,:.'.2. :::4.::!

r,
"' .:.r* . :a
':
ll :'1 I*j_ .,4
e- .l- ,
eJf
q: l
:i
MACROECONOMICS
AN INTRODUCTION TO ADVANCED METHODS

THIRD EDITION

WILLIAM M. SCARTH
Copyright © 2009 William M. Scarth

Printed and bound at McMaster University

For more information contact:

McMaster University Bookstore
1280 Main St W
Hamilton, Ontario, L8S 4L8
‘vvv w.fitlesondemand.ca

All rights reserved. No part of the work covered by the copyright hereon may be
reproduced or used in any form without written permission.

For permission to use material from this text, contact us:

Tel: 905-525-9140
Fax: 905-572-7160

Cover Photo "Bricks via Block" © 2009 Greg Roberts (httpligregrob.ca)

Usedwithprmon
Preface
Thirty years ago, Robert Lucas and Thomas Sargent argued that
conventional macroeconomic methods were "fatally flawed;" since then,
macroeconomics has been a most exciting part of our discipline. After all,
it is stimulating to be involved in the initiation of new research agendas.
But while this activity is exciting for researchers in the field, it can be
frustrating for students and their instructors: Journal articles by original
researchers rarely represent the best pedagogic treatment of a subject,
especially when the analysis bccomes quite technical. Thus, when the first
edition of this text was published 20 years ago, I set out to bridge the gap
between intermediate macro texts and the more advanced analysis of
graduate school.
But I had a second purpose as well — to draw attention to the work
of macroeconomists who have been trying to integrate two quite different
schools of thought. On the one hand, there is the rigour and the explicit
micro-foundations provided by the New Classical approach, and on the
other hand, there is the concern with policy that stems from the Keynesian
tradition of focusing on the possibility of market failure. The problem in
the 1970s and 1980s was that the Classicals achieved their rigour by
excluding market failure, while the Keynesians achieved their supposed
policy relevance by excluding explicit micro-foundations. So both schools
of thought were limited in a fundamental way. My earlier editions drew
attention to, analyses by macroeconomists who saw the merit in both
traditions, and who were, therefore, developing models that could score
well on both the criteria for judging model usefulness (consistency with
empirical observation and consistency with constrained maximization
principles). Happily, this drive toward integration has succeeded. Indeed,
the integrated approach has now acquired a title — the "New Neoclassical
Synthesis" — and it now occupies centre stage in the discipline. It is high
time that a new edition of the book discusses this accomplishment, instead
of talking about the desirability of moving toward this goal.
With the increasing use of mathematics in graduate school,
virtually all undergraduate programs now offer advanced theory courses,
and it has become less appealing to use the same books at both levels. I
have found it best to leave PhD students to be served by the excellent
books that now exist for them, and to focus this book on senior
undergraduate students and MA students in more applied programs. With
this focus, it is appropriate that the book stress policy application, not just
the teaching of techniques. The tradition has been to assume that every
student who enrolls in a course at this level is so motivated that no effort
is needed to Berri to the applicability of the ar.a...:.'scs contrast to
this, I have found that -especially at the senior undersdn,L,%re :e -. el. each
student's ability to Maste7 research methods is still -ch dependent
on her belief that rhi; ana:-sis is fundamentally "relevan -: - fOir. the issues
that she talks about witli non-economists. The book respt-_.-Pls this need in
every chapter.
Twelve years ha•.e passed since the second :d1 :or_. so many
changes have been necessar:. to bring the book up-to-dL, -_;:- The:: is now
more coverage of New Classical work, endogenous growth r2:C-07".. and the
theory behind the natural unemployment rate. The new _:.:a' involves
calibrated versions of the theory that make it possible for readers to see
how modern models and methods can directly inform debate. And
the New Neoclassical Synthesis permits a consistent co:717a:- son of the
"short-term pain" and the "long-term gain" that is part o'f mar-% policy
initiatives.
Here is a brief introduction to the book's structure Chapter 1
provides a concise summary (and extension) of intermediate-level
macroeconomics. It ends with the identification of three shortcomings —
the need for more explicit treatment of dynamics, expectations_ and micro-
foundations. The next three chapters cover the analysis that has emerged
to address each of these issues. Chapter 2 examines the First Neoclassical
Synthesis — a system that involved the Classical model determining full
equilibrium, and a Keynesian model of temporarily sticky prices
determining the approach to that full equilibrium. Chapter 3 gives
extensive discussion of the development of rational expectations, and
Chapter 4 provides the dynamic optimization analysis that is necessary for
models to have a more thorough micro base. The next three chapters cover
models that are not so limited, since they incorporate model-consistent
expectations and optimization underpinnings. Chapter 5 covers the first
school of thought to stress the desirability of keeping these features
central — real business cycle theory. Then, with temporary nominal
rigidity added to a simplified New Classical model, Chapters 6 and 7
explain both the methods needed to analyze the New Neoclassical
Synthesis, and the success we have had in applying this approach to a set
of central policy issues.
The book shifts to long-run issues for the final five chapters.
Chapters 8 and 9 focus on theory and policy issues concerning the natural
unemployment rate, while Chapters 10, 11 and 12 discuss both old and
new growth theory. With the chapters on micro-foundations, New
Classical work, labour markets, and growth, more than half of this edition
is focused on the long run.

ii
 I have tried to maintain the user friendly exposition that has been
appreciated in earlier editions — giving equal treatment to explaining
technical details and to exposing the essence of each result and
controversy. Using basic mathematics throughout, the book introduces
readers to the actual research methods of macroeconomics. But in addition
to explaining methods, it discusses the underlying logic at the intuitive
level, and with an eye to both the historical development of the subject,
and the ability to apply the analysis to applied policy debates. Concerning
application, some of the highlighted topics are: the Lucas critique of
standard methods for evaluating policy, credibility and dynamic
consistency issues in policy design, the sustainability of rising debt levels
and an evaluation of Europe's Stability Pact, the optimal inflation rate, the
implications of alternative monetary policies for pursuing price stability
(price-level vs inflation-rate targeting, fixed vs flexible exchange rates),
tax reform (trickle-down controversies and whether second-best initial
conditions ease the trade-off between efficiency and equity objectives),
theories of the natural unemployment rate and the possibility of multiple
equilibria, alternative low-income support policies, and globalization
(including the alleged threat to the scope for independent macro policy).
I welcome comments from users: [email protected] Indeed,
much useful feedback in the past is reflected in these pages, so it is
appropriate to acknowledge a few debts here. In earlier editions, I have
thanked some of my mentors — individuals who have been instrumental in
my own development. In this edition, I confine my acknowledgements to
two groups — those who have provided helpful discussion concerning
particular topics, and some impressive students who have helped improve
earlier drafts. I thank John Burbidge, Peter Howitt, Harriet Jackson, Ron
Kneebone, Jean-Paul Lam, David Laidler, Tiff Macklem, Lonnie Magee,
Hamza Malik, Thomas Moutos, Tony Myatt, Siy,arn Rafique, Krishna
Sengupta, John Smithin, Malick Souare, Mike Veall, and especially Leilei
Tang and Huizi Zhao for their capable and generous assistance. It should,
of course, be emphasized that none of these individuals can be held
responsible for how I may have filtered their remarks. Despite the real
contributions of these individuals, my greatest debt is to my wife, Kathy,
whose unfailing love and support have been invaluable. Without this
support I would have been unable to work at making the exciting
developments in modern macroeconomics more accessible.

iii
Contents
Chapter 1: Keynes and the Classics 1

1.1 Introduction 1
1.2 Criteria for Model Selection 2
1.3 The Textbook Classical Model:
The Labour Market with Flexible Wages 5
1.4 The Textbook Keynesian Model:
The Labour Market with Money-Wage Rigidity 11
1.5 Generalized Disequilibrium:
Money-Wage and Price Rigidity 14
1.6 Conclusions 17

Chapter 2: The First Neoclassical Synthesis 19

2.1 Introduction 19
2.2 A Simple Dynamic Model:
Keynesian Short-Run Features
and a Classical Full-Equilibrium 19
2.3 The Correspondence Principle 23
2.4 Can Increased Price Flexibility be De-Stabilizing? 26
2.5 Monetary Policy as a Substitute for Price Flexibility 31
2.6 Conclusions 35

Chapter 3: Model-Consistent Expectations 36

3.1 Introduction 36
3.2 Uncertainty in Traditional Macroeconomics 39
3.3 Adaptive Expectations 48
3.4 Rational Expectations: Basic Analysis 51
3.5 Rational Expectations: Extended Analysis 56
3.6 Conclusions 60
iv
Chapter 4: The Micro-Foundations
of Modern Macroeconomics 64

4.1 Introduction 64
4.2 The Lucas Critique 64
4.2 Household Behaviour 70
4.3 Firms' Behaviour: Factor Demands 79
4.5 Firms' Behaviour: Setting Prices 83
4.6 Conclusions 87

Chapter 5: The Challenge of New Classical

Macroeconomics 89

5.1 Introduction 89
5.2 The Original Real Business Cycle Model 89
5.3 Extensions to the Basic Model 93
5.4 Optimal Inflation Policy 101
5.5 Harberger Triangles vs. Okun's Gap 105
5.6 Conclusions 111

Chapter 6: The New Neoclassical Synthesis 112

6.1 Introduction 112

6.2 Phase Diagram Methodology 113
6.3 Stabilization Policy Analysis
with a "New" Phillips Curve 122
6.4 Some Precursors of the New Synthesis 124
6.5 An Integrated Analysis:
a "New" IS Curve and a "New" Phillips Curve 131
6.6 Conclusions 138

Chapter 7: Stabilization Policy Controversies 140

7.1 Introduction 140

7.2 Commitment and Dynamic Consistency
in Monetary Policy 140
7.3 Fixed vs. Flexible Exchange Rates 152
7.4 The Feasibility of Bond-Financed
Government Budget Deficits 156
7.5 An Evaluation of Balanced-Budget Rules
and the European Stability Pact 162
7.6 Conclusions 169
Chapter 8: Structural Unemployment 171

8.1 Introduction 171

8.2 Asymmetric Information in the Labour Market:
Efficiency Wages 171
8.3 Imperfect Competition in the Labour Market:
Unions 175
8.4 Transaction Frictions in the Labour Market:
Search Theory 179
8.5 Related Issues in New Keynesian Economics:
Real vs. Nominal Rigidities 183
8.6 Conclusions 188

Chapter 9: Unemployment and Low Incomes:

Applying the Theory 190

9.1 Introduction 190

9.2 Tax Reform: Direct vs. Indirect Taxation 190
9.3 The Globalization Challenge:
Is Mobile Capital a Bad Thing to Tax? 195
9.4 Low-Income Support Policies in Developed
and Developing Economies 201
9.5 Multiple Equilibria 206
9.6 Conclusions 213

Chapter 10: Traditional Growth Theory 215

10.1 Introduction 215

10.2 The Solow Model 216
10.3 Exogenous Growth with Micro-Foundations 222
10.4 A Benefit-Cost Analysis of Debt Reduction 226
10.5 Natural Resources and the Limits to Growth 232
10.6 Conclusions 235

Chapter 11: New Growth Theory 236

11.1 Introduction 236

11.2 A One-Sector Endogenous Growth Model 236
11.3 Two-Sector Endogenous Growth Models 240
11.4 Other Models of Endogenous Growth 245
11.5 An Evaluation of Endogenous Growth Analysis 252
11.6 Conclusions 255
vi
Chapter 12: Growth Policy 257

12.1 Introduction 257

12.2 Tax Reform: Income Taxes
vs. the Progressive Expenditure Tax 258
12.3 Economic Growth and Subjective Happiness 270
12.4 Unemployment and Growth 273
12.5 The Aging Population and Future Living Standards 278
12.6 Conclusions 281

Reference 283

vi i
Chapter 1

Keynes and the Classics

1.1 Introduction
Almost 70 years have elapsed since the publication of Keynes' The
General Theory of Employment, Interest and Money, yet the controversies
between his followers and those macroeconomists who favour a more
classical approach have remained active. One purpose of this book is to
examine some of these controversies, to draw attention to developments
—
that have led to a important
synthesis of mpoant ideas from both traditions, and to
illustrate in some detail how this integrated approach can inform policy
debates.
At the policy level, the hallmarks of Keynesian analysis are that
involuntary unemployment can exist and that, without government
assistance, any adjustment of the system back to the "natural"
unemployment rate is likely to be slow and to involve cycles and
overshoots. In its extreme form, the Keynesian view is that adjustment
back to equilibrium simply does not take place without policy assistance.
This view can be defended by maintaining either of the following
positions: (i) the economy has multiple equilibria, only one of which
involves "full" employment; or (ii) there is only one equilibrium, and it
involves "full" employment, but the economic system is unstable without
the assistance of policy, so it cannot reach the "full" employment
equilibrium on its own.
We shall consider the issue of multiple equilibria in Chapter 9. In
earlier chapters, we focus on the question of convergence to a full
equilibrium. To simplify the exposition, we concentrate on stability versus
outright instability, which is the extreme form of the issue. We interpret
any tendency toward outright instability as analytical support for the more
general proposition that adjustment between full equilibria is protracted.
In this first chapter, we examine alternative specifications of the
labour market, such as perfectly flexible money wages (the textbook
Classical model) and completely fixed money wages (the textbook
Keynesian model), to clarify some of the causes of unemployment. We
consider fixed goods prices as well (the model of generalized
disequilibrium), and then we build on this background in later chapters.
For example, in Chapter 2, we assume that nominal rigidities are only
temporary, and we consider a dynamic analysis that has Classical

1
properties in full equilibrium, but Keynesian features in the transitional
periods on the way to full equilibrium. Fifty years ago, Paul Samuelson
labelled this class of dynamic models the Neoclassical Synthesis.
In Chapter 3, we enrich this dynamic analysis by exploring
alternative ways of bringing expectations into the analysis. With
expectations involved, it is not obvious that an increased degree of price
flexibility lessens the amount of cyclical unemployment that follows from
a decrease in aggregate demand. By the end of Chapter 3, we will have
identified two important considerations that make macroeconomic
convergence more problematic: firms' reactions to sticky prices and sales
constraints, and expectations.
In Chapter 4, we rectify one major limitation of the analysis to
that point — that formal micro-foundations have been missing. The inter-
temporal optimization that is needed to overcome this limitation is
explained in Chapter 4. Then, in Chapter 5, we examine the New Classical
approach to business cycle analysis — the modern, more micro-based
version of the market-clearing approach to macroeconomics, in which no
appeal to sticky prices is involved. Finally, in Chapters 6 and 7, we
examine what has been called the "New" Neoclassical Synthesis — a
business-cycle analysis that blends the microeconomic rigour of the New
Classicals with the empirical applicability that has always been the focus
of the Keynesian tradition and the original Neoclassical Synthesis.
For the remainder of the book (the final five chapters), the focus
shifts from short-run stabilization issues to concerns about long-run living
standards. In these chapters, we focus on structural unemployment and the
challenge of raising productivity growth.

1.2 Criteria for Model Selection

To ensure a useful selection of macro models, economists rely on two

broad criteria. First, models must be subjected to empirical tests, to see
whether the predictions are consistent with actual experience. This
criterion is fundamentally important. Unfortunately, however, it cannot be
the only one for model selection, since empirical tests are often not
definitive. Thus, while progress has been made in dev'eloping applied
methods, macroeconomists have no choice but to put at least some weight
on a second criterion for model evaluation.
Since the hypothesis of constrained maximization is at the core of
our discipline, all modern macroeconomists agree that macro models
should be evaluated as to their consistency with optimizing underpinnings.
Without a microeconomic base, there is no well defined basis for arguing
2
that either an ongoing stabilization policy, or an increase in the average
growth rate, improves welfare. Increasingly, Keynesians have realized
that they must acknowledge this point. Further, the challenge posed by
New Classicals has forced Keynesians to admit that it is utility and
production functions that are independent of government policy; agents'
decision rules do not necessarily remain invariant to shifts in policy. A
specific microeconomic base is required to derive how private decision
rules may be adjusted in the face of major changes in policy. Another
advantage is that a specific microeconomic rationale imposes more
structure on macro models, so the corresponding empirical work involves
fewer "free" parameters (parameters that are not constrained by
theoretical considerations and can thus take on whatever value will
maximize the fit of the model). It must be admitted that the empirical
success of a model is compromised if the estimation involves many free
parameters.
Despite these clear advantages of an explicit microeconomic base,
those who typically stress these points — the New Classicals — have had to
make some acknowledgments too. They have had to admit that, until
recently, their models have been inconsistent with several important
empirical regularities. As a result, they, like Keynesians, now allow for
some temporary stickiness in nominal variables. Also, since the primary
goal of this school of thought is to eliminate arbitrary assumptions, its
followers cannot downplay the significance of aggregation issues or of the
non-uniqueness problem that often plagues the solution of their models.
These issues have yet to be resolved in a satisfactory manner.
During the 1970s and 1980s, controversy between New Classicals
and Keynesians was frustrating for students. Each group focused on the
advantages of its own approach, and tended to ignore the legitimate
criticisms offered by the "other side." The discipline was fragmented into
two schools of thought that did not interact. In the 1990s however, there
was an increased willingness on the part of macroeconomists to combine
the best features of the competing approaches so that now the subject is
empirically applicable, has solid micro-foundations, and it allows for
market failure — so economic policy can be explored in a rigorous fashion.
Students can now explore models that combine the rigour of the New
Classicals with the policy concern that preoccupies Keynesians.
The purpose of any model is to provide answers to a series of if-
then questions: if one assumes a specified change in the values of the
exogenous variables (those determined outside of the model), what will
happen to the set of endogenous variables (those determined within the
model)? A high degree of simultaneity seems to exist among the main
endogenous variables (for example, household behaviour makes
3
consumption depend on income, while the goods market-clearing
condition makes income depend on consumption). To cope with this
simultaneity, we define macro models in the form of systems of equations
for which standard solution techniques (either algebraic or geometric) can
be employed. A model comprises a set of structural equations, that are
either definitions, equilibrium conditions, or behavioral reaction functions
assumed on behalf of agents. The textbook Classical model, the textbook
Keynesian model, and the "more Keynesian" model of generalized
disequilibrium (all summarized graphically in later sections of this chapter;
are standard examples.
In constructing these models, macroeconomists have disciplined
their selection of alternative behavioural rules by appealing to
microeconomic models of households and firms. In other words, their
basis for choosing structural equations is constrained maximization at the
individual level, without much concern for problems of aggregation. To
keep the analysis manageable, macroeconomists sometimes restrict
attention to particular components of the macroeconomy, considered one
at a time. They record the resulting decision rules (the consumption
function, the investment function, the money-demand function, the
Phillips curve, and so on, which are the first-order conditions of the
constrained maximizations) as a list of structural equations. This series of
equations is then brought together for solving as a standard set of
simultaneous equations in which the unknowns are the endogenous
variables.
In other words, the procedure has two stages:

Stage 1: Derive the structural equations, which define the macro

model, by presenting a set of (sometimes unconnected)
constrained maximization exercises (that is, define and solve a
set of microeconomic problems).
Stage 2: Use the set of structural equations to derive the solution
or reduced form equations (in which each endogenous variable
is related explicitly to nothing but exogenous variables and
parameters) and perform the counterfactual exercises (for
example, derivation of the policy multipliers).

Before 1970, macroeconomics developed in a fairly orderly way,

following this two-stage approach. In recent decades, however, the
discipline has seen some changes in basic approaches following from the
fact that macroeconomists have tried to consider ever more consistent and
complicated theories of household and firm behaviour. That is, the
specification of the constrained maximizations in stage 1 of the analysis
4
has been made more general by allowing for such things as dynamics and
the fact that agents must make decisions on the basis of expectations of
the future.
This expansion has led to some conceptual and methodological
complications. Many analysts now regard it as unappealing to derive any
one component structural equation without reference at stage 1 to the
properties of the overall system. For example, if agents' behavior turns
out to depend on expected inflation, it is tempting to model their forecast
of inflation so that it is consistent with the actual inflation process, which
is determined as one of the endogenous variables within the model. From
a technical point of view, such an approach means that stages 1 and 2
must be considered simultaneously. It also means that the form of at least
some of the structural equations and, therefore, the overall structure of the
model itself depends on the assumed time paths of the exogenous
variables. Thus, it may be a bad practice for economists to use an
estimated model found suitable for one data period as a mechanism for
predicting what would happen in another period under a different set of
policy reactions. We shall consider this problem, which is referred to as
the Lucas critique, in later chapters. Initially, however, we restrict
attention to models whose structures are assumed to be independent of the
behaviour of the exogenous variables. The textbook Keynesian and
classical models (covered in the remainder of this chapter) are examples
3f such models.

1.3 The Textbook Classical Model: the Labour Market with Flexible
,

Wages

The classical macro model is defined by the following equations:

Y = C[(1 — k)Y] + I(r)+G (1.1)

L(Y,r)= MI P (1.2)
Y = F(N,K) (1.3)
W = PFN (N,K) (1.4)
W(1— k)=PS(N) (1.5)

Equations (1.1) and (1.2) are the IS and LM relationships; the

;ymbols Y, C, I, G, M, P, k and r denote real output, household
:onsumption, firms' investment spending, government program spending,
he nominal money supply, the price of goods, the proportional income
ax rate, and the interest rate. Since we ignore expectations at this point,
5
 anticipated inflation is assumed to be zero, so there is no difference
between the nominal and real interest rates. The standard assumptions
concerning the behavioural equations (with partial derivatives indicatec
by subscripts) are: I„ L,. < 0,11 > 0,0 < k, CYO , <1. The usual specificatior
of government policy (that G, k and M are set exogenously) is alsc
imposed. The aggregate demand for goods relationship follows from the
IS and LM functions, as is explained below.
Equations (1.3), (1.4), and (1.5) are the production, labour
demand, and labour supply functions, where W, N and K stand for the
nominal wage rate, the level of employment of labour and the capital
stock. The assumptions we make about the production function are
standard (that is, the marginal products are positive and diminishing)
FA„FK ,Fmc = F/37 >0,Fmv ,F,,c <0. Equation (1.4) involves the assumption of
profit maximization: firms hire workers up to the point that labour's
marginal product equals the real wage. It is assumed that it is not optimal
for firms to follow a similar optimal hiring rule for capital, since there are
installation costs. The details of this constraint are explained in Chapter 4:
here we simply follow convention and assume that firms invest more in
new capital, the lower are borrowing costs. We allow for a positively
sloped labour supply curve by assuming SN > 0. Workers care about the
after-tax real wage, W(1 — k)/P.
In the present system, the five equations determine five
endogenous variables: Y, N, r, P, and W. However, the system is not fully
simultaneous. Equations (1.4) and (1.5) form a subset that can determine
employment and the real wage w = W/P. If the real wage is eliminated by
substitution, equations (1.4) and (1.5) become F,(N,K)= S(N)I(1— k).
Since k and K are given exogenously, N is determined by this one
equation, which is the labour market equilibrium condition. This
equilibrium value of employment can then be substituted into the
production function, equation (1.3), to determine output. Thus, this model
involves what is called the Classical Dichotomy: the key real variables
(output and employment) are determined solely on the basis of aggregate
supply relationships (the factor market relations and the production
function), while the demand considerations (the IS and LM curves)
determine the other variables (r and P) residually.
The model can be pictured in terms of aggregate demand and
supply curves (in price-output space), so the term "supply-side
economics" can be appreciated. The aggregate demand curve comes from
equations 1.1 and 1.2. Figure 1.1 gives the graphic derivation. The
aggregate demand curve in the lower panel represents all those
combinations of price and output that satisfy the demands for goods and
6
assets. To check that this aggregate demand curve is negatively sloped, we
take the total differential of the IS and LM equations, set the exogenous
variable changes to zero, and solve for (dP I dY) after eliminating (dr) by
substitution. The result is

Slope of the aggregate demand curve = (rise/run in P-Y space):

dP I dY =—[L y I,. + L, (1— Cyd (1 k))] 1[1,,M I P2 ] < 0 (1.6)

The aggregate supply curve is vertical, since P does not even

enter the equation (any value of P, along with the labour market-clearing
level of Y, satisfies these supply conditions). The summary picture, with
shift variables listed in parentheses, is shown in Figure 1.2. The key
policy implication is that the standard monetary and fiscal policy variables,
G and M, involve price effects only. For example, complete "crowding
out" follows increases in government spending (that is, output is not
affected). The reason is that higher prices shrink the real value of the
money supply so that interest rates are pushed up and pre-existing private
investment expenditures are reduced. Nevertheless, tax policy has a role to
play in this model, A tax cut shifts both the supply and the demand curves
to the right. Thus, output and employment must increase, although price
may go up or down. Blinder (1973) formally derives the (dP/dk)
multiplier and, considering plausible parameter values, argues that it is
negative. "Supply-side" economists are those who favour applying this
"textbook Classical model" to actual policy making (as was done in the
United States in the 1980s).
From a graphic point of view, the "Classical dichotomy" feature
of this model follows form the fact that it has a vertical aggregate supply
curve. But the position of this vertical line can be shifted by tax policy. A
policy of balanced-budget reduction in the size of government makes
some macroeconomic sense here. Cuts in G and k may largely cancel each
other in terms of affecting the position of the demand curve, but the lower
tax rate stimulates labour supply, and so shifts the aggregate supply curve
for goods to the right. Workers are willing to offer their services at a
lower before-tax wage rate, so profit-maximizing firms are willing to hire
more workers. Thus, according to this model, both higher output and
lower prices can follow tax cuts.
This model also suggests that significantly reduced prices can be
assured (without reduced output rates) if the money supply is reduced.
Such a policy shifts the aggregate demand curve to the left but does not
move the vertical aggregate supply curve. In the early 1980s, several

7
Western countries tried a policy package of tax cuts along with decreased
money supply growth; the motive for this policy package was, to a large
extent, the belief that the Classical macro model has some short-run policy
relevance. Such policies are controversial, however, because various
analysts believe that the model ignores some key questions. Is the real-
world supply curve approximately vertical in the short run? Are labour
Figure 1.1 Derivation of the Aggregate Demand Curve

P2
P1
Po

Figure 1.2 Aggregate Demand and Supply Curves

P S(K,k)
supply elasticities large enough to lead to a significant shift in aggregate
supply? Many economists doubt that these conditions are satisfied.
Another key issue is the effect on macroeconomic convergence of the
growing government debt that accompanies this combination policy of
decreased reliance on taxation and money issue as methods of government
finance. The textbook Classical model abstracts from this consideration.
An explicit treatment of government debt is considered later in this book
(in Chapter 7), and a negative verdict on the possibility of tax cuts paying
for themselves is available in Mankiw and Weinzierl (2006).
Before leaving the textbook Classical model, we summarize a
graphic exposition that highlights both the goods market and the labour
market. In Figure 1.3, consider that the economy starts at point A. Then a
decrease in government spending occurs. The initial effect is a leftward
shift of the IS curve (and therefore, in the aggregate demand curve). At the
initial price level, aggregate supply exceeds aggregate demand. The result
is a fall in the price level, and this (in turn) causes two shifts in the labour
market quadrant of Figure 1.3: (1) labour demand shifts down (because of
the decrease in the marginal revenue product of labour); and (2) labour
supply shifts down by the same proportionate amount as the decrease in
the price level (because of workers' decreased money-wage claims). Both
workers and firms care about real wages; had we drawn the labour market
with the real wage on the vertical axis, neither the first nor the second
shift would occur. These shifts occur because we must "correct" for
having drawn the labour demand and supply curves with reference to the
nominal wage. The final observation point for the economy is B in both
bottom panels of Figure 1.3. The economy avoids ever having a recession
in actual output and employment since the shock is fully absorbed by the
falling wages and prices. These fixed levels of output and employment are
often referred as economy's "natural rates" (denoted here by P.- and N ).
Many economists find this model unappealing for two reasons.
First, they think they do observe recessions in response to drops in
aggregate demand. Second, adjustment within this model involves firms
that are perfectly happy to let inventories accumulate. A series of large
decreases in aggregate demand would cause a dramatic increase in
inventories, yet firms apparently never want to work them down since the
model shows no layoffs.
This implicit build-up of inventories will be particularly acute if
the economy is characterized by the phenomenon that Keynes called a
"liquidity trap". This special case can be considered by letting the interest
sensitivity of money demand become very large: L, > 00. By checking
— —

the slope expression for the aggregate demand curve (equation (1.6)
above), the reader can verify that this situation involves the aggregate
9
demand curve being so steep that it is almost vertical. Thus, when this
curve shifts to the left, it may no longer intersect the aggregate supply
curve anywhere in the positive quadrant. In this situation, falling wages
and prices cannot eliminate the recession. Indeed, no consistent full
equilibrium exists in this case. The Classical model can, however, be
modified to avoid this problem by allowing the household consumption-
savings decision to depend on the quantity of liquid assets available - by
making the consumption function C[(1-k)Y, MIP]. The second term in this
function is referred to as the Pigou effect.

Figure 1.3 The Classical Model

Y Production Function

P• S(N)
(1-k)
S

N
N Y
Labour Market Aggregate Supply and Demand

Macro models focusing on inventory fluctuations were very

popular many years ago (see, for example, Metzler (1941)). Space
limitations preclude our reviewing these analyses, but the reader is
encouraged to consult Blinder (1981). Suffice it to say here that
10
macroeconomic stability is problematic when firms try to work off large
inventory holdings since periods of excess supply must be followed by
periods of excess demand. Thus, it is very difficult to avoid overshoots
when inventories are explicitly modelled.
What changes are required in the Classical model to make the
system consistent with the existence of recessions and unemployment?
We consider the New Classical's response to this question in Chapter 5.
But here we focus on the traditional responses to this question, so that we
can clarify what most economists have assumed to be the major
assumptions behind Keynes' analysis. Keynes considered: (1) money-
wage rigidity; (2) a model of generalized disequilibrium involving both
money-wage and price rigidity; and (3) expectations effects that could
destabilize the economy. The first and second points can be discussed in a
static framework and so are analyzed in the remainder of this introductory
chapter. The third point requires a dynamic analysis, which will be
undertaken in Chapters 2 and 3.

1.4 The Textbook Keynesian Model: The Labour Market with

Money-Wage Rigidity

Contracts, explicit or implicit, often fix money wages for a period of time.
In Chapter 8, we shall consider some of the considerations that might
motivate these contracts. For the present, however, we simply presume the
existence of fixed money-wage contracts and we explore their
macroeconomic implications.
On the assumption that money wages are fixed by contracts for
the entire relevant short run, W is now taken as an exogenous variable
stuck at value W. Some further change in the model is required, however,
since otherwise we would now have five equations in four unknowns — Y,
V, r, and P.
Since the money wage does not clear the labour market in this
case, we must distinguish actual employment, labour demand, and labour
supply, which are all equal only in equilibrium. The standard assumption
n disequilibrium analyses is to assume that firms have the "right to
-nanage” . the size of their labour force during the period after which the
wage has been set. This means that labour demand is always satisfied, and
hat the five endogenous variables are now Y,r,P,N,N s , where the latter
variable is desired labour supply. Since this variable occurs nowhere in
he model except in equation (1.5), that equation solves residually for Ns .

11
Actual employment is determined by the intersection of the labour
demand curve and the given money wage line.

Figure 1.4 Fixed Money Wages and Excess Labor Supply

N
V/-

Figure 1.4 is a graphic representation of the results of a decrease

in government spending. As before, we start from the observation point A
and assume a decrease in government spending that moves the aggregate
demand curve to the left. The resulting excess supply of goods causes
price to decrease, with the same shifts in the labour demand and the labour
supply curves as were discussed above. The observation point becomes B
in both panels of Figure 1.4. The unemployment rate, which was zero, is
now BD/CD. Unemployment has two components: layoffs, AB, plus
increased participation in the labour force, AD.
The short-run aggregate supply curve in the Keynesian model is
positively sloped, and this is why the model does not display the classical
dichotomy results (that is, why demand shocks have real effects). The
reader can verify that the aggregate supply curve's slope is positive, by
taking the total differential of the key equations ((1.3) and (1.4)) while
imposing the assumptions that wages and the capital stock are fixed in the
short run (that is, by setting dW = dK = 0 ). After eliminating the change
in employment by substitution, the result is the expression for the slope of
the aggregate supply curve:

dP I dY =—PF,,,s I FN 2 >0. (1.7)

The entire position of this short-run aggregate supply curve is

shifted up and in to the left if there is an increase in the wage rate (in
12
 symbols, an increase in W). A similar change in any input price has the
same effect. Thus, for an oil-using economy, an increase in the price of oil
causes stagflation — a simultaneous increase in both unemployment and
inflation.
Additional considerations can be modelled on the demand side of
the labour market as well. For example, if we assume that there is
monopolistic competition, the marginal-cost-equals-marginal-revenue
condition becomes slightly more complicated. Marginal cost still equals
W FN , but marginal revenue becomes equal to [1-11(ne)]P, where n
and c are the number of firms in the industry (economy) and the elasticity
of the demand curve for the industry's (whole economy's) output. In this
case, the number of firms becomes a shift influence for the position of the
demand curve for labour. If the number of firms rises in good times and
falls in bad times, the corresponding shifts in the position of the labour
demand curve (and therefore in the position of the goods supply curve)
generate a series of booms and recessions. And real wages will rise during
the booms and fall during the recessions (that is, the real wage will move
pro-cyclically). But this imperfect-competition extension of the standard
textbook Keynesian model is rarely considered. As a result, the following
summary is what has become conventional wisdom.
Unemployment occurs in the Keynesian model because of wage
rigidity. This can be reduced by any of the following policies: increasing
government spending, increasing the money supply, or reducing the
money wage (think of an exogenous decrease in wages accomplished by
policy as the static equivalent of a wage guidelines policy). These policy
propositions can be proved by verifying that dN I dG,dN I dM > 0 and
that dNIdiff <0. Using more everyday language, the properties of the
perfect-competition version of the rigid-money-wage model are:

1. Unemployment can exist only because the wage is "too high".

2. Unemployment can be lowered only if the level of real
incomes of those already employed (the real wage) is reduced.
3. The level of the real wage must correlate inversely with the
level of employment (that is, it must move contracyclically).

Intermediate textbooks call this model the Keynesian system. Many

economists who regard themselves as Keynesians have a difficult time
accepting these three propositions, however. They know that Keynes
argued, in Chapter 19 of The General Theory, that large wage cuts might
have only worsened the Depression of the 1930s. They feel that
unemployment stems from some kind of market failure, so it should be
13
possible to help unemployed workers without hurting those already
employed. Finally, they have observed that there is no strong contra-
cyclical movement to the real wage; indeed, it often increases when
employment increases (see Solon et al (1993) and Huang et al (2004)).

1.5 Generalized Disequilibrium: Money-Wage and Price Rigidity

These inconsistencies between Keynesian beliefs on the one hand and the
properties of the textbook (perfect competition version of the) Keynesian
model on the other suggest that Keynesian economists must have
developed other models that involve more fundamental departures from
the Classical system. One of these developments is the generalization of
the notion of disequilibrium to apply beyond the labour market, a concept
pioneered by Barro and Grossman (1971) and Malinvaud (1977).
If the price level is rigid in the short run, the aggregate supply
curve is horizontal. There are two ways in which this specification can be
defended. One becomes evident when we focus on slope expression (1.7).
This expression equals zero if FNN = 0. To put the point verbally, the
marginal product of labour is constant if labour and capital must be
combined in fixed proportions. This set of assumptions — rigid money
wages and fixed-coefficient technology — is often appealed to in defending
fixed-price models. (Note that these models are the opposite of supply-
side economics since with a horizontal supply curve, output is completely
demand-determined, not supply-determined.)
Another defence for price rigidity is simply the existence of long-
term contracts fixing the money price of goods as well as factors. To use
this interpretation, however, we must re-derive the equations in the macro
model that relate to firms, since if the goods market is not clearing, it may
no longer be sensible for firms to set marginal revenue equal to marginal
cost. This situation is evident in Figure 1.5, which shows a perfectly
competitive firm facing a sales constraint. If there were no sales constraint
the firm would operate at point A, with marginal revenue (which equals
price) equal to marginal cost. Since marginal cost = f(dN / dY)= W I F N ,

this is the assumption we have made throughout our analysis of Keynesian

models up to this point. But if the market price is fixed for a time (at P)
and aggregate demand falls so that all firms face a sales constraint (sales
fixed at Y ), the firm will operate at point B. The marginal revenue
schedule now has two components: 15B and YD in Figure 1.5. Thus,
marginal revenue and marginal cost diverge by amount BC.

14
 Figure 1.5 A Competitive Firm Facing a Sales Constraint
Price
Marginal Cost Marginal Cost

Output

Sales constraint

We derive formally the factor demand equations in Chapter 4 —

both those relevant for the textbook Classical and textbook Keynesian
models (where there is no sales constraint), and those relevant for this
generalized disequilibrium version of Keynesian economics. Here, we
simply assert the results that are obtained in the sticky-goods-price case.
First, the labour-demand curve becomes a vertical line (in wage-
employment space). The corresponding equation is simply the production
function — inverted and solved for N which stipulates that labour
—

demand is whatever solves the production function after the historically

determined value for the capital stock and the sales-constrained value for
output have been substituted in. The revised investment function follows
immediately from cost-minimization. Firms should invest more in capital
whenever the excess of capital's marginal product over labour's marginal
product is bigger than the excess of capital's rental price over labour's
rental price. Using 5 to denote capital's depreciation rate, the investment
function that is derived in Chapter 4 is

I = a[(FK (W I P))I(FN (r + 8)) — (1.8)

The model now has two key differences from what we labelled
the textbook Keynesian model. First, labour demand is now independent
of the real wage, so any reduction in the real wage does not help in raising
employment. Second, the real wage is now a shift variable for the IS curve,
and therefore for the aggregate demand curve for goods, so wage cuts can
decrease aggregate demand and thereby lower employment. (This second
point is explained more fully below.) These properties can be verified
formally by noting that the model becomes simply equations 1.1 to 1.3 but

15
with W and P exogenous and with the revised investment function
replacing I(r). The three endogenous variables are Y, r, and N, with N
solved residually by equation 1.3.
The model is presented graphically in Figure 1.6. The initial
observation point is A in both the goods and labour markets. Assume a
decrease in government expenditure. The demand for goods curve moves
left so firms can only sell y ; the labour demand curve becomes the
line, and the observation point moves to point B in both diagrams.
Unemployment clearly exists. Can it be eliminated? Increases in M or G
would shift the demand for goods back, so these policies would still work.
But what about a wage cut? If the W line shifts down, all that happens is
that income is redistributed from labour to capitalists (as shown by the
shaded rectangle). If capitalists have a smaller marginal propensity to
consume than workers, the demand for goods shifts further to the left,
leading to further declines in real output and employment. The demand for
goods shifts to the left in any event, however, since given the modified
investment function (equation 1.8), the lower wage reduces investment.
Thus, wage cuts actually make unemployment worse.

Figure 1.6 The Effects of Falling Demand with Fixed Wages and Prices

Iv N

Some Keynesians find this generalized disequilibrium model

appealing since it supports the proposition that activist aggregate demand
policy can still successfully cure recessions while wage cuts cannot. Thus
the unemployed can be helped without taking from workers who are
already employed (that is, without having to lower the real wage).
16
However, the prediction that wage cuts lead to lower employment requires
the assumption that prices do not fall as wages do. In Figure 1.6, the
reader can verify that if both the given wage and price lines shift down (so
that the real wage remains constant), output and employment must
increase. The falling price allows point B to shift down the dashed
aggregate demand curve, so the sales-constrained level of output rises
(sales become less constrained). Further, with a less binding sales
constraint, the position of the vertical labour demand curve shifts to the
right.
Many economists are not comfortable with the assumption that
goods prices are more sticky than money wages. This discomfort forces
them to downplay the significance of the prediction that wage cuts could
worsen a recession, at least as shown in generalized disequilibrium
models of the sort just summarized. For this reason, we do not pursue
these models further in this text. (For a brief but excellent survey of this
literature, we refer the reader to Stoneman 1979.)

1.6 Conclusions

In this chapter we have reviewed Keynesian and Classical interpretations

of the goods and labour markets. We have established, among other points,
that unemployment can exist only in the presence of some stickiness in
money wages. Appreciation of this fact naturally leads to a question:
;hould we advocate increased wage flexibility so as to avoid at least some
memployment? There are two general responses. First, one can say that
mivate agents must have adopted the institution of temporarily rigid
wages (contracts) for a reason and that reason must be understood before
me can be confident that increasing wage flexibility is "good." (Some
nicroeconomic theories of wage rigidity are considered in Chapter 4, 6
ind 9.) Second, one can presume that the microeconomic costs of
ncreased flexibility would not be large and, therefore, directly proceed
vith a macroeconomic analysis -of whether the built-in stability of the
werall economy is enhanced by wage flexibility. In section 1.4, we
.eviewed one argument against this possibility. The model of generalized
lisequilibrium supports the proposition that wage cuts could worsen a
ecess ion.
We shall consider other ways in which standard models have
:hanged to generate conclusions that are more in line with Keynes'
:oncerns. In particular, in Chapter 3, we shall consider another
riechanism — expectations — whose presence can make wage cuts during
ecessions undesirable. In addition, we consider market failure in labour
17
markets in Chapter 8. In this setting, involuntary unemployment emerges
as a well-explained equilibrium outcome, and multiple equilibria can exist
(so that the economy can become stuck at a low level equilibrium, just
because agents become pessimistic and expect such an outcome). Many
economists feel that these results provide a modern underpinning for some
important Keynesian ideas that are usually ignored in textbook treatments
of "the" Keynesian model.
Despite the fact that many mainstream economists have used
modern tools to highlight some of Keynes' most central concerns, some
economists (known as Post Keynesians) still reject much of this "Nem,
Keynesian" analysis. One Post Keynesian concern is that mainstream
analysis treats uncertainty in a way that Keynes argued was silly. Keynes
followed Knight's suggestion that risk and uncertainty were
fundamentally different. Risky outcomes can be dealt with by assuming a
stable probability distribution of outcomes, but some events occur sc
infrequently that the relevant actuarial information is not available.
According to Post Keynesians, such truly uncertain outcomes simply
cannot be modelled formally.
Another concern of Post Keynesians is that there are fundamental
inconsistencies in defining an aggregate capital stock that are ignored in
standard analysis (see Cohen and Harcourt (2003)). Mainstream macro-
economists argue that the empirical success that econometricians find
when testing aggregate production relationships must mean that the
empirical relevance of this capital controversy is limited. However.
Shaikh (1974) has demonstrated that these tests have very low power.
An excellent general summary and introduction to the Post
Keynesian approach is contained in Wolfson (1994), and interested
readers are encouraged to pursue this reference. Given our objective of
providing a text that focuses on what is usually highlighted in a one-
semester course, we cannot afford to consider Post Keynesian analysis
further in this book. Instead, we focus on the New Classical revival, and
we cover the research of "New Keynesians" in later chapters.

18
Chapter 2

The First Neoclassical Synthesis

2.1 Introduction

The traditional (that is, pre New Classical) analysis of economic cycles
involved a compact structure that included the textbook Classical and
Keynesian models as special cases. This simple — yet encompassing —
framework was achieved by dropping any explicit treatment of the labour
market (and the production function). Instead, a single summary
relationship of the supply side of the goods market was specified. That
one function was an expectations-augmented Phillips curve — a
relationship that imposes temporary rigidity for goods prices in the short
run, but Classical-dichotomy (natural-rate) features in full equilibrium.
This simple, but complete, model of simultaneous fluctuations in real
output and inflation consisted of two equations: a Phillips curve (the
supply-side specification), and a summary of IS-LM theory — a simple
reduced-form aggregate-demand function. The purpose of this chapter is
to review the properties of this standard dynamic model.

2.2 A Simple Dynamic Model — Keynesian Short-Run Features and a

Classical Full-Equilibrium

As just noted, traditional dynamic analysis combined a simple aggregate

demand function and a Phillips curve, and expectations were ignored. The
aggregate demand function was a summary of the IS-LM system. To
proceed in a specific manner, we assume the following linear
relationships:

y = —ar + fig, the IS function, and

m— p = y y — fir , the LM relationship.

y, g, m and p denoted the natural logarithms of real output, autonomous

expenditure (sometimes assumed to be government spending), the
nominal money supply, and the price level. r is the level (not the
logarithm) of the interest rate, and since (for this chapter) we assume that
expectations of inflation are always zero, r is both the real and nominal
interest rate. The Greek letters are positive slope parameters.

19
 These IS and LM relationships can be combined to yield the
aggregate demand function (by eliminating the interest rate via simple
substitution). The result is

y =0(m— p)+,g (2.1)

where
0 = 1(ay +SI)
= fit /(ay + f2) .

This aggregate demand function is combined with a standard

dynamic supply function (a Phillips curve) in which the "core" inflation
rate is assumed to be zero.

= 0(Y 5) (2.2)

The new notation, .57 and TC, denote the natural rate of output (the
value that emerges in the textbook Classical model, and a value we take as
an exogenous variable in the present chapter) and the core inflation rate.
Since p is the logarithm of the price level, its absolute time change equals
the percentage change in the price level. Thus, p is the inflation rate.
Initially, the core inflation rate is assumed to be zero. Later on in the
chapter, the core inflation rate is assumed to equal the full-equilibrium
inflation rate, and since we assume a constant natural rate of output, this
core inflation rate is simply equal to the rate of monetary expansion:
= m . If we assume the rate of monetary expansion to be zero, there is
no difference between these specifications.
The full equilibrium properties of this system are: y =j7,
p =ir =th and 7 = (fig — 37)1a, so (as already noted) macroeconomists
talk in terms of the "natural" output rate, the "natural" interest rate, and
the proposition that there is no lasting inflation-output trade-off. Milton
Friedman went so far as to claim that inflation is "always and
everywhere" a monetary phenomenon, but this assertion is supported by
the model only if prices are completely flexible (that is, if parameter 40
approaches infinity). In general, the model involves simultaneous
fluctuations in real output and inflation, bringing predictions such as:
disinflation must involve a temporary recession. Such properties imply
that Friedman's claim is accurate only when comparing full long-run
equilibria. Nevertheless, the presumption that the model's full equilibrium
is, in fact, reached as time proceeds, should not be viewed as terribly

20
controversial, since it turns out that this model's stability condition can be
violated in only rather limited circumstances.
Keynes' approach to macroeconomics involved the concern that
convergence to a classical full-equilibrium should not be presumed.
Indeed, Keynes argued that a central job for macro theory was to identify
those circumstances when convergence is unlikely, so that policy can be
designed to ensure that real economies do not get into these
circumstances. So while this traditional model involves sticky prices in
the short run (and from this vantage point, at least, it is appealing to
Keynesians), the fact that — when expectations are ignored — it rejects the
possibility of instability as rather unlikely makes it offensive to
Keynesians. How has this model been altered to avoid this offensive
feature? The answer: by letting expected inflation depend on actual
inflation, and by allowing these expectations to have demand-side effects.
Thus far, we have limited the effects of anticipated inflation to the
wage/price setting process (by allowing the core, or full-equilibrium,
inflation rate to enter the Phillips curve). As an extension, we can allow
the nominal and the real interest rates to differ by peoples' expectations
concerning inflation in the short run. But before we introduce this
distinction, we discuss stability in this initial, more basic, model.
Mathematically, we can focus on the question of convergence to
full equilibrium by taking the time derivative of the aggregate demand
equation, assuming that autonomous spending and the money supply are
not changing in an ongoing fashion (setting g = m = 0) , and substituting
out p by using the Phillips curve (with it = 0). The result is

= -s(), (2.3)

where s is the stability coefficient: s = 00. For the convergence of actual

real output to the natural rate, we require that y rise when it is "too low"
and that y fall when it is "too high". These outcomes are consistent with
equation (2.3) only if parameter s is positive. Thus, the model's stability
condition is s > 0. Since summary parameter 0 is defined as
B9 = a /(ay + Q) , we see that, in general, instability is impossible. The
only problem that can develop is if the aggregate demand curve is not
negatively sloped. It can be vertical if 0 is zero, and this (in turn) is
possible if the economy gets into what Keynes called a "liquidity trap". If
the nominal interest rate approaches its lower bound of zero, the demand
for money becomes limited only by agents' wealth, and the interest
elasticity of money demand approaches infinity (SI —› oo). This situation is
sufficient to make both 0 and s equal to zero. In short, the system does not
21
converge to full employment in this case. Observing that interest rates
were essentially zero in the United States during the 1930s, and in Japan
during the 1990s, many analysts have argued that the economy's self-
correction mechanism broke down in these cases. We certainly did
observe very protracted recessions during these episodes (even the Great
Depression in the 1930s case), so it may well be appropriate to interpret
these periods in this manner. Some recent papers that examine the
implications of a liquidity trap in some detail are Svensson (2003),
Bernanke and Reinhart (2004), Eggertsson and Woodford (2004) and
Coenen et al (2006).

Figure 2.1 Short- Run and Long- Run Equilibria

Long-run
Aggregate
Supply

B Short-run
Aggregate
Supply

Aggregate
Demand

Figure 2.1 illustrates the convergence to full equilibrium in the "normal"

(non-liquidity-trap) case. The long-run aggregate supply curve is vertical
at the natural rate of output — reflecting the fact that — in full equilibrium —
this model coincides with the textbook Classical system. But the
Keynesian feature is that the price level is predetermined at each point in
time, so the instantaneous, short-run aggregate supply curve is horizontal
at that height. Normally, the aggregate demand curve is negatively sloped
(and the shift variables are the nominal money supply and the level of
autonomous spending). We consider a once-for-all drop in exogenous
spending. The aggregate demand curve shifts to the left, and the economy
moves from point A to B instantly. Output is completely demand-
determined in the instantaneous short run. But then, as time begins to
pass, prices begin to fall, and the short-run supply curve moves down to
22
reflect this fact. The economy traces along the B-to-C time path as output
comes back up to the natural rate (as point C is reached). The recession is
only temporary. But this benign conclusion does not emerge if the
aggregate demand curve is very steep (or vertical). If it is very steep, and
it moves significantly to the left, then the price level will fall to zero
before the economy gets back to the natural output rate. Keynesian
economists argue that we should not downplay this non-convergence-to-
full-employment possibility. More classically minded economists are
comfortable interpreting this possibility as just a remotely relevant
pathology. In the next section of this chapter, we see how allowing
inflationary expectations to play a role in this model makes instability a
much more realistic possibility. This is why Keynesian economists always
stress expectations.

2.3 The Correspondence Principle

We now extend our simple dynamic aggregate supply and demand model
by allowing inflationary and deflationary expectations to affect aggregate
demand. We continue to assume descriptive behavioural equations,
leaving the consideration of formal micro-foundations until Chapter 4. We
now distinguish real and nominal interest rates. The former is involved in
the IS relationship, since we assume that households and firms realize that
it is the real interest rate that represents the true cost of postponing
consumption and borrowing. But it is the nominal interest rate that
belongs in the LM equation, as long as we assume that peoples' portfolio
choice is between non-indexed "bonds" (that involve .a real return of
r =i—k ) and money (that involves a real return of -p). The real yield
iifferential is, therefore, i — the nominal interest rate. Notice that, to avoid
laving to specify a relationship between actual and expected inflation, we
lave simply assumed that they are equal. We consider alternative
>pecifications in Chapter 3. In any event, when the nominal interest rate is
liminated by substitution, the IS-LM summary is

y =0(m— p)+ yr 1.3 + fig. (2.1a)

two of the summary aggregate-demand parameters have already been

lefined earlier in the chapter. The coefficient on the new term is

yr = aS2 /(ay +

23
At the intuitive level, the basic rationale for this term is straightforward
aggregate demand is higher, if expected (equals actual) inflation rises
since people want to "beat the price increases" by purchasing the items
now. Similarly, aggregate demand is lower if people expect deflation
since in this case they want to postpone purchases so that they can benefi
from the expected lower prices in the future. Thus, while currem
aggregate demand depends inversely on the current price of goods, ii
depends positively on the expected future price of goods.
On the supply side, as long as we assume (as above) that the
natural output rate is constant, we know that the core (full-equilibrium:
inflation rate is the money growth rate, so the aggregate supply
relationship can remain as specified earlier. The model now consists of
equations (2.1a) and (2.2).
We are interested in determining how this simple economy react s
to shocks such as a once-for-all drop in autonomous spending. What
determines how real output is affected in the short run? Under what
conditions will the economy's self-correction mechanism work (that is.
under what conditions will the short-run effect — a recession — be
temporary and automatically eliminated)? Was Keynes right when he
argued that it is a "good" thing that prices are sticky? That is, is the
magnitude or duration of the recession made worse if the short-run
Phillips curve is steeper (if coefficient (I) is larger)? It is to these questions
that we now turn.
The effect of a change in autonomous expenditure on output is
calculated by substituting equation (2.2) into equation (2.1 a) to eliminate
the inflation rate. Further, we simplify by setting 7r = m = 0. The resulting
at-a-point-in-time reduced form for output is

Y = [1 — OVA {19(m P) w05; + gJ (2.4)

Taking the derivative with respect to g, we have the impact effect:

dy I dg — 0V)

which is positive only if the denominator is positive. Thus, the model only
supports the proposition that a drop in demand causes a recession if the
denominator is positive.
It may seem surprising that — in such a simple model as this one —
our basic assumptions about the signs of the parameters are not sufficient
to determine the sign of this most basic policy multiplier. If we cannot
"solve" this problem in such a simple setting, we will surely be plagued

24
with sign ambiguities in essentially all macro models that are more
complicated than this one. Macroeconomists have responded to this
problem in three ways. First, on the basis of empirical work, theorists
have become more confident in making quantitative assumptions about
the model's parameters, not just qualitative (or sign) assumptions. But
given the controversy that surrounds most econometric work, this strategy
has somewhat limited appeal. The second approach is to provide more
explicit micro-foundations for the model's behavioural equations. By
having a more specific theory behind these relationships, we have more
restrictions on the admissible magnitudes for these structural coefficients.
While this approach limits the model's sign ambiguity problems, as we
shall see in Chapter 3, it does not fully eliminate them. Thus, some
reliance must remain on what Paul Samuelson called the correspondence
principle many years ago. He assumed that the least controversial
additional assumption that can be made concerning the model's
parameters (other than their signs) is to assume that — given infinite time —
the system will eventually converge to its full equilibrium. After all, most
economists presume that we eventually get to equilibrium. To exploit this
belief, Samuelson's recommendation was to derive system's dynamic
stability condition, and then to use that condition as a restriction to help
sign the corresponding comparative static multipliers. Since macro-
economists are assuming eventual convergence implicitly, Samuelson felt
that nothing more of substance is being assumed when that presumption is
made more explicit to sign policy multipliers. This has been standard
procedure in the profession for 60 years, and we will apply the
correspondence principle in our analysis here. But before doing so, we
note that some macroeconomists regard the use of the correspondence
principle as suspect.
The dissenters can see that there is an analogy between
macroeconomists using the correspondence principle and micro-
economists focusing on second-order conditions. Microeconomists use the
second-order conditions to resolve sign ambiguities in their analyses — that
are based on agents obeying the first-order conditions. There is no
controversy in this case, because the second-order conditions are an
integral part of the model, and analysts are simply making implicit
assumptions explicit. But the analogy between second-order conditions in
micro and dynamic stability conditions in macro breaks down since, in
most macro models, analysts have the freedom to specify more than one
set of assumptions for the model's dynamics. Thus, there is an element of
arbitrariness in macro applications of the correspondence principle that is
not present when microeconomists rely on second-order conditions for
additional restrictions. One of the purposes of providing explicit micro-
25
foundations for macroeconomics is to discipline macro model builders so
that they have less opportunity for making what others might regard as
arbitrary assumptions.
A more fundamental problem with the correspondence principle is
that some economists (for example, Keynes) are not prepared to assume
stability. Indeed, some of them can be viewed as arguing that this issue
should be the fundamental focus of research (see Tobin 1975, 1980; and
Hahn and Solow 1986). According to this approach, we should compare
the stability conditions under alternative policy regimes, to see whether or
not a particular policy is a built-in stabilizer. Policy regimes that lead to
likely instability should be avoided. Thus, even though the stability
conditions are not presumed to hold by all analysts, all economists must
know how to derive these conditions. Thus, we now consider the stability
condition for our aggregate supply and demand model.
The stability of the economy is assessed by taking the time
derivative of equation (2.4) and using equation (2.2) to, once again,
eliminate the inflation rate. In this derivation, we assume that there are no
further (ongoing) changes in autonomous spending and that the natural
rate is constant (g = y = 0 ). As in the simpler model which suppressed
the distinction between real and nominal interest rates, the result is
= —s(y — 3,-) , but now the expression for the stability parameter is

s = 0041— 0).

As before, stability requires that s be positive; in this case, we require

v0 < 1. Applying the correspondence principle, we see that this restriction
is sufficient to guarantee that the expenditure multiplier has its
conventional sign. This restriction is the familiar requirement that the
income sensitivity of total spending not be "too large." In this case, the
income sensitivity comes indirectly through the dependence of aggregate
demand on the expected rate of inflation, and through the dependence of
inflation on the output gap.

2.4 Can Increased Price Flexibility be De-Stabilizing?

We are now in a position to assess whether increased price flexibility

leads to more desirable responses to aggregate demand shocks. There are
both "good news" and "bad news" aspects to an increase in the size of
parameter 4. The good news is that the economy's speed of adjustment
back to full equilibrium (following all shocks) is higher (since the speed
26
coefficient, s, rises with (0). There are two dimensions to the "bad" news.
First, the size of the initial recession is larger; second, the likelihood of
outright instability is increased. In terms of this standard model, it appears
that Keynes expected the "bad" news effects to outweigh the "good" news
effect. Should we agree?
Before proceeding, we should consider simple intuition. How is it
possible that more flexible prices can enlarge a recession? The logic of
this result runs as follows. Given that aggregate demand depends on both
autonomous expenditure and inflationary expectations, which are
perfectly anticipated, a decrease in autonomous expenditure has both a
direct and an indirect effect. The direct effect leads to lower output. The
indirect effect follows from the fact that agents realize that the fall-off in
output will reduce inflation; other things being equal, this decrease raises
the real interest rate, so firms reduce the investment component of
aggregate demand. With increased price flexibility, this secondary effect
is larger than it would be otherwise. So the Keynesian proposition that
increased price flexibility has this "bad" aspect is supported by this
analysis which stresses expectations effects.
The other "bad" aspect of higher price flexibility is that it can
make the economy unstable. Some would argue that there is some support
for Keynes on this question as well. It is widely presumed that we have
had longer-term wage contracts since WWII, and this is captured in our
model by a smaller value for coefficient (I). Since many believe that
Western economies have displayed less volatility since WWII, there
would seem to be some evidence to support Keynes. However, Romer
(1986) has noted that — over time — there have been important changes
concerning how we measure GDP. When GDP for recent years is
measured in the same way as we were constrained to measure this
aggregate in pre-WWII days, we find that the economy is not less volatile.
Thus, the jury is still out on this question. Nevertheless, we can pursue
Keynes' conjecture more fully if we presume stability, and calculate the
cumulative output loss following a once-for-all drop in demand.
As noted already, increased price flexibility is not all "bad" in this
model. As long as the economy remains stable, we know that the speed
with which the temporary recession is eliminated is proportional to the
stability and speed parameter s. We have already seen that increased price
flexibility must raise s and so speed up the elimination of output
disturbances.
Assuming eventual convergence, the solid line in the Figure 2.2
shows the output time path following a once-for-all reduction in
autonomous spending. The dashed line indicates how output responds
when parameter 4) is a larger coefficient (on the assumption that the
27
system remains stable). The size of the initial recession is bigger, but that
larger recession is eliminated more quickly. The output time path is closer
to the natural output line in the latter part of the adjustment period (when
prices are more flexible), and Classicals often argue that this may be the
more important consideration.

Figure 2.2 Implications of Increased Price Flexibility

.
, — Output time path with low 0
— — Output time path with high 0

Time, t
Lig occurs at this time

One calculation that supports the Classicals' interpretation is the total

undiscounted output loss that follows a permanent decrease in aggregate
demand. From a geometric point of view, this measure is the area between
the output time path and the natural rate line in Figure 2.2. It can be
calculated as Hy y)dt which, as Buiter and Miller (1982) show,
—

equals the initial output loss (the impact multiplier) divided by the speed
of adjustment. Thus, in this case, the cumulative output loss is 00.
According to this method of weighting the "bad" short-run effect and the
"good" longer-run effect of a larger parameter (I), then, an increased degree
of price flexibility is deemed desirable. Of course, supporters of Keynes
can argue that what matters is a discounted cumulative output loss
calculation – not what we have just calculated. Once the short run is given
more weight than the longer run, it is immediately apparent from Figure
2.2 that the undesirable aspects of an increased degree of price flexibility
could dominate. Overall, this analysis provides at least partial analytical
support for the Keynesian proposition that increased price flexibility may

28
not help the built-in stability properties of the economy. This issue is
particularly important since many policy analysts advocate that
governments use taxes and/or subsidies to stimulate private firms and their
workers to adopt such arrangements as profit-sharing and shorter wage
contracts. One motive for encouraging these institutional changes is a
desire to increase wage flexibility (which would indirectly bring increased
price flexibility) and the proponents of these policies simply presume that
their adoption would be "good."
The opposite presumption seems to be involved at central banks,
such as the Bank of Canada. Analysts there have noted that one of the
"beneficial" aspects of our reaching price stability, is that average contract
length has increased dramatically over the last 20 years. With low
inflation, people are prepared to sign long-term wage contracts. While this
means lower industrial-dispute and negotiation costs, it also means that
the size of parameter 4 is now smaller. This development is "good" for
macroeconomic stability only if Keynes was right. It appears that the
Bank of Canada is comfortable siding with Keynes on this issue.
Before leaving this section, we investigate what things make the
economy's stability condition more or less likely to be satisfied. We
derived above that convergence to full equilibrium occurs only if Ow <1.
To assess the plausibility of this condition being met, we focus on the
detailed determinants of the reduced-form aggregate demand parameters
given earlier. Using these interpretations, the stability condition can be re-
,xpressed as

1+ (ay I SI)> (a0) •

We consider this convergence condition in two separate situations: an

ongoing inflation, such as that of the 1970s, and a deep depression, such
is that of the 1930s. In a period of inflation, the equilibrium nominal
nterest rate is high and the interest elasticity of money demand is very
ow. This situation can be imposed in the stability condition by letting
)arameter S2 approach zero, and the result is that the stability condition
nust be satisfied. The opposite case involves extremely low interest rates
as were observed throughout the world in the deep depression of the
[930s, and in Japan during its prolonged recession of the 1990s). As the
iominal interest rate approaches zero, many macroeconomists believe that
he interest elasticity of money demand, parameter SI, approaches infinity.
['his "liquidity trap" situation can make it very difficult for the stability
:ondition to be satisfied.

29
 So, under monetary aggregate targeting at least, the efficacy of the
economy's self-correction mechanism very much depends on the interest
elasticity of money demand. The intuition behind this fact is straight-
forward. Lower prices have two effects on aggregate demand. Falling
actual prices stimulate demand, and (other things equal) help end a
recession. But expectations of falling prices raise real interest rates, and
(other things equal) dampen demand and thereby worsen a recession. The
stabilizing effect of falling actual prices works through its expansionary
effect on the real money supply, while the destabilizing effect of expected
deflation works through interest rates and the associated increase in
money demand. If real money demand increases more than real money
supply, the initial short-fall in the demand for goods is increased. A
liquidity trap maximizes the chance that the money-demand effect is
stronger — making the economy unstable.
We conclude that this simple dynamic model represents a
compact system that allows for Keynes's worry that wage and price
decreases can worsen a deep recession through expectational effects. Al
the same time, however, it is important to realize that the model does not
suggest that instability must always occur. Indeed, macroeconomists can
appeal to this one simplified model, and consistently argue that
government intervention was justified in the 1930s (to avoid instability or
at least protracted adjustment problems) but was not required in the 1970s
(to avoid hyperinflation). Since any scientist wants a single, simple model
to "explain" a host of diverse situations, it is easy to see why variants of
this model represented mainstream macroeconomics for many years.
Some macroeconomists interpret the economy as having a stable
"corridor." The term corridor has been used to capture the notion that the
economic system may well be stable in the face of small disturbances that
do not push the level of activity outside its normal range of operation (the
corridor). But that fact still leaves open the possibility that large shocks
can push the economy out of the stable range. For example, as long as
shocks are fairly small, we do not get into a liquidity trap. But sometimes
a shock is big enough to make this extreme outcome relevant, and the
economy is pushed out of the stable corridor. This appears to be a
reasonable characterization of the Great Depression in the 1930s. At that
time, individuals became convinced that bankruptcies would be prevalent.
As a result, they developed an extraordinary preference for cash, and the
corresponding liquidity trap destroyed the applicability of self-correcting
mechanism. It is important for Keynesians that instability does not always
occur, so that Keynesian concerns cannot be dismissed by simply
observing that there have been many episodes which have not involved
macroeconomic breakdown. Howitt (1979) has argued that models that do
30
not permit a corridor feature of this sort cannot claim to truly represent
Keynes's ideas. While this discussion of corridors has been instructive, it
has not been completely rigorous. After all, formal modeling of corridors
would require nonlinear relationships, and our basic model in this chapter
has involved linear relationships.

2.5 Monetary Policy as a Substitute for Price Flexibility

Throughout the previous section of this chapter, we have reasoned as if

government policy could choose the degree of price flexibility. While
some countries have institutional structures that might permit this, for
many others, policy options are less direct. The government can only
affect price flexibility indirectly, through such measures as tax incentives
that might stimulate the use of profit sharing arrangements. The question
which naturally arises is whether monetary policy — an instrument over
which the government has much more direct control — can be used as a
substitute for price flexibility. Keynes believed that monetary policy could
be used in this way. Even Milton Friedman (1953) agreed; his advocacy
of flexible exchange rates was based on the presumption that this
monetary policy could substitute for flexible wages and prices.
To investigate this issue formally we examine nominal income
targeting — a monetary policy advocated by many economists recently.
Letting z = p + y denote the logarithm of nominal GDP, we can specify
the following monetary policy reaction function:

x(p+ y—T) (2.5)

where the bars denote (constant) target values and parameter x defines
alternative policies: x —> 0 implies a constant money supply (what we
have been assuming in the previous section); x —> co implies pegging
nominal income. Since the target variables are constant, the equilibrium
inflation rate is zero for both of these monetary policies. This policy
reaction function can be combined with equations (2.1a) and (2.2). Using
the methods already described, the reader can verify that the impact
autonomous spending multiplier is

dy I dg = 01— Ovf +19%),

the stability and the adjustment speed parameter is

31
 s = 600 + /(1 — Ov + ex ),

and the cumulative output loss is 4 1(t90(1+ x)). These results imply that
nominal income targeting (an increase in parameter x) reduces the size of
the impact effect which follows an aggregate demand shock, and it has an
ambiguous effect on the speed with which this temporary output deviation
is eliminated. These effects are not quite the same as we obtained for an
increase in price flexibility, but the net effect on the undiscounted
cumulative output outcome is similar. The overall output effect is made
smaller by nominal income targeting. In this sense, then, a more active
attempt to target nominal income can substitute for an attempt to vary the
degree of wage/price flexibility.
Readers may regard this analysis of monetary policy as somewhat
dated. After all, central banks no longer set policy in terms of targeting
monetary aggregates. Instead, they adjust interest rates with a view to
targeting inflation directly. Their research branches investigate whether
their interest-rate reaction function should focus on the deviation of the
inflation rate from target (the current practice), on the deviation of the
price level from target, or on the deviation of nominal GDP from target.
For the remainder of this section, we recast the analysis so that it can
apply directly to this set of questions. Also, we consider a different
specification of disturbances. Thus far, we have focused on the effects of a
once-for-all change in demand. Now, we focus on an ongoing cycle in
autonomous spending. Since modern analysis views policy as an ongoing
process, not a series of isolated events, this alternative treatment of
disturbances is appealing for an analysis that highlights model-consistent
expectations on the part of private agents who understand that policy
involves an ongoing interaction with the economy.
The revised model involves the IS relationship, the central bank's
interest-rate setting rule, the Phillips curve, and the specification of the
ongoing cycle (a sine curve) in autonomous demand:

y = —ar + fig
r+p +OA - 2(p —0)+(1— 2)(p —0)
P = 0(y — )+0
g = g + ssin(t)

The LM relationship is not used, since its only function in this

specification of monetary policy is to solve (residually) for what value of
the money supply is necessary for the central bank to deliver the nominal
interest rate that is chosen according to the bank's reaction function. We
32
 focus on a central bank that is committed to price stability. This is why
there is a zero core-inflation term in the Phillips curve, and a zero average
inflation rate term in the rate-setting equation. This relationship states that
the bank sets the current nominal interest rate above (below) its long-run
average value whenever either the inflation rate is above (below) its target
value, or whenever the price level is above (below) its target level. As
noted above, one major point of debate among central banks today is
whether they should pursue an inflation-rate target or a price-level target.
We consider this debate in this model by examining alternative values for
parameter X. Inflation targeting is involved if X = 1, while price-level
targeting is specified by X = 0. Parkin (2000) highlights the implications
of these two policy options for the real value of contracts over time.
Inflation targeting permits long-run drift in the price level, while price-
level targeting does not. Thus, if the goal is to preserve the purchasing
power of money, price-level targeting is preferred. Here, we focus on the
implications of this choice for the economy's short-run built-in stability
properties.
We proceed by deriving the reduced form for real output, to see
how the amplitude of the resulting cycle in y is affected by changes in the
monetary policy parameter (X). We choose units so that y = 0 , and
substitute the Phillips curve and interest-rate-setting relationships into the
IS function, to eliminate inflation and the interest rate:

(1— 4(1— A))y —a(1— A)p+ fig —aF.

Next, we take the time derivative and use the Phillips curve again to
eliminate inflation:

(1- coo - 2))5' =-a(1- 2)0y+ fig. (2.6)

We analyze (2.6) by positing a trial solution and using the undetermined

coefficient solution procedure. It is worth a brief aside to explain this
procedure — to clarify that it is likely to be more familiar to economics
students than they realize. Consider the familiar example of compound
interest. The basic relationship (the economic model) of compound
interest is x, (1+ r)x,_„ where r is the interest rate and x the
accumulated value. We know that the solution equation for this dynamic
process is x, = (1+ r)' xo , where x0 is the initial amount that is invested.
Let us pretend that we do not know that this is the answer. To derive this
solution equation, we simply posit a trial solution of the form x, = ,u1 A ,

33
where the arbitrary parameters, and A, are yet to be determined
Substituting the trial solution into the model, we have: µ`A = (1 + r),u' A
or p = (1+ r) . Similarly, substituting t = 0 into the trial solution, we have
A= x0 . As a result, the initially arbitrary reduced form coefficients, and
A, are now determined as functions of the economically meaningful
parameters, r and xo .

We use this same procedure for our macro model here. Following
Chiang (1984, p. 472) the (trial) solution for output can be written as

y= y + B[cos(t)] + C[sin(t)] . (2.7)

Equation (2.7), the time derivative of (2.7), y = —Bsin(t) + C cos(t) , and

the time derivative of the autonomous spending equation, k = gcos(t), are
substituted into (2.6). The resulting coefficient-identifying restrictions are

B =[ot0(1— 2,)C]I[1— ot0(1— 2)]

C =[185(1— 4(1- 2 ))] /[(a0(1— 2)) 2 + (1— a93 ( 1— 2 )) ? ].

Even for this simple model, illustrative parameter values are needed to
assess the resulting amplitude of the cycle in y. Representative values
(considering a period length of one year) are: a = 0.8 and p = 0.2
(autonomous spending is 20% of GDP), 8 = 1, and (I) = 0.2 (see Walsh
(2003a)). With these values, the amplitude of the real output cycle is equal
to one-fifth of the amplitude of the autonomous spending cycle if the
central bank targets the inflation rate (if X = 1). In contrast,' the amplitude
of the output cycle is a slightly larger fraction (0.23 instead of 0.20) of
that of the autonomous spending cycle if the central bank targets the price
level (X. = 0). According to the model, then, the contemplated move from
inflation targeting to price-level targeting is not supported.
The intuition behind this result can be appreciated by considering
an exogenous increase in the price level. With inflation-rate targeting,
such a "bygone" outcome is simply accepted, and future inflation is
resisted. But with price-level targeting, future inflation has to be less than
zero to eliminate this past outcome. That is, only under price-level
targeting is a policy-induced recession called for. So price-level targeting
seems obviously "bad." The reason that this consideration may not be the
dominant one, however, is that the avoidance of any long-term price-level
drift (that is a feature of price-level targeting) has a stabilizing effect on
expectations. For the plausible parameter values considered here, it

34
appears that the former (destabilizing) effect of price-level targeting
slightly outweighs the latter (stabilizing) effect.

2.6 Conclusions

The analysis of this chapter has involved a simple model that summarizes
mainstream macroeconomics before the rational-expectations and new-
classical "revolutions". This version of macroeconomics is called the
Neoclassical Synthesis since it combines important elements of both
Classical and Keynesian traditions. It involves both the long-run
equilibrium defined by the static textbook Classical model, and the
temporary nominal stickiness feature that is the hallmark of the textbook
Keynesian model (as the mechanism whereby the system departs from its
full equilibrium in the short run). We have used this model to explain the
correspondence principle, to examine how several monetary policies
might be used to make up for the fact that prices are not more flexible, and
to establish whether more flexible prices are desirable or not.
We have learned how important it is to add expectations to a
macro model. Initially, expectations was an issue stressed by Keynesians,
;ince it represents a mechanism that makes convergence to full
equilibrium less assured. But, more recently, macroeconomists of all
)ersuasion highlight expectations in their analysis. This is because
;tabilization policy is now modelled as an ongoing operation, not an
solated one-time event, and analysts are drawn to models in which agents
ire fully aware of what the government has been, and will be, doing.
Chus, as far as stabilization policy analysis is concerned, there has been a
;onvergence of views in the sense that all modern analysis focuses on
nodel-consistent expectations (as we did by equating actual and expected
nflation in this chapter).
We extend our appreciation of these issues in the next two
:hapters, by exploring alternative treatments of expectations in the next
:hapter, and by providing more explicit micro-foundations for the
ynthesis model in Chapter 4. Once the rational-expectations (Chapter 3)
md the new-classical (Chapter 5) revolutions have been explored, we will
► e in a position to consider an updated version of the synthesis model —

he so-called "New Neoclassical Synthesis" — in Chapters 6 and 7.

35
Chapter 3

Model-Consistent Expectations

3.1 Introduction

As noted in Chapter 1, early work in macroeconomics involved a bold

simplifying assumption — that economic agents have static expectations
concerning the model's endogenous variables. This assumption facilitated
the separation of the "stage 1" and "stage 2" analyses. The unappealing
thing about this approach is that individuals are always surprised by the
fact that these variables change. By 1970, macro theorists had come to
regard it as unappealing to model households and firm managers as
schizophrenic individuals. On the one hand, we assumed that individuals
took great pains to pursue a detailed plan when deciding how much to
consume and how to operate their firms (focusing on points of tangency
between indifference maps and constraints). However, in that traditional
approach, we assumed that these same individuals were quite content to
just presume that many important variables that affect their decisions will
never change. Why is it appealing to assume that individuals are content
to be so consistently wrong on this score, while they are so careful making
detailed plans in other dimensions of their decision making? And why
should economists be content to build models which explain the time
paths of these variables and yet involve individuals who apparently do not
believe in the relevance of these models (since they do not use them to
help them forecast what will happen)?
In fact, modern macroeconomists have become quite dissatisfied
with this procedure that had been traditional, and they now limit their
attention to models containing individuals whose expectations are
consistent with the model itself. This is the rational expectations approach
(which is also called the perfect foresight approach, if stochastic
disturbances are not involved).
As far as the historical development of the subject is concerned,
we can identify four approaches to modeling expectations:

Static Expectations — individuals are always surprised by any changes,

and so they make systematic forecast errors. This early approach had the
advantage of simplicity (so that model solutions could be derived more
easily) but it had the disadvantage that we assumed irrational agents. After
all, when errors are systematic, it should be straightforward to do better.
36
Adaptive Expectations — individuals forecast each endogenous variable by
assuming that the future value will be a weighted average of past values
for that variable — with the weights summing to unity and getting smaller
as ever earlier time periods are considered. This hypothesis was
"invented" by Cagan in 1956, and it was made popular by Friedman since
it played a central role in his permanent-income theory of consumption.
This approach had the advantage that it did not complicate solution
procedures too much, and it involved long-run consistency (in the sense
that forecasters eventually "get it right"). For example, consider a
situation in which inflation doubles. Under adaptive expectations,
expected inflation eventually doubles. However, all during the (infinite)
time it takes for forecasters to reach this final adjustment, they under-
predict inflation. In other words, this hypothesis still involves systematic
forecast errors, so it does not involve short-run consistency.

Perfect Foresight — individuals are assumed to be so adept at revising their

forecasts in the light of new information concerning the economy's
evolution that they never make any forecast errors. Economists (and other
scientists) often find it useful to "bracket the truth" by considering polar
cases. For example, economists focus on the extremes of perfect
competition and pure monopoly — knowing that most real-world firms find
themselves in industries that are in between these two extremes.
Nevertheless, since formal analysis of these intermediate cases is difficult,
we learn a lot by understanding the polar cases. The hypothesis of perfect
foresight has some appeal, since it is the opposite polar case to static
expectations. This hypothesis was involved in our treatment of the first
Neoclassical Synthesis analysis in Chapter 2.
Many economists find it appealing to adopt an hypothesis that is
between the polar-case extremes of perfect foresight and static
expectations, and in a sense, that is what adaptive expectations is. After all,
perfect foresight coincides with adaptive. expectations itlim very recent
past gets all the weightinthe weighted average, and static expectations
coincides with adaptiveexpectations if theinfinitely_distant past gets all
the weight. But the adaptive expectation approach — with some weight
`

given TOThir time horizons — is still an unappealing way to achieve a

compromise since it involves systematic forecast error addition, there
is no reason for agents to limit their attention to just past values of the
variable that they are trying to forecast. Why wouldn't they use their
knowledge of macroeconomics when forecasting? For example, suppose
individuals were told that the central bank will double the money supply
next year. Most people would raise their expectations of inflation when
presented with this policy announcement. But people following the
37
adaptive expectations approach would make no such change in their
forecast, because the announcement concerning the future cannot change
pre-existing outcomes (and the latter are the only items that appear in a
backward-looking forecasting rule).

Rational Expectations — Like adaptive expectations, this hypothesis is an

attempt to have an analysis that is between perfect foresight and static
expectations. It is used in models that involve the economy being hit with
a series of stochastic shocks, so agents cannot know everything. But the
agents in the model understand the probability distributions that are
generating the shocks, so they can form expectations in a purposeful
manner. Under the rational expectations hypothesis, each agent's
subjective expectation for the value of each endogenous variable is the
same as what we can calculate as the mathematical expectation for that
variable — as we formally solve the model. Thus, agents make forecast
errors (so this hypothesis is more "realistic" than perfect foresight) but
those errors are not systematic. Perhaps a better name for this hypothesis
is model-consistent expectations. In any event, most modern macro-
economists find rational expectations a very appealing approach.

As noted above, the perfect foresight and rational expectations

hypotheses coincide if the variance of the stochastic shocks shrinks to
zero. Since the policy implications of two models that differ only in this
dimension are often the same, macroeconomists often rely on perfect
foresight analyses — despite the fact that this approach "sounds" more
"unrealistic."
We proceed through a series of analyses in this chapter. First, we
introduce a stochastic disturbance term in an otherwise familiar macro
model, by reconsidering the basic model of Chapter 2 when it is defined in
a discrete-time specification. It turns out that — since stochastic differential
--
equations are very difficult analyze and \ since stochastic difference
equations are much simpler — macroeconomists switch to discrete-time
specificationS:Whed thEY—Wlift to s-1-6-§-1 incomplete information. We
consider three aspeCis of uncertainty: situations iff -WhiCE6-66r1O—nifc agents
have incomplete information concerning exogenous variables, situations
in which agents (and the stabilization policy authorities) have incomplete
knowledge concerning the model's slope parameters, and situations in
which the functional fortis of iniportaiifiria-a-5eZm -ioinic relationships are
not known with certainty. Finally, in later sections of the chapter, we
move on to adaptive and rational expectations analyses.

38
3.2 Uncertainty in Traditional Macroeconomics

In the last chapter, we focused on the economy's response to a single one-

time shock, and the resulting time path — a one-time departure of output
from its natural rate, and then a monotonic return to full equilibrium. Our
only analysis of ongoing business cycles, involved specifying that one of
the exogenous variables followed a sine curve (in section 2.5). A more
common method of analyzing ongoing cycles involves specifying that at
least one exogenous variable does not follow a deterministic time path;
instead it is a stochastic variable. We begin this chapter by reconsidering
the static-expectations model of Chapter 2 — this time specified in discrete
time, with a central bank that cannot perfectly control the size of the
nation's money supply at each point in time.
Ignoring the autonomous spending variable and assuming a zero
core inflation rate for simplicity, the model can be defined by the
following equations:

.Y, = 0(mi — A) aggregate demand

(m, — m1_1)= — X(Y,--1 - 5)+u, monetary policy
19, 13,-1=0(Y,_1 - 57)= Phillips curve

The "hat" denotes the output gap. The monetary policy reaction function
involves the money growth rate adjusting to the most recent observation
on the output gap (if x is positive, not zero). Otherwise the money growth
rate is a random variable, since we assume that u is a standard "error"
term, with expected value of zero, no serial correlation, and a constant
variance. Among other things, the model can be used to assess whether
"leaning against the wind" is a good policy (as Keynesians have always
recommended), or whether (as Milton Friedman has long advocated) a
"hands of' policy is better. We can evaluate this "rules vs. discretionary
policy" debate by analyzing whether a zero value for parameter x leads to
a smaller value for output variance (the variable the central bank is
assumed to care about (given the policy reaction function)) or whether a
positive value for parameter x delivers the smaller variance. But we
postpone this policy analysis for the moment, by imposing x = 0. By
taking the first difference of the demand function, and substituting in both
the policy reaction function and the Phillips curve, we have:

A= + Ou, (3.1)

39
where v = (1— 00). Aside from the ongoing error term, there are (in
general) four possible time paths that can follow from a first-order
difference equation of this sort (as shown in Figure 3.1). We observe
explosion if u > 1, asymptotic approach to full equilibrium if 0 < u < 1,
damped oscillations if —1 < u < 0, and explosive cycles if u < -1. The
standard way of using the model to "explain" business cycles is to assume
that 0 < u < 1, and that the stochastic disturbance term keeps the economy
from ever settling down to full equilibrium. With these assumptions, the
model predicts ongoing cycles.
It is useful to compare the stability conditions for continuous-time
and discrete-time macro models. In the former, we saw that overshoots
were not possible, so the stability condition is just a qualitative restriction
(that the sign of parameter s be appropriate). But in discrete time,
overshoots of the full equilibrium are possible, so the stability condition
involves a quantitative restriction on the model's parameters (that the
absolute value of u be less than unity). To maximize the generality of
their analyses, macro theorists prefer to restrict their assumptions to
qualitative, not quantitative, presumptions. This fact clarifies why much of
modern macro theory is specified in continuous time. But, as noted above,
if stochastic considerations are to form the focus of the analysis, we must
put up with the more restrictive stability conditions of discrete-time
analysis, since stochastic differential equations are beyond the technical
abilities of many analysts.
Now let us re-introduce ongoing policy, by considering the
possibility of x > 0. In this case, v = (1 0(0 +,')). Is an increase in x a
—

"good" thing? The answer appears to depend on whether u is positive or

negative before the interventionist policy is introduced. If it is positive,
interventionist policy makes parameter u smaller, and this is desirable
since it either eliminates explosive behaviour, or speeds up the asymptotic
approach path to full equilibrium (if u were initially a positive fraction).
But making x positive could make u become negative, and so the well-
intentioned stabilization policy could create cycles in economic activity
that did not exist without policy. In this case, interventionist policy is not
recommended. Further, if u were already a negative fraction, policy could
make u go beyond the minus one value, so that policy would have caused
damped cycles to be replaced by explosive cycles. These possibilities
support Friedman's concern that even well-intentioned policy can be quite
undesirable.

40
 Figure 3.1 Possible Time Paths for Output

Yr
1. Direct Instability (v> 1)
2. Direct Convergence
(0<v<1)
3. Damped Cycles
(-1 <v<O)
/• /` .‘ 3 4. Explosive Cycles—not
New Y _ ••__/_ I shown (v<-1)
/ -

I , / 2
Initial Y

Time, t

Another way of summarizing this possibility is by focusing on the

asymptotic variance of the output gap. To derive this expression, we take
the expectations operator, E„ through equation 3.1, using the assumption
that – entering each period – the expected value of each error term is zero.
The result is:

E(Y,)= vE(Y„).

When this relationship is subtracted from equation (3.1), and the outcome
is squared, we have

– E(Y,)J 2 = + 6642 +20ou – E(Y,_,)]. (3.2)

The variance of Y is now calculated by taking the expectations operator

hrough equation (3.2). The result is

62 y = [02 1(1 .- 1) 2 )]Cr2 1} . (3.3)

t is important to clarify the information set upon which this expectation is

)ased. We discuss two extreme assumptions. The first is that expectations
tre based on (t-1) period information. If so, E1_, (Y,) = Y,_, in equation (3.2)
tnd so 6 2 }, = o-2u . But this assumption is not appealing if we wish to

41
consider the effects on the economy of a whole series of stochastic shocks
buffeting the system through time (at any moment, the shocks from many
periods continue to have some effect). To capture this ongoing uncertainty
in the variance calculation, we need to assume that expectations for period
t are based on information from period (t j), where j is much larger than
one. The convention is to calculate the asymptotic variance by letting j
approach infinity. In this case, both [Y, — E(Y,)] 2 and [Y,_, — E(y_ 1 )]2 equal
62 y so equation (3.2) leads to (3.3).
,

Another way of explaining the calculation of asymptotic variancc

helps us appreciate this result, and why it is important. The asymptotic
variance captures the ongoing effect on output of an entire series of short-
run disturbances. It can be calculated by evaluating:

Epc—E(Y i+Of .

An expression for Y,„ can be had by recursive substitution:

171+1 etif+1
1-)Y ,

Y„, = v(vY„, + Ou,)+ Bu,,

Y,, = u2 y„ + 0(u„, vu,)

Following this pattern, we have:

Y„ = un+' Y + Oz

where

Z = U, + , UU„n „ + U ntl,.

As long as u is a fraction and n approaches infinity, we have E(Y,, n )= 0,

and

cr 2 y = 02 (1 u2 04 = 02472,,
— U 2 ).

Since an increase in the degree of policy intervention can magnify the

coefficient that connects the variance of national output to the variance of
the temporary shock process, it can reduce the degree to which the
economy is insulated from a series of ongoing disturbances. We conclude
that when stochastic considerations are combined with time lags, it is
42
 difficult to define what institutional arrangements provide the economy
with a "built-in stabilizer."
We can summarize as follows. The relevant coefficient for
evaluating built-in stability in the face of a series of temporary shocks is
1/(1— 0 2 ) , not the deterministic full-equilibrium multiplier that is familiar
from intermediate macroeconomics, 1/(1— v). This latter expression is
relevant only for summarizing the effects of a one-time permanent shock.
The coefficient that relates the variances is the only one that refers to the
extent to which random shocks cumulate through the system. In the case
of our demonstration model, we see that a more Keynesian policy (a less
hands-off Friedman-like monetary policy) must lower 1/(1— v), whether u
is positive or negative, but such a policy may raise or lower the more
relevant reduced-form coefficient 1/(1— v 2 ) . A more Keynesian policy
decreases this coefficient if u is positive, but increases it if u is negative.
This result threatens the standard idea that an interventionist central bank
can do better than one that follows Friedman's hands-off dictum. This
conclusion is interesting in itself because the question of stabilization
policy is important. But there is also a more general message. Simple
intuition can be misleading when both time lags and the consideration of
uncertainty are involved. There appears to be no substitute for an explicit
treatment of these factors.
We now consider uncertainty in two different ways. First, we
refer to Brainard's (1967) more general analysis which allows both the
intercept and slope parameters to be uncertain. Second, we examine the
implications of the policy maker being uncertain about the functional form
of a structural equation.
Brainard was the first to analyze a stochastic macro model in
which the policy maker did not know the structural parameters with
certainty. Brainard considered a stabilization policy authority that tries to
minimize the expected deviation of output from its target value,
E(Y — Y) 2 , subject to the fact that output, Y, is related to the policy
instrument G by a simple macro model: Y = aG + u. Both a and u are
stochastic variables with expected values equal to a and 0, with constant
variances equal to a-2, and cr 2i, , and with no serial or cross correlation.
After substitution of the constraints into the objective function, we
have:
E(a + e)G +u — f) 2
zi2G2 G20. 26. +0_
2a +Y 2

43
Differentiating this objective function with respect to the decision variable
G, we have the optimizing rule, and the optimal value for the policy
variable that emerges is:

- 7 1(.72 +a2 e ).
G* = 0

It is instructive to consider the limiting cases. If the policy multiplier

coefficient is known with certainty (3 -2, = 0), then the optimal value for
the policy variable is the target value for output divided by the multiplier.
But as uncertainty about the multiplier rises (that is — as the variance of E
approaches infinity) G should be set at zero. For intermediate values of
the variance, it is best to set the policy instrument at a value that is
somewhere between the "certainty equivalent" optimum and the "do
nothing" value.
This formal analysis of uncertainty supports Friedman's long-
standing argument that policy makers should attempt less when they are
unsure of the connection between policy and the economy's performance.
In this sense, then, the analysis supports "rules" over "discretion". There
have been a number of extensions of this analysis in recent years. For
example, Basu et al (1990) apply this method to a situation in which a
policy advisor is unsure about whether her advice will be followed. Others
(Brock, Durlauf and West (2003), Majundar and Mukand (2004), Favero
and Milani (2005) and Wieland (2006)) explore why it may be optimal for
the authority to depart temporarily from what would otherwise be the
optimal stabilization policy, just so it can learn more about the structure of
the system, and so be less burdened with uncertainty in the future.
There is an alternative to modelling uncertainty in a stochastic
framework. Uncertainty can take the form of the policy maker not
knowing the curvature involved in a particular structural relationship. To
establish the implications of this form of uncertainty, we now compare
two models that are exactly the same except for the fact that the precise
degree of curvature in one relationship, the Phillips curve, is not known.
The first model is defined by the following three equations:

Y, = 0(mi — 13,) aggregate demand

m, = 171— 2'(P, monetary policy
Pt P,_, = 0(Y,_1 — .1)) Phillips curve

This system is quite similar to the model we have discussed thus far in this
chapter. Indeed, the aggregate-demand and Phillips-curve relationships

44
are exactly the same. The policy reaction function is more up-to-date in
the present specification, with the central bank adjusting its monetary
aggregate with a view to hitting a price-level, not an output, target. For
simplicity, we assume that the value of the level of all variables with bars
above them is unity (so the associated logarithms (the lower-case
variables appearing in this system) are zero). The solution equation for the
output gap then is

where v=1-00(1+ 2). As was the case with our earlier analysis, this
model reduces to a first-order linear difference equation. Convergence to
full equilibrium occurs if u is a fraction. The model cannot predict an
ongoing cycle of constant amplitude. Is a policy of price-level targeting
recommended? Not necessarily: an increase in parameter x can make u
exceed unity, so more aggressive price-level targeting can be destabilizing.
But this policy can never change the fact that output gravitates to its
natural rate value (unity) as long as instability is avoided. We now show
that this property is lost if the curvature involved in the Phillips curve is
altered just slightly. Indeed, the classical dichotomy is lost in this case.
These assertions can be defended by replacing the Phillips curve
in this simple model with the following:

This one change leads to a reduced form for output that is slightly
different:

— —00(1 + 2')(17,_, —1).

Since the left-hand side of this equation can be approximated by

((Yi — Y,_,), and since we can define X, = (G6 —1)1 f3g, where
13=1+ 00(1+ 2,), this reduced form can be re-expressed as

fiX,(1— X,). (3.4)

This reduced form is a first-order non-linear difference equation; it is the

simple logistic function involved in basic chaos theory. One purpose of
this section of the chapter is to relate standard stabilization policy analysis
to the chaos literature in a simple and clear fashion. We will see that the

45
average rate of output observed in the long run (a real variable) is affected
by the policy reaction coefficient (that is, by monetary policy).
Despite its simplicity, equation (3.4) is consistent with many
different time paths for output, as is summarized in Table 3.1. Two of the
possible dynamic patterns are shown in Figure 3.2. The general shape of
equation (3.4) that is shown in both parts of the figure can be verified by
noting that both the values 0 and 1 for current X imply that next period's X
is zero, and that the slope of the relationship decreases as current X rises
(and is zero when current Xis 0.5).
The dashed lines in Figure 3.2 indicate the time sequence.
Convergence to the natural rate (which involves X equal to 0.5 and Y
equal to 1) occurs with the "low" value for 13, but a limit cycle (one that
maintains a constant amplitude forever) occurs with the "high" value for 0.
In the right-hand panel of Figure 3.2, "full equilibrium" involves the
economy shifting back and forth between points A and D (every other
period) forever. The average level of economic activity is represented by
point C. Since point D corresponds to the natural rate of output, point C
involves an average activity level for the economy that is below the
natural rate. Thus, with a non-linear reduced form relationship, it is
possible that output remains below the natural rate (on average) even in
the long run.

Table 3.1

Range for 13 Nature of Time Path

X must approach 0
113<3 X must approach (/-1)/fl so Y approaches
one
3f3<3.449 X follows a 2-period limit cycle
3.449..11<3.549 X follows a 4-period limit cycle
3.549.13<3.57 X follows even-numbered-period limit
cycles of ever greater number (8, 16, 32, ...)
Xis subject to chaotic fluctuations
13>4.0 X must approach negative infinity

46
 Figure 3.2 Convergence and Cycles in a Nonlinear Model

Xt+I Xt4-1

This analysis is a deterministic variant of Lipsey's (1960) non-

linear aggregation hypothesis. The interesting features of this variant are
that stochastic elements are not required to generate a cycle that maintains
its amplitude as time passes, and that it is variations in monetary policy
that can make these permanent output effects important.
In this model, a more aggressive attempt to target the price level
:a larger value for parameter x) can lower the average rate of output.
Since uncertainty concerning the appropriate functional form for one of
he model's structural equations is what leads to the possibility that
something as basic as the classical dichotomy can break down (even in a
iatural-rate setting), it must be admitted that the standard practice of
inear approximations in macroeconomics is a definite limitation.
One final point worth explaining is how Figure 3.2 is altered
vhen the macroeconomy is chaotic. Chaotic cycles occur with larger
'alues for f3 than are assumed for Figure 3.2, since a higher I3 makes the
•duced-form relationship pictured in Figure 3.2 more non-linear. (The
tnd points are unaltered, but the height of the hump is increased.) Chaotic
:ycles involve an "observation box" like that formed by points A and B in
'igure 3.2, but with the location and size of the box moving around the
ilane forever, with no regular pattern ever emerging.

47
 There are three reasons for integrating basic macro analysis and
chaos theory. First, we can appreciate that ongoing changes in exogenous
variables do not need to be assumed to explain business cycles. Second.
we can see that intuitively appealing policies can create cycles (both
regular ones and chaotic ones) — so once again — built-in "stabilizers" can
be destabilizing. Finally, we can appreciate that the entire nature of the
full equlibrium for real variables — whether it involves asymptotic
approach to the natural rate or a limit cycle which averages out to an
output level that is less than the natural rate (to take just two of the
possibilities) — can depend on the short-run targeting strategy for
implementing monetary policy. Keynesians have become particularly
excited about this implication of non-linearity, since it implies that they
need not follow what has been the convention — conceding the full-
equilibrium properties of macro models to the Classicals, and limiting the
Keynesian contribution to a discussion of approach paths. With non-
linearities, the short and long-run properties of models cannot be so
readily separated.
Two useful surveys of work on non-linear dynamics are
Mullineux and Peng (1993) and Rosser (1990). It has proved difficult for
econometricians to distinguish between the traditional view of business
cycles (linear systems with stochastic shocks maintaining the amplitude of
cycles that would otherwise dampen out) and this alternative view
(endogenous deterministic cycles in a non-linear system that never die
down). Despite this inability to reject the relevance of the non-linear
approach, it has not become too popular. Mainstream macroeconomists
have limited their attention to linear stochastic systems, as we do for the
remainder of the chapter.

3.3 Adaptive Expectations

Real world business cycles are not the two-period saw-tooth cycles that
we see in Figure 3.1; they are more irregular. But if mainstream analysis
does not rely on exogenous variables following sine-curve time paths, and
it simplifies by ignoring non-linearities, how does the mainstream
approach generate realistic-looking cycles in the model's endogenous
variables? The answer is by extending the linear model so that it is
consistent with cycles that appear more like a sine curve. This can be done
if the model is complicated to the point that the reduced-form for output
deviations is a second-order (or higher) linear difference equation. One
way of doing this is by allowing for adaptive expectations. Other ways are

48
discussed in Chapter 5, where we explore New Classical macro-
economics.
Ignoring all policy variables and error terms for simplicity, the
revised demand and supply functions are:

Y,= -0A+ WTI

Pt — Pt -1 = 0(Y1-1 — Y)+

where it stands for expected inflation. As explained in Chapter 2, the

revised demand function comes from distinguishing the nominal from the
real interest rate. The revised Phillips-curve involves the assumption that
the "core" inflation rate is the adaptively-formed expected inflation rate.
The adaptive-expectations hypothesis is defined by the following equation:

gt-i= 2 (A — A-1 — 7t _1).

This hypothesis is often called the error-learning model, since it specifies

that the change in the agents' forecast is equal to a fraction of the previous
forecast error (as long as parameter A, is assumed to be a positive fraction).
An alternative interpretation of this hypothesis can be seen if it is
rewritten for several time periods:

= (1 — 2Ap„
1-1 = (1 — 2)7r t-2

so that successive substitution leads to

71. = 2{ AP + (l - 2)A13,_, + (1 — 2 )2 AA-2+ ...]

This last formulation of the hypothesis states that It is a weighted average

of past actual values, with less weight given to the more distant past. The
weights decline geometrically. Both the error-adjustment interpretation
and the weighted-average interpretation suggest a certain plausibility for
the adaptive expectations hypothesis, as does its long-run consistency
property (discussed in the Introduction to this chapter). Thus, it is not
surprising that this hypothesis was the mainstream way of modelling
expectations in macroeconomics from the mid 1950s to the mid 1970s.
The three equations of this simple model can be combined to
yield a second-order difference equation (y, as a function of both j),_ i and
In particular, we proceed through the following steps. We begin by
49
first-differencing the demand function, and using the Phillips curve and
the adaptive expectations equations to eliminate the change in the price
level and the change in expectations (respectively). The result is

Yr = (I+ 0(v2 - O))Y' -1 - e7rf-i.

This relationship itself is first-differenced, and the resulting

term is eliminated by combining lagged versions of the Phillips-curve and
the expectations equations: (7r,_, - 205),_2 . The result is

yt =1)15),- 1 +1-)25.)!--2

where 0, = 2 + Ø(y'2- 0) and 02 = —(1+ q2(0 +0-00). As in the first-

order linear difference equation case, there are four possible time paths (as
is explained in all standard math-for-economists texts such as Chiang
(1984, p. 576)). The only difference is that, in this case, the cycles involve
damped and explosive sine curves, not two-period saw-tooth cycles. With
slope parameter values that lead to damped cycles, and exogenous
disturbances to keep those cycles from ever dying down, we have the
standard approach to business cycles.
It turns out that the condition for cycles to occur is that 0 1 2 < 402 ,
and the stability conditions are: 0 2 <1, 0 + 02 +1 > 0, and vl v2 < 1. It is
, —

left for the reader to verify that the bigger is the role that adaptive
expectations play in the system (that is, the bigger is parameter X), the
more likely it is for the system to display cycles, and for the system to
display instability. It is for this reason that Keynesian economists have
emphasized expectations. And, as we discovered in Chapter 2, similar
conclusions emerge when a model-consistent form of expectations
(perfect foresight) is assumed, instead of adaptive expectations.
While the adaptive-expectations hypothesis has some appealing
features, it has an unappealing one as well. It makes little sense to analyze
the effectiveness of any government policy that is intended to improve
agents' economic welfare if that analysis does not permit those agents to
understand both the environment they live in and the effect of the ongoing
policy intervention within that environment. Our modeling has allowed
for all these considerations only in the perfect-foresight case — not in the
adaptive-expectations case.
A less "unrealistic" version of model-consistent expectations
(compared to perfect foresight) is the "rational expectations" hypothesis.
According to this hypothesis, agents make forecast errors, but (since they
50
are aware of the probability distributions that generate the random shocks
that hit the macro economy), they make their forecasts of the endogenous
variables as if they were calculating the best forecast possible from the
formal model. Thus, forecast errors occur, but they are not systematic.
The remaining sections of the chapter explain how we can analyze models
of this sort, and what some of the policy implications are.

3.4 Rational Expectations: Basic Analysis

The basic idea behind rational expectations can be explained in an

extremely basic setting — the simple fixed-price fixed-interest-rate
income-expenditure model that students encounter in their introductory
economics course. That model can be defined by the following
relationships:
Vl C C-7
Y, = C,+G
cy
C, = cYe,

and one of

=
Ye, = Et _1

The first two equations define goods market clearing, and that private
demand is proportional to agents' expectations concerning what income
they will receive that period. The remaining equations define two
hypotheses concerning how those income expectations may be formed.
The former can be thought of as either static expectations (the forecast for
today is equal to what was actually observed yesterday), or a degenerate
version of adaptive expectations (with the entire weight in the weighted
average put on the most recent past). The second hypothesis is rational
expectations. According to this hypothesis, agents' subjective forecast is
the same as we can calculate by evaluating the mathematical expectation
of actual current income (as determined in the model).
In the static expectations case, the at-a-point-in-time solution
equation for current GDP is:

= cy_i + G.

51
Consider a once-for-all increase in autonomous spending, G. The solution
equation for Y indicates that GDP rises by the same amount as the
increase in G in that very period. Then, Y keeps rising by_smaller and
smaller amounts each period, with the overall increase in output (given
infinite time) being:
^ whys
Y = (1 /(1 c))G.

In the rational-expectations case, the same substitution of the second and

third equations into the first results in:

Yt =cE,_1 (Y,)+G.

This is not a solution equation for actual output, Y, since there are still two
endogenous variables (Y and its expectation)_in this one equation. We
need a second equation — one for the forecast of Y.. This can be had by
taking the expectations operator through this "almost reduced-form". The
result is

E,_1 (y)=G 1(1—c).

When this expression is substituted back into the almost-reduced-form, we

end with the full reduced-form (solution equation) for actual output:

Y, = (11(1— c))G.

By comparing the two solution equations (for static expectations

and for rational expectations), we can determine how the model's
properties are affected by embracing this model-consistent forecasts
hypothesis. _We see that the impact effect of a once-for-all increase in
autonomous spending_ is bigger in the rational expectations case. Indeed,
the impact effect on Y is now equal to the eventual effect. With forward-
looking agents, it makes sense that there not be a. gradual increase in
income. In this case, agents correctly see that this higher income is
coming. They raise their consumption immediately as a result, and this (in
turn), since output is demand-determined in this model, makes overall
income rise immediately. In short, what used to be called the model's
long-run properties become the model's short-run properties when agents
have rational expectations). Worded in this general fashion, this summary
applies to all rational-expectations models.

52
 We now explore the properties of a slightly more complicated
rational-expectations model. The system we now consider is similar to
that in the previous section in that it involves descriptive aggregate
demand and supply functions. But there are several extensions here. For
one thing, we allow for variable prices and interest rates. Also, we focus
on monetary (not fiscal) policy. Initially, we continue to suppress any
distinction between nominal and real interest rates, and we specify an
arbitrary monetary policy reaction function. The initial model is defined
by the following four equations.

y, = a — yr, + v, (3.5)
Pt — Pt_1=0.Y1+ Pe PI-1 ±Ut (3.6)
r, = r+ y(ioe, — 0) (3.7)
P = Et_1(13t) (3.8)

y and p stand for the natural logs of real output and the price level. Both
the natural rate of output and the central bank's target for the price level
are unity, so the logs of these variables are zero (and y = a — yr: = 0) . v
and u are stochastic shocks — drawn from distributions that involve zero
means, constant variances and no serial correlation. r is the interest rate,
not its logarithm.
Equation (3.5) is a standard (descriptive) IS relationship, which is
also the aggregate demand function since the central bank sets the interest
rate. Since we focus exclusively on monetary policy in this discussion,
fiscal variables are constant (and erril_ec
) ided in thejiiiircePt). Equation
;3.6) summarizes the supply side of the goods market. It is a standard
expectations-augmented Phillips curve. Equation (3.7) is the central
yank's reaction function. The bank raises (lowers) the interest rate above
:below) its long-run average value whenever the bank's forecast for the
jog of the) price level is above (below) target. There is no need to specify
m LM equation, since its only function is to determine what value the
yank had to adjust the money supply to (in order for the public's demand
or money function to be satisfied at this interest rate). Since the money
;upply enters none of the other equations of the model, we can afford to
gnore this consideration.
Equation (3.8) defines rational expectations; it specifies that the
igents' subjective expectation for price is equal to what we (as model
nanipulators) can calculate as the mathematical expectation of price. The
ime subscript for the expectations operator denotes that agents know all ,

ralues for the stochastic shocks up to and including the previous period

53
(time period t-1). Agents (and the central banker) must forecast the current
shock — at the end of the previous period before it is revealed — on the
basis of all past information. Agents know the exact structure of the
economy (the form of all equations and all slopeciietyclients). The only
thing they do not know is what the current and all futureNT—M-6)s of the
additive error terms are.
To solve the model, we first eliminate the interest rate and the
expected price variables by simple substitution. The results are:

Y, = — V7E,--i(P,)+vt (3.9)
P, OY, E,--1(P,)±u, (3.10)

Taking the expectations operator through these two equations, we realize

that the expected value for both p and y is zero. Substituting this value for
the expected price level back into (3.9) and (3.10), we end with very
simple reduced forms:

yl = v,
P, = OY, + 14,

These reduced forms lead to the following volatility expressions:

var(y) = o-
var(p) 0=262v +6z ,4

Since neither variance is a function of the policy maker's parameter, y,

monetary policy is irrelevant. This makes sense; after all, the central bank
must set its instrument variable (the interest rate) before the current
shocks are known — just as the private agents must commit to setting their
nominal variables — before the current shocks are known. There is nothing
that the central bank can do for the private agents that they cannot do for
themselves.
We can remove this "policy irrelevance" result by changing the
model in at least two ways. For one thing, we could have some of the
private agents constrained to set their nominal variables more than one
period in advance. For another, we could let the central bank wait until the
current shocks are known before the interest rate is set. In either case, the
central bank would be able to do more than can agents on their own, so
monetary policy should matter. We choose the second alternative here,
and consider multi-period private sector nominal contracts in later

54
chapters. We verify that policy irrelevance is a very model-specific result
by changing equation (3.7) to the following

ri = 7 + 21(P, 0) (3.7a)

and re-deriving the output and price variances. It is left for the reader to
duplicate the earlier steps, and to verify that the revised solutions are:

y, =Ogyu,+v,)1(1
19, =0Y, + u,

These results imply that a more aggressive price-level targeting policy (a

bigger value for parameter y) is desirable since it makes demand shocks
have a smaller effect on real output, but it is undesirable since it makes
supply shocks have a larger effect on real output. So the monetary
authority faces a permanent volatility trade-off, even though it does not
face a permanent trade-off between the average level of real output and
inflation (or the price level).
These results make sense at the intuitive level. The natural-rate
feature (the Classical dichotomy) is a built-in feature of the Phillips-curve
specification, so it is not surprising that there is no lasting trade-off
between inflation and the level of real activity. But this does not mean that
central bankers should ignore the effect of their policies on real variables.
Even within the class of policies that deliver long-run price stability
within a natural-rate model, there is a basis for preferring one specific
monetary-policy rule to another, since there are lasting differences in the
volatility of real activity about its natural-rate value.
When a negative demand shock hits, the aggregate demand curve
shifts left. If the short-run aggregate supply curve is positively sloped,
price falls. An aggressive price-targeting central bank reacts vigorously to
reverse this effect — by shifting aggregate demand back to the right. Since
the exogenous shock and the policy variable both affect the same curve
(demand, not supply), the central bank cannot help but limit output
volatility as it pursues price stability. But things are different when the
shock is on the supply side. If the aggregate supply curve shifts left, there
is pressure for the price level to rise, and for real activity to fall. To limit
the rise in price, the central bank must shift the demand curve to the left.
The problem is that this policy accentuates the output fall, and so there is
a volatility trade-off.
The fact that we can appreciate these outcomes so easily at the
intuitive level makes it seem as though there is no need to work things out

55
formally. But such a conclusion is not warranted, as we will see in the
next section — where we consider a more complicated model. In such
settings, it is almost impossible to sort things out without the precision
that accompanies a formal solution.

3.5 Rational Expectations: Extended Analysis

We now consider a more complicated model — one that involves both

yesterday's expectations of today's outcomes and today's expectations of
tomorrow's outcomes. A more involved solution procedure — known as
the undetermined coefficient solution method — is required in such a
setting. The following model is fully consistent with the micro-
foundations that modern macroeconomists insist upon. (Readers will be
able to verify this when they read Chapter 4. For the remainder of this
chapter, however, we simply take this model as "to be defended later.")
The model involves a proper distinction between nominal and real interest
rates, IS and Phillips curve equations that are based on inter-temporal
optimization, and a central banker who optimizes. To simplify slightly, we
set the supply shock to zero in all time periods. The revised IS, supply,
and policy-reaction functions are:

Y, = Ei(Y,+1) — v[i, -(E,(P,+1) - P,)]+ v, (3.11)

P1 — P,--1= OY, 13,) (3.12)

In the first instance, we assume price-level targeting on the part of the

central bank. Thus, the bank sets the nominal interest rate (i) each period
(on the basis of (t-1) information) so that E,_1 (p,)= 0. The implications
can be seen by taking the expectations operator through the IS equation,
solving the result for the interest rate, and imposing the central bank's
goal (both in the resulting interest-rate-setting equation and throughout the
model). This involves our simulating how the central bank determines
what value of the interest will be needed to deliver a value for GDP that
will ensure (at least expectationally) that its target for the price level will
be met. As above, there is no need to specify an LM equation, since its
only function is to determine the money supply (residually).
We now perform the steps that were outlined in the last paragraph.
so that we can explicitly derive the interest-rate-setting equation. First, we
take the E,_1 (.) operator through equation (3.11), second, we set
E,_1 (p,)= E1_1 (vg ) = 0, and finally (since the central bank sets the current

56
interest rate equal to the value that it expects will deliver its goal) we set
E,_,(i,). The result is:

4 = -(1 / V)E,_1(Y, — Yi+1).

When this relationship is substituted back into the IS function, and the
Phillips curve is simplified slightly, we have the model in the following
revised form:

Y, E,--1(Y,+i)]+ E,-1(y,) P, v, (3.1 la)

Pt = (1 / 2 )A 1 + (0/ 2 ).Y,
-
(3.12a)

We now proceed with solving the model, using trial solutions and the
undetermined coefficient method. Since there is only one pre-determined
variable in the model (the previous period's price) and one exogenous
variable (the demand shock), there can only be these two terms in the
solution equations for real output and the price level. Thus, we assume the
following trial solutions:

y, = ap,_, + bv, (3.13)

p,=cp,_,+dv, (3.14)

We use the undetermined coefficient solution procedure to determine how

the reduced-form coefficients (a, b, c and d) depend on the underlying
structural parameters that have economic meaning (y and 4)).

We now substitute the trial solutions into (3.11a) and (3.12a). First, after
substituting (113) into (3.12a), we have

p, = ((1 + a0)12)p,_,+(b012)v, (3.15)

Equations (3.16) and (3.15) are the same if and only if

c = ((1 + a0)12)
and
d =(b012).

We need two more identifying restrictions. These are obtained by using

the trial solutions to get expressions for all terms on the right-hand side of

57
(3.11 a) except the error term. Once these are substituted in and the result
is compared to (3.13), the remaining two identifying restrictions emerge:

a=a—vc
b= ad — yid +1.

The four identifying restrictions are now solved explicitly for a, b, c and d:
so the reduced forms contain only primitive parameters:

a = —11 0
b = 2 /(3 + v0)
c=0
d = 1(3 +

The expressions for price and output volatility (the asymptotic variance
for each variable) follow directly from the trial solutions:

0.2 p = {d2 /0 Acr 2 v (3.16)

= „„2 1,2 0_2
y v p it/ (3.17)

Illustrative values for these volatility expressions can be had if

representative parameter values are inserted. As we shall see in Chapter 4,
the most straightforward defense of the IS relationship in this model
involves assuming that households have a utility function equal to the
logarithm of consumption. If this is assumed, our parameter xi' must be
unity, so we insert this value here. Walsh (2003a, p. 576) has used 0.2 as
an estimate of the slope of the short-run (annual) Phillips curve, so for our
illustrative calculations here, we follow his lead and take •:1) = 0.2. With
these representative values,

o-2 = 0.562„ and 62 = 0.004o-2v . (3.18)

We now compare these outcomes to what occurs if the central bank

pursues a different target. Suppose the bank's goal is to set the interest
rate so that the expected value for nominal GDP is right on target (equals
zero). In this case, all the steps in the solution procedure are repeated, but
with E,_, (p,) = —E,_1 (y,) , not E'1_, (p,) = 0, in this case. The derivations
are messier with this monetary policy, so we simplify by imposing w = 1

58
from the outset. It is left for the reader to verify that the revised
expressions for the identifying restrictions are:

a =—c
b =1—d
d=01(2 — c+0)
c=[(0+2)±V(0+2)2 —4]/2

This last restriction implies that there are two rational-expectations

solutions. The standard practice in this case involves the presumption that
agents choose the solution that involves stability (finite values for the
asymptotic variances). This involves the presumption that the absolute
value of coefficient c be less than one, which (in turn) implies that we
restrict our attention to the negative square root possibility in the final
identifying restriction. As noted in Chapter 2, this practice — of rejecting
unstable solutions as economically inadmissible — is often "justified" on
the basis of the correspondence principle. If the policy that is being
examined has never been tried, however, it is not clear how the fact that
instability has not been observed in the past (when other policies were in
place) can be relevant.
An alternative line of argument involves justifying the
presumption of stability by arguing that only this assumption is consistent
with the first-order conditions of the underlying optimizations being
satisfied. But since this solution ambiguity exists at the overall market
level — not at the level of an individual decision-maker — it is not clear
what leads decentralized individuals to focus on only the stable solution.
In any event, in some cases, both solutions involve stability. Taylor (1977)
ias suggested that the solution that involves the minimum variance be
;elected, and the other rejected. However, since some models involve a
volatility trade-off (as we discovered in the last section of this chapter,
Ind as was initially emphasized in Taylor (1979a)) it is not clear which
endogenous variable the minimum variance criterion should be applied to.
AcCallum (1983) has suggested that analysts should avoid this
lonuniqueness problem by including in the trial solutions only the
ninimal set of state variables that is warranted by considering the
structure of the basic model. The idea is that only "fundamentals" should
natter. However, when selecting our trial solutions ((3.13) and (3.14)) we
bllowed this advice, and we have still obtained two solutions. So, all
hings considered, it must be admitted that There has been no fully
.atisfactory solution of the nonuniqueness problem in rational
:xpectations models.

59
Chapter 11

New Growth Theory

11.1 Introduction

The Solow growth model (and all the extensions we considered in Chapter
10, except the last one involving non-renewable resources) has the
property that savings/investment policies cannot permanently affect the
economic growth rate. This property of the model stems from the fact that
the man-made item that raises labour's productivity — physical capital
accumulation — involves diminishing returns. Given our standard
assumption of diminishing returns, as more capital is accumulated, its
marginal product must fall. This makes it ever less tempting for house-
holds to acquire more capital. As this reaction sets in over time, the
temporary rise in productivity growth that initially accompanies a rise in
saving must wither away. The only way we can have a model that avoids
this prediction is by changing our assumption about the production
process. We need to build a model in which there is a man-made factor of
production that does not involve diminishing returns. The obvious
candidate is "knowledge". There seems to be no compelling reason why
we need assume that the more knowledge we have, the less valuable more
knowledge becomes. Thus, to have a theory of endogenous productivity
growth, economists have built models involving constant returns to scale
in the production process. We consider three of these models in this
chapter.

11.2 A One-Sector Endogenous Growth Model

The most straightforward way of eliminating decreasing returns is to

assume a production function that is a straight line ray emerging from the
input-output origin: Y= AK. For this reason, this approach is known as the
"AK Model". The problem with assuming such a production function at
the level of the individual firm is that — with constant returns — the size of
the individual firm is indeterminate, so formal micro-foundations cannot
be provided. Given our desire to keep the analysis free from the Lucas
critique, this is unacceptable. This problem is usually solved by assuming
that there is an externality. Each individual firm is assumed to have a
"normal" production function:

236
 = Kte

where the i subscripts demote individual-firm items and B is the worker-

effectiveness index. Worker effectiveness is assumed to be proportional to
the "state of knowledge" in the economy, and — to avoid the complications
of specifying a separate "knowledge-producing" sector — it is assumed
that the amount of knowledge possessed by the workers in each individual
firm is simply proportional to the overall level of economic development
(as measured by the aggregate physical capital stock that has been
accumulated):

B. = aK.

We can derive an aggregate version of the first equation by substituting in

the second equation and defining Y = qY „ and similar equations for the
two inputs (where q is the number of firms). Ignoring population growth
(assuming N is constant), the result is Y= AK, where A =
At the aggregate level, the full model consists of the following
relationships:

Y= AK
wN = (1 — a)Y
(r + 8)K = aY
e I C = r(1— r)— p
k =Y -C -G- SK
G = rrK +twN

The second and third equations follow from profit maximization; firms
hire each factor up to the point that the marginal product equals the rental
cost. Combining the first and third equations, we have: a A = (r + (5) .
Since three variables in this relationship are technology coefficients, this
equation fixes the pre-tax rate of interest. This simplifies the interpretation
of the remaining three relationships.
The fourth equation is the standard micro-based consumption
function (where there is an interest-income tax, z). For simplicity we have
assumed infinitely lived family dynasties (for whom the probability of
death is zero). The fifth equation is the GDP identity — indicating that the
capital stock grows whenever more output is produced, than is consumed.
The final equation defines a balanced government budget. Spending is

237
financed by the revenue that is collected from the interest-income tax (rate
r) and the wage-income tax (rate t).
When a balanced growth path is reached, capital, output and
consumption all grow at the same rate, n. This growth rate in living
standards is also the labour productivity growth rate; otherwise "effective"
labour would not be growing at the same rate as all other aggregates. We
re-write the final three equations to focus on this equilibrium growth rate.
In doing so, we divide the fifth equation through by K, define x and g as
C/K and GIY respectively, and we use the second and third equations to
simplify. The results are:

n = r (1 — r) — p
n= A(1— g)— x —
g = ra — (r8 I A) + t(1— a)

We assume that the government sets g and r exogenously, so the model

determines the wage tax. The system is recursive. The first equation
determines the growth rate. Then, given n, the second equation determines
x. Finally, the third equation determines t.
Since one of the policy variables enters the growth-rate
determining equation, policy can have permanent effects on the growth
rate in this model. To illustrate, we can take dnI dr =—r from the first
equation. A pro-savings initiative (a cut in r) raises the growth rate. The
size of this response is particularly noteworthy. With a 5% interest rate
and a 10 percentage point cut in the tax rate (r = .05, dr = — 0.1), we
conclude that the annual productivity growth rate can be increased by one
half a percentage point.
This is a very big effect. This interpretation can be defended if we
calculate the present value of the benefit of living in a setting that involves
a growth rate that is one half of one percentage point higher. The first step
is to evaluate how long it takes for the economy to reach its new higher-
growth-rate steady state. It turns out that it takes absolutely no time at all
to reach this outcome. For simplicity, we abstract from the government as
we consider the transitional dynamics. In this simplified case, we can use
the time derivative of the definition of x = CIK,iIx= eic_ K IK . The
model can be summarized by just two equations:

C/C=r—p
ilx=r+x-F45—p— A.

238
The total differential of this system is:

di= .7dx

where I/ and .7 are the steady-state values of the growth rate and the
consumption-capital ratio. Since both these entities are positive, both
these first-order differential equations involve an unstable dynamic
process. If we drew a phase diagram in C-x space, all arrows showing the
tendency for motion would point to outright instability. There is not even
a saddle path to jump on to. But since both variables are "jump" variables,
the economy is capable of jumping immediately to the one and only stable
point — the full equilibrium. In keeping with the convention that we limit
our attention to feasible stable outcomes, we assume that the economy
jumps instantly to its new steady state — the moment the exogenous shock
occurs to make the steady-state growth rate higher. In short, there are no
transitional dynamics in this growth model. We exploit this fact in our
growth policy analysis in the final chapter. We rely on this fact here as
well, as we illustrate the significance of a slightly higher growth rate.
Assume that today's GDP is unity. If n and r denote the GDP growth rate
and the discount rate, the present value (PV) of all future GDP is

PV = f e-( r-" dt =11(r — n) .

For illustrative parameter values (r = 0.05 and n = 0.03), PV = 50. How

much is this PV increased if the growth rate is increased by one half of
one percentage point? When r = 0.05 and n = 0.035 are substituted in, the
answer is PV = 66.7. Thus, the once-for-all equivalent of living in a world
with a one-half-point higher growth rate is a gift equal to 16.7 times the
starting year's GDP.
To have some feel as to whether this is "big" or "small" we
compare this estimate to what was a hotly debated issue in Canada almost
20 years ago — the free trade arrangement with the United States. In that
policy application, the benefits for Canada were estimated to be a per-
period level effect of 3% higher living standards every year into
perpetuity. The one-time lump-sum equivalent of this annuity is (.03)/(r —
n)= .03/.02 = 1.5. Thus, in the past, Canadians have treated a policy that
delivers this much as a big issue. The pro-savings tax cut that we are
examining here involves a benefit that is more than 11 times as big. When
contemplating results such as this, Robert Lucas (1988) argued that it is
hard to get excited about anything else in life!
239
113 Two-Sector Endogenous Growth Models

Before immediately sharing Lucas' excitement, we should check how

sensitive this "big response" result is to the particular features of the AK
model. To pursue such a sensitivity test, we explore the two-sector
literature. In this approach, a separate knowledge-producing sector of the
economy is specified. This knowledge is called human capital, H, and it is
distinct from physical capital, K.
It is assumed that both forms of capital are important in the
manufacturing sector — that part of the economy that produces
consumption goods and the physical capital. A standard production
function is assumed in this sector:

Y = AK' (bH)''.

The total stock of human capital in the economy is H. Proportion b is

employed in the manufacturing sector. The output, Y, can take the form of
either consumption goods or physical capital accumulation. In other words,
the usual accumulation identity for physical capital applies:

=Y —C —G — SK (11.1)

Following Lucas (1988), it is assumed that physical capital is not

important in the education sector. Only human capital is needed to
produce knowledge, and the simplest form of a constant-returns
(production function and accumulation identity) relationship is assumed:

H = B[(1— b)H]-8H. (11.2)

B is the gross rate of return on each unit of human capital employed in the
education sector. The total amount that is employed there is proportion
(1— b) of the total. The net rate of return is (B — 8), since, for simplicity,
we assume that physical and human capital depreciate at the same rate.
It is not obvious how we should interpret the education sector. On
the one hand, it can be thought of as "home study" where individuals
simply refrain from paid employment, and use the time to learn more. On
this interpretation, these individuals are receiving no income (when away
from the manufacturing sector) so there is no taxable wage income
generated in the knowledge sector. This interpretation is appealing when
thinking of university students. The other way of thinking about the
education sector is that it is a "research institute" that employs researchers.
240
On this interpretation, everyone in society is automatically aware of all
existing knowledge. This sector's job is to produce new knowledge. This
interpretation of the education sector is appealing when thinking about
professors, since professors receive wages which the government can tax.
We proceed in two stages. First, we follow the "home study" inter-
pretation; then, later in this section, we shift to the "research institute"
interpretation. Unfortunately, the policy implications of the two models
are very different.
Households are assumed to optimize. One outcome is a standard
consumption function (listed below). Another implication is that
individuals arrange their portfolio of assets so that — in equilibrium and at
the margin — they are indifferent between their three options: holding
physical capital employed in the goods sector, holding human capital
employed in the goods sector, and holding human capital employed in the
education sector. Since there is no tax on home study activity, the after-tax
yield and the before tax yield on human capital in that sector is the same.
We denote that return by

r* =(B —(5).

It is clear from this relationship that, since there is no taxation in the home
study sector, r* is totally pinned down by two technology parameters.
The net of depreciation (but before tax) rate of return on physical
capital in the goods sector is

r =(aY I K)— (11.3)

For portfolio equilibrium, the after-tax yield, r(1 — r), must equal what is
available on human capital placed in the other sector:

r(1— r) = (11.4)

The model consists of the four numbered equations plus the consumption
function (equation (11.5) below). Here we generalize (compared to the
simple AK model) by re-introducing population growth and overlapping
generations (a positive probability of death). In this case, the growth rate
for aggregate consumption (C) is the sum of the growth rate for per capita
consumption (c) plus the population growth rate (C IC =n=y+z). From
Chapter 4 (section 4.3), we know that the consumption function is

r =(r*-P)-013+z)(P+ P))I x (11.5)

241
 Before deriving the full-model properties, we simplify equations (11.1)
and (11.2). We start by dividing (11.1) by K and (11.2) by H. Next we
substitute in the full-equilibrium condition that the growth rate for both K
and H is (y + z). Finally, we define x = C/K, and we assume a common
tax rate on all factor earnings in the goods sector (t = z). With government
spending being determined residually (G = t(Y — 8K) , the GDP identity
can be simplified further. The compact versions of equations (11.1) and
(11.2) become:

y+z+x+8(1—t)=.(1—t)(r+8)la (11.1a)
y + z B(1— b)— (11.2a)

The five equations determine r, r*, b, x and the productivity growth rate y.
In simplifying these five relationships, we have substituted in the
assumption of balanced steady-state growth — that the growth rates of C, K
and Hare all equal. Thus, we are limiting our attention to full equilibrium,
and for a simplified exposition, we are ignoring the transitional dynamics
(that are not degenerate in this case, as they were for the AK model).
Readers interested in an analysis of the between steady states dynamics
should read Barro and Sala-i-Martin (1995, p 182).
It is worth taking stock of the model's structure. The taste
parameter (p), the demographic parameters (p and z), the technology
parameters (a, B and 8), and the exogenous policy variable (t) are all
given from outside. We can use the model to determine the effect on the
steady-state growth rate of a shift to a smaller government. If there is an
exogenous cut in the tax rate (and a corresponding drop in the level of
government program spending), we would expect — on intuitive grounds —
that there would be a higher growth rate. First, the lower tax rate can be
expected to encourage growth since interest taxation is a disincentive to
accumulate capital. Further, with lower government spending, there is less
wastage. After all, in this model, since G enters neither the utility function
nor the production function, society loses whenever it has any government.
We do not intend to argue that this is the most satisfactory way of
modelling government activity. Instead, we are just pointing out that it is
in this setting that the move-to-smaller-government initiative should be
expected to be at its most powerful. It is alarming, therefore, that we find
such tiny growth rate effects (below).
Of course, to derive quantitative results, the model has to be
calibrated. For illustration, we use the following set of representative
parameter values:

242
 r net of depreciation return on capital 0.075
t tax rate 0.20
initial productivity growth rate 0.02
z population growth rate 0.02
annual death probability 0.02
6 depreciation rate 0.04
p rate of time preference 0.03
physical capital output share — manufacturing 0.33

All these parameter values are standard and appealing on empirical

applicability grounds. Further, when they are substituted into the model's
equations, they imply reasonable values for the division of market output.
The (C/Y), (EY) and (G/Y) ratios are: 0.60, 0.23, and 0.17.
With this representative calibration, when the tax rate is cut in
half (to 0.10), the productivity growth rate rises permanently, but only by
four one hundredths of one percentage point. This outcome is much
smaller than that which excited Lucas so much — the effect we calculated
from the one-sector AK model in the previous section. But before seeking
perspective concerning this vast disparity in results, we consider the
"research institute" version of this two-sector model.
We now interpret the education sector as involving market
activity, with the income being taxable. The two production functions are
exactly as before, but the net-of-tax return on human capital in the
education sector is now:

r* (B — 8)(1— t)

where t is the tax on wage income. (Since we are now highlighting the
taxability of income in this sector, we revert to our earlier practice
(followed in section 11.2 involving the AK model) of allowing for
different rates of tax on the income derived from physical and human
capital.)

Profit maximization in the goods sector requires:

aY I K rK + and (1 — a)Y I bH = ril + g

Portfolio equilibrium requires equal yields:

rK (1— ) = r* and rn (1 — t) = r* .

243
Eliminating the pre-tax rates of return from these relationships by
substitution, we have

(Y / K) = [r * +8(1— r)]l[a(1— r)]

(Y 1 1 -1)= b[r * +8(1— t)]1[(1— a)(1— t)]

These relationships (along with G = rr,K +trH H + AC, the government

budget constraint) are needed to proceed through the substitutions that
lead to the same compact expression of the model that was our focus in
the "home study" version of the two-sector model. That compact set of
equations is:

r* = (B — 8)(1— t)
y+ z = B(1— b)— 8
y+ z+ x+ 8 =[(1— g)(r* +8(1— r))]l[a(1—r)]
g = ar + (1— a)t 1 b+ a(1—r)(Ax— k8)•
[r* +8(1— r)-8t(1— a)(1—t)]1(r* +8(1—t)]
y = (r * —p)—(p+ z)(p+ p)1(x(1+ A))

The new variable is ?c, the sales tax rate. The five equations determine r*,
b, x, y and one of the three tax rates. The taste parameter (p), the
demographic parameters (p and z), the technology parameters (a, B and 6),
and the exogenous policy variables (g and two of the three tax rates) are
all given from outside. To illustrate, we consider using the model to
determine the effect on the steady-state growth rate of a cut in the tax on
wage income (financed by an increase in the tax on the earnings of
physical capital). From the first equation, we see that the cut in t can have
the same large increase in r* that we got in the AK model. Then, focusing
on the first term on the right-hand side of the fifth equation, we see that
the productivity growth rate moves one-for-one with r* (again, as in the
AK setting). The presence of the other term on the right-hand side of the
fifth equation means that things are a little more complicated.
Nevertheless, with a long life expectancy (that is, with a small value for
the death probability) this other term is not very important quantitatively.
Hence, we are back to the Lucas level of excitement. With this version of
the two-sector model, we can expect big and permanent growth-rate
effects from fiscal policy.
But there is a critical difference. In this model it is the wage-
income tax that needs to be cut to stimulate growth, not the interest-
income tax. This is because — of the two man-made inputs that are part of
244
the production processes — it is human capital that is the central one. Its
production is the process that involves constant returns to scale (and it is
the presence of constant returns to scale that makes endogenous growth
possible). This means that the growth-retarding tax is the one that lowers
the incentive for people to acquire human capital. Even when the cut in
this tax is financed by a higher tax on interest income, growth is
significantly stimulated. Simulations with a calibrated model verify that
growth is stimulated a little more if the wage-tax cut is financed by a
higher consumption tax instead of a higher interest-income tax. But this
difference is very small. Thus, the model provides only the smallest
support for the common presumption in applied tax-policy circles — that it
is physical capital that is the bad thing to tax.

11.4 Other Models of Endogenous Growth

Of course, there are many other analyses of endogenous growth, and to

pursue this literature, readers are encouraged to consult Aghion and
Howitt (1998), Jones (2002) and Helpman (2004). The analyses in these
books share the goal of the two-sector literature, in the sense that more
structure concerning the creation of productivity enhancements is offered
(compared to the reduced-form or "black box" approach of the AK model).
Perhaps the most popular of these more elaborate analyses is the class of
models that focus on investment in research. Firms in these models take a
profit-maximizing approach to investment in research. Successful research
is a stochastic process, and more research raises the probability that a
worthwhile outcome occurs. These models highlight externality problems.
On the one hand, firms invest "too little" in research because they cannot
hang on to all the benefits (higher profits) that follow from a successful
innovation. This is because, after a lag, the new knowledge becomes a
public good, and all the firm's competitors can benefit. But, in another
sense, each firm invests "too much" in research, since no individual
inventor puts any weight on a negative externality — that her invention
destroys the remaining rents that would have continued to accrue to the
previous successful inventor (had this new project not been successful).
This literature has brought a new level of rigour to Hayek's famous
concept — "creative destruction". When these micro-based models are
calibrated with realistic parameter values, they suggest that the under-
investment problem outweighs the over-investment problem (so that a
policy of supporting research and development is called for).
The creative-destruction model is not the only part of the
endogenous growth literature that has attracted much attention in recent
245
years. There has been a big revival of interest in what Arthur Okun called
"the big trade-off' many years ago. The conventional wisdom has been
that we face a trade-off between our equity and efficiency objectives. In
common parlance, if (through government policy) we try to redistribute
the "economic pie" by giving a bigger slice to the poor, the rich react to
the higher taxes by producing less, so the total size of the pie shrinks.
Policy analysts get particularly excited if they can identify an exception to
this standard trade-off. For an initiative to be both efficiency-enhancing
and equity-enhancing, we must be starting from a "second best" situation.
We encountered such a case when examining low-income support policy
in a setting that involves involuntary unemployment (in section 9.3).
Endogenous growth models open up similar possibilities.
It all boils down to whether higher inequality is good for growth
or not. In the early stages of economic development, the particularly
scarce input seems to be physical capital. Acquiring more of this input
requires saving. Since only the rich do much saving, inequality is good for
growth, and (in this setting) Okun is correct to presume that we face a
trade-off. But as development proceeds, diminishing returns for physical
capital sets in, with the result that limited physical capital ceases to be the
binding constraint. The increased availability of physical capital also
raises the return on human capital, and its scarcity starts to become the
factor that is more important for limiting further development. The
important fact of life at this stage seems to be that there are imperfections
in capital markets. In particular, human capital is rather imperfect
collateral, and loans are particularly necessary for most people since
investments in human capital are rather lumpy. (You do not gain much by
going to university for just a couple of months.) The more unequal is the
distribution of income, the higher is the proportion of the population that
is blocked by these imperfections in the loan market. In this case, income
redistribution can lower the number of loan-rationed individuals, and so
relieve these constraints. The result is both lower inequality and higher
economic growth.
Saint-Paul's (2006) recent study is less optimistic on this score.
He focuses on the relationship between wages and productivity. For a long
time, wages for less-skilled individuals have increased with advances in
productivity, while more recently this correlation appears to have turned
negative. Saint-Paul's model is designed to offer an explanation for this
pattern, and it highlights the fact that people shift their demand away from
the products that use low-skilled labour intensively as they become more
wealthy. As a result, as growth continues, there is an ever-decreasing
demand for unskilled labour — in relative terms — and this outcome stems
from a change in the pattern of demand as the growth process continues,
246
not from skill-biased technical change. But the implication is the same —
disappointment for those wanting to pursue both equity and growth.
There is an extensive multiple-equilibria literature that highlights
the possibility of escaping an equity-efficiency trade-off (see Aghion et al
(1999) and Das and Ghate (2005)). Some studies, such as Galor and Zeira
(1993) stress the possibility of "poverty traps." With nonlinearities stem-
ming from such phenomena as increasing returns, there can be two
equilibria: both a low-growth and a high-growth outcome. Quite a few
development economists have become convinced of the applicability of
these models, and so they argue that limited programs of foreign aid may
be wasted. According to this view, the North must offer the South enough
aid to move the trapped economies all the way to the neighbourhood of
the higher-growth equilibrium. Some recent empirical work (Graham and
Temple (2006)) suggests that about one-quarter of the world's economies
are stuck in a low-growth equilibrium.
We turn now from policy implications to specification issues.
There is controversy concerning the most appealing way of modelling the
growth process. Some models (such as Romer (1990)) focus on R&D
spending by firms. Romer stresses that research output is different from
human capital since it becomes common knowledge that is not lost to the
production process if labour is unemployed. Despite this difference,
Romer's model has an important similarity with the human-capital model,
since Romer assumes that each year's increase in knowledge is pro-
portional to the pre-existing level of knowledge. It turns out that the
number of useful discoveries depends on the size of the population. The
factor of proportionality linking the new output of knowledge to the pre-
existing level of knowledge is a function of the number of people. More
people mean more research centres, so there is a bigger chance of
successful research emerging. This result is known as the "scale effect" —
the bigger is the level of the population, the higher is the productivity
growth rate.
Jones (2003) is critical of Romer's R&D model for two reasons.
Despite Kremer's (1993b) influential study of growth over the last one
million years, Jones argues that the empirical evidence does not support
the scale-effect prediction. Second, he argues that it is quite arbitrary to
locate the linearity assumption that is necessary for non-explosive
endogenous growth within a production function. Why should the input-
output relationship in either the human-capital model or the R&D model
be precisely linear? After all, data on individual establishments does not
seem to support the assumption of absolutely constant marginal product
relationships. Jones argues that it is more plausible to assume a linear
relationship in the population growth part of the model. Defining b and d
247
as the birth and death rates, and L as the size of the population, we can
specify L. (b — d)L , which implies L / L = z, and this is not arbitrary.
Indeed, this relationship is true by definition. As a result, Jones argues that
this equation forms a more plausible, less arbitrary, basis for endogenous
growth theory. Nevertheless, there has not been a convergence of views
on this controversy. For one thing, Solow (2003) has noted that there is
considerable evidence in favour of conditional convergence among OECD
countries. These countries would have to be converging in demographic
patterns for Jones' model is to explain this outcome.
Given the central nature of the issues that are being explored in
endogenous growth theory, and the fact that substantial differences remain
concerning the relative appeal of several modelling strategies, it is
understandable that many economists are involved in this investigation —
even though (as we have seen above) the quantitative relevance of policy-
induced changes in the growth rate in some models can be questioned.
For the growth policy analysis in the next chapter, we rely on a
human capital model that involves several features. First, we follow Barro
and Sala-i-Martin (1995, pp 144-146), and collapse the two sectors (the
physical and human capital producing sectors) into one. This yields a
system which has the simplicity of the AK model, but which has a richer
interpretation. Second, we follow (and extend slightly) Mankiw and
Weinzierl (2006) by introducing imperfect competition. This extension
makes it possible for us to consider creative-destruction effects in a
rudimentary fashion — while avoiding the complexity of the research-and-
development class of endogenous growth models. (We do not endogenize
the degree of monopoly power.) In the next chapter, we extend this
framework, to allow for a subset of households to be short-sighted. Two
things follow from this extension. First, these individuals never save, so
they remain at a lower level of living standards. With both "rich" and
"poor" in the model, we can investigate the effect of redistributive policies
on the overall growth rate. The second implication of short-sightedness is
that it makes these individuals prone to shirking on the job. Employers
react by paying efficiency wages, and unemployment results. Within this
setting, we can investigate the effect of job-creation policies on growth.
As just suggested, compared to the literature surveyed in the last
section of this chapter, the model we now outline involves a middle-of-
the-road assumption concerning factor intensities. The standard AK model
specifies that knowledge is proportional to the aggregate stock of physical
capital, while the Lucas two-sector framework makes human capital the
engine of growth. The former specification leads to the policy that
interest-income taxes should not exist, while the second leads to the
proposition that wage-income taxes should not exist. The specification
248
suggested by Barro and Sala-i-Martin is an appealing intermediate speci-
fication. It involves the assumption that physical capital and human
capital are used in the same proportions when producing all three items:
consumer goods, new physical capital goods, and new knowledge (human
capital).
To outline this framework concisely, we suppress the possibility
of monopolistic competition initially. National output, Y, is either
consumed privately, C, consumed in the form of a government-provided
good, G, or used to accumulate physical or human capital, k + H :

Y=C+G+k +H.

For simplicity, we ignore depreciation of capital. Both forms of capital,

and consumer goods, are produced via the following production function:

Y = F(K,H)= fiK" 11 1-a

with a a positive fraction. By defining B = H/K and A= fie" , the

production function can be re-expressed as Y = AK.
Profit maximization on the part of firms results in factors being
hired to the point that marginal products just equal rental prices (r is the
rental price of physical capital; w that for human capital):

aA = r
(1-a)Y1H = w.

Households maximize utility:

f[ln C, + lnG,]e-P1 dt
subject to C + k + H = rK + wH. As usual, p is the rate of impatience.
For simplicity at this stage (until the next chapter), we ignore the taxes
that are necessary to finance government spending. Households choose
the time paths for C, K and H. The first-order conditions are:

n= r — p
= W.

n is the percentage growth rate of consumption (which equals the

percentage growth in all other aggregates (Y, K, H and G) in the balanced
249
growth equilibrium). The intuition behind these behavioural rules is
standard. Households save as long as the return on capital exceeds their
rate of impatience, and this saving generates the income necessary to have
positive growth in consumption. In equilibrium, households must be
indifferent between holding their wealth in each of the two forms of
capital, and the equal-yields relationship imposes this equilibrium
condition.
A compact version of the model can be specified by defining g =
G/Y and c = C/K, by equating the two marginal product expressions, and
by dividing the resource constraint through by K:

n(1 + B)= A(1 — g)— c

B = (1 — a)la
A = 1811"
r = aA
n=r—p

Among other things, we add taxes to this system in the next chapter, to
determine the method of government finance that is most conducive to
raising growth and household welfare. In the present chapter, we focus on
the implications of imperfect competition.
The easiest way of allowing for some market power is to specify a
two-stage production process. Intermediate goods are created by primary
producers (competitive firms) employing both forms of capital. Then,
final goods are produced by the second-stage producers. This second stage
involves fixed-coefficient technology (each single final good requires one
intermediate product and no additional primary factors). The final good is
sold at a mark-up over marginal cost:

price of final good = (mark-up)(marginal cost)

Letting the price of final goods be the numeraire (set at unity), the mark-
up be m > 1, and marginal cost be denoted by MC, this standard pricing
relationship becomes

1 = m(MC).

Given competitive conditions among primary producers, the marginal cost

of the second-stage firms is

MC = rl FK = wl FN .

250
Combining these last relationships, we have

r F, I m
w = FN M.

The final sales of the second-stage producers constitutes the GDP.

Measuring this aggregate from the expenditure side of the national
accounts, we know that it equals C + K + H + G. Measuring GDP from
the income side, it is the sum of primary-factor incomes, wH + rK, plus
the profits of the second-stage producers, it. Given the factor pricing
equations, and the assumption that the production function involves
constant returns to scale, the national income measure is (F(K,H)Im) + 7C.
Profits are defined as the excess of final sales F(K,H) over what must be
paid to the primary producers, F(K,H)/m:

it = F(K, H)[(m —1)1 m].

Substituting this definition of profits into the national income definition,

we end with C + K + H + G = F(K, H). We conclude that no re-
specification of the economy's resource constraint is appropriate with this
simple specification of imperfect competition.
To complete the model, we allow for the possibility that the
economy-wide productivity parameter, fl, is a function of the share of
GDP that goes to profits. The idea is that higher profits make more
innovation possible, so we specify

= 0[71- 1(Ka )]° = —1)1

where 0> 0. We do not formally model the imperfect competition. Instead,

we simply interpret m as a parameter that can be affected by competition
policy. Further, each individual firm takes /3 as given. Thus, we consider a
slightly revised production function:

Y = 0[(m —1)/ mr[Kaill' ].

This is a short-cut or "black box" specification of the essence of the R&D

models of endogenous growth. No optimizing basis for investment in
research is offered here; instead there is the straightforward assumption
that an exogenous increase in the availability of profits (relative to
primary factor earnings) results in a higher level of technological ability.
251
 Removing the government for a more simplified exposition, we
have the revised compact listing of the model:

n(1 + B) = A — c
B= (1 — a)I a
A = 0131-a Rm —1)1 /ni g
r = aAl m
n=r—p

This model determines A, B, r, n and c, and we are interested in the effect

of the degree of monopoly on the growth rate:

do I dm = [a A I (m —1)][0 — (in —1)].

If profits do not act as an incentive for innovation, as assumed by Mankiw

and Weinzierl (2006) and Judd (2002), and which we can impose by
setting 6 = 0, then a more competitive economy involves less pulling
scarce resources away from accumulating capital, and therefore higher
growth (dn/dm < 0). But if profits do increase innovation (which is
derived as a possibility in quite elegant models such as Aghion and Howitt
(1992), rather than just assumed here (if 0 > 0)), then a more competitive
economy involves lower ongoing growth. In this case, the pro-growth
effect of pulling resources back into the production of more capital is
dominated by the anti-growth effect of there being less technical change.
Most applied competition-policy discussions highlight this, tension, and it
is similar to the creative-destruction tension in the imperfect competition
models that offer an explicit structure for modelling the decision to invest
in research. In the next chapter, we use this framework to explore how the
existence of imperfect competition affects the implications of endogenous
growth theory for tax policy.

11.5 An Evaluation of Endogenous Growth Analysis

We have considered several endogenous growth models in this chapter.

Since the results are so different, we need some sort of common-sense
base for relating some of this work to the real world. Thus, in this section
of the chapter, we provide a "back of the envelope" estimate of the payoff
from increased investment in education — an estimate that is not based on
any specific or formal macro model.

252
 By comparing peoples' incomes — people with and without further
education — many labour economists have estimated the returns to more
education. A typical result from these cross-section regression equations
(involving earnings as a function of years of schooling) is that the annual
return to education is in the 7.5% range. We use this return to estimate the
higher standards of living that we might enjoy if everyone were more
educated. This is a controversial thing to do, since it may be that much of
the estimated return in the cross-section regressions just reflects a signal —
that smarter people can make it through school and less clever people
cannot. So, even if school does nothing but visibly separate people into
these two groups, employers will find it useful to use this signal of native
intelligence to save decision-making costs. The education process itself
may not raise productivity. If this is the case, and the high private return
reflects just signaling, then we should not apply it in an economy-wide
thought experiment. After all, a signal has no discriminating power if
everyone has the credential. Despite this controversy, we assume that
education is not just a signal. This means that the following calculation is
biased toward finding too big a pay-off from more investment in
education. This bias only supports the conclusion that we reach, however,
since the estimated pay-off is very small — despite the upward bias.
Consider transferring 1% of GDP out of current consumption, and
into education. This reallocation is like buying an equity that pays a
dividend of 7.5% of 1% of GDP forever. The present value of the stream
of dividends that accompanies this year's equity purchase is therefore

(.075)(0.01)(a + a 2 + a3 +...)

where a = (1 + n)1(1 + r),and today's GDP is unity. This present value

expression can be simplified to

(.075)(.01)V

where V = (1 + n)1(r — n).

The cost of this year's equity purchase is the lost consumption, which is
0.01. The net benefit, NB, is therefore

NB = (0.01)[(.075) V— 1].

If society embarks on a permanent shift of resources into the education

sector, we are making one of these equity purchases every year. The
present value, PV, of this entire sequence of purchases is (NB)(V). We are
253
interested in the percentage change in living standards that is achieved by
moving from a PV value of V to the new PV value of (NB)(V). The ratio of
these two outcomes is simply NB:

NB = (0.01)[((0.075)(1 + n)I(r — n)) — 1].

Illustrative parameter values are r = 0.075 and n = .04. These values make
NB= 1.23%
Given the amount that most western countries are spending on
education at this time — something in the order of 5 or 6 percent of GDP
GDP going to—apermntopercnagoitncreas th reof
education has to be regarded as a very big investment. Yet this investment
is estimated to bring a series of annual benefits that is equivalent to a one-
time pay-off equal to 1.23%. The interesting question is what increase in n
(from a starting value of 0.04, with an r value of 0.075) is required to raise
V = (1 + n)I(r — n) by 1.23? The answer is that n must rise to 0.0406. In
other words, when measured in terms of the equivalent permanent annual
growth rate increase, a truly large investment in education can be expected
to raise living standards by just four one-hundredths of a percentage point.
This is just what we obtained from the "home study" version of the two-
sector endogenous growth model in section 11.3. This back-of-the-
envelope exercise suggests that this particular model's predictions may be
the most relevant for assessing real-world policy options. In turn, this
suggests that we temper our enthusiasm — compared to Lucas'. For an
alternative defense of this same recommendation, see Easterly (2005).
We close this section with a brief reference to some empirical
studies. Have economists found any evidence that policy can permanently
affect growth rates? At a very basic level, the answer surely has to be
"yes". The experience of the formally planned economies has shown
dramatically that a market system cannot work without a set of supporting
institutions (such as a legal system that creates and protects property
rights). These institutions — that induce people to be producers and not just
rent seekers — are the outcome of policy.
But at a more specific level, there is evidence against policy
mattering. An example pointed out by Romer (2001) involves three
countries. During the 1960-1997 period, the United States, Bolivia and
Malawi had essentially the same growth rates, and very different fiscal
policies. These differences have resulted in very different levels of living
standards (as predicted by exogenous growth theory) but not different
growth rates. Another problem concerns estimated production functions.
As these estimates vary the definition of capital from a narrow physical
capital measure to an ever more broad measure of capital (including
254
 human capital), capital's estimated share ranges between one-third and
eight-tenths. No study has found a value of unity (which is required to
support the linear differential equation that lies at the core of the
endogenous growth framework).
Finally, Jones (1995) has presented some interesting projections.
He used the pre-1940 growth experience of the United States as a basis for
making a simple projection of how much per capita incomes would grow
over the next 45 years. He replicated what someone might have predicted
back then — without taking any account of later developments. He
expected these projections to under-estimate actual growth, since they
ignore the vast increase in human capital that has been accumulated
during the later period. For example, the share of scientists and engineers
in the labour force has increased by 300%. In addition, now 85% of
individuals, not 25% finish high school. But despite all this, the simple
projections over-estimated the actual growth in living standards. This
exercise has created a challenge for new growth theory to explain.
What advice can be given to policy makers — given all this
controversy? It would seem that — if a policy involves definite short-term
pain and uncertain long-term growth-rate gains — it may be prudent to
postpone implementing that policy until the insight from further research
can be had. This cautious strategy does not mean that nothing can be
recommended in the meantime. After all, a number of policies can be
present-value justified if they deliver a (much less uncertain) one-time
level-effect on living standards. This may not be as "exciting" as a
permanent growth-rate effect, but it is still very worthwhile. For example,
the increase in living standards that we estimated to follow from
government debt-reduction policy (see section 10.4) is a very worthwhile
outcome. This analysis assures us that there is a large scope for relevant
policy analysis — even if it is too soon to base advice on an area of inquiry
— endogenous growth theory — that is still "in progress".

11.6 Conclusions

We are leaving our introduction to growth theory in a somewhat untidy

state. For one thing, we have had to acknowledge that there are empirical
issues that both exogenous-growth and endogenous-growth models have
trouble explaining. For another, our coverage of the two-sector endo-
genous growth models has been simplified by our focusing entirely on
steady-state outcomes; that is, by our ignoring what happens along the
transition path between steady states. Finally, we have not attempted a
detailed exposition of the much more elaborate research-and-development
255
based endogenous-growth models. We make no attempt to rectify this
latter shortcoming in the book's final chapter, for two reasons. First, there
are other more advanced books (such as Aghion and Howitt (1998)) that
serve this need. Second, we need the space to provide an extensive
application of the more basic growth analysis to a set of policy questions:
Does a pro-growth tax structure hurt those on low incomes? Do policies
that reduce unemployment retard growth? Does an aging population lead
to lower growth? It is to these questions that we turn in our final chapter.

256
Chapter 12

Growth Policy

12.1 Introduction

The purpose of this chapter is to assess the analytical support for several
propositions advocated by policy advisors, using a simple version of
endogenous growth theory. We proceed through the following steps. First,
we consider the optimal tax question by comparing income taxes to
expenditure taxes. Without any market distortions allowed for in the
analysis, we show that expenditure taxes are recommended. This basic
analysis has been very influential. For example, the President's Advisory
Panel on Federal Tax Reform in the United States (issued in 2005) argues
for a wholesale replacement of the income tax with an expenditure tax.
Our analysis, indicates how the analytical underpinnings for this proposal
are sensitive to the existence of other sources of market failure. For
example, in section 2, we consider a second-best setting in which the
government is "too big." In this situation, it is appropriate for the
government to levy distorting taxes, so the income tax should not be
eliminated. Many policy analysts argue both that government is "too big"
and that income taxes are "bad." This section of the chapter challenges
these analysts to identify the analytical underpinnings of their views. A
related finding follows from an extension that allows for two groups of
households, one so impatient that these individuals do not accumulate
physical capital. Progressive taxation is analyzed by taxing only "rich"
households to finance transfers to the "hand-to-mouth" group. The
analysis does not support replacing a progressive income tax with a
progressive expenditure tax.
Other second-best considerations are considered in later sections
of the chapter. In section 3, we focus on consumption externalities. This
consideration pushes the conclusion for tax policy in the opposite
direction; it strengthens the case for the expenditure tax. Some analysts
(such as Frank (2005)) argue that consumption externalities are very
important from an empirical point of view. How else, asks Frank, can we
explain the fact that survey measures of subjective happiness show no
increases over a half century — when this period has involved significant
increases in the standard measures of economic growth (such as per capita
GDP)?

257
 Unemployment is added in section 4 of the chapter. In this
second-best situation, an employment subsidy is supported as a
mechanism for simultaneously lowering unemployment and raising the
growth rate. This result is a challenge for policy analysts who argue that
we face a trade-off in the pursuit of low-income support policy (our equity
objectives) and higher-growth policy (our efficiency objective). It is
interesting that basic endogenous growth theory can provide examples
such as this one — that illustrate the relevance of what Alan Blinder has
called "percolate up" economics. In this class of models it seems
relatively easy to find settings in which a fiscal policy that is designed to
help those lower down on the "economic ladder" has indirect benefits for
those up the ladder. This is similar in spirit, but opposite in direction, to
the more widely known approach known as "trickle down" economics —
in which a fiscal policy designed to help the rich generates indirect
benefits for those further down the economic ladder.
Finally, in section 5 of the chapter, we outline what basic growth
models imply about how future living standards may be affected by a
major demographic event that is much discussed. We consider the
increase in the old-age dependency ratio that will accompany the aging of
the population that will occur as the post-war baby-boom generation
moves on to retirement.

12.2 Tax Reform: Income Taxes vs. the Progressive Expenditure Tax

Many economists and tax-reform panels have called for a shift in tax
policy: a decreased reliance on income taxation and an increased reliance
on expenditure taxes. The standard analytical underpinning for this tax-
reform proposal is endogenous growth theory. While less emphasis is
given to equity considerations in the theory (since standard growth theory
involves a single representative agent), policy analysts sometimes call for
a progressive expenditure tax — to avoid the regressivity that would
otherwise accompany the use of sales taxes. The purpose of this section of
the chapter is to use a simple version of endogenous growth theory to
review this debate.
We begin with the proposition that total output produced each
period, Y, is used in two ways. First, it is purchased by households to be
consumed that period, C. Second, it is purchased by households to add to
their stock of capital, K. K refers to that period's increase in capital. The
supply-equals-demand statement is

(12.1)
258
Next we specify the production process; in this initial case, it is very
simple — output is proportional to the one input that is used in the
production process, K:

Y = AK. (12.2)

A = r, the rate of return on capital, is the amount that output increases as

additional units of the input are hired (the average and marginal product of
capital). A is the pre-tax return that households earn on their saving. There
are two potentially controversial aspects of this specification of the
nation's input-output relationship. First, there is no diminishing returns (as
more capital is employed, its marginal product does not fall). As
explained in chapter 11, the economy's equilibrium growth rate is
independent of fiscal policy if this assumption is not made, so we follow
that standard practice in presenting the standard policy-oriented analysis
here. Second, once the output emerges from this production process in the
form of consumer goods, it is costless for society to convert that new
output into additional capital. Again, this assumption is made to keep the
analysis consistent with standard practice. Also, for simplicity, we ignore
depreciation of capital.
In this initial specification, the government has just one function;
it levies a proportional income tax rate, 1-, on the income that households
earn by employing their capital, and a proportional sales tax rate, s, on
household consumption spending. The government uses this revenue to
finance transfer payments, R, to households. The government's balanced-
budget constraint is

R= rrK + sC. (12.3)

This specification of government is rather contrived, since all households

are identical. There seems little point in having the government impose
taxes that can distort each household's decisions, when the only use for
that revenue is to transfer the funds back to that same household. We
make this assumption simply because it is standard in the tax-policy
literature. Also, it serves as a base upon which we can build a more
interesting analysis of government policy in later sections of the chapter.
The only other relationship that is needed to complete this initial
analysis is the one that describes how households make their savings-vs-
consumption decision. We assume that their objective is to maximize their
utility, which is defined by the following standard function:

259
 utility = ln Cicli

subject to their budget constraint:

C+K=rK+R-rrK-sC.

This constraint states that households pay for their consumption

and capital accumulation (saving) by spending all their after-tax income
and their transfer payment. As we discovered in Chapter 4, the decision
rule that follows from this optimization is

eic. r(1— r) — p (12.4)

The intuition behind this behavioural rule is straightforward. Households

save as long as the after-tax return on capital exceeds their rate of
impatience, and this saving generates the income that permits positive
growth in consumption.
We now write the four numbered equations in a more compact
form. We focus on a balanced-growth equilibrium in which all aggregates:
consumption, output, and the capital stock all grow at the same rate
n=0IC =1'. IV = K / K, with the tax rates, r and s, the ratio of
consumption-to-capital, c = C/K, and the ratio of transfer payments-to-
GDP, z = R/Y, all constant. To accomplish this, we divide (12.1) and (12.3)
through by K and Y respectively, and use (12.2) and the balanced growth
assumption to simplify the results:

r=c+n (12.5)
z=r+scIr (12.6)
n = r(1—r)— p (12.7)

Equations (12.5) through (12.7) solve for the economy's growth rate, n,
the consumption-to-capital ratio, c, and one of the tax rates (we assume
that the sales tax rate, s, is residually determined by the model). The
equations indicate how these three variables are affected by any change
we wish to assume in the technology parameter, r, the taste parameter, p,
and the government's exogenous instruments — the transfers-to-GDP ratio,
z, and the income tax rate, r.
Given the existing debate on tax reform, we focus on cutting the
income tax rate, and financing this initiative with a corresponding increase
260
in the sales tax rate. The effect on the growth rate follows immediately
from (12.7): dnldr.—r< 0. Thus, shifting to the sales tax raises the
ongoing growth rate of living standards. But there is a one-time level
effect on living standards (consumption) as well, and from (12.5) we see
that this outcome is adverse: dcl dr =—dnI dr =r>0. So a shift away
from income taxation toward an increased reliance on expenditure taxes
shifts the time path of household consumption in the manner shown in
Figure 12.1. (We noted in section 11.2 that this model involves no
transitional dynamics; that is, the time path for living standards moves
immediately from its original equilibrium path to its new one.)

Figure 12.1 Shifting to Expenditure Taxation —

Short-term Pain for Long-term Gain

Living Standards
(In C)

Time

Given that there is short-term pain to achieve long-term gain, it is not

immediately clear that this tax substitution is supported. But we can
address this issue by using the household utility function. It is shown later
in this section that when the household utility integral is worked out,
overall social welfare (SW) is

SW =Elnc+(n1 p))1 p (12.8)

where C0 = cic and K0 is the initial capital stock, at the time when the
tax substitution takes effect. From (12.8), we can determine the effect on
overall welfare of the tax substitution:

261
 dSW 1 dr =[(11 c)(dc 1 dr)+ (11 p)(dn1 dr)11 p

Using the one-time-level and ongoing-growth-rate effects reported above,

we have

dSW 1 dr =r(p—c)1(cp 2 )

which can be simplified by using (12.5) and (12.7):

dSW 1 dr =—(rr 2 )1(cp 2 )< 0.

The fact that this expression is negative indicates that the government can
raise peoples' material welfare by cutting the income tax rate — all the way
to zero. This is the standard proof that we should rely on expenditure, not
income, taxes to finance the transfer payments. Because the generosity of
the transfer does not affect the household's consumption-saving choice,
there is no such thing as an "optimal" value for transfers. Whatever level
is arbitrarily chosen has to be financed by expenditure taxes if the
government wishes to maximize the material welfare of its citizens.
We now consider a sensitivity test, by asking how the optimal-tax
conclusion is affected by our replacing the government transfer payment
with a program whereby the government buys a fraction of the GDP and
distributes it free to users (as in the case of government-provided health
care). We continue to assume that no resources are needed to convert
newly produced consumer goods into new capital or (flow) into the
government service. Thus, the economy's production function remains
(12.2), and the supply-equals-demand relationship becomes

Y=C+K+G (12.1a)

where G is the level of the government-provided good each period. The

government budget constraint becomes

G= rrK +sC. (12.3a)

Finally, we assume that households value the government-provided good,

so there are two term's in each period's utility function:

utility = f [ln C, + y ln
i=0

262
Parameter y indicates the relative value that households attach to the
government service. Since the government imposes the level of
government spending, individual households still have only one choice to
make; they must choose their accumulation of capital with a view to
maximizing the present discounted value of private consumption. The
solution to that problem is still equation (12.4).
The model is now defined by equations (12.1a), (12.2), (12.3a)
and (12.4). Defining g = GIY as the ratio of the government's spending to
GDP, these relationships can be re-expressed in compact form:

r(1 g) , c+ n (12.5a)
g=r+scIr (12.6a)
n = r(1— r)— p (12.7)

and the modified overall material welfare function is

SW = [ln C0 + ln Go + (1 + y)(n I p)]I p (12.8a)

It remains true that dc I dr = —dn / dr = r, but before we make use of these

outcomes, we focus on the question of the optimal level of government
program spending. The criterion used to set government spending
optimally is to arrange the outcome so that the following condition holds
for households:

(MU / price) private consumption good = (MU / price) g overnment provided good

With the price of both goods being unity, and the utility function that has
been assumed, this condition requires 1/C = y/G; that is, the optimal
program-spending-to-GDP ratio is g* = yclr. This definition is used to
simplify the overall welfare effect of varying the income tax rate. The
result is:

dSW / dr =[r2 1(cp 2 )][(g — g*)—

As before, material welfare is maximized when this expression is zero,

and this requires no income tax (1- = 0) only if government is set optimally
(g = g*). If, however, government spending is "too big" (g > g*), there
should be an income tax.
What is the intuition behind the result that income taxes should be
used to finance some of the government when government is too big? The
263
answer hinges on the fact that this resource misallocation is in terms of a
fixed proportion of a growing GDP. With growth, the magnitude of the
distortion is being magnified. In this case, then, growth has a "bad"
dimension. This is why it becomes "good" to rely, to some extent, on the
anti-growth instrument for raising tax revenue (the income tax rate). Barro
and Sala-i-Martin (1995, pp 144-161) and Marrero and Novales (2005)
consider similar models. Their government good is infrastructure, and it
enters the production function instead of the utility function. Nevertheless,
the same dependence of the optimal tax question on second-best con-
siderations — whether the government has set its expenditure "properly" —
is stressed. A related result is also available in Garcia-Penalosa and
Turnovsky (2005) in a developing economy setting. The fact that the
government cannot levy taxes on workers in the traditional sector creates
a second-best situation, with the result that capital taxation is appropriate.
Returning to our specific analysis, we conclude that we cannot
know whether the actual economy involves the "preferred" ratio of
income to sales tax rates or not, unless we know whether the government
is "too big" or not, and by how much. Many pro-growth policy analysts
defend two propositions: (i) that income taxes are too high relative to sales
taxes, and (ii) that the government sector is too big. Our analysis has
shown that the more correct these analysts are concerning the second
proposition, the less standard growth theory supports their first
proposition. One purpose of this section of the chapter is to identify this
tension that policy analysts do not appear to be aware of.
Thus far, there has been just one input in the production process,
and we have interpreted that input quite broadly — as including both
physical and human capital. For most of the rest of this chapter, we wish
to contrast patient households who plan inter-temporally and acquire
physical capital (as discussed above) with households that always spend
their entire labour income and transfer payments. To make this
comparison explicit, we must distinguish physical capital (K) from human
capital (II). As outlined in section 11.4, the supply-equals-demand
equation becomes

Y -= C + + + G

since output is either consumed privately, consumed in the form of a

government-provided good, or used to accumulate physical and human
capital.
Both forms of capital, the government service, and consumer
goods, are produced via the standard Cobb-Douglas production function:

264
 Y Ka H I'

By defining B = H/K and A= 13B 1' , the production function can be re-
expressed as Y = AK — the same form as that used above. But in this
setting, tax policy affects A.
As explained in section 11.4, it is assumed that households rent
out their physical and human capital to firms (that are owned by other
households). Profit maximization on the part of firms results in factors
being hired to the point that marginal products just equal rental prices (r is
the rental price of physical capital; w that for human capital):

aA = r
(1 — a)Y1H = w.

Households maximize the same utility function. In this case, the constraint
involved is: (1 + s)C + K + H = r(1— r)K + w(1— r)H . In addition to the
familiar consumption-growth rule, n= r(1—r)— p , the optimization leads
to r = w. In equilibrium, households must be indifferent between holding
their wealth in each of the two forms of capital, and the equal-yields
relationship imposes this equilibrium condition. Finally, the government
budget constraint is G = zrK + rwH +sC, and the compact version of the
full model is:

A= /3B' -a
r = bA
n = r(1 — r)— p
B = (1 — a)1 a
n(1 + B) = A(1 — g) — c
g= r+ scl A.

The endogenous variables are: A, B, r, n, c, and one of the tax rates. It is

left for the reader to verify that all the conclusions that we noted earlier in
this section continue to hold in this slightly extended setting. The reason
that the distinction between physical and human capital does not yield
different conclusions is because the economy uses the two forms of
capital in exactly the same proportions when producing all four items:
consumer goods, the government good, new physical capital goods, and
new knowledge (human capital).

265
 Perhaps the only result that readers might need help with concerns
the effect of the tax substitution on social welfare. The social welfare
function is the utility function of the representative household:

SW = [ln C + yln G]e-mdi

We know that C; = Coe"' and that G, = "' so the social welfare

expression can be simplified to

SW = [(ln Co + y In Go ) / p]+ (1 + y) ln(em

The integral in this last equation can be re-expressed as n times fie'di,

and this expression, in turn, can be simplified by using the standard
formula fuck = uv — Jvdu, to yield

(-1 / p)[i + di] = —e-1(1+ pi) / p2

p 2 r,

When this solution is evaluated at the extreme points in time, and the
result is substituted into the expression for SW, equation (12.8a) above is
the result. When differentiating that equation with respect to the tax rate,
we use c = C / K, which implies that (dC0 / C0 ) / dr = (dc / c) / dr since K
cannot jump ( Ko is independent of policy). Also, since G0 = grK0 , we
know that dG0 / dr = 0 as well. The g* = yc I r result emerges from
maximizing SW subject to the resource constraint, that is, by evaluating
dSW I dG = 0.
We now move on to consider income-distribution issues. Mankiw
(2000) has suggested that, for the sake of realism, all fiscal policy
analyses should allow for roughly half of the population operating as
infinitely-lived family dynasties and the other half operating hand-to-
mouth. Thus far in this tax-policy analysis, we have not followed this
advice. Further, since all households have been identical, progressive
taxes could not be considered. We now extend our analysis so that we can
follow Mankiw's suggestion — both to increase the empirical applicability
of the analysis, and to make it possible to investigate progressive taxes.
Since the poor consume a higher proportion of their income than
do the rich, a shift from a proportional income tax to a proportional
expenditure tax makes the tax system regressive. For this reason, policy
266
analysts are drawn to the progressive expenditure tax. It is hoped that this
tax can avoid creating an equity problem, as we take steps to eliminate
what is perceived to be an efficiency problem. For the remainder of this
section of the chapter, we use our otherwise standard growth model to
examine shifting between income and expenditure taxes — when both
taxes are strongly progressive. To impose progressivity in a stark fashion
we assume that only "Group 1" households — the inter-temporal
consumption smoothers who are "rich" — pay any taxes. "Group 2"
households — the "poor" individuals who live hand-to-mouth — pay no
taxes. In this section, we revert to our original specification concerning
the expenditure side of the government budget. Once again we assume
that there is no government-provided good; the taxes collected from the
rich are used to make a transfer payment to the poor.
Group 2 households have a utility function just like the Group 1
utility function, except Group 2 people are more impatient. Their rate of
time preference is 0, which is sufficiently larger than the other group's
rate of impatience, p, that Group 2 people never save voluntarily, so they
never acquire any physical capital. It is assumed that they simply have to
do some saving, in the form of acquiring the human capital that is
absolutely required for employment. But beyond that "compulsory"
saving, they do none. Thus, this group's consumption function is simply
their budget constraint. Using E to denote total expenditure by this portion
of the population (assumed to be one half of the total), we have
E = R + (wH — 1.1)1 2, since this group receives no interest income and it
pays no taxes, but it does receive the transfer payment. Since only Group
1 pays either tax, both the income and the expenditure tax are progressive.
The government budget constraint is R = r[rK + (wH / 2)] + sC. We re-
express the Group 2 expenditure relationship and the government budget
constraint by using the optimization conditions for firms and the rich
households, and by defining e = El K and z = R/Y . The compact form of
the model is:

A = flea
r = bA
n = r(1 — p
B (1 — a)I a
n = r — a(c + e)
e = zA +[(1 — a)A — Bn]I2
z = ((1 + a)712) + (scl A)

267
We assume that the initial situation involves only an expenditure tax (t =
0 initially), and we use the model to determine the effects of moving away
from this initial situation by introducing the income tax (and cutting the
expenditure tax by whatever is needed to maintain budget balance). If the
conventional wisdom (that a progressive expenditure tax is preferred to a
progressive income tax) is to be supported, the analysis must render the
verdict that undesirable effects follow from the introduction of the income
tax. The results are

(dn 1 dr) = -r <0

(dc 1 dr) = r(1+ a)1 2a >0
(de 1 dr) = r(1- a)1 2a > 0

As usual, moving toward a heavier reliance on the income tax brings

favourable consumption-level effects (in this case for both groups), but an
unfavourable growth-rate effect. As above, we combine these competing
effects in overall welfare calculations. After some simplification, using
the initial conditions and welfare expression (12.8) for Group 1, we have

(dSW, 1 dr) = [r / 2 p2 ac][p(1+ a) - 2ac]

and for Group 2 (using a similar welfare function with 0 replacing p):

(dSW., I dr) = [r / 202 ae][(0(1— a) — 2ae]

The model must be calibrated to assess the sign of these overall material
welfare effects. To illustrate the outcomes, we assume the following
illustrative values:

Physical capital's share of GDP (a) 33%

Rate of return on capital 12%
Growth rate 2%
Initial income tax rate 0%
Initial sales tax rate 10%
Total consumption-to-GDP ratio 80%
Group 1 consumption-to-GDP ratio 48%
Group 2 consumption-to-GDP ratio 32%
Investment-to-GDP ratio 20%

These values imply a rate of impatience for the "poor" that is twice that of
the "rich" 10%), and the following initial conditions: c = 0.182 and e
268
= 0.118. When these representative parameter values are substituted into
the welfare change expressions, we see that both are positive. This implies
that both rich and poor are made better off by having the lower growth
rate that results from replacing the progressive expenditure tax with the
progressive income tax. There appears to be no support for conventional
wisdom, at least for these representative parameter values.
The intuition behind this finding runs as follows. From the selfish
perspective of the rich, the optimal transfer to the poor is zero. Thus, with
a positive transfer, the government is "too big," and since this problem is
a constant proportion of a growing GDP, less growth is preferred. Thus,
we have an additional application of the general principle that emerged
earlier in this section. Finally, as far as the poor are concerned, in this
illustrative calibration, they are impatient enough to prefer the favourable
consumption-level effect that accompanies the shift to the income tax.
The analysis in this section has involved numerous simplifying
assumptions. With a view to providing a little balance, we consider one
such assumption in the remainder of this section. By doing so, we re-
iterate Judd's (2002) point that the presence of imperfect competition
strengthens the case for an increased reliance on expenditure taxation.
Here we add taxes to the monopolistic competition model that was
explained in section 11.4. We assume that there is no government
provided good, that the government returns all revenue to private agents
as a lump-sum transfer, and that all forms of income (including monopoly
profits) are taxed at the same rate. As a result, the government budget
constraint is R = z-Y + sC. The compact listing of the model is quite
similar to that discussed in earlier parts of this section, except that the
monopoly mark-up parameter, m, appears here in the fourth equation:

n(1 + B) = A — c
B = (1 — a)1 a
A= fie"
r = calm
n = r(1 — r)— p
z= r scIA

This model determines A, B, r, n, s and c, and we are interested in the

effect of a change in the income tax rate, r, on n, c, and social welfare.
Thus we derive do /dr = —r < 0 and dc 1 dr =r / a>0, and substitute
these results into

dSW 1 dr = (1 / p)[(11 c)(dc 1 dr)+ (1 / p)(dn / dr)].

269
After simplifying, the result is

dSW I dr .(r 2 I cap2 )(1—r —m).

The best value for the income tax rate is that which makes this expression
zero. Since m exceeds unity, that best value for r is a negative number.
Not only should positive income taxation be eliminated; the ownership of
capital should be subsidized. This outcome follows from another second-
best situation. With imperfect competition, there is the opportunity for
profit income. This possibility means a reduced incentive to earn income
by acquiring and employing capital (compared to what exists in a
competitive economy). If there is no other source of market failure, it is
optimal for the government to remove this incentive — by subsidizing the
non-profit source of income (that is, by subsidizing capital accumulation).
In the real economy, there are likely several market failures — for
example, both imperfect competition and an over-expanded government
(at least as viewed from the perspective of the rich). As we have seen in
this section of the chapter, the former distortion calls for a negative r
value, while the latter distortion calls for a positive r value. Without
further empirical analysis, it seems difficult to defend the proposition that
the best value for the income tax rate is zero.

12.3 Economic Growth and Subjective Happiness

In this section, we address Frank's (2005) concern that standard growth

theory ignores the findings of the subjective happiness literature. Despite
the dramatic increase in statistics such as per-capita GDP and per-capita
consumption over the last 50 years,. citizens in all developed economies
are telling us in surveys that they do not feel any happier. Frank's
explanation for these results is that much of our higher consumption
expenditure has been on "positional goods." If everyone acquires more of
these goods, any one person's relative position is not improved by
economic growth. We offer a simple formalization of this argument in this
section. As with imperfect competition, increased support for a shift
toward expenditure taxation emerges.
Thus far (except when we considered a government-provided
good), we have assumed that only one thing affects utility — the house-
hold's own level of consumption. We now extend this specification by
allowing utility to depend on the amount of leisure that the household

270
chooses. Further, we allow one household's utility to depend inversely on
the level of consumption of other households. This last feature is
necessary if we are to explore Frank's (2005) suggestion that positional
goods be taken more seriously in otherwise standard analysis. To simplify
the exposition, and to highlight these extensions, we introduce them into
the most basic model that was considered in the first few paragraphs of
section 12.2. In particular, there is perfect competition, only one
composite form of capital, and no hand-to-mouth individuals or govern-
ment provided good.
For the present discussion, it is best to think of capital as human
capital. Letting x denote the fraction of each household's time that is spent
at work, xH is employed capital, and the remainder, (1 — x)H, is what is
devoted to leisure. The resource constraint, the production function, and
the government budget constraint are:

Y=C+H
Y = rxH
R= TrxH + sC

The remaining equations that define the model are the first-order
conditions that emerge from the following optimization. Households are
assumed to maximize

f [ln(C / C 2 ) ln((1 — x)H)]e -Pi di

subject to (l+s)C+11= r(1— r)xH + R, by choosing C, H and x. Parameter

allows for positional goods. C is the average level of consumption of
other households. If 2 = 0, no goods are positional, and each individual
simply enjoys the direct benefit of what she consumes. But if 2> 0, there
is some positional element to her consumption, since any given level of
consumption is less valuable to her, the higher is everyone else's level of
consumption (meaning that she is relatively less well off). If 2= 1, higher
consumption brings this individual no utility if others' consumption rises
by the same amount as hers. The standard analysis in the previous section
of this chapter implicitly assumes that 2= 0. In terms of this formalization,
Frank recommends an analysis of the 2> 0 case (which we now provide).
The second argument in the utility function can be interpreted as
leisure. Utility is higher, the larger is the fraction of time that the
household spends away from work. Since human capital is embedded
within each individual, taking leisure means withholding capital from the
271
production process. It is assumed that the ability to enjoy leisure rises with
the opportunities that are available, and that this range of abilities and
options is proportional to the individual's stock of human capital. This
specification makes it possible for both arguments in the utility function to
grow over time in a balanced fashion.
Optimization yields e/ C = r(1— r) — p and vc(1+ s)= r(1—r)(1—x),
so the entire model can be written in compact form as follows:

n = r(1— r)— p
tit c(1 + s) = r(1— r)(1— x)
rx = c + n
z=r+sclrx

where c is now CI H. These four equations define the behaviour that takes
place in the decentralized market economy. They can be compared to the
four equations that define the outcome that would obtain if a benevolent
planner were in charge. The planner would maximize the same utility
function, but she would recognize that C could not differ from C.
Formally, this optimization involves maximizing

f [1n(C-2 ) + yr ln((1 — x)H]e-P' di

subject to the economy's resource constraint, C + H = rxH , by choosing

C, H and x. This optimization, along with the associated definitions as
before, leads to the following set of relationships:

n=r—p
wc = r(1— .1)(1 — x)
rx = + n
z=r+sc/rx

How can we ensure that the planner's outcome and the decentralized
market outcome coincide? We answer this question in two stages. First, if
goods are not positional at all (2= 0), the outcomes are the same only if r
= s = z = 0. Intuitively if there is no market failure, there is no role for
government. Second, if goods have a positional feature (2 > 0), the
outcomes are the same only if r= 0, s = 21(1 — 2) , and z > 0. Intuitively, if
there is a market failure (a negative externality arising from consumption),
consumption should be discouraged. There is no role for an income tax.

272
 Ignoring all the other issues that have been considered in earlier
sections of this chapter, then, we can conclude as follows: if goods have a
positional feature, then the government should rely on expenditure taxes,
not income taxes. But (just as we noted at the end of section 12.2) it is
difficult to reach a firm conclusion regarding tax policy if there are two
sources of market failure in the economy.

12.4 Unemployment and Growth

As noted in previous chapters, there has been a revival of interest in what

Arthur Okun (1975) called "the big trade-off' between our equity and
efficiency objectives. As a review, consider an employment subsidy — an
initiative we considered in Chapter 9 — that is intended to lower
unemployment, and thereby help the poor. The government has to finance
this set of payments, and if this is accomplished by an increase in interest-
income taxes, there is an increase in the tax burden on the rich. The rich
react to the higher taxes by investing and producing less, so that there is a
reduction in the average person's material living standards. Okun
introduced the metaphor of the leaky bucket to describe the loss of
efficiency that is part of the redistribution process — if there are no pre-
existing market failures that redistribution can alleviate. Okun argued that
"money must be carried from the rich to the poor in a leaky bucket. Some
of it will disappear in transit, so the poor will not receive all the money
that is taken from the rich." This loss involved in the redistribution
process is thought to be particularly high if the economy's ongoing
productivity growth rate is reduced by higher taxation.
Endogenous growth theory has shown that we may not have to
face the big trade-off. The purpose of this section of the chapter is to
argue that employment subsidies that help the poor turn out to raise — not
lower — the productivity growth rate in an entirely standard model. Thus,
modern analysis suggests that policy makers can pursue egalitarian
measures of this sort with much more confidence than is usually thought
appropriate. Trade-offs can be avoided if the economy starts from a
second-best situation. In the analysis that follows (just as that in section
9.3) the second-best problem is involuntary unemployment. When
labour's price is "too high" and firms employ "too little" labour, an
employment subsidy can increase overall efficiency — even when that
initiative must be financed in a way that introduces a distortion. This is
because the prospect of being unemployed reduces an individual's
incentive to acquire human capital. An employment subsidy reduces this
problem, and this is why policies that help the poor — by lowering
273
 unemployment — can be pro-growth after all. We pursue this simple idea
for the remainder of this section of the chapter.
As in earlier sections, our analysis relies on a very simple version
of endogenous growth theory. We follow Scarth (2005), and examine an
employment subsidy paid to each firm (that lowers unemployment and so
increases the incentive to accumulate human capital). The subsidy is paid
for by levying a general income tax. Since this levy includes the taxation
of interest income, the financing of the employment subsidy reduces the
incentive to save. A formal model is needed to determine whether the pro-
growth feature of the higher employment dominates, or is dominated by,
the anti-growth influence that is part of the higher tax on interest earnings
that is levied to finance the employment subsidy.
As usual, we start with the straightforward proposition that supply
equals demand: Y = C + E + k + H . The supply of goods produced is
either consumed by each of two groups of households — the "rich" and the
"poor" — or it is used to accumulate physical and human capital. C is
consumption spending by the rich (the Group-1 households), and E is the
consumer expenditures of the poor (the Group-2 households).
Both forms of capital, and consumer goods, are produced via a
standard production function. The inputs are the utilized stocks of physical
and human capital. We follow the convention of standard growth theory
by abstracting from short-run business cycles. Thus, physical capital is
fully utilized. But, because there is structural unemployment, human
capital is not. The utilized portion of the stock of human capital is (1 — u)
where u is the unemployment rate. The Cobb-Douglas production function
is Y = fiKa ((1— u)H)' which is re-expressed as Y = AK since we
define B = (1— u)H / K, and A= fie" . Compared to standard endo-
genous growth theory, there is an important difference here. In this model,
A is not a technologically determined constant. Instead, it rises if
structural unemployment is lower, since physical capital has more labour
to work with. Thus, parameter A is affected by the government's un-
employment policy.
As usual, it is assumed that households rent out their physical and
human capital to firms (that are owned by other households). Profit
maximization on the part of firms results in factors being hired to the
point that marginal products just equal rental prices. The expression for
the marginal product of each factor input follows from the Cobb-Douglas
production function: aA = r and (1— a)Y /(1— u)H = w(1— 0), where r, w
and 0 are the rental prices of physical and human capital, and the
employment subsidy rate.

274
 The remaining relationships that are needed to define the model
are the ones that describe how households make their consumption-vs-
saving (investment in capital) decision, how the unemployment rate is
determined, and how the government finances its employment-creating
initiative. We discuss each of these issues in turn.
As in section 12.2, we follow Mankiw (2000) and assume that
there are two groups of households — with each representing one half of
the population. One group is patient and the other is not. The patient
households save as long as the after-tax return on capital exceeds their rate
of impatience, and this saving generates the income that is necessary to
yield a positive percentage growth rate in consumption. This growth in
living standards equals the economy's productivity growth rate. The
simplest version of this outcome is the straightforward proposition that
the productivity growth rate equals the excess of the after-tax interest rate
over the household's rate of impatience: e/ C = r(1— r)— p. The other
condition that follows from household optimization concerning capital
accumulation is that both physical and human capital must generate the
same rate of return per unit, so households are indifferent between holding
their wealth in each of the two forms of capital. This condition is
r = (1 — u)w.
These forward-looking households make two separate decisions.
As a group, each family makes the capital-accumulation decision by
following the consumption-growth relationship that was just discussed.
Following Alexopoulos (2003), we can think of this decision being
executed by the family matriarch, who takes the labour market outcomes
of the various family members as exogenous to her planning problem. She
chooses the optimal capital-accumulation plan, and allocates the
corresponding amount of consumption each period to each family member.
Each family member is free to augment that level of consumption by
adjusting her labour market involvement. The workers at each firm are
assumed to rely on a group representative to negotiate wages with their
employer. The negotiator pursues a wage that exceeds the workers'
outside option, but only to a limited degree since the negotiator values a
high level of employment as well. As explained in section 8.3,
unemployment emerges in this setting. Specifically, we have
u = (a(1— v) / v)(1— 9), where (as already noted) 0 is the employment
subsidy and v is the exponent on employment in the labour negotiator's
Cobb-Douglas objective function. (1 — v) is the weight on wages. Of
particular interest here is that the unemployment rate varies inversely with
the level of the employment subsidy. Clearly, the growth model is not
needed to arrive at this conclusion. What the model does is facilitate an

275
examination of how an employment subsidy affects the growth rate of
living standards (the productivity growth rate).
Some macroeconomists might find the separate-decisions format
for specifying the family unappealing. After all, for some time now, the
goal when providing micro-foundations for macroeconomics has been to
specify one overall optimization that simultaneously yields all behavioural
equations for the agents. But, in the interest of tractability, it is now
common to separate certain decisions. As already noted, in her model of
efficiency wages and endogenous growth, Alexopoulos (2003) adopts
precisely this same separation of the asset-accumulation and labour-
market-involvement household decisions. Similarly, in the New Neo-
classical Synthesis approach to stabilization policy analysis, modellers
specify two separate firms: one to hire the factors and to sell
"intermediate" products, and the other to buy the intermediate products
and to sell final products to households. These "final goods" sellers have
no costs at all, other than the menu costs that are incurred when changing
prices. The two-stage process involving two separate firms is adopted
solely to separate the optimization for gradual price adjustment from the
optimization for determining factor demands. Thus, while it can be
regarded as contrived, this practice of separating certain decisions appears
to be accepted as a necessary way of proceeding in even the most central
areas of study in modern macroeconomics. As noted, we follow this
practice here.
The second group of households is impatient. They have such a
high rate of time preference that they never save — beyond the investment
in human capital that is necessary to have a job. As a result, this group
simply consumes all their income — which is half the after-tax labour
income generated each period, plus a transfer payment that the
government pays to this lower-income group, minus their spending on
acquiring human capital. This group interacts with employers in the same
way as was described in the previous paragraph. Thus, since this group
constitutes half the population, they represent half the unemployed. They
are relatively poor since, by never acquiring any physical capital, they
receive no "interest" income. The spending function for Group-2
households is given by E = R+[w(1— 0(1— u)H — MI 2, where R is
transfer-payment receipts.
Finally, we note the government's balanced-budget constraint. It
is Olv(1— u)H + R = z-rK + zw(1—u)H , a relationship which states that the
income tax revenue pays for general transfers and the employment
subsidy. Balanced growth is assumed, so e/ C= kIK =H / H = n. The
equations determine the responses of n, c, e, T, r, w, u, A and B, when the

276
employment subsidy is introduced (6) is increased). c and e are defined as
C/K and E/K, and it is assumed that the government fixes the transfer-
payments-to-GDP ratio, RIY. For the remainder of this section, we discuss
four properties of this system — that n, c and e all rise, and that u falls — as
9 becomes positive. It is left for the reader to use the equations to verify
the results, if desired.
Introducing the employment subsidy has a direct effect in the
labour market — lower unemployment. As a result, physical capital has
more labour to work with, and this raises physical capital's marginal
product, and so raises the interest rate. Thus, there is an increased
incentive to save. Financing this initiative with a higher income tax rate
shrinks, but does not eliminate, this increased incentive to save. The value
of the formal model is that it allows us to see that between these
competing effects on the after-tax return on saving — the rise in the pre-tax
rate of return and the rise in the tax rate applied to that return — the former
must dominate. Further, the model clarifies that there is no short-term pain
involved (in either the richer or the poorer households having to cut
current consumption) in order to secure this long-term gain (higher
productivity growth). Indeed, there are "good news" outcomes on three
fronts: unemployment falls, the level of consumption rises, and the
ongoing growth rate of living standards rises. We conclude that basic
endogenous-growth analysis can support initiatives designed to reduce
structural unemployment.
As we noted when reporting a similar finding in section 9.3, this
conclusion will not be regarded as too surprising, if one recalls the
Bhagwati/Ramaswami (1963) theorem — a proposition which states that
we have the best chance of improving economic welfare if the attempt to
alleviate a distortion is introduced at the very source of that distortion.
Many prominent economists such as Phelps (1997), Solow (1998),
and Freeman (1999) have advocated employment subsidies. In practical
terms, they call for a major enlarging of the earned income tax credit
policy in the United States. As surprising as it may seem, given the high
profile that is enjoyed by these advocates of employment subsidies, the
investigation of this broad strategy within an endogenous-productivity-
growth setting has not been researched at all extensively. This section of
the chapter has been intended as a partial filling of this gap. Of course,
much sensitivity testing is needed to see if similar results emerge in other
formulations of endogenous growth. Also, it will be instructive to extend
this analysis to an open-economy setting. Van Der Ploeg (1996) and
Turnovsky (20002) provide useful starting points for pursuing this agenda.

277
12.5 The Aging Population and Future Living Standards

There is widespread concern about the living standards that will be

available for the generation that follows the baby-boom cohort. According
to conventional wisdom, when the baby-boomers are old, the much
smaller number of workers in the next generation will face high tax rates
(and consequently lower living standards) if the pay-as-we-go public
pension and public health care programs are to be maintained. In this
section of the chapter, we introduce readers to how our overlapping
generations model can be adapted to make it useful for addressing these
questions.
As noted, the challenge posed by the aging baby boomers is that
the old-age dependency ratio is due to rise noticeably. To analyze this
development within the context of the overlapping generations model that
was introduced in section 4.2, we must introduce a retirement age. Nielsen
(1994) has extended Blanchard's (1985b) model in this way. He shows
that — with a constant probability of death (equal to p) and a constant
overall population (equal to unity) — a retirement age of 2 means that the
proportion of the population that is in retirement must be CA P and the
proportion of the population that is of working age is (1— CA P). As long
as the model is calibrated so that the retirement age (2) is less than each
individual's life expectancy (1/p), each individual will make her life's
plans on the assumption that she will need to finance her consumption
needs during a retirement period. The old-age dependency ratio is
[CA P 1(1—C 2P)]. Within this framework, we can consider a higher old-age
dependency ratio in the model economy by imposing a reduction in the
retirement age.
It is a bit difficult to calibrate this model. For example, p = 0.02
and 2= 40 are reasonable specifications in and of themselves. (Recall that,
in this framework, there is no separate "youth" period, so individuals are
born at working age (say 20 years). p = .02 then implies a life expectancy
of 20 + 50 years.) The unappealing thing is that these parameter values
imply an old-age dependency ratio that is unrealistically large (roughly
equal to the overall dependency ratio). Of course, it should not be
surprising that a model which specifies that life expectancy is independent
of age, and that people begin working at birth, has some difficulty fitting
the facts perfectly. Why, then, do we use such a model? Since the
alternatives have some unappealing features as well.
One alternative is Diamond's (1965) two-period overlapping-
generations model. The problem with this specification is that each period
is the length of one generation, say 35 years. This forces the modeller to
278
specify that it takes 35 years for some output that has been produced but
not consumed to actually be used in the production process as new capital.
For this reason, Barro and Sala-i-Martin (1995) advise against using this
specification. Despite this advice, many analyses are based on this
framework. One accessible example is Scarth and Souare (2002). A
number of studies (for example, Auerbach and Kotlikoff (1987)) have
overcome the limitation of Diamond's two-period specification by
analyzing a multi-period, discrete-time, overlapping-generations model.
The disadvantage in this case is that analytical solutions are impossible, so
complete reliance on numerical simulation is required. Since it is very
difficult for other analysts to undertake sensitivity tests of this work, these
other analysts tend to regard the conclusions as having been generated
from a "black box." As a result, meaningful debate can be limited. It is for
these reasons that there is widespread interest in extending Blanchard's
continuous-time version of an economy with overlapping generations.
One such extension is by Jensen and Nielsen (1993). They allow
some dependency of life expectancy on age, since they specify that the
probability of death is zero for all individuals who are younger than the
exogenous retirement age. Then, there is a constant probability of death
once the retirement age is reached. This specification preserves our ability
to generate analytical solutions, but not quite as easily as in Nielsen
(1994). Other extensions of Blanchard's framework — designed to
examine aging populations and remain analytically tractable — are Faruqee
(2003) and Bettendorf and Heijdra (2006). Another variation is offered by
Gertler (1999). In his model, while individuals are working, they face no
probability of death, but they are subject to a constant probability of being
retired. Then, once retired, there is a constant probability of death. To
keep his model tractable, Gertler must avoid an implication of logarithmic
utility — that the degree of income risk aversion and the preference for
inter-temporal substitution is pinned down by the same parameter. As a
result, he assumes a more complicated preference function, and this
allows him to aggregate across the different age cohorts without having to
assume that wages are constant — and with only one additional state
variable entering the dynamic analysis. Nevertheless, since this
framework is much messier than Nielsen's, We rely on the latter for the
remainder of this section of the chapter.
Recall from section 4.2, that (ignoring taxes) the aggregate
consumption function in the perpetual-youth version of the overlapping-
generations model is specified by the following aggregate relationships:

C = (p + p)(K + H)
k=rK+w—C
279
 H = (r + p)H — w

These relationships can be combined to yield the standard aggregate

consumption function:

C = (r — p)C — p(p + p)K.

The Ramsey (1928) specification is nested within this model; it is the

special case that exists when p = 0. Nielson's contribution is to add life-
cycle features to this model, so that Blanchard's model becomes a special
case as well. In Nielsen's set-up, the first two equations are the same as in
Blanchard' s :

C = (p + p)(K + H)
k=rK+w—C

It is the accumulation identity for human capital that is altered by the

existence of retirement. By adding up over cohorts of all ages, and over
the remaining years in each individual's life, this accumulation identity
becomes:

H=(r+p)H—w(1— (w 1 r)[perA P (1— C h. )].

The second term on the right-hand side is smaller than previously, since
only a portion of the population is working (and therefore receiving wage
income). The (new) third term on the right-hand side accounts for the fact
that each individual receives the wage for only a portion of her remaining
life. For a detailed derivation of this aggregate human capital
accumulation identity, the reader must consult Nielsen (1994). Evaluation
of the double integral (across cohorts of different ages and across the
remaining years in any one person's life) is only possible if wages are
constant. Perfect international capital mobility and constant-returns-to-
scale technology are sufficient assumptions to deliver this independence
of wages from changes in demography. As a result, it is prudent to limit
our application of this particular model to small open economies. In any
event, when the three relationships are combined, we have the revised
aggregate consumption function:

C = (r — p)C — p(p + p)(K — SI)

SZ = (w/r)(1— Ar )(e- P ) 1(1— C aP).
280
As can be seen, the aggregate consumption function is altered when a
portion of the population is retired. This is because retirement lowers the
aggregate stock of human capital (since human capital is embodied within
workers), and it is also because people have an incentive to save more
when they have to plan for a period of lower income later in life. As
asserted above, this more general consumption function nests the earlier
models. For example, when retirement is eliminated (when parameter X
goes to infinity), Nielsen's consumption function reduces to Blanchard's
overlapping generations relationship. Then, when the probability of death
falls to zero (p = 0), Blanchard's model reduces to Ramsey's (1928)
analysis of the infinitely lived representative agent. The most general of
these three frameworks is needed to consider rising old-age dependency.
As explained in Scarth (2004), a calibrated version of this model
can be applied to Canada. The results indicate that Canadians need the
government debt-to-GDP ratio to fall to about 20% by 2020 if material
living standards are to be increased by about the same amount that the
aging population can be expected to (other things equal) lower living
standards. Of course, not all dimensions of the aging population are
captured in this model. For example, as Van Groezen et at (2005) have
argued, since the elderly consume a higher proportion of services, and
since it is more difficult to have productivity increases in the service
sector, aging can be expected to bring lower growth, and this mechanism
is not included in this model. Despite this fact, the Canadian federal
government has accepted the 20% value as its target for the debt ratio, and
officials have stated that this target is based on the desire to insulate living
standards from the coming demographic shock. So it certainly is the case
that policy makers do pay attention to the models that we have examined
in this book.

12.6. Conclusions

How can our tax system be designed to promote both an increase in

fairness (as we compare rich and poor today) and an increase in average
living standards over the years to come? How can our structural
unemployment problems be addressed without jeopardizing our desire to
have more rapidly rising material living standards? What is needed to
limit the threat to living standards that is posed by the aging population?
These questions are among the most disputed topics in public policy
analysis today. It is the job of the policy-oriented economist to use the
analytical structure of our discipline to help inform policy makers on these
matters. It is important to identify both the trade-offs and the "free
281
lunches" that are possible, when confronting these issues. This chapter has
been designed to help readers meet this challenge, by explaining how
basic growth theory can be directly applied to help us understand these
topical questions.

282
References
Abel, A., N.G. Mankiw, L.H. Summers and R.J. Zeckhauser (1989),
"Assessing Dynamic Efficiency: Theory and Evidence," Review
of Economic Studies 1-20.
Adelman, I. and F.L. Adelman (1959), "The Dynamic Properties of the
Klein-Goldberger Model," Econometrica 27, 596-625.
Aghion, P. and P. Howitt (1992), "A Model of Growth through Creative
Destruction," Econometrica 60, 323-351.
Aghion, P. and P. Howitt (1998), Endogenous Growth Theory
(Cambridge, Mass.: MIT Press).
Aghion, P., E. Caroli and C. Garcia-Penalosa (1999), "Inequality and
Economic Growth: The Perspective of the New Growth
Theories," Journal of Economic Literature 37, 1615-1660.
Akerlof, G.A. and J. Yellen (1985), "A Near Rational Model of the
Business Cycle, with Wage and Price Inertia," Quarterly Journal
of Economics 100, 823-838.
Alexopoulos, M. (2003), "Growth and Unemployment in a Shirking
Efficiency Wage Model," Canadian Journal of Economics 36,
728-746.
Alexopoulos, M. (2004), "Unemployment and the Business Cycle,"
Journal of Monetary Economics 51, 277-298.
Alvi, E. (1993), "Near-Rationality, Menu Costs, Strategic Comple-
mentarity, and Real Rigidity: An Integration," Journal of
Macroeconomics 15, 619-625.
Amato, J.D., and T. Laubach (2003), "Rule-of-Thumb Behaviour and
Monetary Policy," European Economic Review 47, 791-831.
Ambler, S. and L. Phaneuf (1992), "Wage Contracts and Business Cycle
Models," European Economic Review 36, 783-800.
Andres, J., J. D. Lopez-Salido, and E. Nelson (2005), "Sticky-Price
Models and the Natural Rate Hypothesis," Journal of Monetary
Economics 52, 1025-1053.
Auerbach, A. and L. Kotlikoff (1987), Dynamic Fiscal Policy
(Cambridge: Cambridge University Press).
Ball, L. (1994), "What Determines the Sacrifice Ratio?" in N. G. Mankiw
(ed.) Monetary Policy (Chicago: University of Chicago Press).

283
Ball, L. (1995), "Disinflation with Imperfect Credibility," Journal of
Monetary Economics 35, 5-23.
Ball, L., D.W. Elmendorf, and N.G. Mankiw (1995), "The Deficit
Gamble," Journal of Money, Credit and Banking 30, 699-720.
Ball, L. and N G Mankiw (1994), "A Sticky-Price Manifesto," Carnegie-
Rochester Conference Series on Public Policy 41, 127-151.
Ball, L., N.G. Mankiw, and R. Reis (2005), "Monetary Policy for
Inattentive Economies," Journal of Monetary Economics 52, 703-
725.
Barbie, M., M. Hagedorn, and A. Kaul (2004), "Assessing Aggregate
Tests of Efficiency for Dynamic Economies," Topics in
Macroeconomics 4.
Barlevy, G. (2004), "The Cost of Business Cycles Under Endogenous
Growth," American Economic Review 94, 964-991.
Barro, R.J. (1977), "Unanticipated Money Growth and Unemployment in
the United States," American Economic Review 67, 101-115.
Barro, R.J. and D.B. Gordon (1983), "Rules, Discretion and Reputation in
a Model of Monetary Policy," Journal of Monetary Economics
12, 101-122.
Barro, R.J. and H.I. Grossman (1971), "A General Disequilibrium Model
of Income and Employment," American Economic Review 61, 82-
93.
Barro, R.J. and X. Sala-i-Martin (1995), Economic Growth (New York:
McGraw-Hill).
Basu, S., M.S. Kimball, N.G. Mankiw and D.N. Weil (1990), "Optimal
Advice for Monetary Policy," Journal of Money, Credit and
Banking 22, 19-36.
Benhabib, J., R. Rogerson and R. Wright (1991), "Homework in Macro-
economics: Household Production and Aggregate Fluctuations,"
Journal of Political Economy 99, 1166-1187.
Bernanke, B.S. and V.R. Reinhart (2004), "Conducting Monetary Policy
at Very Low Short-Term Interest Rates," American Economic
Review Papers and Proceedings 94, 85-90.
Bettendorf, L.J.H. and B.J. Heijdra (2006), "Population Ageing and
Pension Reform in a Small Open Economy with Non-Traded
Goods," Journal of Economic Dynamics and Control 30, 2389-
2424.
Bhagwati, J. and V, Ramaswami (1963), "Domestic Distortions, Tariffs,
and the Theory of Optimum Subsidy," Journal of Political
Economy 71, 44-50.
Blackburn, K. and A. Pelloni (2005), "Growth, Cycles, and Stabilization
Policy," Oxford Economic Papers 57, 262-282.
284
Blanchard, O.J. (1981), "Output, the Stock Market, and Interest Rates,"
American Economic Review 71, 132-143.
Blanchard, O.J. (1985a), "Credibility, Disinflation and Gradualism,"
Economics Letters 17, 211-217.
Blanchard, O.J. (1985b), "Debt, Deficits and Finite Horizons," Journal of
Political Economy 93, 223-247.
Blanchard, O.J., and N. Kiyotaki (1987), "Monopolistic Competition and
the Effects of Aggregate Demand," American Economic Review
77, 647-666.
Blanchard, O.J. and L.H. Summers (1986), "Hysteresis and the European
Unemployment Problem," NBER Macroeconomics Annual 1986
1, 15-78.
Blanchard, 0. and P. Weil (2003), "Dynamic Efficiency, the Riskless
Rate, and Debt Ponzi Games under Uncertainty," Advances in
Macroeconomics 3.
Blinder, A.S. (1973),"Can Income Taxes Be Inflationary? An Expository
Note," National Tax Journal 26, 295-301.
Blinder, A.S. (1981), "Inventories and the Structure of Macro Models,"
American Economic Review Papers and Proceedings 71,11-16.
Blinder, A.S. (1987), "Keynes, Lucas, and Scientific Progress," American
Economic Review Papers and Proceedings 77, 130-135.
Bohn, H. (1995), "The Sustainability of Budget Deficits in a Stochastic
Economy," Journal of Money, Credit and Banking 27, 257-271.
Bouakez, H., E. Cardia, and F.J. Ruge-Murcia (2005), "Habit Formation
and the Persistence of Monetary Shocks," Journal of Monetary
Economics 52, 1073-1088.
Bouakez, H. and T. Kano (2006), "Learning-by-doing or Habit
Formation?" Review of Economic Dynamics 9, 508-524.
Brainard, W. (1967), "Uncertainty and the Effectiveness of Policy,"
American Economic Review 57, 411-425.
Brock, W.A., S.N. Durlauf, and K.D. West (2003), "Policy Evaluation in
Uncertain Economic Environments," Brookings Papers on
Economic Activity, 235-322.
Bryant, J. (1991), "A Simple Rational-Expectations Keynes-Type Model,"
in N.G. Mankiw and D. Romer (eds.), New Keynesian Economics
Volume 2 (Cambridge, Mass.: MIT Press).
Buiter, W.H. and Miller, M. (1982), "Real Exchange Rate Overshooting
and the Output Cost of Bringing Down Inflation," European
Economic Review 18, 85-123.
Burbidge, J.B. (1984), "Government Debt: Reply," American Economic
Review 74, 766-767.

285
Cagan, P. (1956), "The Monetary Dynamics of Hyperinflation," in M.
Friedman, ed., Studies in the Quantity Theory of Money (Chicago:
University of Chicago Press).
Calvo, G.A. (1983), "Staggered Prices in a Utility-Maximizing
Framework," Journal of Monetary Economics 12, 383-398.
Calvo, G.A. and F.S. Mishkin (2003), "The Mirage of Exchange Rate
Regimes for Emerging Market Countries," The Journal of
Economic Perspectives 17, 99-118.
Calvo, G.A. and M. Obstfeld (1988), "Optimal Time-Consistent Fiscal
Policy with Finite Lifetimes," Econometrica 56, 411-432.
Campbell, J., and N.G. Mankiw (1987), "Are Output Fluctuations
Transitory?" Quarterly Journal of Economics 52, 857-880.
Carroll, C.D. (2001), "Death to the Log-Linearized Consumption Euler
Equation! (And Very Poor Health to the Second-Order
Approximation)," Advances in Macroeconomics 1.
Chiang, A.C. (1984), Fundamental Methods of Mathematical Economics
Third edition (New York: McGraw-Hill).
Christiano, L. and M. Eichenbaum (1992), "Current Real-Business-Cycle
Theories and Aggregate Labor-Market Fluctuations," American
Economic Review 82, 430-450.
Christiano, L., M. Eichenbaum, and C. Evans (2005), "Nominal Rigidities
and the Dynamic Effects of a Shock to Monetary Policy," Journal
of Political Economy 113, 1-45.
Coenen, G., A. Orphanides, and V. Wieland (2006), "Price Stability and
Monetary Policy Effectiveness when Nominal Interest Rates are
Bounded at Zero," Advances in Macroeconomics 6.
Cohen, A.J. and G.C. Harcourt (2003), "Retrospectives: Whatever
happened to the Cambridge Capital Theory Controversies?"
Journal of Economic Perspectives 17, 199-214.
Cover, J.P. and P. Pecorino (2004), "Optimal Monetary Policy and the
Correlation between Prices and Output," Contributions to
Macroeconomics 4.
Cooley, T.F. and G D Hansen (1991), "The Welfare Costs of Moderate
Inflation," Journal of Money, Credit and Banking 23, 483-503.
Cooper, R., and A. John (1988), "Coordinating Coordination Failures in
Keynesian Models," Quarterly Journal of Economics 103, 441-
464.
Cooper R., and A. Johri (2002), "Learning By Doing and Aggregate
Fluctuations," Journal of Monetary Economics 49, 1539-1566.
Das, S.P. and C. Ghate (2005), "Endogenous Distribution, Politics, and
the Growth-Equity Trade-off," Contributions to Macroeconomics
5.
286
Devereux, M., A.C. Head and B.J. Lapham (1993), "Monopolistic
Competition, Technology Shocks, and Aggregate Fluctuations,"
Economics Letters 41, 57-61.
Devereux,M.B., P.R. Lane, and J. Xu (2006), "Exchange Rates and
Monetary Policy in Emerging market Economics," The Economic
Journal 116, 478-506.
Diamond, P.A. (1965), "National Debt in a Neoclassical Growth Model,"
American Economic Review 55, 1126-1150.
Diamond, P.A. (1984), A Search-Equilibrium Approach to the Micro
Foundations of Macroeconomics (Cambridge, Mass. and London,
England: MIT Press).
Dixon, H.D. and E. Kara (2006), "How to Compare Taylor and Calvo
Contracts: A Comment on Michael Kiley," Journal of Money,
Credit, and Banking 38, 1119-1126.
Domeij, D. (2005), "Optimal Capital Taxation and Labour Market
Search," Review of Economic Dynamics 8, 623-650.
Dornbusch, R. (1976), "Expectations and Exchange Rate Dynamics,"
Journal of Political Economy 84, 1161-1176.
Drazen, A. and P.R. Masson (1994), "Credibility of Policies Versus
Credibility of Policymakers," Quarterly Journal of Economics 90,
735-754.
Driskill, R. (2006), "Multiple Equilibria in Dynamic Rational
Expectations Models: A Critical Review," European Economic
Review 50, 171-210.
Edwards, S. and Y.E. Levy (2005), "Flexible Exchange Rates as Shock
Absorbers," European Economic Review 49, 2079-2105.
Eggertsson, G.B. and M. Woodford (2004), "Policy Options in a Liquidity
Trap," American Economic Review Papers and Proceedings 94,
76-79.
Easterly, W. (2005), "National Policies and Economic Growth," in P.
Aghion and S. Durlauf (eds.), Handbook of Economic Growth
(Amsterdam: Elsevier).
Estralla, A. and J. Fuhrer (2002), "Dynamic Inconsistencies: Counter-
factual Implications of a Class of Rational Expectations Models,"
American Economic Review 92, 1013-1028.
Farmer, R.E.A. (1993) The Macroeconomics of Self-Fulfilling Prophecies
(Cambridge, Mass.: MIT Press).
Faruqee, H. (2003), "Debt, Deficits, and Age-Specific Mortality," Review
of Economic Dynamics 6, 300-312.
Fatas, A. (2000), "Do Business Cycles Cast Long Shadows? Short-Run
Persistence and Economic Growth," Journal of Economic Growth
5, 147-162.
287
 Fatas , A. and I. Mihov (2003), "The Case for Restricting Fiscal Policy
Discretion, " Quarterly Journal of Economics 118, 1419-1447.
Favero, C.A. and F. Milani (2005), "Parameter Instability, Model
Uncertainty and the Choice of Monetary Policy," Topics in
Macroeconomics 5.
Fillion, J-F. (1996), "L'endettement du Canada et ses effets sur les taux
d'interet reels de long terme," Bank of Canada Working Paper 96-
14.
Fischer, S. (1980), "Dynamic Inconsistency, Cooperation, and the
Benevolent Dissembling Government," Journal of Economic
Dynamics and Control 2, 93-107.
Fischer, S. (1995), "Central-Bank Independence Revisited," The American
Economic Review Proceedings 85, 201-206.
Fischer, S. and L. Summers (1989), "Should Nations Learn to Live with
Inflation?" American Economic Review Papers and Proceedings
79, 382-387.
Fleming, J.M. (1962), "Domestic Financial Policies Under Fixed and
Floating Exchange Rates," International Monetary Fund Staff
Papers 9, 369-379.
Frank, R.H. (2005), "Positional Externalities Cause Large and Preventable
Welfare Losses," American Economic Review Papers and
Proceedings 95, 137-141.
Freeman, R. (1999), The New Inequality: Creating Solutions for Poor
America (Boston: Beacon Press).
Friedman, M. (1948), "A Monetary and Fiscal Framework for Economic
Stability," The American Economic Review 38, 245-264.
Friedman, M. (1953), "The Case for Flexible Exchange Rates," in Essays
-,0,0//C, (Chicago: University of Chicago Press).
Friedman, M. (1957), A Theory of the Consumption Function (Princeton,
N.J.: National Bureau of Economic Research and Princeton
University Press).
Friedman, M. (1959), A Program for Monetary Stability (New York:
Fordham University Press).
Frydman, R. and E. Phelps, (1983), Individual Forecasting and Aggregate
Outcomes: Rational Expectations Examined (Cambridge:
Cambridge University Press).
Gali, J. and M. Gertler (1999), "Inflation Dynamics: A Structural
Econometric Analysis," Journal of Monetary Economics 44 195-
222.
Gali, J., M. Gertler, and J.D. Lopez-Salido (2005), "Robustness of the
Estimates of the Hybrid New Keynesian Philips Curve," Journal
of Monetary Economics 52, 1107-1118.

288
Galor, 0. and J. Zeira (1993), "Income Distribution and Macro-
economics," Review of Economics Studies 60, 35-52.
Garcia-Penalosa, C. and S.J. Turnovsky (2005), "Second-best Optimal
Taxation of Capital and Labour in a Developing Economy,"
Journal of Public Economics 89, 1045-1074.
Gertler, M, "Government Debt and Social Security in a Life-Cycle
Economy," Carnegie-Rochester Series on Public Policy 50, 61-
110.
Geweke, J. (1985), "Macroeconometric Modeling and the Theory of the
Representative Agent," The American Economic Review
Proceedings 75, 206-210.
Giannitsarou, C. (2005), "E-stability Does Not Imply Learnability,"
Macroeconomic Dynamics 9, 276-287.
Goodfriend, M.and R.G. King (1997), "The New Neoclassical Synthesis
and the Role of Monetary Policy," Macroeconomics Annual 12,
231-283.
Goodfriend, M. (2004), "Monetary Policy in the New Neoclassical
Synthesis: A Primer," Federal Reserve Bank of Richmond
Economic Quarterly 90, 21-45.
Gorbet, F. and J. Helliwell (1971), "Assessing the Dynamic Efficiency of
Automatic Stabilizers," Journal of Political Economy 79, 826-
845.
Graham, B.S. and J.R.W. Temple (2006), "Rich Nations, Poor Nations:
How Much Can Multiple Equilibria Explain?" Journal of
Economic Growth 11, 5-41.
Gravelle, J.G. (1991), "Income, Consumption, and Wage Taxation in a
Life-Cycle Model: Separating Efficiency from Distribution,"
American Economic Review 67, 985-995.
Hahn, F.H. and R.M. Solow (1986), "Is Wage Flexibility a Good Thing?"
in W. Beckerman (ed.), Wage Rigidity and Unemployment
(Baltimore: The Johns Hopkins University Press).
Hall, R.E. (2003), "Modern Theory of Unemployment Fluctuations:
Empirics and Policy Applications," American Economic Review
Papers and Proceedings 93, 145-150.
Hansen, G.D. (1985), "Indivisible Labor and the Business Cycle," Journal
of Monetary Economics 16, 309-327.
Hansen, G.D., and R. Wright (1992), "The Labor Market in Real Business
Cycle Theory," Federal Reserve Bank of Minneapolis Quarterly
Review 16, 2-12.
Hartley, J.E. (1997), The Representative Agent in Macroeconomics (New
York: Routledge).

289
Helpman, E. (2004), The Mystery of Economic Growth (Cambridge,
Mass.: Harvard University Press).
Honkapohja, S. and K. Mitra (2004), "Are Non-fundamental Equilibria
Learnable in Models of Monetary Policy?" Journal of Monetary
Economics 51, 1743-1770.
Howitt, P. (1978), "The Limits to Stability of Full Employment
Equilibrium," Scandinavian Journal of Economics 80, 265-282.
Howitt, P. (1985), "Transaction Costs in the Theory of Unemployment,"
American Economic Review 75, 88-100.
Howitt, P. (1986), "The Keynesian Recovery," Canadian Journal of
Economics 19, 626-641.
Howitt, P. and R.P. McAfee (1992), "Animal Spirits," The American
Economic Review 82, 493-507.
Huang, K.X.D., L. Zheng, and L. Phaneuf (2004), "Why Does the
Cyclical Behavior of Real Wages Change Over Time," The
American Economic Review 94, 836-857.
Huang, K.X.D. and L. Zheng (2005), "Inflation Targeting: What Inflation
Rate to Target?" Journal of Monetary Economics 52, 1435-1462.
Ireland, P.N. (2003), "Endogenous Money or Sticky Prices?" Journal of
Monetary Economics 50, 1623-1648.
Jackson, A.L. (2005), "Disinflationary Boom Reversion," Macroeconomic
Dynamics 9, 489-575.
Jensen, H. (2002), "Targeting Nominal Income Growth or Inflation,"
American Economic Review 94, 928-956.
Jensen, S. and S. Nielsen (1993), "Aging, Intergenerational Distribution
and Public Pension Systems," Public Finance 48, 29-42.
Johnson, L. (1977), "Keynesian Dynamics and Growth," Journal of
Money, Credit and Banking 9, 328-340.
Jones, C.I. (1995), "Time Series Tests of Endogenous Growth Models,"
Quarterly Journal of Economics 91, 495-525.
Jones, C.I. (2002), Introduction to Economic Growth, Second Edition
(New York: W.W. Norton).
Jones, C.I. (2003), "Population and Ideas: A Theory of Endogenous
Growth," in P. Aghion, R. Frydman, J. Stiglitz and M. Woodford
(eds.) Knowledge, Information, and Expectations in Modern
Macroeconomics (Princeton: Princeton University Press.
Judd, J.P. and B. Trehan, "The Cyclical Behavior of Prices: Interpreting
the Evidence," Journal of Money, Credit and Banking 27, 789-
797.
Judd, K.L. (2002), "Capital Income Taxation with Imperfect Compe-
tition," American Economic Review Papers and Proceedings 92,
417-421.
290
Keynes, J. M. (1936), The General Theory of Employment, Interest and
Money (London: Macmillan).
Kim, J. and D.W. Henderson (2005), "Inflation Targeting and Nominal-
income-growth Targeting: When and Why Are They Sub-
optimal?" Journal of Monetary Economics 52, 1463-1495.
King, R.G. (1993), "Will the New Keynesian Macroeconomics Resurrect
the IS-LM Model? Journal of Economic Perspectives 7, 67-82.
King, R.G. (2000), "The New IS-LM Model: Language, Logic, and
Limits," Federal Reserve Bank of Richmond Economic Quarterly
86, 45-103.
King, R.G. (2006), "Discretionary Policy and Multiple Equilibria,"
Economic Quarterly, Federal Reserve Bank of Richmond 92, 1-
15.
Kirman, A.P. (1992), "Whom or What Does the Representative Individual
Represent?" Journal of Economic Perspectives 6, 117-136.
Kirsanova, T., C. Leith, and S. Wren-Lewis (2006), "Should Central
Banks Target Consumer Prices or the Exchange Rate?" The
Economic Journal 116, F208-F231.
Knight, F.H. (1921), Risk, Uncertainly and Profit (London: London
School of Economics).
Koskela, E. and R. Schob (2002), "Why Governments Should Tax Mobile
Capital in the Presence of Unemployment," Contributions to
Economic Analysis and Policy 1.
Kremer, M. (1993a), "The 0-Ring Theory of Economic Development,"
Quarterly Journal of Economics 108, 551-575.
Kremer, M. (1993b), "Population Growth and Technical Change: One
Million B.C. to 1990," Quarterly Journal of Economics 108, 681-
716.
Kydland, F.E. and Prescott, E.C. (1977), "Rules Rather than Discretion:
The Inconsistency of Optimal Plans," Journal of Political
Economy 85, 473-492.
Kydland, F.E. and Prescott, E.C. (1982), "Time to Build and Aggregate
Fluctuations," Econometrica 50, 1345-1370.
Lam, J-P. and W. Scarth (2006), "Balanced Budgets vs. Keynesian Built-
In Stabilizers," mimeo.
Lipsey, R.G. (1960), "The Relationaship Between Unemployment and the
Rate of Change of Money Wage Rates in the U.K. 1862-1957: A
Further Analysis," Economica 28, 1-31.
Lipsey, R and K. Lancaster (1956), "The General Theory of the Second
Best," Review of Economic Studies 24, 11-32.
Loh, S. (2002), "A Cold-Turkey versus a Gradualist Approach in a Menu
Cost Model," Topics in Macroeconomics 2.
291
Lucas, R.E.,Jr. (1976), "Econometric Policy Evaluations: A Critique," in
K. Brunner and A.H. Meltzer, eds., The Phillips Curve and the
Labor Market (Amsterdam, New York, and Oxford: North
Holland).
Lucas, R.E.,Jr. (1987), Models of Business Cycles (Oxford: Blackwell).
Lucas, R.E., Jr. (1988), "On the Mechanics of Development Planning,"
Journal of Monetary Economics 22, 3-42.
Lucas, R.E.,Jr. and T.J. Sargent (1979), "After Keynesian
Macroeconomics," Federal Reserve Bank of Minneapolis
Quarterly Review 3, 1-16.
Malik, H.A. (2004), Four Essays on the Theory of Monetary Policy, PhD
Dissertation in Economics, McMaster University.
Malik, H.A. and W.M. Scarth (2006), "Is Price Flexibility De-stabilizing?
A Reconsideration," mimeo.
Malinvaud, E. (1977), The Theory of Unemployment Reconsidered
(Oxford: Blackwell).
Majundar, S. and S.W. Mukand (2004), "Policy Gambles," American
Economic Review 94, 1207-1223.
Mankiw, N.G. (1985), "Small Menu Costs and Large Business Cycles: A
Macroeconomic Model of Monopoly," Quarterly Journal of
Economics 100, 529-537
Mankiw, N.G. (1992), "The Reincarnation of Keynesian Economics,"
European Economic Review 36, 559-565.
Mankiw, N.G. (2000), "The Savers-Spenders Theory of Fiscal Policy,"
American Economic Review Papers and Proceedings 90, 120-
125 .
Mankiw, N.G., D. Romer and D.N. Weil (1992), "A Contribution to the
Empirics of Economic Growth," Quarterly Journal of Economics
88, 407-437.
Mankiw, N.G. and R. Reis (2002), "Sticky Information Versus Sticky
Prices: A Proposal to Replace the New Keynesian Phillips
Curve," Quarterly Journal of Economics 17, 1295-1328.
Mankiw, N.G. and M. Weinzierl (2006), "Dynamic Scoring: A Back-of-
the-envelop Guide," Journal of Public Economics 90, 1415-1433.
Manning, A. (1990), "Imperfect Competition, Multiple Equilibria, and
Unemployment Policy," Economic Journal Supplement 100, 151-
162.
Mansoorian, A. and M. Mohsin (2004), "Monetary Policy in a Cash-in-
Advance Economy: Employment, Capital Accumulation, and the
Term Structure of Interest Rates," Canadian Journal of
Economics 37, 336-352.

292
Marrero, G.A. and A. Novales (2005), "Growth and Welfare: Distorting
versus Non-Distorting Taxes," Journal of Macroeconomics 27,
403-433.
McCallum, B.T. (1980), "Rational Expectations and Macroeconomic
Stabilization Policy: An Overview," Journal of Money, Credit
and Banking 12, 716-746.
McCallum, B.T. (1983). "On Non-Uniqueness in Rational Expectations
Models: An Attempt at Perspective," Journal of Monetary
Economics 11, 139-168.
McCallum, B.T. (1984), "Are Bond-Financed Deficits Inflationary? A
Ricardian Analysis," Journal of Political Economy 92, 123-135.
McCallum, B.T. (1986), "On 'Real' and 'Sticky-Price' Theories of the
Business Cycle," Journal of Money, Credit and Banking 18, 397-
414.
McCallum, B.T. (1995), "Two Fallacies Concerning Central-Bank
Independence," American Economic Review Proceedings 85, 207-
211.
McCallum, B.T. (2001), "Monetary Policy Analysis in Models Without
Money," Federal Reserve Bank of St. Louis Review 83, 145-160.
McCallum, B.T. (2003), "Multiple-Solution Indeterminacies in Monetary
Policy Analysis," Journal of Monetary Economics 50, 1153-1175.
McCallum, B. and E. Nelsen (1999), "An Optimizing IS-LM Specification
for Monetary Policy and Business Cycle Analysis," Journal of
Money, Credit and Banking 27, 296-316.
McDonald, I. and R.M. Solow (1981), "Wage Bargaining and
Employment," The American Economic Review 71, 896-908.
McGrattan, E.R. (1994), "A Progress Report on Business Cycle Models,"
Federal Reserve Bank of Minneapolis Quarterly Review 18, 2-16.
Mehra, Y.P. (2004), "The Output Gap, Expected Future Inflation and
Inflation Dynamics: Another Look," Topics in Macroeconomics
4.
Metzler, L.A. (1941), "The Nature and Stability of Inventory Cycles,"
Review of Economics and Statistics 23, 113-129.
Milborne, R.D. D.D. Purvis and D. Scoones (1991), "Unemployment
Insurance and Unemployment Dynamics," Canadian Journal of
Economics 24, 804-826.
Minford, P. and D. Peel (2002), Advanced Macroeconomics: A Primer
(Northampton: Edward Elgar).
Moutos, T., and W. Scarth (2004), "Some Macroeconomic Consequences
of Basic Income and Employment Subsidies," in J. Agell, M.
Keen and A. Weichenrieder (eds.) Labor Market Institutions and
Public Regulation (Cambridge, Mass.: MIT Press).
293
Mullineux, A. and W. Peng (1993), "Nonlinear Business Cycle
Modelling," Journal of Economic Surveys 7, 41-83.
Mundell, R.A. (1963), "Capital Mobility and Stabilization Policy under
Fixed and Flexible Exchange Rates," Canadian Journal of
Economics and Political Science 29, 475-485.
Mussa, M. (1981), "Sticky Prices and Disequilibrium Adjustment in a
Rational Model of the Inflationary Process," American Economic
Review 71, 1020-1027.
Nelson, E. (2003), "The Future of Monetary Aggregates in Monetary
Policy Analysis," Journal of Monetary Economics 50, 1029-1059.
Nielsen, S. (1994), "Social Security and Foreign Indebtedness in a Small
Open Economy," Open Economies Review 5, 47-63.
Nordhous, W.D. (1992), "Letharo Model 2: The Limits to Growth
Revisited," Brookings Papers on Economic Activity 2, 1-43.
Oh, S. and M. Waldman (1994), "Strategic Complementarity Slows
Macroeconomic Adjustment to Temporary Shocks," Economic
Inquiry 32, 318-329.
Okun, A.M. (1962), "Potential GNP: Its Measurement and Significance,"
Proceedings of the Business and Economic Statistics Section of
the American Statistical Association, 98-104.
Okun, A.M. (1975), Equity and Efficiency: The Big Trade-Off
(Washington: The Brookings Institution).
Parkin, M. (2000), "What Have We Learned About Price Stability?" in
Price Stability and The Long-Run Target for Monetary Policy
Bank of Canada Conference Proceedings.
Pemberton, J. (1995), "Trends Versus Cycles: Asymmetric Preferences
and Heterogeneous Individual Responses," Journal of
Macroeconomics 17, 241-256.
Persson, T. and G. Tabellini (1994), Monetary and Fiscal Policy
(Cambridge, Mass.: MIT Press).
Pesaran, M.H. (1982), "A Critique of the Proposed Tests of the Natural
Rate-Rational Expectations Hypothesis," Economic Journal 92,
529-554.
Pesaran, M.H. (1987), The Limits to Rational Expectations (Oxford:
Blackwell).
Phelps, E. (1997), Rewarding Work (Boston: Harvard University Press).
Pissarides, C.A. (1998), "The Impact of Employment Tax Cuts on
Unemployment and Wages; The Role of Unemployment Benefits
and Tax Structure," European Economic Review 42, 155-183.
Ramsey, F.P. (1928), "A Mathematical Theory of Saving," Economic
Journal 38, 543-559.

294
Ravenna, F. and C.E. Walsh (2006), "Optimal Monetary Policy with the
Cost Channel." Journal of Monetary Economics 53, 199-216.
Rebelo, S. (2005), "Real Business Cycle Models: Past, Present and
Future," The Scandinavian Journal of Economics 107, 217-238.
Roberts, J.M. (2006), "How Well Does the New Keynesian Sticky-Price
Model Fit the Data?" Contributions to Macroeconomics 6.
Rogerson, R.D. (1988), "Indivisible Labor, Lotteries and Equilibrium,"
Journal of Monetary Economics 21, 3-16.
Romer, C.D. and D. Romer (1989), "Does Monetary Policy Matter? A
New Test in t Spirit of Friedman and Schwartz," NBER
Macroeconorr,,- s Annual 4, 121-170.
Romer, C.D. and D 71-1 ?,,orner (2004), "New Measure of Monetary
Shocks: 1/...---7- ,ackum and Implications," American Economic
Review 94. .
Romer, D. (1993). 7-me Keynesian Synthesis," Journal of Economic
Perspective$ . 5-21
Romer, D. (2001). A—r,agved Macroeconomics, Second Edition (New
York: MeGrro
Romer , P. (1990). "Frog -venous Technical Change," Journal of Political
Economy 91-1, . S - :-SIO2.
Rosser, J.B., Jr. « i $. "Chaos Theory and the New Keynesian
Economies.' :-P)e Manchester School 58, 265-291.
Rowe, N. (1987), -Ar. Extreme Keynesian Macro-economic Model with
Formal Mic7.-oeconomic Foundations," Canadian Journal of
Economics 2(..1. 306 -320.
Rudd, J. and K. v‘Ielan (2005), "New Tests of the New-Keynesian
Phillips Cur. e: Journal of Monetary Economics 52, 1167-1181.
Rudd, J. and K. Whelan (2006), "Can Rational Expectation Sticky-Price
Models Explain Inflation Dynamics?" American Economic
Review 96. 303.
Rudebusch, G.D. (2005), "Assessing the Lucas Critique in Monetary
Policy Models," Journal of Money, Credit, and Banking 37, 245-
272.
Saint-Paul, G. (2006), "Distribution and Growth in an Economy with
Limited Needs: Variable Markups and the End of Work," The
Economic Journal 116, 382-407.
Sargent, T.J. (1978), "Rational Expectations, Econometric Exogeneity,
and Consumption," Journal of Political Economy 86, 673-700.
Sargent, T.J. (1982), "Beyond Demand and Supply Curves in Macro-
economics," American Economic Review Papers and Proceedings
72, 382- 389.

295
Sargent, T.J. and N. Wallace (1981), "Some Unpleasant Monetarist
Arithmetic," Federal Reserve Bank of Minneapolis Quarterl;
Review 5, 1-17.
Scarth, W. (2004), "What Should We Do About the Debt?" in C. Ragan
and W. Watson (eds.) Is The Debt War Over? (Montreal: Institute
for Research on Public Policy.
Scarth, W. (2005), "Fiscal Policy Can Raise Both Employment and
Productivity," International Productivity Monitor 11, 39-46.
Scarth, W. and M. Souare (2002), "Baby-Boom Aging and Average
Living Standards," SEDAP Research Paper No268.
Shaikh, A. (1974), "Laws of Production and Laws of Algebra: The
Humbug Production Function," Review of Economics and
Statistics 56, 115-120.
Smets, F. and R. Wouters (2003), "An Estimated Stochastic Dynamic
General Equilibrium Model of the Euro Area," Journal of the
European Economic Association 1, 1123-1175.
Smyth, D.J. (1974), "Built-in Flexibility of Taxation and Stability in a
Simple Dynamic IS-LM Model," Public Finance 29, 111-113.
Soderstrom, U. (2002), "Targeting Inflation with a Role for Money."
Contributions to Macroeconomics 2.
Solon, G., R. Barsky and J.A. Parker (1994), "Measuring the Cyclicality
of Real Wages: How Important is Composition Bias?" Quarterlti
Journal of Economics 109, 1-25.
Solow, R.M. (1956), "A Contribution to the Theory of Economic
Growth," Quarterly Journal of Economics 70, 65-94.
Solow, R.M. (1998), Monopolistic Competition and Macroeconomic
Theory (Cambridge: Cambridge University Press).
Solow, R.M. (2003), "General Comments on Part IV, " in P. Aghion.
R. Frydman, J. Stiglitz and M. Woodford (eds.) Knowledge.
Information, and Expectations in Modern Macroeconomics
(Princeton: Princeton University Press).
Stiglitz, J.E. (1979), "Equilibrium in Product Markets with Imperfect
Information," American Economic Review Papers and
Proceedings 69, 339-346.
Stiglitz, J.E. (1992), "Capital Markets and Economic Fluctuations in
Capitalist Economies," European Economic Review 36, 269-306.
Stoneman, P. (1979), "A Simple Diagrammatic Apparatus for the
Investigation of a Macroeconomic Model of Temporary
Equilibria," Economica 46, 61-66.
Summers, LB. (1986), "Some Skeptical Observations on Real Business
Cy Theory," Federal Reserve Bank of Minneapolis Quarterly
Rev' 10, 23-27.

296
st Summers, L.H. (1988), "Relative Wages, Efficiency Wages and
ly Keynesian Unemployment," American Economic Review Papers
and Proceedings 78, 383-388.
Svensson L. (1999), "Price Level Targeting vs. Inflation Targeting,"
Journal of Money, Credit and Banking 31, 277-295.
Svensson, L. (2003), "Escaping from a Liquidity Trap and Deflation: The
d Foolproof Way and Others," The Journal of Economic
Perspectives 17, 145-166.
e Swan, T.W. (1956), "Economic Growth and Capital Accumulation,"
Economic Record 32, 334-361.
Taylor, J.B. (1977), "Conditions for Unique Solutions in Stochastic
Macroeconomic Models with Rational Expectations," Econo-
metrica 45, 1377-1385.
Taylor, J.B. (1979a), "Estimation and Control of a Macroeconomic
Model with Rational Expectations," Econometrica 47, 1267-1286.
Taylor, J.B. (1979b), "Staggered Wage Setting in a Macro Model," The
American Economic Review 69, 108-113
Tobin, J. (1969), "A General Equilibrium Approach to Monetary Theory,"
Journal of Money, Credit and Banking 1, 15-29.
Tobin, J. (1975), "Keynesian Models of Recession and Depression," The
American Economic Review Proceedings 65, 195-202.
Tobin, J. (1977), "How Dead Is Keynes?" Economic Inquiry 15, 459-468.
Turnovsky, S.J. (2002), "Knife-Edge Conditions and The Macro-
economics of Small Open Economies," Macroeconomic
Dynamics 6, 307-335.
Van Der Ploeg, F. (1996), "Budgetary Policies, Foreign Indebtedness, the
Stock Market and Economic Growth," Oxford Economic Papers
48, 382-396.
Van Groezen, B., L. Meijdam, and H.A.A. Verbon (2005), "Serving the
Old: Ageing and Economic Growth," Oxford Economic Papers
57, 647-663.
Van Parijs, P. (2000), "A Basic Income For All, Boston Review," 25.
Walsh, C. (2003a), Monetary Theory and Policy, Second Edition
(Cambridge, Mass.: MIT Press).
Walsh, C. (2003b), "Speed Limit Policies: The Output Gap and Optimal
Monetary Policy," American Economic Review 93, 265-278.
Wieland, V. (2006), "Monetary Policy and Uncertainty about the Natural
Unemployment Rate: Brainard-Style Conservatism versus
Experimental Activism," Advances in Macroeconomics 6,
Woglom, G.(1982), "Under-Employment Equilibrium with Rational
Expectations," Quarterly Journal of Economics 96, 89-107.

297
Wolfson, P. (1994), "A Simple Keynesian Model with Flexible Wages: a
Succinct Synthesis," Journal of Macroeconomics 16, 129-156.
Woodford, M. (1991), "Self-Fulfilling Expectations and Fluctuations in
Aggregate Demand," in N.G. Mankiw and D. Romer (eds.), New
Keynesian Economics (Cambridge, Mass.: MIT Press).
Woodford, M. (2003a), "Inflation Stabilization and Welfare," Contri-
butions to Macroeconomics 3.
Woodford, M. (2003b), "Comment on: Multiple-Solution Indeterminacies
in Monetary Policy Analysis," Journal of Monetary Economics
50, 1177-1188.
Wright, S. (2004), "Monetary Stabilisation with Nominal Asymmetries,"
The Economic Journal 114, 196-222.

298
 Readers who wish to pursue the nonuniqueness issue coulc
consult McCallum (2003), and the critical commentary on this me]
offered by Woodford (2003b). Another development in this literature
focuses on which of the multiple equilibria do agents converge to, if they
start with incomplete information and have to gradually learn acquire
what we have assumed they have from the outset — rational expectations.
The idea is that only equilibria agents can get to via gradual learning
should be deemed admissible. Two recent references for this literature are
Honkapohja and Mitra (2004) and Giannitsarou (2005). A general survey
on the nonuniqueness problem in monetary policy models is available in
Driskill (2006).
Despite the incompleteness in the methodology for dealing with
multiple equilibria, we proceed in the present circumstance by simply
rejecting the unstable outcome. Thus, we follow convention and focus on
the fractional value for parameter c. When the same illustrative parameter
value 034= 0.2) is inserted into this version of the model, we have

cr2 , = 0.77c2 v and 62, = 0.028a2 . (3.19)

By comparing (3.18) and (3.19), we see that — according to this

analysis — there is no volatility trade-off. By shifting from nominal-GDP
targeting to price-level targeting — which is motivated by a desire to have
lower price volatility — we get lower real-output volatility as well.
Svensson (1999) has refereed to this outcome as a "free lunch" — progress
made with respect to two goals, and this has been achieved by adjusting
just one policy instrument. Walsh (2003b) has shown that' this desirable
outcome is much less likely to emerge when the model involves a Phillips
curve which allows past inflation, not just expected future inflation, to
play a role. We return to this issue in Chapter 6, where we provide this
additional perspective on the "free lunch" possibility. Our fundamental
purpose in this section has been to explain the actual solution procedures
involved with rational expectations models.

3.6 Conclusions

This chapter has surveyed alternative approaches to modelling uncertainty

and business cycles. A primary focus has been on the alternative
treatments of expectations that have featured in the development of
modern macroeconomics. Initially, expectations was an issue stressed by
Keynesians, since it represents a mechanism that makes convergence to

60
 full equilibrium less assured. But, more recently, macroeconomists of all
persuasion highlight expectations in their analysis. This is because
stabilization policy is now modelled as an ongoing operation, not an
isolated one-time event, and analysts restrict their attention to models in
which agents are fully aware of what the government has been, and will
be, doing. Thus, as far as stabilization policy analysis is concerned, there
has been a convergence of views in two senses. First, all modern analysis
focuses on model-consistent expectations. A central task for this chapter
has been to equip readers to be able to execute the required derivations.
The second dimension of convergence among macroeconomists — whether
they come from either new Classical or more traditional Keynesian
traditions — is that they all emphasize micro-foundations. Without starting
from a specification of utility and a well-identified source of market
failure, there is no way we can argue that one policy is "better" than
another. Understanding these micro-foundations is our task for the next
chapter.
The analyses covered in the latter sections of this chapter are
examples of current research in macroeconomics. Unfortunately, space
constraints force us to exclude many interesting studies. For example,
further work on the theory of monetary policy has focused on more
general objective functions for the central bank. Instead of assuming that
the bank restricts its attention to hitting a price-level target in just the
current period, for example, a vast literature considers central banks with
an objective function involving both current and future values of several
macroeconomic outcomes. A central consideration in this literature is a
credibility issue. Since current outcomes depend on agents' expectations
of the future, it can be tempting for the central bank to promise a
particular future policy (to get favourable expectations effects at that time)
and then to deliver a different monetary policy once the future arrives
(since agents cannot then go back and change their earlier expectations).
Walsh (2003a, chapter 11) surveys this literature.
Other interesting work enriches the theory of rational expectations.
In one strand of the literature, agents are assumed to know some, but not
all, of the current endogenous variable values, when forecasts of the other
variables are being made. For example, when forecasting this period's real
GDP, we usually know more than just the lagged values of all variables.
We know a couple of current variable values — for example, for the
interest rate and the exchange rate (which are reported in the news media
every day, and which are never revised). We do not know enough to
figure out the current values for all the current "error terms," but we can
make somewhat better forecasts than we assumed agents could make in

61
 our analysis above. Minford and Peel (2002) provide a good summary of
this class of rational expectations models.
As already noted, another interesting line of investigation
involves agents who have to use least-squares forecasting techniques to
learn about a change in the economy's structure. In these models, agents'
forecasts gradually converge to the version of rational expectations that
we have assumed from the outset, and examined, in this chapter. While
the level of technical difficulty rises quite dramatically as these extensions
are pursued, they represent important developments. One reason for this
importance is the fact that some empirical work has been "unkind" to the
basic version of rational expectations that we have considered here.
One awkward bit of evidence for the rational-expectations
hypothesis is that surveyed series on expectations differ from the
associated actual series in systematic ways. Another source of tension
follows from sectoral econometric studies, such as estimated consumption
functions. For example, when Friedman (1957) tested his permanent-
income theory of consumption, he tested that hypothesis and an additional
one — that permanent income is related to actual measured income
according to the adaptive expectations scheme — simultaneously. This
package of hypotheses was not rejected by the data. Rational expectations
theorists have criticized Friedman's work on the grounds that the adaptive
expectations part of the package gave him a "free" parameter. Since there
was no restriction based on the theory for the coefficient of expectation
revision to be anything other than a positive fraction, Friedman's
computer was free to pick a value for that parameter that maximized the
goodness of fit of the overall package of hypotheses.
When Friedman's critics re-estimate the permanent-income theory
of consumption — as a package with rational expectations instead — they
find much less support. It can be argued that these later studies (for
example, Sargent (1978)) have gone too far in replacing the free-
parameter problem with a very-restrictive assumption concerning the rest
of the economy. Since rational expectations has to be implemented as a
full-model proposition, the consumption function cannot be estimated
without specifying the entire rest of the macro model. Since this is done in
a boldly simplified way in any study this is focusing on just one
relationship (for example, the consumption function), imposing the
associated cross-equation restrictions can essentially destroy any chance
that the original theory might have had to get a "passing grade." In other
words, as it is often applied, empirical work that combines inter-temporal
optimization by one group of agents, an over-simplified specification of
the rest of the economy, and rational expectations may involve too few —
not too many — free parameters. Carroll's (2001) concern goes even
62
 further. He generates fictitious data from sets of simulations involving
hypothetical consumers who behave exactly according to the theory (the
strict permanent income hypothesis plus rational expectations). He finds
that fictitious researchers who use this data still reject the theory. So there
is certainly more research to do on how to test both our inter-temporal
models and the hypothesis of rational expectations. Beyond acquiring an
initial awareness of these challenges, it is hoped that readers have
acquired two things by studying this chapter: both an ability to solve and
interpret basic rational-expectations models, and a sense of perspective
concerning this literature.
We end this chapter by summarizing the prerequisites for any
analysis of stabilization policy to be deemed acceptable by modern
macroeconomists — whether their background is Classical or Keynesian.
These features are: consistent with inter-temporal optimization, allowing
for some short-run nominal stickiness, and involving model-consistent
expectations. In addition to these features, if the analysis is to be used to
directly inform actual policy debates, it has to be highly aggregative and
fairly simple (that is, involve a limited number of equations). The
framework that appears to satisfy all these prerequisites has been called
the New Neoclassical Synthesis. We considered the basic version of this
framework in section 3.5 in this chapter. In the next chapter, we discuss
the micro-foundations of this framework. Then, in the following two
chapters, we explore in a much more complete manner, the development
of this new synthesis.

63
Chapter 4

The Micro-Foundations
Of Modern Macroeconomics

4.1 Introduction

The traditional analysis that was reviewed in previous chapters involved

some appeal to micro-foundations. For example, in the textbook classical
model (in Chapter 1), it is customary to refer to a static theory of
household utility maximization to "justify" the household labour supply
function. (We assumed that households maximize a utility function
containing two arguments — after-tax real income and leisure — subject to a
simple budget constraint — that income equals the after-tax real wage
times the amount of time worked.) Also, we assumed that firms maximize
their profits, and this was the rationale behind the standard optimal hiring
rule for labour — that workers be hired until the marginal product of labour
is pushed down to the rental cost of labour (the wage). But we did not
assume a similar optimal hiring rule for the other factor input — capital.
This different treatment is explained later in the present chapter — in the
section on the micro-foundations of the firms. But first, we consider the
micro basis for the equations of a standard macro model that summarize
the behaviour of households. And even before that, in the next section of
the chapter, we motivate the drive for more explicit micro-foundations in
modern macroeconomics.

4.2 The Lucas Critique

Before the early 1970s, virtually all macro policy analyses were
inconsistent with the principles of microeconomics. The households and
firms who operated within the macro model followed the same decision
rules no matter how their environment was altered by changes in policy
regime. Economics is often defined as the subject that explores the
implications of constrained maximization. But this description did not
apply to traditional macroeconomics. Since this was forcefully first
pointed out by Nobel laureate Robert Lucas in 1976, and since
macroeconomists have been working hard to avoid this criticism ever
since, we begin this chapter by explaining what has become known as the
"Lucas critique" of more traditional macroeconomics. The material in this
64
 section provides a compact summary of Lucas' argument — with an
explicit applied example — by relying on a simple theory behind the
Phillips curve. A fuller (inter-temporal) version of this theory of sticky
prices is available in McCallum (1980) and Mussa (1981). Here, for
simplicity, we consider just a one-period optimization.
Firms face negotiation costs when setting wages. When workers
are dissatisfied with the prospect that wage changes will be too low, they
work to rule or strike, and this industrial action lowers output (and raises
firms' costs). When firm owners are dissatisfied with the prospect that
wage changes will be too high, they lock out workers, and these actions
also reduce output and raise costs. If prices are a mark-up on wages, this
same reasoning implies that these adjustment costs are incurred whenever
price increases are either above or below what full-equilibrium
considerations dictate would be an "appropriate" price change. To capture
these considerations, we assume that the price-setter's optimization
involves balancing the costs of fast versus slow price adjustment. Fast
adjustment lessens the cost of being away from full equilibrium, while
slow adjustment lessens the adjustment costs. The following cost function
captures these considerations:

(p, - A)
z
+PRP, -P,_)-(7.1-75,_)1 2
-p- is the full-equilibrium price — the one that would have the firm
operating at its natural rate of output (at the minimum point of its long-run
average-cost curve). The first term in the cost function captures the cost of
being away from this desirable long-run level of operations. The second
term captures the adjustment cost — that exists whenever the actual change
in wages (and therefore prices) is either above or below what is dictated
by the firm's equilibrium considerations. Parameter 13 defines the relative
importance of these adjustment costs. To keep this demonstration of the
Lucas critique simplified, we take 13 to be a "primitive" parameter.
economics starts with a specification of tastes and technology. If one
Arants to explore the determination of tastes, one becomes a psychologist,
lot an economist, and if one is interested in understanding technology,
one becomes an engineer, not an economist. So taste and technology are
he primitive constructs in our discipline. Clearly, parameter 13 is not
vimitive in this ultimate sense. However, since the relative cost of
adjustment can be assumed to be an aspect of the given "technology"
aced by firms, we interpret 13 in this way, in the interests of a simplified
!xposition.

65
 Since the equilibrium considerations and previous history are
beyond the control of this period's price-setter, costs are minimized by
differentiating the cost function with respect to the current actual price
and setting the result equal to zero. After a bit of manipulation, we have

19,__)= — 15,1)

where y =1 /(1+ /1). This relationship can be written in continuous time as

=P - r(P - T9)•

A standard version of the short-run Phillips curve is

= /5 + 0(Y—Y)

and this form must now be related to the price-change relationship just
derived, via an aggregate demand relationship. For simplicity, let us
assume that transactions technology dictates that each level of output
requires a certain amount of money. This is captured in a quantity-theory
specification for aggregate demand (where parameter 0 is also "primitive"
— that is — unaffected by changes in the policy regime):

y = O(m — p).

We add a simple monetary policy reaction function:

m = — Z(P

According to this relationship, the central bank raises (lowers) the (log of
the) nominal money supply — above (below) its long-run average value —
whenever the price level is below (above) its target value. If parameter x
is zero, the central bank is a monetary-aggregate targeter; if x approaches
infinity, the bank is a price-level targeter.
Substituting the policy reaction function into the quantity theory
equation, we obtain the nation's aggregate demand function:

y = 0(n — 13)-0(1+ x)(p — 17)•

66
Next, we define the full-equilibrium price as that value for p that makes
output equal its natural rate. This value can be determined by setting
y = y and p = T) into the demand function:

y =19(in - T2).

This equation can be used in two ways. First, if the natural rate is constant,
the time derivative of this equation states that /-z5. = th ; in other words, the
core inflation rate is the money growth rate. The second use of the last
equation is that it can be subtracted from the demand function. The result
is:

(y —3;)= —0(1+ x)(p— T ).

Substituting this last relationship into the price-change equation that was
derived above, we end with:

io = + 0(Y — 57 )
where 0 = y 40(1 + x)).

This Phillips curve has two important features. First, it states that
temporary deviations of inflation from the monetary growth rate correlate
with temporary deviations of real output from the natural rate. This
implies that there is no inconsistency in the view that inflation is
ultimately a purely monetary phenomenon, and the view that the Phillips
curve is an important ingredient in a theory of the short-run interaction
between inflation and real activity. As noted earlier chapters, this has long
been a feature of standard macroeconomics.
Second, we conclude that the summary parameter 4 cannot be
considered a primitive parameter. According to this derivation, y and 0 are
technology parameters, since they define the adjustment-cost and
transactions processes. Thus, these parameters are not affected by
aggregate demand policy — that is, by changes in the value of monetary-
policy parameter x. But precisely because these "private sector" response
coefficients are policy invariant, it must be the case that the slope of the
short-run Phillips curve, 4, does depends on monetary policy. Thus, the
slope of the short-run Phillips curve does not represent purely "supply-
side" phenomena.

67
 Standard practice in applied economics (in all fields — not just
macroeconomics) involves estimating a model, and then using those
estimated coefficients to simulate what would happen if policy were
different. The Lucas critique is the warning that it may not make sense to
assume that those estimated coefficients would be the same if an
alternative policy regime were in place. The only way we can respond to
this warning is to have some theory behind each of the model's equations.
We can then derive how (if at all) the coefficients depend on the policy
regime. This derivation of the short-run Phillips curve illustrates two
things. First, the Lucas critique does apply to the Phillips curve — its slope
is a policy-dependent coefficient. Second, the derivation illustrates how
we can react constructively to the Lucas critique. Just because the slope
coefficient depends on policy, it does not mean that legitimate counter-
factual experiments cannot proceed. The value of micro-foundations is
evident. They do not just expose the non-primitive nature of the Phillips
curve slope; they also outline precisely how to adjust that parameter to
conduct theoretically defensible simulations of alternative policy rules.
One illustration of how the answers to policy questions change
when the micro-foundations are respected can be had by reconsidering the
results that were reported in Chapter 2 (Section 2.4). In that analysis —
which involved essentially exactly this model but without the micro
underpinnings — we determined that the speed of adjustment between full
equilibria (parameter s in p= —s(y — y)) could be derived in a
straightforward fashion, and in this case we have s = 00(1+ x). Without
considering micro underpinnings, we conclude that more aggressive
policy increases the economy's adjustment speed back to full equilibrium
following disturbances. But, according to the Lucas critique, we should
take account of the micro basis of the Phillips curve, and substitute out the
policy-dependent Phillips curve slope parameter (by using the
= y/(0(1+ x)) result) before conducting the policy analysis. When this
is done, the expression for the adjustment speed parameter becomes s = y .
This result has dramatically different policy implications. It says that
monetary policy has no effect on the economy's adjustment speed. So
respecting the Lucas critique does not just affect the smaller details of
policy analysis; it can change the analysis in fundamental ways.
But is Lucas critique actually important in the real world? It
appears so, if the history of recent decades is considered. Our micro-based
analysis predicts that a move toward more direct price-level targeting on
the part of the central bank can be expected to decrease the slope of
estimated Phillips curves. This effect is not usually noted, so (for example:
many people were surprised that the monetary-policy-induced recessions

68
 of the early 1980s and the early 1990s were as long and deep as they were.
These outcomes are less surprising when this underlying theory is
considered. In the 1980s and 1990s, central banks were increasingly
gearing policy to price level-targets. With a flatter Phillips curve emerging
as a result, the contraction in demand that was part of the disinflation
policy had a noticeably larger effect on real output than the previous
estimates of the Phillips curve had suggested. Viewed through the analysis
of this section, these outcomes are not surprising after all. While this
rather dramatic change in monetary policy clearly illustrates the
importance of the Lucas critique, some authors (such as Rudebusch
(2005)) have noted that the empirical evidence concerning less bold
changes in policy does not support the conclusion of structural breaks in
the estimated reduced forms of macro models.
Despite the value of micro-foundations (that has been illustrated
in a policy-relevant setting above), one consideration has kept some
macroeconomists from working toward a more elaborate microeconomic
base for conventional macro models. This problem is aggregation — an
issue which was not addressed in the Phillips curve analysis just given.
The conclusion which emerges from the aggregation literature is that the
conditions required for consistent aggregation are so rigid that constrained
maximization at the individual level may have very few macroeconomic
implications — that is, very few useful insights for aggregative analysis.
This presents a problem since the only way to solve the Lucas critique is
to use optimizing underpinnings to go "behind demand and supply
curves" (Sargent 1982), and to treat only the ultimate taste and technology
parameters as primitive (policy-invariant). If aggregation issues prevent
these individual optimizations from imposing any restrictions on
macroeconomic relationships, the Lucas critique cannot be faced. Yet
only a few commentators (for example, Geweke 1985) emphasize that
ignoring aggregation issues can be as important as ignoring the Lucas
critique.
Thus, we are on the horns of a dilemma. Economists should
ignore neither aggregation problems nor optimizing underpinnings. Yet
the current convention is, in essence, to ignore aggregation issues by
building macro models involving no differences between any individuals
— the so-called representative-agent model. The only justification for this
approach is an empirical one — that the predictions of the macro models,
which are based on such a representative agent, are not rejected by the
data. Thus macroeconomists have reacted to this dilemma in a pragmatic
way. Since aggregate models seem consistent with the macroeconomic
"facts," then no matter how restrictive the aggregation requirements seem,
not too much seems to be lost by assuming that the economy operates as if
69
these restrictions are appropriate. Some macroeconomists find this
pragmatic approach unconvincing, and they draw attention to inherent
logical difficulties within the representative-agent methodology (see
Kirman (1992) and Hartley (1997)).
In later chapters (10 and 12), we make use of an over-lapping
generations macro model that involves both appealing optimization
underpinnings and explicit aggregation across cohorts of different ages.
Indeed, we introduce readers to this model that generalizes the single
ever-lasting representative-agent framework later in the present chapter.
The over-lapping generations model certainly respects the Lucas critique
without ignoring the aggregation challenge. Nevertheless, since much of
the modern macro literature that we need to survey in this book follows
the representative-agent convention, we focus on this simpler model as
well.

4.3 Household Behaviour

The standard (descriptive) IS relationship involves households that are

liquidity-constrained; current consumption is limited by current income.
In fact, many households can borrow and lend, so current consumption
can exceed current income for many periods (as long as the household is
solvent — in the sense that the present value of its debts can be covered by
the present value of its assets). Many young families borrow as they
purchase a home and then use future income to cover both the purchase
price and the interest on the mortgage. Households do this because they
are impatient; other things equal, they get higher utility if consumption of
any item can occur sooner rather than later. The following model captures
this behaviour.
Assume that the household utility function is:

utility = ±(11(1+ p))' In C,

i is the index of time periods, p is the rate of time preference (the higher is
p, the more impatient people are), and the logarithmic form for the utility
function at each point in time is consistent with two propositions — that a
certain minimum amount of consumption (defined to be one unit) is
needed to live (to receive any positive amount of utility), and that (beyond
that one unit) there is diminishing marginal utility of consumption.

70
 The household maximizes this utility function subject to the
constraint that the present value of the entire stream of consumption be no
more than the present value of its disposable income. As readers will have
learned in basic micro theory, to achieve utility maximization, the
household must arrange its affairs so that the ratio of the marginal utilities
of any two items is equal the ratio of the prices of those two items. In this
case, the two items are the levels of consumption in any two adjacent time
periods. If the price of buying one unit of consumption today is unity, then
the price of one unit of consumption deferred for one period is less than
unity since the household's funds can be invested at the real rate of
interest for one period. Thus, the price of one-period-deferred
consumption is (1/(1+r)). With this insight, and the knowledge that
marginal utility for a logarithmic utility function is the inverse of
consumption, we can write the condition for utility maximization as:

(MWMU,_ /)= ((price of C in period i)/(price of C in period i-1))

[(1/(1+p)) 1(1/0]/[(1/(1+p)r i (1/C,. 1 )] = 1/(1+r)

C,= [(1+r)1(1+AGI

By subtracting lagged consumption from both sides of this last

representation of the household's decision rule, and by noting that (1+p) is
approximately equal to unity (a reasonable value for the annual rate of
time preference is something like 0.04), the consumption function can be
simplified further:

AC = (r p)C, —

Re-expressing this relationship in a continuous-time format, we have:

C = (r — p)C.

We now show how this lifetime-wealth-based consumption function —

which is consistent with inter-temporal optimization — is just another way
3f thinking about Friedman's (1957) model of permanent income.
According to Friedman, consumption is proportional to broadly-defined
wealth — the sum of non-human assets, A, and human wealth, H:

C = p(A +

71
The factor of proportionality is the rate of time preference. To see that this
is the same theory as the one just derived, we need to know how both
human and non-human wealth change over time. In the case of non-
human wealth, the specification is familiar. If each individual's level of
employment is one unit, she acquires assets when the sum of her wage
income, w, and interest income, rA, exceeds current consumption:

A=w+rA — C.

Human wealth is the present value of all future after-tax wage income.
Since intuition is more straightforward in a discrete-time specification,
initially we write human wealth as:

Hi = [w,l(l+r)]+ [w,+11(1+01+

Writing this relationship forward one period in time, we have:

11, 1 = [w,+1 41 +0] + [14'1+241+02]+ ...

Multiplying this last equation through by (1/(1+r)), and subtracting the

result from the H, equation, we have:

H,±1 — H,= rH,— w„ or

AH= rH— w.

In a continuous-time specification, then, Friedman's model can be

summarized by the following three relationships:

C = p(A+ H)
A=w+rA—C
H = rH —w

By taking the time derivative of the first relationship, and substituting the
other two equations into the result, we have:

e (r p)C .
—

This derivation proves that the permanent-income hypothesis and the

inter-temporal-optimization theory are equivalent. As a result, the inter-
temporal utility-maximization approach is supported by the extensive
72
e mpirical work that has established the professions' confidence in the
applicability of the permanent-income model, and Friedman's approach
gains from this equivalence as well, since it is now seen as more
consistent with formal optimization than had been previously thought.
Before proceeding with some extensions to this basic framework,
it is useful to indicate how the consumption function can be derived more
formally. We do so in two stages — first with time treated as a discrete
variable, and second with time treated as a continuous variable.

Households are assumed to maximize

E (11(1
70

utility = p))' In
.0

subject to

C, + (A, +, — A) , rA,+ w,.

After eliminating consumption by substitution, differentiating with respect

to the ith period value of A, and setting to zero, we have

as derived less formally above.

The same result can be derived in a continuous time setting if the
calculus of variations is used. This is how Ramsey (1928) originally
defined this analysis. In this case, the specification is that households
maximize

fin Ce-Ps dt
0

subject to

C=rA+w—A.

Since many readers will not have been taught how to deal with a situation
in which the objective function is an integral and the constraint is a
differential equation, a simple "cook book" rule is given here. Whenever

73
readers confront this situation, they should write down what is known as
the Hamiltonian. In this case it is

A ="ln[rA+w— A].

Optimal behaviour then follows from

A A —A A = 0,

where subscripts stand for partial derivatives. It is left for the reader to
verify that, in this case, following this procedure leads to exactly what was
derived earlier:

a (r — p)C .

Since the reader already knew the answer in this case, he/she can feel
reassured that the "cook book" method for dealing with continuous-time
specifications does "work".

Figure 4.1 Ricardian Equivalence

Future Income
and Consumption

Present Income
and consumption

74
 To have more intuition about this model of household behaviour,
consider Figure 4.1. For simplicity, the diagram refers to a planning
horizon with only two periods: the present (measured on the horizontal
axis) and the future (measured on the vertical axis), and taxes on interest
income are ignored. The household's endowment point is A. Since the
household can borrow and lend at rate r, the maximum amount of
consumption in each of the two periods is marked on both axes. The line
joining these two points is the inter-temporal budget constraint, and the
household chooses the point on this boundary of its feasible set that allows
it to reach the highest indifference curve (point B).
What happens if the government raises taxes today to retire some
government bonds? Since this just amounts to the government substituting
current for future taxes (with the present value of the household's tax
liabilities staying constant), all that happens is that the endowment point
shifts in a northwest direction along a fixed budget constraint to a point
such as C. The household simply borrows more, and remains consuming
according to point B. Thus, the Ramsey model involves what is known as
"Ricardian Equivalence" — the proposition that the size of the outstanding
government debt is irrelevant.
Some OECD governments have been very proud of themselves in
recent years — as they have been working down their debt-to-GDP ratios.
In a similar vein, policy analysts have regularly referred to George W.
Bush administration's fiscal policy as irresponsible, since the United
States government debt-to-GDP ratio has been rising dramatically. It is
clear that these OECD Ministers of Finance and these commentators on
US policy do not believe in Ricardian Equivalence. Perhaps this is
because they know that some "real world" individuals are liquidity
constraint — that is, they cannot borrow. We can focus on this situation in
Figure 4.1, by realizing that such an individual faces a budget constraint
given by DAE. Such an individual would be at point A — a corner solution
— initially. Then, after an increase in current taxes (which retires some
bonds and, therefore, cuts future taxes), the budget constraint becomes
DCF. The individual moves to point C. Since current consumption is
affected by the quantity of bonds outstanding, Ricardian Equivalence does
not apply.
Allowing for liquidity constraints is just one way to eliminate the
Ricardian Equivalence property. Another consideration is that people may
discount the future since they expect to die. The Ramsey model assumes
that the decision-making agent lives forever. If agents are infinitely lived
dynasties (households who have children), this assumption may be quite
appropriate. Nevertheless, some households have no children, so we
should consider the case in which agents do not live forever. The model
75
has been extended (by Blanchard (1985b)) by assuming that each
individual faces a constant probability of death, p. With this assumption,
life expectancy is (1/p) and (as derived in Blanchard and Fischer (1989),
chapter 3) the aggregate consumption function becomes

C = (p + p)(A + 11).

Aggregation across individual agents of different ages (to obtain the

aggregate consumption function from the first-order condition derived at
the individual level) is messy. Nevertheless, it is feasible — given the
assumption that life expectancy is independent of age. Not surprisingly,
individuals who expect to die, consume a bigger proportion of their
broadly defined wealth. Saving is less appealing when there is a larger
probability that the individual will not live to enjoy the spoils. The overall
rate of time preference becomes (p + p) — the sum of the individual's rate
of impatience and her probability of death. As in the simpler set-up, the
consumption function can be re-expressed as a consumption-change
relationship. The time derivative of the permanent-income representation
is taken, and the aggregate version of the accumulation identities for both
human and non-human wealth are substituted in. In this case, these
identities are:

A=w+(r+p)A—C—pA
H=(r+p)H—w.

The new terms in the wealth-accumulation identities stem from the fact
that people realize that the present value of their future wage income is
smaller when death is a possibility. The most convenient way to think of
the arrangements for non-human wealth is that there is a competitive
annuity industry in the economy. It provides each individual with annuity
income on her holdings of A throughout her lifetime, and in exchange the
individual bequeaths her non-human wealth to the annuity company when
she dies. Since a new person is born to replace each one that dies (so that
the overall population size is constant), in aggregate, both these payments
to and from the annuity companies are pA each period. When these
identities are substituted into the level version of the aggregate
consumption function, the result is:

C= (r — p)C — p(p + p)A.

76
 This consumption function collapses to Ramsey's when the death
probability is zero. Since government bonds are part of non-human wealth
(variable A), empirical workers have utilized this formulation to test
Ricardian Equivalence (the proposition that p = 0). In pooled time-series
cross-section regressions with future consumption regressed on current
consumption and non-human wealth, empirical researchers are able to
reject the null hypothesis that the A variable's coefficient is zero. We
respect these empirical results when using an overlapping-generations
analysis to evaluate deficit and debt reduction, and the implications of an
aging population, in Chapters 10 and 12, by allowing for p > 0 in that
analysis. However, since the New Neoclassical Synthesis approach to
stabilization policy analysis simplifies by setting p = 0, we follow this
convention as we report on that literature in Chapters 6 and 7.
Thus far, our theory of households has been simplified by
assuming exogenous labour supply. If households can vary the quantity of
leisure they can consume, inter-temporal optimization leads to the
derivation of both the consumption function and a labour-supply function.
With endogenous leisure — both current goods consumption and leisure
turn out to depend on permanent income. As far as labour income is
concerned, permanent income depends on both the current wage rate and
the present (discounted) value of the future wage rate. Thus, current
labour supply depends positively on both the current wage and the interest
rate, and negatively on the future wage. This aspect of the inter-temporal
model of household behaviour plays a central part in New Classical
macroeconomics (Chapter 5) — the first generation of fully micro-based
macro models. Also, the theory behind the "new" Phillips curve (that
Forms an integral part of the New Neoclassical Synthesis) relies very
much on this same labour supply function. To have a micro base for the
\lew Synthesis that is internally consistent, we want the labour supply
function to emerge from the same theory of the household that lies behind
he consumption function. Thus, for several reasons, we must now extend
:hat earlier analysis to allow for a labour-leisure (labour supply) choice.
As noted, for simplicity and to be able to report the New
neoclassical Synthesis literature as it is, we revert to the ever-lasting
:amity-dynasty version of the theory of the household, for this extension.
[hese household dynasties live forever and they choose a saving plan that
s designed to smooth consumption over time. Here we assume that
iouseholds maximize the discounted value of a utility function that is a
)ositive function of consumption and a negative function of labour
;upplied:

77
 f [ln C — (1 /(1 + s))N i+le-Pi dt

N is employment, and parameters turns out to be the inverse of the wage

elasticity of labour supply. The constraint is the standard accumulation
identity for non-human wealth — that asset accumulation equals total
income minus consumption:

C =rA+(W 1 P)N — A.

The Hamiltonian for this problem is:

A= [In C — ( N 1' 1(1+ s))+ A,(rA + (W I P)N —C —

where 2,, is a Lagrange multiplier. By following the method explained

above, readers can verify that households must follow the two rules
specified below (which are the consumption and labour supply functions,'
if they, in fact, maximize utility:

elC=r—p and
N =(W IPC)".

We use the labour supply function in the derivation of the Phillips curve
below. In this section of the chapter, we focus on the IS relationship. If
for simplicity, we ignore investment and government spending, we know,
that C = Y. Using lower-case letters to denote logarithms (for all variables
except the interest rate) and using r to represent the rate of time
preference, the consumption function can be re-expressed as j) = (r —7).
Except for adding an autonomous component of expenditure, this
completes the derivation of the "new" IS relationship. To add
component of demand that is independent of interest rates we take
logarithmic approximation of the economy's resource constraint (Y = C 4
G): y = y + ac + (1— a)g, where a is the full-equilibrium ratio of interest-
sensitive consumption to output; we end with the following slightly more
general micro-based IS relationship:

= a(r—7) + (1— a)k.

78
The final term is eliminated (by setting a = 1) when analysts are not
focusing on fiscal policy. We consider the policy implications of this
"new" IS relationship in Chapters 6 and 7.

4.4 Firms' Behaviour: Factor Demands

The consumption function is not the only behavioural relationship that is

embedded within the IS function. The investment function is an integral
part of the IS-LM system as well, so (to have a micro base for this
relationship) we consider the inter-temporal theory of the firm in this
section. Since firm managers may not know the time preference rate of the
firm owners, the best thing that managers can do for their owners (who
consume according to the permanent-income hypothesis) is to deliver a
flow of income that has the maximum present value. Thus, firms are
assumed to maximize the present value of net revenues

fe -"[F(N ,K)—wN — I —bIldt

subject to the standard accumulation identity for the capital stock:

I = K + 8K. S is capital's depreciation rate. The final term in the
objective function captures the "adjustment costs." It is assumed that — to
turn consumer goods into installed new machines — some output is used
up in the installation process. The quadratic functional form is the
simplest that can ensure that the installation costs for capital rise more
than in proportion to the amount of investment being undertaken. This
assumption is needed if investment is to adjust gradually to changing
conditions, and it is necessary for the production-possibility frontier
(between consumption and capital goods) to be the standard bowed-out
shape.

The Hamiltonian for this optimization is

A = e-"[F(N ,K)— wN — I —b1 2 q(I — k - MTh

where q is a Lagrange multiplier that denotes the value to the firm of a

slight relaxation of the constraint — that is, the value of a bit more capital.
Thus, q is the relative price of capital goods in terms of consumption
goods (and if equity markets function in an informed manner, q also
equals the value of equities). By using the "cook book" rule (applying it

79
for N, K and 1), the reader can verify that the following rules must be
followed for the firm to maximize its affairs.

FN =w
I = a(q —1) a=112b
r=FK lq-8+41q

The first rule is familiar; it stipulates that firms must hire labour each
period up to the point that its marginal product has been driven down to its
rental cost (the real wage). The second rule is intuitive; it states that firms
should invest in acquiring more capital whenever it is worth more than
consumption goods (alternatively, whenever the stock market values
capital at more than its purchase price). The third equation states that
individuals should be content to own capital when its overall return equals
what is available on alternative assets (interest rate r). The overall return
is composed of a "dividend" plus a "capital gain". When measured as a
percentage of the amount invested, the former term is the gross earnings
(capital's marginal product divided by the purchase price of capital) minus
the depreciation rate. The final term on the right-hand side measures the
capital gain. In our specification of the firms' investment function in
Chapter 1 (that was embedded within the basic IS relationship), we
assumed static expectations (4 = 0) . In Chapter 6 we are more general;
we examine what insights are missed by assuming static expectations. But
to simplify the exposition concerning installation costs for the remainder
of this section, we assume static expectations.
It is useful to focus on the implications of this theory of the firm
for the several models that were discussed in Chapter 1. In the textbook
Classical and Keynesian models, firms were assumed to have investment
and labour demand functions just like those that we have derived here. As
a result, we can now appreciate that the implicit assumption behind these
models is that firms maximize profits. That is, they pick factor inputs
according to cost minimization, and they can simultaneously adjust
employment and output to whatever values they want. With no installation
costs for labour, the standard marginal-product-equal-wage condition
applies. But, with adjustment costs for capital, a gap between the marginal
product of capital and its rental cost (the interest rate plus capital's rate of
depreciation) exists in the short run. The optimal investment function is:

I = a[(FK l(r + 8)) —1].

80
This relationship states that investment is proportional to the gap between
capital's marginal product and its rental cost. Since capital's marginal
product is positively related to the employment of labour, this result
"justifies" assuming that investment depends positively on output and
negatively on the rate of interest. (This is standard in traditional IS-LM
theory, although sometimes (as in Chapter 1 above), analysts simplify by
excluding the income argument.)
If firms encounter a sales constraint (that is, if they pick factor
inputs with the goal of achieving cost minimization without being able to
simultaneously adjust employment and output to whatever values they
want), the optimal investment function is:

I = aRFK (W I P)) I (FN (r + 8))— 1].

Readers can verify this by re-specifying the Hamiltonian to

A= [F(N ,K)— wN — I — bI 2 + q(I — k - SK)+ — F(N , K))],

where Y and represent the sales-constrained level of output and the

Lagrangian multiplier attached to that constraint. The revised investment
function emerges after A, is eliminated by substitution — using the first-
order conditions for labour. Some of the implications of this revision in
the investment function were explained in our analysis of the extended
Keynesian model of generalized disequilibrium in Chapter 1.
Returning to a setting with no sales constraints, it is worth
drawing attention to the fact that the investment function that is consistent
with profit maximization can be written in several different ways. We
have already seen that it can be written as I = a(q —1) and
I = aRFK I (r + —1] . A third alternative is also possible. Let K* denote
the optimal holding of capital in full equilibrium. In a no-growth setting,
investment in full equilibrium is just for replacement purposes, so we have
I = 8K * . When this equation is combined with the accumulation identity
for capital, I= k+ 8K, we have K = 8(K * —K). This way of
summarizing optimal behaviour says that firms should follow a partial-
adjustment rule when setting net investment. Net investment should equal
a constant fraction of the gap between the actual and the desired capital
stocks. Empirical workers have made this assumption for many years. The
contribution of the formal micro-foundations is that we can appreciate that
the partial adjustment coefficient should not be treated as a free parameter.

81
For the theory to receive empirical support, we must find that the
estimated partial adjustment coefficient must equal the depreciation rate.
It was mentioned above that we could interpret the Lagrange
multiplier, q, as the value of stocks. In fact, it can also be interpreted as
the slope of the nation's production-possibilities curve as well. We defend
both interpretations now. In a well-functioning stock market, the value of
equities equals the present value of the income derived from owning
capital. If capital is held into the indefinite future, its per-period earnings
will be national output, PF(N ,K), minus the wage bill, WN. To obtain
the present value of this flow, it is discounted by the sum of the real
interest rate and the depreciation rate. (Not only must the future be
discounted to calculate present value, but also the capital stock must be
maintained.) Assuming constant returns to scale, we have
F(N,K). FK K + FN N. Using this fact and PFN = W, the market value
of equities can be re-expressed in nominal terms as PFK K l(r + 6). We
can define q as the ratio of the market's valuation of capital to its actual
purchase price, PK. This means q = FK /(r + 8). This is most appealing.
When shares can be sold for such a price that capital can increase at the
same rate as the ownership of the company is being diluted, and there are
some additional funds left over, the existing owners should approve
expansion. This was the intuition behind the Keynes-Tobin (1969)
approach to the investment function. It is reassuring to know that we can
embrace a model of investment that is consistent with both this Keynesian
intuition and formal inter-temporal optimization.
Without adjustment costs, there is no difference between
consumption goods and investment goods. In that case, the economy's
production possibilities curve (drawn in C—I space) is a straight
(negatively sloped) 45-degree line. But with adjustment costs, this is not
the case. Ignoring government spending for a simplified exposition, the
amount of goods that are available for consumption is given by
C = F(N ,K)— I — bI 2 . The slope of the production possibilities curve is
had by taking the total differential of this definition, while imposing the
condition that factor supplies are constant (dN = dK = 0). The result is
dC =—qdI, since we know from our earlier derivations that q=1+2b1.
Thus, Tobin's valuation ratio can also be interpreted as the relative price
of investment goods in terms of consumption goods. As long as macro
theorists specify the resource constraint to include the installation cost
term, the production possibilities curve has its normal bowed-out shape,
and the model involves a consistent aggregation of the two kinds of goods
(even though it appears as simple as a one-sector model).

82
 Other versions of the firms' investment function emerge if we
specify alternative installation-cost functions. For example, with an
installation-cost specification that normalizes for the size of the firm (such
2
as b1 2 1K or b1 I Y) we get slightly different investment functions:
I 1 K = a(q —1) and / / Y = a(q —1) . In this last specification, since the
installation process involves labour, there is also an important revision in
the labour demand function: FN (1 + b(I I Y) 2 ) = W / P. With this
specification, the separation of demand and supply-side fiscal policy
instruments is blurred. Anything, such as program spending, that can
affect interest rates and therefore investment, is a policy variable that
causes a shift in the position of the labour demand function. (With higher
interest rates, fewer workers are needed for installing capital (at any real
wage).) Since the labour market is what lies behind the aggregate supply
curve for goods, G is a policy variable that shifts the position of both the
demand and supply curves for goods. While some New Keynesian
economists (such as Stiglitz (1992)) use models of equity rationing to give
their model this very feature, space constraints limit our ability to pursue
further these models involving interdependent aggregate supply and
demand curves.
Finally, before closing this section, it is worth noting what
happens if there were no adjustment costs. In this case, parameter b would
be set to zero, and the first-order conditions imply that q would always
equal one, and that FN = r + 8. would hold at all times. In such a world,
.7.apital and labour would be treated in a symmetric fashion. Both factors
could be adjusted costlessly at each point in time so that — even for capital
- marginal product would equal rental cost. There would be no well-
lefined investment function in such a world, since firms would always
lave the optimal amount of capital. As a result, firms would passively
nvest whatever households saved. We examine macro models with this
'eature, when we discuss economic growth in Chapters 10-12. But all
nodels that focus on short-run fluctuations (that is, both Classical and
(eynesian models of short-run cycles) allow for adjustment costs.

1.5 Firms' Behaviour: Setting Prices

[he purpose of this section is to explore the micro-foundations of what

ias come to be called the "new" Phillips curve. This relationship now
orms an integral part of both New Classical macroeconomics and the
few Neoclassical Synthesis. The particular micro model of sticky prices

83
that is favoured in the literature is Calvo's (1983), so we explore that
analysis here.
A full treatment would be very complicated. We would need to
derive the firms' factor demand functions (labour demand and investment)
and the firms' optimal price-setting strategy simultaneously — within one
very general inter-temporal optimization. This is rarely attempted in the
literature. Instead, it is assumed that there are two separate groups of firms
The first produces an "intermediate" product, and it is these firms who
demand labour and capital. The second group of firms buys the
intermediate product and (without incurring any costs, but subject to a
constraint on how often selling prices can be changed) they sell them as
final goods. This two-stage procedure is an ad hoc simplification that is
intended to "justify" our not integrating the optimal factor-demands
problem with the optimal-price-setting problem. Even with this separation
a full treatment of the price-setting problem is more complicated than
what is presented here. The fuller treatment involves product
differentiation across firms. Strictly speaking, this is necessary, since each
individual firm must have some monopoly power to have a price-setting
decision. To ease exposition, however, the present discussion suppresses
the formal treatment of monopolistic competition. By comparing this
analysis to King (2000) and Goodfriend (2004), the reader can verify that
the "bottom line" is unaffected by our following the many authors who
take this short-cut.
Prices are sticky. Specifically, a proportion, 0, of firms cannot
change their price each period. One way of thinking about the
environment is to assume that all firms face a constant probability of
being able to change price. That probability is (1— 0), so the average
duration of each price is (1/0). Then, if p denotes the index of all prices
ruling at each point in time and z denotes what is set by those who do
adjust their price at each point in time, we have:

p, = Op,_1 + (1-0)z,.

The objective function for each firm is to minimize a quadratic cost

function — the discounted present value of the deviations between (the
logarithms of) its price and its nominal marginal cost, mc. The discount
factor is (1/(1+p)), and the cost function is:

E(1/(14-p))10/E,(p,
,=0

84
It is shown in the following paragraphs that the first-order condition for
this problem can be approximated by the following equation (if the
discount factor is set to unity):

(Pt — P,_1) = (A+1 — Pr ) e p(mc, — 10),

where the e superscript denotes expectations. Assuming perfect foresight

and switching to a continuous time specification, we can write

= - p, ).

To appreciate how we can reach these outcomes, we must first motivate

the initial objective function, and then outline the derivation. Firms
discount the future for two reasons: the normal rate of time preference
applies, and (as time proceeds) there is an ever smaller probability that
they are stuck with a fixed price. Using x =1/(1+p) as the discount
factor, the cost function for the ith firm that was introduced above can be
simplified to

E x`Of E,(p„—mc,, j )2

=Ej= 0
Of E,(p 2

We differentiate this objective function with respect to the firm's choice

variable, p„ and obtain

=Ex j0jE,(mc,, j ), so

p„=(1-0x)EziOlE,(mc,,i ).
.T=0

We assume that a symmetric equilibrium holds in each period, so that

pi, = z, for all firms. To summarize the derivation thus far, we have:

p, = Op,_,+(1— 0)z,. (4.1)

z,=(1-0x)E2,1 9/ Ei (mc,, i ) (4.2)

.fro

We simplify (4.2) as follows:

85
 z, = (1— 0x)mc, + (1— 0x)02iE 1 (mc, +1 ) + (1— 0X) 02 2'2E,(mc1.,2) +

z, = (1— 0x)mc, + (1— 0x)0xi0- E t (mci+j+i)

z1 = (1— 0x)mc, + (1 — 0x) 64,1E,(zt+1) (4.3)

The next steps in the derivation are as follows. First, write (4.1) forward
one period in time. Second, take the expectations operator, E, , through
the result. Third, substitute the result into (4.3), then that result into (4.1).
Finally, simplify what remains, using the definition of the inflation rate:
7T, = pe — p,_ 1 . We end with

71", = xE,(z, +1 )+[(1— 0)(1— 0x)I 0]rmc, (4.4)

where rmc stands for (the logarithm of) each firm's real marginal cost:
rmc, = mc1 — p t . Thus,

7r, = ,u(rmci )

where p = — 0)(1— 0x))1 0. We approximate this relationship below by

setting the discount factor, x, equal to unity.
The final step in the derivation involves replacing the real
marginal cost term with the output gap. This substitution can be explained
as follows. We start with the definitions of nominal marginal cost, total
product, and the marginal product of labour, assuming that the production
function is Cobb-Douglas:

MC = dTC I dY =d(WN)1 dY =W(dN I dY)=W I MPL

Y = N'
MPL= aY I N

We combine these relationships with the household labour supply

function (which was derived above, and which has been modified by
replacing C with Y): N =(W 1 PY) Il e . After eliminating MPL, N and W by
substitution, we have (MC / P) = Y° / a, where SI = (1 + e) / a. Re-writing
this in logarithms, we have

(mc— p)= — in 6.

86
We pick units so that, in full equilibrium, price equals marginal cost
equals unity. This implies that firms have monopolistic power only when
they are out of long-run equilibrium, and it implies that the logarithm of
both price and marginal cost are zero. Thus, the full-equilibrium version
of this last equation is

(rriE — )7) = — In cr.

Subtracting this last relationship from the previous one, we end with

— P) = n(Y

When this relationship is combined with either the discrete-time or the

continuous-time relationships derived above

(Pi — = (Pi+1 )e p(mc, — P1), or P = - 1-1(MCI

the final result is the "new" Phillips curve:

(P1 — /3,1) = — + 0(Y - 37)1 or p = —0(y — ,

where q$ = pn. This completes the derivation of the closed-economy

version of the "new" micro-based Phillips curve.

4.6 Conclusions

The hallmarks of modern macroeconomics are model-consistent

expectations, and micro-based, not simply descriptive, behavioural
reaction functions on the part of private agents. Chapter 3 contained the
analysis that is needed to allow readers to solve macro systems involving
model-consistent expectation, and the present chapter contains the
analysis that is required to permit readers to understand the inter-temporal
optimization that lies behind modern macro models.
The remaining task is to develop an understanding of how the
policy implications of the "new" macro models differ from the more
traditional descriptive models. We took a first pass at exploring this
question in section 3.5 in the last chapter. The reader is now in a position
to appreciate and verify that the system examined there is "new" — it
involves both the "new" IS and Phillips-curve relationships, and it
involves rational expectations. We now want to pursue this set of issues in
87
more detail, and to do so, we proceed in two stages. First, we explore the
work of the first group of macroeconomists to take micro-foundations
seriously — the New Classicals — in the next chapter. Then, we consider
the extension that New Classicals have embraced in recent years — sticky
prices — in Chapter 6. New Classicals refer to this extended model as a
simple dynamic general equilibrium model with sticky prices. Others refer
to it as a simple New Keynsian macro model. Still others refer to it as the
"New Neoclassical Synthesis." Whatever the label, this compact structure
now represents mainstream macroeconomics, and that is why we study its
properties in some detail in Chapters 6 and 7.

88
Chapter 5

The Challenge of
New Classical Macroeconomics

5.1 Introduction

Our analysis thus far may have left the impression that all macro-
economists feel comfortable with models which "explain" short-run
business cycles by appealing to some form of nominal rigidity. This may
be true as far as macro policy-makers are concerned, but this has not been
a good description of the view of many macro theorists. These theorists
are concerned that (until recently) the profession has not been very
successful in providing micro foundations for nominal rigidity. Even those
who have pioneered the "new synthesis" of Keynesian and Classical
approaches (for example, King (2000)) have expressed concern that some
of its underling assumptions concerning sticky prices can be regarded as
"heroic."
The response of New Keynesian academics is to work at
developing more convincing models of "menu" costs (the costs of
changing nominal prices), and to elaborate how other features, such as
real rigidities and strategic complementarities, make the basing of
business cycle theory on seemingly small menu costs appealing after all
(see Chapter 8). But there has been another. reaction — on the part of New
Classicals. Their reaction to the proposition that menu costs "seem" too
;mall to explain business cycles is to investigate whether cycles can be
explained without any reference to nominal rigidities at all. They have
Deen remarkably successful in demonstrating that this is possible. Further,
;ome revolutionary conclusions have been derived from this
`equilibrium" approach to cycles. Perhaps the most central result is that
he estimated benefit to society of completely eliminating business cycles
nay be trivial! This chapter explains this so-called real business cycle
approach, and how it leads to this strong verdict regarding stabilization
)olicy.

5.2 The Original Real Business Cycle Model

['he unifying theme in New Classical work is "equilibrium" analysis.

vlarkets always clear; agents make intelligent forward-looking plans; and
89
expectations are rational. By adopting this framework, New Classicals car
rely on the existing general equilibrium analysis provided by micro-
economists over more than a century. New Classicals feel that wher
Keynesians focus on sticky prices and disequilibria, they lose the ability tc
benefit from this long intellectual heritage. New Classicals see this as
big price to pay to make macroeconomics "relevant." This view becomes
particularly firm when the New Classicals feel that they may have
demonstrated that the equilibrium approach may be just as relevant — ever
for explaining short-run cycles. One purpose of this chapter to permit the
reader to form an independent decision on whether this claim of the Nev,,
Classicals can be sustained.
According to standard intermediate textbooks, Keynesian analysis
explains the business cycle by assuming rigid money wages. Demand
shocks push the price level up in booms and down in recessions. This set
of outcomes makes the real wage rise in recessions and fall in booms, and
the resulting variations in employment follow from the fact that firms
slide back and forth along a given labour demand curve. Thus, the model
predicts that the real wage moves contracyclically (and this is not
observed), and because the labour market is not clearing, it makes the
labour supply curve irrelevant for determining the level of employment.
New Classicals prefer to assume that wages are flexible and that
the labour market always clears. To their critics, a model which assumes
no involuntary unemployment is an "obviously" bad idea. But if this is so,
say the New Classicals, it should be easy to reject their theory. While the
approach has encountered some difficulties when comparing its
predictions to the data, it has turned out to be much more difficult to reject
this modelling strategy than had been first anticipated by Keynesians.
New Classicals explain business cycles by referring to real shocks
(shifts in technology) so the approach is called real business cycle analysis
The basic idea is that workers make a choice concerning the best time to
work. If something happens to make working today more valuable (such
as an increase in today's wage compared to tomorrow's) workers make an
inter-temporal substitution; they work more today and take a longer than
usual vacation tomorrow. Similarly, an increase in interest rates lowers the
present value of future wages, and so leads individuals to work more
today.
If there is a positive technology shock today, the higher marginal
product of labour implies a shift to the right of the labour demand curve.
Similarly, a negative development in the technology field next period
shifts the labour demand curve back to the left. Thus, according to this
view, business cycles are interpreted as shifts in the position of the labour
demand curve, not movements along a fixed labour demand curve. Cycles
90
trace out the points along the labour supply curve, and it is the reaction of
the suppliers of labour which is central. Since the observation point in a
graph of the labour market is never off the labour supply curve, any
unemployment that occurs in the low-activity period must be interpreted
as voluntary. Critics of the New Classical view are uncomfortable with
this interpretation of unemployment.
It should be noted that the New Classical model cannot explain
both a wide fluctuation in employment and a very mild fluctuation in real
wages over the cycle unless the wage elasticity of labour supply is very
large. Micro panel data suggest, however, that this elasticity is quite small.
Before addressing the several methods that New Classicals have used to
try to get around this problem, we explain the basic approach in more
specific terms, by reviewing Hansen and Wright's (1992) version real
business cycle model. It is defined by the following equations:
utility = Et (11(1+ p))I [ln C + /3 ln L,]
, (5.1)
1=0

L, + N, =1 (5.2)
Y,=e; Ka ,N ( ''), (5.3)
z, =ft + z, (5.4)
z, = 01-1+ vi (5.5)
K, +, = (1— 45)K, + (5.6)
Y, =C,+ (5.7)

Equation (5.1) is the household utility function; E, p, C and L denote

expectations, the rate of time preference, consumption, and leisure.
Equation (5.2) is the time constraint (N is employment). Formal utility
maximization is used to derive consumption and labour supply functions.
Equations (5.3), (5.4) and (5.5) define the production function and the one
stochastic variable in the system — the technology shock, z. An ongoing
trend is involved, and the stochastic part is a normally distributed error
term with zero mean and constant variance, and 4) is the coefficient of
serial correlation in this exogenous process. Formal profit maximization is
used to derive the demands for capital and labour. finally, equations (5.6)
and (5.7) define the accumulation of the capital stock (8 is its depreciation
rate and I is gross investment). New Classicals refer to the capital stock
accumulation identity as a "propagation" or "persistence-generation"
mechanism. Like the Keynesian assumption of staggered overlapping
wage or price contracts, this mechanism imparts persistence to the
system's dynamics, and this gives the model a serious chance to fit the
facts.
91
 While it has now become common practice for practitioners to
estimate simple aggregative models like this one, New Classicals initially
preferred to pick plausible values for the few parameters, and then to use
the model to generate data (by simulation). They compared the moments
of the time series that were generated by stochastic simulations from the
calibrated model, to the moments of the various time series from a real
(usually the US) economy. If the model's data "looked like" the real
world data, researchers concluded that this simple approach has been
vindicated. Of course, to avoid this exercise being circular, analysts must
choose the parameter values on the basis of empirical considerations that
are not econometric papers involving aggregate time series data.
Hansen and Wright report quarterly simulations with the
following chosen parameter values. The rate of time preference is set at
0.01 per quarter, which implies 4 percent on an annual basis. We have
learned that (with infinitely lived agents) this parameter should be the
same as the average real interest rate, so 4 percent is a plausible value.
Utility function parameter 13 is chosen so that the average proportion of
discretionary time spent working is one-third — a value consistent with
time-use studies. The production function exponent a is set at 0.36, a very
plausible value for capital's share of output in the United States (as
observed in the national accounts). A similar reference justifies that 10
percent of the capital wears out annually (so 8 = 0.025). Finally, the
technology shock process was calibrated by assuming a standard deviation
for the error term of 0.007 and a serial correlation coefficient of 0.95.
These values were borrowed from one of the original contributions in this
field (Kydland and Prescott (1982)) who in turn obtained these values by
estimating the combination of equations (3) - (5) with time series data for
the United States. By estimating the Cobb/Douglas production function
with data for Y, N and K only, the estimated residuals — the so-called
"Solow residuals" — can be taken as data for z. Thus, except for the
parameters which define this error process, all coefficient values are
chosen from sources other than the time series that the modellers are
trying to replicate. Nevertheless, since the coefficients for equation (5.5)
were chosen so that the times series properties of GDP would be
simulated well, we cannot count that feature of the results as any victory
at all.
But there is an achievement nonetheless, since other important
stylized facts of the business cycle are well illustrated by the "data" that is
generated from this very simple structure. For example, the simulations
show that consumption is less variable than income, and that investment is
more volatile than income. This outcome follows from the assumption of
diminishing marginal utility of consumption. If a positive technology
92
shock occurs today, people can achieve higher utility by spreading out this
benefit over time. So they increase consumption by less than their income
has increased initially, and as a result, some of the new output goes into
capital accumulation. Since this additional capital makes labour more
productive in the future, even a one-time technology shock has an impact
for a number of periods into the future. It is impressive that these very
simple calibrated models can mimic the actual magnitude of investment's
higher volatility relative to consumption — and this has been accomplished
without researchers allowing themselves the luxury of introducing "free"
parameters.
Despite these encouraging results, however, this model does not
generate the wide variations in employment and the very low variability in
real wages (that we observe in real data) without an implausibly high
value for the wage elasticity of labour supply (an unacceptable value for
parameter p). Thus, in the next section of the chapter, we consider a
number of extensions to the basic New Classical model that have been
offered as mechanisms that can make the "data" that is generated from the
calibrated models more representative of the actual co-movements in real
wages and employment.

5.3 Extensions to the Basic Model

The first extension we consider is an alteration in the utility function. The

one given in equation (5.1) embodies the assumption that the marginal
utility of leisure in one period is not affected by the amount of leisure
enjoyed in other periods. When that function is changed so that this
separability is removed, it makes agents more willing to substitute their
leisure across time, and so respond more dramatically to wage changes.
Keynesians regard this extension as ad hoc. They say that the Classicals
are introducing "free" parameters (like the Classicals accuse the
Keynesians have been doing when they assume arbitrary elements of
nominal rigidity) just to make the model fit. If the whole point of the New
Classical approach is to have a simple and standard market-clearing model
that fits the facts, then such an adjustment made after it was found to be
necessary indicates failure to some New Keynesians. In any event, the
modern approach to adapting the utility function (central to much recent
New Classical work) is to specify that today's utility depends on both
today's level of consumption and today's level of habits. Habits evolve
over time according to the following relationship:

h,,, — h, = A(c,—h,).
93
 This relationship is an additional persistence-generation mechanism that
helps both calibrated and estimated models match real-world data (see, for
example, Bouakez et al (2005)). But since the habit-adjustment process is
identical to the adaptive expectations formula, critics argue that New
Classicals cannot simultaneously argue that rational expectations is
fundamentally more appealing than adaptive expectations, and that this
extension to the standard utility function is not ad hoc. Of course, as a
technical matter, New Classicals can safely ignore such criticism. Since
our subject starts from a specification of tastes and technology, every
assumption about such matters is necessarily "arbitrary," and it seems to
demonstrate a misunderstanding of the bounds of our discipline to call any
such assumption "ad hoc." On the other hand, since the hypothesis of
adaptive expectations concerns the relationship between actual and
forecasted values of endogenous variables, not exogenous items such as
the definition of tastes, it is legitimately viewed as an arbitrary (ad hoc)
specification.
It is not useful for us to get bogged down in methodological
dispute. At the practical level, two considerations are worth mentioning.
First, since it is difficult to find any non-time-series-econometrics
evidence to use as a basis for choosing an "appropriate" value for the
habit-persistence parameter, 2k,, it has been difficult for practitioners to
avoid some proliferation of the free-parameter problem as they implement
this extension. The second point worth noting is that this generalization of
the utility function does not adequately repair the real wage-employment
correlations, so other changes to the basic model have become quite
prevalent in the literature as well. It is to some of these other
modifications that we now turn.
One such extension is indivisible labour. Some New Classical
models make working an all-or-nothing choice for labour suppliers. At the
macro level, then, variation in employment comes from changes in the
number of people working, not from variations in average hours per
worker. This means that the macro correlations are not pinned down by
needing to be consistent with evidence from micro studies of the hours
supplied by each individual (that show a very small elasticity). For more
detail, see Hansen (1985) and Rogerson (1988).
Another extension focuses on non-market activity. Statistical
agencies in OECD countries have estimated that, on average, households
produce items for their own consumption equal in value to about one third
-

of measured GDP. Thus, "home" production is a very significant amount

of real economic activity. Benhabib, Rogerson, and Wright (1991) have
shown that when the real business cycle model involves this additional
margin of adjustment, its real wage-employment correlations are much
94
more realistic. The basic idea is that the amount of leisure consumed can
remain quite stable over the cycle — even while measured employment in
the market sector of the economy is changing quite dramatically — when
households have the additional option of working at home (doing chores
for which they would otherwise have paid others to do).
Other New Classicals, for example Christiano/Eichenbaum (1992)
and McGrattan (1994) have introduced variations in government spending.
This work involves adding an additional (demand) shock (in addition to
technology shocks) to the model; for example,

G, = (1— 7)G- + yG,_, +u, (5.8)

is added to the system, and the market clearing condition is changed to

Y, = C, +I, +G, (5.7a)

When representative values for the additional persistence-generating

parameter y and for the variance of this additional error term are included
in the simulations, the resulting real wage-employment correlations look
much more realistic. It is straightforward to understand why. Increases in
government spending raise interest rates, and higher interest rates decrease
the present value of working in the future. With the relative return of
working in the current period thereby increased, the current-period labour
supply curve shifts to the right. With these supply-side shifts in the model,
some of the variations in employment are explained by shifts along a
given labour demand curve. With this additional source- of employment
fluctuation, the magnitude of the technology shocks does not need to be as
great for the calibrated model to generate realistic changes in employment.
[n addition, with both labour supply and demand curves shifting back and
7orth, wide employment variations can easily occur with very modest
;hanges in the real wage. That is, as long as the labour supply curve shifts
)ack and forth enough, its steepness is no longer a concern. Needless to
;ay, while Keynesians view the transmission mechanism quite differently,
hey are delighted to see the new school of thought relying on autonomous
;xpenditure variations — a central concept in traditional Keynesian models
- to improve the new model's predictions.
It is instructive to depict the New Classical model in terms of
aggregate demand and supply curves in price-output space. That picture
tppears just as in Figure 1.1 in the first chapter, but the list of shift
nfluences for the aggregate demand and supply curves are different from
vhat variables cause shifts in the textbook version of the classical

95
framework. For simplicity, in the present discussion, we assume static
expectations, so no separate expected future consumption and expected
future wage rates need to be considered. But when the labour supply and
demand equations are combined (to eliminate the real wage by
substitution), we are left with a relationship that stipulates output as a
positive function of the interest rate. We can use the IS relationship to
replace the interest rate. The resulting summary of the labour demand,
labour supply, production-function, and IS relationships is a vertical line
in P-Y space with government spending and tax rates — in addition to the
technology shock — as shift influences.
The real business cycle model allows no role for the money
supply, so — to have the model determine nominal prices — we must add
some sort of LM relationship to the system. Initially, to avoid having to re-
specify the household's optimization, this relationship was assumed to be
the quantity-theory of money relationship (L(Y) = M / P) — justified as a
specification of the nation's trading transactions technology. This
relationship is the economy's aggregate demand function, and the nominal
money supply is the only shift influence. Thus, the New Classical model
still exhibits the classical dichotomy as far as monetary policy is
concerned. But fiscal policy — even a change in program spending — has a
supply-side effect, so it has real output effects. Readers can draw the
appropriate aggregate demand and supply diagram to compare the effects
on output and the price level of variations in autonomous expenditure
across several models — those examined in Chapter 1 and the New
Classical model of the present chapter. Since the traditional Keynesian
model involves the prediction that the price level rises during business-
cycle booms, while this New Classical model has the property that the
price level falls during booms, various researchers (for example, Cover
and Pecorino (2004)) have tried to exploit this difference in prediction to
be able to discriminate between these alternative approaches to
interpreting cycles.
Care must be exercised when pursuing this strategy, however,
since there is no reason to restrict our attention to an LM relationship that
does not involve the interest rate. As we see in section 5.4 below, it is not
difficult to extend the household-optimization part of the New Classical
model to derive such a more general LM relationship from first principles.
When this more general specification is involved in the model, the IS
function is needed to eliminate the interest rate both from the labour-
market relationships (to obtain the aggregate supply of goods function)
and from the LM relationship (to obtain the aggregate demand for goods
function). As a result, changes in autonomous spending shift the position

96
 of both the aggregate supply and demand curves, and the model no longer
predicts that the price level must move contra-cyclically.
Other extensions to the basic New Classical model can be
understood within the same labour supply and demand framework that we
have just discussed. Any mechanism which causes the labour demand
curve to shift over the cycle decreases the burden that has to be borne by
technology shocks, and any mechanism which causes the labour supply
curve to shift accomplishes the same thing — while at the same time
decreasing the variability of real wages over the cycle. The variability of
price mark-ups over the cycle is an example of a demand-shift mechanism,
and households shifting between market-oriented employment and home
production is an example of a supply-shift mechanism. Regarding the
former, we know that the mark-up of price over marginal cost falls during
booms because of the entry of new firms. This fact causes the labour
demand curve to shift to the right during booms. Devereux, Head, and
Lapham (1993) have shown how this mechanism can operate within a real
business cycle model involving imperfect competition.
There are still other reasons for the labour demand curve to move
in a way that adds persistence to employment variations. Some authors
(such as Christiano and Eichenbaum (1992)) introduce a payment lag. If
firms have to pay their wage bill one period before receiving their sales
revenue, the labour demand function becomes EN = ( W I P)(1 + r) so
variations in the interest rate shift the position of the labour demand curve.
Further, firms may encounter adjustment costs when hiring/firing labour,
and learning-by-doing may be an important phenomenon (see Cooper and
Johri (2002)). In the latter case, tomorrow's labour productivity is high if
today's employment level is high. Simulations have shown that the
consistency between the output of calibrated equilibrium models and real-
world time series is increased, when persistence-generation mechanisms
such as these are added. It can be challenging to discriminate between
some of these mechanisms. For example, there is a strong similarity
between the habits extension and the learning-by-doing extension. But
despite this, Bouakez and Kano (2006) conclude that the habits approach
fits the facts better.
As noted above, it is interesting to note the convergence involved
with parts of New Keynesian and New Classical work. Keynesians have
been taking expectations and micro foundations more seriously to
improve the logical consistency of their systems, while Classicals are
embracing such things as autonomous expenditure variation, imperfect
competition and payment lags to improve the empirical success of their
models. Despite this convergence, however, there is still a noticeable
difference in emphasis. Classical models have the property that the
97
observed fluctuations in employment have been chosen by agents, so there
is no obvious role for government to reduce output variation below what
agents have already determined to be optimal. This presumption of social
optimality is inappropriate, however, if markets fail for any reason (such
as externalities, moral hazard, or imperfect competition).
Hansen and Wright (1992) have shown that when all of these
extensions are combined, a simple aggregative model can generate data
that reflects fairly well the main features of the real-world real wage-
employment correlations after all. But still the model does not fit the facts
well enough, so even pioneers of this approach (for example, Goodfriend
and King (1997) have called for the New Neoclassical Synthesis in which
temporarily sticky prices are added to the real business cycle model. We
consider this synthesis in some detail in Chapter 6.
It may seem surprising that New Classicals have embraced the
hallmark of Keynesian analysis — sticky prices. Why has this happened?
Perhaps because there is one fact that appears to support the relevance of
nominal rigidities — the well-documented correlation between changes in
the nominal money supply and variations in real output. Either this is
evidence in favour of nominal rigidities, or it is evidence that the central
bank always accommodates — increasing the money supply whenever
more is wanted (during an upswing). In response to this reverse causation
argument, Romer and Romer (1989) have consulted the minutes of the
Federal Reserve's committee meetings to establish seven clear episodes
during which contractionary monetary policy was adopted as an
unquestionably exogenous and discretionary development. The real
effects that have accompanied these major shifts in policy simply cannot
be put down to accommodative behaviour on the part of the central bank.
Further evidence is offered in Romer and Romer (2004), and further
support is provided by Ireland's (2003) econometric results. This evidence
— especially that which is derived from independent evidence of the
central bank's deliberations — removes the uncertainty that remains when
only statistical causality tests are performed. One final consideration is
that real and nominal exchange rates are very highly correlated. Many
economists argue that there appears to be no way to account for this fact
other than by embracing short-run nominal rigidities.
Keynesians welcome this convergence of research approaches, yet
(at the conceptual level) they remain concerned about the lack of market
failure involved in the classical tradition. Also, they have some empirical
concerns, and a few of these are summarized in the next few paragraphs.
The real business cycle approach is based on the notion of inter-
temporal substitution of labour supply. However, micro studies of
household behaviour suggest that leisure and the consumption of goods
98
 are complements, not substitutes (as assumed in New Classical theory).
Another awkward fact is that, in the United States at least, only 15 percent
of actual labour market separations are quits. The rest of separations are
layoffs. In addition, the data on quits indicates that they are higher in
booms. The real business cycle model predicts that all separations are
quits and that they are higher in recessions. Finally, it is a fact that a high
proportion of unemployment involves individuals who have been out of
work for a long time — an outcome that does not seem consistent with the
assumption of random separations.
Other interpretation disputes stem from the fact that it is
impossible to observe the technology shocks directly. In Solow's original
work, the residual accounted for 48 percent of the variation in the output
growth rate. Later work, which measured inputs more carefully, avoided
some aggregation problems, and allowed for a variable utilization rate for
both labour and capital, pushed the residual's contribution down to 3
percent. It is no wonder that New Classicals can explain a lot with the
original Solow residuals; they contain a lot more than technology shocks.
Incidently, many analysts find it reassuring that the residuals are now
perceived to be much smaller. Surely, if there are both positive and
negative technology shocks, disturbances at the individual firm or industry
level would largely "cancel out" each other, so that, in the aggregate, there
would not be large losses in technological knowledge. Possibly it would
be better if New Classicals interpreted real shocks more broadly, and
included such things as variations in the relative price of raw materials, as
well as technology shocks, in what they consider as supply-side
disturbances.
Quite apart from all these specific details, and the even the
general question of real-wage-employment correlations, some economists
regard the inter-temporal substitution model as outrageous. In its simplest
terms, it suggests that the Great Depression of the 1930s was the result of
agents anticipating World War II and deciding to withhold their labour
services for a decade until that high labour-demand period arrived.
Summers' (1986) remarks that even if workers took such a prolonged
voluntary holiday during the 1930s, how can the same strategic behaviour
be posited for the machines that were also unemployed?
Given these problems, why does real business-cycle theory appeal
to many of the best young minds of the profession? Blinder (1987)
attributes the attraction to "Lucas's keen intellect and profound
influence," but it also comes from the theory's firm basis in
microeconomic principles and its ability to match significant features of
real-world business cycles. Rebelo (2005) provides a clear and balanced
assessment of both the successes and some of the challenges that remain
99
 for the research agenda of real business cycle theorists. It is likely that
both this ongoing willingness to address these challenges, and the shift of
the New Classicals from calibration to estimation, have strengthened the
appeal of this approach to young researchers. Finally, perhaps another
consideration is that this school of thought's insistence on starting from a
specification of utility makes it possible for straightforward normative
(not just positive) analysis to be conducted. Since the theory is so
explicitly grounded in a competitive framework with optimizing agents
who encounter no market failure problems, the output and employment
calculations are not just "fairly realistic"; they can be viewed as optimal
responses to the exogenous technology shocks that hit a particular
economy.
Using this interpretation, economists have a basis for calculating
the welfare gains from stabilization policy. Lucas (1987) has used data on
the volatility of consumption over the cycle and an assumed degree of
curvature in the utility of consumption function that appears to fit some
facts to calculate how much business cycles lower utility. He concludes
that "eliminating aggregate consumption variability entirely would ... be
the equivalent in utility terms of an increase in average consumption of
something like one or two tenths of a percentage point." Lucas'
conclusion has been influential; it is one of the reasons macroeconomists
have shifted their emphasis to growth theory (Chapters 10-12) in recent
years.
It is interesting to interpret the simulations produced by modern
real business cycle theorists as an up-dated version of Adelman and
Adelman (1959). These authors performed a similar stochastic simulation
experiment with a small econometric model; their intention was to show
that a standard Keynesian model (with just a few numerical parameters)
could mimic the actual US data. Since both groups (Old Keynesians and
New Classicals) have established that their models are consistent with
significant parts of actual business-cycle data (and therefore should be
taken seriously), how can any one of them argue that its preferred
approach should have priority in the profession's research agenda (for this
reason alone)? Even New Keynesians, for example Ambler and Phaneuf
(1992) have shown that an updating of the original Adelman/Adelman
study gives the same support to the New Keynesian approach. Now that
all groups have proved that their approach has passed this basic test — to
be taken seriously as one of the legitimate and contending schools of
thought — it seems that either some additional criteria for choosing among
the different approaches is required — or a synthesis of the New Classical
and sticky-price approaches should be embraced. Indeed, as we explore in
the next chapter, this synthesis has been just what has developed.
100
 Before ending this chapter and moving on to the synthesis model,
we do two things. First, we use a particular version of New Classical
theory to illustrate how this school of thought can be used to contribute to
policy debates. In particular, we show how it facilitates our estimating the
long-term benefits of adopting a low-inflation policy. Second, we pursue
Lucas's proposition that the value of stabilization policy is trivial.

5.4 Optimal Inflation Policy

To justify a zero-inflation target, many analysts make the following

argument. Since inflation is a tax on the holders of money, and since taxes
create a loss of consumer surplus known as an "excess burden" — that tax
(inflation) should be eliminated. The problem with this argument is that if
one tax is eliminated, the government must raise another tax (and that
other tax creates an excess burden of its own). Recognizing this, the
standard approach in public finance is to recommend the "inverse
elasticity rule" for setting taxes. Since excess burdens are bigger when
taxes cause large substitution effects, the efficiency criterion for judging
taxes stipulates that the largest tax rates should apply to the items that
have the smallest price elasticities of demand. Since the interest rate is the
(opportunity) cost of holding money, and since the estimated interest
elasticity of money demand is very small, the inverse elasticity rule can be
used to defend a relatively large tax on money. In other words, it supports
choosing an inflation rate that is (perhaps well) above zero.
The full-equilibrium benefits of low inflation are independent of
the complexities of the transition path that the economy takes to reach the
long run (such as those caused by nominal rigidities). As a result, all
analysts agree that a basic version of the New Classical model is the
appropriate vehicle to use for estimating the benefits of that policy. It is
true that a model with nominal rigidities is needed to assess the short-term
costs of reducing inflation, and that is why we use such a model for
addressing this issue in Chapter 7. But here, since our focus is on the long-
term benefits of low inflation, we use an example of New Classical work
that highlights money — Mansoorian and Mohsin (2004).
Households maximize a standard utility function. As usual,
instantaneous utility is a weighted average of leisure and consumption
(a In(1— N) + (1— a) ln C), and there is a constant rate of time preference,
p. The budget constraint is

C=wN+rK+r—.rm—A.

101
Consumption is the sum of wage and employment income, plus the
transfer payments received from the government, T, minus the inflation
tax incurred by holding real money balances and minus asset
accumulation. Since A= K + m and i = r + r, we can re-express the
constraint as C = wN + rA + r — im — A. Households do not focus on the
fact that their individual transfer payment may depend on how much
inflation tax the government collects from them — individually. That is,
they do not see the aggregate government budget constraint (given below)
as applying at the individual level. However, there is an additional
complication that confronts households. This novel feature is the "cash-in-
advance" constraint: each period's consumption cannot exceed the start-
of-period money holdings. Since the rate of return on bonds dominates
that on money, individuals satisfy the financing constraint as an equality:
C = m. Thus, we replace the m term in the constraint with C. Finally, for
simplicity, capital does not depreciate, and since there is no growth or
government program spending, investment is zero in full equilibrium, and
so (in full equilibrium) total output and C are identical. Firms have a
Cobb-Douglas production function (with capital's exponent being 0), and
they hire labour and capital so that marginal products equal rental costs.
The full-equilibrium version of the model is described by the following
equations:

r=p
(N 1(1— N)) = (1— a)(1— 0)1(a(1+ r + r))
C = Ke N"
OC I K =r

The first equation is the Ramsey consumption function when consumption

growth is zero (that follows from the household differentiating with
respect to variable C). The second equation is what emerges when the
labour supply function (that follows from the household differentiating
with respect to variable IV) is equated with the firms' labour demand
function. The third equation is the production function, and the fourth is
the other relationship that follows from profit maximization — that capital
is hired to the point that its marginal product equals the interest rate. The
final three equations can be used to determine how consumption,
employment and the capital stock respond to different inflation rates. As
noted, the government budget constraint is

r = zC

102
This equation states that lump-sum transfer payments, T, are paid to
individuals, and in aggregate, these transfers are financed by the inflation
tax.
With T endogenous, cutting inflation is unambiguously "good".
Lower inflation eliminates a tax that distorts the household saving
decision, and no other distortion is introduced by the government having
to levy some other tax to acquire the missing revenue. Consumption,
employment, output and the capital stock all increase by the same
percentage when the inflation rate is reduced. Specifically,

(dC 1 C) = N)1(1+ r + g))thr

Mansoorian and Mohsin calibrate the model with standard real-business-

cycle assumptions: a = 0.64, p = 0.042, 0 = 0.30 and they assume an
initial inflation rate of zero: IC = 0. These assumptions allow them to
evaluate the inflation multiplier. The result is that creating inflation of 2%
lowers steady-state consumption by 1.37%. This outcome is an annual
annuity. With no growth and a discount rate of 0.042, the present value of
this annual loss in consumption is 0.0137/0.042 = 32% of one year's level
of consumption. Most analysts regard this magnitude as quite large. The
policy implication is that even a two-percent inflation rate should not be
tolerated.
While we do not use a formal model to derive an estimate of the
transitional costs of lowering the inflation rate in this chapter, it is worth
pointing out what the magnitude of these costs turns out to be. Experience
has shown (see Ball (1994)) that we suffer an increase in the GDP gap of
about two percentage points for about 3.5 years to lower the inflation rate
by two percentage points. This means that we lose approximately 7
percentage points of one year's GDP to lower steady-state inflation by
this amount. The benefit-cost analysis says that the present value of the
benefits exceeds this present cost (32% exceeds 7%), so disinflation is
supported.
This is the standard defense for targeting zero inflation. It is an
application of the neoclassical synthesis. The long-run benefits of lower
inflation are estimated by appealing to our theory of the natural rate (New
Classical macroeconomics). The short-run costs are estimated by
appealing to a more Keynesian model of temporary deviations from that
natural rate that stem from temporary nominal rigidities (our Chapter 3
model if micro-foundations are not stressed, our Chapter 6 model if they
are). In diagrammatic terms, the reasoning is illustrated in Figure 5.1,
where we continue to abstract from any ongoing growth. New Classical
analysis is used to estimate the shift up in the potential GDP line that
103
accompanies disinflation, and the short-run synthesis model (in which
potential GDP is exogenous) is used to calculate the temporary drop (and
then later recovery) in actual GDP.

Figure 5.1 Implications of Lower Inflation

time

One unappealing aspect of this estimate of the long-term benefits of low

inflation is that it involves the assumption that the monetary authority can
dictate to the fiscal policy maker (and insist that the latter must cut
transfer payments in the face of disinflation). How is the analysis affected
if we assume that this is not possible? To explore this question, we add a
tax on wage income. Assuming (as we have already) that the wage equals
the marginal product of labour, we re-express the government budget
constraint as:

= 71-C + 4(1— O)C 1 N)N

There is one change in household budget constraint, since it must now

stipulate that households receive only the after-tax wage. This results in
only one change; the third equation in our list of the model's relationships
becomes:

(N 1(1— N)) = (1— a)(1 — 8)(1— t) /(a(1 + r + g)).

It is left for the reader to re-derive the effect on consumption of a change

in the inflation rate, with t being the endogenous policy instrument instead
of T. It is more difficult for disinflation to be supported in this case, since
one distortion (the wage-tax) is replacing another (the inflation tax).
Indeed, in this case, for the calibration assumed by Mansoorian and
Mohsin, it turns out that the "benefit" of lower inflation has to be negative!
104
Disinflation forces the fiscal authority to rely more heavily on a more
distortionary revenue source than the inflation tax. Similar results in more
elaborate calibrated models are reported in Cooley and Hansen (1991). It
would therefore appear that the public-finance approach (that simplifies
by using a New Classical model with no ongoing growth) does not lead to
a solid under-pinning for a zero inflation target.
A somewhat more reliable argument for choosing a very low
inflation rate (perhaps zero) as "best," concerns the effect of inflation on
savings in a growth context. Most tax systems are not fully indexed for
inflation. To appreciate why this is important, suppose you have a $100
bond that gives you a nominal return of 10%. Suppose that inflation is 5%,
and that the interest rate on your bond would be 5% if inflation were zero.
At the end of the year, the financial institution sends you a tax form
indicating that you received $10 of interest earnings, and the government
taxes you on the entire $10. In fact, however, only $5 of the $10 is interest
earnings. The other $5 is compensation for the fact that the principal value
of your investment has shrunk with inflation. An interest income tax
system should tax only interest income, not the saver's depreciation
expenses. An indexed tax system would do just this. Non-indexed tax
systems make inflation amount to the same thing as a raising of the
interest-income tax rate. Thus, inflation reduces the incentive to save, so
that individuals living in the future inherit either a smaller capital stock
(with which to work) or a larger foreign debt to service, or both.
According to this analysis, then, to avoid a lowering of future living
standards, we should pursue a "zero" inflation target. We evaluate this
line of argument more fully in the economic growth chapters (10-12) later
in this book. At this point, we simply assert what will be derived and
explained there. If there are no "second-best" problems, growth theory
supports the removal of all taxes on saving (such as inflation when the tax
system is not indexed). But if there are second-best problems, this policy
is not necessarily supported. Again, the case for zero inflation is far from
complete.

5.5 Harberger Triangles vs. Okun's Gap

Thus far, much of this book has focused on explanations of the business
cycle and an evaluation of stabilization policy. It has been implicit that
there would be significant gains for society if the business cycle could be
eliminated. The standard defence for this presumption can be given by
referring to Figure 5.2 (where we now allow for ongoing growth by
drawing the (log of the) GDP time paths with a positive slope.
105
 Without business cycles, actual output, y, would coincide with the
natural rate, 5; , and both series would follow a smooth growth path such
as the straight line labelled 37 in Figure 5.2. But because we observe
business cycles, the actual output time path is cyclical — as is the wavy
line labelled y in the figure. Traditionally, Keynesians equated the natural
rate with potential GDP, and Figure 5.2 reflects this interpretation by
having the actual and natural rates coinciding only at the peak of each
cycle. Okun (1962) measured the area between the two time paths for the
United States for a several decade long period of time, and since the
average recession involved a loss of at least 5 percent of national output,
the sum of the so-called Okun gaps was taken to represent a very large
loss in material welfare. The payoff to be derived from a successful
stabilization policy seemed immense.

Figure 5.2 Output Gaps

time

Before explaining how the New Classicals have taken issue with
this analysis, it is useful to note how the likely payoff that can follow from
successful microeconomic policy was estimated back when Okun was
writing. As an example, consider a reduction in the income tax rate, which
(since it applies to interest income) distorts the consumption-savings
decision. In section 4.3, we derived the consumption function that follows
from inter-temporal optimization on the part of an infinitely lived agent
who is not liquidity constrained; the decision rule is:

r(I—t)— p,

106
where C, r, p, and t denote consumption, the real interest rate, the rate of
time preference, and an income tax rate that does not exempt interest
income (as we assumed in the previous section).
Traditional applied microeconomic analysis involved focusing on
full equilibrium without growth (that is, on the r(1 —t)= p relationship,
and combining this supply of savings function with the full-equilibrium
demand for capital ( FK = r + 8, where F(K,N) = Y is the production
function and 8 is the depreciation rate for capital). Assuming a fixed
quantity of labour employed, a Cobb/Douglas function, Y = K ° N" , and
that the rates of time preference and capital depreciation are independent
of tax policy, these relationships imply

(dY / Y) = [0n7((r + 8)(0 —1)(1— t))](dt / t).

With representative parameter values (t = 0.3, 0 = 0.36, r = 0.03, 8 = 0.1)

a 10 percent reduction in taxes (dt/t = — 0.1) involves an increase in
national output of just one-half of one percent. This once-for-all gain in
material welfare is just one-tenth the size of what Okun estimated to be
the benefit of avoiding one recession. Since the analysis of micro
distortions was often presented geometrically as a consumer surplus
triangle by public finance specialists such as Harberger, Tobin (1977,
page 468) concluded that "it takes a heap of Harberger triangles to fill an
Okun's gap." Thus, traditional Keynesians have felt confident that
stabilization policy was more important than microeconomic policy.
There has been a major change in thinking on these issues in
recent years. For one thing, analysts no longer feel comfortable with
equating the natural rate of output and potential GDP. A strict
interpretation of real business cycle theory involves the presumption that
there is no difference between the actual and the natural rate of output.
There are simply variations in the level of output that are caused by
stochastic elements in the production process. Some take a slightly less
doctrinaire view of the New Classical approach. According to this view,
there are differences between the natural and the actual output rates, and
the natural rate is what can be sustained on an average basis. The
economy operates below this level during downturns, and above this level
during booms. Thus, a proper drawing of Figure 5.2 involves shifting the
smooth natural rate line down so that it cuts through the mid point of each
up and down portion of the actual output time path. Using such a revised
graph to calculate the total Okun's gap over a period of years gives a very
different answer. Output losses are still incurred during recessions, but
these losses are approximately made up for by the output gains during
107
booms. A perfect stabilization policy would eliminate both the output
losses and the output gains. Thus, it is possible that, on balance, the net
benefit of (even a perfectly successful) stabilization policy is close to zero.

Figure 5.3 Diminishing Marginal Utility and Risk Aversion

Utility

Wage Income
$50 $100 $150

Even if gains and losses did cancel out, there would still be some
benefit to individuals as long as they are risk averse. That is, two income
streams with the same present value are not evaluated as equal in utility
terms if one income stream involves volatility. Figure 5.3 illustrates this
issue. It shows that with risk aversion, an individual refuses a fair bet — for
example, she refuses to pay $100 for the right to play a game in which
there is a 50-50 chance of receiving either $150 or $50. The expected
value of the game is $100, but — given the uncertainty — the utility that can
be derived from this expected value is not as big as what is enjoyed when
the $100 is certain. Thus, an individual with diminishing marginal utility
is willing to give up an amount of utility equal to distance DB to eliminate
the variability in her income stream. If the degree of risk aversion is very
slight, the utility of income function is almost linear, and distance DB is
very tiny. This is the reasoning that Lucas (1987) used in arriving at his
estimate of the value of stabilization policy. Using a time-separable utility
function with a constant coefficient of relative risk aversion (for which
empirical demand systems yield an estimate), Lucas was able to quantify
the benefits of eliminating variability, and as already noted, he concluded
that they were trivial.

108
 Keynesians have made three points in reacting to Lucas. The first
concerns whether there is market failure. According to New Classicals,
unemployment is voluntary, so when output is low it is because the value
of leisure is high. What is so bad about an "output loss" if it is just another
word for a "leisure gain"? But Keynesians think that a significant
component of unemployment is involuntary, since it stems from some
market failure such as asymmetric information, adverse selection, or
externalities in the trading process. (We examine these possibilities in
Chapter 8.) According to this view, smoothing is not the only result to
follow from stabilization policy. Indeed, just the commitment to attempt
stabilization may be sufficient to shift the economy to a Pareto-superior
equilibrium in models that involve both market failure and multiple
equilibria. Thus, stabilization policy can affect the mean, not just the
variance, of income. In terms of Figure 5.2, stabilization policy can both
reduce the wiggles in the y line and shift up its intercept. Lucas'
calculations simply assume that this second effect is not possible.
A second point concerns the distribution of the gains and losses
over the business cycle. A relatively small proportion of the population
bears most of the variability, so these individuals are sliding back and
forth around a much wider arc of their utility function (than Lucas
assumed). Even staying within Lucas' framework and numerical values,
Pemberton (1995) has shown that this distributional consideration can
raise the estimated benefit of stabilization policy by a factor of eight. An
even bigger revision is called for if a different utility function is used.
Pemberton notes that many experimental studies have cast fundamental
doubt on the expected utility approach. Indeed, the equity-premium puzzle
implies that we cannot have confidence in utility functions like the one
Lucas used. When some of the alternatives are used to rework Lucas'
calculations, it turns out that business cycles do involve significant
welfare implications.
Thus far, our discussion of the relative size of Okun's gap and
Harberger triangles has ignored two things: what are the effects of tax
changes before full equilibrium is reached? and what are the effects (if
any) on the economy's average rate of growth? These issues can be
clarified with reference to Figure 5.4. As we have seen, a cut in the
interest-income tax stimulates savings. As a result, current consumption
must drop, as shown by the step down in the solid-line time path in the
left-hand panel of Figure 5.4. Individuals must suffer this lower standard
of living for a time, before the increase in the stock of capital (made
possible by the higher saving) takes place. Our illustrative calculations
have estimated the long-term gain (the step up in the dashed line in the
left-hand panel of the figure) but not this short-term pain. Thus, the
109
comparison of Okun gaps and Harberger triangles is not complete without
a dynamic analysis of tax policy (which is provided in Chapters 10 and
12). But we must also note that a tax policy which stimulates savings may
not just cause a once-for-all increase in the level of living standards. It
may raise the ongoing growth rate of consumption, as shown in the right-
hand panel of Figure 5.4 There is still a period of short-term pain in this
case, but the effect on the present value of all future consumption can be
much more dramatic. Whether tax policy can have any effect on the long-
run average growth rate has been much debated in recent years, and this
debate is covered in the final two chapters of the book. But if it can, we
would have to conclude that the size of Harberger triangles may be far
bigger than earlier analysts had thought.
Does this mean that Lucas is right after all — that microeconomic
policy initiatives are more important than stabilization policy? Not
necessarily. As Fatas (2000), Barlevy (2004) and Blackburn and Pelloni
(2005) have shown, there is a negative correlation between the variance of
output growth and its mean value. It seems that a more volatile business
cycle is not conducive to investment, and so it contributes to a smaller
long-run growth rate than would otherwise occur. Thus, endogenous
growth analysis raises the size of both Harberger triangles and Okun gaps.

Figure 5.4 Effects on Consumption of Lower Interest-Income Taxes

In C In C

Time Time
Old Growth Theory New Growth Theory

It seems that a prudent reaction to the Okun gap vs. Harberger triangle
debate is to take the view that the profession should allocate some of its
resources to investigating both stabilization policy and long-term growth

110
policy (eliminating distortions). As we shall see in later chapters, the same
analytical tools are needed to pursue both tasks.

5.6 Conclusions

Real business cycle theorists have convinced all modern macroeconomists

of the value of explicit micro-foundations, and as a result they have made
modern work much more rigorous than what preceded their challenge to
that earlier literature. And now that New Classicals have acknowledged
that some form of nominal rigidity needs to be part of their model, we
have a convergence of views. Over the last several years, a particular
version of nominal rigidity (Calvo (1983)) — that involves micro-
foundations — has been added to the New Classical approach. The
resulting "New Neo-Classical Synthesis" model (a real business cycle
system with temporarily sticky prices) is quite similar to what New
Keynesians had been developing independently. Indeed, when some other
key features of New Keynesian work (for example, real wage rigidity
stemming from incomplete information (discussed in Chapter 8)) are
added to the real business cycle framework, the empirical applicability of
the synthesis approach is enhanced even more. It is to these developments
that we turn our attention in the next several chapters — before shifting our
focus to long-term growth theory.

111
Chapter 6

The New Neoclassical Synthesis

6.1 Introduction

In this chapter, we analyze what has been called the "New Neoclassical
Synthesis" in macroeconomics. As noted in earlier chapters, this approach
attempts to combine the best of two earlier schools of thought. First, it is
consistent with the empirical "fact of life" that prices are sticky in the
short run (the Keynesian tradition). Second, it is based on the presumption
that the Lucas critique must be respected. That is, it is in keeping with the
demands of the New Classicals; both the temporary price stickiness and
the determinants of the demand for goods must be based on a clearly
specified inter-temporal optimization.
This synthesis involves the basic (infinitely lived representative
agent) version of the inter-temporal theory of the household (derived in
Chapter 4, section 3) to re-specify the IS relationship, and the similar
theory of the firm (derived in Chapter 4, section 5) as a basis for a re-
specified Phillips curve. The traditional (or "old") IS relationship involves
the level of aggregate demand (output) depending inversely on the interest
rate, and the traditional Phillips curve involves the level of the inflation
rate depending positively on the output gap. When we derived the micro-
based IS and Phillips curve relationships in Chapter 4, they appeared
rather different: j). = (r — 7) and p -, 0(y — T)) . Thus the "new" IS
relationship involves the change in aggregate demand depending
positively on the interest rate, and the "new" Phillips curve involves the
change in the inflation rate depending inversely on the output gap.
We investigated one aspect of monetary policy in a model
involving these "new" relationships in a rational-expectations setting in
Chapter 3 (section 5). Since that analysis was rather messy, in this chapter
we simplify in three ways. First, we ignore stochastic shocks, so that
rational expectations becomes the same thing as perfect foresight. Second,
we use a continuous-time specification, so that a geometric approach —
phase diagrams — can be used instead of algebra. Third, for most of the
chapter, we consider only one aspect of the new model at a time — initially,
the new IS relationship with a traditional Phillips curve, and then a new
Phillips curve with a traditional IS relationship. In each case, we wish to
explore how (if at all) these changes in the model's specification affect the
answer to a standard stabilization policy question: what happens to output
when the central bank embarks on a disinflation policy?
112
6.2 Phase Diagram Methodology

Chapter 2 focused on the first Neoclassical Synthesis — a model that

combined traditional IS and Phillips curve relationships that were
descriptive, not based on formal inter-temporal optimization. In somewhat
modified notation, when a monetary-policy reaction function is added,
that system can be defined by equations (6.1) — (6.3):

.) = —V(r — (6.1)
= 0(y - 5) + (6.2)
r+p=7+±+2(p—x) (6.3)

The first equation (the aggregate demand function) states that output falls
below its full-equilibrium value when the real interest rate rises above its
full equilibrium value. The second equation (the dynamic aggregate
supply function) states that inflation exceeds the authority's target
inflation rate whenever the actual rate of output exceeds the natural rate.
The third equation states that the central bank raises the nominal interest
rate above its full equilibrium value whenever the price level exceeds the
bank's target value for the price level, x. The slope parameters (the three
Greek letters) are all positive.
We focus on a contractionary monetary policy; the central bank
lowers its target value for the price level in a once-for-all, previously
unexpected, fashion. Further, we assume that — both before and after this
change — the bank did maintain, and will then revert to maintaining, a
constant value for that target variable (x). If we were to graph this
exogenous variable — the level of x as a function of time — it would appear
as a horizontal line that drops down in a one-time step fashion at a
particular point in time. At that very instant, the slope of the graph is
undefined, but both before and after that point in time, the slope, z, is
zero.
We are interested in knowing what the time graphs for real output
and the price level are in the face of this one-time contractionary monetary
policy. We learned how to answer this question in Chapter 2. We were
able to re-write the model as a single linear differential equation in one
variable, and from that compact version of the system, we could derive
both the impact effect on real output, and the nature of the time paths after
the policy change had occurred. Specifically, we learned that (as long as
the system is stable) there is a temporary recession (which is biggest at the
very instant that the target price level is cut). The output time path then
starts rising asymptotically back up to the unaffected natural rate line, and

113
 the temporary recession is gradually eliminated (see Figure 2.2, p. 28).
There is no jump in the price level; the Keynesian element of the synthesis
is that the price level is a sticky variable. But while it cannot "jump" at a
point in time, it can adjust gradually through time. In this case, it
gradually falls to (asymptotically approaches) the new lower value of x.
These properties represent the base for comparison in the present chapter.
We want to know if the output and price-level time paths follow these
same general patterns in a series of modified models.
The first modified model is defined by equations (6.1 a), (6.2) and
(6.3). The only change is that the traditional IS relationship is replaced by
the "new" IS function:

y = (r —7) (6.1a)

To analyze this system, we are unable to use the methods of Chapter 2.

This is because, with a second differential equation in the system, we
cannot reduce it down to anything simpler than a set of two first-order
linear differential equations. The purpose of this section is to explain how
this system can be analyzed — first graphically (in what is known as a
phase diagram) and then more formally.
The first step in deriving the phase diagram is to reduce the
system to just two differential equations that contain only the two
endogenous variables that we most want to focus on. In this case, since we
wish to highlight the output and price-level effects, we use the policy
reaction function to eliminate the interest rate in the new IS function (after
having used to Phillips curve to eliminate the inflation 'rate from the
central bank's reaction function). The result is:

= %(p - x ) -0 (y -y). (6.4)

Equations (6.2) and (6.4) represent the compact version of the model.
These relationships contain no endogenous variables other than the two
we are focusing on — y and p — and also the time rates of change of no
endogenous variables other than these same two. This is exactly the
format we need, if we are to draw a phase diagram with y and p on the two
axes. It is customary to put the "jump" variable — in this case, y — on the
vertical axis, and the sticky-at-a-point-in-time variable — in this case, p
—ontherizalxs.Wnowepihtusqaon(6.2)d
(6.4) to derive the phase diagram.
The goal is to draw two "no-motion" lines in a y-p space graph.
The (p = o) locus is all combinations of y and p values that involve p not

114
changing through time. The (j) = 0) locus is all combinations of y and p
values that involve y not changing through time. The model's full
equilibrium involves neither variable changing, so — graphically — full
equilibrium is determined by the intersection of the two no-motion loci.
Only that one point involves no motion in both variables. To draw each
no-motion locus, we must determine its three properties: What is the slope
of the locus? What precise motion occurs when the economy is observed
at a point that is not on this line? and How does this locus shift (if at all)
when each exogenous variable is changed? We now proceed to answer all
three questions for both no-motion loci.
The properties of the (p = 0) locus can be determined from the
/5 relationship — equation (6.2). When p = 0, this relationship reduces to
y = y-- , and this fact is graphed as the horizontal line in Figure 6.1,
labelled (p = o) . So we have already answered question one; the slope of

Figure 6.1 The Properties of the p = 0 Locus

/3 > 0
A.

/3 = 0
y
/3 < 0

the (p = 0) locus is zero. What motion takes place when the economy is
at a point off this line? The best way to answer this question is to assume
that that we are at such a point, say point A in Figure 6.1, and then
determine what equation (6.2) implies about point A. At A, actual output
exceeds the natural rate, and (according to equation (6.2)) p must be
positive at point A as a result. This is just a mathematical way of saying
that "p is rising", so we draw in a horizontal line pointing to the right in

115
this upper region of the diagram to show this rising price level. It may
seem tempting to put an upward pointing arrow in the graph, since we are
talking about "rising" prices. But we must remember that we are graphing
p on the horizontal axis, so a rising value means an arrow pointing east,
not north. Similar reasoning leads to our inserting a western pointing
arrow below the (p = 0) locus. We have now summarized what is
happening (with respect to the price level, at least) at every point in the
plane. p is rising when it is observed at values above the line, falling when
observed at points below the line, and not moving at all when observed at
points on the line. The third (and final) question of interest concerning the
(p = 0) locus: is Does this line shift when the central bank lowers the
value of x — its price-level target? Since this exogenous variable does not
appear in equation (6.2), the answer is simply "no"; this policy does not
shift the position of this no-motion locus. We now proceed to ask and
answer these same three questions for the (5; = 0) locus.
The properties of the (p = 0) locus follow from equation (6.4). To
determine the slope of this no-motion locus, we substitute in the definition
of no motion (p = 0) , and solve for the variable that we are measuring on
the vertical axis in our phase diagram, y:

Y = (2 1 0)P + — (2 1 0)x].

The slope expression in this equation is the coefficient on the variable that
is being measured on the horizontal axis. Since this coefficient is (X/4) > 0,
we know that the (5, = 0) locus is positively sloped. The intercept is the
term in square brackets. Within this term, the coefficient of x is — (X/4)) < 0,
so we know that the reduction in x must increase the vertical intercept of
the (5, = 0) locus. Thus, we know that the contractionary monetary policy
moves the positively sloped (p = 0) locus up on the page. To answer the
remaining question — what motion is involved when the economy is
observed at a point that is not on this locus, we must revert to equation
(6.4) — that does not involve our having imposed P = 0 . From (6.4), we
see that ay = -0 < 0, which means that the time change in y goes from
zero to a negative value as we move to a point off the line and above the
line. Thus, we label all points above the line as involving p < 0, and this
is what justifies the arrows pointing south in this region of Figure 6.2.
Similarly, we label all points below the line as involving p > 0, and, as a
result, we insert a northern pointing arrow — indicating this rising y motion
for all observation points below the line in Figure 6.2.

116
 Figure 6.2 The Properties of the Y = 0 Locus

y =o

Figure 6.3 Combining the y = 0 Locus and the p = 0 Locus

=0

/3 - 0

4.111-4
..** Saddle Path
p

There is an alternative way to realize that these are the appropriate arrows
of motion, and this is by considering points that are to the right or left of
the line, instead of above or below the line. To proceed in this way, we
derive aj,/op = 2 > 0 from equation (6.4). This sign means that the time
change in y goes from zero to a positive value as we move off the line to
the right. Thus, we label all points in this region as having the property

117
 (j' > 0), and we show this with the arrow that is pointing north. When
deriving a phase diagram, we can select either a vertical-displacement
thought experiment, or a horizontal-displacement from the line thought
experiment — which ever appears to involve the simpler algebra.
We are now ready to put both no-motion lines — with their
corresponding arrows that indicate what changes occur when the economy
is observed at points off these lines — in the same graph. This is done in
Figure 6.3. Point E is full equilibrium, since it is the only point in the
plane that involves no motion for either endogenous variable. The two loci
divide the rest of the plane into four regions, and the combined forces that
operate at all these other points are shown. These forces show that the
system has both convergent and divergent tendencies. If the economy is
ever observed at points in the north-east or the south-west regions of
Figure 6.3, for example, the system diverges from full equilibrium, and
we conclude that the system is unstable. This same conclusion is
warranted for many points in the north-west and south-east regions of
Figure 6.3. Readers can verify this by selecting an arbitrary "starting"
point in one of these regions and tracing the economy's possible time path.
If that time path misses point E, the trajectory enters either the north-east
or the south-west regions and instability is assured once again. But since
trajectories that start in the north-west and the south-east regions might
just hit (and therefore end at) point E, stability is possible. The dotted
negatively sloped line — labelled the saddle path in Figure 6.3 — shows this
possible stable path. So if this economy is ever observed on the saddle
path, there will be convergence to full equilibrium.
At this point, the outcome seems quite arbitrary; the system may
or may not involve stability. To resolve this problem, we apply the
correspondence principle. We observe, in the real world, that the
instability prediction seems not to apply. Thus, if we are to relate the
model to this reality, we must reject the unstable outcomes as
mathematically possible, but inadmissible on empirical applicability
grounds. Hence, we simply assume that the economy manages to get on
the saddle path. This is possible, since history does not pin down a value
for one of the two endogenous variables. The price level is pre-determined
at each point in time, but the level of output has been assumed to be a
variable that is free to "jump" at a point in time. These assumptions reflect
the Keynesian feature of the synthesis model in the (instantaneous) short
run. Even though prices are flexible as time passes, the quantity of output
can change faster than its price in the short run. We can summarize this
given-history constraint that operates on the price level, but not on output,
by adding an "initial conditions constraint" line in the phase diagram. This
is done in Figure 6.4.
118
 Figure 6.4 Jumping on the Saddle Path

Initial =0
Conditions
Constraint

/3 = 0

j
' 'Saddle Path
p

There are four loci in Figure 6.4. The intersection of two — the
(5/ = 0) and the (p = 0) lines — determines the full equilibrium of the
system (the long-run outcome — point E). The intersection of the other two
lines — the saddle path that cuts through the long-run equilibrium point
and the pre-existing initial conditions constraint — determines the short-
run outcome (point A). The saddle path is the line we need to jump on to if
outright instability (which is presumed to be irrelevant on empirical
grounds) is to be avoided, and the initial conditions line is what we can
jump along. The intersection (point A) is the only point in the plane that is
both feasible (given the historically determined starting price level) and
desirable. While this methodology offers no discussion of a decentralized
mechanism that would help individual agents coordinate to find point A, it
assumes — on the basis of instability not being observed — that agents
somehow achieve this starting point. While this seems somewhat arbitrary,
it must be realized that some additional assumption is needed to complete
the model. After all, the two-equation system involves three endogenous
items: p, y and y. With y being a jump variable, both its level and its
time derivative are determined within the model. So an additional
restriction is needed to close the model, and this additional restriction is
that the system always jump on to the relevant saddle path, the moment
some previously unexpected event takes place. Any other assumption
renders the system unstable, and therefore unable to be related to the
(presumed to be stable) real world.
It is easier to appreciate how the model works by actually
following through a specific event. We do just this by referring to Figure
119
6.5. The economy starts in full equilibrium at point A — the intersection of
the initial no-motion loci. Then a once-for-all, previously unexpected,
drop in x occurs, as the central bank performs this unanticipated monetary
contraction. We have determined that this event shifts the (j) = 0) line up
to the left — to what is shown as the dashed line in Figure 6.5. The new
full equilibrium point is C, but the economy cannot jump immediately to
this point, because the price level is predetermined at a point in time. The
initial conditions line (always a vertical line if we follow the convention
of graphing the jump variable on the vertical axis) is the vertical line
going through the initial equilibrium point A. Since y is a jump variable,
the economy can move — instantaneously — to any point on this line. To
determine which point, we draw in the saddle path going through the new
full equilibrium point C. The intersection of this line with the initial
conditions line determines the immediate jump point — B.

Figure 6.5 Dynamic Adjustment Following a Reduction in x

Initial 0 after 14
Conditions
Constraint Initial y = 0

e
I e

p = 0 both before
and after . L x

B0 . Saddle Path
Through New
Full Equilibrium

The full solution is now summarized. The observation point jumps

immediately from A to B; then it travels gradually through time thereafter,
from B to C. So the contractionary monetary policy involves a temporary
recession, and that recession gets ever smaller thereafter as real output
asymptotically returns back to the natural rate. This is the same outcome
that emerged in the first neoclassical synthesis model (Chapter 2), yet we
have a new, not an old, IS relationship in the present system. This
similarity in results is "good news." Since many policy makers were
educated several decades ago — before the new synthesis with its more
thorough micro-foundations was developed — they find it difficult to shift
120
to a new paradigm. It appears that this may not be a serious problem. The
new synthesis analysis indicates that at least this one aspect of the earlier
policy analysis is robust. (This is not true for all policies, but at least it is
for some initiatives.)
Before analyzing related models with the phase diagram
methodology, it is worthwhile noting how a more formal approach can
yield slightly more precise predictions. In other words, it is useful to be
able to calculate impact and adjustment-speed effects in models of this
sort formally, rather than simply illustrating them geometrically. The
remainder of this section is devoted to explaining how this can be done.
As with all differential equations, the form of the solutions (for
the system defined by equations (6.2) and (6.4)) are:

p, = x+zi e6'` +z,e6" and y, = + z3 e6'' +z 4 e64( (6.5)

where the Ss are the characteristic roots, and the zs are determined by
initial conditions. Since we are restricting attention to saddle path
outcomes, we know that one characteristic root must be positive, and the
other negative. Let us assume 82 > 0 and gi < 0. By presuming a jump to
the saddle path, the unstable root is being precluded from having influence,
so z, = z, = 0 are imposed as initial conditions.
With a linear system, the equation of the saddle path must be
linear as well:

y, - .T ) = r (p, - x) (6.6)

where y is the slope of the saddle path. It is immediately clear that

equations (6.5) and (6.6) are consistent with each other only if z3 =yz,.
Thus, once on the saddle path, all motion is defined by

p,=x+zl e°1` and 3), = y + yz l e61: (6.7)

We substitute equations (6.7), and their time derivatives, into (6.2) and
(6.4) and obtain

1 = Or
(5
and (6; + = (6.8)

These two equations can be solved for the stable root and the slope of the
saddle path. The absolute value of the stable root defines the speed of
adjustment, while the slope of the saddle path is used to calculate impact
121
multipliers. For example, the impact effect on real output of the
unanticipated reduction in the target price level follows from the saddle
path equation (6.6):

(dy I dx)=(d) I dx)+ y[(dp I dx) —1]

This result states that the impact effect on y is a weighted average

involving the full-equilibrium effect on y (which we know, in this model,
is zero), and the impact effect on p (which is also zero in this model).
Thus, the impact effect on real output simplifies to — 7, the speed of
adjustment is — 8,, and the cumulative output loss is the former divided
by the latter. Using equations (6.8), this cumulative output loss expression
— which, for contractionary monetary policy is usually called the sacrifice
ratio — can be simplified to (1 /4)).
It is left for the reader to verify (using the methods of Chapter 2)
that when the new IS relationship that is involved in this model is replaced
by the old IS function, the sacrifice ratio involved with this contractionary
monetary policy is exactly the same. This fact adds to the assurance noted
earlier that the old and new synthesis models can have — but do not always
have — quite similar policy implications.

6.3 Stabilization Policy Analysis with a "New" Phillips Curve

The analysis in the previous section involved a model with a new IS

function and an old Phillips curve. In this section, we consider a model
with the opposite dimension of novelty — one with an old IS function and a
new Phillips curve. Thus, the model in this section is defined by equations
(6.1), (6.3) and (6.2a), where the revised equation is:

ij = —0( y — -
.)
7)
(6.2a)

In this case, it is most convenient to draw a phase diagram in ic-p space,

where it is the inflation rate. We put it on the vertical axis (since it is a
jump variable) and p on the horizontal axis. The definition of

=(Ir —0)

is the equation we use to derive the properties of the (p = 0) locus. It is a

horizontal line — drawn at the point where TC = 0 on the vertical axis in
Figure 6.6. There are rightward pointing arrows above this line, and
122
leftward pointing arrows below this line, indicating the forces of motion
when the economy's observation point is not on this line. Since x does not
enter this equation, this locus is not shifted when the central bank cuts its
target price level.

Figure 6.6 Properties of the 5; = 0 Locus and the p = 0 Locus

it = 0 Locus

0 /3 = 0 Locus

.1-1
\
Saddle Path
p

To obtain the equation that is needed to determine the properties of the

= 0) locus, we substitute (6.3) into (6.1) to eliminate the interest rate,
and then substitute the result into (6.2a) to eliminate the output gap. After
noting that p = it, the result is

= Ovg p - x
( ) - iffOr (6.9)

This equation has exactly the same format as (6.4), with it playing the role
in (6.9) that y plays in (6.4). This means that the (7i - = 0) locus in the
present model has all the same properties as the (5) = 0) locus had in the
model of the previous section (it is positively sloped, there are downward
pointing arrows of motion above the locus, and the position shifts left
when x is cut). Thus, the entire policy discussion proceeds in Figure 6.6 as
an exact analogy to that which accompanied Figure 6.5 above. To save
space, we leave it to the reader to review this discussion if necessary. We
simply close this particular analysis with a short discussion that links the
impact effect on it to the impact effect on y. Equations (6.1) and (6.3)
combine to yield

(.1' — TO= — VA(P "-

123
The phase diagram (Figure 6.6) proves that it falls when x falls. Given this
outcome, and the fact that p cannot jump, this last equation implies that y
must jump down. Thus, the predictions of this new-Phillips-curve model
align with those of the new-IS-curve model (of the last section) and of the
first synthesis model (of Chapter 2): there is a temporary recession that
immediately begins shrinking as time passes beyond the impact period.
We conclude, as in the previous section, that the predicted effects of
monetary policy are not specific to any one of these models.

6.4 Some Precursors of the New Synthesis

In this section, we outline a series of further alternative specifications for

the aggregate demand (IS) and supply (Phillips curve) relationships. This
survey allows readers to appreciate that versions of the new synthesis
were present in the literature before the "New Neoclassical" label became
widespread. The first variation is based on the fact that the new IS
function involves a fundamental limitation. The only component of
aggregate demand is consumption spending by households. As a
sensitivity test, we now consider a model that involves the opposite
extreme specification — that aggregate demand is driven totally by firms'
investment spending. As has been our practice in earlier sections of this
chapter, we compare this new specification to the original synthesis model.
Thus, we specify this investment-oriented demand model by a set of four
equations: the same old Phillips curve and central bank reaction functions
as used already (equations (6.2) and (6.3)) along with the following set of
two equations to replace equation (6.1).

(y — )7) = a(q —1) (6.1b)

r + fl(y — .57)+ q (6.1c)

These equations were explained in our derivation of firms' investment

behaviour in Chapter 4 (section 4). The first relationship states that
aggregate demand depends positively on Tobin's stock market valuation
ratio, q, and the second relationship states that the overall yield on stocks
is the sum of a dividend (the first two terms in (6.1 c)) and a capital gain
(the final term). In long-run equilibrium, the dividend is the "natural" real
rate of interest; with cycles in the short run, dividends are higher in booms
than they are in recessions.
To achieve a more compact version of this model, we proceed
with the following steps. Substitute (6.2) into (6.3) to eliminate the
inflation rate; substitute the result into (6.1c) to eliminate the interest rate;
124
take the time derivative of (6.1b) and use the result to eliminate the
change in the stock price from (6.1 c). We end with

= aA(p — x)— a(0 Y) (6.10)

The solution proceeds by drawing a phase diagram based on equations

(6.2) and (6.10). As before, a straightforward comparison of this set of
two equations to (6.2) and (6.4) indicates that the phase diagram is exactly
as drawn in Figure (6.5). Thus, as long as we confine our attention to
unanticipated monetary policy, there is no need to repeat the analysis.
However, since much can be learned by considering anticipated policy
initiatives, we proceed with this new use of the same phase diagram.

Figure 6.7 Anticipated Monetary Policy

y Initial k = 0 after 4,x

(9) Conditions .
Constraint :
,
.' Initial k =o
. ,
I , I
I -
I ,

e
/
e I
D e p = 0 both before
y •
e
e
e
and after 4-x

\ Saddle Path
Through New
Full Equilibrium
p

Initially the economy is as pictured in Figure 6.7 — at point A, the

intersection of the original no-motion loci. While the variable that is
calibrated and labelled on the vertical axis is national output, at some
points in the following discussion we interpret movements up and down in
the graph as increases and decreases in the stock market. This is
permissible since, given equation (6.1b), y and q move one-for-one
together. Assume that — starting from point A — the central bank cuts its
target for the price level. As usual, the (j, = 0) locus shifts left, point D
appears as the new full-equilibrium outcome point, and the saddle path
drawn through point D becomes relevant. If this monetary policy were
unanticipated, the economy would jump to the point of intersection
125
between this saddle path and the initial conditions line. Stock prices would
drop as soon as individuals realized that a temporary recession has been
created and — since this hurts profits and dividends — it is not rational to
pay as much for stocks as had been previously the case. The stock market
gradually recovers — as does the economy.
How does this scenario differ when the monetary contraction is
pre-announced? For example, suppose the central bank states that in
exactly one year's time, it will cut the target price level. Forward-looking
investors know that this will cause a drop in the stock market, and to
avoid these anticipated capital losses, they sell stocks at the very moment
that the future policy is announced. Each individual is trying to sell her
shares before anyone else does, to (at least in principle) avoid any capital
loss. But everyone is just as smart as everyone else, so this one-sided
pressure in the stock market to sell causes a drop in stock prices
immediately. Since the monetary policy has not yet been implemented,
there is no reason for stock prices to fall as much as they do in the
unanticipated policy case. Thus, stock prices (and the level of GDP) fall
by some intermediate amount — say to point B in Figure 6.7 — the moment
the policy is announced.
What happens during the one-year time interval between the
announcement of the policy and its implementation? First, since the
observation point of the economy is not on either of its no-motion loci, all
endogenous variables — real output, stock prices and goods prices — will
start to change. By locating point B in Figure 6.3, readers will see that the
joint force for motion is in the north-west direction. As a result, the
economy tracks along in this fashion until the observation point crosses
the (j) = 0) locus in Figure 6.7. Then, because Figure 6.3 involves south-
west motion in this region, the economy's time path bends down as it
approaches point C in Figure 6.7. With reference to the original full
equilibrium (point A) the economy is following an unstable trajectory. But,
if agents have perfect foresight (as we are assuming), they will have
chosen the location of point B so that the policy is implemented just as
point C is reached. At that very moment, the saddle path through point D
becomes relevant, and the economy begins to track along it from C to D,
after the one year time interval has expired. To summarize: the economy
moves instantly from A to B upon announcement of the policy, it moves
gradually from B to C between announcement and implementation, and
then gradually from C to D after implementation.
Of course, in reality, agents do not have perfect foresight. Thus, it
can turn out that the economy is not at point C the moment the policy is
implemented. In this case, the economy continues to follow an unstable
trajectory until people realize that the stock market is departing by an
126
ever-widening amount from the "fundamental" determinants of stock
values. At this point there is a "correction" in the stock market. If the
correction is in the downward direction, analysts refer to the "bubble"
bursting. By assuming perfect foresight, we are assuming that there are
never any bubbles in asset markets. This is not fully realistic, and it means
we may miss an extra wiggle or two in the actual time path. Nevertheless,
to maintain precision, it is customary to abstract from such bubbles. An
implication of doing so is that the implementation of the policy is not
"news". Everyone knows from the announcement that it will be
implemented, so its effect must be already fully capitalized into stock
values. This is why the assumption of perfect foresight forces us to ignore
bubbles.
This analysis suggests that we should exercise caution when
running Granger causality tests. These tests establish causality between
two time series by defining the causing influence as the item that happens
first. In this situation, the stock market is observed to cycle — first a drop,
then a rise, then another drop and a final rise. GDP follows the same
sequence. The money supply follows a very different time path. It shows
no action at all while q and y go through their first down-then-up-then-
down motion. After that, the money supply falls (this is what the central
bank does to raise interest rates) when the policy is implemented, and
immediately both the stock market and GDP rise. Time series evidence
would appear to be totally against Friedman's view — that monetary policy
is a fundamental determinant of variations in the stock market and
economic activity. The time series evidence would appear to support the
Keynesian view that the "animal spirits" of investors (as evidenced by the
swings of optimism and pessimism in the stock market) drive the business
cycle. But this interpretation of the evidence is entirely wrong. We know
that, in this model — by construction — monetary policy has caused
everything. Clearly, if forward-looking behaviour is important, the
validity of standard causality tests is threatened.
It is worth pointing out that there are other ways of bringing
forward-looking behaviour and asset markets into macroeconomics. One
branch of literature has distinguished long and short-term interest rates.
For this discussion, we define r to stand for the short-teim interest rate —
the one that is determined in the monetary-policy reaction function. We
use R to denote the long-term interest rate — the one that firms care about
if they use bonds, not stocks, to finance real investment. In this setting,
another equation is needed to close the macro model — a term-structure
relationship that pins down the arbitrage equilibrium that would satisfy
forward-looking investors. Letting the price of long-term bonds be q = 1/R,
we can derive that arbitrage condition. Investors have two options: they
127
can hold a short-tern bond to maturity (and earn r) or they can hold a long-
term bond for that same short period (and earn R-F4Ig). There is a
capital gain (or loss) term in the yield of a long-term bond over a short
period, since the bond does not mature within that holding period. Risk
neutral investors are not in equilibrium if the two bonds have different
yields. Imposing that equality, we have the term-structure equation we
need to close the model: RI R = R — r. The full system involves equations
(6.2 and 6.3) along with

— 7V)= —W(R — k)
R =k(R—r) .

The first of these relationships appears to be an old IS curve. The second

is a linear approximation of the arbitrage condition just derived (R = F is
the full-equilibrium value of the interest rate). It is left for the reader to
analyze this model with a phase diagram, as is required in one of the
practice questions suggested for this chapter. Before proceeding, however,
it is worth noting that this specification of aggregate demand essentially
nests what we have been referring to as the "old" and "new" IS
relationships. To appreciate this fact, take the time derivative of the first
of the equations on this page. Use the second to eliminate the time
derivative of R, and use the original version of the first to then eliminate
the level of R. The result is

= R[yf(r — 7)+(y — y,)].

There are three terms in this general specification; the last two are
involved in the "old" IS function, while the first two are involved in the
"new" IS function. It is interesting that models involving this quite general
relationship were being analyzed for a number of years before the New
Neoclassical synthesis was formally introduced in the literature (for
example, see Blanchard (1981)).
Another analysis that involves forward-looking agents and asset
prices within the demand side of the model is Dornbusch's (1976) model
of overshooting exchange rates. The open-economy equivalent of the
"old" IS-LM system is the Mundell (1963)-Fleming (1962) model.
Dornbusch extended this framework to allow for exchange-rate
expectations and perfect foresight. His goal was to make the model more
consistent with what had previously been regarded as surprising volatility
in exchange rates. Dornbusch's model involved the prediction that the

128
exchange rate adjusts more in the short run than it does in full equilibrium
— in response to a change in monetary policy. The following system is a
version of Dornbusch's model.
The small open economy model involves equations (6.2 and 6.3)
along with

(y — j7) , —v(r — T-) + 0(e — p)

r+P=7-Fe

The first of these open-economy equations is an old IS relationship; in

addition to the usual interest rate determinant of aggregate demand, there
is a terms of trade effect stemming from the net export component of
spending. e is the nominal exchange rate — the (logarithm of the) value of
foreign currency. The other new relationship defines interest arbitrage: the
domestic interest rate equals the foreign interest rate (and we have
assumed zero inflation in the rest of the world) plus the (actual equals
expected) depreciation of the domestic currency. The properties of this
model can be determined via a phase diagram. The compact version of the
system (needed to derive the phase diagram) is achieved as follows.
Combine the interest arbitrage relationship and (6.3) to yield e = .1(p — x).
Combine the open-economy IS function with (6.2) and (6.3) to yield
(1— y/0)(y — j7).—y/A,(p — x) + 0(e— p). Finally, take the time derivative
of this last relationship, and substitute in the second last equation. The
final result is

= (141- 0 ))[/10(p - x) - (0 + w2)0(y - 3)1 (6.11)

The phase diagram can be constructed from (6.2) and (6.11). As we have
encountered several times already, the form of (6.11) is the same as (6.4)
so the phase diagram analysis is very similar to what was presented in
detail in section 6.2. For this reason, we leave it for interested readers to
complete this specific open-economy analysis. However, before moving
on, we do note that (6.11) can be re-expressed, using (6.3), as a
relationship that involves just three variables: y (r —7) and (y — j7). Just
,

as we concluded with the interest-rate term-structure model, then,

Dornbusch's system appears to be a forerunner of the "old" vs. "new" IS
curve debate, since this last relationship nests both versions.
The final study that we include in this brief survey of work that is
related to the New Neoclassical Synthesis is Taylor's (1979b) model of
multi-period overlapping wage contracts (specified in change form, not in
levels as was the case in the most widely used version of Taylor's work).
129
This specification yields a model that is similar to one that involves
Calvo's (1983) specification of the new Phillips curve, yet it relies on a
descriptive, not a micro-based, definition of sticky wages. (For a more
detailed comparison of the Taylor and Calvo versions of the Phillips curve,
see Dixon and Kara (2006).)
Let w, denote the log of all wages contracted in period t; let t be
the proportion of contracts that are of one-period duration; let -41-T) be
the proportion of contracts that are of two-period duration; and so on.
These definitions imply that p„ the log of the overall wage index, is

Pt = r[w, + (1-1-)wi_t + (1— r) 2 w,_2 + ...J.

Writing this equation lagged once, multiplying the result through by (1-t),
and then subtracting the result from the original, yields

13,-1=n(wi — P1)

where 52 = r 1(1— r). With constant returns to scale technology, units can
be chosen so that the marginal product of labour is unity; thus p stands for
both the wage index and the price level. Each contracted w is set with a
view to the expected (equals actual) price that will obtain in the various
periods in the future, and to the state of market pressure in all future
periods (with the weight given to each period in the future depending on
the number of contracts that will run for two, three or more periods). Thus.
we have

= r[P, + (1— r)Pt+1 + — Pt+2 +... +0[Yt + ( 1 r)Yi+i + •••]

]

since we define the natural rate as zero. Writing this last equation forward
one period, multiplying the result by (1-T), and subtracting this last
equation from that result, we have

=C2(w, — — 0y1).

Continuous-time versions of these price and wage change relationships

can be written as:

=n(w- p) and
P 0Y) •

130
The full model consists of these two relationships and equations (6.1) and
(6.3). In continuous time, the length of one period is just an instant; thus, y,
r, and w are jump variables, while p is predetermined at each instant.
Defining v = w —p as an additional jump variable, we can re-express the
system (using the by-now familiar steps) as

V =12 tRa[2,(p — x)— Qv]

P =f2v

The first of these equations that define the compact system has the very
same form as equation (6.9). The second is analogous to the definition of
TE in section 6.3. This symmetry implies that this descriptive model of
overlapping wage contracts and the micro-based model of optimal price
adjustment have the same macro properties.

6.5 An Integrated Analysis: a "New" IS Curve and a "New" Phillips

Curve

Thus far, this chapter has proceeded in a piecemeal fashion; we have

analyzed models with either a new IS function or a new Phillips curve, but
not with both new relationships. If we put both new features in the same
system, it involves a first-order differential equation, p = (r , a
second-order differential equation, p = —0(y — , and one static
relationship — the central bank interest-rate setting equation. If we arrange
this system in the same format that we have followed in earlier sections of
this chapter, we would have to deal with a system involving three first-
order differential equations. Graphic analysis would then require a three-
dimensional phase diagram involving two jump variables and one sticky
variable. Such a graphic analysis is too cumbersome to pursue. We could
use an extended version of the mathematical approach that was explained
in the final paragraphs of section 6.2, but pursuing this more technical
approach has been rejected as beyond the intended- level of this book.
Even if readers did invest in mastering this more advanced mathematical
analysis, the model would still have important limitations. In particular,
we would still be unable to consider ongoing variations in the exogenous
variables — as opposed to one-time changes.
What other options do we have? One is to switch to a discrete-
time specification. We have already examined the full new synthesis
model with rational expectations methods in Chapter 3 (section 5). But

131
 that analysis was very messy as well. There is one other option — to stay in
continuous time and allow for ongoing changes in exogenous variables,
but to rely on the undetermined-coefficient solution method that was
explained in Chapter 2 (section 5) and used again in Chapter 3. We take
this approach in this remaining section of this chapter.
The exogenous variable that we want to model as involving
ongoing changes is autonomous spending. Thus, we must start by
indicating how the new synthesis model is altered slightly — compared to
what was derived in Chapter 4 — when this component of aggregate
demand is considered. It is left for readers to rework the Chapter 4
analysis to verify that — with an exogenous component of aggregate
demand, g, the new IS function becomes 5) a(r — + (1— a)g and the
"new" Phillips curve becomes /6 = —Ø(y — 57) + 18(g — . (1 — a) is the
full-equilibrium ratio of autonomous spending to total output (that is,
a = C I f, Y = C + G) , = 2(1— 0) 2 / + fi [(1— 0) 2 (1— a)]1(0a),
and 0 is the proportion of firms that cannot change their price each period.
While there seems to be little to justify assuming that there is no
exogenous component to aggregate demand, (a = 1), this is the common
assumption in the literature. We extend that literature in this section — first
by focusing on the basic new synthesis model, and then by considering
what has been called a "hybrid" model — that blends old and new
approaches.
To keep this section self-contained so that readers do not have to
keep referring back to equations that were defined much earlier in the
chapter, the full set of the model's relationships are listed together here.
The basic version of the model is defined by equations (6.12) through
(6.15). These equations define (respectively) the "new" IS relationship
(aggregate demand), the "new" Phillips curve (aggregate supply),
monetary policy (the interest-rate setting rule), and the exogenous cycle in
the autonomous component of demand.

= a(r — 7) + (1— a)g (6.12)

)3 = -- Ø(Y — TO+ fl(g — (6.13)
r+ F + 0 + .1(fi — 0) + (1 — .1)(p — 0) (6.14)
g = g + asin(t) (6.15)

The first two equations have already been discussed. Equation (6.14) is
the central bank's reaction function. We continue to focus on a bank that
is committed to price stability and this is why there is a zero inflation-rate
target term in the interest-rate-setting equation. But some of the remaining

132
details are slightly different here compared to earlier sections of this
chapter. We have chosen units so that both the (now constant) target price
level and the natural rate of output are zero. This monetary-policy reaction
function states that the bank sets the current nominal interest rate above
(below) its long-run average value whenever either the inflation rate is
above (below) its target value, or whenever the price level is above (below)
its target level.
As noted in earlier chapters, one major point of debate among
monetary policy analysts is whether central banks should pursue an
inflation-rate target or a price-level target. We consider this debate here by
examining alternative values for parameter k. Inflation targeting is in-
volved if X. = 1, while price-level targeting is specified by k = 0. We focus
on the implications of this choice for the economy's short-run built-in
stability properties. The existing literature has provided a thorough
investigation of this policy choice in the face of supply shocks, but there
has been nothing reported regarding demand shocks. It is partly to address
this gap in the literature that we focus on business cycles that are caused
by exogenous variations in autonomous demand, as defined by the sine
curve in equation (6.15). We proceed by deriving the reduced form for
real output, to see how the amplitude of the resulting cycle in y is affected
by changes in the monetary policy parameter (X).
To analyze the model, we first substitute (6.14) into (6.12) to
eliminate the real interest rate gap. Then, we take the time derivative of
the result, and use (6.13) to eliminate the change in the inflation rate.
Finally, we take one more time derivative and use (6.13) again, to
eliminate the remaining inflation-change term. The result is:

)7 = a(1— /1)00' — + a(1— .1)fl(g — g - g) + - (xi . (6.16)

As in Chapter 2 (section 5), we analyze (6.16) by positing a trial solution

and using the undetermined coefficient solution procedure. The solution
for output must take the following form:

y = j-; + B[cos(t)] + C[sin(t)] . (6.17)

Equation (6.17), the time derivative of (6.17), y = B sin(t) + C cos(t) , its

—

third time derivative, .5; = B sin(t) — C cos(t), and equation (6.15) and its
first and third time derivatives, k = S cos(t) and k = —8 cos(t), are all
substituted into (6.16). The resulting coefficient-identifying restrictions
are

133
 B = a(l - 2)[/35 ?E] /[I + 4( 1- 2 )]
C = 8[(1— a)(1+ aO(1— 2)) +(afl(1- A)) + fliba 2 (1— 2 )2 l 1
[(a0(1 - 2))2 + (1 + GOO- 2 W].
Illustrative parameter values are needed to assess the resulting amplitude
of the cycle in y. Representative values are: a = 0.8, (I) = 0.2 (which, given
the restrictions noted in the third paragraph of this section, imply (3 =
0.022) and 8 = 1.0. With these values, the amplitude of the real output
cycle is about 15% larger if the central bank targets the inflation rate (X =
1), than it is if the central bank targets the price level (X. = 0). According to
the model, then, the contemplated move from inflation-targeting to price-
level targeting is supported. Long-run price stability is achieved in either
case, and there is a small bonus with price-level targeting — there is a
slight reduction in real output volatility.
The reason why this issue requires a formal analysis is that there
are competing effects. These can be readily appreciated at the intuitive
level (as was noted in Chapter 2). Consider an exogenous increase in the
price level. With inflation-rate targeting, such a "bygone" outcome is
simply accepted, and only future inflation is resisted. But with price-level
targeting, future inflation has to be less than zero to eliminate this past
outcome. That is, only under price-level targeting is a policy-induced
recession called for. The reason that this consideration may not be the
dominant one, however, is that the avoidance of any long-term price-level
drift has a stabilizing effect on expectations. For plausible parameter
values, it appears that the former (destabilizing) effect is outweighed by
the latter (stabilizing) effect. As noted in Chapter 3 (section 5), Svensson
(1999) has interpreted this result as representing a "free lunch." The
present analysis supports Svensson.
But it was also noted in Chapter 3 that this free lunch proposition
is not robust. Walsh (2003b) has shown that it disappears when the
analysis shifts from the basic new synthesis model (that we have relied on
above) to a "hybrid" model — at least as far as supply-side shocks are
concerned. Walsh excluded demand shocks from his analysis. For the
remainder of this section, we outline what is meant by a hybrid model,
and then we determine whether Walsh's conclusion is appropriate for
demand disturbances (which is our focus here). We conclude that it is.
It is important that we consider some reformulation of the new
Phillips curve. This is because the strict version of this relationship makes
a prediction that is clearly refuted by empirical observation. Recall that
the strict version is j5 = —Ø(y — . Consider what this equation implies
during a period of disinflation — a period when the inflation rate is falling
134
(that is, when jj is negative). This Phillips curve predicts that the output
gap must be positive in such situations; that is that we must enjoy a boom
— not a recession — during disinflations! This is clearly not what we have
observed. Thus, while the new Phillips curve is appealing on micro-
foundations grounds, it is less so on empirical applicability grounds. To
improve the applicability without losing much on the optimization-
underpinnings front, macro-economists have allowed past inflation, not
just future inflation, to play a role in the new Phillips curve. This
extension avoids the counter-factual prediction concerning booms during
disinflations. Jackson (2005) has shown that a substantial degree of such a
"backward-looking" element is needed in this regard. Since hybrid
versions of the Phillips curve provide this feature, we consider them now.
Hybrid versions of the new synthesis model involve IS and
Phillips curve relationships that are weighted averages involving both the
new forward-looking specification and a backward-looking component.
The latter is included by appealing to the notion of decision-making costs.
It is assumed that there is a proportion of agents who find it too expensive
to engage in inter-temporal optimization. In an attempt to approximate
what the other optimizing agents are doing, the rule-of-thumb agents
simply mimic the behaviour of other agents with a one-period lag.
For example, on the supply side (using it to denote the inflation
rate and writing relationships in discrete time), the optimizers set prices
according to

irt = 7r,+1 ± — fl(gi — g),

while the rule-of-thumb agents set prices according to

A hybrid Phillips relationship can be had by giving each of these

component equations a weight of one half in an overall equation for TC.
After doing just that, and replacing first differences with time derivatives
to return to continuous time, we have the hybrid Phillips curve

= -.T)+ fl(g - (6.13a)

Several authors (Amato and Laubach (2003), Gali and Gertler (1999),
Walsh (2003b), Estralla and Fuhrer (2002), Smets and Wouters (2003),
Christiano et al (2005) and Jensen (2002)) have derived versions of a
hybrid that are equivalent to (6.13a) but which involve a more elaborate
135
derivation. Some lead to the proposition that the coefficient on the output
gap should be bigger in the hybrid environment than it is in the simpler
Calvo setting, while others lead to the opposite prediction. At least one,
Jensen, supports exactly (6.13a). There is a mixed empirical verdict on
alternative versions of the new hybrid Phillips curve. Gali et al (2005)
argue that most of the weight should be assigned to the forward-looking
component, while the results of Rudd and Whelan (2005) lead to the
opposite conclusion. Roberts' (2006) empirical findings support our 50-50
specification. Rudd and Whelan (2006) are quite negative on the general
empirical success of the hybrid models, while Mehra (2004) reaches a
more positive conclusion. Mankiw and Reis (2002) argue that a "sticky
information" approach is more appealing than any of the sticky price
models, and their approach is followed up in Ball et al (2005) and Andres
et al (2005). Given our inability to establish a clear preference for any one
of these hybrid price-change relationships, it seems advisable to use the
intermediate specification that (6.13a) represents, with wide sensitivity
testing on its slope parameters when reporting calibrated results with the
model.
We now explain a representative specification of a hybrid IS
relationship. The log-linear approximation of the resource constraint is
common to both the forward-looking and backward-looking components

y, = ac, + (1 a)gi .—

The optimizers follow the Ramsey rule

c, = c,,, — (r —7),

and the rule-of-thumb agents mimic what other agents did in the previous
period

C, = C,_1 .

Giving a one-half weight to each of these two decision rules, and

replacing first differences with time derivative as we switch to continuous
time, we arrive at the hybrid IS curve

jj = a(r — 17 ) + ( 1 — a)k. (6.12a)

As with the hybrid relationship on the supply side, the differences

between several authors' particular models — concerning the size of the
136
slope parameters, not the form of the equation — suggests that researchers
should allow for a fairly wide set of sensitivity tests when reporting
numerical simulation results.
We do not expect readers to derive the revised expressions for the
reduced-form parameters (B and C) in the trial solution (that are relevant
when our inflation-rate-vs-price-level targeting analysis involves both
hybrid functions). While the analysis is no more involved on conceptual
grounds in this hybrid case, the actual derivation is quite tedious. (With
the extra time derivatives involved in both the demand and supply
relationships, the analogue of equation (6.16) turns out to be a fifth-order
differential equation in this case.) Suffice it to say that the analysis
supports Walsh's contention, that the free lunch that is supposed to
accompany a switch from inflation-rate targeting to price-level targeting
seems to disappear in this hybrid setting (even with demand shocks).
Since empirical studies give stronger support for the hybrid relationships,
it seems advisable that we put more weight on this result — than on the
free lunch proposition that requires the original new synthesis model —
with no backward-looking behaviour involved — for analytical support.
Before ending this chapter, we briefly discuss several general
issues concerning the new synthesis model. First, it may strike readers as
odd that the modern analysis of monetary policy makes no reference
whatsoever to the money supply. McCallum (2001) has addressed this
concern. He allows money to play a role in reducing transactions costs (in
a manner that is more general than the strict cash-in-advance specification
that we considered in Chapter 5 (section 4)). When the household's
optimization is worked out in this setting, a money demand function
emerges as one of the items that is derived along with the consumption
function. The IS relationship contains real money balances as one of its
terms, and this creates an additional channel through which the interest
rate can affect spending. Using estimates of the interest elasticity of
money demand, McCallum calibrates the model and concludes that being
more complete along these lines makes only a trivial difference to the
model's properties and policy implications. Others who have investigated
this issue are Soderstrom (2002) and Nelson (2003).
Some researchers have raised the issue that there is more than one
channel for the monetary policy transmission mechanism. As noted in the
last chapter, higher interest rates can have a cost-increasing effect, not just
a demand-reducing effect. This can be important. A number of studies that
are based on the standard hybrid new synthesis model show that nominal-
GDP targeting can dominate inflation-rate targeting (for example, see
Kim and Henderson (2005)). Malik (2004) and Ravenna and Walsh (2006)

137
have shown that these policy implications can be affected when there is a
direct effect of interest rates in the new Phillips curve.
Finally, readers may be surprised by the following feature of the
new synthesis framework. Much attention is paid to the proposition that
the analysis starts from a clear specification of the agents' utility function.
Then, when policies are evaluated with the model, analysis seem to revert
to the "old" habit of ranking the outcomes according to which policy
delivers the smallest deviations of output and the price level from their
full-equilibrium values. Shouldn't an analysis that is based on an explicit
utility function (involving consumption and leisure as arguments) use that
very same utility function to evaluate alternative policies? Indeed, this is
the only approach that could claim to have internal consistency. It is
reassuring, therefore, that Woodford (2003a) has derived how the standard
policy-maker's objective function — involving price and output deviations
— can be explicitly derived as an approximation that follows directly from
the private agents' utility function that underpins the analysis.

6.6 Conclusions

When the New Classical revolution began in the 1970s, strong statements
were made concerning "old fashioned" macroeconomics. For example,
Lucas and Sargent (1979) referred to that work as "fatally flawed", and
King (1993) argued that the IS-LM-Phillips analysis (the first Neoclassical
Synthesis) was a "hazardous base on which to ... undertake policy
advice". But, as Mankiw (1992) has noted, that analysis has been
"reincarnated" (not "resurrected" — since the "new" IS-LM-Phillips
analysis certainly involves important differences in its structure from its
predecessor), and we are left with a framework that is now embraced as a
very useful base for undertaking policy advice. It is "good news" that this
analytical framework involves roughly the same level of aggregation and
abstraction as the older analysis, since this facilitates communication
between actual policy makers (who were brought up in the older tradition)
and modern analysts. In short, there is reason for much more optimism
than there was 30 years ago.
This is not to say that there is no controversy remaining. Indeed,
some macroeconomists still actively debate the relative merits of the "old"
and "new" specifications of the IS and Phillips curve relationships. This is
because, as noted above, the "old" relationships appear to be much more
consistent with real-world data, despite the fact that the "new" rela-
tionships are more consistent with at least one specific version of
microeconomics. It is frustrating to have to choose between the two
138
criteria for evaluating macro models — consistency with the facts and
consistency with optimization theory. It is for this reason that many
researchers are focused on developing the hybrid models that share some
of the features of both the "old" and "new" approaches.
One purpose of this chapter has been to make the reader aware of
these developments, and to thereby impart some perspective. But since we
wished to limit the technical demands on the reader, for the most part, we
limited the analysis to a set of "partially new" models. We have found that
some of the properties of these models are very similar. This is fortunate.
It means that policy makers do not have to wait until all the controversies
within modern macroeconomics are settled — before taking an initiative. It
is true that the answers to some questions are model-specific. For example,
the question as to whether the central bank should target the inflation rate
or the price level receives different answers as we vary the model. In
Chapter 3, using an "old" specification, we concluded that inflation-rate
targeting is preferred. In section 5 of the present chapter, we saw two
things — that a "new" model supports the opposite conclusion, and that a
hybrid model swings the support back in the direction of inflation-rate
targeting. Nevertheless, despite the sensitivity of our recommendations to
changes in model specification, policy makers can still proceed on this
issue. This is because, as already noted, the hybrid model seems to score
the best when evaluated according to both model-selection criteria
(consistency with both constrained maximization theory and data).
The main methodological tool that is used in this chapter is the
phase diagram. We will use this tool in later chapters as well. One of the
most interesting things that has emerged from our use of phase diagrams
is that the impact of government policies can be very different —
depending on whether private agents did or did not anticipate that policy.
This sensitivity can sometimes make it very difficult for macroeconomists
to test their models empirically.

139
Chapter 7

Stabilization Policy Controversies

7.1 Introduction

This is the second of a pair of chapters on the New Neoclassical Synthesis

approach to macroeconomic policy questions. The first chapter in this
sequence (Chapter 6) was concerned mostly with deriving the necessary
methods of analysis. This preparation makes possible the focus in this
chapter — we apply the new approach to four central questions of
stabilization policy. We begin by investigating inflation policy; in
particular, we consider whether disinflation should be applied in a "cold
turkey" fashion, or whether a more gradual approach to lower inflation is
more desirable. The second policy issue is also a monetary policy question
— the central one for open economies: How can the country achieve a
higher degree of built-in stability? By adopting a flexible exchange rate
policy or by forming a currency union with one's major trading partners?
The third and fourth questions are matters of fiscal policy. We investigate
the feasibility of ongoing bond-financed budget deficits, and the
desirability of balanced-budget rules, such as those embodied in Europe's
Stability Pact. This final analysis involves a similar focus to the optimal
exchange rate section. We are asking what leads to more built-in stability:
a budget balance that varies over the cycle (deficits during recessions and
surpluses during booms), or a budget balance that is independent of
variations in economic activity. We now proceed with each of these
analyses in turn.

7.2 Commitment and Dynamic Consistency in Monetary Policy

Perhaps the central question in monetary policy is Why should the central
bank target zero (or perhaps some low) rate of inflation? We analyzed this
question in Chapter 5 (section 4). We noted there that estimates of the
"sacrifice ratio" involved in disinflation have averaged about 3.5. This
means that a country incurs a one-time loss of about 3.5 percentage points
of GDP — in some year or other — for each percentage point of reduction
in the full-equilibrium inflation rate. We argued that it is difficult to be
dogmatic about what constitutes the optimal inflation rate, however, since
estimates of the benefits of lower inflation are not at all precise, and they
140
range both above and below this cost estimate. We add to that earlier
analysis in this section in two ways, and both extensions are motivated by
empirical considerations. For one thing, empirical studies suggest that the
short-run Phillips curve is flatter at low inflation rates. For another,
estimated sacrifice ratios vary a great deal, and one of the considerations
that appears to be important is how abrupt or gradual the disinflation
episode is (see Ball (1994)). In this section, we highlight analyses that
address these issues.
One of the reasons for flatter Phillips curves at low inflation rates
may be the fact that wages appear to be more "sticky" in the downward
direction than they are in the upward direction. The implications of this
proposition can be appreciated by considering the short run Phillips curves
shown in Figure 7.1. They are "curved" — that is, flatter in the lower
quadrant than they are in the upper quadrant, to reflect relative downward
rigidity.

Figure 7.1 Implications of a Non-Linear Short-Run Phillips Curve

Inflation

Natural (irs.„. 5%)

unemployment
rate Unemployment

Consider a set of shocks that moves the economy's observation point

back-and-forth along a given short-run trade-off line in this setting —
alternating between positive and negative inflation shocks of equal
magnitude. If an inflation rate of (say) 5% is chosen, and the shocks never
push the economy's observation point below the horizontal axis, the
average inflation rate will be 5% and the average unemployment rate will
be the natural rate. If a zero inflation target is chosen, however, shocks

141
will push the economy's observation point into the lower quadrant for half
of the time. Even if inflation averages zero, the fact that the Phillips curve
is flatter in the lower part of the diagram means that the unemployment
rate will average more the natural rate. The moral of this story is that —
with nonlinear short-run Phillips curves and stochastic shocks — there is a
long run trade-off after all. When we are "too ambitious" and aim too low
-

on the inflation front, we risk having a higher average unemployment rate.

This risk can be limited by choosing something like 2% as the
target inflation rate. This consideration, along with the limited support for
pushing inflation all the way to zero that was considered earlier in section
5.4, suggests that the common practice of adopting an inflation target in
the 2% range has some merit. (Yet another consideration that supports
such a target is the upward bias in the consumer price index. Since it is a
fixed-weight index, this measure "assumes" that individuals do not have
the opportunity of shifting away from relatively expensive items. Since
individuals do have this degree of freedom, the standard price index gives
an exaggerated impression of the true rise in the cost of living. The size of
this bias has been estimated to be about one-quarter to one-half of one
percentage point on an annual basis.)

Figure 7.2 The Superiority of Stabilization Rules Without Feedback

For expected
inflation of AC Short-run
trade-off
For expected lines
inflation of zero

142
The remainder of this section reviews Barro and Gordon (1983) who
apply the concept of dynamic consistency in policy-making to monetary
policy. Their model abstracts from the non-linearities and stochastic
shocks that have been the focus of the analysis in the last three paragraphs.
It emphasizes that monetary policy should deliver an entirely predictable
inflation rate — whatever target value is selected. The reasoning is best
clarified by considering the following two equations:

u = u * —a(g- — ge)
L = (ti — ju*) 2 + b(71- — 0) 2 .

The first relationship is an inverted expectations-augmented Phillips curve,

which stipulates that the unemployment rate falls below its natural value,
u*, only if actual inflation exceeds expected inflation. This relationship is
illustrated in Figure 7.2. The vertical line at u = u* indicates the set of
outcomes that are consistent with full equilibrium (that is, that
expectations be realized). The fact that this line is vertical means that
there is no trade-off between unemployment and inflation in the long run.
The family of negatively sloped lines (each with a slope equal to (— 1/a))
are the short-run trade-off curves. They indicate that — for as long as
individuals' expectations of inflation are fixed — higher inflation will
generate lower unemployment. As long as the central bank cares about
unemployment, this short-run Phillips relationship will represent a
temptation. By raising inflation, the bank can reduce unemployment. The
problem is that — by raising inflation above what people were previously
forecasting — the bank will cause problems in the longer run. Rising
expectations of inflation shift the economy to a higher short-run trade-off
line. In the end, citizens receive short-term gain (lower unemployment) in
exchange for long-term pain (higher inflation).
A more formal analysis of this dynamic choice can be had by
defining a specific objective function for the central bank. That is done in
the second equation. It states that the bank suffers losses (L) whenever the
unemployment rate differs from its "best" value (ju*), and whenever
inflation differs from its "best" value (assumed to be zero). To ensure that
losses are incurred whether these targets are missed on both the high and
low sides, the loss function stipulates that it is the square of each deviation
that matters. Parameter b represents the relative importance citizens attach
to low inflation — as opposed to sub-optimal levels of unemployment. We
assume that parameter j is a positive fraction. This ensures that the
invisible hand has not worked. The unemployment rate that the market
generates (on average, u*) is higher than the value that is socially optimal

143
 (ju*). Only if this is assumed can we interpret the unemployment as
"involuntary." (The interpretation of unemployment as involuntary is
discussed in much greater detail in Chapter 8. The analysis in that chapter
provides a rigorous defense for the proposition that it is reasonable to
specify parameter j as a fraction. This is important, since the analysis in
the present section breaks down ifj is not a fraction.)
The loss function is illustrated in Figure 7.2 by the indifference
curves. Curves that are closest to the point marked "best" represent higher
utility. Suppose that the economy is at point A with an undesirable level
—

of unemployment (the natural unemployment rate) and the desirable

amount of inflation (zero). The central bank will find it tempting to move
to a point like B. This point is feasible — at least for a while — since it is on
the currently relevant short-run trade-off line. This point is also desirable
— since it is on a preferred indifference curve than is point A. So the
central bank will be tempted to print money to generate the inflation rate
indicated by the height of point B.
The trouble with this strategy is that point B is not sustainable. It
involves actual inflation (positive) exceeding expected inflation (zero).
Individuals will be smart enough to revise upward their forecast of
inflation. This will move the economy to a higher trade-off curve, and so
to less desirable indifference curves. Actually, if individuals are clever,
they will be able to see immediately the point where the economy will
come to rest. After all, a full equilibrium requires two things. First, it must
be a point of tangency between a short-run Phillips curve and an
indifference curve — since (by choosing point B) the central bank will have
demonstrated that this is how it sets policy. Second, it must be a point
where actual and expected inflation are equal; otherwise private agents
will revise their forecast and that means the economy has yet to reach a
state of rest. The vertical line at u = u* represents all points that involve
actual inflation equaling expected inflation. Thus, the long-run
equilibrium is the point on this long-run Phillips curve that meets the
central bank's tangency requirement — that is, point C in Figure 7.2.
This reasoning proves that it is not advisable for the central bank
to try for point B by falsifying the expectations that individuals had at
—

point A. Once this attempt to move to B gets going, the system gravitates
to point C — which is on the least desirable of the three indifference curves.
The central bank will make the right decision if it focuses on the vertical
long-run Phillips curve. If it does so, and tries to achieve the point of
tangency between this constraint and the indifference map, the bank will
realize that point A is that best tangency point. By focusing on the long
run — and ignoring the fact that it can (but should not) exploit the fact that
individuals have already signed their wage and price contracts and so are
144
already committed to their inflationary expectations — the bank can deliver
point A (and this is better than point C). The moral of the story is that the
bank should focus on the long run, and not give in to the temptation to
create inflation surprises in the short run.
A formal proof of this proposition can be developed as follows.
The central bank has two options. First, it can be myopic and inter-
ventionist by revising its decision every period — capitalizing on the fact
that the inflationary expectations of private agents are given at each point
in time. Second, it can take a long-term view and be passive — just setting
inflation at what is best in full equilibrium and not taking action on a
period-by-period basis. In each case, the appropriate inflation rate can be
calculated by substituting the constraint (the Phillips curve equation) into
the objective function, differentiating that objective function with respect
to the bank's choice variable, the inflation rate, and setting that derivative
equal to zero. In the myopic version, inflationary expectations are taken as
given (and that fact can be exploited), while in the passive version, the
bank subjects itself to the additional constraint that it not falsify anyone's
expectations. In this case, the 7T = Ire constraint is imposed before, not
after optimization by the bank.

In the myopic case the first-order condition is

L / r7ir = 2(u — ju*)(—a) + 2bn- = 0,

and when the = 7z- e (which implies u = u * ) full-equilibrium condition is

substituted in, this policy generates an equilibrium inflation rate equal to
rc = au * (1— j)1 b . This expression gives the height of point C in Figure
7.2. When this value for inflation and u = u * are substituted into the loss
function, we see that losses under the myopic interventionist approach are:

L(myopic) = (u * (1— AY (1 + a' b).

In the passive case the first order condition is (by substituting = ,e and
u = u * into the loss function before differentiation):

oL 1 diz- =2bn- = 0,

so this policy generates 7z- = 0 (the height of point A in Figure 7.2) and
losses equal to:

145
 L(passive) = (u * (1— j)) 2

By inspecting the two loss expressions, the reader can verify that losses
are smaller with the "hands off' approach to policy.
Despite what has just been shown, individuals know that central
banks will always be tempted by point B. When the central banker's boss
(the government) is particularly worried about an electorate (that is
frustrated by unemployment), there is pressure on the bank to pursue what
has been called the "myopic" policy. Following Fischer and Summers
(1989), let parameter c be the central bank's credibility coefficient. If
people believe that there is no chance that the bank will be myopic, its c
coefficient is unity. If people expect the myopic policy with certainty, the
bank's c coefficient is zero. All real-world central banks will have a
credibility coefficient that is somewhere between zero and one. Thus, in
general, social losses will be equal to:

L = cL(passive) + (1— c)L(myopic), so

L = (u * (1— j)) 2 (1 + (1 — c)a2 I b).

This expression teaches us three important lessons: that losses will be less
with a higher c, a lower a, and a higher b. Let us consider each of these
lessons in turn.
Anything that increases central bank credibility is "good." This
explains why all central bankers make "tough" speeches — assuring
listeners that the bank cares much more about inflation than it does about
unemployment. It also explains why some countries (such as New
Zealand) link the central banker's pay to his/her performance (he/she loses
the job if the inflation target is exceeded). In addition, it explains why
particularly conservative individuals are appointed as central bankers -
even by left-wing governments, and why central banks are set up to be
fairly independent of the government of the day (such as in the United
Kingdom). Finally, it explains why developing countries, and countries
that have had a history of irresponsible monetary policy, often choose to
give up trying to establish credibility — and just use another country's
currency instead of their own.
Anything that makes the family of short-run Phillips curves
steeper (that is, makes the slope expression (— 1/a) bigger) is "good." Thus.
the analysis supports the wider use of profit sharing and/or shorter wage
contracts.
Finally, increased indexation is "bad." Twenty-five years ago,
when various western countries were debating the wisdom of fighting
146
inflation, some argued that it would be best to avoid the temporary
recession that accompanies disinflation. These individuals argued that it
would be better to include a cost-of-living clause in all wage, pension and
loan contracts. With indexed contracts, unexpected inflation would not
transfer income and wealth in an unintended fashion away from
pensioners and lenders to others. Thus, some people argued that
embracing indexation and "living with inflation" would have been better
than fighting it. Our analysis can be used to evaluate this controversy
since widespread indexation would change parameter b. If we had indexed
contracts, any given amount of inflation would result in smaller social
losses. Thus, more indexation means a smaller parameter b. But the final
loss expression indicates that a smaller b is "bad." This is because
indexation makes it more sensible for the central bank to be myopic.
There is always the temptation of point B, and indexation makes the cost
of going for B seem smaller.
We can summarize as follows. More indexation is "good news"
since it makes any given amount of inflation less costly. But more
indexation is "bad news" since it tempts the central bank to choose more
inflation. It is interesting (and perhaps unexpected) that — according to the
formal analysis — the "bad news" dimension must be the dominant
consideration. (This is why economists sometimes use formal analyses.
Algebra is a more precise tool than geometry, and sometimes this means
we can learn more from algebraic analysis. This analysis of indexation is
an illustration of this fact.) Overall, the analysis supports disinflation over
"living with inflation."
Before leaving this analysis of central bank credibility, it is worth
noting several things. First, the analysis does not prove that zero inflation
is best; it assumes it. Another point that is worth emphasizing is that what
this central bank credibility analysis rules out is policy that creates
surprises. It does not rule out discretionary policy that is well-announced
in advance. This distinction can be clarified as follows. Let the inflation
term in the loss function be the average inflation rate, g*, instead of just
this period's inflation:

L = (u — ju*) 2 + b(,r * —0)2 .

Assume that the central banker follows an ongoing reaction function:

z= * +d(u —u*),

147
which states that the bank raises aggregate demand (and therefore
inflation) whenever unemployment rises above the natural rate. It is
assumed that this reaction is announced in advance — as an ongoing
decision rule. Thus, it is a policy that is not based on causing private-
sector forecasts to be incorrect. If we substitute this policy reaction
function into a standard Phillips curve:

u =u*—h(g. — ze),

the result is

u =u*—a(n- * —ze)

where a = h 1(1 + dh). The analysis proceeds just as before — using the
revised loss function and this revised Phillips curve. The only thing that is
new is the revised interpretation of parameter a — that is defined in this
paragraph. As before, anything that makes a small is "good," and we see
that embracing discretionary policy that is announced in advance (moving
to a higher value for coefficient d) is therefore "good." Overall, then, this
credibility analysis supports discretionary policy — as long as that policy
is a pre-announced ongoing process, not a one-time event. This is why the
policy analysis in earlier chapters of this book have involved policy
reaction functions that were assumed to be fully understood by the agents
in the model.
More on central bank independence is available in Fischer (1995)
and McCallum (1995). As a result of this basic model of policy credibility,
we now have an integrated analysis of disinflation which combines
important insights from both Keynesian and Classical perspectives.
Keynesians have emphasized rigidities in nominal wages and, prices to
explain the recession that seems to always accompany disinflation. New
Classicals have stressed the inability of the central bank to make credible
policy announcements. Ball (1995) has shown that both sticky prices and
incomplete credibility are necessary to understand the short-run output
effects of disinflation. Indeed if incomplete credibility is not an issue, Ball
shows that disinflation can cause a temporary boom, not a recession. Only
a fraction of firms can adjust their prices quickly to the disinflationary
monetary policy. But if the policy is anticipated with full credibility, firms
that do adjust know that money growth will fall considerably while their
prices are in effect. Thus, it is rational for them to reduce dramatically
their price increases, and this can even be enough to push the overall
inflation rate below the money growth rate initially. This is what causes
148
the predicted temporary boom. However, since we know that such booms
do not occur during disinflations, we can conclude that policy credibility
must be a central issue.
The general conclusion to follow from this monetary policy
analysis is that the only dynamically consistent — that is, credible — policy
approach is one that leaves the authorities no freedom to react in a
previously unexpected fashion to developments in the future.
Discretionary policy that is not part of an ongoing well-understood rule
can be sub-optimal because "there is no mechanism to induce future
policy makers to take into consideration the effect of their policy, via the
expectations mechanism, upon current decisions of agents" (Kydland and
Prescott (1977, p 627)).
This argument for "rules" (with or without predictable feedback)
over "discretion" (unpredictable feedback) has many other applications in
everyday life. For example, there is the issue of negotiating with
kidnappers. In any one instance, negotiation is appealing — so that the life
of the hostage can be saved. But concessions create an incentive for more
kidnapping. Many individuals have decided that the long-term benefits of
a rule (no negotiations) exceed those associated with discretion (engaging
in negotiations). Another application of this issue, that is less important
but closer to home for university students, concerns the institution of final
exams. Both students and professors would be better off if there were no
final examinations. Yet students seem to need the incentive of the exam to
work properly. Thus, one's first thought is that the optimal policy would
be a discretionary one in which the professor promises an exam and then,
near the end of term, breaks that promise. With this arrangement, the
students would work and learn but would avoid the trauma of the exam,
and the professor would avoid the marking. The problem is that students
can anticipate such a broken promise (especially if there is a history of
this behaviour), so we opt for a rules approach — exams no matter what.
Most participants regard this policy as the best one. (For further
discussion, see Fischer (1980).)
We close this section by returning to disinflation policy. In this
further treatment of the issue, we integrate political and macroeconomic
considerations. We include this material for two reasons. First, it shows
how political uncertainty can be a source of incomplete credibility in
economic policy making. Second, it provides a framework for examining
whether gradualism in policy making is recommended or not.
Blanchard (1985a) has presented a simple model which highlights
this interplay. His analysis focuses on the possibility of two equilibria —
both of which are durable since both are consistent with rational
expectations. The essence of this model is that disinflation policy only
149
works if agents expect low money growth in the future. If a "hard-line"
political party is in power and it promises disinflation, two outcomes are
possible. On the one hand, agents can expect that the disinflation will
succeed and that the anti-inflation party will remain in power. Since both
these expectations will be realized, this outcome is a legitimate
equilibrium. On the other hand, agents can expect that the disinflation will
fail, and that (perhaps as a result) the anti-inflation party will lose power.
Since the change in government means that the disinflation policy is not
maintained, once again, the results validate the initial expectation, and this
outcome is another legitimate equilibrium. Blanchard formalizes this
argument, to see what strategy is advised for the hard-line party.
The basic version of the model is straightforward. Unemployment
depends inversely on the excess of the money growth rate, g, over the
inflation rate, Tc: u = —y(g — R -). The inflation rate is a weighted average of
the current actual money growth rate and what growth rate is expected in
the future, ge : it = cog +(1—yo)ge The hard-line party (which is in power
initially) sets g = g . The political opponent, the "soft" party, sets g = g*
where g* > k-.
If agents expect the hard-liners to stay in power, it = T. - so
unemployment is zero. If agents expect the soft party to gain power,
it = cok + (1— co)g * . Combining this relationship with the unemployment
equation, we have a summary of how unemployment and inflation interact:

u = —(1— co)(y/ yo)(7r — g*).

Any of these infinity of combinations of unemployment and inflation rates

are equilibria assuming the soft party wins power. In the simplest version
of his model, Blanchard assumes that the soft party does win power if
inflation is pushed so low that unemployment is deemed the more serious
problem; specifically the soft party wins if flu > tr. Using u = (1/ fl )11- to
eliminate the unemployment rate from the previous equation, we
determine the minimum inflation rate that the hard-liners can afford to
promise and still be certain of remaining in power:

it* =[(fly(1—yo))1(yo+ fly(1— v))1g*.

To recap — for inflation rates between k and z*, the hard-liners stay in
power and disinflation succeeds. For all inflation rates lower than 114 ,
there are two equilibria. If the hard-liners are expected to remain in power,

150
any chosen inflation rate below g* and u = 0 are the outcomes and the
hard-liners do remain in power. However, private agents may expect the
hard-liners to lose, and in that case we observe the same chosen inflation
rate along with positive unemployment (the value given by the second last
indented equation) and the hard-liners do lose power.
Three reactions to this two-equilibria feature are possible. One is
to interpret it as a theoretical under-pining for gradualism: the hard-liners
can avoid any chance of losing power (and therefore ensure that the two-
equilibria range of outcomes never emerges) by only disinflating within
the k to 71- * range. (This is Blanchard's reaction.) A second reaction is to
argue that the model should be re-specified to eliminate the two-equilibria
possibility. (Blanchard notes that this is achieved if the probability of the
hard-liners retaining power is less than unity even if unemployment never
exceeds zero.) The third reaction is to argue that we must develop more
explicit learning models of how agents' belief structures are modified
through time. (This is Farmer's (1993) reaction, as he exhorts macro-
economists to directly focus on self-fulfilling prophecies rather than to
avoid analyzing these phenomena.)
Blanchard's analysis is not the only one to provide a rationale for
gradualism. Our study of stabilization policy when the authority is
uncertain about the system's parameter values (Chapter 3, section 2)
showed that uncertainty makes optimal policy less activist. Drazen and
Masson (1994) derive a very similar result to Blanchard's, in an open
economy model that stresses the duration of the hard-line policy. It is
commonplace to assume that a central bank's credibility rises with the
duration of its hard-line position. This presumption does not always hold
in Drazen and Masson's study, since the larger period of (temporary)
suffering can so undermine support for the central bank's policy that it
becomes rational for everyone to expect that a softening of its stand must
take place. The bank can avoid this loss of credibility only by being less
hard-line in the first place. Drazen and Masson compare their model to a
person following an ambitious diet. When such a person assures us that he
will skip lunch today, is the credibility of that promise enhanced or
undermined by the observation that the individual has not eaten at all for
three days. On the one hand, there is clear evidence of the person's
determination; on the other hand, the probability is rising that the
individual will simply have to renege on his promise. Again, credibility
may be higher with gradualism.
For a more thorough sampling of the literature which integrates
macro theory, credibility and considerations of political feasibility, see
Persson and Tabellini (1994). A recent study that is similar to Blanchard's
is Loh (2002). Also, King (2006) provides a very general and concise
151
explanation of why discretionary policy creates a multiple equilibria
problem that a rules approach avoids.

7.3 Fixed vs. Flexible Exchange Rates

This section of the chapter uses the new neoclassical synthesis approach
to consider the implications for output volatility of alternative exchange-
rate regimes. A more detailed version of this material is available in Malik
and Scarth (2006).
The model is defined by equations (7.1) through (7.5). These
equations define (respectively) the new IS relationship, the new Phillips
curve, interest parity, monetary policy (assuming flexible exchange rates),
and the exogenous cycle in autonomous export demand. The definition of
variables that readers have not encountered in earlier chapters is given
following the equations.

=a(r —T)+12(/* +e-13)+ pa (7.1)

= i7)+ rOf - f + (e — (p — 13)) (a --a) (7 . 2)
r=T-+:f+e- 15 (7.3)
p + cluy = (7.4)
a =a- + S sin(t) (7.5)

The variables that may be less familiar are: a — autonomous export

demand, e — nominal exchange rate (the domestic currency price of
foreign exchange), and f — the price of foreign goods. As usual, p is the
price of domestically produced goods.
This open-economy version of the new IS relationship is based on
a log-linear approximation of the economy's resource constraint,
y = ac + /3a + (1— a — /3)x, where c is the log of domestic consumption
expenditure, a is the log of the autonomous part of exports and x is the log
of the part of exports that is sensitive to the real exchange rate. As is now
customary, the Ramsey model is used to define the behaviour of forward-
looking domestic households. Following McCallum and Nelson (1999),
the rest of the world is not modeled; it is simply assumed that
x = ( f + e — p). Equation (7.1) follows by taking the time derivative of
the resource constraint, substituting in the domestic consumption and the
export functions, and interpreting parameter n as 40 - a —(3).
As derived in Chapter 4, equation (7.2) is based on Calvo's (1983)
model of sticky prices. In this open-economy setting, we assume a
152
Leontief production relationship between intermediate imports and
domestic value added. It. is the unit requirement coefficient for
intermediate imports, and 0 is labour's exponent in the Cobb-Douglas
domestic value added process (so that y = On, and the marginal product
of labour, MPL, equals BY/N). The associated definition of marginal cost,
MC = W 1[MPL(1— (13(FE 1 P))] , is used in a derivation just like that
reported in section 4.5 to arrive at equation (7.2). We report only one
detail from that derivation; it yields the following restriction concerning
the summary parameters: Al / yi =1+ (a / 2)(2 — 0). This restriction
implies that 1(33.i. /W must exceed unity, and we rely on this fact below. For
a closed economy with no autonomous spending term, only the output gap
appears in the new Phillips curve. But in this open-economy setting, there
are direct supply-side effects of the exchange rate (stemming from the
intermediate imports) and a direct supply-side effect of the exogenous
variation in exports (stemming from using the resource constraint to
establish the link between real marginal cost and the output gap).
There is much discussion of the importance of exchange-rate
"pass-through" in the literature that compares the efficacy of fixed and
flexible exchange rates. One advantage of specifying imports as
intermediate products is that no independent assumption concerning
exchange-rate pass-through needs to be made. Indeed, the Calvo nominal
flexibility parameter also stands for the proportion of firms that pass
changes in the exchange rate through to customers at each point in time.
Equation (7.3) defines interest arbitrage. With perfect foresight,
the domestic nominal interest rate, r /9, must exceed the foreign
nominal interest rate, F. + f , by the expected depreciation of the domestic
currency, e. Price stability exists in the rest of the world (f = 7, f = 0),
so the domestic central bank can achieve domestic price stability in two
ways. One option is to fix the exchange rate. This option is best thought of
as opting for currency union, so it does not generate speculation about
changes in the value of the peg. This option is imposed in the model by
assuming e = e = e=0 and by ignoring equation (7.4)). The second option
is to peg a linear combination of the domestic price level and domestic
real output (imposed in the model by assuming equation (7.4)). Equation
(7.4) encompasses two interesting cases: targeting the price level Oa = 0),
and targeting nominal GDP (p.= 1).
We assume that business cycles in this small open economy are
caused by exogenous variations in export demand, as defined by the sine
curve in equation (7.5). We now proceed to derive the reduced form for
real output, to see how the amplitude of the resulting cycle in y is affected
153
by picking one exchange-rate regime or the other. We explain the
derivation of the reduced form for real output in the flexible exchange rate
case (and we assume that readers can use similar steps to verify the result
that we simply report for fixed exchange rates).
First, we simplify by setting f =7 = 1.0. Next, we take time
derivatives of (7.4) and use the result to eliminate the first and second
time derivatives of p. Then, we substitute (7.3) into (7.1) to eliminate the
interest rate, and use the result to eliminate the term involving the time
derivative of the exchange rate in the time derivative of equation (7.2).
The result is:

—,u)7 = + A26 (7.6)

AI = yp — A.+ (y 1(a + 51))(1— p(a + SI))

4 (y13 1(a + SI)) .

As we did in sections 3.5 and 6.5, we use the undetermined coefficient

solution procedure. The trial solution for output can be written as

y = 57+ B[cos(t)] + C[sin(t)] . (7.7)

The time derivatives of (7.7), y = —B sin(t) +Ccos(t) , =Bsin(t)—Ccos(t),

and the derivative of (5), a = Scos(t), are substituted into (7.6). The
resulting coefficient-identifying restrictions are B = 0, and

Cflex = 641(44 + ,u(a + (7.8)

A 3 = yi(a + CI) — yfl

A4 = A.(a + — y .

The similar expression for the fixed exchange rate version of the model 1 ,

fix = 8(A 3 +/3)/(A4 +1). (7.9) C

Since reduced-form parameter C represents the amplitude of the cycle

real output, it is our summary measure of output volatility. To evaluate the
relative appeal of flexible vs. fixed exchange rates, we compare
expressions (7.8) and (7.9). Subtracting the former from the latter, we
have:
154
 Cfi, — C , = [g(a + S2)(1+ pyt (a + + /3(1— y))(,82 — tif)]1
+ 114)(A4 (7.10)

The reader can verify that a sufficient, though not necessary, condition for
A4 to be positive is that intermediate imports be less than half of GDP.
Taking this to be uncontroversial, we make this assumption. Then, a
sufficient, though not necessary condition for expression (7.10) to be
positive is (112 tif)> 0. As noted above, the micro-foundations imply
that this restriction must hold. We conclude that the model supports a
flexible exchange rate. It is noteworthy that this conclusion is warranted
for quite a range of monetary policies followed under flexible exchange
rates — all the way from a hard-line approach (absolute price-level
targeting) to a policy that pays significant attentions to real-output
outcomes (nominal GDP targeting). Since each one of these options
delivers lower output volatility — with no trade-off in terms of long-run
price stability — the flexible exchange rate policy is supported.
There is a vast pre-new-synthesis literature on this question. The
original analysis — the Mundell-Fleming model — predicted that a flexible
exchange rate would serve as a shock absorber in the face of demand
shocks. As complications such as exchange-rate expectations and supply-
side effects of the exchange rate were added as extension to this
descriptive analysis, the support for flexible exchange rates — as a
mechanism for achieving lower output volatility — was decreased but (in
most studies) not eliminated. The analysis in this section allows for all
these extensions, and for the dynamic structure that is imposed by the
micro-foundations and inter-temporal optimization that forms the core of
the new synthesis approach. For those who have maintained a preference
for flexible exchange rates throughout this period, the robustness of our
conclusion is reassuring. As we found in the previous chapter, both old
and new analyses often complement each other — at the level of policy
implications.
A number of recent studies have focused on monetary policy in
small open economies. Kirsanova et al (2005) focus on the question of
which price level should the central bank attempt to keep on a target — the
price of domestically produced goods or the overall consumer price index.
The former is a sticky variable, while the latter has a jump dimension
since it is affected directly by the exchange rate (through its effect on the
price of imports). Similar closed-economy analyses have addressed the
question: should the price level or the wage level be targeted by monetary
policy? The general answer to these questions is that the central bank
should target whichever nominal variable is the stickiest. After all, welfare
155
costs arise when the sticky variable cannot adjust to its equilibrium value.
If the central bank never lets that variable get very far from that desired
value, its stickiness cannot matter much.
Devereux et al (2006) apply the new synthesis approach to models
of developing economies; they provide a detailed account of how optimal
monetary policy depends on the extent of exchange-rate pass-through.
Calvo and Mishkin (2003) offer a less model-specific discussion of the
options for developing countries. They argue that the choice between
fixed and flexible exchange rates is likely to be less important than
whether or not the country possesses good institutions and a set of sound
fiscal policies. Finally, Edwards and Levy (2005) provide a summary of
the empirical evidence. By comparing the group of countries that have
adopted flexible exchange rates with those that have not, they find that —
despite the differences in monetary policy strategies within the first group
— a flexible exchange rate acts as a shock absorber.

7.4 The Feasibility of Bond-Financed Government Budget Deficits

Government debt has accumulated in many western countries — both in

absolute terms and as a proportion of GDP. Many governments are trying
to contain this development in recent years, as they stabilize, and
sometimes reduce, their debt ratios. We explore these attempts in this
section.
Government debt would not have accumulated if governments
had taken the advice that follows from the standard theory of fiscal policy.
That advice is to run a deficit during the recessionary phase of each
business cycle and to run a surplus during the boom half of each business
cycle. According to conventional wisdom, this policy would help to
balance the economy over the cycle, and it would balance the budget over
the interval of each full cycle. As a result, this would be sufficient to keep
the debt from increasing over the long term.
For the last 30-40 years, governments have followed this advice
during recessions — indeed, they have appreciated the fact that economists
have condoned deficit spending during these periods. But governments
have chosen to disregard the other half of the advice from economists, for
they have not run surplus budgets for anything like half the time. By the
1990s, many governments had started to react to their exploding debt-to-
GDP ratios. Many have become so preoccupied with stopping the growth
in debt that they no longer make any attempt to help balance the economy.
Indeed, Europe's Stability Pact has been much criticized for forcing this
outcome on member governments.
156
 We provide a micro-based estimate of the long-term benefits of
deficit and debt reduction in Chapter 10. These benefits turn out to be
substantial. Here, we restrict our attention to an investigation of how it has
turned out that some governments have been able re-establish control of
their debt levels, while others have not. The basic issue is that short-term
pain must be incurred to achieve debt reduction unless we can somehow
"grow our way out" of the problem. It would seem as if this might be
possible, since the denominator of the debt-to-GDP ratio is growing
through time, and even with small annual deficits, the numerator may
grow at a slower rate. The governments that have achieved a turn-around
in their financial affairs are the ones that did not expect too much from
this "less painful" approach of hoping that growth could substitute for
tough budget decisions. To defend this assertion, we now compare two
fiscal policies — one that keeps a tight reign on the primary deficit, and the
other that keeps the overall budget deficit as an exogenous target. How do
such targets affect the stability of government finances?
To answer this question, we simplify by ignoring money issue, the
tax revenue collected on bond interest, and short-run deviations of real
output from its natural rate. If we define G and T as real program spending
and real taxes, i as the nominal interest rate, and B as the number of non-
indexed government bonds outstanding. The nominal deficit, D, is then
defined by

D = PG — PT + iB .

Using lower-case letters to define proportions of GDP: d = DIPY, g = GIY,

t = TIY, b = BIPY, we have

d=g—t+ib (7.11)

Under pure bond-financing, the deficit is financed through bond issue:

= D. Using the time derivative of the definition of the debt ratio, and
introducing notation for the inflation rate and the real GDP growth rate:
= P I P, n = Y I Y, this accumulation identity can be re-expressed as

= d — (n- + n)b (7.12)

We consider two fiscal regimes — first, one that involves the government
maintaining a fixed structural, or primary, deficit-to-GDP ratio. This
policy means that (g — 1) remains an exogenous constant, and the overall

157
deficit, d, is endogenous (determined in (7.11). Substituting (7.11) into
(7.12) to eliminate d yields

(g — 1)+ (r — n)b

where r=i—n- is the real interest rate on government bonds. Assuming

Ricardian equivalence, both the interest rate and the growth rate are
independent of the quantity of bonds outstanding. With this simplification,
we can use this last equation to determine aiiiab. Convergence to a
constant debt ratio occurs only if this expression is negative, so the
stability condition is n > r. With a fixed primary deficit, then, a program
of bond-financed fiscal deficits is feasible only if the economy's long-run
average real growth rate exceeds its long-run average real interest rate.
We assess the likelihood of this condition being met for much of the
remainder of this section. But first, we consider an alternative fiscal
regime.
Fiscal authorities can adopt a target for the overall deficit instead
of the primary deficit. This regime makes d an exogenous variable. In this
case the two-equation system defined by (7.11) and (7.12) becomes
segmented. The dynamics of the debt ratio is determined by (7.12) alone.
It implies that stability is assured if the nominal growth rate of GDP is
positive. As long as the central bank does not try for deflation, the
nominal growth rate is surely positive as a long-run-average proposition
(for example, any positive n along with 71" = 0 will do). Thus, the debt
ratio converges to b = d /(n + 7r). We assume that this fiscal regime is in
place when we use a calibrated model to estimate the size of the benefits
of deficit and debt reduction in Chapter 10.
The Canadian situation provides an interesting application of this
analysis. For the 1968-1983 period, the Liberal government of Pierre
Trudeau ran large primary deficits every year, and as this analysis
suggests, the debt ratio exploded. Then, for the next ten years (1983-1993),
the Conservative government of Brian Mulroney maintained an average
primary deficit of zero, hoping that this contraction in fiscal policy would
be sufficient to permit the country to grow its way out of its debt-ratio
problem. However, since interest rates exceeded the economy's growth
rate, the analysis suggests that this hope would not be realized. Indeed, it
was not; the debt ratio continued to increase in dramatic fashion. Finally,
the Liberals (under Jean Chretien and Paul Martin) returned to power for
thirteen years (1993-2006). During this period, the Liberals adopted a
strict target for the overall deficit ratio. After bringing that target down
over a five-year period, it remained constant (at a small surplus of about
158
 one half of one percentage point of GDP) thereafter. Again, just as the
analysis predicts, the debt ratio has fallen (by about 35 percentage points).
So with this policy, we can grow our way out of the debt problem. But
this more appealing outcome required much pain initially since getting the
overall deficit to zero was much tougher than eliminating the primary
deficit. In any event, we conclude that — as simple as it is — a model that
contains nothing but the basic accounting identities for government
budgets and debt provides an excellent guide for interpreting actual policy.
A final note about the Canadian episode is worthwhile. By the
turn of the century, the focus was on a much rosier scenario. People were
speaking of the "fiscal dividend," which referred to the extra room in the
budget created by the shrinking interest payment obligations. With the
prospect of the debt ratio falling from its peak to the government's
announced long-run target by 50 percentage points, analysts expect
significant interest payment savings. A drop in the debt ratio of 0.5 times
an interest rate of 0.05 means a saving of 2.5 percent of GDP every year.
There has been no shortage of people ready to advise the government
about how to spend this fiscal dividend — some arguing for better funded
programs and others advocating tax cuts. To have some idea of the
magnitudes involved in this debate, we should note that an annuity of 2.5
percent of GDP would permit a permanent elimination of one-third of the
federal government's total collection of personal income taxes. With so
much at stake, we can expect ongoing debate about how best to use the
fiscal dividend.
One question remains: Was the fact that interest rates exceeded
the economy's growth rate in the Canadian episode an unusual situation,
or should we expect that as normal? An extensive literature addresses this
question — known as the question of dynamic efficiency — and we discuss
that analysis for the remainder of this section.
McCallum (1984) analyzes a model involving agents with infinite
lives and money that enters the utility function; his conclusion is that the
conditions of optimizing behaviour preclude the growth rate's exceeding
the interest rate. Burbidge (1984) reaches the same conclusion with an
overlapping-generations model. The logic of this result follows from the
fact that in full equilibrium, the marginal product of capital equals the real
after-tax interest rate (as long as we ignore depreciation). If the growth
rate exceeds capital's marginal product, society has sacrificed too much
and over-accumulated capital. In such a situation, the return on capital is
an amount of future consumption that is less than the amount that can be
had by just allowing for ongoing growth, and the current generation of
consumers can gain (and no other generation of consumers will lose) by
consuming some of the nation's capital stock. An equilibrium based on
159
optimization involves exhausting such opportunities until the marginal
product of capital is driven upward.
The economy is said to be dynamically efficient if it is impossible
to make one generation better off without making at least one other
generation worse off. As we note in Chapter 10, Abel, Mankiw, Summers
and Zeckhauser (1989) have tested for dynamic efficiency. The essence of
their test can be explained as follows. Dynamic efficiency obtains if the
return on capital, FK , exceeds the real growth rate, n. Since all aggregates
must grow at the same rate in a balanced growth equilibrium, we have
K / K = n. Thus, the dynamic efficiency inequality, FK > n, can be re-
expressed as FK > K/K or FK K > K. In words, this last representation is:
profits must exceed investment. Abel et. al. compare profits and
investment for many countries and for many years, and they find that the
former exceeds the latter in every case. Thus, their analysis supports
dynamic efficiency. While aspects of this empirical analysis have been
questioned recently (Barbie et al (2004)), it is worth stressing two policy
implications that follow from the observation that profits exceed
investment. First, our economies have not accumulated enough capital to
maximize steady-state consumption per head, and (given this objective
function) any policy that stimulates saving can be defended on efficiency
grounds. (Much more detailed discussion of these issues is given in
Chapters 10-12.) The second implication of dynamic efficiency is that, as
we have noted already, bond-financed budget deficits must render the
macroeconomy unstable — at least if the rate of interest that is relevant for
government policy decisions is of the same order of magnitude as the
marginal product of capital.
It is clear that applied microeconomic policy decisions involve the
government using interest rates of this order of magnitude. Indeed, in this
context it is not at all appealing to assume that the growth rate exceeds the
interest rate. If such a proposition were taken as true by people who carry
out cost-benefit studies, they would calculate the present value of many of
the component costs and benefits as infinity! To see this point, suppose
that some cost or benefit item is estimated to be x percent of GDP. The
CO

present value of this item is then ixe ("-R) dt, where R is the after-tax real
interest rate. Clearly, if n > R, this present value is infinite. To avoid such
a situation in cost-benefit studies, analysts always assume R > n.
Does internal consistency demand that we make the same
assumption (R > n) when considering macroeconomic policy (for example,
when assessing the stability or instability of bond-financed budget deficits)

160
as we do when conducting microeconomic policy analysis (for example,
when conducting cost-benefit studies)? If so, we must conclude that bond-
financing of budget deficits with an exogenous primary deficit involves
instability, and so this fiscal regime should be avoided. But a number of
analysts answer this question in the negative. They note that different rates
of interest are involved in both firms' and the government's investment
decisions on the one hand, and in the financing of the government debt on
the other. Investment decisions involve risk, and it is customary to
account for risk in cost-benefit calculations by using a "high" discount
rate. Thus, it is not surprising that estimates of the marginal product of
capital are noticeably higher than the rate of return paid on a safe
government bond. It is possible for the average real growth rate to be
consistently below the former and consistently above the latter. If so, it
may be possible to argue that there is no inherent inconsistency involved
in following standard cost-benefit practice, while at the same time
maintaining that a bond-financed deficit — with exogenous program
spending and tax rates — is feasible.
It is difficult to react to this proposition formally, since our
discussion has not involved uncertainty, and so it contains no well-defined
mechanism for maintaining any gap between the marginal product of
capital and a risk-free yield. We can say that during the final quarter of the
twentieth century, government bond yields exceeded real growth rates in
many countries. In this environment, the entire distinction between safe
and risky assets ceases to be important since there is no doubt that bond
financing (along with an exogenous primary deficit) involves instability.
The literal take-off of government-debt-to-GDP ratios during much of that
time is, of course, consistent with this interpretation.
In the end, over long time intervals at least, it seems fair to say
that there is serious doubt concerning the relative magnitude of the growth
rate and the level of government bond yields. Bohn (1995) and Ball,
Elmandorf and Mankiw (1995) and Blanchard and Weil (2003) have
developed models of bond-financed deficits that involve explicit
modelling of uncertainty. Ball et. al. consider historical evidence for the
United States since 1871. They estimate that this experience is consistent
with being able, with a probability in the 80-90 percent range, to run
temporary deficits and then roll over the resulting government debt
forever. The country can likely grow its way out of a large debt problem,
without the need for ever having to raise taxes. As a result, the welfare of
all generations can be improved. This result is not certain. Nor does it
violate the evidence in favour of dynamic efficiency — since there is still a
positive probability that some generations will be worse off if growth is
insufficient to lower the debt -to-GDP ratio.
161
 Ball et. al. use their result to argue that the usual metaphor
concerning government deficits — that they are like termites eating one's
house — is inappropriate. The termite analogy suggests very gradual but
inevitable disaster. They advocate thinking of deficits like a homeowner's
decision not to buy fire insurance. Living standards can be higher as long
as no fire occurs, but there is a large adverse effect if the fire does occur.
They prefer the fire insurance metaphor because the occurrence of a fire is
not inevitable. Of course, reasonable people can disagree about the
likelihood of the fire. Since we are not sure what caused the productivity
slowdown in the mid-1970s, for example, it may be that extrapolating the
more rapid average growth rates of the previous century into the future is
ill-advised. Similarly, with world capital markets now integrated, and with
the high demand for savings stemming from the developing markets in
Asia and elsewhere, world (real) interest rates may rise above historical
norms for a prolonged period of time. In short, it may be quite imprudent
to put much weight on the 1945-1975 historical experience (which was an
historical anomaly since it involved n > r). If so, the termite analogy is not
so bad after all.

7.5 An Evaluation of Balanced-Budget Rules and the European

Stability Pact

Should the debt ratio be allowed to vary over the business cycle? One of
the central lessons of the Great Depression was that adjusting annual
spending and taxation with a view to maintaining a fixed budget-balance
target "come hell or high water" increases output volatility: spending has
to be cut and taxes raised as the economy slows down, which is exactly
the time we do not want that to happen. The Keynesian message was that
it is better to help balance the economy by balancing the budget over the
time horizon of one full business cycle, not over an arbitrary shorter
period such as one year. Thus, for at least a half century following the
Depression, it was assumed that a rigid annually balanced budget
approach was "obviously" to be avoided. But the Keynesian message has
been increasingly ignored in recent years. As the "hell or high water"
quotation from the Canadian finance minister in 1994 indicates,
governments have reverted to annual budget-balance targets that permit
only very small departures from a more rigid regime. Adoption of the
"Growth and Stability Pact" in Europe has applied similar pressure. As
The Economist magazine has editorialized:

162
 as the euro area faces the possibility of its first recession ... the
stability pact must not only preclude any fiscal easing but even
trammel the operation of fiscal 'automatic stabilizers.' That could
mean that these countries are required to increase taxes or cut public
spending even as their economies slow. That smacks of 1930s-style
self-flagellation" (Aug. 25, 2001, p.13).

Are The Economist's editorial writers correct or are they putting too much
stock in an "old" analysis that has not been modified to make it consistent
with modern standards of rigour? Is it, or is it not, appropriate for the
government to allow cyclical variation in its debt ratio by running deficits
during recessions and surpluses during booms?
There is a long literature assessing the usefulness of Keynesian-
style "built-in stabilizers." For example, 35 years ago, Gorbet and
Helliwell (1971) and Smyth (1974) showed that these mechanisms can
serve as de-stabilizers. Running a deficit budget during a downturn may
well decrease the size of that initial recession. But over time the
government debt must be worked back down, so the overall speed of
adjustment of the economy is reduced. The initial recession is smaller, but
the recovery takes longer. While this literature identifies this trade-off
between a favourable initial impact effect and an unfavourable persistence
effect, it is rather dated, in that expectations are not highlighted, and the
behavioural equations are descriptive, not formally micro-based.
We can re-interpret some of our earlier modelling to make this
analysis somewhat more up-to-date. Consider ongoing shocks, in the
context of the simple model involving perfect foresight and descriptive
behavioural relationships (discussed in Chapter 2, pages 32-34). If we
focus on the inflation targeting (X, = 1) case, we see that the amplitude of
the real-output cycle is C = Pa A Keynesian fiscal policy involves
taxation depending positively on output, and this (in turn) makes IS-curve
parameters a and 13 larger. We can appreciate this by considering the
relationship that lies behind the IS specification: Y = C((1—t)Y)+ 10+ G.
The total differential of this relationship implies that the coefficients in
y = —ar + fig must be interpreted as

a = dy 1 dr = dY 1 Ydr = —(1'I Y)/(1— C(1— t))

Q = dy I dg = (dY I Y)1(dG I G)=(G I Y)/(1— C'(1— t))

As is readily seen, the steeper is the tax function (the larger is t), the
smaller are summary coefficients a and 13. As a result, allowing for the
"built-in stabilizers" does lower output volatility. Thus, with a somewhat
163
more up-to-date analysis of fiscal policy (one that allows for the modem
treatment of expectations), support for Keynesian stabilizers re-emerges.
But, as we have seen concerning monetary policy, the analytical
under-pinning for any policy is now viewed as quite incomplete if that
analysis does not involve micro-based behavioural relationships. Lam and
Scarth (2006) have investigated whether the (undesirable) "increased
persistence" property of Keynesian debt policy is bigger or smaller than
the (desirable) decreased impact-effect property, when that regime is
compared to a rigid regime (involving a constant debt ratio) — in a setting
that respects the requirements of the modem approach to business cycle
theory. Here is a summary of their analysis.
Except for the presence of the (log of) government spending, the
model's new IS and Phillips curve relationships are standard:

= a(r(1- r) - p) + (1- a)g. (7.13)

= -0(Y - v(g (7.14)

The remaining equations of the model define how monetary and fiscal
policies are conducted. For monetary policy, an important consideration is
simplification. There is one time derivative in the new IS relationship, two
time derivatives in the new Phillips curve, and one time derivative needed
to define Keynesian fiscal policy (the change in bonds equals the current
deficit). Thus, we need a monetary policy that can reduce the order of
dynamics that is inherent in a macro model that involves both forward-
looking dynamically optimizing agents, and the dynamics of bond-
financed deficits. To this end, we assume that the central bank has a
constant target value for the price level, I-3-, and that the bank adjusts the
interest rate by whatever it takes to ensure that the actual price level
approaches this target according to the following relationship:

= --2(p - (7.15)

The bank ensures that the percentage change in the actual price is
proportional to the percentage gap between the actual and target price
levels. When the time derivative of this policy rule is substituted into the
new Phillips curve, (7.14), we have a convenient static relationship:

22 (15- - = 0(Y — v(g (7.16)

Another aspect of this specification of monetary policy is that it eliminates

the new IS relationship from having any fundamental role to play. Its
164
function is simply to determine (residually) the value for the interest rate
that the central bank has to set to deliver its monetary policy.
We consider fiscal policy in two stages — first with the govern-
ment always balancing the budget, then with the government following
Keynes' suggested approach. To be explicit about the government's
accounts, we need to define government bonds in a specific way. It is
convenient to assume that they are indexed consols. As a result, since each
bond is a right to receive one unit of goods every period forever, variable
B denotes both the quantity of bonds outstanding and the interest-payment
obligations of the government. The market price of each bond is (1/r).
With a balanced-budget rule, the bond stock stays constant, so
B = B , h= 0, and G = TY — (1 — r)B. T is the tax rate, and since we
take it as fixed (in both fiscal regimes), the balanced-budget rule forces
the government to allow program spending to vary over the cycle as real
output changes through time. In full equilibrium, government spending is
given by G = rY — (1 — r)B. Expressing the spending relationship as a
percentage deviation from equilibrium, we have

(g-g)=v(y-i-,) (7.17)

where = r /(G / FT) = r 1(1 — a). The fact that this relationship involves
government spending falling whenever the economy is in recession is
what leads the analysis to reject a balanced-budget rule and to support the
Keynesian approach.
We now turn to the specification of fiscal policy in a Keynesian
setting. In this regime, temporary budget deficits and surpluses are
permitted. But there is the possibility of instability in the debt-to-GDP
ratio, since the economy has a positive interest rate, and no ongoing real
growth. To avoid this problem, we specify that the Keynesian government
reduces spending — whenever debt rises — by what is necessary to avoid
instability. This policy is specified by G =G — y(B — TO. Recalling that
G = rf — (1 — r)B, we can specify government spending in the
Keynesian regime by

(g — = — 0(b — b) (7.18)

where f2 = 7(B 1 Y)I(G If). The government covers the temporary

deficits and surpluses by issuing and retiring government debt:

165
 (11 r)13 = G + (1— r)B —TY.

After substituting in the equations for both the actual and full-equilibrium
values for spending, and using the fact that b = B / B, this bond-issue
equation can be re-expressed as

[B I iir]b = (1 r y)(b
— — — (rY I B)(y — .T)).

If a Taylor-series expansion is taken of the left-hand side of this budget

identity, the coefficient of all the deviations that emerge from the square-
bracket part all have as their coefficients the full equilibrium value of
which is zero. Thus, the bond-issue equation can be simplified to

= ,y(b — 0(y — (7.19)

where x = F(1 — y) and 0 = Frf / B. This relationship is part of the

model when the Keynesian regime is in place. The fact that it involves
government debt being issued whenever the economy is in recession is
what forces the government to work that debt back down later on. This
need makes the recession last for a longer time, and (other things equal) is
what leads the analysis to reject the Keynesian approach.
Our strategy is to examine the output effects of a contractionary
monetary policy — a once-for-all unanticipated cut in the target price level,
p. We calculate the undiscounted sum of the deviations of real output
from its natural rate — that results from this contraction in demand (in each
fiscal regime). The regime that delivers the smaller overall output effect is
deemed to be the one that provides more built-in stability.
The model involving the balanced-budget rule is defined by
equations (7.15), (7.16) and (7.17), which determine y, g and p at each
point in time. The impact effect of the monetary policy on real output can
be had by substituting (7.17) into (7.16) to eliminate the government
spending variable, and then differentiating:

dy I clj5 = %2 1 (0 - 0v). (7.20)

We compare expression (7.20), the size of the initial recession caused by

the monetary policy, to the similar multiplier derived in the Keynesian
case (below). But first, we consider how rapidly this recession dissipates

166
when the rigid fiscal regime is in place. Since the three-equation system is
linear, all variables involve the same speed of adjustment. Given equation
(7.15), that adjustment speed is 2. The undiscounted output loss is the
impact effect divided by the adjustment speed, so the overall output
deviation is

[sum of output deviations] = 21(0 — ov). (7.21)

•

The model involving Keynesian fiscal policy is defined by equations

(7.15), (7.16), (7.18) and (7.19), which determine y, g and p and b at
each point in time. The impact effect of the monetary policy on real output
can be had by substituting (7.18) into (7.16) to eliminate the government
spending variable:

22 (I) — P) = 0(Y + vc-20 (7.22)

and then differentiating the result, (7.22), to get

dy 1 = 1 0. (7.23)

By comparing expressions (7.20) and (7.23), we see that the

contractionary monetary policy causes a bigger initial recession in the
balanced-budget regime. Now we compare the adjustment speed across
the two environments. To this end, we substitute (7.22) into (7.19) to
eliminate the output-gap term, and arrange the result and (7.15) in the
following matrix format:

b]' = A[(p — p) (b — b)1 1

—2
A 0
922 /0
fi = x +(OA /0)

There are no jump variables involved in this dynamic process, so both

characteristic roots must be negative for stability. Since the determinant
and trace of matrix A are the product and the sum of these two roots
(respectively), stability requires both det(A) > 0 and trace(A) < 0. Thus, 13
< 0 is necessary and sufficient for stability, and (as noted above) we

167
assume that this condition is met. The adjustment speed in this case is
— trace(A)12 = (2 — 16)1 2. This expression is definitely smaller than X,
the adjustment speed that obtains in the rigid fiscal regime.
We can summarize as follows: the Keynesian approach has both
desirable and undesirable features — it involves a smaller initial recession
but a slower speed with which that recession is dissipated. Since the
overall sum of the output deviations is a measure that gives weight to both
these features, we compare this expression across fiscal policy regimes. In
this Keynesian case, we have

[sum of output deviations] = 22 40(1 — (13 / 2))). (7.24)

By comparing expression (7.21) and (7.24), we see that reference to

representative parameter values is necessary if we are to assess which
fiscal regime delivers more built-in stability. A full analysis is provided in
Lam and Scarth. Here we simplify by assuming that the full-equilibrium
involves no debt, and by assuming that — when in the Keynesian regime —
the fiscal authority comes as close as possible to keeping its spending
constant by letting parameter get close to zero. Further, from the
calibration we explained in section 6.5, we know that it is safe to take
parameter v as approximately zero. These assumptions imply that the
cumulative output loss is twice as big when the Keynesian fiscal regime is
in operation. For plausible parameter values, then, we can conclude that
the adoption of rigid budget-balance rules may not involve any loss of
built-in stability after all — indeed this policy can improve matters. This
result is consistent with the empirical results of Fatas and Mihov (2003),
who study data for 91 countries and conclude that institutional constraints
on governments actually decrease output volatility. It appears that the
basic new neoclassical synthesis can rationalize these empirical findings.
As noted in earlier chapters, there is a debate in the theory of
monetary policy about whether a price-level targeting strategy represents
a "free lunch" — in that it may provide improved outcomes in terms of
both lower inflation volatility and lower output volatility. We have seen in
this section that a similar debate has emerged in the theory of fiscal policy.
A "free lunch" may be possible in this realm of policy as well if budget-
balance rules can deliver both more stable debt-ratio time paths in the
longer term and lower output volatility in the short term. As we have seen,
the aspect of the Keynesian approach that makes this counter-intuitive
outcome emerge is that the Keynesian policy creates a need for the
government to work the debt ratio back to its full-equilibrium value
following temporary disturbances. This need slows the economy's speed

168
of adjustment back to the natural rate. Historically, it has been difficult for
macroeconomists to evaluate dynamic issues when their models have had
only limited micro-foundations. With its solid grounding in inter-temporal
optimization, however, the new neoclassical synthesis gives analysts more
confidence in their ability to assess dynamic considerations of this sort.
The fact that the speed of adjustment outcome can be the quantitatively
dominant consideration in this fiscal policy debate means that what is now
mainstream macroeconomics offers support for maintaining Europe's
Stability Pact.

7.6 Conclusions

Some of the most central questions that have been debated in the
stabilization policy field over the years are: Is following a rule preferred to
discretionary policy? Is a flexible exchange rate a shock absorber? Are
rigid budget-balance rules required to avoid an exploding debt-to-GDP
ratio? Does the adoption of a budget-balance rule decrease the
effectiveness of fiscal built-in stabilizers? Now that modern macro-
economics has a new paradigm — the New Neoclassical Synthesis —
analysts are returning to these long-standing issues, and checking to see if
the conventional wisdom — which is based on earlier, less micro-based
models — is threatened by the sensitivity test that the new modelling
approach makes possible. The purpose of this chapter has been to explain
how these questions have been pursued within this new framework.
Some very central insights have emerged — such as the whole
question of dynamic consistency in policy making. Part of the chapter
focused on the financing of government budget deficits. We learned how
one common specification of bond-financed fiscal deficits is infeasible.
Not only does this help us interpret fiscal policy, it can bring useful
insights concerning monetary policy. For example, some economists have
relied on this result to offer an explanation of why the disinflation policy
of the early 1980s took so long to work. Sargent and Wallace (1981) note
that if the fiscal deficit is exogenously set and if bond financing can be
used only temporarily, a reduction in money issue today must mean an
increased reliance on money issue in the future. They show that, therefore,
rational agents would not necessarily expect the inflation rate to fall along
with the current money-growth rate. Even in the short run, in the Sargent
/Wallace model, inflation can rise as money growth falls.
The ultimate issue raised by Sargent and Wallace is which policy
variables — the fiscal or the monetary instruments — should be set
residually. In the 1980s, macroeconomic policy in many countries
169
involved imposing monetary policy as a constraint on fiscal decisions — an
approach that led to extreme reliance on bond-financed deficits and a
dramatic increase in debt-to-GDP ratios. The analysis in part of this
chapter indicates that it might have been more appropriate for monetary
policy to be determined residually, as Friedman originally proposed in
1948. As long as fiscal parameters are set so that the budget deficit
averages zero, this arrangement would control the longer-run growth in
the money supply and, therefore, the underlying inflation rate. It would
also have the advantage of avoiding the instabilities associated with
inordinate reliance on bond financing. But these benefits can be had only
if the fiscal policy record is consistent and credible. Friedman gave up
expecting this in 1959, and he switched to supporting an exogenous
monetary policy. Yet the analysis of this chapter suggests that this
exogenous monetary policy can involve a decrease in the built-in stability
properties of the macro-economy. Increased reliance on bond financing
can lead to instability (an unstable time path for the debt-to-GDP ratio).
This problem can be avoided by embracing a rigid rule for the annual
budget balance, and as we have seen, the rigid fiscal regime may not
worsen short-run macroeconomic performance appreciably.

170
Chapter 8

Structural Unemployment
8.1 Introduction

In earlier chapters, we examined cyclical (temporary) unemployment. For

example, when actual output falls below the natural rate of output,
unemployment is temporarily above its natural rate. In this chapter, we
ignore cycles; instead, we focus on equilibrium, or permanent,
unemployment. We assume that unemployment can persist in full
equilibrium for three reasons.
• Asymmetric information problems cause the equilibrium wage to
exceed the market clearing level.
• The market power of unions causes the equilibrium wage to exceed
the market clearing level.
• Transaction costs and friction in the labour market result in the
simultaneous existence of both unemployed individuals and unfilled
job vacancies.
We discuss each of these approaches to modelling the labour market in a
separate section of the chapter. Much of this work has been pursued by a
group of economists who are often called "New Keynesians" —
economists who believe that there is market failure in labour markets, but
who also believe that explicit micro-foundations are essential to rigorous
macroeconomics. Some New Keynesians have shown that real rigidites —
such as the phenomena we study in this chapter — can make any given
degree of nominal price rigidity more important from a quantitative point
of view. We end this chapter with a discussion of this complementarity
between the several strands of New Keynesian work. In the next chapter,
we return to the three alternative (non-competitive) models of the labour
market, and we focus on identifying fiscal policy initiatives that can be
expected to lower the natural unemployment rate in all three settings.

8.2 Asymmetric Information in the Labour Market: Efficiency Wages

To start, we consider structural unemployment that results from the first

of the three considerations listed above: incomplete information. Workers
know whether they are trying their best on the job, while employers can
monitor this worker effort only incompletely. This differential access to
171
information may lead firms to use wage policy as a mechanism fcr
inducing higher productivity from their employees. Since a qtapp:tl
worker is a good worker," firms can have workers who put forth hieic-
effort if they make it the case that workers really want to keep thcr
current job. But one price — the wage rate — cannot clear two marken.
Thus, if firms use the wage to "clear" the "effort market," society canna
use the wage to clear the workers market. Selfishness, in the form
profit maximizing firms (the invisible hand) does not deliver what un -
knowing social planner would arrange (full employment) because — at the
individual level — firms care about the productivity of their workers — nor.
the nation's unemployment rate. In this setting, therefore, the invisf: ,:t
hand breaks down.
This section of the chapter is devoted to following up on 141,,,-
simple idea. Because this approach focuses on worker efficiency.
known as efficiency-wage theory. Firms think of two considerations whe.-.7..
deciding what wage to set. The direct effect on profits is that a hiczh
wage raises labour costs, so high wages are undesirable. But there is
indirect effect on profits, since higher wages raise worker productiN
and so raise profits. If this second effect dominates — at least over
range of wage rates — it is in the firm's interest to raise the wage aby.:. =
the competitive level. When all firms do this, the market involi. es
unemployed individuals. This unemployment is involuntary, sirict
individuals cannot convince firms to accept an offer to work for less .
Firms reject such offers, since they represent a request that firms
wage that is inconsistent with profit maximization. Thus, one appealmz.
feature of efficiency-wage theory is that we have some rigour behind
common notion that unemployment is a public-policy problem.
We pursue these ideas more carefully by explaining Sumner s
(1988) theory of efficiency wages. Summers defines profits as follows:

profits = F(qN) — wN.

Profits equal sales of goods (produced via the production function. F

minus the wage bill (all items measured here in real terms).. For simplizi7
in this exposition, we follow Summers by ignoring capital, until
extended analysis in the next chapter. The overall quantity of labour is
measured in efficiency units — as the product of the number of workers... 1%.
times the effort expended by each worker, q. This worker quality index is
assumed to be higher, the bigger is the gap between what the worker gets
at this firm, w, and what she receives if she separates from this firm. We

172
denote this outside option as b, and we define it more fully below. Thus,
the index of worker effort is defined as:

q =((w b)1 b)°

—

If we wish to ignore the variability in worker effort in what is derived

below, we can set parameter a equal to zero. This makes q = 1, and the
entire mechanism of variable worker productivity is suppressed. But if a
is a positive fraction, the work-effort index falls below unity, and effort
varies directly with the gap between the current wage and the outside
option.
Firms have two choice variables — what quantity of workers to
hire, and what wage to set. We can determine the best values for both
these things by: substituting the variable-worker-effort constraint into the
definition of profits, differentiating the resulting objective function with
respect to the firm's two choice variables, N and w, and setting these
partial derivatives to zero. We now do just that. To decide on the best
level of employment, we work out

O profits 1 = — w = 0,

which states that firms should hire labour up to the point that its marginal
product has been pushed down to the rental price of labour (the wage).
This is the standard optimal hiring rule (encountered in basic price theory)
except that here, the marginal product expression involves the work-effort
index. To decide on the best wage, we work out

0 profits 1 Ow= F'Na((w— b)1 b)° -1 1 b — N = [F'qa 1(w — b)] —1= 0.

To see the intuition behind this wage-setting rule, we use the other first-
order condition to substitute out F'q = w and get

w = b/(1— a).

According to this rule, firms must set the wage equal to their workers'
outside option if there is no variability in worker effort (if a equals 0).
This is what is assumed in the standard competitive model of the labour
market. But with variable worker productivity (a > 0), it becomes optimal
to set the wage above the outside option — to induce workers to work hard
(to lower the probability of getting fired by shirking less).
173
 The implications for the unemployment rate can be determined
once the workers' outside option is defined. We assume

b = (1— u)w + ufw.

With u standing for the unemployment rate, there is a probability equal to

the employment rate (1 — u) that fired workers will find a job with another
firm (that in full equilibrium pays the same wage). There is a probability
equal to the unemployment rate that the individual will be without a job.
We assume a simple unemployment insurance program in which
individuals receive fraction f of their wage as an unemployment-insurance
benefit (without any waiting period or time limit) in this eventuality. The
b equation above defines the outside option as this. weighted average.
When this relationship is substituted into the wage setting rule, we have

u = a 1(1— f).

This solution for the unemployment rate teaches us three things. First,
unemployment is zero if there is no variability in worker effort (that is, if
parameter a is zero). Second, increased generosity in the unemployment
insurance system (a higher value for parameter .I) raises unemployment.
This is because higher unemployment insurance shrinks the relative pay-
off that individuals get by keeping their jobs. As a result, they choose to
shirk more. Knowing this, firms raise the wage in an attempt to lessen this
reaction of their employees. With higher wages, firms shift back along
their (downward sloping) labour demand curve, and hire fewer workers.
(By the way, this does not mean that unemployment insurance is "bad."
After all, with unemployment insurance, any one unemployment spell
hurts the individual less. It is just that there is a trade-off; this beneficial
effect induces an increased frequency of unemployment spells.) The third
implication of the solution equation for the unemployment rate follows
from the fact that it does not include a productivity term, F'. Thus, the
proposition that investment in education (that raises overall productivity)
would lower unemployment is not supported by this analysis. Higher
productivity is desirable because it raises the wages of those who already
have jobs, not because it brings more people jobs. This prediction is
consistent with centuries of economic history. Vast productivity growth
has led to similar increases in real wages, without any significant long-
term trend in the unemployment rate.
We have focused on this model so that we have at least one
rigorous framework for arguing that some unemployment is involuntary.
174
We will use the theory to evaluate how fiscal policy might be used to
lower unemployment in the next chapter. Before closing the present
discussion of efficiency-wage theory, however, it is useful to extend
Summers' analysis by considering optimization on the part of households
(not just firms). It is preferable that the household variable-work-effort
function be derived, not assumed. This can be accomplished by assuming
that households maximize (71w-i-(1--71)b— fibct ). it is the proportion of time
that the individual is employed in her current job, and this proportion
rises with worker effort: 7 = e. The first two terms in the objective
function define the individual's income; she receives w if she keeps her
current job, and she receives b if she does not. The final term defines the
disutility associated with putting effort into one's job. To be compatible
with the income components of the objective function, this term is scaled
by b. Since it is reasonable to specify that higher work effort increases the
probability of keeping one's job, but at a decreasing rate, W must be less
than one. Similarly, since it is appealing for higher effort to decrease
utility at an increasing rate, y must exceed unity. Household behaviour
follows from substituting in the constraint and differentiating the
objective function with respect to q. The result is the effort function given
above, if a is interpreted as 1 /(y — v) and units are chosen so that
1 yfl =1.
We pursue several policy implications of this model of efficiency
wages in the next chapter. But for much of the remainder of this chapter,
we consider alternative ways to model the labour market.

8.3 Imperfect Competition in the Labour Market: Unions

The literature on the macroeconomic effects of unions focuses on two

different models, which are referred to as the non-co-operative and the co-
operative theories of union/firm interaction. The non-co-operative model
involves firms maximizing profits defined by F(N) — wN. This is achieved
when the familiar condition F` = w holds. This labour demand function is
shown in Figure 8.1, as the locus of all points that are at the apex of an
iso-profit curve. The slope of the iso-profit lines can be derived by setting
the total differential of the definition of profits equal to zero. The slope of
an iso-profit line is dw I dN = [F'(N)— w]l N, which is positive for low
levels of N and negative for high levels of N (when the marginal product
is low).

175
 Unions are assumed to maximize the income of a representative
member of the group. This can be defined as w(N I L) + IT,(L — N) I L,
where N and L denote employment and size of union membership
respectively, w is the wage paid to union members if they are employed,
IT is the wage they receive if they are unemployed (which could be
interpreted as unemployment insurance), and NIL is the probability of
employment. The slope of the union's indifference curves is derived by
setting the total differential of the expected income definition to zero. The
result is dw I dN = —(w — cv-)I N, which is negative since the union wage
cannot be less than the individual's reservation wage.

Figure 8.1 Models of Union/Firm Interaction

Union
Indifference
curves

The union achieves the highest indifference curve by picking the wage
that corresponds to the point at which the labour demand curve is tangent
to an indifference curve (point A in Figure 8.1). Once the wage is set,
firms are free to choose employment. But since the union has taken the
firm's reaction into account, its members know that point A will be
chosen. We can derive what this model predicts concerning the real wage
and employment by including a shift variable — such as A in a revised
production function: AF(N). Comparative static predictions are calculated
by taking the total differential of the labour demand curve and the equal
slopes condition. It is left for the reader to verify that the model does not
predict real wage rigidity. One purpose of examining this model was to
see whether the existence of unions in labour markets leads to wage
ridigity and/or involuntary unemployment. Since the model contains no

176
explicit explanation of why individuals deal with the firm exclusively
through the union (are they forced to? did they choose to?), it is not
possible to say whether lower employment means higher involuntary
unemployment.
Let us now investigate whether wage rigidity occurs in the co-
operative model of union/firm interaction. The outcome in the previous
model is inefficient since there are many wage/employment outcomes —
all the points within the shaded, lense-shaped region in Figure 8.1 — that
can make both the firm and the union better off than they are at point A.
The co-operative model assumes that the two parties reach an agreement
and settle at one of the Pareto-efficient points that lie along the contract
curve. Completing the model now requires some additional assumption
that defines how the two parties divide the gains from trade. The
additional assumption that is most common (see McDonald and Solow
(1981)) is that the two bargainers reach a Nash equilibrium. Without
specifying some rule of this sort, we cannot derive any predictions about
how the wage level responds to shifts in the position of the labour demand
function.
Employment effects can, however, be derived without any such
additional specification. The equation of the contract curve is had by
equating the slope expressions for the iso-profit and the indifference
curves. With the shift variable for labour's marginal product inserted, this
equal slopes condition is (AF' — w) / N = —(w —171)1 N, or AF' = IT. The
contract curve is vertical since w does not enter this equation. From this
equation of the contract curve, we see that we can determine the effects
on employment of changes in A and IT without having to specify the
bargaining model that is required for the model to yield any real wage
predictions. It appears that this co-operative union model does not support
the hypothesis of real wage rigidity, but we can derive the employment
effects that follow from this theory if we impose real wage rigidity in a
macroeconomic context. For macroeconomic employment effects, it is as
if real wages were fixed.
We complete our analysis of the co-operative model by using the
standard Nash product to derive the condition which determines the
division of the rents between the union and the firm. The function that is
delegated to the arbitrator to maximize involves the product of two items:
first, what the firm can earn in profits if co-operation is achieved (minus
what it gets with no co-operation — zero), and second, the similar
differential for workers. This product can be written as [(w —
where V is profits: V = AF(N) — wN, 0 is the bargaining power parameter

177
 (0 = 1: unions have all the power; 0 = 0: firms have all the power), and NJ
is a union preference parameter (w = 1: the union is utilitarian in that it
values all its members (not just the currently employed); w = 0: the union
is seniority oriented (only the wages of those currently employed are
valued)).
After differentiating the arbitrator's objective function with
respect to w and N and simplifying, we have two labour market equations
(which are the imperfect competition analogues for supply and demand
curves) to determine wages and employment. McDonald and Solow call
these two relationships the efficiency locus and the equity locus. The
equation that defines the contract curve is the efficiency locus. It is
AF' = Tv" with a utilitarian union, while it is AF' = w with a seniority-
based union. With the seniority-based union, the negotiations pay no
attention to employment and the union indifference curves are horizontal
lines. Thus, in this case, the labour demand curve and the contract curve
coincide, and this version of the system essentially replicates the non-co-
operative model. Finally, the equation which defines the division of the
rents (the equity locus) is CV = (1— 6)(w — TON, whether unions are
utilitarian or not.
Pissarides (1998) uses a model of union-firm interaction that
combines features of the two different approaches that have just been
summarized. His model follows the "right to manage" aspect of the non-
co-operative approach, in that it is the firm that chooses the employment
level after the wage , has been determined. But the model involves a key
feature of the co-operative approach as well, since the wage is not set
unilaterally by the union. Instead, it is the result of a bargaining process
involving both the union and the employer. A simplified version of
Pissarides' model (involving risk neutrality on the part of the union and a
Cobb-Douglas production function) is explained here. In the first stage,
an arbitrator is appointed to choose the wage which maximizes the
following Nash product function: (I —1) 9 (V — 0)" . I is the index of the
workers' net benefit from the contract. Pissarides' definition of this net
benefit is I = wN + (L — N)[(1—u)w* +ulT]. As above, N is employment
in this firm, and L is union membership. It is assumed that those who do
not find employment in this firm seek employment elsewhere. These
individuals face probabilities equal to the employment rate, and the
unemployment rate, concerning whether they secure another job (and are
paid w*) or whether they are without work (and receive employment
insurance equal to IT) . T is what individuals receive if employment at
this firm, N, is zero. Thus, (I = N(w — (1— u)w* ulT). As far as the
—

178
firm's profit is concerned, we have V = Y — wN and the production
function is Y
Differentiating the arbitrator's objective function with respect to
w, and then substituting in the equations that define full equilibrium
(w = w*) and the unemployment insurance system ( 4. 3= fw), we have:
u = a 1(1— f) , where a =[0(1— y) I y(1— 0)1 We see that this version of
imperfect competition in the labour market yields the same equation for
the natural unemployment rate as did our efficiency-wage model. Despite
this similarity, there are some new results embedded within this
alternative derivation of this reduced form. For example, in this
union/firm interaction interpretation, we see that the natural
unemployment rate is predicted to rise, the higher is the degree of union
power. Thus, if lower structural unemployment is the goal, we might view
the model as providing some support for legislation that is designed to
limit workers' rights.
The more general point is that many policies that are designed to
lower the natural unemployment rate (some of which we stress in the next
chapter) receive equivalent analytical support — whether one appeals to
efficiency-wage theory or union/firm interaction theory. We can have
more confidence about making applied policy advice when the underlying
rationale for that policy proposal is not dependent on just one
interpretation of the labour market or the other.

8.4 Transaction Frictions in the Labour Market: Search Theory

We continue our examination of "natural" unemployment rate models by

considering one more source of market failure. In this case, instead of
incomplete information (and efficiency wages) or market power (and
unions), we focus on frictions in the labour market. This analytical
framework highlights the fact that job seekers and employers meet in a
series of decentralized one-on-one settings. Since matches do not occur
instantaneously, this setting generates an equilibrium amount of frictional
unemployment. This section follows Romer (2001, pages 444-453) and
Hall (2003) very closely in outlining as simple a summary of search
theory as is possible.
First, we define some notation; E, U and L stand for the number
of individuals who are employed, unemployed, and in the labour force.
The labour force is taken as an exogenous constant. In this section of the
chapter, we use f and s to denote the job finding rate and the job
separation rate. This notation can be used to define equilibrium. The
179
labour market is in equilibrium when the unemployment rate is constant,
and this, in turn, requires that the number of individuals leaving the
unemployment pool each period be just balanced by the number who are
entering that pool. In symbols: fU = sE. Since E = L — U, and since the
unemployment rate, u, is UIL, this equilibrium condition can be solved for
the unemployment rate:

sl(s + f).

What search theorists have done is to build a model that makes the job
finding rate, f, endogenous (and based on optimizing behaviour). The
resulting solution for f is then substituted into the unemployment-rate
equation just presented, so that the determinants of the natural
unemployment rate are then exposed.
We need additional notation to define this theory (and we use
Romer's): A — the marginal product of each employed worker, C — the
fixed cost of maintaining a job, and w — the wage rate. It is assumed that
there is no cost of posting a vacancy. The profit associated with each
filled job is given by (A — C — w), and the profit associated with each job
vacancy is (— C). We assume static expectations, so that we can represent
the present value of receiving these flows indefinitely, by simply
multiplying them by (11r).
The technology of the matching process is specified by:

M = at Ifir ,

where M is the number of matches. If + y <1 , it is said that there is

congestion or crowding in the labour market (the more individuals and
firms there are in the market, the less chance there is that any one match
can be made). If f3+y>1, it is said that there is a "thick market"
externality involved (the more individuals and firms there are in the
market, the more likely it is that a good match will be found). This is the
assumption made by Howitt (1985) and Diamond (1984) in the multiple-
equilibria models that are discussed in the next chapter. Finally, if
+ y =1, we rule out both negative and positive externality effects. Since
the empirical evidence weakly favours this case, we proceed with this
assumption below. Specifically, defining the vacancy rate, the
unemployment rate, and the ratio of these two measures as v = VIL, u =
UIL, and x = u/v, we can define the job finding rate as

180
 f = MIU = aU " V" = a(u 1 v)" = 0(x).

Similarly, the job filling rate can be represented as

= M I V = aU fl T7-fi = a (u I v)fl = a(u 1 v)(u 1 v)" = xq5(x).

We complete the model by defining the utility function of individuals. For

simplicity, it is specified that individuals are risk neutral and that they do
not save, so utility equals current income. Thus, the level of utility is w if
an individual is employed, and utility is zero if she is unemployed.
Dynamic programming is used to define optimal behaviour. In
this approach, we consider (for example) the "annual return" that an
individual receives from being employed. This value is denoted by rVE ,
and it is equal to the sum of a "dividend" and a "capital loss":

rVE = w—s(V E —Vu ). (8.1)

This equation defines the annual dividend of having a job as the wage
received, and the capital loss as the difference between the value of a job
and the value of not having one (being unemployed). The probability of
sustaining this capital loss is the job separation rate. The annual return
that is associated with other states are defined in a similar manner.
The value to a firm of maintaining a filled job is

rVF= (A — C — w)— s(V F —VV ) (8.2)

The firm's annual profit is the "dividend" and the difference between the
value of a filled job and the value of an unfilled vacancy is the "capital
loss". Again, the separation rate is the probability of sustaining this
capital loss.
The annual return of being unemployed is given by

rVu =0 + f (V, —Vu ) (8.3)

since, without unemployment insurance, the dividend is zero, the capital

gain is the increase in utility that comes with the possibility that the
individual becomes employed, and the probability of receiving this gain is
the job finding rate.

Finally, the annual return to the firm of maintaining an unfilled vacancy is

181
 rVv = —C +0(VF — Vv ) (8.4)

The dividend is negative (the cost of maintaining any position), the

capital gain is the extra profits that are received if the position is filled,
and the probability of receiving this gain is the job filling rate.
The remaining equations define different aspects of full
equilibrium. As noted above, a constant unemployment rate requires that
the flows out of, and in to, the unemployed pool must be equal:

M= sE, or JU = sE (8.5)

Given that the cost of posting a vacancy is zero, firms must have posted a
sufficient number to ensure that the marginal benefit is zero:

rVv = 0 (8.6)

Finally, the wage must be set so that the gains of the match are distributed
between individuals and firms in a way that is consistent with the "market
power" of these groups. As in our models of union/firm interaction, we
assume a Nash equilibrium, and for simplicity here, we assume equal
levels of bargaining power. This implies that the wage is set so that an
individual's gain from a match is equal to the firm's gain from the
relationship:

( VE — yU)= (VF — Vv)• (8.7)

The solution proceeds as follows. Subtracting (8.3) from (8.1), and (8.4)
from (8.2), we have

VE - Vu = 147 1(f + s + r) (8.8)

VF - Vv (A— w)1(19 + s + r) (8.9)

Substituting (8.8) and (8.9) into (8.7), and solving for w, we have

w= A(f +s+r)1(f +0+2(s+r)) (8.10)

and substituting (8.9) into (8.4):

rVy = —C +[B(A— w)]1(6) + s + r) (8.11)

182
Finally, substituting (8.10) into (8.11) to eliminate w, imposing Vv = 0,
and substituting in f = 0(x), B = x0(x), and z = AIC, we end with

((z —1)x —1)0(x) = 2(s + r) (8.12)

which is a nonlinear equation in x, the ratio of the unemployment rate to

the vacancy rate. Hall (2003) picks values for s, r, z, a and 13, and solves
for x. Once x is known, Hall solves for the unemployment-rate:-

u = s i(s + 0( x )).

Part of Hall's calibration is that 13 = 0.5, so (8.12) becomes a quadratic

equation. Hall chooses representative values for the other parameters as
well, and as a result (8.12) becomes:

(0.62)y 2 — (0.0913)y —1= 0

where y = . Of the two solutions, only one is positive (that is,

economically admissible). This one solution is used — in a version of the
model that is extended to allow for several taxes — in the next chapter. As
with the other models of market failure in the labour market, we identify
fiscal policies that can be expected to have favourable effects on the
natural unemployment rate.

8.5 Related Issues in New Keynesian Economics: Real vs. Nominal

Rigidities

Some of the labour market models that were surveyed in the three
previous sections of this chapter provide support for the hypothesis of
real wage rigidity, but a fundamental problem for business-cycle theory is
to explain why purely nominal shocks have real effects. The Keynesian
and new neoclassical synthesis approaches to this question rely on
nominal wage and/or price rigidity. Can the models of this chapter apply
in any way to this question?
Some analysts have argued that the theories of real wage rigidity
can apply to nominal wages in an indirect way. When wages are set,
bargaining is about the real wage. But the item that is actually set as a
result of these decisions is the money wage. It is set at the level intended
to deliver the desired real wage, given inflationary expectations. With this
183
interpretation, we can argue that the theories apply to money wage-
setting, although an additional assumption regarding indexation is
required. We would expect agents to set the money wage with a full
indexing clause so there would be no need to incur errors in inflationary
expectations. But given that catch-up provisions can roughly replace ex
ante indexing formulae, and given that households and firms want to tie
wages to different price indexes, the costs of full indexation are probably
not worth the benefit (as McCallum (1986) has stressed).
Quite apart from_the preceding argument, some analysts are
uneasy about applying an adjustment-cost model (such as the one we
explored in Chapter 4) to explain sticky goods prices. Negotiation costs
between buyers and firms are not a feature of reality for many
commodities. Of course the sale of many commodities involves posting
prices, but it does not seem compelling to rest all of sticky-price
macroeconomics on an item that seems rather trivial (such as the cost of
printing new prices in catalogues — the so-called "menu" cost — and the
cost of informing sales staff about price changes). The response that one
can make to the charge that adjustment costs for many nominal prices
cannot be "that important" is simply to demonstrate that even explicitly
small price-change costs can lead to large welfare losses. Akerlof and
Yellen (1985) and Mankiw (1985) provide analysis that is intended to
support this view.
Let us examine a brief summary of the argument (along the lines
suggested by Romer (1993)). Consider a monopolist who must set her
nominal price before the relevant period but who can change that price
later (during the period) at a "small" cost. A situation in which the price
has been set at too high a value is illustrated in Figure 8.2. When the firm
set its price, it did not guess the then-future position of its demand curve
perfectly. As it enters the period analyzed in Figure 8.2, it has already
posted a price equal to OA, but the appropriate price is OB. The firm must
now decide whether making the change is worthwhile. As far as private
profits are concerned, the firm loses an amount equal to area FGH by not
lowering its price to OB. The cost to society of not adjusting the price is
area DGHE — potentially a much bigger amount. It is quite possible for
even quite small adjustment costs to be larger than area FGH but much
smaller than area DGHE. Thus, the social gains from price adjustment
may far exceed the private gains.

184
 Figure 8.2 Menu Costs
Price

Posted price

Marginal cost
G
Marginal Demand
revenue
Quantity/time

This analysis suggests that Keynesians may not have to assume

"large" nominal price-adjustment costs to sustain their simultaneous
claims that:

1. Prices may be sticky.

2. The welfare losses that follow from sticky prices may be
large (at least if general equilibrium expectational effects — such
as those stressed in Chapter 2 — are ignored).
3. Government policy may be justified since individual firms
consider only the private benefits of price flexibility.

There are a couple of reasons why this analysis may not support such
sweeping conclusions. For one thing, Figure 8.2 shows just the case
involving prices remaining too high. It is left for the reader to draw a
diagram in which the existing price is set at too low a value. It is still true
that firms incur a small private cost (in terms of foregone profits) that may
easily be dominated by the menu cost, if they do not raise prices. And it is
still true that the implications of not adjusting price are much larger for
society. But this time, that large area is a gain in welfare. Since prices
may be too low just about as often as they are too high, it may be roughly
the case that menu costs lead to no significant net effect on society
welfare. While this nonnative issue has been overstated by some

185
Keynesians, the positive point remains — even seemingly trivial menu
costs may dominate the private benefits of incurring them.
A second issue that has been glossed over in our discussion of
Figure 8.2 is a more detailed focus on how the demand and cost curves
may have shifted to create an initial situation such as that illustrated. One
possibility is that the vertical intercept of the demand curve and the height
of the marginal cost curve shifted down by the same amount. But if real
wages are rigid, and the nominal price does not fall (given menu costs),
the position of the marginal cost curve should not be shifted down at all.
Such a redrawing of Figure 8.2 shrinks the size of area FGH, and so
makes it all the more likely that even small menu costs can be the
dominant consideration. In short, real wage rigidity may increase the
applicability of the menu-cost model to such an extent that the hypothesis
can be said to play a central (if indirect) role in providing the micro
foundations for nominal rigidities.
The intuition behind this result is perhaps best appreciated by
considering an oligopoly. Each firm finds it costly to change its relative
price, since a higher relative price is immediately noticed by its current
customers, while a lower relative price is not widely noticed by the other
firms' customers. Thus, there is a real rigidity — in relative prices. Even if
the nominal rigidity — the actual cost of changing its own nominal price —
is very small, all firms will behave as if this is not the case — because of
the real rigidity. Thus, real rigidities magnify the importance of a little
nominal rigidity.
Alvi (1993) has presented a simple proof that this proposition is
quite general, which we now summarize. Alvi assumes that each firm's
profit function can be written as V(P/i5,M /3 ), where P, P and M
denote the firm's own price, the economy-wide average price, and the
nominal money supply. The first argument in this function captures
imperfect competition, while the second captures aggregate effects. The
fact that firms care only about relative prices and the real value of money
means that no money illusion is involved. Each firm's optimum outcome
can be written as

P/i5 =--H(M//5), (8.13)

which indicates that the best value for the firm's price is simply a function
of the two items which the firm takes as parameters: P and M. Note that
real rigidity (that is, relative price rigidity) is prevalent if H' is small. We

186
assume P = P = M =1 initially, and that H' = h . Then, the total
differential of (8.13) implies:

dP I dM h + (1— h)(di" I dM). (8.14)

Let z = dP I dM be an index of nominal flexibility (so z = 0 implies

complete nominal rigidity and z = 1 implies completely flexible nominal
variables). Let proportion q of the firms be subject to nominal menu costs
to the extent that their prices are not adjusted in response to a change in
the money supply. Given this notation, we know that

dP / dM = (q)(0) + (1— q)(dP / dM) = (1— q)z.

Substituting this relationship into (8.14), we obtain

z = h 1(1— (1— q)(1— h)).

This final equation implies that z = 1 if there are no menu costs (q = 0).
With menu costs, however, it implies az/oh > 0, so a given amount of
menu cost results in a high degree of overall nominal rigidity if real
rigidities are prevalent. We conclude that it is not necessarily
unreasonable to base a theory of business cycles on small nominal
"menu" costs.
Research continues on the microeconomics of menu costs. To
some, the most appealing model of price changes at the individual level is
known as the two-sided (S,$) adjustment rule. It involves the firm only
incurring the fixed cost of adjustment when the gap between the desired
price and the existing one exceeds a critical value (S on the high side, s on
the low side). Heterogeneity among firms can take various forms, such as
differing initial positions within common (S,$) bands, or firm-specific
shocks. As is usual in the aggregation literature, not all specifications lead
to well-defined macro implications.
Ball and Mankiw (1994) draw the distinction between time-
contingent adjustment models and state-contingent adjustment models.
The theory we covered in Chapter 4 is an example of the former, while
the (S,$) models are examples of the latter. Ball and Mankiw note that no
robust conclusions have emerged from the literature on state-contingent
adjustment, but that this state of affairs is not necessarily upsetting. This
is because time-contingent adjustment is optimal if the main cost is
gathering information about the state rather than making the actual price

187
adjustment. Also, in economies with two groups of firms — one making
each kind of adjustment — it turns out that the sluggish adjustment on the
part of the time-contingent firms makes it rational for those monitoring
developments continuously according to the state-contingent model to
behave much like the other group. Thus, it may well be that the quadratic
adjustment cost model of Chapter 4 is not such a bad approximation of a
theory with much more detailed structure.
Another issue that is being researched is whether it matters to
specify explicitly that it is the gathering of information, not the re-setting
of prices, that is costly. Mankiw and Ries (2002) have shown that a
"sticky information" version of an expectations-augmented Phillips curve
may fit the facts better than a "sticky" price" version does. Further, they
demonstrate that this version of a new synthesis model can lead to
different conclusions regarding the relative appeal of alternative monetary
policies.

8.6 Conclusions

In Chapters 1 through 4, we focused on the deviations of real output from

its natural rate. We assumed that the natural output rate was unique, and
that we could think of it being determined by the Classical model.
Further, we assumed — but had not formally shown — that if we added
some friction to the specification of the labour market, we could generate
a unique, non-zero value for the associated natural unemployment rate.
The primary task of this chapter has been to provide that more detailed set
of analyses for the labour market. We have found that market failure can
occur in the labour market for several different reasons — incomplete
information, imperfect competition, and transactions costs. In some of
these cases, these features allow us to interpret the resulting
unemployment as involuntary, and so it is reasonable to investigate
whether policy can improve the outcome. We proceed to address this very
question in the next chapter.
There was one other purpose in surveying the several leading
models that generate real rigidities in labour markets. We have seen that
real rigidities — such as those highlighted in this chapter — magnify the
importance of nominal rigidities. This is important since both New
Keynesians, and now all those who have adopted the new neoclassical
synthesis, have relied on this proposition to "justify" their having their
approach to business cycles depend on seemingly small "menu" costs in
an important way. Even New Classicals — such as Alexopoulos (2004),

188
who has added a version of efficiency-wage theory to the real business
cycle framework — are relying have relied on this proposition to generate
more persistence in real variables within their models.
There are three tasks that we address in the remaining chapters.
First, as just noted, we use the models developed here to analyze a series
of policy proposals designed to lower structural unemployment and to
raise the economic position of those on low income. Second, we use some
of these models to investigate the possibility of multiple equilibria. If
theory leads to the possibility that there are two "natural" unemployment
rates — both a high-employment equilibrium and a low-employment
equilibrium — there is an "announcement effect" role for policy. It may be
possible for the government to induce agents to focus on the high-activity
outcome if agents know that the policy maker stands ready to push the
system to that outcome if necessary. It is possible that no action — just the
commitment to act — is all that may be necessary. Third, we would like to
see whether policies that are geared to reducing structural unemployment
have an undesirable long-run implication. Might these initiatives retard
the productivity growth rate? We examine the first two issues in the next
chapter, and then devote the final three chapters to an analysis of long-
term growth.

189
Chapter 9

Unemployment and Low Incomes:

Applying the Theory

9.1 Introduction

In the last chapter, we summarized three approaches to modelling the

labour market — standard analyses of what determines the level of
structural unemployment. While this is often called the "natural"
unemployment rate, this term is unfortunate, since all these theories
suggest that the full-equilibrium unemployment rate can be affected by
fiscal policy. The primary task for this chapter is to investigate precisely
how. We consider several fiscal policies:

• replacing the income tax with an expenditure tax in section 9.2,

• taxing physical capital owners to finance a tax break for wage earners
in section 9.3,
• introducing low-income support policies, 'such as employment
subsidies and a guaranteed annual income, in models of both
developed and developing economies in section 9.4, and
• investigating how policy can both create and react to multiple
equilibria in section 9.5.

9.2 Tax Reform: Direct vs. Indirect Taxation

We begin by illustrating the possibilities for fiscal policy within the

Summers (1988) version of efficiency-wage theory. First, we add an
income tax to the model, which can be interpreted in two ways. First, it
could be an employee payroll tax (that is levied on wage income but not
on unemployment-insurance benefits). Second, it could be part of a
progressive personal income tax, that involves no tax on low levels of
income (such as what individuals receive if all they have access to is the
unemployment-insurance benefit). With this tax, the specifications of the
efficiency index and the outside option change compared to how they
were specified in Chapter 8. These relationships are now given as
q = ((w(1— t) — b) / b)" and b = (1— u)w(1— t) + ufw, where t is the wage-

190
income tax rate. It is left for the reader to verify that these modifications
change the unemployment rate solution to

u = a /[1— (f /(1

This equation implies that an increase in the tax rate raises the natural
unemployment rate. This occurs because higher taxes reduce the relative
pay-off individuals receive from work. To lessen the resulting increase in
worker shirking, firms offer a higher wage, and they hire fewer workers at
this higher price.
The importance of taxes can be illustrated by considering some
illustrative parameter values. Realistic assumptions are: u = .05,f = .50, t
= .15. These representative values are consistent with this model only if a
= .02, which we therefore assume. Now consider fixing a and f at 0.02
and 0.5 respectively, while higher tax rates are considered. The reader can
verify that the unemployment rate rises by one percentage point (to u =
.06) when the tax rate rises by 10 percentage points to 0.25, and the
unemployment rate rises by much more (2 and 2/3 percentage points, to
0.0867) when the tax rate rises by an additional 10 percentage points to
0.35. This thought experiment indicates that one does not need to have
ultra right-wing views to be concerned about efficiency in government.
Only with such efficiency can we have the many valuable services of
government with the lowest possible taxes, and (as this numerical
example suggests) high taxes can very much raise unemployment.
It is instructive to examine the effects that several other taxes
have (or more precisely, do not have) within this basic version of the
efficiency-wage model. With an employer payroll tax, T, the firms' wage
bill becomes wN(1 + -r), and with a sales tax, X, the wage that concerns
households is w* = w/(1 + X). It is left for the reader to verify that, when
these changes are made in the specification of the efficiency-wage model,
there is no change in the solution equation for the unemployment rate. It
is useful to review the intuition behind why employee payroll taxes do,
but these other taxes do not, affect unemployment. As already noted, both
a more generous unemployment insurance system and a higher employee
payroll tax increase unemployment. Both these measures lower the
relative return from working. To compensate for the deterioration in work
effort that results, firms must raise wages, and this makes a lower level of
employment optimal.
The other taxes do not change the relative return of work
compared to being unemployed. For example, sales taxes must be paid
simply because goods are purchased; it makes no difference how the
191
purchaser obtained her funds. This is why the natural unemployment rate
is unaffected by the sales tax. Similar reasoning applies to the employer
payroll tax. A cut in this levy increases both the ability of the worker's
employer to pay higher wages and the ability of all other firms to pay that
individual higher wages. Competition among firms for workers forces this
entire increase in ability to pay to be transferred to those already working
(in the form of higher wages). As a result there is no reduction in
unemployment. The same outcome follows for anything that shifts the
labour demand curve without having any direct effect within the workers'
effort function. This is why we stressed in the previous chapter that
increases in general productivity raise wages — and do not lower
unemployment — in this efficiency-wage setting.
These results imply that we can have a lower natural
unemployment rate if we rely more heavily on a sales tax, instead of an
income tax. They also imply that investments in training and education
lead to higher wages, but not to lower unemployment. But before we can
have confidence in such strong predictions, and exhort real-world
authorities to act on this advice, we need to know whether they are
supported by the other theories of the natural unemployment rate.
To check the effects of various fiscal policies in our models of
union-firm interaction, we add a wage-income tax (which, as above, can
also be interpreted as the employee payroll tax), an employer payroll tax,
and a sales tax. As in Chapter 8, the function that is delegated to the
arbitrator to maximize involves the product of two items: first, what the
firm can earn in profits if co-operation is achieved (minus what it gets
with no co-operation — zero), and second, the similar differential in
returns for workers. This product is [(((1— t)w— 17)41 + 2))N" J8 V" .
V = AF(N) — wN(l+z) is profits. 0 is the union bargaining power
parameter, and w is the union seniority parameter.
After differentiating the arbitrator's objective function with
respect to w and N and simplifying, we have the tax-included versions of
the two labour market equations that determine wages and employment.
The equation that defines the contract curve is (1 —OAF' = W(1+ r) with a
utilitarian union, while it is AP =14(1+z) with a seniority-based union.
The equity relationship is OAF(N) = w(l + z)N —(1— 0)(1+ z)vTAT 1(1— t)
whether unions are utilitarian or not. In both cases, the level of
employment is unaffected by sales taxes, but it is affected by both the
employer and the employee payroll tax (an increase in either tax raises
unemployment).

192
 We add the same set of taxes to the Pissarides (1998) model of
union-firm interaction that combines features of the co-operative and non-
co-operative approaches. The arbitrator's objective function is still
(/ — T)' (V —O)' -B , and the production function is still Y = ANr. There are
several changes: (I — I ) = ((w —(1— u)w*)(1— t) — uW)N 1(1+ 2),
V = Y — wN (1 + r), and = w(1 + r). Proceeding with the same
steps as we followed in Chapter 8, we arrive at the revised solution
equation for the unemployment rate:

u =a/[1— fl(1—t)]

where, as before, a =[0(1— y)]/[y(1— 0)]. The policy implications are a

little different from those that followed from the other models of union-
firm interaction, but they are the same as those that followed from the
efficiency-wage model. For all the models, we have found that the natural
unemployment is increased by higher employee payroll taxes, but it is not
increased by a higher sales tax. It appears that there is one general
conclusion that has emerged from both efficiency-wage theory and
union/firm interaction theory: if a lower unemployment rate is desired, we
should reduce employee-payroll and wage-income taxes, and finance
these tax cuts by imposing a higher sales tax. It is reassuring that the
support for this move toward a heavier reliance on indirect taxes receives
the same analytical support from both theories about labour markets. We
can have more confidence about making applied policy advice when the
underlying rationale for that policy proposal is not dependent on just one
interpretation of the labour market or the other.
But our examination of this policy proposal is not complete; we
must derive its implications in the search model as well. To pursue this
sensitivity test, the same set of taxes is added to that model. It is left for
the reader to verify that equations (8.10) and (8.12) are altered:

w .[A(f + s + r)]I[(f + s + r)(1+ r) + ((0 + s + 0(1— t) 1(1+ 2))]

= —C +[0(A — w(1+ r))] l(0 + s + r).

Proceeding with the solution, and Hall's (2003) calibration, we reach the
several policy conclusions. Some are similar to the outcomes that we
discovered in our analysis of efficiency wages and unions. For example,
an increase in the employee payroll tax increases unemployment. Even
the magnitude of this response is comparable to our earlier findings. (If t
is raised from zero to 0.1, the unemployment rate rises by about one half
193
 of one percentage point.) This finding means that our earlier conclusion is
robust across alternative specifications of the labour market. For this
policy, at least, it appears not to matter that there is controversy
concerning how best to model structural unemployment. Policy makers
can proceed without needing to wait for this controversy to be resolved.
But this assurance does not apply to all policy initiatives, since
some of the implications of search theory are different from the polic y
theorems that followed from the other models. For example, in this
specification, both the employer payroll tax and the interest rate affect the
natural unemployment rate — predictions that are at odds with both the
efficiency-wage model and Pissarides' model of union/firm interaction.
But not all of these differences are important. For example, while the
interest rate matters in the present specification (since a higher interest
rate lowers the benefit of having a job and so raises equilibrium
unemployment), the practical significance of this effect is non existent.
The reader can verify that, when Hall's calibration is used, and when the
annual interest rate is raised by even two or three percentage points, the
effect on the unemployment rate is truly trivial. Hence, some of the
differences across models of the labour market are irrelevant for policy
purposes, and we can proceed with the policy prescriptions that
accompanied the earlier specifications.
However, not all the differences across natural unemployment
rate models can be dispensed with in this way. For example, in this search
model, the unemployment rate is increased by the existence of a sales tax.
Again, for Hall's calibration, we find that increasing the variable from
zero to 0.1, makes the unemployment rate rise by about one half of one
percentage point. This is a non-trivial effect, and it differs markedly from
the zero response we discovered with efficiency wages and unions.
This different outcome is important for the general debate on
whether we should follow the advice of many public-finance practitioners
— that we should replace our progressive personal income tax with a
progressive expenditure tax. According to growth theory (models which
usually involve no unemployment, which we examine in chapters 10-12),
this tax substitution should increase long-run living standards. According
to efficiency-wage and union theory, this tax substitution should bring the
additional benefit of lowering the natural unemployment rate. But as just
noted, this fortuitous outcome is not supported by search theory.
However, this search model indicates that the cut in the wage income tax
can be expected to lower unemployment by about the same amount as the
increase in the expenditure tax can be expected to raise unemployment.
Thus, even this model does not argue for rejecting the move to an

194
expenditure tax. In this limited sense, then, the labour market models give
a single message: with respect to lowering the natural unemployment rate,
we either gain, or at least do not lose, by embracing a shift to expenditure-
based taxation.

9.3 The Globalization Challenge: Is Mobile Capital a Bad Thing to

Tax?

One of the primary concerns about the new global economy is income
inequality. Compared with many low-wage countries, the developed
economies (often referred to as the North) have an abundance of skilled
workers and a small proportion of unskilled workers. The opposite is the
case in the developing countries (the South). With increased integration
of the world economies, the North specializes in the production of goods
that emphasize their relatively abundant factor, skilled labour, so it is the
wages of skilled workers that are bid up by increased foreign trade. The
other side of this development is that Northern countries rely more on
imports to supply goods that require only unskilled labour, so the demand
for unskilled labour falls in the North. The result is either lower wages for
the unskilled in the North (if there is no legislation that puts a floor on
wages there) or rising unemployment among the unskilled in the North (if
there is a floor on wage rates, such as that imposed by minimum wage
laws and welfare). In either case, unskilled Northerners can lose income
in the new global economy.
There is a second hypothesis concerning rising income inequality.
It is that, during the final quarter of the twentieth century, skills-biased
technical change has meant that the demand for skilled workers has risen
while that for unskilled workers has fallen. Technical change has
increased the demand for skilled individuals to design and program in
such fields as robotics, while it has decreased the demand for unskilled
workers since the robots replace these individuals. Just as with the free-
trade hypothesis, the effects of these shifts in demand depend on whether
it is possible for wages in the unskilled sector to fall. The United States
and Europe are often cited as illustrations of the different possible
outcomes. The United States has only a limited welfare state, so there is
little to stop increased wage inequality from emerging, as indeed it has in
recent decades. Europe has much more developed welfare states that
maintain floors below which the wages of unskilled workers cannot fall.
When technological change decreases the demand for unskilled labour,
firms have to reduce their employment of these individuals. Thus, Europe

195
has avoided large increases in wage inequality, but the unemployment
rate has been high there for many years.
Most economists favour the skill-biased technical change ex-
planation for rising income inequality. This is because inequality has
increased so much within each industry and occupation, in ways that are
unrelated to imports. The consensus has been that only 11% of the rising
inequality in America can be attributed to the expansion of international
trade. But whatever the causes, the plight of the less skilled is dire.
Even if globalization is not the cause of the low income problem
for unskilled individuals in the North, it may be an important constraint
on whether their governments can do anything to help them. This is the
fundamental challenge posed by globalization. Citizens expect their
governments to provide support for low-income individuals so that
everyone shares the benefits of rising average living standards. The anti-
globalization protesters fear that governments can no longer do this. The
analysis in this section — which draws heavily on Moutos and Scarth
(2004) — suggests that such pessimism is not warranted. To address this
question specifically, let us assume that capitalists (the owners of capital)
are "rich" and that they have the ability to re-locate their capital costlessly
to lower-tax jurisdictions. Also, we assume that labour is "poor" and that
these individuals cannot migrate to other countries. Can the government
help the "poor" by raising the tax it imposes on the capitalists and using
the revenue to provide a tax cut for the workers? Anti-globalization
protesters argue that the answer to this question is "obviously no." They

Figure 9.1 A Tax on Capital

Quantity of
0 tput

Quantity of Capital

expect capital to relocate to escape the higher tax, and the result will be
less capital for the captive domestic labour force to work with. Labour's
196
 living standards could well go down — even with the cut in the wage-
income tax rate. It is worthwhile reviewing the standard analysis, since it
is the basis for recommending that we not tax a factor that is supplied
perfectly elasticly (such as capital is for a small open economy). Figure
9.1 facilitates this review. The solid lines represent the initial demand and
supply curves for capital. The demand curve is the diminishing marginal
productivity relationship that is drawn for an assumed constant level of
labour employed. The supply curve is perfectly elastic at the yield that
owners of capital can receive on an after-tax basis in the rest of the world.
Before the tax on capital is levied to finance a tax cut for labour, the
economy is observed at the intersection of these solid-line demand and
supply curves, and GDP is represented by the sum of the five regions
numbered 1 to 5.
When the government raises the tax on capital, capitalists demand
a higher pre-tax return — an amount that is just enough to keep the after-
tax yield equal to what is available elsewhere. Thus, the higher (dashed)
supply curve in Figure 9.1 becomes relevant. Domestically produced
output falls by regions 1 and 3. Capital owners do not lose region 1, since
they now earn this income in the rest of the world. Labour loses regions 3
and 4, but since the tax revenue is used to make an unconditional transfer
to labour, their net loss is just region 3. But this is a loss, so the analysis
supports the propositions that capital is a bad thing to tax, and that it is
impossible to raise labour's income.

Figure 9.2 A Tax on Capital to Finance a Wage-Income Tax Cut

Quantity of
Output

Quantity of Capital

But this standard analysis involves the assumption that the policy has no
effect on the number of men and women employed. If the level of

197
 employment rises, capital can be a good thing to tax after all. If there is
unemployment in the labour market, and no similar excess supply in the
capital market, the economy involves a distortion before this policy is
initiated. The existence of involuntary unemployment means that, before
the policy, society's use of labour is "too small," and that (from society's
point of view) profit maximization has led firms to use "too much" capital
compared to labour. A tax on capital induces firms to shift more toward
employing labour and this helps lessen the initial distortion. But can this
desirable effect of the policy package outweigh the traditional cost (the
loss of income represented by region 3 in Figure 9.1)? Figure 9.2 suggests
that this possible. As long as the wage-income tax cut results in lower
unemployment, each unit of capital has more labour to work with, and so
it is more productive. This is shown in Figure 9.2 as a shift up in the
position of the marginal product of capital curve (shown by the higher
dashed demand curve). In this case, the total income available to labour is
affected in two ways. It is reduced by the shaded triangle, and it is
increased by the shaded parallelogram.
If the gain exceeds the loss, the low-income support policy is
effective after all. It lowers unemployment, it raises the total income of
the "poor" (labour) and it does not reduce the income of the "rich" (the
owners of capital). This approach to low-income support is not a zero-
sum game, in the sense that labour is not helped at the expense of
capitalists. This is because the size of the overall economic "pie" has been
increased by policy. Labour receives a bigger slice, and capitalist get the
same slice as before. And all of this appears possible — despite the fact
that the government faces the constraints that are stressed by the anti-
globalization protesters. The same result is stressed in Koskala and Schob
(2002). In their model, the unemployment results from unions, not
asymmetric information, as is the case in our specification below. Related
work, involving search theory instead of either efficiency wages or
unions, is available in Domeij (2005).
There are two crucial questions: First, is it reasonable to expect
that a cut in the wage-income tax rate will lower the long-run average
unemployment rate? We addressed that question in the previous section
of this chapter, and we discovered that the answer is "yes." The second
question concerns whether it is reasonable to argue that the gain can be
bigger than the loss. It is straightforward to answer this question by
combining: one of our models of unemployment (we choose the
efficiency-wage model), a production function that involves both capital
and labour as inputs, a government budget identity, and the hypothesis of
perfect capital mobility. We now define just such a model, and derive the

198
condition that must be satisfied for this revenue-neutral tax substitution to
provide the Pareto improvement that we have just discussed.

-r K r
Y = (gN)'
q =[(w(1— t) — b) I br
h = (1— u)w(1— t)+ ufiv
(1 — y)Y I N = w
yY I K r
u = a(1— 1) 1(1— t — f)
N =1— u
r(1— r) = r*
G + .fwu = rrK + twN

The equations are explained briefly as follows. The first is a Cobb-

Douglas production function, which indicates that output is determined by
the quantity of inputs — capital and effective labour. The labour effective-
ness index is defined in the second equation — as in our efficiency-wage
model in Chapter 8. By combining the second, third and sixth equations,
readers can verify that worker productivity turns out to be an exogenous
constant — independent of tax and unemployment-insurance generosity
policy: (q=(a1(1—a))a). This fact simplifies the derivations that are
referred to below. The third equation defines the average income of a
labourer (which is equivalent to the outside option in the efficiency-wage
model). The next three equations are the firms' first-order conditions for
profit maximization. Firms hire each factor up to the point that the
marginal product equals the rental cost. Also, when the firms' optimal
wage setting relationship is combined with the optimal hiring rule for
labour, the solution for the unemployment rate (the sixth equation)
emerges.
The next two equations define factor supplies. Labour supply is
inelastic (at unity) so employment is one minus the unemployment rate.
Capital is supplied completely elastically at the rate of return that this
factor can earn in the rest of the world. This perfect capital mobility
assumption is what imposes the globalization constraint — that capital can
avoid paying any tax in this small open economy. Finally, the last
equation defines a balanced government budget. The uses of funds (on
program spending and unemployment insurance) must equal the sources
of funds (the taxes on capital and wage incomes). In this section of the

199
chapter, T refers to the tax on the earnings of domestically employed
capital, not an employer payroll tax.
The equations determine Y, N, u, w, b, K r, q and T for given
values of the other variables and g= GIY. We use this system to derive
the effects on the unemployment rate, u, and the average income of a
labourer, b, of a cut in the wage tax rate, t, that is financed by a change in
(presumed to be an increase in) the tax on capital, T. To accomplish this,
we take the total differential of the system, and eliminate the other
endogenous variable changes by substitution. The goal is to sign duldt
and dbldt.
It turns out that the second of these policy multipliers has an
ambiguous sign. Nevertheless, we can show that the average income of a
labourer must rise, as long as the government does not encounter a
"Laffer curve" phenomenon. What this means is that the government
must raise one tax rate when the other is cut. Laffer believed that the
opposite might be true — that a cut in one tax rate might so increase the
level of economic activity (the overall tax base) that overall revenue
collected would increase — despite the fact that the tax rate was reduced.
Most economists read the evidence as being against this proposition, and
so have concluded that tax cuts do not more than finance themselves. That
is, most analysts are comfortable assuming that the other tax rate would
have to be raised. We assume that here. To make use of this non-
controversial assumption, we need to work out dr/dt and to assume that
this expression is negative. It is left for the reader to verify that this
assumption is necessary and sufficient to sign the average-income
response (to ensure that dbldt is negative). The unemployment-rate
response is unambiguous in any event.
We conclude that low-income support policy by governments in
small open economies is quite feasible — despite the constraint imposed
by globalization — as long as the revenue that is raised from taxing capital
is used to lessen the pre-existing distortion in the labour market. Since a
transfer to labour that is not conditional on employment status does not
meet this requirement, using that instrument (in an attempt to provide
low-income support) fails. Nevertheless, the fact that a Pareto
improvement is found with the wage-income tax cut policy, means that
the anti-globalization protesters have been premature in their verdict
concerning the inability of governments in small open economies to raise
the economic position of the low-income individuals within their
countries.
Before closing this section, it is worth reviewing why a Pareto
improvement is possible. For an initiative to be both efficiency-enhancing
200
and equity-enhancing, the economy must be starting from a "second best"
situation. Involuntary unemployment involves just this kind of situation.
We can clarify by recalling an example introduced in the original paper
on this topic (Lipsey and Lancaster (1956)). In a two-good economy,
standard analysis leads to the proposition that a selective sales tax is
"bad". With a tax on the purchase of just one good, the ratio of market
prices does not reflect the ratio of marginal costs, so decentralized
markets cannot replicate what a perfect planner could accomplish —
achieve the most efficient use of society's scarce resources. Society is
producing and consuming "too little" of the taxed good, and "too much"
of the untaxed good. But this conclusion assumes that there is no pre-
existing market distortion — before the tax is levied. A different verdict
emerges if it is assumed that there is an initial market failure. For
example, if one good is produced by a monopolist who restricts output
and raises price above marginal cost, a similar inefficiency is created
(with society consuming "too little" of this good and "too much" of the
competitively supplied good). There are two policies that can "fix" this
problem. One is to try to use the Competition Act to eliminate the
monopoly; the other is to levy a selective excise tax on the sale of the
other product. With this tax, both prices can be above their respective
marginal costs by the same proportion, and society gets the efficient
allocation of resources — even with the monopoly.
So the verdict concerning the desirability of a selective sales tax
is completely reversed, when we switch from a no-other-distortions
situation to a with-other-distortions setting. The analysis in this section
shows that this same logic applies in macroeconomics to factor markets.
With incomplete information in the labour market, labour's price is "too
high" and firms employ "too little" labour. By stimulating employment,
we can increase overall efficiency — have higher GDP — as we improve
equity (by lowering unemployment). This sort of outcome is what led to
the Bhagwati/Ramaswami (1963) theorem. This proposition concerns a
second-best setting, and it states that we have the best chance of
improving economic welfare if the attempt to alleviate the distortion is
introduced at the very source of that distortion. Since the distortion in this
case is that wages are "too high" to employ everyone, one would expect
that the government can improve things by pulling the wage that firms
have to pay to hire labour back down. This takes place in our analysis
since the employee payroll tax cut lessens workers' wage claims. Another
way of saying essentially the same thing is to note that the second-best
problem is the existence of asymmetric information in the labour market,
which leads to a level of employment that is "too low." By directly

201
stimulating employment, the employee payroll tax cut partially removes
the original distortion at source, and this is why the analysis supports this
initiative. ,

9.4 Low-Income Support Policies in Developed and Developing

Economies

In the previous section, we examined one way of providing support to

those on low incomes — taxing the "rich" to finance a tax cut for the
"poor". Other measures are either being used, or are being actively
debated, in many countries — measures such as employment subsidies, and
a guaranteed annual income. We consider these initiatives in this section.
Also, since the problem of inadequate incomes is most acute in the
developing countries, we indicate some ways in which our analysis of the
labour market can be altered to increase its applicability in that setting.
There are three broad ways that governments can offer support to
those on low incomes. One method is to provide unemployment
insurance. This method makes the government support conditional on the
individual being unemployed, so it leads to higher unemployment. The
second method is to provide basic income (see Van Parijs (2000)) — a
guaranteed income for everyone that is not conditional on one's
employment status. Proponents argue that the basic income policy is
better than unemployment insurance, since it does not involve the
incentive that leads to higher unemployment. Other macroeconomists,
such as Phelps (1997), Solow (1998) and Freeman (1999) have taken this
line of reasoning further and advocated that it is best if the government
uses the funds it could have used for these other support programs to
provide subsidies to firms to employ low-skill individuals. By making the
support policy conditional on employment, this policy is intended to be
the one that provides the lowest unemployment rate.
We analyze the employment-subsidy and the basic-income
proposals by modifying the efficiency-wage model of the previous section
(and this provides a simpler but extended version of Moutos and Scarth's
(2004) study). The model is defined by the following equations.

Y = (qN) 1-7 Kr
q =[(w(1—t + p)—b)1 br
b = (1— u)w(1— t)+ pw
(1— y)Y I N = w(1— s)

202
 yY/K=r
u = a(1— t — s)I(1— t)
N =1—u
r(1—r) .= r*
G + pw+ swN =rrK + twN

For simplicity, we have removed the unemployment insurance program.

The new policy parameters are s and p. Firms receive a subsidy of S per
employee, and the government makes this subsidy proportional (at rate s)
to the average wage in the economy. Individuals receive a payment of P
as a guaranteed annual income, and the government makes this payment
fraction p of the average wage. Everyone receives this basic income,
whether she is working or not. Because this receipt is independent of
employment status, it does not affect the unemployment rate. But because
firms get a bigger employment subsidy, the larger is their work force, the
employment subsidy does affect the unemployment rate. The reader can
verify the revision in the unemployment-rate equation by re-deriving the
firm's profit maximization. In this case, profits equal Y — wN — rK + SN.
After optimization, we simplify by substituting in S = sw. As in the
previous section, it is assumed that both policy initiatives are financed by
an increase in the tax on capital.
The results are rather messy, but it is left for the reader to take the
total differential of the model to evaluate how each individual's average
labour income, variable b, is affected by these policies. The government
introduces either the employment subsidy or basic income (raising either s
or p above an initial value of zero), and we assume that both tax rates are
equal initially. Only one quantitative assumption needs be made to sign
one of the outcomes. The signs of all other responses can be determined a
priori. The employment subsidy definitely lowers the unemployment rate,
and it raises average labour income as long as y > t /(1 — t). This condition
is satisfied for plausible parameter values (for example, capital's share of
income equal to one-third and any tax rate up to 25%) so the model
supports the introduction of employment subsidies. A rather different set
of results emerges with the guaranteed annual income policy. It has no
effect on the unemployment rate, and it must reduce average labour
income. It is true that each individual is better off because she receives
basic income, but she is worse off since her market wage falls. The
increase in the tax on capital that is necessary to finance the basic income
policy drives enough capital out of the country to make labour noticeably

203
less productive. This negative effect must dominate, so the analysis does
not support the introduction of a guaranteed annual income.
The overall conclusion is that the employment subsidy can be
defended — even when the model highlights the "globalization constrain(
(the fact that the financing of the initiative requires a higher tax rare
which scares away capital). However, the basic income proposal canner
be defended. The intuition behind this difference in outcomes is the same
as that which applied in the previous section. The employment subsid:.
addresses a distortion (asymmetric information in the labour market) a:
source, while the guaranteed annual income does not.
We pursue the analysis of employment subsidies in one additional
way. To motivate this further investigation, we note that our model has
not involved any specialized features that would make it particularly
applicable to developing economies. Development economists have
stressed two things about production possibilities in the lesser developed
countries that we now insert into our analysis. First, they have stressed
that developing countries often have a limited supply of some crucial
input — a problem that cannot be highlighted if we restrict our attention to
the Cobb-Douglas production function (that allows firms to produce each
level of output with any ratio of factor inputs). Second, they have stressed
that workers can be so under-nourished that their effectiveness on the job
can be compromised. The following adaptation of the earlier model
allows for these considerations.

Y = min(V, L 1 0)
V = (qN) I-7 Kr
q =[(w — 6)1 b"
b = (1— u)w
(1— v0)(1— y)Y / N = w(1— s)
(1— vO)yY I K = r
u = a(1— s)
N =1— u
v = v*
r(1— r) = r*
G + swN = rrK

The first two equations define the production process, and this two-part
specification follows suggestions made by Moutos. The first equation is
the overall production function. It is a Leontief fixed-coefficient rela-
tionship which states that output is equal to the minimum of two inputs —
204
 skilled labour, L, and remaining value-added, V. The latter is a standard
Cobb-Douglas function of unskilled labour, N, and capital, K. Skilled
labour is the "crucial" input; each unit of output requires 0 units of this
input. The remaining value added can be produced with an infinite variety
of unskilled-labour-to-capital ratios. Development economists refer to this
type of specification as an "0-ring" theory of production. This label is
based on the NASA disaster in which the travelers in the Columbia
spacecraft perished all because of one tiny flaw — a damaged 0-ring
sealer. The basic` ideais that — no matter how many and how good all
other inputs are — if one is missing, the entire enterprise amounts to
nothing. Skilled labour is the analogue to the 0-ring in our case, and this
is a concise way of imposing the notion that the modern world involves
knowledge-based economies. With profit maximization, firms do not hire
unused factors, so we proceed on the assumption that Y = V = (1 / OW
The third equation is the unskilled worker effort index. It is
different from what was specified above in two ways. The non-essential
way is that — for simplicity — we have removed the basic-income and the
unemployment-insurance policies in this specification, and we have also
set taxes on both forms of labour to zero. The novel feature in the effort
relationship is the second argument on the right-hand side. We can think
of this as a "nourishment effect"; with parameter n> 0, it is the case that —
other things equal — the higher is the unskilled labour wage, the more
healthy, and therefore, the more productive are these individuals. There is
no variable-worker-effort function for skilled labour. For one thing, it is
assumed that their wage is high enough for there to be no concern about
their basic health and nourishment. Further, since these individuals have
"good" jobs, there is no reason for them to consider shirking; they enjoy
their work too much. Thus, only the unskilled become unemployed.
The fifth, sixth and seventh equations are the first-order
conditions that follow from profit maximization. Profits are defined as
Y — wN — vL — rK + SN. The next three equations define factor supplies;
unskilled labour is stuck within the country (inelastic supply), and the
other two factors are perfectly mobile internationally. Skilled labour can
earn wage v*, and capital can earn rent r*, in the rest of the world. The
last equation is the government budget constraint. Program spending and
the employment-subsidy expenses (paid to firms for hiring unskilled
labour) are financed by a tax on capital.
We do not expect readers to work out the formal results of this
model. It's structure is spelled out just so that readers are aware of how to
adapt the analysis to a developing economy setting. We simply assert the
result that emerges: the subsidy to firms for hiring unskilled labour brings
205
both good and bad news. The good news is that the unemployment rate is
reduced. The bad news is that sufficient capital is pushed out of the
country (due to the higher tax levied on capital to finance the employment
initiative) for the average income of an unskilled individual, b, to fall.
Thus, it is harder for governments to provide low-income support in the
very set of countries where pursuing this objective is most compelling.
To keep exposition straightforward, we followed Moutos'
specification of the 0-ring feature in the production function (which is
much simpler than the standard specification, as in Kremer (1993a)).
Given this departure from the literature, it is useful to provide some
sensitivity test. To this end, we report a different, less thorough-going,
method of decreasing substitution possibilities within the production
process. We revert to just the one (unskilled) labour and capital
specification, but we switch from Cobb-Douglas to a CES production
function with an elasticity of factor substitution equal to one half (not
unity as with the Cobb Douglas). The production and factor-demand
functions become

= &N) - ' + (1— oci

ay I a(qN)= 0(Y I qN) 2
ay l arc = ( 1- o)(y. 1102
Again, we simply report the results, without expecting readers to verify
them, since the derivations are quite messy. (Easier tests of the reader's
ability to perform derivations are available in the practice questions.) The
results for this specification of limited factor substitution are very similar
to what has been reported for the 0-ring model. This fact increases our
confidence that the disappointing conclusion reached in that case is likely
to be relevant for actual developing economies.

9.5 Multiple Equilibria

The model of efficiency wages presented in Section 8.2 is a convenient

vehicle for illustrating the possibility of multiple equilibria. Sometimes,
as formerly in Canada, the generosity of the unemployment insurance
system is increased (up to a maximum) for regions of the country that
have had high unemployment in previous periods. We can model this
policy by specifying that the unemployment-insurance generosity
parameter, f, be higher if the previous period's unemployment rate is

206
higher: f = au,_, if the previous period's unemployment rate is below
some upper limit (u,_, < /7), while f =7 once that maximum upper limit
is reached /7) . Since the solution equation for the unemployment
rate is it, = a 1(1 — f,), in the simpler version of the model with no taxes,
the unemployment rate follows a first-order nonlinear difference equation,
as long as it is below the upper bound.

Figure 9.3 Two Values for the Natural Unemployment Rate

Ut-i
1 2 4

This relationship is shown as the heavy line in Figure 9.3. Since

full equilibrium involves u, =u,_, there are three values for the natural
unemployment rate — given by the points A, B and C. But only points A
and C represent stable outcomes. To see this, suppose the economy starts
at point A. Consider a "small" shock that makes the unemployment rate
rise from 1 to 2. The time path beyond that first period is shown in the
diagram by the steps back to point 1. Now consider a "large" shock that
makes the unemployment rate rise from 1 to 3. The time path in this case
is the set of steps from 3 to 4. Thus, when near A or C, the economy
converges to these points; convergence never occurs to point B. As
stressed by Milbourne, Purvis and Scoones (1991), models of this sort
have awkward implications for disinflation policy. The "temporary"
recession involved in disinflation may be permanent if the recession is a
"large" shock.
207
 Through almost all of the first seven chapters of this book, we
have assumed that the economy has just one long-run equilibrium (the
natural rate of output). Often we have restricted our investigation of the
role for government to questions of how alternative policy rules affect the
speed with which the economy approaches that full equilibrium or how
these rules affect the asymptotic variance of variables about their full-
equilibrium values. Thus, our analysis has followed Tobin's (1975)
suggestion that Classical models be accepted as descriptions of the long-
run outcome and that Keynesian models be recognized as very helpful
descriptions of the adjustment paths toward that full equilibrium. Now,
however, we can see why Keynesians find multiple-equilibria models so
exciting. They suggest that Keynesians should no longer concede the long
run to the Classicals. With more than one natural rate, there is an
additional role for policy — to try to steer the economy to the "preferred"
full equilibrium.
One advantage of the multiple-equilibria model that we have just
discussed is its simplicity. This simplicity permits explicit derivations.
But one disadvantage is that the existence of multiple equilibria depends
on the presence of a particular government policy. The response of the
New Classicals is simply to suggest that the policy maker avoid such
policies. So this dependence of the multiple equilibria on policy itself has
to temper the enthusiasm about reclaiming some relevance concerning the
economy's full equilibrium on the part of Keynesians. But as we shall see
in the following brief (non-technical) review of other multiple-equilibria
models, not all owe their existence to particular government policies.
Diamond (1984) and Howitt (1985) examine search theories,
which analyze how households and firms interact to evaluate whether the
individual is hired (or wants to be hired). A key feature of their versions
of these models is the presence of a trading externality. The probability of
obtaining a positive match between workers and jobs depends on the
amount of resources firms devote to recruitment. But much of the benefit
of increasing the information flow emerges as a general social benefit — a
lower equilibrium level of frictional unemployment. There is no direct
link between this general benefit and the individual decision process of
any particular firm. It is rational for the individual firm to think that if it
increases expenditures on the hiring and searching process, the overall
level of frictional unemployment will be unaffected. In any event, the
firm is unable to appropriate any general benefits that occur. Since the
private return to recruitment from the firm's point of view is less than the
social return, the economy reaches an equilibrium that is inefficient.

208
 Figure 9.4 Multiple Equilibria with Increasing Returns to Scale

One way to understand this class of search models is by focusing

on the very simplified narrative suggested by Diamond. Consider a group
of individuals who live on an isolated island by eating coconuts.
Religious custom precludes eating the coconuts that each individual has
collected herself. Before deciding how much to produce (that is, how
many trees to climb and coconuts to collect), each individual must form
an expectation about the probability that she will find another individual
with whom to trade. To have some other traders is very valuable to the
worker/trader, but once,there are a reasonable number, more traders bring
only a very small additional benefit (since by then she is already almost
certain to find a partner). Thus, with y denoting her own output, and x her
expectation of each other trader's output, she will set her own activity
level according to the y = f (x) relationship like that shown in Figure 9.4
— which captures the increasing returns to the trading process in the AB
range of the graph. Let expectations be adaptive: x = A(a — x) where a is
the actual behavior of others. We can evaluate di I dx in the region of full
equilibrium (when everyone is at the same activity level: a = y .
Evaluating at a = y = f (x), we have di / dx = —2(1— f'), so equilibrium
is stable only if f' <1. Equilibrium involves a = y = x and these points
occur along the 45-degree line in Figure 9.4. Points A and C are the two
stable equilibria. Given that individual traders do not receive the full
social benefit that follows from their entering the otherwise "thin" market,
there is an externality problem. This market failure allows us to rank the
two stable equilibria; C is preferred. In principle, government
involvement could switch the outcome to the Pareto-superior equilibrium.
209
 Howitt (1986, page 636) summarizes the outcome as follows —
stressing that both the more-preferred and the less-preferred equilibria
involve rational expectations: "if everyone believes that markets will be
inactive they will anticipate a high cost of transacting; this will
discourage them from undertaking transactions, and the initial beliefs will
be self-fulfilling." What Diamond and Howitt have done is to provide
modern standards of analytical rigour to defend the very early summary
of Keynesian economics — that, without policy intervention, the economy
could remain stuck indefinitely with a sub-optimal amount of
unemployment.
Woglom (1982), Blanchard and Kiyotaki (1987), and Rowe
(1987) have also constructed models involving multiple equilibria. One
feature in some of these models is that firms face kinked demand curves.
The reason for the kink is not the traditional oligopoly consideration
based on rivals' reactions. Instead, the kink is based on Stiglitz's (1979)
assumption of asymmetries in the dissemination of information to
customers. Customers learn immediately about any price change at the
firm with which they have been trading, but they learn only slowly of
other price changes. Price increases are noticed by a firm's own
customers (so an elastic response can occur), but price decreases are not
noticed by customers who have been buying elsewhere (so an inelastic
response occurs). The resulting kink in the demand curve causes a
discontinuity in firms' marginal revenue curves. If marginal cost cuts
through this discontinuity, almost any such point can be an equilibrium.
Individual firms face a free-rider problem when adjusting prices. They
can understand that when there is a general reduction in all firms'
demand, it would be desirable to have all firms lower price. This would
stimulate aggregate demand and avoid a recession. But if only one firm
lowers price, this general benefit is not forthcoming since one firm is too
small to matter. Every firm wants to keep its own price high while having
all others pass on to it the "macro" benefit of lowering their prices. This is
the standard "unstable cartel" or "prisoners' dilemma" problem.
Expansionary policy can internalize this externality problem.
It is not just that imperfect competition can lead to multiple
equilibria; it is that the equilibria can be formally ranked. The high-
activity equilibrium is preferred since it lessens the standard efficiency
cost associated with monopoly. Another implication of imperfect compe-
tition is noteworthy. Manning (1990) has focused on the increasing
returns that characterize a natural monopolist. Increasing returns can
make the labour demand curve positively sloped. As a result, it can
intersect a positively sloped wage-setting locus more than once, and so

210
there are multiple equilibria. Farmer (1993) has also considered
increasing returns to scale — examining how the New Classical model is
affected by this extension. Multiple stable equilibria emerge.
"Strategic complementarity" is a game-theoretic term which has
been used to interpret many of the multple-equilibria models. As Cooper
and John (1988) have noted, there is a general reason that coordination
fails in many of these New Keynesian models. The general feature is that
the larger is aggregate production, the larger is the incentive for each
individual to produce. They show that this feature provides a general
underpinning for Keynesian multiplier effects, Oh and Waldman (1994)
explain that it is a basis for slow adjustment, and Alvi (1993) proves that
strategic complementarity can (along with real rigidities) accentuate the
importance of any nominal rigidities that are present in the system.
The most general notion of multiple equilibria is found in models
that involve hysteresis. The simplest model of this sort — provided by
Blanchard and Summers (1986) — is based on the idea that the more
senior members of a union (the "insiders") are the ones who make the
decisions on wages. These workers are assumed to give no weight to the
preferences of members who are no longer seen — having become
unemployed. The insiders' power stems from median-voter
considerations. The wage is set equal to the value that makes the firm
want to hire just the number of workers who were employed in the
previous period. Thus, the expected employment in time t, denoted as
E(N,) , equals the last period's employment, N1_, . An expression for
expected employment can be had by specifying a labour demand function.
Blanchard and Summers assume a simple aggregate demand function for
goods, Y, = c(M, — Ps ) , and constant returns to scale in production; thus, if
units are chosen so that labour's marginal product is unity, Y, = N, and
P, = W, (where Y stands for output, M for money supply, P for price, and
W for the wage rate). The implied labour demand function is
N, = c(M, — W,) . If the expectations operator is taken through this
relationship and the resulting equation is subtracted from the original, we
have E(N,) = N,—c(M,— E(M,)) , since wages are set so that
W, = E(W,)). Replacing expected employment by N,_ 1 , the time path for
employment becomes

1V, = E(M,)).

211
This model is consistent with both the random-walk observation
concerning output and employment rates (Campbell/Mankiw 1987) and
the "money surprise" literature (Barro 1977). Unexpected changes in
aggregate demand affect employment, and there is nothing to pull the
level of employment back to any particular equilibrium (because the
preferences of laid-off workers no longer matter for wage-setting).
Blanchard and Summers consider several variations of this and other
models to test the robustness of the hysteresis prediction. Some of these
extensions allow the "outsiders" to exert some pressure on wage-setting,
with the effect that the prediction of pure hysteresis is replaced by one of
extreme persistence.
Another source of multiple equilibria is "the average opinion
problem" in rational expectations. The economy has many equilibria —
each fully consistent with rational expectations — and each one
corresponding to a possible view of what all agents expect all the others
to take as the going market price (see Frydman and Phelps (1983)).
Ultimately, models such as these lead us to the proposition that
the belief structure of private agents is part of the "fundamentals" — much
like tastes and technology — so that economists should study the several
equilibria rather than search for some rationale to treat all but one as
inadmissable. (Readers saw how common this practice is when learning
phase-diagram methods in Chapter 6.) This plea for further study
inevitably forces analysts to explore how agents gradually achieve
rational expectations. For example, consider even a very limited aspect of
learning — can agents grope their way to knowing the actual values of a
model's structural coefficients if all they start with is knowledge of the
form of the model? Pesaran (1982) surveys some of the studies which
pose this class of questions. Some plausible adaptive learning schemes
converge to unique rational expectations equilibria in some contexts, but
not always. Despite the assumption that agents incur no decision-making
costs, these plausible learning models sometimes lead to cycles and/or
divergence from rational expectations equilibria. With decision making
costs, Pesaran (1987) has stressed that agents can become trapped in a
kind of vicious circle of ignorance. If agents expect further learning is not
economically worthwhile, insufficient information will be accumulated to
properly test that initial belief and therefore to realize that the original
decision may have been mistaken. This implies that systematic forecast
errors may not be eliminated with economically rational expectations.
Most studies of multiple equilibria do not question the entire
concept of rational expectations; instead, they stress how the economy
might shift between them — resulting in fluctuations in aggregate demand

212
that are ongoing due to the self-fulfilling cycle of revised expectations (as
in Woodfood (1991). This class of models is quite different from both
traditional macroeconomics and New Classical work, where cycles are
caused by exogenous shocks to fundamentals (such as autonomous
spending in Keynesian models or technology in the real business cycle
framework). In the standard approach, it is almost always the case that it
is optimal for agents to absorb these shocks (at least partly) by permitting
a business cycle to exist. After all, stochastic shocks are a fact of life. As
we have seen in earlier chapters, attempts by the government to lessen
these cycles can reduce welfare. But if cycles result solely from self-
fulfilling expectations, then it is much easier to defend the proposition
that the elimination of cycles is welfare improving. Indeed, government
may not need to actually do anything to eliminate the cycles other than
make a commitment to intervene to stabilize if that were ever necessary.
Knowledge of that commitment may be sufficient to cause agents to
expect (and therefore achieve) a non-cyclical equilibrium.
Howitt and McAfee (1992) build a similar model of endogenous
self-fulfilling cycles. It is based on the theory of search behaviour in the
labour market covered in section 8.4 — one that involves a supposedly
"non-fundamental" random variable called (in deference to Keynes)
"animal spirits." A particularly interesting feature of the analysis is that
the equilibrium involving ongoing cycles between the optimistic and
pessimistic phases is stable in a learning sense. Baysian updating induces
convergence to this equilibrium with positive probability, even though
agents start with no belief that animal spirits affect the probability of
successful matches in the labour market search activity. Models such as
this one provide a solid modern pedigree for even the most (apparently
non-scientific) of Keynesian ideas — animal spirits.
There are many more models of multiple equilibria in the
literature that focus on other topics. But enough has been covered for
readers to appreciate how many public-economics terms — externality,
incomplete information, missing markets, non-convexity, moral hazard,
market power — appear in New Keynesian analyses. The intention is to
meet the challenge posed by the New Classicals — have firmer micro
foundations for macro policies — that can then be motivated on the basis
of some well-identified market failure (second-best initial condition). This
means that the principles that underlie normative analysis in
macroeconomics are becoming consistent with the principles that underlie
microeconomic policy analysis — an outcome much applauded by New
Classicals.

213
9.6 Conclusions

The purpose of this chapter has been to use some of the micro-based
macro models of the natural unemployment rate that were developed in
the previous chapter to assess several policies that have been used or
advocated for reducing structural unemployment and/or raising the
incomes of unskilled individuals. Here, we summarize a few of the key
findings.
First, there is considerable analytical support for a policy of
decreasing our reliance on income taxation and increasing that on
expenditure taxes. This tax substitution can be expected to lower the
natural unemployment rate. Second, involuntary unemployment creates a
second-best environment in the labour market. In such a setting, it can be
welfare-improving to impose a distorting tax — even one levied on capital
that is supplied perfectly elastically — if the revenue can be used to reduce
the pre-existing distortion in the other factor market (the labour market).
This environment makes low-income support possible — even for the
government of a small open economy that faces the "globalization
constraint." This second-best analysis was extended so that the appeal of
competing anti-poverty policies — providing employment subsidies to
firms or providing basic income to individuals — could be compared.
Finally, we explored how natural-unemployment-rate analyses
could be modified to consider some of the additional constraints that
confront policy makers in developing economies, and to consider the
possibility of multiple equilibria. The possibility of multiple equilibria
suggests an "announcement effect" rationale for policy. With both a high-
employment equilibrium and a low-employment equilibrium possible —
and with both involving self-fulfilling rational expectations — policy can
induce agents to focus on the high-activity outcome if agents know that
the policy maker stands ready to push the system to that outcome if
necessary. It is quite possible that no action — just the commitment to act
— is all that may be necessary.
The natural unemployment rate is a long-run concept. There is
another long-run aspect of real economies that we have ignored thus far in
the book. This feature is the fact that there is ongoing growth — a long-run
trend in the natural rate of output. We focus on this issue — productivity
growth — in the remaining chapters of the book.

214
Chapter 10

Traditional Growth Theory

10.1 Introduction

In our discussion of stabilization policy, we focused on short-run

deviations of real GDP from its long-run sustainable value. We now shift
our focus to the determinants of the trend in GDP, and away from a focus
on the deviations from trend. We care about the trend, since we wish to
explore what policy can do to foster rising long-run average living
standards. To highlight this issue, we now abstract from short-run
deviations altogether. In this chapter, therefore, we consider a longer term
analysis in which it is reasonable to assume completely flexible wages and
prices. In such a world, there is no difference between the actual and the
natural rates. We focus on the long-run determinants of per capita
consumption.
All western governments try to stimulate saving. Some of the
initiatives are: taxing consumption instead of income (partially replacing
the income tax revenue with the expenditure taxes), allowing lower taxes
on capital-gain and dividend income, allowing contributions to registered
retirement saving plans to be made with before-tax dollars, keeping
inflation low, and deficit reduction. One of the purposes of this chapter is
to review the traditional economic analysis that is viewed as supporting
initiatives such as these.
Growth theory is often described as "old" or "new." The "old"
analysis refers to work that involves two features: (i) descriptive
behavioural functions for agents (that are not explicitly based on inter-
temporal optimization), and (ii) productivity growth specified as an
exogenous process. As a result, there are two sets of literature that qualify
as "new" growth theory. The first continues to specify productivity
growth as exogenous, but since micro-based decision rules are involved,
the analysis is immune to the Lucas critique. The second branch of new
growth theory — which is new in both senses — involves micro-based
behavioural functions and endogenous productivity growth. This "new-
new" analysis is also called endogenous growth theory. We examine old
growth theory and the first class of new models in this chapter. We move
on to endogenous productivity growth analysis in Chapter 11.

215
10.2 The Solow Model

The standard model of exogenous growth is due to Solow (1956) and

Swan (1956). It is defined by the following relationships:

Y = F(N,K)
S
S =sY
I = k +8K

The first equation is the production function: output is produced by

combining labour and capital. We assume that the production function is
constant returns to scale. One implication of this assumption is that a
doubling of both inputs results in an exact doubling of output. This
assumption is necessary if we want the full equilibrium of the system to
be what is referred to as a balanced growth path. This refers to a situation
in which all aggregates (the effective labour supply, capital, output,
consumption, and investment) grow at the same rate — the sum of the
population and productivity growth rates. This outcome implies that
wages grow at the productivity growth rate and that the interest rate is
constant. The constant-returns-to-scale assumption is an appealing one if
the economy is already big enough for all the gains of specialization to be
exhausted, and if factors that are fixed in supply (such as land and non-
renewable raw materials) are of limited importance. Standard growth
theory abstracts from these issues; we consider them briefly in section
10.5 below.
The notation is standard; the second equation stipulates goods
market clearing (saving equals investment), and the next two relationships
define saving and investment. There are no adjustment costs for capital;
firms invest whatever output is not consumed. N denotes labour measured
in efficiency units, which is always fully employed (flexible wages are
assumed implicitly). Labour grows at an exogenous rate, NI N=n=y+z,
where y is the productivity growth rate and z is the rate of population
growth. As noted above, since y is taken as exogenous, this is called an
"old" growth model involving exogenous technological change. The
fourth equation states that the capital stock grows whenever gross
investment, I, exceeds the depreciation of pre-existing capital, 6K.
Since the production process involves constant returns to scale,
we re-express it in what is called the "intensive" format. We use lower-
case letters to denote each variable on a per-effective-worker basis. For

216
example, assuming a Cobb-Douglas function: Y = K" N j- " , we have
y= Y / N = (K / N)" = f (k). Using the time derivative of the k = KIN
definition, the entire model can be summarized in a single differential
equation:

= sf (k) — (n + g)k

Figure 10.1 The Solow Growth Model and the Golden Rule

(5 +n)k

k"

This dynamic process is stable if ak ak < 0; in other words convergence

to equilibrium requires sf (k)— (n + 8) < 0. We evaluate this requirement
using full equilibrium values. In full equilibrium, ylk=(n+8)1s, so
convergence requires that the average product of capital exceed the
marginal product of capital. This condition is satisfied for any well-
behaved production function such as that shown in Figure 10.1; since the
tangent at any point such as C is flatter than the ray joining point C to the
origin, the marginal product is less than the average product.
This definite stability property is easily seen in Figure 10.1. The
production function is graphed (in intensive form) as the f(k) curve. The
lower curve, labelled sj(k), is the nation's savings function — defined on a
per-effective-worker basis. The final line in the figure is the ray coming

217
out from the origin with a slope equal to (n + 0. This line can be
interpreted as the "required" investment line, if required refers to what is
necessary to keep the capital stock growing at the same rate as the
effective labour supply. With no investment, the capital stock is shrinking
through depreciation at rate 8. Thus, to make up for this, and to have
capital grow at effective labour's growth rate, n, capital must grow at a
rate equal to the sum of these two factors for the system to achieve
balanced growth. Capital-labour ratio k, is the equilibrium, since it marks
the intersection of the actual per-effective-worker saving/investment
schedule with the required saving/investment schedule.
Suppose the economy starts with a capital-labour ratio that is
smaller than value k, (that is, we start to the left of the equilibrium). In
this region of the figure, the height of the actual saving/investment curve
is greater than the height of the required saving/investment line. Thus, the
economy is accumulating more capital than is necessary to keep the
capital-labour ratio constant, and that ratio must, therefore, rise. The
economy moves inexorably toward the k, level of capital intensity.

Figure 10.2 Per-Capita Consumption and Lower Interest Taxation

In C

r „ - -

Time
1 2

Since convergence to k = 0 is assured, we know that, in full equilibrium,

output and capital must grow at the same percentage rate as does labour
(measured in efficiency units). But since that growth rate, n, is an
exogenous variable, it cannot be affected by policy (which in this compact
structure must be interpreted as variations in the savings rate, s). A tax
218
policy which permanently raises the propensity to save pivots the sf(k)
curve up in Figure 10.1, making the equilibrium point move to the right.
The economy settles on a higher capital-labour ratio. Assuming that this
higher level of capital intensity raises per capita consumption, the
response is as shown in Figure 10.2. Consumption falls initially, since
people save a bigger share of a given level of income. But through time,
the higher saving means that workers have more capital to work with, and
there is higher output, and this is what permits both higher saving and
higher consumption. Figure 10.2 shows both the short-term pain and the
long-term gain. It also shows that the growth rate rises — but only in a
transitional way. The lasting effect of the higher savings policy is an
increase in the level of the capital/labour ratio (and therefore in the level
of per capita consumption), not in the growth rate. Nevertheless, this can
still represent a very significant increase in material welfare.
But can we be sure that there is long-term gain? By considering
Figure 10.1, this appears to be possible when the initial equilibrium
(before the pro-savings initiative is instituted) is on the left side of the
diagram (as we have assumed). But it does not appear to be possible if the
initial equilibrium is already well over to the right. This ambiguity raises
the question: how we can determine the optimal value for the savings rate
— at least as long as we restrict our attention to full-equilibrium
considerations? At low capital/labour ratios, labour does not have enough
capital to work with to achieve maximum efficiency. At high
capital/labour ratios, diminishing returns with respect to capital sets in to
such an extent that most of the extra output is not available for
consumption — it is needed just to maintain the capital stock, and to keep it
growing at the same rate as labour. (With a fixed depreciation rate, a large
capital stock requires a lot of replacement investment each period.)
The key to determining the "best" savings rate is to realize that
per capita consumption is pictured in Figure 10.1 as the vertical distance
between the output per capita curve,f(k), and the required (for a constant k)
investment per capita line, (n + 8)k. That gap is maximized at the
capital/labour ratio identified as k* where the tangent at point C is parallel
to the (n + 8)k line. Thus, the rule which must be obeyed to maximize
steady-state consumption per head — the so-called "golden rule" — is that
the marginal product of capital equal the sum of the depreciation rate and
the rates at which the population and the state of technical knowledge are
growing.
Is this condition likely to be met in modern developed economies?
It is clear that all policy makers assume that it is not, and that we are at a
point to the left of the golden rule (as in Figure 10.1). If this were not the
case, the analysis cannot support the universal drive among policy makers
219
to stimulate saving. Thus, the presumption appears to be that — in actual
economies — the marginal product of capital exceeds (n + g) . As long as
we believe that firms maximize profits, we have to believe that — in long-
run equilibrium — they must be equating the marginal product of capital to
its rental cost, (r + g) . Thus, the presumption must be that (r + 8) must
exceed (n+8), or that r must exceed n=kIK. As noted in Chapter 10,
Abel et al (1989) have tested this presumption by comparing net profits,
rK, and net investment, k, by consulting the national accounts for many
countries. For every country and every year, they found that profits
exceeded investment. This finding has been taken as strong support for the
proposition that all these economies are under-capitalized. Thus, Figure
10.2 correctly shows the time path for per-capita consumption following a
pro-savings initiative.
So higher saving involves short-term pain during the time interval
between points 1 and 2 in Figure 10.2, since — without the policy — per
capita consumption would have been higher (following along the lower
dashed line). But after point 2 in time, there is long-term gain. Old people
are hurt by the pro-savings policy, since they could die within the 1-to-2
time period. But the young, especially those who are not born until after
point 2 in time, are made better off. From the point of view of the elderly,
higher saving is a policy of following the golden rule — doing unto others
(the young) what elderly would like others to have done for them.
While the Solow model is the analytical base for pro-savings
policy initiatives, it is not without its critics. For one thing, it seems
incapable of explaining the vast differences in living standards that we
have observed, both across time and across countries at a point in time.
Roughly speaking, the challenge is to explain the fact that citizens in the
developed economies have a living standard that is 10 times the level that
was observed 100 years ago in these same countries. Similarly, the
developed countries enjoy living standards that are roughly 10 times what
the poorer countries are making do with today. Differences of these
magnitudes seem beyond the Solow model. To appreciate this fact, take
the total differential of the steady-state version of the model's basic
equation, replace k with y (by using the Cobb-Douglas production
function given above), and evaluate the result at steady-state values. That
result is:

(dy / y) = (a 1(1— a))(ds / s).

If we consider a plausible value for capital's share of output, a = 0.33, this

result implies that per-capita output is increased by a mere 5% when a
220
 substantial increase in the savings ratio (10%) is undertaken and sustained
indefinitely. This quantitative outcome is not in the league of what we
need to explain.
A second empirical issue concerns the speed of convergence to
full equilibrium. The model's speed of adjustment is the absolute value of
the coefficient in the stability condition (the absolute value of the aid a
expression). When this is evaluated at full-equilibrium, we see that the
speed of adjustment measure is (n + 8)(1— a). Taking one year as the
period of analysis, plausible values are: n = .02, 8= .04 and a = 0.33, so
the speed coefficient is 0.04. From the "rule of 72", it takes (72/4) = 18
years for the model to get half way from an initial steady state to a new
steady state. Most empirical workers argue that adjustment speed in the
real world is much faster than this, so there is concern about the
applicability of the Solow model on this score as well.
Related to this, a great deal of empirical work has been done to
test the "convergence hypothesis" — an implication of the Solow growth
model when it is combined with several assumptions concerning similar
modes of behaviour across countries. Consider two countries for which
the values of n, a, 8 and s are the same. The Solow model implies that
these two economies must eventually have the same levels of per-capita
income, no matter what the size of their initial capital/labour ratios.
Initially "poor" countries will grow faster than the initially "rich"
countries, and they must converge (or "catch up") to the same standard of
living. Growth rates should correlate inversely with the initial level of per-
capita income. Comparable data sets for some 138 countries (annual data
since 1960) have been constructed recently, and this data appeared
(initially at least) to reject this convergence hypothesis. This finding was
one of the things which stimulated the new theories of endogenous growth
— some of which do not imply convergence.
Mankiw, Romer, and Weil (1992) have shown that the Solow
model is not necessarily threatened by this lack of convergence evidence.
After all, countries do have different savings rates and population growth
rates, so they are approaching different steady states. After accounting for
this fact, and for the fact that countries have invested in human capital to
different degrees, Mankiw, Romer and Weil find stronger evidence of
convergence. A number of studies have questioned the robustness of the
Mankiw, Romer, and Weil conclusions. Thus, while a new approach to
growth modeling may not be needed to rationalize the cross-sectional
evidence, many economists remain dissatisfied with the fact that the
Solow model does not attempt to endogenize, and therefore explain, the
steady-state growth rate. We explore the simplest versions of some of

221
these models in Chapter 11. Before doing so, however, we investigate
how basic exogenous growth theory has been modified to respect the
Lucas critique.

10.3 Exogenous Growth with Micro-Foundations

The following equations define a model that has well-defined

underpinnings based on constrained maximization.

= (r(1— r) — p — n)c — p(p + p)k

(k)= r+
k+ nk = (k)— c— gic — g
g = rrk + tw

All variables are defined on a per-effective-unit-of-labour basis. This

means, for example, that in full equilibrium, c will be constant. Per-capita
consumption will be growing at the productivity growth rate, but per-
capita consumption measured in efficiency units, c, will not be growing.
The first equation was explained in Chapter 4; it involves inter-
temporal utility maximization by finitely lived agents (individuals who
face a constant probability of death, and a life expectancy of (1/p)).
The second equation follows from profit-maximizing firms that
incur no adjustment costs while installing capital; capital is hired so that
its marginal product equals the rental cost. For simplicity, there is no
labour/leisure choice. The aggregate labour force is fixed at unity, so the
population growth rate, z, is zero. Total labour income, w, is determined
residually (and the associated equation is not listed above).
The third equation is the goods market clearing condition (often
referred to as the economy's resource constraint); net investment is output
minus household spending, replacement investment expenditure, and
government purchases.
Finally, the fourth equation is the government budget constraint. r
and t are the tax rates levied on interest and wage income (respectively).
Government spending is taken as exogenous and constant. The wage
income tax rate is determined residually by this equation, to balance the
budget — given the permanent reduction in the interest-income tax rate, r.
It is noteworthy that all the model's parameters are "primitive" in
the Lucas sense. Assuming the same Cobb-Douglas production function
as used above, the model involves seven parameters: p is a taste parameter,
p, a and Sue technology parameters, and t, r and g are policy parameters.
222
There are no parameters in the equations defining private-sector behaviour
that could be mixtures of fundamental taste and technology coefficients
and the parameters that define alternative policy regimes (such as "s" in
the Solow model). Thus, more legitimate policy analysis is possible in the
present setting.
Before proceeding with a formal analysis of the revenue-neutral
tax substitution in this model, it is useful to consider some intuition
concerning its full equilibrium. Ignoring ongoing productivity growth, the
cost of forgoing consumption for a period is the rate of time preference,
while the benefit of forgoing that consumption is the amount of output
that an additional piece of capital can generate (the marginal product of
capital). As a result, the stock of capital should be expanded to the point
that its marginal product equals the agents' rate of time preference. Is this
condition satisfied in a decentralized economy? Firms ensure that the (net
of depreciation) marginal product of capital equals r, while households
ensure that the rate of time preference equals r(1- r). The social optimum
is reached by decentralized agents only if r is zero. Thus, the interest-
income tax should be zero in the steady state. But should we not pay some
attention to what occurs during the transition to that full equilibrium? To
answer this question, we analyze a phase diagram.
Taking a total differential of the first three equations, eliminating
the change in the interest rate by substitution, and evaluating coefficients
at steady-state values, we have

[ = o[dk +[0 —cr]i dr

where

r (r —n) —1
0=
Lc(1— Of" p(p + p) r(1—r)—p— n

c can jump at a point in time, while k is predetermined at each instant, so

unique convergence to full equilibrium requires a saddle path (which
obtains as long as we assume that the determinant of 4 is negative, and
this is fully consistent with representative values of the model's
parameters). The reader can use the entries in the (1) matrix to pursue the
methods explained in section 6.2, and verify the particulars of the phase
diagram shown in Figure 10.3.

223
 Figure 10.3 Phase Diagram

c
e = 0 Locus

Figure 10.4 Dynamic Adjustment Following Lower Interest-Income Taxes

c Initial e=0 ,e = 0 After i r

Conditions
Constraint i
I / Saddle Path
1 /
1 ....
..
I / k=0
1 1/
1 -i 3

A cut in interest taxation shifts the a = 0 locus to the right; the economy
moves from point 1 to point 2 immediately, and then from point 2 to point
3 gradually. The formal analysis confirms that there is short-term pain
(lower c initially) followed by long-term gain (higher c in the new full
equilibrium), so the result is similar to that pictured in Figure 10.2. The
only difference stems from the fact that c in the present analysis is per-
224
effective-worker consumption, not per-person consumption (what was the
focus in Figure 10.2). As a result, there is no positive slope to the trend
lines in the version of Figure 10.2 that would apply to variable c.
Nevertheless, the formal analysis accomplishes two things: it confirms
that there is short-term pain followed by long-term gain outcome, and it
facilitates a calculation which determines whether the short-term pain is or
is not dominated by the long-term gain. To answer this question, we
calculate dPVIdr, where PV is the present value function:

PV = Se' In c,di
0

and is the social discount rate. This welfare function is based on the
instantaneous household utility function that was involved in the
derivation of the model's consumption function, so there should be no
controversy about this general form of social welfare function. But there is
controversy concerning what discount rate to use.
One candidate is 2= r(1—r)— n , the economy's net of tax and
growth rate of interest. This is what is used in standard applied benefit-
cost analysis — based as it is on the hypothetical compensation principle.
Another candidate is 2 = p , each individual's rate of time preference. For
internal consistency, this option must be used if agents live forever (that is,
ifp = 0). But in the over-lapping generations setting that is our focus here,
the 2 = p assumption is not so obviously appealing. It is not without any
appeal in this context, however, since it has been shown that this discount
rate is an integral part of the only time-consistent social welfare function
to be identified in the literature as consistent with this over-lapping
generations structure (Calvo and Obstfeld (1988)). Given the uncertainty
concerning what discount rate to use in public policy analysis, it is
instructive to consider both these options. Two rather different
conclusions emerge (and readers can verify this by following the
procedure outlined in section 6.2).
If the time preference rate of any one generation is used as the
social discount rate, we find that — even allowing for the short-term pain —
agents are better off if the interest-income tax is eliminated. But if the net
market interest rate is used, there is less support for this initiative. This is
because, with the net interest rate exceeding the time preference rate in the
overlapping generations setting, this decision rule involves discounting
the long-term gain more heavily. It turns out that aPv or =0 in this case,
so the pro-savings policy is neither supported nor rejected when the
hypothetical compensation criterion is used. This result is consistent with
225
Gravelle (1991) who argues that this tax substitution has more to do with
distribution than efficiency.
We return to this issue — the support or lack thereof for pro-
savings initiatives — in Chapter 12. In that chapter we will add interesting
features to the tax-substitution analysis. First, the equilibrium growth rate
will be endogenous, so fiscal policies will affect growth permanently.
Second, we will consider two groups of households — one very patient and
the other much less so. With each group having its own rate of time
preference — one above the economy's net interest rate, and the other
below it — we can be more explicit when we compare the short-term pain
and the long-term gain of fiscal policies that raise national savings. In the
meantime, we extend the present exogenous-growth model with just one
class of households so that we can consider a policy of government budget
deficit and debt reduction, for a small open economy.

10.4 A Benefit-Cost Analysis of Debt Reduction

Thus far, we have used the optimization-based exogenous-growth model

to examine a balanced-budget tax substitution — a revenue-neutral switch
in the tax system that involves the government relying more heavily on
wage taxation and less heavily on interest taxation. In recent decades,
however, many governments have not been balancing their budgets, and
the pro-savings initiative that has become the centre of attention has been
deficit and debt reduction. The purpose of this section is to bring our
analysis to bear on precisely this topic. To do so, we must add the deficit
and debt variables (and another differential equation — the accumulation
identity for the nation's foreign debt) to the model. This additional source
of dynamics creates a problem. We do not wish to attempt the drawing of
three-dimensional phase diagrams. Nor do we want to solve the set of
differential equations in algebraic form. As a result, we make no attempt
to analyze the entire time path of the economy's response to fiscal
retrenchment. Instead, we confine our attention to a comparison of the
initial, and the final, full equilibria. Since we wish to illustrate the
importance of fiscal retrenchment in quantitative terms, we calibrate by
selecting empirically relevant values for the model's slope coefficients.
The extended model is simplified slightly by our continuing to
abstract from population growth, and by our now dispensing with interest
taxation. The system is defined by the following five equations:

a = (r — p — n)c — p(p + p)(k + b — a)

f'(k) = r+8
226
 = (r — n)a —[f (k)— c — —(n + g)k — g]
b = d — nb
d=rb+g—t

As above, all variables are defined as ratios — with the denominator being
the quantity of labour measured in efficiency units. Units of output are
chosen so that, initially, these ratios can be interpreted as ratios to GDP as
well. The new variable is a, the nation's foreign debt ratio.
The first equation is the private-sector consumption function —
very similar to what has been discussed above. There is one difference
here; there is an additional component to non-human wealth. In addition
to the domestically owned part of the physical capital stock (k — a), there
is the stock of bonds issued by the government to domestic residents (b).
The second equation states that firms maximize profits and hire
capital until the marginal product equals the rental cost. For this
application, we focus on a small open-economy setting. As a result, the
interest rate is determined from outside (it is pinned down by the foreign
interest rate and the assumption of perfect capital mobility). Since k is the
only endogenous variable in the second equation, this optimal-hiring-rule
relationship pegs the capital stock. Thus, in analyzing domestic policy
initiatives, we take k as a constant (and set k = 0).
The third equation combines the GDP identity with the
accumulation identity for foreign debt (both written in ratio form). This
version of the accumulation identity states that the foreign debt-to-GDP
ratio rises whenever net exports falls short of the pre-existing interest
payment obligations to citizens in the rest of the world. The interest
payment term reflects the fact that — even when net exports are zero — the
foreign debt ratio rises if the growth in the numerator (the interest rate
paid on that debt this period) exceeds the growth in the denominator (the
GDP growth rate — n). Net exports are defined by the expression in square
brackets in the third equation.
The final two equations define the government accounting
identities (and they were discussed in Chapter 7 (section 4)).
As noted, to avoid advanced mathematics, we ignore the details
involved in the dynamic approach to full equilibrium, and focus on the
long run. Once full equilibrium is reached, all aggregates are growing at
the same rate as overall GDP (at rate n), so ongoing changes in the ratios
(all the dotted terms in these five equations) are zero. The five equations
then determine the equilibrium values of c, k, a, b and one policy variable
— which we take to be t. We take the total differential of the system, we set

227
 all exogenous variable changes except that in d to zero, and we eliminate
the changes in a, b and t by substitution. The result is:

dc I c =[(r — n)p(p + p)dd]l[nc((r — n)(r — n— p)— p(p + p))].

We evaluate this expression by substituting in the following representative

values for the parameters: r = 0.05, p = 0.025, n = 0.02, and p = 0.02.
These values ensure a rising consumption-age profile for each individual
generation. Also, they involve a life expectancy (lip) once an individual
reaches her initial working age of 50 years.
The illustrative calculation requires that an assumption be made
concerning the amount that the full-equilibrium deficit-to-GDP ratio is
reduced. We take Canada as an example of a small open economy. Since
Canada's debt-to-GDP ratio peaked in the mid-1990s, the government has
announcer/ a large/ he reached by 2020) that is SO percentage points
below the peak value. Thus, as an illustration, we consider a reduction in d
that is just sufficient to generate this reduction in b. Since d = nb in full
equilibrium, this 50-percentage-point drop in b requires a drop in d equal
to n times 0.5 (so the imposed change in d is -0.01).
Finally, the illustrative calculation requires an assumption
concerning the initial ratio of private consumption to GDP. We assume
0.585, since this value emerges from two sets of consideration. First, it is
consistent with Canadian data. Second, it is consistent with the full-
equilibrium restrictions involved in this model and other parameter value
assumptions that we need to make — those noted above and the following.
We have assumed that: the aggregate production function is Cobb-
Douglas with capital's share parameter equal to 0.3; the equilibrium
capital-to-GDP ratio is 3.0; capital depreciates at rate 0.05 per year; the
initial foreign debt-to-GDP ratio is 0.5; and the size of government (as a
proportion of GDP) is representative of the Canadian federal government
in the early 1990s.
With all these assumptions, the percentage change in consumption
turns out to be 3 percent. This means that, according to the model, the
fiscal retrenchment that is about half completed as this book goes to press
can be expected to raise Canadians' standard of living by 3 percent. In
2006 dollars, that amounts to an increase in material living standards of
more than $4000 per year for each family of four, and this is an annuity —
an amount that would be received every year into the indefinite future.
This analysis assumes that the country's risk premium is not reduced
through deficit reduction — an assumption that is consistent with evidence
(Fillion (1996)). If deficit reduction did lead to a lower domestic interest
rate, the long-term benefit of debt reduction would be even larger.
228
 As far as it goes, this analysis supports the policy of many
western governments in recent years — that of government budget deficit
and debt reduction. Of course, it must be remembered that all that has
been established here is that the long-term benefits are significant. Two
other considerations are integral parts of a complete benefit-cost analysis
of debt reduction. First, there is the short-term pain that must be incurred
to achieve this long-term gain. In the short-run (while some prices are
sticky) there will be a temporary recession. This may not be too large,
however,- since a flexible exchange rate can largely insulate real output
from aggregate demand shocks (see Mundell (1963), Fleming (1962), and
one of the practice questions suggested for this chapter). Even if there
were no temporary recession involved, consumption would have to fall in
the short run to "finance" the higher national saving (as discussed in
section 10.3). So there is short-term pain involved in debt reduction. The
second broad issue concerns the distribution of the long-term gain. We
shall now see that — unless the government uses the fiscal dividend to
improve labour's position — all of the gain from debt reduction can be
expected to go to the owners of capital. This prediction may seem so
alarming to some policy makers that it would be enough to cause them to
reject debt reduction — on distribution grounds — even if the aggregate
benefits (long-term gains) exceed the aggregate costs (short-term pain) by
a wide margin.
The proposition that the benefits of higher national saving extend
to individuals on lower incomes is often referred to as "trickle-down"
economics. We now evaluate this proposition.

Figure 10.5 Trickle-Down Economics in a Closed Economy

Percent

Demand
(Marginal Product)

Quantity of Capital

229
 Pro-savings initiatives often take the form of the government
using the tax system to stimulate private saving, instead of boosting
public-sector saving via deficit reduction. Some people oppose these tax
initiatives on equity grounds; they believe that only the rich have enough
income to do much saving, and so they are the only ones who can benefit
from these tax policies. Those who favour pro-savings tax concessions
argue that this presumption is incorrect; indeed they argue that most of the
benefits go to those with lower incomes. But since the process by which
these benefits "trickle down" the income scale is indirect, they argue,
many people do not understand it and thus reject these tax initiatives
inappropriately. We now examine whether this trickle-down view is
correct, first in a closed economy, and then in a small open economy.
Figure 10.5 shows the market for capital; the demand curve's
negative slope shows diminishing returns (the fixed quantity of labour
being shared ever more widely), and the supply curve's positive slope
captures that fact that savings is higher with higher returns. Equilibrium
occurs at point E. With just two factors of production, capital and labour,
the area under the marginal product of capital curve represents total output.
Thus, GDP is the entire area under the demand curve up to point E in
Figure 10.5. Further, since each unit of capital is receiving a rate of return
equal to the height of point E, capital's share of national income is the
rectangle below the horizontal line going through point E. Labour gets the
residual amount — the triangle above this line. A tax policy designed to
stimulate savings shifts the capital supply curve to the right, as shown in
Figure 10.5. Equilibrium moves from point E to point A, and total output
increases by the additional area under the marginal product curve (that is,
by an amount equal to the shaded trapezoid in Figure 10.5). So the pro-
saving tax initiative does raise per-capita output. But how is this
additional output distributed? The owners of capital get the dark shaded
rectangle, and labour gets the light shaded triangle. So even if capitalists
do all the saving and are the apparent beneficiaries of the tax policy, and
even if workers do no saving, labour does get something.
Furthermore, labour's benefit is not just the small shaded triangle
in Figure 10.5. Being more plentiful, capital's rate of return has been bid
down to a lower level. Since that lower rate is being paid on all units of
capital, there has been a transfer from capital to labour of the rectangle
formed by the horizontal lines running through points E and A. So, after
considering general-equilibrium effects, we see that capital owners may
not gain at all; labour, on the other hand, must gain. We conclude that, in
a closed economy (that can determine its own interest rate), the benefits of
pro-savings tax initiatives do trickle down.

230
 Figure 10.6 The Size and Distribution of Income: An Open Economy

Percent

Foreign
Supply
Demand
(Marginal Product)

Quantity of Capital

Figure 10.7 The Size and Distribution of Income: An Open Economy

Percent

Foreign
Supply
Demand
(Marginal Product)

Quantity of Capital

We now consider whether this same conclusion is warranted in a

small open economy. In Figure 10.6, the marginal product of capital curve
and the domestic supply curve appear as before; but the open-economy
graph contains one additional relationship — a line that shows the supply
of savings on behalf of lenders in the rest of the world (to this economy).
For a small economy, this world supply curve is perfectly elastic, at a
yield equal to what capital can earn when it is employed in the rest of the
world (assumed to be the height of the horizontal supply curve in Figure
10.6. Equilibrium occurs at point E, and GDP is the entire trapezoid under
the marginal product curve up to point E.
231
 The rate of return for each unit of capital is given by the height of
the foreign supply curve, so capital owners earn the shaded rectangles
under that line, while labour receives the upper triangle. The domestic
supply of savings curve indicates the proportion of capital's income that
goes to domestically owned capital (the dark shaded rectangle) as well as
the proportion that is the income of foreign owners of capital (the light
shaded rectangle).
We now consider, as before, a tax break for domestic capitalists if
they save more. This policy shifts the domestic supply curve to the right,
as shown in Figure 10.7. In this case, equilibrium remains at E. There is
no growth in the amount of goods produced. But there is growth in the
amount of income that domestic residents receive from that activity,
because the income of domestic individuals who own capital increases by
the amount of the shaded rectangle between points A and B in Figure 10.7.
Since labour is still working with the same overall quantity of capital,
labour's income — the unshaded triangle — is unaffected. So the entire
increase in national income goes to the owners of capital, and we must
conclude that the critics of trickle-down economics are justified if they
apply their critique to a small open economy.

10. 5 Natural Resources and the Limits to Growth

In all our models considered thus far in the book, we have assumed that
there are only two inputs in the production process, and that both can
expand without limit. Many concerned citizens see this as a fundamental
limitation of mainstream economics. This section briefly considers this
issue — in two stages. In the first, we introduce a third factor — land. It is
assumed that the quantity of land cannot grow, and we ask whether this
fact necessarily brings the growth in living standards to a halt in the long
run. In the second stage, we consider a bigger challenge. Instead of
considering a factor that cannot grow, we focus on one whose supply
continually shrinks. This is the appropriate assumption for non-renewable
resources. Again, the focus of the analysis is on whether ongoing growth
remains possible, and if so, is it likely. Our treatment follows that in Jones
(2002, 170-177) very closely.
Fixed land (denoted by 7) is included in the following (otherwise
standard) production function:

Y = BK"T fi

232
B is the exogenously determined productivity growth index (so AIB= y).
Land is fixed, and labour (the population) grows at a constant rate (T = 0
and k I N = z). We wish to focus on balanced growth paths (where output
and capital grow at the same rate), so we re-express the production
function in such a way that highlights the capital-output ratio. After
dividing through by Ya we have

Y B(1/(1-a))(K yprio-anTo/o-a »wo-c841-am

After taking logs and then time derivatives, and noting that T and (KIY)
are constant, we get an expression for the output growth rate (that is, for
the sum of the per capita output growth rate plus z):

Output growth rate = y * +(1 — fi*)z

where 7* = y /(1 — a) and /3* = fi/(1 — a). Subtracting z from both sides,
we end with

Growth in per capita output = y *—fl* z

This equation indicates that there is a never-ending race between technical

progress and the diminishing returns that is imposed by the fixed factor.
The race is necessarily lost if there is no technical progress. This is the
Malthusian (dismal science) prediction. In this case there is negative
growth in per capita income, and as time proceeds, the level of per capita
income approaches zero. The strength of this negative effect is enhanced,
the faster is the rate of population growth, since the fixity of land matters
more in that case.
As can be expected on intuitive grounds, things are even more
discouraging when the third factor is not only unable to grow, but is
actually shrinking. Let E denote such a factor — the amount of energy used
up each period. The production function in this case is

Y = BICEPN 1'-/3

If R denotes the remaining stock of energy (and E is what is used up each

period), the differential equation that defines the resources that are lost is
R = —E. Micro models of optimal resource depletion often lead to the
proposition that the amount used each period should be a constant
proportion of the remaining stock. We assume that this policy is followed,
233
so we assume El R= 0. We can think of 0 as a policy variable. In the
basic Solow model, the policy variable was a constant investment rate. In
this model, the policy variable is a constant dis-investment rate. The last
two relationships imply a linear differential equation: R = —0, and the
solution of this equation is R= Roe- '. This solution gives us an equation
for E that can be substituted into the production function: E =0Roe-*
After substituting in, and proceeding with the same steps that were
involved in the analysis of land, we get

Y= B" -"(K / Y) ("' (OR 0 e -°' )( '" 1-a AT(1-6641-a")

andweith

Growth in per capita output = 7* —fi* (0 + z)

This result indicates that the drag on growth is bigger when the natural
resource is shrinking. Nordhous (1992) has estimated these parameters,
and he concludes that the annual growth-retarding effect (the second term
on the right-hand side) is about one-third of one percentage point. The
implication is that, as long as the productivity growth rate can be expected
to exceed this amount, we can enjoy rising living standards — even though
we are running out of non-renewable resources. A second reassuring point
can be made. If resources really are running out as we have assumed in
this analysis, the real prices of these resources should be increasing. In
many cases they are not. In other words, the rate of discovery of new
supplies appears to be dominating our rate of using the resources. These
observations suggest that the model may have focused on a case that is
more stringent than what we have yet had to confront. On the other hand,
the Cobb-Douglas function implies an elasticity of factor substitution that
may very much overstate how easy it is for firms to make do with less of
the dwindling resources, and there has been no mention of pollution
externalities. Thus, it is worrisome that concern about non-renewable
resources is left out of the standard analyses of fiscal policy and growth.
Despite this concern, space limitations make it necessary for us to follow
this convention in the remaining two chapters of the book.
It is worth noting that this discussion of non-renewable resources
has involed our switching from exogenous-growth to endogenous-growth
analysis, since in this case, the full-equilibrium growth rate is affected by
a policy parameter, 0. We pursue a more systematic exploration of
endogenous growth in the next chapter.
234
10.6 Conclusions

The purpose of this chapter has been to position readers so that they can
benefit from our exploration of "new" growth theory in the remainder of
the book. Traditional growth analysis is "old" in that (originally) it lacked
formal optimization as a basis for its key behavioual relationship — the
savings function, and (even today) it involves a rate of technological
progress that is exogenous. In the analysis that has been covered in this
chapter, we have seen that the first dimension of oldness has been
removed through the addition of micro-foundations. We consider ways of
endogenizing the rate of technical progress in the next chapter.
The basic policy prescription that follows from both the original
and the micro-based version of traditional analysis is that pro-savings
policies can be supported. These initiatives result in a temporary increase
in the growth rate of consumption, and a permanent increase in the level
of per capita consumption. Calibrated versions of these models support
the conclusion that, even without a permanent growth-rate effect, higher
saving leads to quite substantial increases in average living standards. If
higher saving leads to capital accumulation (as it does in a closed
economy), even a poor labourer who does not save benefits from a pro-
savings fiscal policy. Such individuals benefit indirectly, since they have
more capital with which to work. But if the higher saving leads only to
increased domestic ownership of the same quantity of capital, the poor
labourer is not made better off. So, in some circumstances, the increase in
average living standards does not involve higher incomes for everyone.
Our task in the next chapter is to establish whether this basic conclusion —
that a pro-savings policy is supported in models that do not stress income
distribution issues, but it may not be otherwise — holds in an endogenous
productivity growth setting.

235
As a concise survey of the developments in modern macroeconomics, this book bridges
the gap between intermediate-level texts and advanced material. By highlighting the New
Neoclassical Synthesis, it draws attention to recent work which simultaneously emphasizes the
rigour of the New Classical approach and a focus on market failure that is the essence of the
Keynesian tradition. In addition to stabilization policy issues, there is extensive coverage of natural
unemployment rate theories, and both old and new growth analysis. At the upper undergraduate
level, the book can be used on its own; at the introductory post-graduate level, it represents
a much-needed complement to journal readings and the more advanced research monographs.

The user friendly exposition gives equal billing to explaining technical derivations and to exposing
the essence of each result and controversy at the intuitive level. Calibrated versions of the models
allow readers to appreciate how modern macroeconomics can inform central policy debates.
Many topical issues are highlighted: the Lucas critique of standard methods for evaluating policy,
credibility and dynamic consistency issues in policy design, the sustainability of rising debt
levels and an evaluation of Europe's Stability Pact, the optimal inflation rate. the implications of
alternative monetary policies for pursuing price stability (price-level vs inflation-rate targeting.
fixed vs flexible exchange rates), tax reform (trickle-down controversies and whether second-best
initial conditions ease the trade-offbetween efficiency and equity objectives), theories of the natural
unemployment rate and the possibility of multiple equilibria, alternative low-income support
policies. and globalization (including the alleged threat to the scope for independent macro policy).

Using basic mathematics throughout, the book introduces its readers to the actual research
methods of macroeconomics. But in addition to explaining methods, the author presents the
underlying logic at the common-sense level, and with an eye to the historical development
of the subject. As with the earlier editions, both instructors and students will welcome
having the exciting developments in modern macroeconomics made more accessible.

About the author: William Scarth is professor of economics at McMaster University. where
he has been awarded the President's Award for Best Teacher, and the McMaster Student
Union Lifetime Teaching Award. In addition to publishing many articles in academic journals
(in the areas of macroeconomics, labour economics. international trade and public finance),
Professor Scarth has authored three other textbooks. and he is a Research Fellow at the C.D.
Howe Institute. Canada's leading nonprofit policy institute. Professor Scarth's recent work
concerns how both globalization and a commitment to high productivity growth affect the
ability of governments to provide low-income support policy within small open economies.

INNOVATION PRESS