, thus, another determinant

choice between

I(dA/dt)

is

marginal rate
costs

Gersovitz 1981b
negative

noncooperation,

since

likely to correlate positively with the difference

between the
borrowing

cooperation and
of

return

(Folkerts-Landau
: 8).

on

1985

The hypothesis,

relationship

exists

noncooperative behavior

330
thus,

between

and the

capital

the

Eaton

and

states

the

difference

and
that

probability
between

the

a
of
two

rates (6).
In Table
of cooperative

1 the set of hypotheses on the determinants

and noncooperative

behavior on the part of the

debtor are summarized.

Table 1 - Hypotheses on Debt Repudiation Risks (Summary)
Signs of Impacts
Repudiation
Determinants of Noncooperation
Risk
I(dA/dt)
I(dS/dt)
I(iA)
I(eA)
HI (WRGDYL)
H2 (AND)
H3 (POP)
H4 (EXPVARL)
H5 (GDPVARL)
H6 (GDL)
H7 (IMPQL)
H8 (DWRYL)
H9 (AIDCRQL)
HI0 (DZKPL)

-/+
+

+/+
+
+
-/+
+

+/-

+
+

+
+
+

Note : The variables in parentheses are the proxies used in the

empirical part.
The signs

indicate whether the probability of debt repudiation

and

components

the

negatively with

of

condition

the determinants

(1)

vary

denoted in

positively

or

parentheses. The

exact definitions of the proxies used for the empirical testing

below are given in the Appendix.

(6) One may note. however, that from the portfolio theory point
of view a higher rate of return may simply be a compensation
for the
higher systematic
risk associated with an LDC
investment (Sharpe 1970: 97). The implicit assumption here is
that borrowers from LDCs accept a certain trade-off between
risk and return.

339

2. The Case of Foreign Direct Investments

In order
hypotheses on
was to

to understand

noncooperative behavior

provide the

is comparable

Similar attempts

have been

Schneider

hypotheses are

were reviewed. The goal

theoretical basis

macro-level that
(1986),

the expropriation issue, ten

and

for expropriations on a

to the

debt repudiation case.

undertaken

Frey

labelled as

(1985),
Gil to

by

Burton

and

G20.

Frey

In

and

Inoue

(1985).

The

the case of rival

interpretations a letter is added.

(G1)

The

stability (7).

to the

the basic
: 9).

refers

subject

to

political

presence

of

to
large

forces

foreign

political
internal

may,

capital

among
in

the

It was argued that expropriation risks depend upon the

visibility of
over the

be

conflict between

relate

country.

hypothesis

Countries may

tensions. The
others,

first

foreign investments

popular concern

in the

host country, where

is that foreigners may gain control

destiny of the residents (Sherk 1974 : 99, Mason 1974

In a

stable political

acknowledge the
presence of

balance of

environment a
the benefits

government

would

it may gain from the

FOI and the resentments against foreign dominance.

This balance

may translate

into some stable ratio of "foreign

to nationally owned capital" (Hanson 1979 : 616) or of "foreign

productive capital" (Feder and Regev 1975 : 322).

and national
Stoever talks

about the

emergence

towards foreign investment (1981

If it

could be

stable capital
adjustment of

love-hate

attitude

: 8).

shown, for instance, that there is a

at the same time a stable pattern of

the actual

FOI levels in response to changes of

clear

then one

about

its

noncooperative behavior
cases where

ratio and

total investments,
is quite

of

the

upon the

adjustment to

new ratios

dealing

with

FOI.

The

risk

of

is then considered to be lower than in

internal

tensions may,

could argue that the host country

tensions

are

large,

because

occasion of a change of power,

(Basche 1979

translate into

stop-and-go policies.

I(eA) captures

the effect

discussed.

: 6-7),

the

require

and they

may

In terms of equation (I),

The

hypothesis,

thus,

(7) See Jones (1980), Kobrin (1978), Robock (1971), Feierabend

and Feierabend (1966), Oavies (1981) and Gurr (1968).

340
states

that

stability and

negative

relationship

consistency of

policies

exists
towards

between
FDI

probability of noncooperative behavior (Root 1968

The

(G2)

withdrawal of

next hypothesis

aid to
: 102-7

payments

; i(eA

(included

follows

the

70-71).

the

potential

1985:

may have
may be

191-2). The

hypothesis

level is,

the support

foreign

foregone future aid

offset by
The

gains an

confiscating

from

l(dS/dt.

in

The higher

bilateral or the multilateral level

Jodice

expropriating government
assets (IliA)

and

expropriating countries by affected home

governments, either on the

(Knudsen 1972

relates to

the

reads

as

the lower is the

risk of noncooperative behavior on the macro level.

This

(G3)

hypothesis is

relates government
economic policy

behavior to

G1

in

that

it

the national constituency. The

performance of
of doing

similar to

a government is generally hard

to assess.

One way

so, however,

is to

look

at

country as

a member of a reference group of similar countries.

The conjecture put forward here is that governments in LDCs are

assessed by

their consistuencies

performance within

the LDC

noncooperative behavior
performance is

on the basis of the relative

country group.

The probability of

is assumed to rise if the own economic

lagging behind,

since then

the

policy-makers

will at least pretend to do something about it in order to stay

in power.

Multinational cooperations

the scapegoat
talked about

or whipping boy function (Davies 1981 : 6). Gurr

"relative deprivation"

drive governments
called it

and their FDls take over

(1968:

as a major force that may

1104), and Feierabend and Feierabend

"systematic frustration" (1966 : 250).

terms of

equation (1),

r may

In the formal

be large and the time horizon F

short, so that future events get largely discounted (8).

(G4)

The

question

was

may

as

trigger

to

whether

unanticipated

income

behavior. The

argument put forward by Lachler (1985 : 31-2)

that there

increases

raised

noncooperative
is

is a

positive relationship between the probability

of expropriation

and unexpectedly high realized incomes, since

in these

situations the

value

of

the

assets

expropriating

(8) To be sure, governments are not conceptualized as to decide

necessarily in the best interest of the country.

~1

governments may

get hold

definition residual

of is

larger. Equity

claims are by

claims which depend on income and profits.

In formal terms, the rate of return on investment i

(1)

may

then

be

translates into
value. Since

assumed

a lower

be

relatively

and potentially

the focus

argument matters

to

is

more if

on

which

negative net present

unexpected

the time

in equation

large,

developments,

the

horizon F is short, and if

the time preference rate r is high.

(G5)

Income

influence on
(1979:

profit

the probability

10) argued

the prospects
it is

and

expectations

may

have

an

of expropriation as well. Basche

that the expropriation risk is higher when

for future

profits are high (G5a).

In this case

more likely that the value of the assets seized and thus

the rate of return as indicated by i in equation (1)

which lowers

the net

present

value.

is larger,

Alternatively,

it

was

argued by Jones (1984 : 83) that expropriation risks are larger

in cases

of sluggish economic performance with poor prospects,

since in

these situations

profits I(iA),
foreigners in
There are,

a government

which might
order to

thus, two

may wish

to take the

have otherwise been transferred to

stimulate economic
rival hypotheses

performance (G5b).

that

are

related

to

future economic performance.

(G6) Similar to the debt repudiation case in H5, high
income variations
expropriation as
argued

that

may also

create an

strategy.

large

In

the present
variations

income

respective economy

is mono-structured.

income variability

is to

attract new

get rid of external investors.

assumed to

be larger

the hypothesis
behavior

is

incentive not to choose

context it
indicate
One

way

may be

that
of

the

reducing

business rather than to

In formal terms,

I(dA/dt) may be

under these circumstances. Consequently,

states that
smaller

in

the probability
the

presence

of noncooperative
of

large

income

variability.
(G7)

The

noncooperative

next

behavior

hypothesis
with

on

respect

cooperative
to

foreign

and
direct

investments picks up the size of a country and, thus, finds its

parallel in

the debt

counteract potential
much better

case

as

well.

sanctions on

larger

the part

country

of the

may

investors

than a small country, since it may playa decisive

342

role in

other areas of interest in the international arena,

not in the present, then at some point in future.

be argued

that the

expropriation risk

the weight

of a

assumed to

vary negatively

country.

In

It may, thus,

varies positively with

formal terms,
with the

if

I(dS/dt)

importance of

may

be

a country

and, thus, make the net present value smaller, everything being
equal.
(G8) The
also be

assumed to

stock invested
was

probability of

argued

in

reference is
country and

vary positively

in an

economy,

G1,

explanation is

an expropriation

event may

with the absolute capital

if domestic investments are, as

reasonable

benchmark

for

FOI.

This

introduced because in the empirical tests below

made to

an absolute number of expropriations per

year to

define

expropriation

as

macro-event

comparable to that of debt repudiation. The hypothesis relates,

thus, to the term e,
(G9)

just as does hypothesis Gl.

Another

cooperative and

payments pressures.
first one

determinant

of

the

noncooperative behavior

is that

between

Two rival interpretations have merits. The

in situations

payments problems

choice

relates to balance of

of extraordinary balance-of-

pressures could

be reduced

by transferring

the ownership

of FOI to residents, since then foreign exchange

drains caused

by the

The hypothesis,
expropriation
(G9a)
6)

thus, states
increases

(Stoever

(9).

repratiation of

1982:
at

thus, states
lower at

it

times

of

(G9b).

of foreign

that the

times of
In

horizon chosen

macro-

pressures

may
large

well

be

external

argued
imbalances

i.e., expropriations
exchange. The

probability of

that

may cause

set
a

counterhypothesis,

macro-expropriations is

extraordinary balance-of-payments pressures

formal

interpretations,

probability of

balance-of-payments

11 , LLoyd 1974 : 29 , Truitt 1974 : 45-

counterintentional signals,
further drainage

that the

with

Alternatively,

expropriations

profits may be reduced.

it

terms,
appears,

and the

the
is

difference
a matter

between
of the

the

two

actual time

time preference rate r of the relevant

decision-making unit.

(9) It may be noted that residents may also wish to transfer

foreign exchange proceeds abroad.

343
(Gl0) The
refers to

the developmental

this hypothesis
creates a
may be

last hypothesis

there is

income economy

risks

stage of an economy. According to

some level of per capita income that

crucial reference.

proxied by

on expropriation

The magnitude

the mean

per capita

of this reference

income of the middle-

group of the World Bank (1985 : 174). A country

with an

income close

said to

choose the

countries which
capita incomes

to that level. either below or above.

expropriation strategy

have either
A

much higher

more
or

readly

much

curvilinear relationship

is

than

lower

per

is said to exist

(Burton and Inoue 1984 : 412-3). Basically it is argued that by

approaching the
made by

reference income both the potential gain to be

instrumentalizing FDI

development process
rising, whereas
firms is

and the

in order to direct the national

capacity to

the instrumental

run those

usefulness

of

firms

is

nationalized

assumed to decrease beyond the reference income since

economic interactions

get more

terms,

to

I(

eA)

seems

be

and more
a

complex.

function

of

In

the

formal

stage

of

development.
The summary table for the hypotheses on expropriation
risks is

presented as

whether the
to

vary

Table 2.

The signs

displayed indicate

probability of an expropriation action is supposed

positively

discussed in

or

hypotheses Gl

negatively
to Gl0.

with
The

the

determinants

definitions

of

proxies in the parentheses are displayed in the Appendix.

the

344

Table 2 - Hypotheses on Expropriation Risks (summary)

Signs of Impacts
Determinants of Noncooperation
I(dA/dt)
I(dS/dt)
I(iA)
I<eA)

Expropriation
Risk
(DUM)
(AIDCRQ) (WRYM)
+
(DWRY)
(WRGDY)
+/(GDPVAR) (POP)
(IDBSR)
(BOP)
+/Gl0 (DGDPKA)-

Gl
G2
G3
G4
G5
G6
G7
G8
G9

Note:

+
_

+
+

-/+

The variables
in the parentheses are used as proxies.
MNote that a large r puts a larger discount on flows in
the more distant future.
It can also be read as
reflecting a
relative short time horizon on the part of
the agent.

C. Logit Model as a Multivariate Test Procedure

It

has

already

theoretical model

been

exists that

said

that

no

would capture

comprehensive
all the

effects

discussed above

simultaneously. The single determinants of the

choices between

cooperative and

all related

to the

crucial variables

noncooperative behavior
relationships were
form. For

debt

the

the

specified in

repudiation

probability

dependent variable

in

equation

were

condition

for

(1).

the

but

functional

purpose the various determinants of the

(Gl-Gl0). respectively.
where

displayed in

not explicitly

the present

probability of

noncooperative behavior

(Hl-Hl0)

and

expropriation

were put into a multivariate framework

of

noncooperative

and proxies

were chosen

behavior

is

the

to operationalize

the sets of hypotheses.

Basically.
purpose. The
can be
was set

logit analysis was applied for the present

dependent variable

measured in
"1" if

is a

noncooperative behavior

cooperative behavior

discrete variable which

either-or-categories only.
was observed.

i.e . the dummy

prevailed. and "0" if

Logit analysis has several

~5

advantages in comparison to other methods. For example, the use

of a simple linear regression may generate outcomes for the
dependent variable (probabilities) below "0" and above "I" if
used for predictions (Pindyck and Rubinfeld 1981
275-80). The
non-linear transformations suggested by probit and logit models
avoid these

problems. While

probit and

logit

analysis

comparable (Altman,
logit

analysis

iteractions are

the
for

equal

Avery, Eisenbeis
offers

results

data

sets

for

are

and Sinkey 1981

computational

necessary. An

comparable studies

estimation

quite
31-3),

advantages

alternative technique

of

since
used

in

is the multiple discriminant analysis (e.g.

Frank and

Cline 1971). The major advantage of logit and probit

models in

comparison to multiple discriminant analysis is that

it avoids

a priori classifications into either cooperative and

noncooperative agents,

and that it allows a direct test of the

significance of the coefficients estimated (10).

The logit analysis was done on the basis of pooled
data, i.e., a data sample was put together from cross-national
and time-series information. A sample of ten countries notably
from Latin America and Asia was drawn for the present purpose
(II). Of course, the data shortages for testing two sets of ten
different
severe.

hypotheses
While

this

relatively small,
example, were

(HI-HIO,
explains

countries like

not taken

tensions during

the time

GI-GIO)
in

simultaneously

part

why

Nicaragua

into account,

the
and

because

period covered

sample
Turkey,

the

country

reference to
thus, part

capture potential
of the

set of

is
for

internal

by this analysis had

reached a level where empirical testing on the ground

regular behavioral incentives did not seem plausible.
Argentina, Brazil and Mexico
sample, since hypothesis H2

were

of

are not part of the

required an external

parallel behavior.

determinants of

They

are,

the choice between

(10) The functional form of the logit model applied is P(I) =

[I + exp (b o + ... + bJX J ... + bnXn)l exp - I, where "1"
denotes the noncooperative event, XJ the various determinants,
and b J the regression coefficients.
(II) The country sample consists of Colombia, Costa Rica,
Ecuador, Honduras,
India, Korea, Morocco, Peru, Sri Lanka and
Thailand.

346
cooperative and

noncooperative behavior.

mentioning that

one half

seventeen most

of the

heavily indebted

Finally,

country set

it is worth

belongs to

countries (World

the

Bank 1986 :

XXV), while the other half does not.

Moving from

the theoretical

level to actual testing

means also that one has to make definite decisions as to how to

define the variables and proxies. One major problem is posed by
the

dependent

variable.

rescheduling events
been triggered

First,

data

are

available

for

(RSCOG), yet these events may have equally

by solvency

or liquidity considerations. There

is, thus, a bias in this measures for the dependent variable in

the debt

case.

However,

possible explanations
contrary,
the

it

on

the

of default

is argued

hypotheses

in

present

analysis

were taken

not

all

up. Quite on the

below that for the time period covered

willful

default

do

in

fact

provide

reasonable explanation.
Second,

in

order to

dependent variable,
dimensions, one
case.

In

had to

absence of

expropriations of
considered.

get the magnitudes right of the

the reschedulings
decide on

data on

FOI, the

all

had

macroeconomic

the equivalent

the actual

in the

volumes involved

FOI
in

number of expropriation events was

In practice there are formal expropriations, extra-

legal transfers

of foreign

assets by public or private actors

(interventions), forced sales in whole or in part, and contract

negotiations which
ownership (Kobrin

finally lead
1980:

used which

were made

all

forms

four

whenever three
year, the

to an

68-9).

effective transfer

of

In the present study data are

available by F.N. Burton. They relate to

of

or more

condition for

involuntary

divestment.

expropriation events
macro-expropriation

Specifically,

were listed
was

assumed

per
to

prevail (12). Finally, the time periods covered are 1976-85 and
1961-77 in the debt and the FOI case, respectively.
With respect

to the

variable specifications used in

(12)
This assumption seems plausible,
since expropriation
actions may upon occasion be motivated by other than macropolitical configurations.
It may be noted that the empirical
results did not change much when the critical value was raised
to 4 or more expropriation actions per country and year. For a
thorough discussion on measuring involuntary divestiture see
Williams (1975), Kobrin (1980), Burton and Inoue (1984).

347

this analysis the reader is again referred to the variable list

in the

Appendix. Several

cases the
G8)

in

reciprocal of

order to

model more
linear

the variable was used (e.g.

meaningful. Since

transformation

in

approach zero.

variable name

the logit

the

sure that

noncooperative behavior

appropriate.

in some

in H6 and

make the economic interpretation of the logit

transformations make
and G8

remarks are in place. First.

first

model implies
place.

the partial

approach zero
Second. an

these

variable

probabilities

if GDL

"L"

a non-

was

of

and IDBSR in H6
attached

to

the

where information and decision lags seemed to be

The

choices

should

be

taken

as

first

approximation.
Third. with
of political

respect to

instability on

the hypothesis on the impact

the probability

(G1). additional

econometric calculations

well-known stock

adjustment

stocks of

FDI. Let

model

SFDT(t) be

of expropriation

were necessary. The

(Koyck)

was

applied

for

the stock of foreign assets at

time t.

and SFDI*(t) be the desired stock of foreign assets at

time t.

According to

change in

the stocks.

the stock

adjustment model

or the net FDI in t.

the

actual

is proportional to

the gap between the current desired level SFDI*(t) and the past
actual level. which is SFDI(t-1)
(2 )

SFDI(t) - SFDI(t-1)

(equation 2).

T(SFDI*(t) - SFDI(t-1

where 0 < T < 1

The conjecture

along the

lines of Feder and Regev (1975). and

Hanson (1979). was that there may be a desired level of foreign

owned assets in relation to the domestic capital stock. Here it
is hypothesized that
( 3) SFD I * It )

6K(t)

Equations (2) and (3) can be rearranged to get (4).

(4) SFD I ( t )

(1-T)SFDI(t-1) + T6K(t)

Taking the first differences one gets

348
(5) FDI (t)

(l-T)FDI(t-l) + T6I(t)

where K(t)

K(t-l) = I(t). and SFDI(t) - SFDI(t-l)

Equation (5)

sample.

In

was tested

for

the

FDI(

selected

t)

country

cases where the significance level of the estimates

was sufficiently
country to

high. the

reflect stable

cases the dummy

dummy value

"1" was assigned to a

investment conditions.

in all other

was set "0" (13).

D. Empirical Evidences
The

findings

for

the

debt

repudiation

case

are

presented in

Table 3. The overall quality of the regression is

indicated by

the average

reasonable (0.86).
corresponding to

likelihood (AVG.LH.), which is quite

The t-ratios
hypotheses HI

for the
to HI0

single

coefficients

are displayed

in

the

parentheses.

(13)
The criteria actually applied were
(a)
correct and
plausible signs for the parameters. and (b) an adjusted RSquare
of 0.5 or more as an
indication of
the overall
explanatory
power of
the stock adjustment model. Morocco was excluded due
to an
insufficient data base. for Sri Lanka and Thailand the
expected mean adjustment
lags E(I)
were considered to be
unrealistically high and low.
respectively. The procedure as
well as the classification derived should be taken as a first
approximation only.
The regression results are provided by the
author on request.

349

Table

Logit Estimates
Repudiation

(Constant)

the

Probability

of

Debt

2.57
(0.71 )
-0.83
( -1 .89) 2.88
( 1 .88)8.36
(0.78)
19.22
(2.19)--23.02
(-2.06)--12.99
(-1.31)
-13.22
(-1.39 )
-0.07
(-0.44)
0.03
(0.10 )
0.05
(0.42)
0.86

Hl
(WRGDYL)
H2
(AND)
H3
(POP)
H4
(EXPVARL)
H5
(GDPVARL)
H6
(GDL)
H7
( I MPQL)
H8
( DWRYL)
H9
( AIDCRQL)
HI0
( DZKPL)
(AVG. LH. )
Note

for

For definitions see the text and the variable list in

the Appendix
; -denotes a significance level of 0.10,
and -- one of 0.05.

Source

See variable list in the Appendix.

Applying

significant at
hypotheses

the 10

have

to

insignificant. The
data confirm
prospects

two-tai led

the debt

be

which

negatively

with

(willful

as

the

are
other

statistically

stated
the

debt

that

high-income

probability

defaults)

are

that
chosen.

governments tend indeed to be encouraged to choose

repudiation strategy

Mexico, Brazil
balance of

hypotheses

or better

considered

Hla

noncooperative actions
Second, LDC

percent level

four

main findings are the following. First, the

hypothesis
vary

test,

and Argentina

if

other

- take

the

governments
lead

(H2).

here
Third,

payments problems seem to have a positive impact on

the probability

of noncooperative

hypothesis that

domestic income variability as a proxy for the

interest in

smoothing consumption

negative impact
confirmed.

behavior (H4).

Fourth, the

intertemporarily (H5) has a

on the probability of willful debt default was

350

Further,
positively with
indicated by
competing

size

the

of

the probability

a negative

hypotheses

the

external

debt

varies

of debt repudiation, which is

sign of the coefficient. Given that a

were

advanced

as

H6a

and

H6b,

the

significance level of 0.20 is not too bad. Finally, there is at

least some

evidence for

probability of
the agent

on imports

measure what

is at

export credits,
sanction in

a negative

correlation

noncooperative behavior

and the

between

the

dependence of

(H7). The variable IMPQL was supposed to

stake with respect to the denial of future

or else,

the case

if a

trade embargo

of noncooperative

is imposed as a

behavior.

All

other

hypotheses did not pass the test (14).

Turning now
for the

to the

expropriation case

estimated logit regression model

(Table 4),

it becomes immediately

clear that the overall quality of the regression as measured by

AVG.LH.

is

much lower

only three
in a

than in

hypothesis were

the debt case (0.76). Further,

confirmed by the present data base

two-tailed significance

test at

a level 0.10 or better,

one of which is the constant.

Starting the interpretation with the constant, one is
easily lead
the part

to conjecture

of the

that rational

decision-makers as

economic calculus on

implied in all hypotheses

presented is a misspecification (15).

If only the constant were

significant,

hypotheses

then

obviously

the

advanced

as

potential explanations of expropriation decisions would not add

much to the understanding of the observed phenomenon.

(14)
The
population size variable
introduced to capture
hypotheses H3 was not adjusted for the fact that India had a
population more as ten times as high as that of Thailand, which
followed next.
(15) Note again that the condition of whether the cooperative
or the noncooperative strategy
is chosen in equation (1) does
not assume that anyone maximizes social welfare.

351

Table 4 - Logit Estimates for the Probability of Expropriation

(Constant)
Gl
(DUM)

G2

(AIDCRQ)
G3
(WRYM)
G4
(DWRY)

G5

(WRGDY)

G6

(GDPVAR)

G7

(POP)

G8

(IDBSR)
G9
(BOP)
GlO
(DGDPKA)
AVG.LH.
Notes

-3.51
(-1.75)-1.84
(-1.19 )
-0.25
(-0.81)
-0.21
(-2.02)-0.28
( 1 .79)0.17
( 1 .07)
-2.35
(-0.75)
-0.31
(-0.06)
-0.43
(-0.60)
0.06
(0.36)
-4.39
(-0.33)
0.76

For definitions see the text and variable list in the

Appendix;
- denotes a
significance level of 0.10,
and -- one of 0.05.

Sources

See variable list in the Appendix.

The only hypothesis confirmed at a satisfactory level

was G3.

This hypothesis

performance in
makers in
become a

stated that

a poor relative economic

the LDC country group puts government decision-

a situation where FDIs by multinational corporations

target. Worth

mentioning is G4 which stated that the

probability of expropriation is higher when the actual economic

performance is
expected. Note,

relatively strong

in comparison

to

what

was

however, that the corresponding coefficient is

significant only at the 0.10 level.

As in

the debt

case, all

other hypotheses have not

been confirmed by the present data set. One is, thus,

the impression

that the

knowledge about

the

left with

agent's

choice

between cooperation and noncooperation is very limited (16).

(16) The partial logit estimates for HI to HIO, and GI to GIO

as well
as the correlation matrices of the explanatory
variables are provided by the author on request.

352
III. The

Effect

of

Cross-Default

Clauses

on

the

Relative Sovereign Risk Costs

A. Empirical Background and Hypothesis
The empirical
and

FDI

in

the

repudiation can

previous

be better

indicators for
which are

section

have

predicted than

shown

that

debt

expropriation.

The

this assessment were, first, the AVG.LH. values

0.86 in

second, the

estimates of the logit models for debt

the debt

number of

understanding

of

and 0.76

in the

hypotheses which

the

choice

FDI

case

and,

contribued towards an

between

cooperative

and

noncooperative behavior at acceptable significance levels.

It is
innovation in

striking that
the 1970s,

obviously the

i.e.,

the

major

introduction

financial
of

cross-

default clauses,

had been extremely effective. Out of the four

hypotheses which

were confirmed

previous section,
threat of

not less

denying future

markets (Table

by the data set chosen in the

than three
access to

3). Cross-default

were

related

to

the international

the

credit

clauses specify, to be sure,

that a loan will be considered to be in default if the borrower

defaults on any other loan (World Bank 1985 : 114). This result
is remarkable,

since

enforcement of

property rights had emerged in absence of third

party enforcement,
debt case

it

means

which both

(see condition

(1

that

mechanism

reduced sovereign
and

made them

of

self-

risks in the

more manageable

(predictable) from the principal's point of view. No comparable

development

was

observed

in

the

case

of

foreign

direct

investments as the major substitute for credit.

In what

follows it

cross-default clauses

is shown

translated into

that the
a change

adoption
of the

of

price

charged for accepting sovereign risks, which is crucial for the

choice between
the

the two

determinants

(Section B).

of

types of
sovereign

LDC finance.
risk

costs

In a first step
are

presented

353

B. The Bayesian Approach to Sovereign Loss Premiums

In the
there were

initial set-up in the introduction (Figure 1)

investors) and

the agents

decision problem
whether or

for a

(debtors or

principal is

(creditors or

direct

host governments). The

not only

to

decide

on

not to engage in any form of business with an agent

who mayor may

game he

the principals

two players,

not cooperate later on. Obviously, within this

has also

to assess

noncooperation occurs,

what the

and whether

costs are

and at

if

actually

which costs he can

(individually) exit from an existing contract,

i.e., whether he

can pass on the contractual rights and related risks to someone

else. For

any principal who is willing to pass on capital to a

sovereign

agent,

required which

compensation

covers the

charge

(loss

expected costs

premium)

is

associated with the

possibility of noncooperative (debt repudiation, expropriation)

behavior.
Ignoring risk-aversion
for

now,

what

matters

configuration of

in

the cooperative

are chosen

by the

prediction of

to

first,

the

choices are,

with which

agent

in the portfolio-theory sense

addition
and the

and,

the agent's

the

costs

of

any

prior probabilities

noncooperative strategies

second,

the

quality

of

the

actual behavior once a contract has

been concluded in the first place.

In formal
agent's initial
by Al

and 81

monitoring

exist for

alternative without
situation

where

the

the

agent

the principal
contract.

alone

and

the
i.e.,

costs

may

be

in comparison to the best

(e,). State e2 denotes the
decides

to

move

to

the

Consequently, e2 denotes the expected

if he

had erroneously

adhered to the

Alternatively, the principal may wish to

Thereby,

e3

principal if

the agent

stayed with

(Al), and

represent

the loss,

e4

principal's

information

the principal

cooperative strategy.
exit

and

sovereign risk

noncooperative strategy.
loss for

the

described by cooperation,

in Figure 2. Even if there is cooperation on both

sides, additional
assumed to

terms, suppose

positions are

denotes

the

loss

the cooperative
if

he

had

moved

to

the

strategy
to

the

354

noncooperat i ve strategy (A2)

(17).

Figure 2 - Sovereign Losses for a Given Capital Transfer (C)

Principal's
Strategy

81
82

From the

Agent's Strategy

Al

A2

principal's perspective all possible states

outlined in Figure 2 are perceived with some probability, where

the

assumption

is

conjunction with
logit analysis.
in

Figure

that

recognizing that

(d j

the

Consequent ly,

rules

(d j

the

decided to

probability
move

from

moves from

cooperative

(1-[3)

R (A2, d j
(6)

(7)

to

A2

Al

(beta-error).

are the probabilities of making

Agent's Strategy
Al
A2
cd d j )
(1-[3(d j
[3(d j )
(I-cdd j

to assess

the costs

of

engaging

in

the

business, one can calculate the expected losses

decision rule

Al

not

81 to 82, although the agent

strategy

and (I-cd

81
82

In order
sovereign risk
i.e.,

in

The Probabilities of Making Right and Wrong Choices

by Applying a Decision Rule (d j )

Principal's
Strategy

for any

of

Al

the correct predictions by using some decision rule d j

Figure 3

in turn, denotes the probability that the

principal erroneously
with

some

of each state are presented

denotes

the agent

(alpha-error). Beta,
stays

follows

The probabilities
Alpha

3.

he

a forecasting or early-warning procedure like

and AZ,

dj

for

which are

any given
the risk

choice of the agent,

functions R(AI,d j

and

Al

R(Al,d j

el(I-[3(d j

+ e 3 [3(d j

A2

R(AZ,d j

ezo:(d j

Suppose now

+ e4

that data

l-o:(d j
are available

on

the

prior

probabilities of
cooperative (AI)
and noncooperative behavior
(AZ) on the part of the agent. and denote these probabilities
with pI
and pZ
(I-pI). Then the expected loss for any given
(17) Note that the
losses refer to absolute reductions of the
present value of an existing contract,
i.e.,
a
loss could
exceed the initial face value of a given capital transfer.

355

decision rule

may

be

defined

as

the

weighted

average

of

R(Al,d j ) and

R(A2,d j ), where the weights are the probabilities

of Al and A2,

i.e., pI and (I-pI), which is (8).

loss B(d j )

The expected
for decision
and

(Schlaifer 1961 : 173-86). Since R(A1,d j )

rule d j

R(A2,d j )

are

absolute

probabilities, B(d j )
of B(d j )

to the

is called the Bayesian risk

costs,

is measured

and

pI

and

(l-p1)

are

in currency units. The ratio

initial capital

transferred (C), thus,

is an

expression for the required sovereign loss premium.

From Figure

3 and equations (6) to (8)

it is obvious

that the alpha- and the beta-error probabilities are a function

of the

chosen decision

forecasting or
logit functions

rule d j

estimated

in

Section

probabilities of

simply

to

explanatory

up an

fill

variables

probabilities of
such a

is

related

in
in

used

for
One

the

II

can

be

appropriate

order

to

get

noncooperative behavior

infinite number of rules d j

cutoff-probabilities

principal should
probability

of

some

noncooperative behavior.
data
the

of the

stay

with

critical cutoff-rate

q*j, and

the

agent. Given
one can set

which pertain to alternative

qj.

The

strategy

noncooperative

for

predicted

technique like a logit regression function,

critical

to

sampling procedure. One may now recall that the

predicting the
needs

which

rules

Bl.

action

is

if

say
the

smaller

that

predicted
than

the

that he should move to strategy

B2 otherwise.
Errors are
rules.

If

one moves

one choses

unavoidable by

upwards with the critical cutoff-rate qj'

a different

critical cutoff-rate

following anyone of such

decision

upwards,

the

rule

dj .

By

alpha-error

shifting

the

probabilities

necessarily rise while the beta error probabilities necessarily

fall. Thus,

a trade-off

beta-errors. The
a(d j )) and

avoiding

alpha-

and

probabilities of making the right choices (1-

(1-~(dj))

It is

exists between

change accordingly.

obvious that

for any

given set of loss costs

(e. to e4) and prior probabilities of cooperative and noncooperative behavior by the agent (pI and (I-pI)). the Bayesian risk

356
in equation

(8) changes.

This

one substitutes (6) and (7)

The Bayesian

becom~s

immediately obvious.

if

into (8).

procedure now requires that the cutoff-

rate q" j. and hence the decision rule d"j is selected such that
the Bayesian

risk B(d"j)

is minimized. The Bayesian procedure

applied to the present context means that a standard is derived

from

which

the

calculated both

required
for the

sovereign

loss

premiums

can

be

debt and the FDI case. The assumption

is that competition forces the principals to select the minimal

Bayesian risk

B(d"j). and

accordingly.
course.

The

calculate the required cost mark-up

corresponding

sovereign

loss

premium.

of

is calculated by relating B(d"j) to the initial capital

transfer C.
C. The

Decline

in

the

Relative

Costs

of

Debt

Repudiation'
Suppose now
form of

either debt

the optimal
e4. 1 .

and

debt

risks as

Bl(d"j.l) and

to

e 4 .2.

The

prior

probabilities

of

noncooperative action are pl.l and (l-pl.l)

case.

and

pl.2

and

(l-pl.2)

in

Similarly. the

alpha- and

beta-errors depend

decision

d"j.l

d"j.2

rule

B2(d"j.2).

corresponding losses along Figure 2 are el.l

el.2

cooperative and
the

or foreign direct investments. and denote

Bayesian loss

respectively. The
to

that the capital transfer C can take the

and

chosen.

the

FDI

on the

in

case.
optimal

respectively.

Then

equation (9) displays the

(9)

I B2 ( d" j

B 1 ( d" j

[pl.l

(el.l(l-~(d"j.l)

1)

2 )

+ e3.1~(d"j.l

(l-pl.l)(e2 . 1a(d"j.l) + e4.1(l-a(d"j.l]

/[pl.2 (e 1 .2 (l-~(d"j.2)

+ e3.2~(d"j.2 +

( l-p 1 . 1 ) (e 2 . 2a ( d" j .2) + e4. 2 ( l-a ( d" j .2) ) ]

determinants of the relative sovereign loss premium per unit of
capital

(R) .

It is
induced by

obvious that

the change

of the

price

ratio

the emergence of cross-default clauses in the 1970s

as discussed above crucially depends on the relative magnitudes

357

of the losses involved (e , . , to e4.1)' The derivative d(R) with

respect

to

d(a(d- j . 1 ))

innovation. while

and

p1 and

induced

(~(d-j.l))

by

this

hence (1 - p1) are kept constant. is

given by (10). Since

(10) d(R)

both

[p1.1(e3.1-el . )d(~(d-j.ll) + (1-p1.1)

(ez.l-e4 . )d(a(d-j . ))]/B2(d-j.z)
and a(d- j . )

~(d-j.l)

of increased
constant for
of the

go down simultaneously as a result

predictability. and p1.1 and (1 - p1.1) are kept

the moment. a sufficient condition for a decline

relative sovereign

loss price R is that e 3 . is larger

than e .. and that e Z . 1 . is larger than e 4 . 1 . Both assumptions

are very reasonable. as a quick look at Figure 2 and the
economic meaning

of the loss configurations suggest. The first

condition means that the loss when exit is chosen. although. the
country did

not move

"excess" monitoring
an unanticipated
Since adjustment

to willful

default.

is

larger than the

costs. And the second condition means that

willful default cost more than a prior exit.

of the decision rule (d*j.l) will only happen

if the costs are reduced further. the decline of R can only be

increased.
In order to complete the analysis. suppose now that
the alpha- and beta-error probabilities are given. and that the
prior probabilities

of cooperative and noncooperative behavior

have changed

(p1 and (1-p1)). The condition for noncooperative

behavior

equation

in

expression for
would be
part of

(11

has

forgone future

higher in

the case

the creditors.

shown

that

capital inflows
of an

Assuming that

I(dA/dt)
(if

as

an

positive)

effective sanction on the

the relevant

decision-

making unit acts rationally. this change should have lead to an

increase in the probability of cooperative behavior on the part
of the

indebted agents.

The

partial

derivative

of

with

respect to p1.1 reads as follows:

(11) dR/dp1.1

[e , . , (l-~(d*j.l)) + e 3 . ~(d*j.l)
- e Z . 1 a(d*j.l) - e 4 . 1 (1-a(d*j.ll]/B2(d*J.z)

It can easily be seen from the loss matrix (Figure 2)

358

that the

term is

alpha- and

beta-errors as

smaller than
again

negative for

e2.1 and

quite

would be

e4.1

strategy (A2).

(sufficient

and

e3.1

are

of
both

condition). These are

since

otherwise

the

with the cooperative strategy (A1)

Note that the assumption e3.1 < e4.1 means that

in the

secondary market (18). Thus,

the present case the risk function

a smaller value than R(A2,d j
negative in

combinations

those which come with the noncooperative

an asymmetry

Now,

e1.1

assumptions,

which come

larger than

there is

long as

reasonable

expected losses

all possible

if

(R(A1,d j

in (6)

in

generates

in (7).

the dR/dp1,1 can reasonably be assumed to be

the present

application, then any increase in the

prior probability of cooperative behavior p1,1, as suggested by

the impact of the cross-default clauses on condition (1),

leads

to a decline in the sovereign risk ratio R.

The overall

result of

this analysis,

thus,

is that

under very

reasonable assumptions with respect to the relative

magnitudes

of

Figure 4

the

loss

matrix

(Figure 4),

the

Sufficient Conditions for the Loss Matrix (Ranking)

Principal's
Strategy

A1
1
2

B1
B2
of

emergence

relevant

cross-default

Agent's Strategy

clauses

A2
4
3

reduced

the

relative

sovereign loss premium in favor of debt finance. The first line

of the

argument referred

to the more efficient application of

forecast techniques, while leaving the prior risks (p1 and (1 p1) unchanged.

The second

cross-default clauses
of cooperative
debtor, while

reasoning took

into

account

that

had an impact on the prior probabilities

and noncooperative
leaving the

actions on

forecast

the part of the

technique

unchanged.

Of

course, both went together in the same direction.

Empirically,
loans were

1980s, although
(World Bank

the

indeed falling

so-called
from the

spreads

on

early 1970s

syndicated
to the early

there was a temporary increase peaking in 1976

1985:

119). While this observation is well-known

(18) For a further discussion see Picht (1988)

359
and consistent
magnitudes of

with the

prediction derived

above, the actual

the spreads do not completely reflect the actual

risk charges, even if additional credit fees are accounted for.

The risk

charge (ignoring

assumptions about

risk-aversion) for

the loss

quite plausible

estimated at 3,85 % on

matrix was

average for the country sample chosen (19).

IV. The

Impact

External

of

Sovereign

Capital

Loss

Structure

Costs

on

the

and

Economic

Performance.
A. The Hypothesis
The hypothesis
basically of

three

suggests that
relation to

if the

in

First,

this

section

simple

consists

economic

calculus

price of one of two substitutes falls in

the other,

conditions change
to the

advanced

parts.

then rational actors will under normal

their demand in favor of the former.

Applied

present context this means that a relative price change

for incurring
finance of

debt has

the country.

section that
relative to

an impact on the structure of external

Given

the sovereign

the

result

loss premium

of

the

for debt

previous
went

down

the one required in the FDI case, one would expect

that the FDI-Debt ratio was reduced in response.

Second,
structure of
performance of
classic

the

question

external finance
the agent's

Modigliani-Miller

arises

as

is irrelevant

to

whether

the

for the economic

country, one may simply recall the

theorem

in

the

business

finance

literature (Modigliani and Miller 1958 : 261-96), or whether it

matters.

If

it would

matter,

in

which

direction

would

the

(19) The actual prior probability of cooperative behavior is

p1,1 = 84.00 percent.
The loss premium is estimated on the
basis of e1. 1 = 0.5, e Z 1
= 80, e3.1 = 10, and e4.1 = 15 as
conceivable losses of a sovereign loan contract of a face value
of 100. The numbers are'~uestimates" based on data presented in
Watson et al. (1986)
on recent loan losses,
average credit
terms and maturities,
and the assumption that
losses of
principal are tax-deductable (roughly 1/3).
The alpha- and
beta- error combinations for alternatives cutoff-probabilities
are provided by the author on request. The estimation is robust
notably with respect to e Z . 1 '

360

impact go

The hypothesis advanced here is, first,

that the

structure of external finance matters and, second, that a shift

towards debt

relative to

equity capital

impairs the economic

performance of the agent's country.

In order
second part

of the

debt finance

displays the

percent

it

reasoning

behind

this

is helpful to realize that

debt

finance,

and

that

essentially private in nature.

magnitude of

public involvement

in LDC

1970 to 1984 for all developing countries. About

of

all

guaranteed. Hence,
the external

the

is public

investments are

borrowing from
80

hypothesis,

in practice

foreign direct
Table 5

to understand

debt

incurred

any reasonable

is

puhlic

or

publicly

hypothesis on the impact of

capital structure on economic decision-making and

aggregate outcomes needs to start from the assumption that debt

refers primarily
calculus, while
entrepreneurs.

to public

decision makers

foreign direct

and

investments relate

the

related

to priyate

361

- Share of Publ ic and Publicly Guaranteed Debt in

Percent of Total Debt in All Developing Countries (in
Bi 11 ion US Dollars)

Table 5

Total All Countries

Share of
Public/Publicly
Guaranteed
Debt

Public and
Publicly
Guaranteed

Private
NonGuaranteed

Total

1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984

49.49
60.13
68.16
83.72
103.87
125.72
154.82
193. 14
246.58
292.33
351.76
393.32
443.81
513.07
552.96

16.96
14.87
22.52
25.52
31.53
36.46
40.04
46.37
55.54
62.75
75.30
96.67
103.09
110.05
112.53

66.45
75.00
90.67
109.24
135.40
162. 18
194.86
239.52
302. 11
355.08
427.06
489.99
496.41
622.12
665.49

Sources

World Debt Tables, various issues

In what

follows the

government agents

quite popular

0.74
0.80
0.75
0.77
0.77
0.78
0.79
0.81
0.82
0.82
0.82
0.80
0.89
0.82
0.83

view that

governments and

spent money less efficiently is underlined a

number of arguments.
First, De
Alchian and

Alessi

Demsetz that

has

argued

in a

along

the

classical firm

lines

of

centralized

contractual agent (the owner-entrepreneur) specializes to meter

input productivity

and rewards

i.e., the

shirking information

solved by

assigning residual

entrepreneur
agents,

in

(1980
turn, there

directly from
gain from
since

perquisites, i.e.,

civil-service

suggests that

team production process,

rights to
For

is hardly

such profits.

residual incomes
expense of

15).

in a

problem of

team production is

income to

politicians
a

and

possibility

the ownergovernment
to

benefit

Instead, there is an incentive to

non-monetary

regulations

preclude

forms
the

of

income,

payment

of

(profits). The comparative perspective, thus,

increased reliance

(private) foreign

upon (public)

direct investments

debt

at

the

reduces

the

362
aggregate rate of return on capital (20).
Second, Fama
time horizon
quality

of those

of

the

of ordinary

hold for
Hence,

increased

It

horizon of

business

men.

to western-type

regimes who

broader sense

assets has an impact on the

decisions.

the time

certainly applies

(183: 331) argued that the

who manage

investment

acknowledged that
than that

and Jensen

has

widely

been

politicians is shorter
While

this

assessment

of democracies,

it may also

must fear to be removed sooner or later.

public involvement

would indicate

in

the

economy

in

the

that the capital borrowed is put

into less productive uses.

A third
emerge directly
introduction

favor

of

cross-default

individual borrowers
is supported

may

covering

publicly

blurred the differences in

within

country,

since

by others to avoid triggering

the cross-default clause" (1985 : 326).

much incentive

hypothesis

observation, that "the

clauses

have significantly

deliquent borrower
is not

the

from Folkerts-Landau's

of

guaranteed debt
risk among

argument in

In such a setting there

on the part of the individual borrowers

to use the capital most efficiently, because others, eventually

the government,

will have

particularly strong
in a

syndicate to

to bail

them out. There is also no

incentive on the part of the leading banks

monitor the

borrowing

agent

effectively.

Rather there is the incentive to reduce the monitoring efforts,

because with

a given

the additional
leading

compensation the costs would only reduce

discretionary profit

banks,

monitoring would

while

the

accrue to

margin that

overall
all

benefits

banks

goes to
of

the

effective

participating

in

the

syndicate.
Fourth,

in

most developing

centralization within

government is

countries the

degree of

already quite high,

i.e.,

federal elements are the exceptions rather than the rule (Hicks
1978). Since
concern of

external relations
the central

intermediate and

government.

lower levels

(20) Wallich
(1986:
part of the borrowed
investment.

are traditionally a matter of

its

role in comparison to

is increased

by

the

extended

70). for
instance. argued that a large
money went
into consumption rather than

command over
is true

additional resources by external borrowing.

that a

money more

decentralized system

efficiently (Nelson

borrowing would

also mean

of government spends the

1987),

that the

If it

then

extended

quality of

the

public
spending

decisions is negatively affected.

And fifth,
external debt

note

that

substantial

part

of

the

incurred may be assumed to have left the country

in obscure ways by capital flight (The World Bank 1985 : 63-5).

It is

questionable

productive use

whether

abroad due

operations. But

the

capital

to the

is

illegal

put

into

character

most

of

the

anyway, the earnings on flight capital are not

available for servicing debt.

All five arguments presented, thus, point to the same
direction. A
debt

shift away from foreign direct investment towards

suggests

that

the

capital

received

is

spent

less

efficiently (21).
The two

hypotheses presented in this section may now

be combined.

A relative decline in the relative sovereign loss

premium does

not only

structure away

lead to a shift in the external capital

from FOI to debt finance,

it also causes a drop

in the efficiency of the use of the capital received.

The third
far the

part of

the hypothesis is a matter of how

first two parts go. Following Kindleberger's notion of

solvency that

"(t)he only determinant of the capacity to repay

is the loans contribution to the productivity of the economy as

a whole"

(1958:

265), a

situation may arise where countries

run into a severe solvency problem which may have been minor in
the first

place (22).

default clauses
make it

was to

While the

primary

purpose

of

cross-

reduce the willful default risk and to

more manageable, the counterintentional outcome in the

(21) It may be noted that

if one is willing look at public
decision-makers as if they behaved like owner-entrepreneurs in
a classical firm setting (e.g. Jensen and Meckling 1976
Stiglitz 1972),
then the argument can be made that the shift
towards external debt sets an incentive to choose investments
with higher expected return. The critical assumption thereby is
that the owner-managers are risk-neutral.
(22) This concept assumes that the principal can continuously
be rolled over, and that the existing debt
is
in effect
perpetual. For a similar concept see Cline (1983 : 45-6).

long-run

may

be

that

solvency

risks

are

substituted

for

sovereign risks

instead. But unfortunately then the principals

may be

lead to

switch to the noncooperative strategy (from Bl

to B2)

as displayed in Figure 1, irrespective of the intent of

the agents.

In

turn,

advantageous in

for

the

agents

it

may

then

become

the light

of this move to definitively change

from the

cooperative (AI)

to the noncooperative (A2) strategy

as well.

In the

collapses for

extreme case the international capital market

LDGs and severe restrictions on trade may emerge

since barter terms apply subsequently.

B. Empirical Evidences
The empirical analysis is done by using data from the
country sample

selected above.

external finance
investment to

as measured

In Table
by the

6 the

structure

ratio of

of

foreign direct

debt is displayed (unweighted average). There is

a clear trend in the figures up to 1982. The FDI-Debt ratio has

substantially declined throughout the 1970. Following the first
part of

the hypothesis stated above, a decline in the relative

sovereign loss
Debt ratio.
The use

premium should

The hypothesis

of cross-default

relative loss

have lead to a drop in the FDI-

is, thus,

supported by the facts.

clauses has

premiums, but

not

has also

only

lead to

reduced

the

a shift in the

structure of external finance towards debt.

Table
performance of
first proxy

second and

displays

the country
there is

also

data

on

group (unweighted

moving averages

(WRGDY). Again,
that the

of

GDP

growth

a substantial

the

economic

average). As
were

calculated

trend which indicates

economic performance had gone down continuously. As a

more appropriate

proxy the aggregate real marginal

rates of

returns on

(see the

Appendix for the method applied and a comparison with

data

generated

in

unweighted averages

capital were
the

estimated for

International

are presented

as RR.

Monetary

each country
Fund).

These figures

suggest a substantial decline over the period covered.

The
also

365

Table 6 - Selected Data for the Country Sample

Public Debt
and Publicly
Guaranteed
Debt in Percent of Total
(average) (average) (average)

FDI/Debt

0.41
0.48
0.35
0.10
0.38
0.46
0.22
0.16
0.01
0.17
0.07
0.08
0.04
0.19
NA
NA
NA

1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
Note

RR

20.08
15.0
16.0
15.5
15.0
15.6
13.8
13.3
13.3
12.5
11.9
12.0
12.3
16. 1
NA
NA
NA

NA
0.71
0.73
0.72
0.81
0.81
0.72
0.84
0.86
0.83
0.89
0.86
0.86
0.85
0.88
0.88
NA

Growth
Prospects
WRGDY

Libor

(average)

(real)

6.8
6.3
6.3
6.7
6.7
6.0
5.4
5.8
5.6
5.8
4.7
3.6
2.7
2.1
2.4
2.3
NA

NA
NA
3.28
0.97
-3.68
-7.90
-2.25
0.97
-0.05
1.05
-0.41
0.09
7.77
11.29
8.42
8.54
13.40

For the calculation of RR see Appendix.

Sources : See variable list in the Appendix.

Following now

the second part of the hypothesis. one

would expect that the decline in the FDI-Debt ratio had lead to
a less
Table 6

favorable economic
support this

coefficients

are

performance. The

hypothesis.

displayed.

In

They

data presented in

Table 7
are

the correlation

large

and

highly

significant. The hypothesis that the change in the structure of

external finance had a negative impact on the performance of an
economy is.
that the

thus. strongly supported by the data. One may note

share of

public and publicly guaranteed debt has not

only been high at the outset. 0.71 percent in 1970. but that it
increased continuously

up to 0.88 percent in 1984. which means

that public involvment has even further increased.

366

Table 1

- Correlation between the External Capital Structure

and the Economic Performance (Time-Series)

FDI/Debt
(AVR)
FDI/Debt
(AVR)

RR
(AVR)

WRGDY
(AVR)

0.692
( 13)
P=0.004
0.637
( 13)
P=0.010
1.00

0.520
( 13)
P=0.034
1.00

1.00

WRDGY
(AVR)
RR
(AVR)
Note:

P-Values denote
covers 13 years

Source

the significance levels

the analysis

Table 6.
One might now argue that time-series data are not the

appropriate basis
forces might
on the
the

for

testing

hypothesis,

since

other

have been at work. The countrywise assembled data

external capital

estimated

the

structure, the

aggregate

rates

of

GDP growth rates, and

return

on

capital

were

therefore averaged over the sample period (1969-82), and crosscountry correlation
are displayed

coefficients were

in Table

8. Although

calculated. The results

the sample size is small,

the results remain the same.

Table 8

- Correlation between the External Capital Structure

and the Economic Performance (Cross-Country)
FDI/Debt
(M)

FDI/Debt
( M)

WRDGY

1. 00

WRGDY
(M)

0.294
( 10)
P=0.205
1.00

(M)

RR

RR

(M)

0.840
( 10)
P=0.001
0.431
( 10)
P=0.107
1.00

(M)

Note
P-Values denote the significance levels; (M) denotes
the country means (1969-82). The parentheses indicate that 10
countries were looked at.
Sources: See variable list in the Appendix.

367

The correlation
average real

rate of

significant at
prefer.

between the

return on

FDI-Debt ratio

and the

capital is at 0.84. and it is

the 1 percent level. Whatever procedure one may

the

time-series

analysis

or

the

cross-sectional

approach. the hypothesis that the change in the external

capital structure has had a negative impact on the economic
performance of the agent's country is strongly supported.
The

data

presented

in

Table

do

even

provide

evidence for the third part of the hypothesis that increased

solvency problems
have been substituted for the initial
reduction of

sovereign risks.

return on

capital RR

1980s, it

also reached

whether

the

to

come up

returns

foreign capital.

rate

of

up to the early

where doubts

generate

attract new

marginal

only decreased

a level

potential

sufficient to

The average

had not

on

as

to

capital

is

Taking the London

Interbank Offer Rate (US deposits. deflated) as a proxy for the

risk-free rate

on capital,

one can

see that

this

rate

has

substantially increased from the 1980s onwards. Adding mark-ups

for expected losses and premiums for the systematical or nondiversifiable risk, since risk-aversion cannot realistically be
assumed away

(Picht 1988),

the stage
are

may have

not

able

already

reached where

many countries

capital. That

is, their economies are no more able to generate

enough return

on the

capital invested

to

service

been
new

as seen necessary from

the principal's point of view. Recall from Section III that the
the minimal
estimated at
be high

Bayesian (loss)

risk in

the debt

case alone was

around 3.85 %. At the same time RR still seems to

enough to

yield sufficient net returns on the capital

invested if the risks were low or absent.

If

this

distinction between
solvency

hypothesis

holds.

sovereign

risks

however.
on

the

then

one

hand.

the
and

risks

other hand
becomes
empirically
on the
inoperational for new capital. One may recall that the initial

discussion

had

focussed

on

cooperative

and

noncooperative

behavior on both sides of the game, the agent and the principal
(Figure 1). Even if one were to assume
that the agent is
willing to play the cooperative strategy, principals will still
expect excess premiums over the risk-free lending rate to cover
the expected

sovereign risk

loss

and

the

non-diversifiable

368
risk.

If

the economies

enough return
return on

on the

do not

seem to

capital invested

investments including

be able
to meet

adequate risk

to

generate

the

required

compensations,

principals move from the cooperative (Bl) to the noncooperative

strategy (B2) with respect to fresh capital (Figure 1).
This exactly has been the case. Voluntary lending has
become

quite

consequence,

marginal
in

(World

turn, the

Bank

1986)

(23).

But

as

agents have moved more readily from

the cooperative (Al) to the noncooperative (A2) strategy, since

they are

cut-off

equation (1)

from

future

capital

anyway

(I(dA/dt)

in

drops).

Both

the

principals

and

the

agents

have

acted

rationally, yet as a result the financial market got trapped in

a noncooperative constellation (Cline 1983 : 77 ; Guttentag and
Herring 1985

: 4).

The irony of this move from cooperation to

noncooperation of

both sides

much a

reduced trust,

matter of

lack of

of the game is that it is not so

sufficiently attractive

LDCs covering

but -

in the meantime - the

investment

opportunities

in

all risk premiums which hampers the move back to

the cooperative solution (Figure 1).

V. Conclusions
While the international banking community was able to
organize collective

action quite

effectively during the 1970s

by applying cross-default clauses as a form of market sanction,

the direct

investors in

LDCs were

not. As a consequence, the

relative price of both types of finance had changed in favor of

debt, and

debt gained

ground at

the cost

of private foreign

direct investments.
The unfortunate
public debt
generate

effect of this shift notably towards

was that the ability of the recipient countries to

sufficient

severely affected.

returns
In the

on

the

early 1980s

capital

invested

the solvency

was

problems

(23) Note that the FDI-Debt ratio for the country sample chosen
has increased substantially in 1982 (Table 6) and onwards due
to a drastical reduction of new bank loans (see also World Bank
1985 : 112).

369
induced began
mechanism.

to spill

The

suggested not
even

more

back to

reduced
only a

the inventors of the sanction

profitability

of

LDC

investment

withdrawl from existing commitments. but

important

from

future

cooperation

(voluntary

lending).
In turn. LDC governments were lead to minimize losses
by moving

to the

point of

noncooperative strategy

as well. From their

view the difference between getting no more credit as

a sanction

related to
a steady

to generate

cover the

enough to
does not

willful default. or because the ability

flow of

interest services which is high

required risk

premium on future capital.

seem to make too much of a difference.

In either case

their access to future credit is limited.

It

was

thereby

recurrent reschedulings
outward form
sanctions

certainly

of noncooperative

other

than

applicable.

It

the market

sanction of

advantageous

rather than

behavior. since

foreclosing

was shown

in the

to

choose

outright default
fresh

as

the

in this

money

are

case
hardly

analysis that in addition to

foregone future

credit in the case of

default. political sanctions may be assumed to have some impact

on the

choice between

strategy. Within
rational for

the

the cooperative
noncooperative

the agents

to keep

and the noncooperative

strategy

it

the sanction

is.

thus.

costs down

by

obscuring the option actually chosen.

But it

has also

been advantageous

for the

lending

banks to deal with reschedulings rather than outright defaults,

since with
as well

respect to

as the

the financial stability of single banks

stability of

the banking system as a whole it

was crucial to get time for adjustment (Sachs 1982 ; Grauwe and
Fratianni 1984
In

the

creditors move
strategy?

Claassen 1986) (24).

present
from

the

Obviously it

situation.

how

noncooperative

can
to

debtors

the

and

cooperative

is the solvency issue. which needs be

addressed from the agents' and the principals' side.

The LDC

governments need

to find

ways

which

make

(24) See Euromoney (1987

: 86)
for a
recent survey on the
relation of bank reserves and losses on LDC debt at market
terms for 15 major US banks.

370

investments more
may infer,

profitable. From

first,

investments has

that

an opening-up
of course,

was much

uses than

guaranteed debt

debt. Second,

may increasingly

be

The

shirking-problem

coverage

case

of

failure

incentive is

set towards

first place.

Third, and

the efficiency
to be

put into

converted

into

associated

would

be

publicly
private

with

reduced.

public

Thus,

an

most crucial, within the public realm

decisions and

public management

In particular, the capital imported needs

productive rather than consumptive uses.

these suggestions

chance that

public and

more efficient use of capital in the

of investment

be enhanced.

much as

direct

make sense in this respect,

obligations.

need to

towards foreign

evidence that equity capital is put into

more productive

in

analysis one

a positive effect on the economic performance.

Debt-for-equity swaps,
since there

the preceding

are taken

In as

seriously, there is some

capital flight can be stopped and that even a part

of the capital can be repatriated.

For the
credit claims

banks an

will

increase in

lower

the

the

required

tradeability

(future)

return

of
on

capital invested, first, by making the exit option markets less

costly.

The

Bayesian

presented in
lower in

this case.

easier to

approach
Second, by

diversify at

lowers the

to

sovereign

loss

premiums

the analysis shows that the required of return is

faciliating the

least a

required rate

trade it

is

part of the risk, which again

of return

and,

hence,

lowers

the

solvency problem (25). Third, the monitoring techniques applied

to sovereign
lending

be

may increasingly

lending. Both
since the

be improved, and techniques like on-

lending may
lower the

substituted

required premium

principals are

in a

for

unconditional

for solvency risks,

better position to control the

use of the capital.

But even
and

the

related

if the
devices

"active debt management" techniques

just

outlined

have

merits,

the

fundamental challenge of noncooperative behavior on the part of

the agent country remains in absence of contract enforcement by

(25)
For
a detailed discussion on the
sovereign risk management see Picht (1988).

possibilities

of

371

a third

party (26). It is indispensable that the mechanisms of

contract enforcement in the international realm are generally

improved in order to pave the way for a return to cooperation.
In particular, efforts must be directed towards
establishing institutions
direct investment. They
external

capital

investment at

(27).

structure

high rather

inflows. The
advanced by

effectively protect
support the change

in

favor

the World
may

Bank is

bonding
serve

of

than shrinking

foreign

levels

direct

of

but one step in this direction

activities

the

foreign
on the

capital
Guarantee Agency (MIGA)

Multilateral Investment

Voluntary

governments

which
would

same

on

the

purpose,

part

since

of

LOG

brand-name

capital is built up which functions as a collateral for further

capital inflows.
LOG governments could even offer unilaterally
to forego
expected future
aid flows from national and
international sources as a form of collateral. In fact, the
analysis suggests that foreign direct investment has become
complementary to foreign debt rather being a substitute for it.
Finally, the liberalization of international trade
seems to

be crucial

behavior of

LOG governments.
get integrated
imports, the
welfare

the

back

towards

cooperative
countries and

suggests that the more the LOGs

world

economy,

the

lower

is

the

One reason is that with a rising dependence on

potential for

credit is
gains

move

from industrialized

The analysis
into

sovereign risk.
to future

for the

both investors

from

market sanctions by denying access

increased. Another
liberalized

trade

reason
may

is
be

that
assumed

the
to

stimulate the income prospects. While both effects of trade

liberalization may be assumed to favor cooperative behavior on
the part
of LOG
governments,
the
market
access
to
industrialized countries is largely out of their control. Thus,
the governments of the industrialized countries are called to
contribute towards the stabilization of the international

(26) For an introduction into debt management techniques see

Watson et al (1986).
(27) See Hauser (1987) and Shihata (1987) for a most recent
discussion on the Multilateral Investment Guarantee Agency
(MIGA).

372

financial system

by opening

their economies

for imports from

LDes.

APPENDIX
1. Variable List
AIDeRQ

Total ODA from all sources

per unit of GOP.
Sources
OEeD,
Geograph.
Distribution,
various
issues, and
IMF, International Financial Statistics
Yearbook, various issues.

AND

Dummy set "1" if either Argentina, Brazil or Mexico

has rescheduled,
otherwise "0".
Source: The World
Bank 1985 ; 1986.

BOP

Balance of payments deficit divided by the threeyear moving average of balance of payments deficit.
Source:
IMF, Balance of Payments Yearbook, various
issues.

DGDPKA

Absolute distance of per capita

income at
1980
exchange rates
to equivalent
of 700
US
$,
transformed to
[1/Distance + 11.
Source
IMF,
International Financial Statistics Yearbook, various
issues.

DUM

Dummy set" 1 "

if the FDI stock adjustment model
applies to a country, "0" otherwise.

D\/RY

Actual GOP (i n prices of

1980) growth rate minus
IMF,
WRGDY.
Source
International
Financial
Statistics Yearbook, various issues.

DZKP

Estimated real rate of return on capital (RR) minus

LIBOR on US deposits
(3-months,
deflated by GOP
price index
(US)). Source: IMFInternational Financial
Statistics Yearbook, various issues.

EXPR

Three or more expropriations per year as defined in

the text.

EXPVAR

Variation coefficient of five-year moving average

of exports
(in prices of
1980).
Source
IMF,
International Financial Statistics Yearbook, various
issues.

GDL

Reciprocal
of either public or publicly guaranteed
debt, outstanding and undisbursed, in nominal US $.
Source:
The World Bank, World Debt Tables, various
issues.

GDPVAR

Variation
average

coefficient
(in prices

of GDP-five-year
of
1980).
Source

moving
IMF,

373
International Financial Statistics Yearbook. various
issues.
IDBSR

Reciprocal
of gross
investments
(in prices of
1980). in millions.
Source
IMF.
International
Financial Statistics Yearbook. various issues.

IMPQ

Imports
per unit
of GOP.
Source
IMF.
International Financial Statistics Yearbook. various
issues and World Bank 1985.

Attached to variable name means a lag of one year.

LIBOR

London
Interbank Offer Rate on US deposits
(3months. deflated by GOP price index (US. Source:
IMF. International Financial Statistics Yearbook.
various issues.

POP

Reciprocal
of population.
in mil~ions.
Source
IMF. International Financial Statistics Yearbook.
various issues.

RSCDG

Reschedulings either in
Club. or both. Source
1986) .

RR

Aggregate real marginal

rate of return on capital
(RR). See Appendix for methodical explanations.

WRGDY

Growth rate of three-year-moving averages of GOP

(GOP in prices of 1980). Source: IMF. International
Financial Statistics Yearbook. various issues.

WRYM

Actual GOP growth rate minus arithmetic mean of

five best GOP growth rates in the entire LOC group
(GOP figures
in prices of
1980).
Source
IMF.
International Financial Statistics Yearbook. various
issues.

2. Note

on the

the Paris or the London

( 1985
IMF.
World Bank

Estimation Procedure

for the

Aggregate

Real

Marginal Rates of Return on Capital

Two problems
the real

marginal rates

country. First.
estimated and.
to the

had to

the real

of return

to get estimates for

on real capital RR(t) of a

net capital

stock K(t)

had

to

be

second. estimates for the incomes P(t) relating

capital stock

calculated as

be solved

were required.

RR(t)

could

then

been

the ratio of capital income P(t) and the capital

374

stock K(t).
In order to get an estimate of the real capital stock
available for

productive use

in an

economy, one can drawn on

the insight

gained from Solow type of models in modern capital

theory that

macroeconomic variables
product,

gross domestic
effective

labor

progress) all
stocks grow

the same

real investment,

money

balances,

Harrod-neutral

rate in

and

technical

steady states.

It is

the present context that the flows as well as the

at the

same rate.

steady-state rate,
period t,

real

(assuming

supply

grow at

important in

savings,

such as

One can infer from knowing the

and the actual level of net investments for

what the

size of

the underlying

net capital stock

K(t) must have been. Accordingly, the corresponding time series

data (leaving out the labor market for lack of data) were taken
and a

reference year for each country was determined for which

the variation

coefficient (standard

growth rates

of the

variables was
by the

deviation/mean)

three-year-moving averages

of

for

the

the

four

lowest. The time period looked at was determined

data availability

for each country (IMF,

International

Financial Statistics

Yearbook, various issues and tapes). Once

the

estimate

capital

stock

available, all

other capital

adding (sub-tracting)

for

the

stock

the net

reference

data

(total)

were

year

was

generated

investments

for

by
each

following (preceding) year on a cumulative basis.

The

procedure

comparing the
coefficient

suggested

estimates on
with

the

was

the basis

estimates

first,

checked,
of the

based

on

best
the

by

variation

second

best

variation coefficient. The general finding was that the capital

stock estimates

did not

external check

differ

much

from

one

another.

An

of the validity of the estimation procedure was

partially possible,

second,

countries)

most industrialized countries as well. For

some of

included

them net

capital stock

comparison showed

that the

results.

given

Hence,

estimates are

because

the

estimates

procedure

that

not available

sample

for
at all,

are

suggested

most

LDCs

(fifty-four
available.
yields

capital

the procedure

used

good
stock
was

considered as appropriate.
In the second step, the estimate capital incomes P(t)
were estimated.

Data for

the operating

surplus OS,

which is

375
defined as the excess of value added over the sum of (domestic)
compensation of

employees,

indirect taxes minus subsidies, and

fixed capital,

consumption of
(UN, Yearbook

of National

issues). The

were taken

operating surplus

income which,

of course,

employed, and
function of

as a starting point

Account Statistics, Vol.

corresponds to entrepreneurial

relates to

persons

incomes.

Basically,

to property

the Cobb-Douglas

1, various

type was

who

are

self-

production

taken, and regressions

were run where the natural logarithms of OS(t), K(t) and of the
proxies

for

available)

the

number

were

of

entered

Industrial Statistics,

self-employed

as

Vol.

variables

1,

persons

(UN,

(where

Yearbook

various issues).

of

In the actual

tests the parameter constraint implicit in the constant returns

to scale

assumption (Cobb-Douglas)

cases it

was necessary

to run

was not

imposed.

regressions

using

In

some

the

first

differences of the variables to cope with autocollinearity. The

parameters estimated
shares.

In

can directly

the present

(based on

OS) were

be

interpreted

income

case only in the capital income shares

of interest.

Generally,

yielded plausible and stable coefficients.

the implicit

as

the

regressions

It may be noted that

assumption made is that the capital income shares

of OS are constant over time.

From the

actual OS

share estimated,

one directly

level of

interest and

the real

marginal rates

were derived

figures, and

the capital income

gets the numbers for the actual

profit payments P(t). The estimates for

of return

by dividing

on the

real capital stock

the estimate for interest and profit

payments P(t) by the net capital stock estimate K(t).

Again,
countries

in

the
the

magnitudes derived
discard the

inclusion
sample

of

made

it

major
plausible

industrialized
to

check

the

by this procedure. No evidence was found to

method applied,

i.e., the

derived

results

were

plausible.
Finally,
are largely

it

that the estimates derived

in line with unpublished estimates provided by the

External Relations
Fund (Khatkhate
in the

was striking
Department of

1985). For

present study

correlation coefficient

the

International

Monetary

9 out of the 10 countries analyzed

a direct

comparison was

between the

possible.

two data sets is 0.78.

The
It

376
is significant
upward or

at the

0.01 level. There is also no systematic

downward bias

procedure in

in the numbers derived by the present

comparison to

the

estimates

generated

by

the

International Monetary Fund.

REFERENCES
Altman, E.J.,

R.R.

(1981),

Avery,

R.A.

Application

Business, Banking
Economic

of

Eisenbeis

and

Classification

and Finance.

J.F.

Sinkey

Techniques

in

Contemporary Studies in

and Financial Analysis, vol. 3, Greenwich (CT)

: JAl Press.

Basche, J.R.Jr.

(1979), Nationalization: The Experience of US

Companies in the 1970s. The Conference Board Information

Bulletin, n062, August.
Born,

(1986) ,

K.E.

"Erfahrungen

aus

Finanzkrisen der Vergangenheit",

internationale Schuldenkrise.
Historische

Erfahrungen.

Socialpolitik, vol.

internationalen

In A. Gutowski, ed. Die

Ursachen - Konsequenzen -

Schriften

des

ISS, Berlin: Dunker

Vereins

& Humblot,

fur
pp.

9-29.
Burton, F.N.

and H.

Owned Firms

Inoue (1984),
in Developing

Trade Law, vol.

Burton, F.N.
Model

and H.
of

"Expropriations of Foreign
Countries", Journal of World

18(5).
Inoue (1986),

Foreign

Asset

"A Country

Expropriation

Risk Appraisal
in

Developing

Countries", Unpublished Manuscript.

Chakravarty Sh.P. and M.J. Holler (1988),
Agents and
January.

"International Debt:

Chances", Bangor, UK, and Aahrus, DK. Mimeo.

377

(1986),

Claassen, E.H.

"Die

Funktion

des

Kreditgebers

del'

letzten Instanz", In A. Gutowski, ed. Die internationale

Schuldenkrise. Ursachen
Erfahrungen. Schriften
vol.
Cline, W.R.

Konsequenzen
des Vereins

fur

Historische
Socialpolitik,

155, Berlin: Dunckel' & Humblot, pp. 99-104.

(1983),

International Debt and the Stability of the

World Economy.

Institute for

International

Policy Analysis in International

Economics.

Economics n 4, London

: MIT Press (Cambridge).

Davis, W.

(1981),

"Beyond

Political Risk

the Earthquake

Allegory: Managing

Vulnerability", Management International

Review, vol. 21(3), pp. 5-9.

De Alessi,

L.

(1980),

"The Economics

of Property

Rights.

Review of Evidence", Research in Law and Economics, vol.

2, pp.

Eaton, J.

1-47.

and

H.

Gersovitz

Repudiation

(1981a),

Theoretical

"Debt

and

with

Potential

Empirical

Analysis",

Review of Economic Studies, vol. 48(2), pp. 289-309.

Eaton, J.

and H. Gersovitz (1981b),

Private Financial

"Poor-Country Borrowing in

Markets and

the Repudiation

Issue",

Princeton Studies in International Finance, n047.

Euromoney (1987) Special IMF World Bank Issue, September.
Fama,

E.F.

and

H.C.

Jensen

(1983),

"Agency

Problems

and

Residual Claims", Journal of Law and Economics, vol. 26,

pp. 327-49.
Feder, G.

and V.

Regev (1975),

Foreign Investment,

"International Loans,

and Optimal

Direct

Capital Accumulation",

The Economic Record, vol. 51(135), pp. 320-25.

Feierabend,

I.K.

and

R.L.

Feierabend

(1966),

"Aggressive

378

Behaviors within
Study", The

Politics, 1948-1982 : A Cross-National

Journal of Conflict Resolution, vol.

10(3),

pp. 249-71.
Folkerts-Landau, D.
Bank

(1985), "The Changing Role of International

Lending

in

Development

Finance"

International

Monetary Fund Staff Papers, vol. 32(1), pp. 317-63.

Frank, Ch.R.

and

V.R.

Cline

Servicing Capacity
Analysis",

Frey.

An

Journal

1(3),

~p.

B.S.

(1985),

(1971),

of

"Measurement

Application

of

International

of

Debt

Discriminant

Economics,

vol.

327-44.
Internationale

Politische

Okonomie.

Munchen : Vahlen Verlag.

De Crauwe,
of

P. and

H. Fratianni (1984),

International

Research Paper

Lending",

n012,

"The Political Economy

International

Katholieke

Economics

Universiteit

Leuven.

Centrum Voor Economische Studien.

Curr,

(1968),

T.

"A

Causal

Comparative Analysis

Model

Using New

of

Civil

Indices", The

Political Science Review, vol. 62(4), pp.

Cuttentag. J.H.

Strife

American

1104-24.

and R.J. Herring (1985), The Current Crisis in

International

Lending.

Studies

in

International

Economics. Washington : The Brookings Institution.

Hanson,

J.A.

(1974),

Lending", The

"Optimal

International

Borrowing

and

American Economic Review, vol. 64(4), pp.

616-30.
Hauser, H.

(1986),

"Promotion

Developing Countries
Aussenwirtschaft,
Hicks.

U.

(1978),

of Foreign Direct Investment to

An

Exercise

Federalism

Comparative Study.

in

Cooperation",

vol. 41(1), pp. 9-18.

London and

Failure

and

Basingstoke

Success.

MacMillan

379

Press.
IHF,

International Finance Statistics Yearbook, Various Issues.

IHF, Balance of Payments Statistics Yearbook, Various Issues.

Jensen, H.C.

and W.H.

Managerial

Heckling (1976),

Behavior,

Structure", Journal

Agency

"Theory of the Firm:

Costs

of Financial

and

Ownership

Economics, vol. 3(4),

pp. 305-60.
Jodice, D.A.

(1985), Political

Bibliography.

Westport

Risk Assessment.
(ET)

and

An Annotated

London

Greenwood

Press.
Jones, J.Jr.

(1980),

"A

Latin America

Model for Predicting Expropriation in

Applied to Jamaica", The Columbia Journal

of World Business, vol.

Jones, R.J.Jr.
US Oil

(1984),

15(1), pp. 74-80.

"Empirical

Models of Pol itical Risk in

Production Operations in Venezuela", The Journal

of International

Business

Studies,

vol.

15

(Spring/

Summer), pp. 81-95.

(1986),

Khatkhate, D.R.

Interest Rates
Relations

"Assessing

in Less

Department.

the Level

and

Impact

Developed Countries",
International

of

External

Monetary

Fund.

(Document DM/85/8).
Kindleberger,

C.P.

(1958)

Economic

Development.

New

York,

London: McGraw-Hill.
Knudsen,

H.

( 1972) ,

Expropriation

Investments
in Latin America.
~~~~~~~~~~~~~.

of

Foreign

University of

Private
Oregon

Ph.D. Dissertation.
Kobrin, S.J.

(1978),

"When Does Political Instability Result in

Increased Investment
World Business, vol.

Risk 7"
13(3), pp.

The Columbia
113-22.

Journal

of

380

Kobrin. S.J.

(1980),

in LDCs" ,

"Foreign Enterprise and Forced Divestment

International Organization,

vol. 31(1),

pp.

65-88.

U.

Lachler.

(1985),

"Debt

Versus

Equity

in

Development

Finance", Kiel Working Paper n0248.

B.

Lloyd.

(1974),

"The

Political Risk

Identification

in the

and

Assessment

of

International Environment", Long

Range Planning, vol. 7(Dec.), pp. 24-31.

Hason. R.H.

(1974) "Conflicts

Between Host

Countries and the

Multinational Enterprise", California Management Review,

vol.

17( 1), pp. 5-14.

Hodigliani. F.

and H.H.

Hiller (1958),

Corporation Finance,

"The Cost of Capital,

and the Theory of Investment", The

American Economic Review, vol. 48(June). pp. 261-96.

Nelson. H.A.

(1987), "Searching

Extension", The
pp.

American Economic

Comment

Review, vol.

and

77(1),

198-204.

J.

Niehans,

for Leviathan

(1986) ,

"Internationale

undurchsetzbaren Forderungen" ,
Internationale Schuldenkrise.
Historische

Erfahrungen.

Socialpolitik, vol.

In A.

Kredite

mi t

Gutowski, ed. Die

Ursachen - Konsequenzen -

Schriften

des

Vereins

fur

155, Berlin: Duncker 8< Humblot, pp.

151-79.

OECD

Geographical

Distribution

of

Financial

Flows

to

Developing Countries, Various Issues.

Picht. H.R.

(1988)

"Sovereign

Diversification", Kiel

Loan

Exposure

Institute

of

and

World

Portfolio
Economics,

Kiel (February), Mimeo.

Pindyck. R.

and D.

Rubinfeld (1981),

Econometric Models

and

381

Forecasts, Second

Edition, New

McGraw-Hill Book

York

Company.
(1971>,

Robock, S.H.

"Pol itical

Assessment", The

Identification

Risk

Columbia Journal

of

World

and

Business,

vol. 6(4), pp. 6-20.

Root. F.R.

(1968),

"The

Companies. What
Horizons, vol.
Sachs. J.

(19821,

In P.

Expropriation Experience
happened

to

38

of American

companies",

Business

11(2), pp. 69-74.

"LDC

Debt in the 1980s, Risks and Reforms",

Wachtel, ed. Crisis in the Economic and Financial

Structure. Lexington,

Toronto:

Lexington

Books,

pp.

197-243.
Sachs.

(1983),

J.

"Theoretical

Borrowing", Princeton

Issues

Studies in

in

International

International Finance

n 54.
Sachs. J.

and D.

Cohen (1984),

"LDC Borrowing

with

Default

Risk", NBER Working Paper n0925.

Saini.

K. G.

and

Bates

Ph.S.

Quantitative

(1984),
to

Approaches

"A

Country

Survey
Risk

of

the

Analysis",

Journal of Banking and Finance, vol. 8, pp. 341-56.

(1961),

Schlaifer. R.

Decisions. New

Introduction

to Statistics for Business

York, Toronto, London: McGraw-Hill Book

Company.
Schneider. F.

and

Determinants

B.
of

Development, vol.
Sharpe. V.F.

Frey

(1985)

Foreign
13(2), pp.

(1970) Portfolio

"Economic

Direct

and

Political

Investment",

World

161-75.

Theory and Capital Markets, New

York, San Fransisco : McGraw-Hill Book Company.

Sherk. R.

(1974),

"Foreign

Investment in

Asia: Japan vs the

382

U.S.", The

Columbia Journal

of

World

Business,

vol.

9(3), pp. 95-103.

(1986) ,

Shihata, J.

Multilateral

"The

Role

of

Investment

JSID

Aussenwirtschaft, vol. 41(1), pp.

Stiglitz, J.E.

(1972),

"Some

Corporate Finance

the

Projected

Agency

(MIGA)",

and

Guarantee

105-22.

Aspects of

the Pure

: Bankruptcies

Theory

of

and Take-Overs", Bell

Journal of Economics, vol. 3(2), pp. 458-82.

Stoever, W.A.

(1982),

An Analytical

"Endowments,
Scheme for

Country Policy

Priorities, and Policies:

the Formulation of Developing

Toward Foreign Investment", The Columbia

Journal of World Business, vol.

Truitt,

J.F.

(1974),

Investment,

17(3), pp. 3-15.

Expropriation

of

Private

Foreign

International Business Research Series, n03,

Graduate School of Business, Bloomington.

UN Yearbook of Industrial Statistics, Various Issues.

UN Yearbook

of National

Accounts Statistics,

vol.

I, Various

Issues.
Wallich, H.C.

(1986),

"The

American View
In A.

Borrowing Countries and Lending Banks",

Gutowski, ed.

Ursachen
Schriften

M.,

R.

Die Internationale

Konsequenzen
des

Vereins

Berlin: Duncker
Watson,

International Debt Situation in an

Historische

fur

Erfahrungen.

Socialpolitik,

vol.

155,

& Humblot, pp. 69-82.

Kincaid.,

C.

Folkerts-Landau (1986)
Developments

Schuldenkrise.

and

Atkinson,

E.

Kalter

"International Capital

Prospects",

International

and

D.

Markets.
Monetary

Fund. World Economic and Financial Surveys, December.

Williams,

M.L.

(1975)

Nationalization of

"The

Extent

and

Foreign-Owned Assets

Significance
in

of

Developing

Countries

1956-1972",

Oxford

Economic

Papers,

vol.

27(1}, pp. 260-73.

The World Bank (198S), World Development Report 1985.

The World

Bank (1986),

World Debt

Developing Countries.

Tables. External

Debt

of