Essentials of Advanced Macroeconomic Theory 2012

Downloaded by [Hacettepe University] at 20:58 10 April 2017
Essentials of Advanced
Macroeconomic Theory
Trying to summarize the essentials of macroeconomic theory, in the wake of the

financial crisis that has shaken not only Western economies but also the macro-
economic profession, is no easy task. In particular, the notion that markets are
self-correcting and always in equilibrium appears to have taken a heavy blow.
However, the jury is still out on which areas should be considered as failures and
which constitute the future of research.
The overall aim of this text is to provide a compact overview of the contributions
that are currently regarded as the most important for macroeconomic analysis and
to equip the reader with the essential theoretical knowledge that all master’s degree
students in macroeconomics should be acquainted with.
The result is a compact text that should act as the perfect complement to further
study of macroeconomics: an introduction to the key concepts discussed in the
journal literature, and suitable for students from upper undergraduate level through
to PhD courses.
Ola Olsson is Professor in the Department of Economics at the University of
Gothenburg, Sweden.
Routledge Advanced Texts in Economics and Finance
1 Financial Econometrics 10 Financial Econometrics,

Peijie Wang second edition

Peijie Wang
2 Macroeconomics for Developing
Countries, second edition 11 Development Finance
Raghbendra Jha Debates, dogmas and new
3 Advanced Mathematical directions
Economics Stephen Spratt
Rakesh Vohra
12 Culture and Economics
4 Advanced Econometric Theory On values, economics and
John S. Chipman international business
Eelke de Jong
5 Understanding Macroeconomic
Theory 13 Modern Public Economics,
John M. Barron, Bradley T. Ewing second edition
and Gerald J. Lynch Raghbendra Jha
6 Regional Economics 14 Introduction to Estimating
Roberta Capello Economic Models
7 Mathematical Finance Atsushi Maki
Core theory, problems and
15 Advanced Econometric
statistical algorithms
Theory
Nikolai Dokuchaev
John Chipman
8 Applied Health Economics
Andrew M. Jones, Nigel Rice, 16 Behavioral Economics
Teresa Bago d’Uva and Edward Cartwright
Silvia Balia
17 Essentials of Advanced
9 Information Economics Macroeconomic Theory
Urs Birchler and Monika Bütler Ola Olsson
Ola Olsson
Essentials of Advanced
Macroeconomic Theory
First published 2012
by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
Simultaneously published in the USA and Canada
by Routledge
711 Third Avenue, New York, NY 10017
Routledge is an imprint of the Taylor & Francis Group, an Informa business

c 2012 Ola Olsson
The right of Ola Olsson to be identified as author of this work has been
asserted by him in accordance with the Copyright, Designs and Patent
Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced

or utilized in any form or by any electronic, mechanical, or other means,
now known or hereafter invented, including photocopying and recording,
or in any information storage or retrieval system, without permission
in writing from the publishers.
Trademark notice: Product or corporate names may be trademarks or
registered trademarks, and are used only for identification and explanation
without intent to infringe.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
A catalog record for this book has been requested
ISBN: 978-0-415-68505-4 (hbk)
ISBN: 978-0-415-68508-5 (pbk)
ISBN: 978-0-203-13993-6 (ebk)
Typeset in Times New Roman by Sunrise Setting Ltd, Torquay, UK.

Contents
List of figures ix
Preface xi
1 Introduction 1
1.1 The issues 1
1.2 The national accounts identity 2
1.3 Outline 3
PART I
The Long Run 5
2 The Malthusian World 7
2.1 Introduction 7
2.2 The law of diminishing returns 7
2.3 The Malthusian trap 9
2.4 Endogenous fertility 11
2.5 The collapse of the Malthusian link 14
3 The Solow Growth Model 19

3.1 Introduction 19
3.2 Basic assumptions 19
3.3 Dynamics 20
3.4 Equilibrium 21
3.5 Implications 22
3.6 Extensions 25
4 Endogenous Growth Theory 29

4.1 Introduction 29
4.2 AK model 29
4.3 Endogenous technological change 30
4.4 Romer’s product variety model 34
vi Contents
4.5 Schumpeterian growth models 39
4.6 Innovation versus imitation 40
5 The Overlapping Generations Model 42

5.1 Household optimization 42
5.2 Endogenous saving 47
5.3 Endogenous growth 49
PART II
The Short and Medium Run 51
6 Equilibrium Business Cycles 53

6.1 Technology shocks to production 53
6.2 Labor demand 55
6.3 Households 55
7 Financial Crises 58
7.1 Basic assumptions 58
7.2 Banks 59
7.3 A bank run equilibrium 61
7.4 Foreign credit 62
7.5 Short-term debt 63
7.6 Liberalizing international credit markets 64
8 Consumption and Saving 65

8.1 The Keynesian consumption function 65
8.2 Friedman’s critique 66
8.3 The permanent income hypothesis 66
8.4 An example 68
8.5 The random-walk model 71
8.6 Precautionary saving 73
8.7 Interest rates and time discount rates 75
8.8 Relative consumption 77
8.9 Time inconsistency 78
9 Investment and Asset Markets 80

9.1 The Keynesian investment function 80
9.2 The firm’s investment decision 80
9.3 Adjustment costs 83
9.4 The housing market 85
10 Unemployment and the Labor Market 88

10.1 Labor market disequilibrium 88
Contents vii
10.2 Efficiency wages 89
10.3 The Shapiro–Stiglitz model 90
10.4 Insider–outsider models 94
10.5 Search and matching models 95
PART III
Macroeconomic Policy 99
11 IS–MP, Aggregate Demand, and Aggregate Supply 101
11.1 Aggregate expenditure and the multiplier 101
11.2 The IS–MP model 103
11.3 Aggregate demand 105

11.4 Aggregate supply 106
11.5 Financial intermediation 110
11.6 New Keynesian models 114
12 Public Finance and Fiscal Policy 115

12.1 The government budget identity 115
12.2 Ricardian equivalence 116
12.3 Tax smoothing 117
12.4 Political economy of government debt 118
12.5 Debt financing versus debt forgiveness 121
13 Inflation and Monetary Policy 126

13.1 The quantity theory of money 126
13.2 Inflation and the money market 127
13.3 Time inconsistency in monetary policy 127
13.4 Political business cycles 132
13.5 The Taylor rule 136
13.6 Seigniorage 137
14 The Open Economy 139

14.1 Open economy accounting 139
14.2 A representative agent framework 141
14.3 The Mundell–Fleming model 144
14.4 Exchange rate overshooting 146
14.5 Currency unions 148
15 Mathematical Appendix 150

15.1 Introduction 150
15.2 Derivatives of some basic functions 150
15.3 Differentiation rules 151
15.4 Chain differentiation 152
viii Contents
15.5 Implicit function differentiation 153
15.6 Applications to macroeconomics 153
15.7 Basic properties of exponents and logarithms 154
Notes 156
Bibliography 161
Index 164
List of figures
2.1 The Malthusian trap 10

2.2 Technological progress in the Malthusian model 11
2.3 Equilibrium population size 13
2.4 (a) A Malthusian trap in the evolution of technology and
education in the Galor–Weil model. (b) Escape from the
Malthusian trap in the Galor–Weil model due to an increase in
population (L 1 > L 0 ) 17
3.1 The neoclassical growth diagram 22
3.2 Impact on growth rate per worker of an increase in the savings
rate 23
5.1 A strictly concave utility function and its associated marginal
utility 43
8.1 Example of consumption and income over the life cycle 71
8.2 Precautionary saving and the third derivative of the utility
function 74
8.3 Precautionary saving with an increase in uncertainty 75
8.4 Optimal consumption paths over time for different levels of r
and ρ 76
9.1 Equilibrium price of housing following falls in interest rates and
in income 86
10.1 Unemployment 89
10.2 Example of an efficiency wage equilibrium 91
10.3 Equilibrium in the search and matching model 97
11.1 The Keynesian cross and the multiplier impact of an increase in
government spending 102
11.2 The IS–MP curves 105
11.3 The AD–AS curves with sticky wages 106
11.4 Short- and long-run Phillips curves 108
11.5 Impact of an increase in income on household demand and
supply of lending 111
11.6 (a) Introducing financial intermediaries and credit supply
frictions. (b) Supply and demand for financial intermediation as
a function of the credit spread 112
x List of figures
11.7 Impact of an increase in the credit spread 113
13.1 Equilibrium inflation under discretionary monetary policy 130
13.2 Political business cycle effects over time on output (y) and
inflation (π) of an unexpected (ρ > 1/2) left-wing party election
victory over an incumbent right-wing party 135
13.3 The inflation-tax Laffer curve 138
14.1 Expansionary monetary policy with fixed exchange rates and
perfect capital mobility 146
14.2 Expansionary fiscal policy with fixed exchange rates and
imperfect capital mobility 147
Preface
Trying to summarize the essentials of macroeconomic theory in 2011 is no easy

task. The financial crisis has shaken not only Western economies but also the
macroeconomic profession, and the field has recently been the object of strong
criticism. In particular, the notion that markets are self-correcting and always in
equilibrium, as emphasized by the dominant dynamic general equilibrium (DGE)
tradition, appears to have taken a heavy blow. However, the jury is still out on the
areas that should be considered as failures and the areas that constitute the future
of research.
The overall aim of this text is to provide a compact overview of what are cur-
rently regarded as the most important theoretical contributions to macroeconomic
analysis. It is intended to present the core of advanced macroeconomic theory, the
essential knowledge that all master’s degree students in macroeconomics should
be equipped with. As alluded to above, any such compilation of relevant know-
ledge inevitably has to make difficult judgments on what should be included and
what should not. Views on the proper priority of relevant macroeconomic theory
will very much depend on the orientation of the individual scholar and it is quite
likely that such priority lists vary widely among macro teachers in the profession.
Although I choose to include some theoretical contributions that are not usually
included in macro textbooks and exclude others, I would certainly not claim that
my choices are necessarily the most “appropriate” ones. Rather, they reflect to a
large degree my own orientation and interests.
In this book, I have made the following basic priorities. Compared with many
existing texts, I have emphasized the long run rather than the short run. The reason
is partly that an increasing share of the most recent research in the leading jour-
nals is focused on long-run issues and partly that it appears to me that the theory
of short-run (business cycle) fluctuations in the economy is not in good shape at
the moment. Unlike any other macro textbook that I know of, this text includes
a presentation of the Malthusian growth model. In order to understand the pat-
tern of macro developments in countries like India and China and the historical
pattern in Europe, it seems essential to include a discussion about the interaction
between population and economic growth in economic development as discussed
by Malthusian theory.
xii Preface
Sections on long-run growth are included mainly at the expense of short-run
analysis such as business cycle theory. This text includes one chapter on the key
features of DGE modeling, but I do not delve deeply into this type of framework.
A more extensive section is devoted to a model of financial crises and bank runs.
In the chapters on monetary and fiscal policy, I have further chosen to focus on
models emphasizing political economy and institutional features. The political
economy of market failures and imperfect institutions is currently a very active
research agenda in economics and this text taps into some of that recent literature.
In the chapter on consumption, I have further tried to include some of the most
recent insights gained from research in behavioral economics.
Unlike the standard advanced textbooks in microeconomics, the models that
are surveyed below do not emerge from a core set of assumptions that are then
extended and applied in different directions, neither is the analysis based on a
small number of key equations as in certain macro textbooks. Despite recent
efforts, macroeconomic theory is still not a coherent body of theory in the same
way as microeconomics or econometrics is. The DGE program was clearly an
attempt to provide such a coherent framework, but recent events have put that
effort into a less favorable light.1
The “workhorse” model for most chapters is, however, the well-known two-
period, representative agent model of consumer optimization with a utility func-
tion given by U = u(c1 ) + βu(c2 ) that is maximized subject to varying constraints.
The majority of all models presented are thus “micro-founded”, which should
make the links to microeconomic theory more easily recognizable. Several of the
chapters start off with a typical Keynesian model, which is then contrasted to
models founded in individual household behavior and characterized by rational
expectations and intertemporal optimization.
A key motivation of this text in comparison with the literature in the field is its
condensed form. As a rule, most advanced textbooks in macroeconomics are about
500–600 pages long, mixing theory with somewhat randomly chosen empirical
applications. This text is intended to be less than half as long as a standard textbook
and to serve more or less as a reference source on modern macro theory. It is my
hope that it will direct impatient readers (like myself) quickly to the main results.
Admittedly, this writing approach might run the risk of alienating readers who rely
more on texts focusing on the intuition behind models. Such readers might want
to gather deeper intuitive insights from other sources, for instance from articles or
more comprehensive macroeconomic textbooks.
A further and important delimitation of this work is that it will not discuss
empirical tests of the theories surveyed. The reason is partly that I want to keep
the text compact, but also that it is my impression that researchers seem to be
somewhat more in agreement about what they think are the most relevant models,
as compared with what they consider to be the most successful empirical tests of
those models. Theory also changes more slowly than the stock of empirical results.
This text should ideally be complemented with selected readings on empirical
motivations and applications of the theories presented.
Preface xiii
The book is intended to be suitable for a master’s course in macro theory, last-
ing for about half a semester. Certain sections or chapters might also serve as an
introduction to macroeconomics for nonspecialized graduate students. Readers are
presumed to be relatively well equipped with calculus and algebra. A fairly strong
background knowledge of both micro and macro theory is taken for granted.
The text has emerged from my experience of teaching advanced macro the-
ory at the University of Gothenburg. Special thanks are due to my former teacher
and most ardent reviewer Wlodek Bursztyn for having provided extensive com-
ments on several previous versions (we still do not agree on certain aspects . . .).
I have also benefited from many valuable discussions with Heather Congdon Fors
and Per Krusell regarding macro theory in general. Oded Galor, Halvor Mehlum,
Bo Sandelin, Joachim Smend, Olof Johansson-Stenman, David Weil, and three

anonymous referees have commented on certain sections and provided valuable
input. I am also indebted to several students who have commented on parts of
the manuscript and pointed out errors. The mathematical appendix is a modified
variant of a section prepared by Elias Tsakas. Lastly, I have very much appreci-
ated my ongoing discussions with Rob Langham at Routledge, who encouraged
me to write this book. All comments and suggestions for improvements are much
appreciated.
Ola Olsson
(ola.olsson@economics.gu.se)
Floda, Sweden
May 2011
1 Introduction
1.1 The issues

Macroeconomics is the study of the aggregate economy of a country. It seeks
to understand how the economic decisions of individual persons and firms are
translated through markets into aggregate economic outcomes. Variables of inter-
est in macroeconomic analysis are, for instance, the level and change of gross
domestic product, the aggregate level of investment, government debt, inflation,
and unemployment.
The macro economy affects individuals both directly and indirectly. More or
less all individuals in modern economies are, for instance, subject to income tax-
ation, are affected by bank interest rates, receive some kind of government subsidy,
and control household budgets whose real value depends on the aggregate price
level.
Macroeconomics is distinguished from microeconomics primarily in the sense
that the ultimate dependent variables are different. Microeconomics studies the
behavior of individuals or firms in order to understand individual choices. Macro-
economics also increasingly starts off with the modeling of a “representative”
individual or firm that maximizes utility or profits, but the dependent variable is
aggregate outcomes on a national level.
Macroeconomic outcomes are central to politics within countries. Questions
like those below are discussed at more or less every general election (as well as in
between elections) in the Western world:
• What policies are most effective against unemployment, and how should the
government or the central bank fight inflation?
• How can economic growth be increased?
• How should governments stabilize short-run fluctuations and business cycles?
• What is a sustainable level of government debt?
In an international economy with interwoven markets, macroeconomics is also a

central theme in international politics. A recurrent issue in international economic
policy-making has been the determination of exchange rates between currencies.
Another topic which has an important impact on international relations is current
2 Introduction
account levels and deficits or surpluses in the balance of trade. In summary,
macroeconomics is a central field for anyone with an interest in economic policy
or economic development.
1.2 The national accounts identity

The main building block of macroeconomic theory is the national accounts iden-
tity, which shows the gross domestic product (GDP) of a country. Total GDP,
denoted by Yt , measures the total value of all final goods and services that have
been produced in a country during one year.1 GDP is an example of a flow vari-
able, one that is measured per unit of time. The other major type of variables are
stock variables, such as the level of the capital stock or of accumulated public debt,
which measure the level at a given point in time. We will return to this latter type
of variables later.
Total GDP can be calculated in three ways, which all should yield the same
result. The most common characterization of GDP is to study it from the user side,
i.e. what total GDP is spent on. This is the expenditure approach to measuring total
GDP and can be described by the equation
Yt = Ct + It + G t + X t − Mt (1.1)
In this key equation of macroeconomics, Yt is total GDP as before at time t, Ct

is aggregate private consumption, It is aggregate investment, G t is government
spending on goods and services, X t is total exports from the country, and Mt is
total imports. All these variables are flow variables.
Let us briefly take a closer look at these components of GDP. Aggregate private
consumption Ct is typically the biggest item on the expenditure side and usu-
ally amounts to about half of total GDP. It includes personal spending on durable
and nondurable goods and services during a year.2 It is more specifically “gross
domestic private investment” during one year. Investment is the acquisition of
durable goods (with an expected life of more than one year) to be used as factors
of production in the future, typically including machines and factories. Aggre-
gate investment can in turn be split up into nonresidential investment, residential
investment, and change in business inventories.
Government spending on goods and services G t includes both government
consumption (such as on salaries for teachers and judges) and investment (for
instance, in government buildings). It also includes spending at all levels of gov-
ernment: state, regional, and local. Exports X t is the value of goods and services
produced within the country that are sold to people in foreign countries. Like-
wise, total imports Mt shows the value of goods and services produced outside the
country that are bought by people inside the country.
This expenditure accounting of total GDP must be matched by the total income
that all factors of production in the country earn during a given year. The income
side of GDP is therefore
Yt = wages + rental incomes + profits + interest . . . (1.2)

Introduction 3
These incomes are eventually controlled by households in one way or another and
are used for the expenditures above.
Furthermore, the total value of expenditures and incomes must be matched
by the value added of aggregate production during a year. This is shown in the
production side of the national accounts:
Yt = agriculture + · · · + manufacturing + · · ·
+ professional and business services . . . (1.3)
where the total values of production from all sectors of society are added.
The equation for the user side of GDP in (1.1) is the backbone of macro-
economic theory from which the subsequent analysis is derived and extended in
numerous ways. It also serves as a introduction to an outline of the exposition
below.
1.3 Outline
The following chapters are organized as follows. We start off by analyzing the
long-run determinants of total GDP, i.e. growth theory. Chapter 2 is devoted to
the Malthusian model of growth, Chapter 3 presents the neoclassical (or Solow)
growth model and its extensions, and Chapter 4 deals with endogenous growth
models where technological progress plays a prominent role. In Chapter 5, we
develop the overlapping generations model, which is long-run in nature and which
is used also in the chapters ahead.
After the long run, we take a look at macroeconomic theory in the short and
medium run. In Chapter 6, we study models on the behavior of total GDP and its
components over the business cycle, i.e. a period of roughly five years. Chapter 7
discusses a recurrent phenomenon in capitalist economies: financial crises and
bank runs. We then move on to analyze specifically the constituent parts of the
expenditure side: consumption (and saving) in Chapter 8, investment and asset
markets in Chapter 9, and one of the key markets for understanding the macro
economy, the labor market, in Chapter 10.
In the third main section of the book, we analyze a broad range of topics related
to macroeconomic policy. We begin by presenting the traditional IS–MP, aggre-
gate supply and aggregate demand frameworks and the refinements suggested by
the rational expectations view and the new Keynesian view in Chapter 11. We then
go on to public finance and fiscal policy (Chapter 12), and inflation and monetary
policy (Chapter 13). Lastly, we discuss international aspects of economic policy
in Chapter 14.
Some basic mathematical results that are used throughout the text are provided
in an appendix.
Part I
The Long Run

2 The Malthusian World
2.1 Introduction
In this chapter, we will describe a model of long-run economic growth that was
applicable to all countries in the world up until the industrial revolution and which
still is a highly relevant model for some developing countries. In this “Malthusian
world” there is a strong link between income per capita and population growth, so
that anything that increases aggregate income in a society will soon be neutralized
by an increase in the size of the population. Hence, even despite periods of rapid
technological progress, income per capita will remain at a fairly constant level.
Recent empirical work on historical data has shown that standards of living indeed
appear to have been roughly similar in Assyria around 1500 BC, in Egypt during
Roman times, and in late eighteenth-century England (Clark 2007). This section
is motivated by this stylized fact from economic history.
The main insights behind this model were proposed by Thomas Malthus (1798)
but also critically hinge on the principle of diminishing returns to factors of
production, and on theory that was further developed by David Ricardo. In the
sections below, we will briefly discuss the theory of diminishing returns, the
Malthusian model of long-run stagnation, and reasons for the eventual collapse
of the Malthusian link. We will also show how fertility can be endogenously
determined within a representative household.
2.2 The law of diminishing returns

One of the most fundamental building blocks of economics is the principle of
diminishing returns. In production theory, diminishing returns means that if we
produce a good, for instance shirts, by using two factors of production – say,
workers and knitting machines – then if we keep the stock of machines fixed,
each additional extra working hour will result in a smaller and smaller addition
of new shirts on the margin. The reason is that workers need knitting machines
to produce shirts and there will eventually be crowding effects if more and more
workers have to share the same machines. This type of diminishing returns to labor
or any other factor of production will be present in a single factory as well as in
the economy as a whole, aggregated on a national level.
In order to illustrate this principle more rigorously, let us consider its mathe-
matical properties. The fundamental assumption that all growth models share is
8 The long run
that we can describe the total production in an economy (equivalent to total GDP)
by an aggregate production function. In the pre-industrial era (and still in some of
the poorest developing countries), the two most important factors of production
were land and labor. Let us refer to the total number of workers in the economy as
L and the total available land area as X . For simplicity, let us further imagine that
the size of the labor force equals the size of the total population in a country. Land
X is in fixed supply whereas population levels will change in response to changes
in output, birth rates, and death rates, as will be shown below.
The aggregate production function for total output in the economy as a whole
can be described as
Y = AX α L 1−α
(2.1)
In this expression, Y is total output, A is a technology shift parameter, X is the

fixed quantity of land, and L is the size of the population/labor force. The spe-
cific type of production function shown here is referred to as the Cobb–Douglas
functional form. A might be thought of as capturing more broadly the general
characteristics of the country in terms of, for instance, the state of technological
knowledge or the quality of institutions. The parameter α (0 < α < 1) is formally
referred to as the output elasticity of land. Equivalently, 1 − α is the output elas-
ticity of labor. These elasticities show how total output Y responds to an increase
in the production factor in question.
If we consider labor, its marginal product is defined as
α
∂Y α −α X
= (1 − α)AX L = (1 − α)A >0 (2.2)
∂L L
The marginal product is always positive, indicating that if we increase the labor
force in the country by one person, that person will always produce some extra out-
put. However, as we see from (2.2), the marginal product will fall as we increase
the number of workers since L now appears in the denominator. Formally, we can
also show this by taking the second derivative:
∂ 2Y
= −α(1 − α)AX α L −α−1 < 0
∂ L2
Thus, the Cobb–Douglas functional form ensures that Y has a concave relationship
with L.
Note also that we can express total output per worker or per capita y as
α
Y AX α L 1−α X
=y= =A = Ax α (2.3)
L L L
where x is land per capita. Output per capita is one of the most often used indica-
tors of standards of living in a country and is highly correlated with factors such
The Malthusian world 9
as life expectancy, levels of education, the rule of law, and political freedom. We
will refer to it frequently in the pages ahead.
From the expression in (2.3), it is clear that output per capita will fall with a
greater level of population in the Malthusian model. On the one hand, one more
worker means a little bit more output, but it also means another person to share
total production with. The latter negative effect dominates. The first and second
derivatives show us that
∂y ∂2y
= −α AX α L −α−1 < 0, = α(1 + α)AX α L −α−2 > 0
∂L ∂ L2
which confirms what we just said.

2.3 The Malthusian trap

The key feature of the Malthusian model is that output per capita is strongly linked
to population growth. In order to illustrate this, let us consider a very simple setup
where the level of population at current time t is equal to last year’s population
level L t−1 , plus the number of births during year t, Bt , minus the number of deaths
during the same year, Dt :
L t = L t−1 + Bt (yt−1 ) − Dt (yt−1 ) (2.4)
The central feature is that both Bt and Dt are functions of output per capita
lagged one year, yt−1 . Birth rates increase with yt−1 such that Bt (yt−1 ) > 0

whereas Dt (yt−1 ) < 0. With an increase in output per capita, the supply of food
increases, which allows families to grow. Likewise, the higher food consumption
associated with a larger output at a given size of the population means that people
die to a lesser extent from disease.1
The relationships between births, deaths, and the sizes of population and output
per capita are drawn in Figure 2.1. In the figure, we assume for simplicity that Bt
and Dt are linear functions of y. In the lower graph, we have drawn the negative
convex relationship between y and L as stipulated by (2.3). The main insight from
the figure is that output per capita will tend to converge towards an equilibrium
level given by y ∗ . In the Malthusian model, y ∗ is often referred to as the subsis-
tence level since it is inevitably at a quite low level that is not far above the level
of income that allows people to survive. At this level, population ceases to grow
and L t − L t−1 = Bt (y ∗ ) − Dt (y ∗ ) = 0.
To see that y ∗ is an equilibrium, consider a relatively high level of initial out-
put y 0 . At this level of affluence, many children are born and relatively few people
die of disease. Hence population levels increase, which pushes y to the left in the
figure. The economy comes to rest again at y ∗ where output per capita has fallen
to its subsistence level.
An even more grim situation is of course if the economy instead starts off to the
left of y ∗ . In this situation, people are starving, few children are born, and many
10 The long run
Birth rates, death rates

B(y)
D(y )
Y/L
L* y = A(X/L)α
L0
Y/L = y
y* y0
Figure 2.1 The Malthusian trap.
people die from disease. The level of the population shrinks, which gradually
causes output per capita to rise. Back at y ∗ , normal times resume.
Consider now a positive technological shock such as the use of the plow or the
introduction of windmills in the Middle Ages. Such a shock will appear as an
increase in A in (2.3) and as an outward shift of the y-curve in the lower panel
of Figure 2.2. This will temporarily cause output per capita to rise to y > y ∗ .
However, this new situation of prosperity will soon lead to a higher birth rate and
a lower death rate, which will cause the population to grow. When population
has grown to L ∗,new > L ∗ , income per capita is back at its old level y ∗ . The only
lasting result of a positive technology shock during the Malthusian era is thus a
larger population.
As was mentioned above, several developing economies in the world that are
dominated by subsistence agriculture are still caught in the Malthusian trap. Con-
sider an extended drought of several years, such as happened in the African Sahel
in 1985. In terms of Figure 2.2, such a shift would be like a negative shock to At
and would cause the y-curve in the lower bar to shift leftwards so that income fell
Birth rates, death rates
B(y )
D(y )
Y/L
L*,new
L*
y = Anew(X/L)α
y = A(X/L)α
Y/L = y
y* y′
Figure 2.2 Technological progress in the Malthusian model.
below its subsistence level. If such populations are left to themselves, famine will
set in and people will start dying of starvation. Starvation would, however, cause
income levels per capita to start rising again, according to the Malthusian logic.
Eventually, the economy would return to the old equilibrium level but this time
with a much smaller population.
In the modern era, most Western governments would find such mass starvations
unacceptable and provide emergency aid in the form of food and other relief. Such
an intervention is of course highly commendable from a moral point of view, but
it does not remove the Malthusian trap from the affected economy. If a downward
adjustment of population levels does not take place, the country will be caught
with a too large population that the country itself cannot feed. In this way, many
of the poorest countries become dependent on aid.
2.4 Endogenous fertility

In the model above, we simply assumed that the number of births Bt was positively
associated with the level of income yt−1 during the Malthusian era. Ashraf and
12 The long run
Galor (2010) demonstrate how this association can be derived explicitly from a
model of a utility-maximizing household that can choose between having children
and enjoying a high level of individual consumption.
Let us assume that total output in the economy is the same as in (2.1) and that
the (Cobb–Douglas) utility function of a representative individual at time t is
1−β β
Ut = ct nt (2.5)
where ct = Ct /L t is consumption per capita in the economy, n t ≥ 0 is the number

of surviving children that the individual gives birth to, and β ∈ (0, 1) is a param-
eter reflecting the relative importance of children for the individual’s utility. The
function makes two simplifying assumptions. First, we imagine that this single
(hermaphrodite) individual can choose how many children he or she wishes to
have. Second, we imagine that children can come in nondiscrete amounts (i.e. we
allow n t = 1.2 to be a possibility).
The budget constraint for the individual is
ct + ρn t = yt (2.6)
where ρ > 0 is the relative cost of raising children and yt is income per capita as
specified above. This constraint shows that the individual faces a trade-off between
own consumption ct and having children. If the person decides to have children,
the cost per child is fixed at ρ.
The utility function and the budget constraint together define a utility-
maximization problem:
max Ut subject to ct + ρn t = yt
nt
A straightforward way of solving this problem is simply to rewrite the budget

constraint as ct = yt − ρn t and then replace ct with this expression in the util-
ity function. If we further take logs of the utility function, the problem at hand
reduces to
max ln Ut = (1 − β) ln(yt − ρn t ) + β ln n t
nt
The first-order condition for maximum is2

∂ ln Ut (1 − β)ρ β
=− + =0
∂n t yt − ρn t nt
By rearranging this condition, we obtain
β(yt − ρn t ) = (1 − β)ρn =⇒ (1 − β)ρn t + βρn t = ρn t = βyt

βyt
=⇒ n ∗t =
ρ
Thus, the utility-maximizing (optimal) fertility level n ∗t
is a linear function of the
income level per capita yt . This result is a close analogy to the assumption above
of a function Bt (yt−1 ). The number of children decreases with the cost of having
children ρ. Note that n ∗t > 1 implies that the population is growing.
If we insert the derived level of n ∗t back into the budget constraint, we can solve
for ct∗ :
ct∗ + ρn ∗t = yt = ct∗ + βyt =⇒ ct∗ = (1 − β)yt
The growth of the total level of population between periods t and t + 1 is

given by
βyt L t β AL 1−α Xα
L t+1 = n ∗t L t = = t
= ϕ(A, L t ) (2.7)
ρ ρ
where we have substituted yt = A(X/L t )α from (2.3). The expression in (2.7)

implies that L t+1 is a positive and concave function ϕ of L t and a linear function
of the level of technology At .3 The concavity of L t+1 in L t arises since the first
derivative is ϕ L > 0 whereas the second is ϕ L L < 0. In Figure 2.3, we have drawn
(2.7) in (L t+1 , L t ) space.
The figure includes the function ϕ(A, L t ) as well as a 45◦ line where L t+1 = L t .
The crossing happens at L t = L t+1 = L ∗ . This is the equilibrium population size
in the model. To see why, consider for a moment a situation to the left of L ∗ where
L t < L ∗ . At this level, n ∗t > 1 and the population grows. Analogously, to the right
of L ∗ , n ∗t < 1 and the population shrinks. Only at L t = L t+1 = L ∗ will the economy
come to “rest” in equilibrium.
Lt+1
Lt+1 = Lt (45º)
Lt+1 = ϕ (A,Lt)
Lt
Lt*
Figure 2.3 Equilibrium population size.

14 The long run
If we insert L ∗ into (2.7), we can solve for the equilibrium level of population
density:
1
L∗ 1 βA α
= ∗=
X x ρ
The real-world equivalent of this expression could, for instance, be population

per square kilometer. Equilibrium population density during the Malthusian era
should thus increase with a country’s total productivity or level of technological
sophistication, just like above. Ashraf and Galor (2010) find strong support for
this hypothesis in their empirical study.
2.5 The collapse of the Malthusian link

By 1800, the United Kingdom was the first country in the world to break free of
the Malthusian trap during the industrial revolution. The United Kingdom was
soon followed by other Western countries, and later by yet more countries in
other parts of the world in a process that is still going on. This modern growth
regime was characterized by the following empirically observable features: (1) a
fall in birth rates B and an end to the link with income per capita y (death
rates D continued to fall with y); (2) an overall increase in levels of education;
(3) an increase in the growth of technological knowledge A; (4) as a direct conse-
quence of (1)–(3), a dramatic increase in output per capita far beyond subsistence
levels.
Although most economic historians agree on these stylized facts, their models
of how this change came about and why it was initiated in western Europe after
1800 rather than in, say, thirteenth-century India, differ to some extent. Here we
will only focus on some of the most prominent explanations.
2.5.1 Geography and institutions

One tradition emphasizes the central role of geography and technology. The basic
idea in these works is that geography, for various reasons, was particularly favor-
able in Europe for the steady advance of technological knowledge A. In highly
simplified terms, it is argued that although this growth of technological knowledge
in Europe was generally offset by accompanying increases in population levels in
the standard Malthusian manner, technological sophistication eventually reached
a critical threshold level when the rate of innovation became explosive.
One such technology-driven explanation for the early transition of Europe to
the agricultural and industrial revolutions is given by Olsson and Hibbs (2005).
In the spirit of the work of Jared Diamond (1997), they argue that the basic
process for technological progress during the prehistoric era was a simple func-
tion of the quality of the surrounding environment. In particular, technological
progress was faster in environments that were rich in potentially useful factors of
production, i.e. of suitable species for plant and animal domestication. As con-
vincingly shown by Diamond (1997), Western Eurasia (Europe and the Middle
East) was uniquely endowed with such species, which in turn explained the early
transition to sedentary agriculture, civilization and the rise of states.4
Olsson and Hibbs model the growth of technological knowledge during prehis-
toric times as
At+1 − At = At γ E i
where γ > 0 is a parameter invariant across regions and E i is the “wealth” of

the environment in region i in terms of useful plants and animals for domestica-
tion. Regions that were generously endowed thus had fast growth and eventually
reached a critical level Ā where the transition to Neolithic agriculture happened
early.5 The transition to agriculture then implied that a certain part of the popula-
tion was freed from producing and formed a new elite of specialists that would
be essential for the subsequent rapid advance of science and technology. The
Neolithic revolution thus created the first instance of endogenous technological
growth, i.e. the purposeful creation of new knowledge through the allocation of
labor resources to a new sector.
In Olsson and Hibbs (2005), the growth rate of technology changed shape to
become
At+1 − At
= g(at L t )
At
where at < 1 is the share of the total labor force L t at time t engaged in endogenous
knowledge creation, and g(at L t ) is a function of at L t such that g (at L t ) > 0. In
Diamond’s (1997) informal account and in Olsson and Hibbs’ model, the very
early creation of this knowledge-producing sector is the key to understanding why
Western Eurasia could start to dominate other continents from AD 1500 despite the
fact that Europe was not richer in per capita terms than any other part of the world.
According to this view, the industrial revolution was just a natural extension to a
development that had its roots in the transition to agriculture.
This theory has been criticized by a direction of research emphasizing the cen-
tral role of economic and political institutions, i.e. the fundamental rules that
societies live by. Acemoglu et al (2005) demonstrate that Western colonialism
played an important role in the accumulation of capital in western coastal Europe.
The inflow of capital strengthened the political power of a merchant class in coun-
tries like Britain and the Netherlands, which in turn led to reforms and stronger
institutional constraints against the executive vis-à-vis the citizens. In Spain, how-
ever, the inflow of capital only led to the enrichment of an elite around the
Crown and fostered rent-seeking behavior rather than production and investment.
The institutional changes in Britain and the Netherlands eventually triggered the
industrial revolution.
16 The long run
2.5.2 Galor and Weil’s unified growth model
The models above are not concerned with the more exact mechanism whereby
the link between population growth and income per capita is broken. Galor and
Weil (2000), however, focus explicitly on the dynamics of this process. One of the
main assumptions in their model is that parents face a basic trade-off between child
quantity (number of children) and child quality (children’s level of education).6 In
this section we present an extremely simplified version of Galor and Weil’s model.
Parents divide their time between work and childrearing. Time spent on one
child in period t is in turn divided between basic childrearing τ and education so
that the child’s level of education in the next period is et+1 . Let the family’s total
number of children in period t be denoted by n t and let the total time available to
parents be normalized to unity. Hence, the time budget constraint for parents is
n t (τ + et+1 ) ≤ 1
The time budget constraint shows that parents face a trade-off between having
many children (a high n t ) and giving them a good level of future education et+1 .
This is similar to what Becker and Lewis (1973) refer to as the child quantity/child
quality trade-off. If, for instance, n t (τ + et+1 ) = 1 applies, then in order to increase
et+1 we must necessarily decrease n t in a proportional manner.7 In the full version
of the model, Galor and Weil (2000) derive the optimal allocation of time spent
on the two activities and show that it will to a great extent depend on the parents’
preference structure. It is well known, however, that during most of human his-
tory, the trade-off was such that et+1 = 0. A central element of the breaking of
the Malthusian trap was precisely that levels of education started rising and the
number of children born in each family started falling.
So why did families start to substitute child quantity for child quality? In the
model, Galor and Weil derive the result that the optimal level of et+1 is a function
of the growth rate of technological knowledge such that et = e(gt ). The logic is
simply that the greater the advancement of technological knowledge in a society,
the higher will the optimal level of education be. The derivatives are ∂et /∂gt =
eg ≥ 0 and ∂ 2 et /∂gt2 = egg ≤ 0.
Let us also assume that the growth rate of technological knowledge is given by
At+1 − At
= gt+1 = g(et , L t )
At
In this expression, the growth rate g is a positive function of et and L t . A higher

level of education in the population et is associated with faster technological
progress, and, likewise, a larger population means more potential innovators. In
formal terms, the partial derivatives are assumed to be ∂g/∂et = ge > 0, ∂g/∂ L t =
g L > 0, and ∂ 2 g/∂et2 = gee < 0. Once parents start investing in education for their
children, there might thus be a kind of positive feedback loop between education
and technological progress.
(a) gt
et+1 = e (gt+1)
g′ gt+1 = g (et,L0 )
g″
g ′′′
et
0 = e* = g* e″ e′
(b) gt
et+1 = e (gt+1)
g* gt+1 = g (et,L1)
g′
gt+1 = g (et,L0 )
et
0 e′ e*
Figure 2.4 (a) A Malthusian trap in the evolution of technology and education in the
Galor–Weil model. (b) Escape from the Malthusian trap in the Galor–Weil
model due to an increase in population (L 1 > L 0 ).
The basic relationship between the two variables is shown in Figures 2.4a
and 2.4b. In these figures, we have drawn both the et+1 = e(gt+1) and the gt+1 =
g(et , L t ) curves simultaneously. Their slopes are as derived above.8 Note that
below a certain level of technological progress gt , the optimal level of education et
is zero.
Now consider an economy that starts off at a level of education e > 0. At
this level, next year’s technological progress is fairly high at g > 0. The level
of population is L 0 > 0. However, at this level of g, education in the next period
is discouraged and will fall to e < e , as indicated by the arrow. At this lower
18 The long run
level of education, technological progress will slow down and g will fall further to
g < g . Hence, in the economy at hand, there is a negative feedback loop between
education and technology that eventually causes the economy to reach an equilib-
rium where e∗ = g ∗ = 0. Since no time is spent on education, parents will spend
all their nonworking time on having babies and n t will be large. This is the typical
Malthusian trap scenario.
Consider now another economy (or the same economy at a different date) as in
Figure 2.4b where population is L 1 > L 0 so that the g(et , L 0 ) curve has shifted
upward to g(et , L 0 ). If we start off at e and g , education will be strongly encour-
aged by the high level of g, so that e will increase in the next period. The higher
level of education in turn spurs faster technological progress, and so the economy
enters a positive virtuous circle that settles down at equilibrium levels e∗ , g ∗ > 0.
Since parents’ time budget constraint has not expanded, they will substitute child
quantity n t for child quality et . As total production in society starts rising fast
due to technological progress, the combined effect of a lower n t and a higher gt
makes income per capita rise explosively. The economy has escaped from the
Malthusian trap.
3 The Solow Growth Model
3.1 Introduction
In this third chapter, we will analyze the determinants of wealth and poverty during
the industrial era among nations most of which have managed to escape from
the Malthusian trap. Rather than focusing on short-term phenomena like business
cycles, we will try to understand long-run patterns of development, for example
why some of the richest countries such as Switzerland and Norway have a level of
GDP that is about one hundred times greater than the GDP of some of the poorest
countries such as Niger and Haiti.1
Population growth will play no important role in this framework since it is
assumed to have stabilized at relatively low (exogenous) levels. The key factor
of production is instead physical capital and the key process is that of conver-
gence, showing how economies that initially start off with relatively low levels of
physical capital per worker should grow faster than relatively richer countries.
The neoclassical growth model builds fundamentally on the work of Robert
Solow (1956) and has become one of the most important models in macroeco-
nomic research. We start by deriving the well-known k̇-equation and the most
important implications that follow from it such as convergence. In the final
sections, we present some extensions to the Solow model where we demon-
strate how technological progress and human capital can be included in the basic
framework.
3.2 Basic assumptions

The main contribution of the Solow growth model from a theoretical point of
view is that it makes the accumulation of physical capital endogenous, as will
soon become clear.2
The simplest version of the Solow growth model assumes an aggregate produc-
tion function F of the following form:
Yt = F(K (t), L(t)) (3.1)
where Yt is total GDP measured from the production side, K (t) is total physical
capital that is a function of time t, and L(t) is the aggregate labor force.3 K (t)
20 The long run
might be thought of as the total stock of factories and machines in a country and
L(t) as the total number of workers. K and L (we will henceforth usually drop
the time (t) notation for ease of exposition) are thus production factors or inputs
in the aggregate production process.
We also make the following more technical assumptions:
• Constant returns to scale: F(λK , λ(L)) = λF(K , L).

∂F
• All factors of production have positive but diminishing marginal returns: ∂K =
∂2 F ∂2 F
FK > 0, = FK K < 0 at all levels of K ; ∂∂ FL = FL > 0, < 0 at all levels
∂K2 ∂ L2
of L.
The constant-returns assumption means that if we, for instance, double the levels
of K and L simultaneously, we will get a doubling of total output.4 The second
assumption implies that F has a concave relationship with both production factors
and that the marginal product is always positive. This is the same “Ricardian”
assumption of diminishing returns as was used in the Malthusian model.
It is further assumed that we can transform the aggregate production function
in the following way:

F(K , L) K L K K
=F , =F , 1 = f (k), where k = (3.2)
L L L L L
This transformation is referred to as the intensive form of (3.1) and k is for-

mally referred to as capital per worker. As we shall see, using this intensive-form
expression simplifies the algebra significantly in the sections ahead.
The intensive form of the production function has the same basic properties as
above:
• f (0) = 0; f (k) > 0; f (k) < 0 at all levels of k > 0.
The most often used functional form for the production function in growth
theory is Cobb–Douglas:
Y = F(K , L) = K α L 1−α (3.3)
Dividing through by L gives us the intensive form, or output per worker y:

α
K α L 1−α K
y = f (k) = = K α L −α = = kα (3.4)
L L
3.3 Dynamics
All variables in the model are functions of time, so the next step is to specify their
dynamics or laws of motion. The growth of labor L is in this setting assumed to
The Solow growth model 21
be exogenously given, i.e. not explained by the model:
∂ L(t) L̇
= L̇(t) = n L, where n > 0, implying =n
∂t L
In this expression, n is the (percentage) growth rate of the labor force (or of
population size). Although time derivatives indicate instantaneous changes in the
stocks, we normally think of L̇L , for instance, as the growth rate during a year, as
in the national accounts.5 A typical level for n would thus be 0.01–0.05.
The key dynamic equation in the Solow model is that specifying the rate of
change in the physical capital stock:
K̇ = sY − δ K (3.5)
In this expression, s > 0 is the fraction of total output Y that is being saved and
δ > 0 is the capital depreciation rate, i.e. the fraction of total physical capital that is
worn down every year. (Typical and often observed levels are s = 0.2 and δ = 0.05.)
The same expression can be restated as sY = K̇ + δ K = I , which tells us that total
savings sY can be used for net investment K̇ (leading to actual increases in the
capital stock) and replacement investments δ K (replacing the capital that has been
worn down), and which together sum to total aggregate investment I .
The Solow growth model implicitly assumes a closed economy without trade
and where there is no government. Hence, the user side of the economy contains
only investment and consumption (compare with the fundamental equation (1.1)
above). Therefore, we can write
Y = K̇ + δ K + C = I + C
where C is the aggregate level of consumption.

We want to express K̇ in intensive form, i.e. we want to find k̇. Recall that
k(t) = K (t)/L(t). Hence, by using the chain and quotient rules of differentiation,
we can write
∂k(t) K̇ K
= k̇ = − 2 L̇
∂t L L
K̇ L sY (t) − δ K (t)
= −k = − nk (3.6)
L L L
= s f (k) − (δ + n)k
The expression in the third line of (3.6) is the central equation of the Solow growth
model.
3.4 Equilibrium
The k̇-equation can also be drawn as in Figure 3.1. The vertical axis simply shows
levels whereas the horizontal axis shows capital per unit of effective labor, k.
22 The long run
f (k)
(δ + n)k
(1 – s )f(k) = c sf(k )
k
k*
Figure 3.1 The neoclassical growth diagram.
The two most important lines here are the s f (k) and (δ + n)k curves, where it
is important to note that the former is concave since f (k) < 0. s f (k) is some-
times referred to as the actual level of investment and (δ + n)k as the break-even
level of investment. Beyond these curves, however, there is also the f (k) curve.
Note that the vertical distance between the s f (k) and f (k) curves equals c = C/L,
i.e. consumption per unit of labor.
At low levels of k, k̇ > 0, whereas at high levels of k, k̇ < 0. The only stable
equilibrium in Figure 3.1 occurs at k ∗ , which is defined by the point where the
s f (k) and (δ + n)k curves cross. This is also the level of k where k̇ = 0, implying
that K and L grow at a “balanced” rate.6 k ∗ is often also referred to as the steady-
state equilibrium.
3.5 Implications
3.5.1 Cobb–Douglas functional form
If we assume a Cobb–Douglas functional form for the production function as
in (3.4), we will have the following k̇-equation:
k̇ = sk α − (δ + n)k
From this expression, we can also derive the growth rate per worker (by using the
chain rule):
∂(k α )
ẏ αk α−1 k̇ α k̇
= ∂tα = = (3.7)
y k kα k
sα
= sαk α−1 − α(δ + n) = − α(δ + n)
k 1−α
s′
s
time
t0
time
t0
Figure 3.2 Impact on growth rate per worker of an increase in the savings rate.
In the short run, the growth rate will depend on the initial level of k since
economies with a low level of k will have k̇ > 0 and will grow faster. However, as
k increases, the economy will gradually approach its steady-state level k ∗ where
k̇ = 0. Furthermore, an increase in the savings rate s that happens in a steady state
will temporarily increase the growth rate since it will turn k̇ positive. The long-run
effect, however, should be zero when k has moved to a new (and higher) equilib-
rium level (see Figure 3.2). Similarly, a sudden increase in population growth will
lead to a period of negative growth, until the economy settles at a lower equilib-
rium level of k and an associated lower level of output per worker. This prediction
of temporary effects on the growth rate but permanent effects on the steady-state
levels of output is among the most central predictions from the model and has
been tested numerous times in the empirical literature.
We can also solve for the steady-state level of k:
k̇ = 0 =⇒ s(k ∗ )α = (δ + n)k ∗ (3.8)
1
(δ + n) δ+n α−1
=⇒ (k ∗ )α−1 = =⇒ k ∗ =
s s
1
∗ s 1−α
=k =
δ+n
Note that we want to carry out the last step in order to have the exponent positive
1
( α−1 < 0). This expression for the steady-state level clearly shows that k ∗ will
increase with the savings rate s, and decrease with the capital depreciation rate δ
24 The long run
and with the population growth rate n. Identical results are found by moving the
curves in Figure 3.1. The steady-state level of output per worker is thus
α
∗ ∗ α s 1−α
y = (k ) =
δ+n
3.5.2 Golden rule of capital accumulation

What is the impact on the steady-state level of consumption per unit of effective
labor c∗ from an increase in the savings rate s? Note that
c∗ = (1 − s) f (k ∗ ) = f (k ∗ ) − (n + δ)k ∗ (3.9)
We know from (3.8) that k ∗ increases with s, but in the expression (3.9) there
is both a positive and a negative effect of k ∗ on c∗ . When we take the partial
derivative, we get
∂c∗ ∂k ∗
= [ f (k ∗ ) − (n + δ)]
∂s ∂s
The sign of this expression will be determined by the sign of the term in square
∗
brackets (know that ∂k ∗
∂s > 0 always holds). Since f (k ) is very large at small
∗
levels of k (see Figure 3.1), we can infer that ∂c
∂s > 0 when k is small, whereas at
∂c∗
greater levels of k it will be the case that ∂s < 0. Hence, there is a level of k ∗ ,
∗
referred to as k ∗,gold , when ∂c
∂s = 0:
∂c∗
= 0 when f (k ∗,gold ) = n + δ (3.10)
∂s
Beyond this level of k, a further capital accumulation will decrease intensive con-
sumption c. The level k ∗,gold is referred to as s gold . The intuition about this golden
rule of capital accumulation is that it is only useful to increase the savings rate up
to a certain level.
3.5.3 Convergence
The model above has some strong predictions about convergence, i.e. that coun-
tries that start off with a lower level of k should experience a higher growth rate
of output per worker. Another way of illustrating the convergence property is
described below.7
If we insert a Cobb–Douglas production function f (k) = k α into the model
above, it was shown that we can write the growth rate of output per worker as:
ẏ
= α(sk α−1 − δ − n)
y
1
Since YL = k α = y, we can express k as k =yα. Inserting into the growth equation
above yields

ẏ α−1 s
= α(sy α − δ − n) = α 1−α − δ − n (3.11)
y y α
Variants of this expression form the basis of the cross-country empirical studies on
the determinants of economic growth. The key prediction concerning convergence
is that the growth rate of output per worker should decrease with its initial level, y.
In other words, holding all other factors constant, poorer countries should grow
faster than richer ones. A country’s growth rate during the convergence process
should further increase with its savings rate s, and decrease with the popula-
tion growth rate n and with the capital depreciation rate δ. The growth rate of
rich countries that have reached their steady state will depend on the exogenous
parameter g.
The convergence result suggests that the poorest countries in the world should
experience the highest growth rates. Although we know that many previously poor
countries have experienced very fast growth rates in recent decades – for instance
China, India, and Botswana – other countries have experienced stagnant growth or
even growth collapses. Some countries, such as DR Congo and Zambia, have even
seen their levels of income per capita fall by half. We will return to this issue below.
3.6 Extensions
The simple version of the Solow growth function presented above provides the
basic intuition behind the important convergence property and the central role
played by physical capital accumulation. It abstracts, however, from numerous
factors that are believed to be central for economic growth to occur even in the
medium run. Two of the most important of these factors are technological progress
and human capital accumulation.
3.6.1 Technological progress

Technological progress can be readily included in the basic model above. As in
Chapter 2, let us refer to the level of technological knowledge at time t as At . Most
growth models further assume that technology is primarily labor augmenting, i.e.
that At increases workers’ level of productivity. We will refer to the composite
production factor At L t as effective labor.8 The aggregate production function then
becomes
Y = F(K , AL) = K α (AL)1−α
Let us further assume that the growth rate of technology is exogenously given by
Ȧt
=g>0
At
26 The long run
The intensive form is now written as κ = K /AL and is referred to as capital per
unit of effective labor. The time derivative of κ is
K̇ KL KA
κ̇ = − Ȧ − L̇
AL (AL)2 (AL)2
sY − δ K Ȧ L̇
= − κ − κ = s fˆ(κ) − κ(δ + g + n)
AL A L
where fˆ(κ) = F(K , AL)/AL. As in the section above, the economy will be in a
s 1
steady-state equilibrium when κ̇ = 0, which happens at a level κ ∗ = δ+g+n 1−α .
The equilibrium level of capital per unit of effective labor thus decreases with the
technological growth rate.

Since output per capita can now be expressed as Y/L = K α (AL)1−α /L = Aκ α ,
its growth rate is given by
ˆ
ẏ Ȧ κ̇ s f (κ)
= +α =g+α −δ−g−n
y A κ κ
Note that when κ = κ ∗ , the term inside the parentheses will be equal to zero. The
equilibrium growth rate of output per capita is then simply g > 0. Hence, rich
countries at their equilibrium growth rates will only grow through technological
progress. In the Solow model, this growth rate is exogenously given and we cannot
say anything very interesting about it. The main aim of endogenous growth theory,
which is the topic of the next chapter, is to derive this growth rate as the result of
intentional human investments in research and development (R&D).
3.6.2 Human capital

The level of technological knowledge At is meant to capture the stock of
production-relevant ideas that (in the absence of intellectual property rights) are
free for anyone to use. In the economics terminology, we refer to ideas as nonrival
goods that have the same basic character as most public goods. One idea can be
used in thousands of places and situations at the same time.9
Human capital, on the other hand, is the skills and competencies embodied in
people. The skills of a certain person cannot be used in two offices at the same
time. Human capital is typically enhanced through education or learning and often
depreciates with time due to obsolescence or because people tend to forget. In this
sense, human capital has a lot in common with physical capital that also increases
through investment and decreases through depreciation.
Mankiw et al (1992) extend the basic Solow model to also include human
capital. In their model, the aggregate stock of human capital in an economy H
is thought of as distinct from effective labor AL. They imagine an aggregate
production function of the form
Y = F(K , H, AL) = K α H β (AL)1−α−β

In their model, both physical capital and human capital grow endogenously. Let
κ = K /AL, η = H /AL, and ỹ = K α H β (AL)1−α−β /AL = κ α ηβ . The dynamics
of the two stocks are given by
κ̇ = sκ κ α ηβ − κ(δ + g + n)
η̇ = sη κ α ηβ − η(δ + g + n)
The parameters sκ > 0 and sη > 0 reflect the share of total output per unit of
effective labor ỹ = κ α ηβ that is invested in physical capital and human capital,
respectively. A share 1 − sκ − sη of total income is devoted to consumption. sη
might thus be seen as a country’s rate of investment in education. As discussed in
Chapter 2, this rate was very low in most countries until about 1800. For simplicity,
it is assumed that both stocks have the same depreciation rate δ + g + n.
The steady-state levels κ ∗ and η∗ are found where κ̇ = η̇ = 0. Reaching the
solution involves more algebra than before since we first have to solve for one of
the two steady-state levels and then insert this solution into the expression for the
other. In the first step, we find from κ̇ = 0 that
1
∗ sκ (η∗ )β 1−α
κ =
δ+g+n
and from η̇ = 0 that

1
sη (κ ∗ )α 1−β
η∗ =
δ+g+n
If we then take logs of the two expressions above, we get

∗ 1 sκ β
ln κ = ln + ln η∗ (3.12)
1−α δ+g+n 1−α

1 sη α
ln η∗ = ln + ln κ ∗ (3.13)
1−β δ+g+n 1−β
Inserting the expression for ln η∗ in (3.13) into (3.12), we obtain

1 sκ β sη
ln κ ∗ = ln + ln
1−α δ+g+n (1 − α)(1 − β) δ+g+n
αβ
+ ln κ ∗
(1 − α)(1 − β)
Then manipulating the expression so that ln κ ∗ is isolated on the left-hand side,

∗ 1−β sκ β sη
ln κ = ln + ln
1−α−β δ+g+n 1−α−β δ+g+n
28 The long run
By taking the anti-log of this expression, we can express the closed-form solution
as
β 1−β 1
∗ sη sκ 1−α−β
κ =
δ+g+n
By using the same methodology, we can also solve for η∗ :
1
∗
sκα sη1−α 1−α−β
η =
δ+g+n
A noteworthy feature of these expressions is that κ ∗ increases with sη and

that η∗ increases with sκ . This effect arises since a higher sη increases the level of
η and the level of available resources ỹ, which means that there is a higher total
income to be invested in physical capital.
The steady-state level of output per capita in this economy is
β 1−β α β
∗ ∗ α ∗ β sη sκ 1−α−β sκα sη1−α 1−α−β
y = At (κ ) (η ) = At
δ+g+n δ+g+n
β 1
sηα sκ 1−α−β
= At
(δ + g + n)α+β
Since At is the only factor of production that does not converge to a steady-state
level, the equilibrium growth rate of output per capita will be equal to g, as above.
4 Endogenous Growth Theory
4.1 Introduction
The Solow growth model has been criticized on two grounds: Firstly, the pre-
diction that all countries should converge to the same long run growth rate
does not appear to have materialized. Secondly, the engine of long-run growth,
technological progress g, is left unexplained by the model.
In this chapter, we will take a closer look at endogenous growth theory, which
attempts to explain how technological progress emerges as a consequence of more
or less active choices by the agents in the economy. In particular, we will ana-
lyze how a separate R&D sector interacts with sectors producing intermediate and
final goods. The presence of intellectual property rights means that intermediate
goods-producing firms might in practice obtain a monopoly on the good they are
producing. The later sections analyze the implications of incorporating assump-
tions of imperfect competition. The main contributors to this literature are Paul
Romer (1986, 1990) and Aghion and Howitt (1992).
Due to the focus on technological progress, the current chapter is mainly rele-
vant for relatively advanced developed economies where R&D plays a key role.
Most countries in the world do not carry out any advanced R&D and only imi-
tate the innovations carried out in Western countries. Section 4.6 contains a model
where a social planner in a country can choose between imitation and innovation.
4.2 AK model
In a famous article, Romer (1986) argues that the tendency towards diminishing
returns, which is a central feature of the Solow growth model, is perhaps exagger-
ated. According to Romer, it also appears as if certain types of policy have a more
lasting impact on growth rates than in the Solow model. The simple solution to
these aspects was offered by Rebelo’s (1991) AK model.
Let us assume that the correct specification of the long-run production func-
tion is
Y (t) = AK (t) (4.1)
where K (t) is now interpreted as a broad aggregate of physical and human capital
and A is a fixed technology parameter. As before, the model is driven by a capital
30 The long run
accumulation function
K̇ = s AK − δ K (4.2)
where s is the savings rate and δ is the rate of physical capital depreciation. Capital
per worker is k(t) = K (t)/L(t) and the growth rate of the work force equals the
growth rate of the population, so that L̇/L = n > 0. The dynamics for capital per
worker is then
∂k K̇ K L̇
= − = s Ak − δk − nk (4.3)
∂t L L L
Output per worker is y = AK /L = Ak and its growth rate is
ẏ k̇
= =sA−δ−n (4.4)
y k
The key thing to notice in this expression is that any change in policy-related
parameters like s and n will now have permanent effects on the growth rate. For
instance, an increase in savings rate s will lead to a permanently higher growth
rate. This result stands in contrast to predictions from the Solow model (see (3.7),
where an increase in s only had a transitory effect on growth rates). A drawback
is of course that the factors that drive long-run growth are still exogenous to the
model.
4.3 Endogenous technological change

So far, we have basically neglected the issue of how technological progress comes
about. In the Solow framework above, we simply assumed an exogenous growth
rate g. It is well known that most of the progress in technological knowledge
during the last century has been the result of intentional actions by scientists
and engineers in large R&D departments within profit-maximizing firms. The
famous models by Romer (1990) and Aghion and Howitt (1992) provided pioneer-
ing attempts to illustrate the mechanisms in this process. The subsection below
presents a simplified version of those models.
4.3.1 An R&D sector

A key feature of the later endogenous growth models is the modeling of a separate
R&D sector to which factors of production (mainly labor) are devoted on purpose.
Let us assume that there are two sectors in the economy: a final goods sector,
producing all the normal final goods and services that make up a country’s GDP,
and an R&D sector, producing new ideas or new technological knowledge that
can be used in the final goods sector. The production function for the final goods
sector is
Y (t) = A(t)(1 − a)L(t) (4.5)

Endogenous growth theory 31
where A(t) is the level of technological knowledge at time t as before and a is the
fraction of the total labor force L(t) that is active in the R&D sector. A fraction
1 − a of total labor is thus left for the final goods sector.
The R&D sector has the production function
Ȧ(t) = B(a L(t))γ A(t)θ (4.6)
where B > 0 is a parameter indicating the efficiency of research, γ ∈ (0, 1) shows

the output elasticity of labor in research, and θ ≤ 1 is a parameter describing the
elasticity of existing knowledge A(t) for the production of increases in the stock
of knowledge, Ȧ. For simplicity, physical capital is now left out the model.
Two things are particularly noteworthy about equations (4.5) and (4.6). Firstly,
A is available to the full amount in both sectors. The reason for this assumption
is that ideas are inherently nonrival in character, i.e. one person’s use of an idea
(for instance, the blueprint for a new engine) does not preclude another person’s
simultaneous use of the same idea. Unlike ideas, most factors of production like
capital and labor are rival, for instance in the sense that an actual engine can either
be used in factory 1 or in factory 2, not in both at the same time. Labor is also a
rival production factor since a fraction a is employed in R&D and 1 − a in the
final goods sector.
Secondly, as we will see, the nature of the dynamics in this model will hinge on
the level of the output elasticity of existing knowledge, θ . This parameter should
be thought of as describing how useful older knowledge is for the creation of new
knowledge. θ < 1 implies that there is a kind of “fishing-out” effect in the sense
that there are diminishing returns to the existing stock of technological knowledge
in finding new ideas. We will return to this issue below.
A final assumption in the model is that labor grows at a rate n > 0 as before.
4.3.2 Steady-state equilibrium

From (4.6), we know that the growth rate of technological knowledge g A (t) will
be equal to
Ȧ Ba γ L(t)γ
= g A (t) = Ba γ L γ Aθ−1 = (4.7)
A A(t)1−θ
As in the Solow model, a balanced, steady-state growth rate will exist when g A (t)
is constant over time, i.e. when ∂g∂tA (t) = ġ(t) = 0. From inspection of (4.7), we
know that this will only happen if L(t)γ grows at the same rate as A(t)1−θ . One
way of finding this level is to start by assuming that ġ A = 0. The time derivative
of the growth rate in (4.7) is most easily found by first taking logs,
ln g A = ln B + γ ln(a L) + (θ − 1) ln A
32 The long run
and then taking the time derivative of the log:1
∂ ln ga ġ A L̇ Ȧ
= = γ + (θ − 1) = γ n + (θ − 1)g A
∂t gA L A
Multiplying this expression by g A , and inserting the requirement that ġ A = 0 at

the steady-state growth rate g ∗A , gives us
ġ A = [γ n + (θ − 1)g ∗A ]g ∗A = 0 (4.8)
It is clear that this expression can only be zero if g ∗A = 0 or if γ n + (θ − 1)g ∗A = 0.

What type of equilibrium we will have will thus crucially depend on the level of θ .
In the case where θ = 1, so that there are no diminishing returns to existing
knowledge, as assumed in the original models of Romer (1990) and Aghion and
Howitt (1992), the expression in the square brackets in (4.8) will not be zero and
the only existing steady state in the current setting is g ∗A = 0. Note also that the
(positive) growth rate of technological knowledge in this case will be g A (t) =
B(a L)γ . An increase in the R&D labor force a L, caused by increases in either the
share a or the size of the total labor force L, will thus increase the growth rate of
technology. There will further never be a stagnation in growth rates.
In the case where θ < 1, on the other hand, (4.8) will be satisfied if
γ n + (θ − 1)g ∗A = 0 =⇒ γ n = (1 − θ )g ∗A
γn
=⇒ g ∗A =
1−θ
A steady-state growth rate thus exists and will not depend on the level of the R&D
labor force. This “semi-endogenous” growth result, due to Jones (1995), is thus
free of the scale effect of R&D labor inherent in Romer (1990) and Aghion and
Howitt (1992). Instead, only the growth rate of the population n will matter in
the long run. Since output per capita equals Y/L = A(1 − a), g ∗A will also be the
growth rate of GDP per capita.
4.3.3 Allocation of labor

So far, we have said nothing about a, the proportion of labor employed in each
sector. How should labor be allocated in equilibrium? As usual, if we assume com-
petitive labor markets and freely mobile labor, workers should reallocate between
sectors until they earn the same wage.
More formally, let us think of the final goods sector as being run by a single
representative firm selling final goods at a price equal to unity and with profits
given by
F = A(1 − a)L − w(1 − a)L

The typical R&D firm, in turn, also only uses labor and has a profit function of the
form
R = P A B(a L)γ Aθ − wa L
where P A is the price of the innovations that the R&D firm produces. For now,
we will take that price as exogenously given. The first-order conditions for profit
maximization give us
∂ F
= −AL + wL = 0
∂a
∂ R
= γ P A Ba γ −1 L γ Aθ − wL = 0
∂a
In equilibrium, the wage costs of the two firms should thus be the same. The
conditions above imply that
γ P A Ba γ −1 L γ Aθ = AL
Using this term, we can solve for the optimal sectoral allocation a ∗ :
∗ γ −1 A1−θ L 1−γ
a = =⇒
γ PAB
1−θ 1−γ γ −1
1 1
A L γ P A B 1−γ 1
a∗ = =
γ PAB A1−θ L
Given that γ < 1, the allocation of labor to R&D should thus increase with the
price of innovations P A and with R&D productivity B and decrease with L. The
reason for the last observation is that workers face diminishing returns in the R&D
sector, whereas we assume here constant returns in the final goods sector. Since
we assume that L grows with an assumed rate of n > 0, this should imply that the
share of workers in R&D should fall with time. Similarly, if θ < 1 so that there
are diminishing returns to the existing stock of technological knowledge, then a ∗
decreases with A. Note, however, that if θ = 1 as in the steady-state result above,
then a ∗ will be independent of the level of knowledge.
One of the key points with the result above is that a ∗ will be a positive function
of the price of innovations P A . But how is that price determined if ideas are a
nonrival good? Fundamentally, the price of patents will depend on the institutions
in society for intellectual property rights. If such property rights are weak, the
price will be low and there will be weak incentives for people and firms to engage
in R&D. On the positive side, without property rights to ideas, there would be
no monopoly power for users of a particular innovation and anyone could com-
mercialize any existing technological idea. According to the model, however, as
34 The long run
P A goes to zero, a ∗ will also tend towards zero, so that no new ideas would be
produced.
In the next section, we will extend the model and show how the price of inno-
vations is determined in a market with monopolistic intermediate goods producers
who buy patents on innovations from an R&D sector.
4.4 Romer’s product variety model

In this section, we will recapitulate some of the details of the model in Romer
(1990). The aim is to get a better understanding of what determines the price of
innovations P A and how the particular market structure with monopolistic inter-
mediate goods firms affects the model. We will make the simplifying assumption
that labor is only used in the final goods sector.2
4.4.1 Three sectors

The model assumes that there are three sectors:
• A final goods sector that employs all L workers as well as intermediate capital
goods as factors of production.
• An intermediate goods sector using no labor but producing intermediate goods
that are used as factors of production (physical capital) in the final goods sector.
The patents for the intermediate goods are bought from the R&D sector and
each intermediate goods producer is a monopolist on the good that they have
acquired a patent on.
• An R&D sector that produces patents for intermediate goods and sells them to
the intermediate goods producers.
Final goods (equivalent to total measured output in the economy) are produced
by a representative firm according to the function

A
Y = L 1−α (X j )α = L 1−α (X 1α + X 2α + · · · + X αA ) (4.9)
j =1
where X j is the amount of an intermediate capital good of vintage j (one might,

for instance, think of X j as the quantity of machine j ). Let us for simplicity
assume that L is now a constant. There are A > 0 intermediate goods in the econ-
omy. The number of intermediate goods is equivalent to the level of technology in
this model. A noteworthy feature of the production function (4.9) is that marginal
products of the intermediate goods are all given by
∂Y
= αL 1−α X α−1 for all j (4.10)
∂Xj j
The marginal product of good j is thus independent of the levels of all other
intermediate goods. One might think of the intermediate goods in the model as
major inventions that are neither substitutes for nor complements to each other.
The price of final goods is assumed to be 1.
The representative final goods producer maximizes profits:

A
A
max F = L 1−α (X j )α − Pj X j
X
j =1 j =1
where P j is the price of intermediate good j . The usual first-order conditions yield
∂ F
= αL 1−α X α−1 − Pj = 0 (4.11)
∂Xj j
Condition (4.11) states the familiar result that firms should acquire capital good j
up to the level where its value marginal product equals its marginal cost, which
is P j . Equation (4.11) can be rewritten so as to obtain an expression for the final
goods sector’s demand for good j :
1
α 1−α
Xj = L (4.12)
Pj
Demand increases linearly with the work force but decreases, as expected, with
the price charged by the intermediate firm.
Each intermediate good is produced by a single intermediate goods producer
who has a patent on the good in question for a period of T years and hence a
monopoly on its production. No labor is used in this sector. The marginal cost
of producing one unit of good j is simply assumed to be equal to 1. The profit
function is therefore given by
Ij = (P j − 1)X j (P j ) (4.13)
This expression contains the usual trade-off for a price-setting monopolist:

revenues are increased directly by an increase in P j but demand falls since

X j (P j ) < 0. As will be apparent later, it is convenient at this stage to calculate the
price elasticity of demand, i.e. the increase in demand from a marginal increase in
price, multiplied by the ratio of the price to the level of demand:
1
Pj α 1−α L Pj
X j (P j ) =−
X j (P j ) 1 +1 X j (P j )
(1 − α)P j1−α
X j (P j ) Pj 1
=− =−
(1 − α)P j X j (P j ) 1−α
36 The long run
The price elasticity of demand is thus negative and constant regardless of the level
of demand.
The profit-maximizing price is obtained by taking the first-order condition for
maximum:
∂ Ij
= X j (P j ) + P j X j (P j ) − X j (P j ) = 0
∂ Pj
This is a typical first-order condition for a monopolistic firm. We can rearrange

this condition as X j (P j ) = X j (P j )(1 − P j ) and then multiply both sides by
P j / X j (P j ) to obtain
Pj 1
P j = X j (P j ) (1 − P j ) = − (1 − P j )
X j (P j ) 1−α
Note that we have inserted the derived expression for the price elasticity of demand
on the right-hand side. By rearranging the last equation so as to isolate P j , we
finally obtain the profit-maximizing price for the monopolist and the correspond-
ing level of demand from the final goods sector and profits for the intermediate
goods sector:
1
P j∗ =
α
Since α < 1, we can infer that P j∗ > 1, i.e. the monopolistic intermediate goods
producer will charge a price higher than the marginal cost, which is equal to 1. In
microeconomics textbooks, this phenomenon is known as mark-up pricing.3
When we have solved for P j∗ , the corresponding equilibrium levels of final
sector demand and intermediate sector profits can be derived to be
2 1+α
X j (P j∗ ) = α 1−α L, Ij = (P j∗ − 1)X j (P j∗ ) = (1 − α)α 1−α L (4.14)
Note that the solution in (4.14) implies that X j (P j∗ ) = X j +1 (P j∗+1 ) = · · · =

X A (PA∗ ). Hence, we can express the equilibrium level of total output of final goods
in (4.9) without a summation sign and instead multiply the solution for X j (P j∗ )
by A to obtain
2 2
Y ∗ = L 1−α Aα 1−α L = Aα 1−α L 2−α (4.15)
In accordance with intuition, total output of final goods therefore increases with
the number of intermediate goods A, which in this model is an indicator of the
state of technological knowledge.
Finally, then, we can describe the R&D sector that produces patents and sells
them to intermediate goods producers. The patentable invention of one additional
design for a new intermediate good is made using a production function BA(t),
where B is a productivity parameter as before and A is the existing stock of patents
at time t. The output is certain and always results in a new design that can be
patented.4 As before, we assume that knowledge about previous patents is fully
accessible to the R&D firm. A(t) is thus a nonrival production factor since it is
available to the full extent both in the final goods sector and in the R&D sector,
although it is partially excludable since only one intermediate goods producer has
a patent and is allowed to produce the patented good commercially. There is free
entry into the sector.
If the R&D firm chooses to make the invention, this will entail a fixed cost η > 0.
As we shall see, this fixed cost will be very important for the determination of the
price P jA that the R&D firm charges for the invention.
4.4.2 Solving the model in two stages

In order to find the optimal price P jA , we must first note that it is not obvious that
the R&D firm will choose to invent a new design at all. A necessary condition is
that P jA ≥ η. If not, then the R&D firm is better off if it does not produce at all.
P jA will of course also depend on the intermediate goods sector’s willingness to
pay.
The problem can be solved by thinking of it as a two-stage procedure:
• First stage: The R&D firm decides on whether to invent a new good or not.
• Second stage: Once the new good has been invented, the R&D firm decides on
what price P jA to charge the intermediate goods producing firm j for the patent.
The intermediate firm produces the good and it is sold to the final goods sector,
where it is used in the production of final goods.
As is standard in this type of two-stage analysis, we use backward induction

and start at the second stage. Given that the R&D firm has decided to come up
with a new patent idea, what price should it charge for the patent? This will of
course depend on the value of the patent for the intermediate firm. The intermedi-
ate goods producer can use the patent for T years and hence earn monopoly profits
throughout this period. The equilibrium profits of the intermediate goods producer
are given by (4.14). Such profits can be made for T years. The value of the firm,
which we can refer to as V j , is the discounted value of all future profits until the
patent expires:

T
1+α
T
V j (0) = Ij β t = (1 − α)α 1−α L βt (4.16)
t=0 t=0
where β ≤ 1 is a time discount factor. Clearly, the lower is β, the lower is the
present value of the firm. If we make the greatly simplifying assumption that β = 1,
1+α
then V j (0) = T (1 − α)α 1−α L.
38 The long run
The R&D firm observes the intermediate goods firm’s value function and will
bid up the price of the patent until P jA = V j (0). A higher price P jA > V j (0) is not
possible; in that case the intermediate goods producer will not buy the patent. The
R&D firm will thus squeeze out all the profits that the intermediate goods firm can
make.
There must also be equilibrium in the capital market. In buying the patent, the
intermediate firm makes an investment of size P jA . During a given time period,
the returns to this investment must not be less than the returns that the firm could
have received by investing in a risk-free asset at an interest rate r > 0. Hence, in
equilibrium it must be the case that rP Aj = Ij .
In a general equilibrium model, there must be equilibria on all markets. In the
R&D sector, finally, if P jA > η, then new R&D firms will enter the sector and
squeeze down profits. Hence, an equilibrium price will require that P jA = η. In

this case, the typical R&D firm will still choose to invent in the first stage and all
the other things will happen in the second stage.
The conditions above imply that we can write
1+α
Ij
(1 − α)α 1−α L
r= A = (4.17)
Pj η
We will return to this result below.

To close the model, note that the market value of all firms in the economy at
time 0 is

A(0)
1+α
Ij = A(0)T (1 − α)α 1−α L
j =1
The aggregate market value will increase linearly with the number of existing
inventions A, with the duration of patents T , and with the size of the labor force L.
That market value increases with the number of inventions (and hence of interme-
diate goods producers) is not surprising. The market value increases also with L.
The reason is that the demand for intermediate capital goods in the final goods
sector increases linearly with L. More workers means that the marginal product of
a new machine is high and hence that demand for that machine is high.
T might be thought of here as the strength of intellectual property rights, where
a large T means strong rights.5 What would happen in our model if the strength
of such rights suddenly fell? In the short run, this would cause a disequilibrium,
which would lead to a number of responses. To start with, the market value of
firms V j (0) would fall. This would in turn mean that the price for patents charged
by the R&D sector P jA would be too high for the intermediate goods producers. In
the short run, R&D firms will therefore choose not to produce any new inventions
and exit from the sector. A new equilibrium can only be restored if also the fixed
costs of invention η fall.
A value of T = 0 would be equivalent to saying that intellectual property rights
do not exist. In such a world, any intermediate goods producing firm can use any
invention without paying for it and there would be no monopolies and no mark-up
pricing. This might have positive welfare effects in the short run since it implies
a lower price of intermediate goods (it would fall from 1/α to 1, i.e. the marginal
2
cost of production in (4.13)) and total demand would increase from α 1−α L to L
1
(since α = 1). Total output would also increase to Y = Aα 1−α L 2−α = AL. How-
ever, on the downside, A would cease to grow in our model, since as long there is
some positive cost of invention η > 0, there is no incentive whatsoever to produce
new ideas that cannot be sold. The issue of whether intellectual property rights is
a good or a bad thing thus entails a trade-off in the model as outlined here; the
absence of property rights removes monopolies and increases welfare in the short
run but decreases the production of inventions in the long run.
4.5 Schumpeterian growth models

In Romer’s model, each new intermediate good makes a positive contribution to
the overall economy and total output in (4.15) increases linearly with A. Inter-
mediate goods producers never really go out of business since there is always a
demand for the good that they are producing.
Building on the classical works of Josef Schumpeter (1934), Aghion and Howitt
(1992) outline a model with the same basic features as in Romer (1990) but with
the important difference that there is instead a fixed number of goods to be pro-
duced and that firms compete over quality. As soon as a firm comes up with a new
and better version of the good, this firm will take the whole market and render all
older versions obsolete. Even though the “incumbent”, leading firm in the sector
had a monopoly, this now becomes useless. In the literature, this is referred to as
business stealing.
Schumpeter recognized already in the early 1900s that entrepreneurs in modern
capitalism had this inherent tendency of not only bringing forth new products
but also continuously destroying a large share of rents for the existing firms. He
famously referred to this phenomenon as creative destruction.
In Aghion and Howitt’s (1992) model, total output is given by

N
Y = L 1−α (q λ j X j )α = L 1−α [(q λ1 X 1 )α + (q λ2 X 2 )α + · · · + (q λ N X N )α ]
j =1
(4.18)
Just like before, L is the total labor force and X j is the the quantity used of the
intermediate capital good j . N > 0 is now the fixed number of intermediate goods
in the economy. The key new feature of this production function is the inclusion of
a quality indicator q λ j , where q > 1 and the exponent λ j ≥ 1 is a discrete number
1, 2, 3, . . . reflecting how many times that sector j has undergone innovation. For
simplicity, it is assumed that in each sector, each new innovation improves quality
40 The long run
by a factor of q. Different sectors might have had different innovation intensity so
that, for instance, it is possible that λ j > λ j +1 . As has probably become evident, λ j
is now our indicator of technological progress.
The other key difference is that a new superior innovation in sector j is
assumed to appear after Tλ j ≥ 1 periods. Tλ j is now random and takes on a value
1, 2, 3, . . . with a certain probability distribution.6 When a new innovation of qual-
ity q λ j +1 > q λ j then appears, the existing monopoly ends. The firm is assumed to
be risk-neutral and does not take any extra precautions due to this risk.
Apart from these two new assumptions, the model has the same structure as
in Romer (1990). It can be solved by following the same steps as above. Since
intermediate firms are monopolists, they will charge a mark-up price P j∗ = 1/α.
Through the usual profit-maximization conditions, final sector demand for the
2 αλ j
most recent version of good j can be shown to be X ∗j = Lα 1−α q 1−α , which is the
αλ j
same expression as in (4.14) except that it is multiplied by q 1−α . Note that final
sector demand thus increases with quality. Since the price is the same as before,
this is not surprising.
Let us assume that the R&D firm has come up with a new innovation at time
0 and that the intermediate firm considers whether to buy the patent or not. The
expected discounted value of the intermediate firm at time 0 would then be
E 0 (Tλ j ) E 0 (Tλ j )
1+α αλ j
E 0 (V j ) = Ij β t = (1 − α)Lα 1−α q 1−α βt
t=0 t=0
αλ j
The noteworthy features of this expression are the two terms q 1−α and E 0 (Tλ j ).
Rapid technological progress in this model means that a relatively large number
of innovations happen all the time. On the one hand, this means that at time 0, λ j
should be relatively large, which would imply a relatively high E 0 (V j ). But on the
other hand, rapid technological progress would also mean that the expected dura-
tion of monopoly should be short, i.e. E 0 (Tλ j ) should be small and the expected
value of the firm relatively small. If the latter effect dominates, firms will be dis-
couraged from entering the sector. In this manner, creative destruction does not
have an unambiguously positive effect on the economy.
4.6 Innovation versus imitation

The models reviewed so far in this chapter have all had developed rich countries
in mind that have sufficient economic and human resources to engage in R&D
that expands the world technological frontier. However, it is well known that most
of the countries in the world do not pursue R&D at all but rather aim to imi-
tate the innovations developed elsewhere. In this section, we will briefly present a
model that considers the choice between innovation and imitation along the lines
of Acemoglu et al (2003).
The basic idea in this model is that the level of technological knowledge in
a country can be enhanced either by innovation or by imitation. The total level
of effort in society is unity and an effort level e < 1 is devoted to imitation and
1 − e to innovation. Effort might in this setting be interpreted as working time or
directed investments or policies pursued by the government. Innovation requires a
relatively high level of skills, whereas imitation does not. Let us imagine that the
level of technological knowledge in a country at time t is given by
At = eμ Āt−1 + γ (1 − e)Ht−1 At−1 (4.19)
The current level of technology At is thus a result of a process exploiting last year’s
level of knowledge in the country At−1 as well as the level at the world technology
frontier, Āt−1 . The two terms are multiplied respectively by effort levels and the
parameters μ, γ > 0 that describe the (time-invariant) usefulness of imitation and
innovation for knowledge growth in the country. Furthermore, At−1 is multiplied
by the level of human capital Ht−1, capturing the general level of education in the
country. This reflects the assumption that innovation is inherently more reliant on
skills than imitation.7
If we divide expression (4.19) by At−1 and subtract At−1 /At−1 = 1 on both
sides, we can write
At − At−1
= gt−1 = eμdt−1 + γ (1 − e)Ht−1 − 1
At−1
where dt−1 = Āt−1/At−1 ≥ 1 is a measure of the country’s distance from the world
technological frontier. If dt−1 is large, imitation is relatively effective.
How should a social planner in a country optimally devote effort to imitation
and innovation? The simple answer is that all effort should be devoted to imitation
if the marginal product of imitation exceeds the marginal product of innovation,
i.e. if μdt−1 > γ Ht−1 , and to innovation if the opposite result holds. The main
insight from this expression is that in countries that are far from the technological
frontier (a high dt−1 ) and where human capital levels are relatively low, the growth
of technological knowledge is optimally driven solely by imitation and is given by
gt−1 = μdt−1 − 1. On the other hand, for countries at the technological frontier
where dt−1 = 1, it will be the case that imitation is not possible, that the optimal
level of imitation effort is e = 0, and that gt−1 = γ Ht−1 − 1. Note also that if the
level of human capital were a choice variable for governments, it would in this
model be pointless to increase, for instance, levels of engineering skills (Ht−1) if
the country is mainly an imitator, since such skills are only useful for innovation.
5 The Overlapping Generations Model
The growth models presented so far were representations of how macro vari-
ables and profit-maximizing firms are hypothesized to behave in relation to each
other. None of the results were based on an analysis of household behavior over
time, i.e. on the behavior of and choices made by utility-maximizing individu-
als and households who live for more than one period and who care about the
future. The overlapping generations (OLG) model outlined below now introduces
a micro-founded model of economic growth and intertemporal choice that will be
employed as a “workhorse” model in several chapters to follow.
One of the key advantages of the OLG model is that it allows us to derive
an endogenous saving rate. In the Solow model, the saving rate s was simply
exogenously determined.
The OLG model was initially presented by Diamond (1965) and Blanchard
(1985). We will henceforth carry out the analysis in discrete time, i.e. we con-
sider periods t, t + 1, t + 2, . . . , instead of a continuous time framework. As we
shall see, the insights from a discrete time framework are complementary to those
already shown.
5.1 Household optimization

The model makes the following basic assumptions. Individuals live for two peri-
ods. In the first period of an individual’s life, he or she belongs to a young
generation and in the second generation to an old generation. In each period, there
is a young and an old generation living at the same time (hence the term “overlap-
ping generations”). Let us denote consumption by an individual in generation j
during period t as c j,t , where j = 1 means the young generation and j = 2 means
the old generation and t = 1, 2, 3, . . . is the actual period in question.1 Individuals
are born without any assets except 1 unit of labor supply. We assume that people
gain utility only from their own consumption and that they do not leave any money
for future generations.
In the most general case, let us write the utility function for an individual who
is young in period t as
Ut = u(c1,t ) + βu(c2,t+1 ) (5.1)

The overlapping generations model 43
Slope: u ′(c1) = 1 Slope: u′(c ) = 1/4

2
u (c)
1
u ′(c)
1/4
c
c1 c2
Figure 5.1 A strictly concave utility function and its associated marginal utility.
where β ≤ 1 is a time discount factor. The closer β is to unity, the greater the
patience of individuals. If β instead is very low, the individual discounts the future
a lot and thinks consumption when young is more valuable in relative terms. The
utility function satisfies the usual assumptions of a positive but strictly diminishing
marginal utility, so that u (c j,t ) > 0 and u (c j,t ) < 0 at all c j,t > 0. It is further
time-separable since u (c1,t ) is independent of c2,t+1 .
The utility function and its associated marginal utility curve are shown in
Figure 5.1. In order to understand more intuitively how the curves are linked,
consider for instance the point on u(c) where the slope is equal to unity. The level
of c at which this happens is c = c1 so that u (c1 ) = 1. At some higher level of c,
for instance at c = c2 > c1 , we have that u (c2 ) = 1/4 < u (c1 ). It is important to
understand this basic theme of utility functions because it will appear again and
again in this text: as c increases, u(c) increases whereas u (c) falls.
Individuals earn an income y1,t during young age in period t and an income
y2,t+1 when they are old. What is not consumed in the first period can be saved for
old age. The individual will then get an interest rate rt+1 on their savings, which
amount to st = y1,t − c1,t ≥ 0. However, savings may also be negative, such that
st = y1,t − c1,t < 0. In that case individuals borrow for consumption in the first
period at the same interest rate rt+1 .2 Consumption during old age is thus
c2,t+1 = y2,t+1 + (y1,t − c1,t )(1 + rt+1 ) (5.2)
If we isolate the income terms on the right-hand side, we get what is usually
referred to as the individual’s intertemporal budget constraint:
c2,t+1 y2,t+1
+ c1,t = + y1,t (5.3)
(1 + rt+1 ) (1 + rt+1 )
44 The long run
What this expression says is that the present value of a person’s current and future
consumption must be equal to the present value of his or her lifetime incomes.3
How much should the individual optimally consume in each period? In order to
solve this problem, we can start by inserting (5.2) into the utility function in (5.1)
and take the first-order condition for a maximum:
∂Ut
= u (c1,t ) − βu (c2,t+1 )(1 + rt+1 ) = 0
∂c1,t
By rewriting this expression, we obtain what is usually referred to as the Euler

equation:
u (c1,t
∗ )
∗ = β(1 + rt+1 ) (5.4)

u (c2,t+1 )
This fundamental result will be derived many times in the chapters to come.
It implies that an individual should optimally consume so that the relative
marginal utility of consumption when young, u (c1,t ∗ )/u (c ∗
2,t+1 ), exactly equals
∗
β(1 + rt+1 ), where c j,t denotes an optimal level. What does this imply for the
optimal relative level of consumption c1,t ∗ /c ∗
2,t+1 ?
Consider, for instance, a case where individuals are patient so that β is close to
unity and where rt+1 is relatively high. In that case, it is likely that β(1 + rt+1 ) > 1.
According to (5.4), we must then optimally have chosen consumption levels such
that u (c1,t
∗ )/u (c ∗ ∗ ∗
2,t+1 ) > 1 applies as well. Note that u (c1,t ) > u (c2,t+1 ) must
(due to diminishing marginal utility as illustrated in Figure 5.1) imply that c1,t ∗ <
∗
c2,t+1 . Hence, if individuals are patient and if interest rates are high, consumption
should be greater in old age than during youth, and vice versa.
With a general utility function like that in (5.1), we cannot get any further than
interpreting the optimality condition in (5.4). We are not able get any explicit
solution for either c1,t∗ , c∗
2,t+1 , or the level of saving. If we want to achieve closed-
form solutions, we need to assume a more specific utility function.
5.1.1 Example 1: Logarithmic utility

One of the very simplest cases is to assume logarithmic utility such that u(c j,t ) =
ln c j,t . Logarithmic utility implies that the utility function is concave and that the
individual is risk-averse, since u (c j,t ) = 1/c j,t > 0 and u (c j,t ) = −1/c2j,t < 0.
The Euler condition (5.4) then simply becomes
∗
c2,t+1
∗ = β(1 + rt+1) (5.5)
c1,t
From this condition, we can achieve closed-form solutions for our intertempo-
∗
ral choice of consumption. Insert c2,t+1 ∗ into the left-hand side
= β(1 + rt+1 )c1,t
of (5.2) and solve for ∗
c1,t to find
∗ y2,t+1 + y1,t (1 + rt+1 )

c1,t =
(1 + β)(1 + rt+1)
When we have solved for c1,t , it is of course easy to also solve for c2,t+1 :
∗ β[y2,t+1 + y1,t (1 + rt+1 )]

c2,t+1 =
1+β
These results have some straightforward interpretations. First, optimal con-

sumption in both periods increase with income levels y1,t and y2,t+1 . Second,
∗
we already know from before that c2,t+1 increases with people’s patience β and
with the interest rate rt+1 . However, note that an increase in rt+1 with incomes
∗ . This might be referred to as a substitution effect: the
held constant will lower c1,t
individual substitutes consumption in young age for consumption in old age.
We can also solve for the optimal level of savings:
y1,t β(1 + rt+1 ) − y2,t+1

st∗ = y1,t − c1,t
∗
=
(1 + β)(1 + rt+1)
Not surprisingly, savings increase with y1,t , β and with rt+1 . Note, however, that,
all else being equal, a net increase in y2,t+1 will lead to a decrease in savings.
5.1.2 Example 2: CRRA utility

Let us now assume that the utility of an individual is given by a function character-
1−θ
ized by constant relative risk aversion (CRRA) where u(c j,t ) = (c1t − 1)/(1 − θ )
and β = 1/(1 + ρ) ≤ 1:
1−θ
c1,t −1 1 c2,t+1 − 1
1−θ
Ut = + (5.6)
1−θ 1+ρ 1−θ
Here ρ ≥ 0 is a time discount rate and θ is the Arrow–Pratt measure of relative risk
aversion (individuals are risk-averse). The CRRA utility function is particularly
useful since it can encompass several different types of utility, depending on the
value of θ . A θ < 0 means that the individual is risk-loving (since marginal utility
c−θ increases with the level of c), θ = 0 means risk neutrality, whereas θ > 0
implies a risk-averse individual. It can further be shown that if θ = 1, then u(c j,t ) =
ln c j,t .4 The case most often studied will be θ ∈ (0, 1).
We now assume, for simplicity, that individuals earn a wage income when
young, y1,t = wt , and that an old individual has zero income, y2,t+1 = 0. Hence,
young individuals must save some of their labor income when young in order to
consume when old:
st = wt − c1,t (5.7)
46 The long run
When old, individuals consume their savings from the previous period plus the
interest earned from these savings:
c2,t+1 = (1 + rt+1)st = (1 + rt+1 )(wt − c1,t ) (5.8)
In this expression, rt+1 is the interest rate on savings as before.

The intertemporal budget constraint reduces to
c2,t+1
c1,t + = wt (5.9)
1 + rt+1
The objective function in (5.6) and the constraint in (5.9) define a constrained
maximization problem, which can be solved by forming a Lagrangian function

1−θ 1−θ
c1,t −1 1 c2,t+1 − 1 c2,t+1
= + + λ wt − c1,t −
1−θ 1+ρ 1−θ 1 + rt+1
where λ is a Lagrange multiplier.5

The first-order conditions for c1t and c2t+1 are
∂ −θ
= c1,t −λ=0
∂c1,t
−θ
∂ c2,t+1 λ
= − =0
∂c2,t+1 1 + ρ (1 + rt+1)
−θ (1+r
−θ c2t+1 t+1 )
Since c1t = λ and 1+ρ = λ, we can write
−θ ∗ 1
−θ
c2,t+1 (1 + rt+1 ) c2,t+1 1 + rt+1 θ
c1t = =⇒ ∗ = = 1 + gc,t (5.10)
1+ρ c1t 1+ρ
In other words, the relative consumption of the individual when in old age can be
described as 1 + gc,t , where
1 1
c2,t+1 − c1t (1 + rt+1 ) θ − (1 + ρ) θ
gc,t = = 1
c1t (1 + ρ) θ
is the growth rate of consumption between t and t + 1. Note that gc,t can be pos-
itive or negative. gc,t will depend positively on the interest rate rt+1 , negatively
on the time discount rate ρ, and negatively on the risk-aversion parameter θ . The
greater is ρ, the greater is the individual’s impatience and hence the lower is their
consumption when old. A θ that approaches 1 further means that the utility func-
tion has a markedly concave curvature and that the marginal utility of consumption
diminishes fast (θ = 1 implies that u(c j,t ) = ln c j,t ). Such individuals will also
consume relatively less in the future than if they entertained a θ that is close to 0.
The result in (5.10) is often referred to as the “discrete Ramsey” result.6 It
will play an important role in the sections ahead. Note that the requirement for
a positive growth rate is that rt+1 > ρ. The next step is to derive an expression
for rt+1 .
5.2 Endogenous saving

In this section, we will now extend the model to include factor markets. From
(5.7), (5.8), and (5.10) we know that
∗
c2,t+1 (1 + rt+1 )st
∗ = = 1 + gc,t
c1,t (wt − st )
which implies that
(1 + rt+1 )st + (1 + gc,t )st = (1 + gc,t )wt
and that
(1 + gc,t )wt wt
st = =
2 + rt+1 + gc,t (1 + rt+1 )/(1 + gc,t ) + 1
Inserting the expression for 1 + gc,t from (5.10) yields
wt
st = θ−1 1
(5.11)
(1 + rt+1) θ (1 + ρ) θ +1
Saving will thus be a positive function of wt and of rt+1 (since θ < 1). Both are
unknown at this stage. However, by introducing firms into the analysis, we might
also derive wt and rt+1 . Not surprisingly, saving decreases with the time discount
rate ρ.
5.2.1 Firms
Firms produce according to the standard neoclassical aggregate production
function
Yt
Yt = F(K t , L), where = f (kt ) and Yt = L f (kt )
L
48 The long run
We assume that labor and capital are paid their marginal products:
∂Yt 1
rt = = L f (kt ) = f (kt )
∂ Kt L
∂Yt Kt
wt = = f (kt ) − L f (kt ) 2 = f (kt ) − f (kt )kt
∂L L
Furthermore, the total capital stock at t + 1 equals st (individual saving in period t)

times the number of individuals L, so that K t+1 = st L.7 This in turn implies that
kt+1 = st .
5.2.2 The steady state

By inserting the derived terms for rt+1 and wt into (5.11), taking into account that
kt+1 = st , we can write
f (kt ) − f (kt )kt

kt+1 = θ−1 1
[1 + f (kt+1 )] θ (1 + ρ) θ +1
This expression is in general form and, although it shows kt+1 as a function of kt ,

it is not very informative. We therefore make two additional assumptions: First,
that the production function f (kt ) is Cobb–Douglas, i.e. f (kt ) = k α . Second, in
the utility function, we assume as before that θ = 1 so that the utility function has
a logarithmic form.
Armed with these two assumptions, the equation for kt+1 simplifies
considerably:
ktα − αktα ktα (1 − α)

kt+1 = α−1 0
= = st (5.12)
(1 + αkt+1 ) (1 + ρ)1 + 1 2+ρ
In a steady state, it must be the case that kt+1 = kt = k ∗ = s ∗ . Thus
1
(k ∗ )α (1 − α) 1 − α 1−α
k∗ = =⇒ k ∗ =
2+ρ 2+ρ
Note that this is the equivalent of the steady-state level of k derived in the Solow
model. The key difference is that ρ enters this expression since we have now
explicitly taken into account an individual’s preferences. k ∗ will decrease with ρ
just like the growth rate of consumption in (5.10) decreases with ρ. Hence, the
greater the individual’s patience, the lower is ρ and the greater is the steady-state
level of the capital stock. It is further easily derived that the equilibrium level of
α
output per capita is y ∗ = (k ∗ )α = ( 1−α
2+ρ )
1−α .
5.3 Endogenous growth
A key result from the OLG model is that the optimal intertemporal growth rate of
consumption is given by
1
c2t+1 − c1t ∗ 1 + rt+1 θ
= gc = −1
c1t 1+ρ
In the OLG model, rt+1 was found as in the analysis above. Let us now assume
instead a world that is described by Romer’s R&D model in Section 4.4 with
the only exception that we now also specify household optimizing behavior as in
Section 5.1 of the OLG model. Recall from (4.17) that in a general equilibrium, the
intermediate firm’s annual level of profit Ij from investing a sum P jA in buying a

patent lasting for T years must be equal to the risk-free annual return in the capital
market on that same investment rP Aj . Remember further that the R&D firms will
charge a price P jA = η for the patents, where η is the cost of developing the patent.
This general equilibrium implies
1+α
Ij = (1 − α)α 1−α L = r P jA = r η
If we combine these expressions, we can substitute in r and express the steady-

state growth rate of optimal consumption in the Romer model as
1−α L 1
1 + 1 (1 − α)α 1+α
η θ
gc∗ = −1
1+ρ
The growth rate will increase with the number of workers L. As explained above,
this “scale effect” in the Romer model arises since more workers means a greater
demand for intermediate goods. The growth rate will fall with the cost of inventing
new goods, η. As before, the growth rate decreases with the level of the behavioral
parameters θ and ρ.
Note that the equation above further implies that gc∗ > 0 only if η1 (1 − α)
1+α
α 1−α L > ρ. If people are generally impatient so that ρ is high and if the cost
of invention η is high, then this criterion for positive growth might not be satis-
fied. It further suggests that a certain population size is necessary for R&D to be a
viable path to consumption growth.
Part II
The Short and Medium Run

6 Equilibrium Business Cycles
From the long run, we now move to the short and medium run, by which we will
here mean time periods of less than five years. Long-run growth is the average
growth rate over decades, but in the shorter run there can be substantial cycli-
cal variations in GDP, referred to as business cycles. The theory of real business
cycles (RBCs) is associated mainly with Kydland and Prescott (1982) and Long
and Plosser (1983). See also the overview in Rebelo (2005).
Early RBC models were in part inspired by the empirical record showing that
business cycles did not appear to follow any systematic cyclical patterns. There
was also a dissatisfaction with the Keynesian type of explanations emphasizing
market failures in the form of, for instance, wage rigidities. Keynesian theory fur-
ther lacked a micro foundation, i.e. the modeling was not based on the decisions
of optimizing individuals. The RBC models instead proposed a framework based
on individual behavior where the engine of the cyclical behavior was real shocks,
induced by technological change and government spending, for instance, rather
than nominal or monetary effects as emphasized by Keynesian theory. A further
difference from previous modeling was the introduction of leisure in the utility
function of the individual. This novelty made it possible to derive implications for
the intertemporal substitution of labor supply.
6.1 Technology shocks to production

Let us assume that aggregate output in the economy is produced according to the
Cobb–Douglas production function
Yt = K tα (At L t )1−α (6.1)
where K t is total physical capital, At is the level of technological knowledge, and

L t is labor. Let us further assume for simplicity that K t is exogenous to the model
so that K t = K . Labor is given by L t = Nt lt , where Nt is the size of the working
population and lt is the number of working hours that each worker provides. For
simplicity, Nt = N. The key determinant of L t is thus lt . These assumptions mean
54 The short and medium run
that we can rewrite (6.1) in log form as
ln Yt = α ln K + (1 − α) ln N + (1 − α)(ln lt + ln At ) (6.2)
= + (1 − α)(lnlt + ln At )
where is the exogenous part. The sources of variation in output will therefore
be working hours lt and technology At .
The development of technological knowledge is assumed to follow the follow-
ing process:
ln At = Ā + gt + Ãt (6.3)
In this expression, Ā is the initial level of technology, g is the trend growth rate of
technological knowledge (as in the Solow model), t is time, and Ãt is a stochastic
shock to the trend. More specifically, the stochastic component is given by
Ãt = ρ A Ãt−1 + t (6.4)
where ρ A ∈ (0, 1) is a parameter indicating the persistence of past shocks for the
current level of technology and t is an error term such that E(t ) = 0 for all t.
A positive technology shock might, for instance, be the sudden appearance of a
new and drastically improved computer program or a breakthrough in transporta-
tion technology. A negative technology shock to trend growth could arise if the
implementation of some existing technology unexpectedly stalled during some
period.
In technical terms, the process described above follows a first-order autore-
gressive process (AR(1)) since Ãt depends on the corresponding level one period
before. Inserting (6.4) into (6.3) gives us
ln At = Ā + gt + ρ A Ãt−1 + t
where it might be noted that the expected level is E t (ln At ) = Ā + gt + ρ A Ãt−1

and where the expected growth rate of At is thus g.
In order to see the effects on output, suppose, for instance, that a positive
technology shock happens in period 1 so that 1 = ¯ > 0 and that 2 = 3 = 0.
Assume also that Ã0 = 0. In that case, Ã1 = ¯ and Ã2 = ρ A ¯ . By period 3,
we have Ã3 = ρ A ( Ã2 ) = ρ 2A ¯ . The level of technological knowledge is therefore
ln A3 = Ā + 3g + ρ 2A ¯ . The shock in period 1 is still felt in period 3 by an amount
ρ 2A ¯ . As time goes by, the effects of the shock approach zero since ρ A < 1.
Finally, comparing the level of total output in the economy with and without the
shock, we can infer from (6.2) that the shock causes a cyclical upturn since total
output in period 3 is higher than usual by an amount (1 − α)ρ 2A ¯ > 0.
Equilibrium business cycles 55
6.2 Labor demand
One of the key consequences of technology shocks is the impact they have on the
labor market. Let us imagine a representative firm i in the economy that produces
according to the production function in (6.1) and sells its good at a price Pt .1 For
simplicity, we will set price equal to unity so that Pt = 1. The labor market is
competitive and workers are hired until the marginal product of labor equals the
market wage rate, wt . The profit function for the representative firm is
it = (K ti )α (At L it )1−α − wt L it − rt K ti (6.5)
where K ti is the physical capital held by firm i at time t, L it is the labor employed
by firm i , At is nonrival technology available to all firms, and rt is the interest rate
on capital.
The first-order condition for profit maximization for labor is
∂it
= (K ti )α A1−α
t (1 − α)(L it )−α − wt = 0
∂ L it
where the left-hand side in the middle expression is the marginal product of labor.2
Rearranging terms gives us an expression for firm i ’ s demand for labor, L i,D :3
1
(1 − α)A1−α α
L i,D
t = Kt
i t
(6.6)
wt
What is immediately evident from this expression is that labor demand will
increase with a positive technology shock that increases At . Analogously, a
negative technology shock will decrease labor demand. Labor demand will also
increase with the firm’s stock of capital K ti since that increases the marginal
product of labor.
RBC models assume a perfectly functioning economy where all markets are
in equilibrium. Hence, an increase in labor demand should generally mean
that aggregate employment (in terms of working hours lt ) should increase. In
the empirical literature, the prediction about a positive relationship between
technology shocks and employment has been extensively discussed.
6.3 Households
Labor supply is determined by the households and involves a trade-off. On the
one hand, more work means more income and higher consumption. On the other
hand, more work means less leisure, which is now assumed to be a part of the
individual’s utility function. The lifetime utility of an individual who lives for two
periods and who receives utility from consumption and leisure is given by the
function
U = U (c1 , c2 , 1 − l1 , 1 − l2 ) (6.7)
= ln c1 + b ln(1 − l1 ) + β[ln c2 + b ln(1 − l2 )]
where ct is consumption in period t = {0, 1}, lt is working hours in period t, b > 0

is a parameter showing the relative weight given to leisure, and β ≤ 1 is a time
discount factor. For simplicity, we have normalized total working hours to unity,
so that the amount of leisure equals 1 − lt .
First-period consumption might be saved for the second period and the indi-
vidual works in both periods for a wage wt . Second-period consumption must
therefore equal savings plus second-period income from work:
c2 = (1 + r )(w1l1 − c1 ) + w2l2
The individual receives a (real) interest rate r on savings (w1l1 − c1 ). Rewriting

this condition yields an intertemporal budget constraint:
c2 w2 l 2
c1 + = w1 l 1 + (6.8)
1+r 1+r
By utilizing (6.7) and (6.8), we can set up the individual’s maximization problem
as a Lagrangian:
= ln c1 + b ln(1 − l1 ) + β[ln c2 + b ln(1 − l2 )]

w2 l 2 c2
+ λ w1l1 + − c1 −
1+r 1+r
In this optimization problem, we let labor supply l1 and l2 be the choice variables.
The first-order conditions are
∂ b
=− + λw1 = 0 (6.9)
∂l1 (1 − l1∗ )
∂ bβ λw2
=− + =0 (6.10)
∂l2 (1 − l2∗ ) 1 + r
By isolating the λs on the right-hand sides in (6.9) and (6.10), we can write
b bβ(1 + r )
λ= =
(1 − l1 )w1 (1 − l2∗ )w2
∗
If we cancel the bs, this can in turn be rewritten as
1 − l1∗ 1 w2
= (6.11)
1 − l2∗ (1 + r )β w1
Equilibrium business cycles 57
What we have achieved in (6.11) is an expression for the utility-maximizing rela-
tive leisure in period 1. Since it is not possible to obtain explicit solutions for
l1∗ and l2∗ , we have to base our analysis on (6.11).4
Let us assume that there will be a known fall in the relative wage w2 /w1 .
Such a fall means that (1 − l1∗ )/(1 − l2∗ ) optimally must fall too, implying that
relative labor supply l1∗ /l2∗ must rise. In other words, a fall in the relative wage
in period 2 means that individuals will work more in period 1, thus causing a
temporary increase in total output in (6.2).
As another example, consider an increase in the real interest rate r . Just as in
the previous case, such an increase means that relative first-period labor supply
l1∗ /l2∗ must increase. The intuition is that a rise r makes first-period labor income
relatively more valuable (since only first-period consumption can be saved), which
leads to an intertemporal substitution of labor supply towards more work in the
first period. This rational response to a real shock also leads to a boom in the
first-period total income Yt .
As the examples above have indicated, RBC models are often too complex to
be solved analytically. Therefore, scholars in this tradition use advanced methods
of simulation where realistic parameter values are plugged in before a variation in
some real variable is made.
7 Financial Crises
In real business cycle models of the basic kind as above, all markets work effi-
ciently, including the financial market. According to that logic, there is therefore
no point in modeling a separate financial sector. However, the deep financial crisis
that hit the world economy in autumn 2008 seems not to lend much support to
an efficient financial market hypothesis. A recurring feature of financial crises
rather appears to be bank runs, i.e. that a large number of depositors withdraw
their savings at the same time, potentially causing banks to become insolvent and
collapse.1 In some serious cases, this might even cause a systemic banking crisis
when a country’s whole banking system is close to collapse.2 The macroeconomic
consequences on, for instance, economic growth and fiscal balances are typically
substantial and go beyond normal business cycle downturns (Reinhart and Rogoff
2009a, b).
In this section, we present a model of bank runs from Chang and Velasco (2001),
who in turn build upon Diamond and Dybwig (1983). It is mainly meant to capture
an “emerging market” with an open economy and a fairly advanced banking sector.
Recent events suggest that it might also be used to describe the situation in, for
instance, the United States.
7.1 Basic assumptions

Let us consider an open economy with a large number of identical agents. There
is one good that can be freely traded on world markets and that can be used for
consumption or for investment. The price of the consumption good on world mar-
kets is fixed and set to one (dollar) for simplicity. There are three time periods,
t = 0, 1, 2. The model contains the following key components:
• Domestic individuals are born with an endowment of e > 0 dollars.

• Domestic individuals might invest one dollar in a risk-free investment project
at time t = 0 that yields r < 1 at t = 1 and R > 1 at t = 2. The investment project
is said to be illiquid, since an early liquidation at time 1 causes a loss of 1 − r
dollars.
• Domestic individuals can invest or borrow money on an international capital
market at a zero interest rate. The maximum amount of borrowing is f > 0,
which might be thought of as a domestically imposed credit restriction.
Financial crises 59
• Individuals are either “impatient” with a probability ρ and consume only in
period 1 or “patient” with a probability 1 − ρ and derive utility only from
consuming in period 2. The realization of each individual’s type is private
information.
• c1 denotes consumption in period 1 if the individual turns out to be impatient
and c2 is consumption in period 2 if the individual turns out to be patient.
The expected utility of a representative individual with CRRA preferences and

no time discount rate is therefore3
c11−θ c1−θ
U =ρ + (1 − ρ) 2 (7.1)
1−θ 1−θ
Note that a patient individual who invests all her initial endowment e as well as
the maximum amount she can borrow on the international market f in the risk-free
project will consume e R + f (R − 1) in period 2. Clearly, only individuals who
know that they are impatient will choose to invest in the world market rather than
in the domestic project (since this will give them c1 = e instead of c1 = e(1 − r )).
7.2 Banks
Since individual types are private information, people have an interest in pooling
risk and money. Let us therefore assume that they form a bank together where
they pool all their resources and that aims to maximize the utility of the repre-
sentative individual. The individual can withdraw money for consumption either
in period 1 or in period 2. The total amount (per depositor) that the bank invests
in the investment project is k > 0. Furthermore, the bank can borrow money on
the international capital market. Let d be the amount of “long-term” borrowing
per depositor in period 0 to be repaid in period 2 and b be the equivalent num-
ber in period 1 to be repaid in period 2. Since it is beneficial to borrow abroad
and invest domestically, it will always be the case that d + b = f . The prevalence
of impatient individuals who want to consume in period 1 implies that the bank
needs to liquidate part of the investment project at this time. l < k is the size of
this liquidation.
The timing of events in the model is the following:
• At t = 0: Individuals deposit e in the bank and the bank borrows d on the

international market. The bank invests this money into a long-run project such
that d + e = k.
• At t = 1: In normal times, a fraction ρ of depositors withdraw c1 from the bank.
In order to cover this withdrawal, the bank borrows an additional amount b on
the international market and potentially liquidates part of the project at a loss
such that ρc1 = b + rl. Should a bank run occur, however, all depositors might
want to withdraw their money.
• At t = 2: The part of the project that has not been liquidated (k − l) yields a
return R. This amount needs to be large enough in order for loans b + d = f
to be repaid and for patient consumers to withdraw their money: R(k − l) =
(1 − ρ)c2 + d + b = (1 − ρ)c2 + f. A so-called “incentive compatibility
constraint” necessary for anyone to want to invest long-term is further that c2 ≥ c1 .
What are the socially optimal levels of c1∗ , c2∗ , and l ∗ , i.e. the solutions that
maximize the joint welfare of both patient and impatient individuals? To start with,
recall that early liquidation is associated with a loss, which means that l ∗ = 0 must
be the socially optimal level.4 Hence ρc1∗ = b in period 1, i.e. the bank will satisfy
impatient depositors’ consumption needs only through international borrowing.
In period 2, we will further have that (1 − ρ)c2∗ + f = Rk (recalling that l ∗ = 0).
What we want to do is to rewrite the conditions above into an intertemporal
budget constraint. We can manipulate the expressions above,
(1 − ρ)c2∗ = Rk − f = R(d + e) − f = R( f − b + e) − f
= R( f − ρc1∗ + e) − f
and then rearrange the resulting expression so that c1∗ and c2∗ end up on the left-
hand side:
Rρc1∗ + (1 − ρ)c2∗ = R( f + e) − f = Re + f (R − 1) (7.2)

f (R − 1)
= Rw, where w = e +
R
The term w thus describes the individual’s initial endowment plus the returns to
the investment project and might be interpreted as the economy’s wealth (per indi-
vidual). It also serves as the budget constraint for the individual in its utility
maximization.
In order to maximize the utility function in (7.1) subject to (7.2), we set up the
Lagrangian
c11−θ c1−θ
=ρ + (1 − ρ) 2 + λ[Rw − Rρc1∗ − (1 − ρ)c2∗ ]
1−θ 1−θ
where λ is the Lagrange multiplier, as always. The first-order conditions are
∂
= ρc1−θ − λRρ = 0 (7.3)
∂c1
∂
= (1 − ρ)c2−θ − λ(1 − ρ) = 0 (7.4)
∂c2
The solution to this maximization problem can be found by using the following
procedure. First, by combining (7.3) and (7.4), it can be shown that the equilib-
rium levels of consumption are given by c2∗ /c1∗ = R 1/θ . Inserting the value for c1∗
Financial crises 61
into the budget constraint in (7.2) eventually allows us to express the closed-form
equilibrium levels:5
ρc1∗ = αw, (1 − ρ)c2∗ = w R(1 − α) (7.5)

−1
1−ρ
where α = 1 + <1
ρ R (θ−1)/θ
Not surprisingly, consumption levels in both periods will depend on wealth w.

Wealth, in turn, increases linearly with initial endowments e and with the
international credit limit f , as we know from (7.2).
The key thing to remember from these solutions so far is that the socially opti-
mal situation is that no premature liquidation occurs, i.e. that l ∗ = 0, but this might
still be the actual outcome if too many investors withdraw their money in the first
period. We will analyze this aspect more closely next.
7.3 A bank run equilibrium

Let us imagine that banks have the following contract with their depositors.
In period 0, depositors surrender to the bank their initial endowment e and their
ability to borrow abroad b and d. In return, each individual can withdraw either c1∗
units for consumption in period 1 or c2∗ units in period 2. Note that we will ana-
lyze scenarios when even patient individuals for some reason choose to withdraw
money in period 1 and thus potentially cause a dangerous bank run. In this case,
banks will need to liquidate part of the project (l > 0), although this is not socially
optimal.
Banks are assumed to face a hard budget constraint in the sense that they must
repay their foreign debt in all circumstances. What this means is that the invest-
ment project must at least provide enough returns in period 2 to cover period 0 and
period 1 loans: R(k − l) = b + d = f. Rewriting this condition, and making use of
the solutions above, gives us an expression for the maximum level of liquidation
that the bank can undertake in the first period:
Rk − f (1 − ρ)c2∗
l+ = = = w(1 − α) (7.6)
R R
Imagine now that all depositors come and want to withdraw money in the first
period, perhaps due to a bank panic. The bank pays out c1∗ to each of them, in
accordance with the contract. How long can the bank make such payments?
The answer is that withdrawals (per individual) can be made up to the bank’s
total liquidation value b + rl + . This is simply the maximum amount of cash assets
that the bank can provide in the short term. Should consumption withdrawals
exceed short-term foreign loans b plus the highest attainable (premature) project
liquidation value rl + , the bank becomes insolvent and must close down. The for-
eign loans are then paid back already in period 1, impatient and some patient
individuals get their money back, but a number of patient individuals run the risk
of losing their savings.
More specifically, if all individuals withdraw their money in period 1, the
amount (per individual) will be c1∗ . A bank run equilibrium is said to exist if
z + = c1∗ − (b + rl + ) > 0 (7.7)
If this situation prevails, banks cannot survive a sudden withdrawal of all indi-
vidual savings. z + might therefore be referred to as a measure of the bank’s
illiquidity.
If we insert the equilibrium values derived above into (7.7), we find that
αw
z + > 0 if c1∗ − (b + rl + ) = − [αw + r w(1 − α)] > 0
ρ
It is clear that wealth w is just a scalar in this expression and will not matter for
determining the sign. Removing w and recalling that α is given by (7.5) enables
us to show that
z + > 0 if R (θ−1)/θ > r (7.8)
Clearly, if θ ≥ 1, then a bank run equilibrium will exist, since R > r . In the
often assumed range of θ ∈ (0, 1), (7.8) will not be satisfied and there will not
exist a bank run equilibrium. In other words, the more risk-averse individuals are,
the more likely that a bank run equilibrium will exist. Whether the equilibrium
actually materializes or not depends on individual strategies and on the partic-
ular nature of the strategic interaction between individuals. The analysis above
describes the fundamental characteristics of the economy that need to be in place
for a bank run to be possible.
7.4 Foreign credit

In the scenario above, foreign credit in period 1 (b) was available even if a bank run
occurred. What if this was not possible? We know, for instance, that an important
feature of the financial crisis of 2008 and onwards was that banks in general were
unwilling to lend to each other in the midst of the crisis.
In the model, a failure to receive foreign credit in case of a depositor panic in
period 1 would imply that R(k − l) = d = f, which in turn would imply that the
maximum level of first-period liquidation is
Rk − d
la = (7.9)
R
Financial crises 63
Since d < f , it will be the case that la > l+ (see (7.6)). The bank’s level of
illiquidity is

d
z a = c1∗ − rl a = c1∗ − r k − (7.10)
R

1
= c1∗ + r d − 1 − re
R
By comparing this level with that in (7.7) and using (7.6) and (7.9), we can deduce
that

r
z a − z + = b + r (l + − l a ) = b 1 − >0
R
In other words, if foreign creditors will not provide loans in period 1 in case of
a panic, the bank has a higher illiquidity and is more vulnerable than otherwise,
despite the fact that the maximum level of liquidation is higher. This is of course
well in line with intuition.
It is even possible that an international refusal to lend might cause a bank run.
If the international creditors announce in period 1 that b is not available to the
bank, contrary to expectations, depositors will realize that the bank’s vulnera-
bility has increased and will therefore try to withdraw as much as possible in
period 1, potentially causing the bank to collapse. In this sense, it does not matter
if b becomes unavailable through domestic conditions or due to a crisis abroad.
It also shows how foreign financial crises can have serious repercussions on the
domestic financial sector.
7.5 Short-term debt

In the sections above, even if foreign creditors potentially did not lend during a
bank run in period 1, they would still accept having their period 0 loan d repaid
in period 2 when the bank’s investment project had run its course. But what if
creditors demanded all loans to be repaid and the project to be liquidated already
in period 1 in the event of a run?
In this case, the bank would have to pay its short-term obligations c1∗ + d by
using the liquidation value of the whole project r k:
z b = c1∗ + d − r k = c1∗ + d − r (d + e) = c1∗ + d(1 − r ) − r e (7.11)
Using the same procedure as before and comparing z a and z b , we can easily see
that z b > z a , implying that this type of short-term debt rearrangement increases
the risk of a bank run even further.
7.6 Liberalizing international credit markets
It seems to be an important policy issue to analyze what happens when interna-
tional credit markets are liberalized in the sense that f , the international borrowing
constraint, is loosened so that f rises. Several empirical studies on financial crises
seem to suggest that such major economic upheavals are usually preceded by
financial liberalization.
To start with, we know from (7.2) that the economy’s wealth increases with an
increase in f since foreign credit has a zero interest rate and can earn a return
R > 1 for every dollar invested. From (7.5), we also see that consumption in both
periods will increase. In this sense, a deregulation of the international financial
market would increase social welfare.
However, if we assume that the bank lives in a world of short-term debt where
b = 0 and d = f , as just described, we can rewrite (7.11) as
αw
z b = c1∗ + f − r k = + f − r ( f + e)
ρ
αw
= + ( f − αw)(1 − r ) − r e
ρ

α f (R − 1) 1
= e+ + f 1−α+ − αe (1 − r ) − r e
ρ R R
where we have substituted in the expression for w to the right on the second line.
What comes out very clearly from this equation is that z b increases with f . In
other words, if debts can be easily canceled in the case of a bank run, an increase in
capital inflows will increase financial fragility and increase the bank’s illiquidity.6
In this sense, a liberalization of international financial markets is rational ex ante
because it increases the socially optimal level of consumption, but is typically
regarded as irrational after a financial crisis since increased capital inflows make
banks more vulnerable to bank runs.
8 Consumption and Saving
In this chapter, we analyze in depth the determinants of the largest component on

the user side of the GDP equation: aggregate private consumption Ct . Consump-
tion theory has been a key topic in macroeconomics since the Keynesian revolution
in the 1930s. It is also one of the areas where differences between the Keynesian
paradigm and the rational expectations paradigm of Milton Friedman and Robert
Lucas are most evident.
We start this chapter by briefly reviewing the Keynesian consumption func-
tion. We then present an extensive discussion of the permanent income hypothesis
where a representative agent maximizes consumption in a multi-period frame-
work. Given the centrality of this model in modern macroeconomics, we give this
framework a great deal of attention. We also discuss its offshoot, the random walk
model. We then briefly discuss precautionary saving, as well as how the model
changes when we introduce time discount rates and interest rates. Sections 8.8
and 8.9 discuss two currently very intensive fields of research: the importance of
relative consumption concerns and time inconsistency.
8.1 The Keynesian consumption function

The standard aggregate consumption function in the Keynesian framework relates
current consumption Ct to current disposable income Ytd :
Ct = ca + cmpc Ytd = ca + cmpc (1 − τ )Yt (8.1)
In this well-known expression, ca > 0 is the autonomous level of consumption that

is independent of income, cmpc ∈ (0, 1) is the marginal propensity to consume, τ is
the income tax rate, and Yt is aggregate current household income. The after-tax
net income is thus Ytd = (1 − τ )Yt .
The straightforward implication of this model is that a marginal increase
in Yt will increase consumption by cmpc (1 − τ ) > 0. Also, it is noteworthy that
government policy has a strong and immediate effect. A marginal decrease in
the income tax rate will increase aggregate consumption by cmpc Yt > 0.1 In the
Keynesian world, agents are not forward-looking and only react to changes in
current income.
8.2 Friedman’s critique
In a famous book, Milton Friedman (1957) challenged the Keynesian view by
arguing that agents in the real world most likely were able to distinguish between a
permanent and a transitory component in the flow of disposable incomes. If current
disposable income is Ytd , then each individual distinguishes between the long-
run average income, or permanent income, YtP , and transitory income YtT , where
transitory incomes are positive or negative and where the expected level is zero:
E t (YtT ) = E t (Yt+1
T ) = 0. E is an expectations operator showing expectations at
t
time t. Hence, current disposable income equals
Ytd = YtP + YtT

The key new assumption in the permanent income hypothesis (PIH) is that deci-
sions regarding current consumption are only based on permanent income rather
than on the current income:
Ct = YtP (8.2)
The wider implications of this model will be explored in greater detail in the
next section, where a more complete version of the model is presented.
8.3 The permanent income hypothesis

As in the OLG model, the standard version of the PIH employs a micro-founded
framework with a utility-maximizing representative individual who maximizes
lifetime utility, subject to a lifetime budget constraint. Utility is gained solely from
the individual’s own consumption.2
The utility function is

T
U= β t u(ct ) = u(c0 ) + βu(c1 ) + β 2 u(c2 ) + · · · + β T u(cT ) (8.3)
t=0
where ct is individual consumption and where the function has the standard prop-
erties u (ct ) > 0 and u (ct ) < 0. There is thus always a positive marginal utility
of consumption, but marginal utility is declining as consumption increases. The
individual is now assumed to live for T + 1 > 0 years.3 u(ct ) is referred to as
the instantaneous utility function and is stable over time. β ≤ 1 is a time discount
factor, as before.
We assume for simplicity that the individual has a known stream of exogenously
future incomes y0 , y1 , y2 , . . . , yT and that income tax rates are zero. The lifetime
budget constraint (or permanent income) is

T
ct yt
T
= (8.4)
(1 + r )t (1 + r )t
t=0 t=0
Consumption and saving 67
where r ≥ 0 is a time-invariant interest rate. There is further a perfect credit market
so that individuals can borrow money if yt < ct , as long as the lifetime budget
constraint in (8.4) is satisfied.
Although this intertemporal constraint is really just an extension of the two-
period constraint from the OLG model in (5.3), it is worth elaborating a little
exactly how (8.4) arises. Consider, for example, a case where T = 2. In this case,
consumption in the last period must be
c2 = y2 + (y1 − c1 )(1 + r ) + (y0 − c0 )(1 + r )2

That is, the indidual will consume all net savings from previous periods
(y1 − c1 )(1 + r ) + (y0 − c0 )(1 + r )2 plus last period income y2 . If we isolate the
consumption terms on the left-hand side, we can write

2
c2 + c1 (1 + r ) + c0 (1 + r )2 = ct (1 + r )t
t=0

2
= y2 + y1 (1 + r ) + y0(1 + r )2 = yt (1 + r )t
t=0
Normally, we want to rewrite the budget constraint so that it expresses the present
value in period 0, i.e. when the decision about the consumption plan is made.
By dividing both sides by (1 + r )2 , we end up with the expression in (8.4) when
T = 2, which shows that the present value of lifetime consumption must be equal
to the present value of lifetime incomes.
8.3.1 Utility maximization over the life cycle

The utility function in (8.3) and the budget constraint in (8.4) together form a
maximization problem for the individual where the challenge is to find the levels
of affordable consumption that maximize utility. We therefore set up a Lagrangian
function

T
T yt
T
ct
= β t u(ct ) + λ − (8.5)
(1 + r ) t (1 + r )t
t=0 t=0 t=0
where λ > 0 is a standard Lagrange multiplier. Note that the choice variables for
this optimization are the levels of consumption for each time period: c0 , c1 , c2 , etc.
The first-order conditions for a maximum are thus
∂
= u (c0 ) − λ = 0 (8.6)
∂c0
∂ λ
= βu (c1 ) − =0
∂c1 1+r
...
∂ λ
= β T u (cT ) − =0
∂cT (1 + r )T
These first-order conditions imply that u (c0∗ ) = λ = β(1 + r )u (c1∗ ) =

β 2 (1 + r )2 u (c2∗ ) · · · = β T (1 + r )T u (cT∗ ). Hence, we once again obtain the famil-

iar Euler equation, already derived in the OLG model, but now in a (T + 1)-period
framework:
u (ct∗ )
∗ ) = β(1 + r ) for all t ∈ {0, 1, 2, . . . , T − 1}
u (ct+1
8.4 An example
In order to develop the intuition for the PIH result further, let us make the highly
simplifying assumption that β = 1 and that r = 0. In that case, the Euler equa-
tion reduces to u (c0∗ ) = λ = u (c1∗ ) = · · · = u (cT∗ ). In other words, optimally, the
marginal utility of consumption should be the same in all periods. Since u(ct ) is
strictly concave, there can only be one unique level of ct where marginal utility
(the slope of the utility function) is u (ct ) = λ. Hence, the individual will opti-
mally set a level of consumption such that ct∗ = ct+1 ∗ = c ∗ = · · · = c ∗ . This result
t+2 T
of identical optimal levels of consumption throughout life is sometimes referred
to as consumption smoothing.
The fact that the optimal level of consumption is identical throughout an
individual’s lifetime T means that we can reformulate the budget constraint as

T
T
ct = (T + 1)ct∗ = yt
t=0 t=0
Dividing both sides by life length T + 1 gives us the key expression for
consumption in the PIH:
T
t=0 yt
ct∗ = (8.7)
T +1
The interpretation of this expression is simply that an individual will optimally

consume in every period her average annual income. Clearly, the optimal level
of consumption will increase if income level yt increases. What is the impact of
an increase in T ? That will of course depend on the intertemporal structure of
the income flow. If it is the case that all individuals retire at some t r < T where-
after Yt = 0, then an increase in T (i.e. a longer life) will imply a lower level of
consumption today since the flow of incomes must last longer.
8.4.1 Saving
In the PIH, savings (positive or negative) are used to smooth the time pattern of
consumption. Current income is yt = ct + st . Hence, the expression for optimal
saving in the PIH is
T
t=0 yt
st∗ = yt − ct∗ = yt − (8.8)
T +1
where st∗ is positive if current incomes are relatively large and negative if current
income in period t is lower than the optimal level of consumption.4
An important implication of the expression (8.8) is that whereas the level of
consumption is constant over time and is insensitive to the time distribution of
incomes y1 , y2 , . . . , yT , savings are very sensitive to current incomes. To be more
precise, the derivative of st∗ with respect to yt is 1 − 1/(T + 1) > 0 whereas the
derivative of ct∗ with respect to yt is only 1/(T + 1). The sensitivity of savings with
respect to current incomes is larger if T is large. A prediction from these results
would thus be that savings should display a greater degree of variation over life
than consumption levels.
8.4.2 Mid-term analysis

If we only observe an individual in mid-life, i.e. at some t = τ such that 0 < τ < T ,
then in order to understand the individual’s consumption choice, we must take
into account
past incomes and savings as well as future incomes. Note that since
(T + 1)cτ∗ = t=0
T
yt , we can write

T τ −1

T
cτ∗ = yt − T cτ∗ = yt − τ cτ∗ + yt − (T − τ )cτ∗ (8.9)
t=0 t=0 t=τ

T
= Aτ −1 + yt − (T − τ )cτ∗
t=τ
−1
In this expression, τt=0 yt − τ cτ∗ = Aτ −1 shows accumulated incomes minus total
consumption spendings from t = 0 to τ − 1. If Aτ −1 > 0, the individual has positive
assets at the time of analysis τ , and negative if Aτ −1 < 0.
If we isolate cτ∗ on the left-hand side and then divide both sides by T + 1 − τ ,
we obtain
T
∗ Aτ −1 + t=τ yt
cτ =
T +1−τ
A commonly shown version of this expression is when it is assumed that τ = 1.
The expression then collapses into the very simple form
T
A0 + t=1 yt
c1∗ = (8.10)
T
Note that, in this case, A0 = y0 − c0∗ , so that it shows accumulated savings from
previous periods. The net level of assets A0 can be positive or negative, depending
on the individual’s incomes during young age.
In some of the expositions below, we will work with versions of an intertempo-
ral maximization problem that builds on the logic of expression (8.10).
8.4.3 Life-cycle implications

In the real world, there is of course never perfect information about future incomes.
According to the life-cycle hypothesis of consumption, however, we can make the
following reasonable projection about the time distribution of current incomes.
In the first phase of an individual’s life (let us call it period 0), education and
human capital will be acquired and incomes will typically be low. In this phase,
individuals are net borrowers with negative savings. In “middle age” (period 1),
people will work and usually be able to accumulate substantial savings.5 In the
third phase (period 2), finally, people retire from work and live mainly off the
savings accumulated during working years.6 Consumption should, according to
theory, remain at a fairly constant level throughout the three periods.
To illustrate this insight, let us assume that the flow of incomes from young to
old age is y0 , y1 , and y2 and that y0 < y2 < y1 as in Figure 8.1. Incomes are thus
lowest when people
are young and highest when people are middle-aged. Lifetime
incomes are 2t=0 yt and a consumption smoothing individual should choose to

consume ct∗ = 2t=0 yt /3. Hence, we know for sure that y0 < ct∗ < y1 , i.e. that
young people will tend to consume above their earnings, whereas the reverse is
true for most people of middle age. In the scenario described here, we cannot
predict anything definite about consumption relative to income during old age. In
the example in Figure 8.1, y2 < ct∗ .
8.4.4 Policy implications

Comparing the result in (8.7) with the typical Keynesian aggregate consumption
function Ct = ca + cmpc Yt , a number of very different policy implications will
emerge. A temporary increase in income Yt , perhaps due to a temporary gov-
ernment subsidy, should increase consumption by cmpc Yt . If we, for instance,
ct, yt
yt
ct
time
1 2 3
Figure 8.1 Example of consumption and income over the life cycle.
assume cmpc = 1/2 and Yt = 10, then Ct ≈ 5. However, in the PIH scenario, an
identical increase in income should increase consumption only by Yt /T . If we
assume a middle-aged person in the Western world at 40 years who can expect to
live for another 40 years, the increase in consumption will be only 10/40 = 0.4.7
The more far-reaching implication of this example is that short-run policy
efforts aimed at boosting consumption are likely to be fairly successful if we
assume a Keynesian consumption function, but they will be very ineffective in
the PIH framework.
8.5 The random-walk model

A major weakness of the simple version of the PIH shown above was that individu-
als could make perfect assessments of their lifetime incomes. An important feature
of real-world economic decisions is that there is always a substantial degree of
uncertainty about the future and that we can only make informed guesses about
our future economic situation. In Hall’s (1978) random-walk model, which is an
extension of the PIH framework, a key feature is that individuals form expecta-
tions about uncertain future incomes and that these expectations are continually
updated. Individuals are forward-looking and take into account all available infor-
mation at some time t. In economics terminology, we refer to assessments of this
kind as rational expectations.8
An individual’s expected utility at some (mid-life) year t = 1 is now quadratic:
T

a 2
E 1 (U ) = E 1 ct − ct (8.11)
2
t=1
E t is an expectations operator indicating that expectations are made at time t.
The expectations operator has the standard properties used in statistical the-
ory.9 Note that the expression (8.11) can be written out in full as E 1 (U ) =
E 1 (c1 − a2 c12 ) + E 1 (c2 − a2 c22 ) + · · · + E 1 (cT − a2 cT2 ). It is further central to under-
stand that expectations about future levels of consumption are all made at time
t = 1. For simplicity, we still assume that the time discount factor is β = 1.
The instantaneous utility function (inside the parentheses) u t (ct ) = ct − a2 ct2
is also different from what we have seen before. a > 0 is a constant parameter.
Marginal utility of ct is u t (ct ) = 1 − act > 0 if ct < 1/a. This means that if
consumption levels exceed 1/a, marginal utility of consumption will actually be
negative. Although this might be an interesting scenario, we will assume levels of
consumption such that ct < 1/a. The second derivative is u t (ct ) = −a < 0, which
ensures concavity.
The budget constraint is equivalent to the one implied above with the exception
that we have now introduced uncertainty about lifetime incomes. A0 = y0 − c0∗ has
the same interpretation as above and r = 0:

T
T
E 1 (ct ) ≤ A0 + E 1 (yt ) (8.12)
t=1 t=1
By setting up the Lagrangian as in (8.5) and taking first-order conditions

∂
for a maximum ∂c = E 1 (1 − act ) − λ = 0 for all t = 1, 2, . . . , T , we obtain the
t
familiar Euler equation: 1 − ac1 = E 1 (1 − ac2 ) = · · · = E 1 (1 − acT ). Note that the
expectations operator is absent from the first term since c1 is observed at time 1.
The general result is hence that 1 − ac1 = 1 − a E 1(ct ), where t > 1. By canceling
1 and a from this expression, we get the optimal consumption path:
c1∗ = E 1 (ct∗ ) (8.13)
In other words, at the time of decision, the individual expects to consume the same
amount at all future dates as he or she currently consumes.
Inserting this result into the budget constraint yields the key result of the
random-walk model:

T
1
c1∗ = A0 + E 1 (yt ) (8.14)
T
t=1
This expression is identical to the one in (8.10) except for the expectations opera-
tor E 1 . The interpretation is also similar to the one above: the individual consumes
every period a fraction 1/T of his or her expected lifetime resources. A noteworthy
feature is that the presence of uncertainty does not really
T seem to affect the indi-
vidual’s consumption choice. The individual values t=1 E 1 (yt ) as equivalent to
T
t=1 y t in (8.7). Hall’s random-walk model is therefore sometimes described as
being characterized by certainty equivalence. We will relax this assumption below.
8.5.1 Introducing a stochastic component
In order to describe how uncertainty can enter the model, let us imagine that actual
consumption at some date t is given by
ct = E t−1 (ct∗ ) + et = ct−1

∗
+ et (8.15)
In this simple expression, et is a stochastic error term with the property that
E t−1 (et ) = 0. et reflects that there is an element of uncertainty since actual con-
sumption might turn out to be higher (et > 0) or lower (et < 0) that what was
expected at t − 1. Furthermore, we know from (8.13) that E t−1 (ct∗ ) = ct−1
∗ . Hence,
the prediction is that consumption at t equals consumption at t − 1, plus a random

component that is realized at t.

In the language of time series econometrics, the process in (8.15) is one char-
acterized by a random walk. This implies that the time pattern of consumption is
without any trend and that it does not fluctuate around any mean.10
It is also important to emphasize that at each time period t, individuals use
all available information about the future in forming their consumption decision
(see (8.14)). Indeed, if they did not, they would not have rational expectations.
Each ct∗ is thus a reflection of all available information about future incomes
and other relevant facts, and since individuals have a preference for consump-
tion smoothing, they have reason to believe that they will consume ct∗ throughout
their future lives.
How should we then think about the random component et ? By the logic of
rational expectations, it must be the case that et arises due to some new informa-
tion about our future prospects that was not available at t − 1. Random events that
change our view on our future economic situation indeed happen all the time and,
for instance, include serious injury by accident, unexpected new job opportuni-
ties, or lottery prizes. More formally, if we want to study changes in consumption
between periods 1 and 2 in the framework above so that c2 = c1∗ + e2 , then the
error term is given by

T
T
1
e2 = E 2 (yt ) − E 1 (yt ) (8.16)
T −1
t=2 t=2
In other words, if the error term is different from zero, expectations about the flow
of incomes over the discrete interval from t = 2 to t = T must have changed from
period 1 to period 2.11 Whether aggregate consumption actually follows a random
walk as proposed by Hall (1978) has been a source of numerous empirical studies
in macroeconomics.
8.6 Precautionary saving

A weakness with the models above is that there is no real motivation to “save
for a rainy day”. Even in the presence of uncertainty, as in Hall’s framework,
individuals behave as if the flow of future incomes was certain (which is referred
to as certainty equivalence above). Below we introduce an extension of the model
that helps us explain the prevalence of precautionary saving.
From a theoretical point of view, we know from the Euler equation above that
the condition for optimality is
u (ct ) = E t u (ct+1 ) =⇒
1 − act = E t (1 − act+1) =⇒ 1 − a E t (ct+1 ) = u [E t (ct+1 )]
A key technical aspect of this expression is that the expected marginal utility
of ct+1 equals the marginal utility of the expected value of ct+1 : E t u (ct+1 ) =
u [E t (ct+1 )]. This is a feature that follows from the quadratic utility function in
the random-walk model, and the exact meaning of this will be demonstrated below.
One problem with the quadratic utility function is, as mentioned above, that
it allows for a situation where u (ct ) < 0 at ct > 1/a. A negative marginal
utility of consumption at high levels of consumption is problematic in terms
of realism. Note also that in the quadratic case, u (ct ) = 0. If, however, we
assumed u (ct ) > 0, marginal utility would be a negative and convex function
of consumption, as illustrated in Figure 8.2.
If the only new assumption is that u (ct ) > 0 at all ct > 0, then E t u (ct+1 ) must
be higher than before (due to the convexity of u (ct )). At the old level of ct , we
then have that u [E t (ct+1 )] < E t u (ct+1 ), i.e. a violation of the Euler condition.
In order to restore the equilibrium in the Euler equation, there must be an increase
in u [E t (ct+1 )] = u (ct ). This can only come about through a decrease in ct .
A reduction in consumption at time t means an equivalent increase in that period’s
savings. Hence, the new assumption that u (ct ) > 0 leads to precautionary saving.
u ′(ct )
u ′′′(ct ) > 0
u ′′′(ct ) = 0
ct
1/a
Figure 8.2 Precautionary saving and the third derivative of the utility function.
u ′(c )
Et u′(c )
u ′(c )
c
c** c* c a
cL′ cL cH c′H
Figure 8.3 Precautionary saving with an increase in uncertainty.
A more intuitive explanation is given in Figure 8.3. There we assume that

consumption can assume either a high value c H with a probability of 1/2 or a
low value c L with a probability of 1/2. Hence, the expected level of consump-
tion is E t (c) = ca = (c H + c L )/2. However, at ca , we have u (ca ) < E t u (c) =
[u (c H ) + u (c L )]/2. Only at c∗ will we have that u (c∗ ) = E t u (c∗ ), which will
therefore be the chosen, lower level of consumption.
If uncertainty increases in the sense that c L decreases (to cL ) by exactly the
same amount as c H increases (to cH ), then the expected value ca is preserved.
However, the change in risk causes a drastic increase in E t u (c). This is indeed a
direct consequence of the assumption of u (ct ) > 0. In order to restore the Euler
equation, c must now decrease even further to c∗∗ < c∗ . Hence, the increase in risk
decreases the optimal level of consumption and increases savings. This is the key
insight from models of precautionary saving.
8.7 Interest rates and time discount rates

Let us now reintroduce interest rates and time discount rates into the model and
analyze the implications.
From the OLG model, we borrow the CRRA utility function with a finite time
horizon of T :

T
1 ct1−θ
U= (8.17)
(1 + ρ)t 1 − θ
t=0
As before, θ ∈ (0, 1) is the relative risk-aversion parameter and ρ ≥ 0 is the time
discount rate.
The individual’s intertemporal budget constraint in the multi-period case is

T
ct yt
T
= (8.18)
(1 + r )t (1 + r )t
t=0 t=0
where r is the interest rate.

Setting up the Lagrangian and solving for the first-order conditions in the
standard way, we will arrive at the same “discrete Ramsey result” for relative
consumption as in the OLG model:
t
1+r θ
ct∗ = c0∗ (8.19)
1+ρ
One of the more interesting aspects of this expression is that we are not likely to
have consumption smoothing any more. If, for instance, r > ρ (which should be the
case in dynamic, thrifty societies), then consumption will increase exponentially
over time and will not display a random walk.12 Figure 8.4 shows examples of the
optimal consumption paths for three different scenarios: r > ρ, r = ρ, and r < ρ.
As before, the optimal growth rate of consumption is given by
∗ − c∗ 1
ct+1 1+r θ
= gc∗ =
t
−1
ct∗ 1+ρ
c *t
r>ρ
r=ρ
c0
r<ρ
t
t=0 T
Figure 8.4 Optimal consumption paths over time for different levels of r and ρ.
8.8 Relative consumption
Recent research, inspired by experimental studies, has demonstrated that the indi-
vidual’s utility is determined not only by her own absolute level of consumption
but also by her relative level, compared with some reference point R. The refer-
ence point might, for instance, be the neighbor’s level of consumption – in which
case the individual will have so-called “keeping up with the Joneses” concerns
– or it could be derived from past levels of her own consumption, in which case
“habit formation” plays a role in utility.13
As a simple illustration of these ideas, consider the following utility function
for two periods with properties similar to that in Bowman et al (1999):
U = u(c1 ) + v(c1 − R1 ) + u(c2 ) + v(c2 − R2 ) (8.20)
The function u(ct ) is an instantaneous function of absolute consumption levels

as before, with the standard properties of u (ct ) > 0 and u (ct ) < 0. The v(x)
function (sometimes referred to as the gain–loss function), where x = ct − Rt ,
introduces reference dependence into the model. v(x) is assumed to have the
following properties:
• v (x) > 0 at all x;

• v(0) = 0;
• v (x) < 0 for all x > 0, and v (x) > 0 for all x < 0.
The first assumption means that the individual will always be better off the greater
her own level of consumption is compared to the reference point Rt . The second
assumption simply means that if consumption is at the reference point, utility from
that period will be equal to u(ct ). The third assumption is more far-reaching and
implies that the individual has loss aversion. This means that her gain–loss func-
tion is concave above the reference point and convex below. The deeper meaning
is that individuals are even more sensitive to losses than in the standard concave
utility setting, which is what defines loss aversion.
The reference point might further be endogenous to the decisions made in the
model. It is common to assume an exogenous reference point in the first period R1 .
R2 might, however, depend on c1 :
R2 = (1 − α)R1 + αc1 (8.21)
An α = 0 simply means that the reference point is constant in the two periods.
If α > 0, however, R2 is influenced by c1 . If c1 > R1 , then R2 > R1 and the
individual is subject to habit formation.
A key implication of the prevalence of habit formation is that the intertemporal
consumption choice is no longer time-independent, as in the PIH.14 The individ-
ual’s consumption decision for period 1 will affect utility in period 2 since a very
high level of consumption in period 1 will make the individual “addicted” to a
high level of consumption, which will affect how she values c2 .
In the case where α = 0, so that the reference point is some exogenous factor,
it is usually assumed that the individual will compare her own consumption with
that of some other person, perhaps the average consumption level in the country
or the consumption of neighbors or friends. The exact implications of such rela-
tive consumption comparisons depend on the characteristics of the u(ct ) and v(t)
functions. Research on the impact of reference-dependent utility is currently very
intensive, and it remains to be seen how it will change macroeconomic models of
aggregate consumption.
8.9 Time inconsistency

Another frequently demonstrated anomaly in consumer behavior is that people

appear to have time-inconsistent preferences in the sense that there is a utility
bias towards present consumption that goes beyond normal time discounting. The
model analyzed here follows Laibson (1997), while a general overview is provided
by Frederick et al (2002).
Let us assume as above that we have a standard concave utility function u(ct )
such that u (ct ) > 0 and u (ct ) < 0, and a standard time discount factor β ≤ 1.
Let us, however, make the new assumption that there is a bias towards current
consumption in the utility function such that all future time periods are discounted
by an additional factor γ ∈ (0, 1). In that case, the lifetime utility of an individual is
U = u(c0 ) + γβu(c2 ) + γβ 2 u(c3 ) + · · · (8.22)

T
= u(c0 ) + γ β t u(ct )
t=1
The lifetime budget constraint is still given by (8.4). By setting up the

Lagrangian in the standard way and rearranging the first-order conditions, we
find that the intertemporal marginal rate of substitution between any two periods
t > 0 is
u (ct )
= β(1 + r ) (8.23)
u (ct+1 )
whereas the same calculation for the the initial period 0 and period 1 gives us
u (c0 )
= γβ(1 + r ) (8.24)
u (c1 )
Note that the only difference from the standard result is the inclusion of γ < 1. In
this sense, preferences are clearly time-inconsistent.
If we also assume the usual CRRA utility such that u(ct ) = ct1−θ /(1 − θ ), then
1
we get the standard Ramsey result that ct+1 /ct = [β(1 + r )] θ for all t > 1 but that
c1
c0 = 1
(8.25)
[γβ(1 + r )] θ
Since γ < 1, this implies that the individual will consume relatively more in the
first period than if γ = 1, as in the time-consistent, standard case. In each period
of the individual’s remaining life, there will then be a bias present that will make
the individual consume relatively more in the current period. This will not be
sustainable, since the lifetime budget constraint is the same as before, and sharp
decreases in consumption typically become unavoidable as time approaches T .

9 Investment and Asset Markets
Investments are efforts aimed at increasing a stock of capital, usually physical cap-
ital.1 A country’s investments are closely associated with its savings, as shown in
the neoclassical growth model. In this chapter, we will present the standard neo-
classical model of investment, based on the profit maximization of an individual
firm, and derive central results such as Tobin’s marginal q and the user cost of
capital. We will also discuss the nature of adjustment costs and how their char-
acteristics affect firms’ investment decisions. The account of investment theory
outlined here relies on Branson (1989), Caballero (1997), and Romer (2005).
In Section 9.4 we will briefly consider the housing market and analyze the
determinants of housing demand and equilibrium price levels.
9.1 The Keynesian investment function

In the simplest Keynesian model, aggregate investment in physical capital is a
function of the aggregate level of income Yt and the interest rate rt , denoted
by It (Yt , rt ), where ∂ It ∂Y
(Yt ,rt )
> 0 and ∂ It (Y
∂rt
t ,rt ) < 0. The prediction that invest-
t
ment should increase with Y follows from an assumption of a long-run relation-
ship between the capital stock K t and Yt .2 Higher interest rates rt mean that it
is more expensive for producers to hold capital, which should decrease invest-
ment (just as a higher wage level will induce firms to reduce the number of
workers). The Keynesian investment function, however, is not based on any micro
foundations of firm behavior. This is what we will turn to next.
9.2 The firm’s investment decision

When considering the optimal level of investments, firm owners maximize the
present value of future profit streams:

T
t T
1
V (0) = = [Pt F(L t , K t ) − wt L t − PtI It ] (9.1)
(1 + r )t (1 + r )t
t=0 t=0
In this expression, t is total profits in the economy at time t, r is the interest rate,
Pt is the price (index) of firms’ total produced output Yt = F(L t , K t ) (which is a
Investment and asset markets 81
function of labor L t and physical capital K t ), wt is the wage rate, PtI is the price
of the investment good, and It is total gross investment. The production function
F(L t , K t ) is characterized by ∂ F(L t ,K t ) = F > 0 and F > 0. The lifetime of
∂ Lt Lt Kt
the firm is from 0 to T .
Capital accumulates according to
K t+1 = K t + It − δK t = K t (1 − δ) + It (9.2)
where It is gross investment (It = K t+1 − K t + δ K t , i.e. net investment K t+1 − K t

plus replacement investment δ K t ) and δ is the rate of capital depreciation.
The firm’s maximization problem is

T
t
max V (0) = s.t. K t+1 = K t (1 − δ) + It at each t = (0, 1, . . . , T )
L t ,K t ,It (1 + r )t
t=0
(9.3)
This defines a Lagrangian function

T
1
T
= [Pt F(L t , K t ) − wt L t − PtI It ] + λt [It + K t (1 − δ) − K t+1]
(1 + r ) t
t=0 t=0
(9.4)
The relevant first-order conditions for this problem are
∂ 1
= (Pt FL t − wt ) = 0 (9.5)
∂ L t (1 + r )t
∂ Pt f K t
= + λt (1 − δ) − λt−1 = 0 (9.6)
∂ K t (1 + r )t
∂ PtI
=− + λt = 0 (9.7)
∂ It (1 + r )t
Note that the λt−1 in (9.6) comes from the fact that K t also appears in the
restriction for period t − 1: λt−1 [It−1 + K t−1 (1 − δ) − K t ].
9.2.1 The user cost of capital

From (9.7), we find that
I
PtI Pt−1
λt = , implying that λt−1 =
(1 + r )t (1 + r )t−1
Substituting this result into (9.6) yields
I
Pt FK t PtI (1 − δ) Pt−1
+ − =0
(1 + r )t (1 + r )t (1 + r )t−1
Pt F
Kt
If we first isolate (1+r)t on the left-hand side and then multiply both sides by
(1 + r ) /Pt , we obtain
t
δ PtI + r Pt−1
I − (P I − P I )
t t−1 Ct
FK t = = (9.8)
Pt Pt
What we have derived is the real user cost of capital Ct /Pt . The user cost of
capital Ct contains three terms: the depreciation cost of investing in capital δ PtI ,
plus the interest cost of holding capital during t valued at t − 1, r Pt−1
I , minus the
possible increase in the price of the investment good (which an investor benefits
from), PtI − Pt−1
I . The nominal user cost C is divided by the price index P to
t t
yield the real user cost of capital.
The expression in (9.8) implies that firms should optimally make investments
up to the level where the marginal product of capital FK t equals the real user cost
Ct /Pt . This equilibrium level is implicitly given in (9.8) such that
K t∗ = K (Yt , Ct , Pt )
where Yt = F(L t , K t ) is total output in the economy.
9.2.2 A Cobb–Douglas example

In order to obtain a closed-form solution for K ∗ , let us use the conventional Cobb–
Douglas production function
Yt = AK tα L 1−α
t (9.9)
where A is some positive productivity parameter and 0 < α < 1 is the output elas-
ticity of capital. With this function, we have a marginal product of capital equal to
FK t = AαK tα−1 L 1−α
t = αYt /K t . Inserting this value back into (9.8), we get
αYt Ct
=
K t∗ Pt
which in turn implies that
α Pt Yt
K t∗ =
Ct
In other words, the optimal level of capital increases with the price level Pt and
with total output Yt and decreases with the (nominal) user cost Ct .
9.2.3 Tobin’s q
The expression for the real user cost of capital might be rearranged into
Pt FK t + PtI (1 − δ) − Pt−1
I
(1 + r ) = 0
I (1 + r ) on the right-hand side and then multiply both sides

If we first isolate Pt−1
I (1 + r )], we obtain
by 1/[Pt−1
1 Pt FK t + PtI (1 − δ)
=1 (9.10)
1+r I
Pt−1
The term on the left-hand side is referred to as Tobin’s marginal q after Nobel
Prize laureate James Tobin. It shows the change in the value of the firm at t of a
marginal increase in the capital stock at t − 1, Pt FK t + PtI (1 − δ), divided by the
cost of acquiring that marginal increase Pt−1I , and discounted back to period t − 1,
1/(1 + r ).
The increase in the value of the firm arises from an increase in revenue Pt FK t
and from an increase in the value of its capital PtI (1 − δ). Note that FK t decreases
with K t due to diminishing returns. In equilibrium, the discounted increased value
I . If the firm
of the firm should be equal to the cost of a marginal unit of capital Pt−1
is not in equilibrium so that Tobin’s q is greater (or smaller) than unity, then the
marginal benefit of a small increase (decrease) in the capital stock will exceed the
marginal cost. Thus the capital stock will be expanded until it reaches the optimal
level K ∗ where Tobin’s marginal q is one, as required.
9.3 Adjustment costs

The firm’s costs of investment come not only from the purchase of capital. It is
usually assumed that the main source of adjustment costs is the demands on inter-
nal reorganization of activities. For instance, new computers with new software
require both installation and training of workers, which are costs that go beyond
the purchase costs.
Let us define adjustment costs as A(It ) with the property that A (It ) > 0 so that
adjustment costs increase with the size of investment. If such costs are present, the
new profit function becomes

T
1
[Pt F(L t , K t ) − wt L t − PtI It − A(It )] (9.11)
(1 + r )t
t=0
and the new first-order condition for investment is
∂ −PtI − A (It )
= + λt = 0 (9.12)
∂ It (1 + r )t
while the other conditions are as before. This further means that
PtI + A (It )
λt = (9.13)
(1 + r )t
The expressions indicate that the introduction of adjustment costs will make
the marginal cost of investment higher than before. This in turn implies that
investments should generally be smaller than without adjustment costs.
9.3.1 Types of adjustment costs

The literature discusses three major types of adjustment costs. One key source
of adjustment costs is investment irreversibility. After having purchased a unit of

capital, it might have a very low second-hand value. Though increasing a capital
stock is easy, it is often much more difficult to decrease a capital stock. One exam-
ple is a factory, built so that it will exactly fit the production needs of the firm that
acquired it. Since it is unlikely that an identical firm exists nearby with identical
needs, the factory might be hard to sell should that be necessary.
A simple way to characterize costs of irreversibility is
A(It ) = (P I − P S )It (9.14)
where P I − P S is the difference between purchase and the potential sale prices. If
this difference is large, the firm will only buy the unit of capital if it is quite sure
that it is not going to need to sell it in the near future.
A second type of adjustment costs is one where there is a substantial fixed cost
component:
A(It ) = F + κ(It ) (9.15)
where F > 0 is a fixed cost and where adjustment also has a flexible component
κ (It ) > 0. A fixed cost could, for instance, be that a firm must make a large dis-
crete change in how its goods are produced after the investment. Irreversibilities
and fixed adjustment costs both imply that investments should appear in a dis-
continuous (lumpy) fashion, with long periods of inactivity ended by investment
“spikes”.
A third and quite different kind of adjustment costs are those that are biased
towards small, incremental changes:
a It2
A(It ) = (9.16)
2
In this expression, a > 0 is a parameter. We further assume that It can be negative

if the capital stock is reduced. Taking derivatives, we find that A (It ) = a I > 0
if I > 0 and that A (It ) = a I < 0 if I < 0. Hence, both positive and negative
investments (disinvestments) will incur a positive marginal cost. Furthermore,
A (It ) = a > 0, which means that costs are convex. Convexity in turn implies
that the costs of changing the capital stock will increase more than proportionally
with the size of the change. Hence, in this case, investors will be biased towards
incremental changes. The type of adjustment cost that the firm has will of course
depend on the structure of production and on the nature of the firm.
9.4 The housing market

The single most important investment decision that a typical household makes
over its lifetime is the choice of whether and when to buy a place to live. In
making this decision, the household members need to consider their optimal
mix of consumption of ordinary goods and the utility provided by owning a
house.
If we think of this as a static problem, let us denote the size of the housing
stock as H (perhaps reflecting the number of square meters), which is multi-
plied by p H (the price per unit of housing) in order to get the total value of
the household’s housing stock at a given time, p H H .3 If a house is bought, the
household borrows the whole amount. In each period of time, the household has
to pay r p H H in interest. For simplicity, we assume that no repayments need to
be made during the period under consideration. In addition to this, the household
needs to pay for repair and maintenance an amount δp H H , where δ > 0 is the
depreciation of the housing stock. The total cost of housing in one period is thus
(r + δ) p H H .
The household earns an income y during each period. This income is used for
consumption c and for housing such that
c + (r + δ) p H H ≤ y
For simplicity, there are there are no savings.

Let us further assume that the household gets utility from consumption and from
living in a house. The utility function is loglinear:
U = ln c + η ln H (9.17)
In this function, η > 0 shows the relative importance given to housing. If η > 1,
then the household gets a higher relative marginal utility from housing than from
consumption of other goods.
The household’s choice is the optimum amount of housing to hold. If we insert
the budget constraint into the utility function, we can write
U = ln[y − (r + δ) p H H ] + η ln H
Taking the first-order condition for a maximum, we obtain
∂U (r + δ) p H η
=− + =0
∂H Y − (r + δ) p H H H
This condition can be rearranged to solve for the household’s demand for housing:
ηy
HD= (9.18)
p H (r + δ)(1 + η)
In the short run, the supply of housing is fixed at H S . In equilibrium, demand

will equal supply, so that H S = H D . This implies in turn that the equilibrium price
of housing is
ηy
p H,∗ =
(r + δ)(1 + η)H S
Hardly surprisingly, the price will decrease with the short-run supply of
houses H S . Clearly, the greater the existing stock of houses, the lower the price.
A more interesting implication of the result above, and one that is constantly rel-
evant for millions of households, is the fact that the equilibrium price of houses
decreases with the interest rate r . An unexpected lowering of the interest rate, per-
haps due to a monetary policy intervention by the central bank, will lead to a shift
in the demand curve outwards and a temporary excess demand for housing. This
will eventually cause house prices to rise to a new and higher level ( p H,∗∗ ), as
illustrated in Figure 9.1.
pH
HS
p H,**
p H,*
H D,1
p H,*** HD
H D,2
HS
Figure 9.1 Equilibrium price of housing following falls in interest rates and in income.
Note also the impact of a fall in household income y, perhaps due to an
economy-wide recession. In the short run, prices will typically be slow to adjust
from the initial level p H,∗ and there will be an excess supply of houses that will
be empty. Just like with wages, house prices tend to be sticky downwards. A price
decrease means that household wealth for homeowners decreases and people will
be reluctant to sell houses at a lower price than they paid for them (or for a lower
amount than the household has used as collateral for loans for consumption, as in
the United States). Hence, downward adjustment will be slow. The price should
eventually settle at the new lower equilibrium level p H,∗∗∗ .
10 Unemployment and the
Labor Market
Having discussed capital accumulation, we now move on to discuss the other

major physical factor of production: labor. The market for capital and the market
for labor are remarkably different. Whereas capital markets are generally assumed
to be working fairly efficiently in developed countries so that demand usually
matches supply, this is not at all the case with labor markets. In fact, labor markets
are rarely characterized by a situation where labor demand equals labor supply.
Since Keynesian times, there has been a notion that different types of market
failures typically affect the labor market, as will be discussed below.
The three main topics treated in this chapter are the theory of efficiency wages,
associated with Shapiro and Stiglitz (1984), insider–outsider models of wage set-
ting, and the search and matching framework developed by Peter Diamond, Dale
Mortensen, and Christopher Pissarides.
10.1 Labor market disequilibrium

As shown in Chapter 6, both the demand and supply of labor will depend on
the prevailing wage rate. Let L D (w) be firms’ total demand for working hours (or
∂ L D (w)
workers), which is a function of the wage rate such that ∂w < 0.1 Equivalently,
∂ L S (w)
let total labor supply be L S (w) such that ∂w > 0. If labor were a “normal”
market, we would have a situation where L D (w∗ ) = L S (w∗ ) at an equilibrium
wage rate w∗ . The standard situation in countries around the world is, however,
that the prevailing wage rate is w̃ > w∗ , with the implication that L D (w̃) < L S (w̃),
as shown in Figure 10.1. This excess supply, equal to L S (w̃) − L D (w̃) = μ, is what
we normally call unemployment.
Why is it that wage rates are set at levels too high for an equilibrium to occur?
This is the issue that unemployment theory has been primarily concerned with. In
the Keynesian paradigm, wages were modeled as being “sticky” in the downward
direction but flexible upwards. Keynesian theorists tended to emphasize that the
labor market was very special since it provided a livelihood for the vast majority
of a country’s population.
In the sections below, we will discuss three major explanations advanced in the
literature: firms pay high wages in order to motivate workers and reduce shirking
Unemployment and the labor market 89
LS(w )
w̃
w*
LD(w)
labor
Figure 10.1 Unemployment.
(efficiency wage theory); firms want to cut wages but are prevented from doing
so by contracts and labor unions (insider–outsider theory); and unemployment
is largely due to worker heterogeneity and the substantial problem of allocating
workers to suitable jobs (search and matching models).
10.2 Efficiency wages

The key assumption in efficiency wage theory is that the wage rate affects indi-
vidual worker behavior. More specifically, it is assumed that a higher wage will
induce workers to work harder, or to exert more effort. There may be several rea-
sons why this is true. In developing countries, a higher wage means higher and
more stable food consumption, which in turn has a direct effect on effort. In more
advanced countries, a key link is that workers’ true effort is hard for firms’ man-
agers to monitor. If workers could easily get a similar job elsewhere for the same
wage, they would have an incentive to shirk on the job. A wage somewhat higher
than necessary would thus make shirking – and the associated risk of getting fired
– more costly for the individual worker. A third reason might be that higher wages
improve the quality of job applicants. A higher wage might also build loyalty
among workers and to the firm, whereas lower wages cause anger and shirking.
A very simple formulation of these ideas (which follows the exposition in
Romer 2005) is to assume a firm with a profit function
= F(eL) − wL (10.1)
where F(eL) is the firm’s production function, eL is “effective” labor, w is the

wage rate, and L is the number of people employed. Labor is the only factor of
production and the price of the goods produced is normalized to unity so that
F(eL) is the firm’s total revenue. The production function has the standard fea-
tures F (eL) > 0 and F (eL) < 0. The key assumption in the model, however, is
that effort reacts to the wage level according to a function e(w) where e (w) > 0.
Hence, the wage has both a negative and a positive impact on firm profits: on the
one hand, it constitutes a direct cost in terms of wage payments; on the other hand,
it increases net revenue due to a greater worker effort.
The firm’s maximization problem is to choose the level of the wage w
and the number of people employed L so as to maximize total profits
= F(e(w)L) − wL. The first-order conditions for maximum are
∂
= F (e(w)L)Le (w) − L = 0 (10.2)

∂w
∂
= F (e(w)L)e(w) − w = 0 (10.3)
∂L
w
From (10.3), we find that F (e(w)L) = e(w) . This might be interpreted as a modi-
fied version of the standard optimization criterion: that the marginal product of
w
labor F (e(w)L) should be equal to the (effort-adjusted) marginal cost e(w) .
Substituting this expression back into (10.2) and rearranging terms yields the
optimality condition
w̃
e (w̃) = 1 (10.4)
e(w̃)
The term on the left-hand side is the elasticity of effort with respect to the wage.
The wage at which this equality holds is w̃. Whether such a wage can be found
depends to a large extent on the character of the e(w) function. Figure 10.2 shows
one example where the efficiency wage equilibrium is w̃ > 0.
The firm’s labor demand is in turn implicitly given by the expression in (10.3).
10.3 The Shapiro–Stiglitz model

The model above gave an intuition for why firms might want to set wages at higher
levels than would ensure an equilibrium in total demand and supply. The richer
efficiency wage model below from Shapiro and Stiglitz (1984) allows us to analyze
the determinants of individual behavior in more detail. Let us imagine that workers
have an instantaneous utility function
U (w, e) = w − e (10.5)
where w is the wage received and e is the effort exerted. Utility is a negative function
of effort and a positive function of the wage. We further assume that individuals can
either exert e = 0 or some positive level e = ē > 0. Regardless of behavior, there
is a probability b per unit of time that the worker loses his or her job. In the case
e(w )
e(w)
w
~
w
Figure 10.2 Example of an efficiency wage equilibrium.
of shirking, there is an additional probability q > 0 of being fired. Note that q < 1,
which means that manager monitoring is imperfect; even in the case of shirking,
there is a probability 1 − q that the worker will manage to keep his or her job. If
the worker is unemployed, he or she will receive an unemployment benefit of w̄.
Workers are for simplicity assumed to maximize the expected present discounted
value of utility over an infinite life. The interest rate is r > 0.
The key choice for the worker is whether to shirk or not. The lifetime utility
from being an employed shirker is referred to as VES , whereas the lifetime utility
of an employed nonshirker is VEN . The so-called “fundamental asset equation” for
shirkers shows that the interest rate times the “asset” VES should be equal to the
“flow benefit” at a point in time (equal to the wage w) minus the expected change
in the value of the asset due to unemployment (b + q)(VES − VU ):
r VES = w − (b + q)(VES − VU ) (10.6)
The interpretation of this condition is the following. VES might be thought of as

an asset that can be held by a worker. In order for it to be held by anyone, it should
provide a rate of return during one unit of time equal to the market rate r . During
that same unit of time, an individual is paid a wage w and faces a probability of
b + q of becoming unemployed. In the event of unemployment, the worker loses
the difference between VES and the lifetime utility of being unemployed, VU . Since
VES > VU , shirking might thus entail a “cost” (b + q)(VES − VU ). In equilibrium,
(10.6) should be satisfied.
The lifetime utility of a nonshirking worker is VEN and the asset equation
equivalent to the one above is
r VEN = w − ē − b(VEN − VU ) (10.7)
In this case, the worker does not shirk and exerts an effort ē. Furthermore, the
probability of being fired for shirking q is zero.
From these two equations, one can solve for the levels of VES and VEN :
w + (b + q)VU
VES = (10.8)
r +b+q
w − ē + bVU
VEN =
r +b
A rational individual will not choose to shirk if VEN ≥ VES , a condition that
we will refer to as the “no-shirking condition” (NSC). If we use the expressions
in (10.8), a comparison between the two levels shows that
VEN ≥ VES =⇒
w − ē + bVU
w + (b + q)V
≥ =⇒
r +b+q r +b

q
w + (b + q)VU ≥ (w − ē + bVU ) 1 + =⇒
r +b

−wq r +b+q
≥ (bVU − ē) − (b + q)VU
r +b r +b
Multiplying both sides by −(r + b)/q yields

r +b+q (b + q)(r + b)VU
w ≥ (ē − bVU ) + (10.9)
q q
(ē − bVU )(r + b + q) + (b + q)(r + b)VU
=
q
ē(r + b + q)
= + r VU
q
Expression (10.9) shows the wage that needs to be paid for workers not to shirk.
All firms are aware of the NSC and hence set wages such that VEN = VES = VE so
that no shirking is carried out.
Given that VEN = VES = VE , we can equate the right-hand sides of (10.6) and
(10.7) and insert VE :
w − (b + q)(VE − VU ) = w − ē − b(VE − VU ) =⇒
ē
VE − VU =
q
We can also rewrite the nonshirking asset equation (10.7) as
w = ē + r VE + b(VE − VU ) = ē + r VU + (b + r )(VE − VU ) (10.10)
which will be useful later on.

What we still have not solved for is the lifetime utility of being unemployed, VU .
In an analogous manner as above, we can specify a fundamental asset equation for
an unemployed individual as
r VU = w̄ + a(VE − VU ) (10.11)
where a > 0 is the probability per unit of time that an unemployed worker finds a
new job. As mentioned above, w̄ > 0 is the level of unemployment benefits.
Inserting (10.11) and the result that VE − VU = qē into (10.10) gives us a closed-
form solution to the wage rate that satisfies the NSC:
ē
w = w̄ + ē + (a + b + r )
q
This critical wage increases with the required level of effort ē, with the interest
rate r , with unemployment benefits w̄, with the probability of finding a new job a,
and with the probability of becoming unemployed b, and decreases with the prob-
ability that a worker who shirks is caught. The term w̄ might be thought of as an
indicator of the generosity of unemployment benefits. If unemployment benefits
are high, workers have less to lose from being unemployed, and are therefore more
prone to shirk. Hence, firms must pay higher wages. Likewise, if it is easy for an
unemployed worker to find a new job, the expected cost of being unemployed is
small, the worker is more willing to shirk, and a higher wage has to be paid.
What will the level of employment be? In a steady-state equilibrium, the number
of people who become unemployed will be equal to the number of people who
find new jobs. The number of workers who become unemployed at any given time
is Lb, where L is the aggregate number of employed workers. The number of
unemployed workers finding new jobs at any given time is a(N − L). By setting
these equal to one another and solving for a, we get a = NLb
−L . This in turn implies
N −L
that a + b = N −L . Further, N = μ is the unemployment rate in the economy,
Nb
so the final NSC will be

b ē
w = w̄ + e + +r ≡ w̃
μ q
All this is still on the supply side of labor. On the demand side, we will look
at a representative individual firm, just as in the RBC model. Let us imagine
that the firm has a production function Q = F(L) so that the profit function is
= F(L) − wL. Let us further assume that F (N) > ē, i.e. if all potential work-
ers were employed, they would still have a marginal product above the required
effort level. Full employment would therefore be a good idea from the firm’s point
of view.
The usual profit maximization would give us the first-order condition
F (L) = w, as in the RBC model. In equilibrium the marginal product of labor
must therefore be equal to the NSC, i.e. F (L) = e + ( μb + r ) qe . This defines the
equilibrium level of employment in the labor market. The higher the wage required
to keep workers from shirking, the greater the level of unemployment. Workers
who are unemployed would be willing to work for a wage below w̃, but cannot
credibly commit to not shirking at that wage and hence are not employed.
10.4 Insider–outsider models

In Europe, it is commonly believed that union power could be an important reason
for the higher levels of unemployment than in the United States. The intuition for
understanding how unions affect unemployment is the assumption that unions care
only about their own members, the insiders, who are already employed, and will
aim for wage levels that are beyond a full employment equilibrium. The theory of
insiders and outsiders in the labor market was pioneered by Lindbeck and Snower
(1986).
To see this more formally, let us assume a representative firm with profits
given by
= AF(L I + L O ) − w I L I − w O L O
where A is some random productivity level to be defined below, L I is employment

of insiders, L O is employment of outsiders, and w I and w O are their respective
wage levels. The marginal products of insiders and outsiders are identical, FL I =
FL O > 0, so that insider and outsider labor are perfect substitutes in production.
We further assume that insider power manifests itself partly in the situation that
L I = L̄ I , meaning that insiders are always employed for certain.
A is a random shift parameter, assuming value Ai with probability pi . For sim-
plicity, let us assume that Ai = {A G , A B }, where the good outcome is A G > A B ,
which happens with probability p G = p B = 1/2. Insiders only receive utility from
their own wage level: u I = U (w I ), where U (w I ) > 0.
Union bargaining power appears in two ways. Firstly, they are able to determine
a linkage between their own wages and those that any outsider would get, w O =
Rw I where R ∈ (0, 1). Unions are strong when R is high (or even equal to unity)
since outsiders then cannot offer to work at a lower wage in order to enter the labor
market. Secondly, insiders have a reservation utility u 0 below which they will go
on strike. What is left for the firm to choose is therefore how many outsiders to
employ, L O,i , and the level of insider wages, w I,i , at each state of the economy
i = G, B.2
The maximization problem for the firm is

max pi [ Ai F(L I + L O,i ) − w I,i , L I − Rw I,i L O,i ],
L O, i ,W I, i
i=G,B

subject to pi U (w I,i ) ≥ u 0
i=G,B
The associated Lagrangian function is therefore

= pi [ Ai F(L I + L O,i ) − w I,i , L I − Rw I,i L O,i ]
i=G,B

+λ pi U (w I,i ) − u 0
i=G,B
By taking the first-order conditions with respect to L O,i , W I,i , we obtain
∂ 1
= [ Ai F (L I + L O,i ) − Rw I,i ] = 0 (10.12)
∂ L O,i 2
∂ 1
= [−(L I + RL O,i ) + λU (w I,i )] = 0 (10.13)
∂w I,i 2
From (10.12), we find that the optimal level of outsider employment should sat-
isfy Ai F (L I + L ∗O,i ) = Rw I,i , which is the standard result that marginal product
should equal marginal cost. Comparing the good and the bad business cycle out-
comes, we see that the left-hand side will be larger when Ai = A G . Hence, in order
to restore the first-order condition, F (L I + L ∗O,i ) must fall, which it can only do
if L ∗O,G > L ∗O,B . Thus, firms will employ more outsiders in good times, which
makes sense. Importantly, an increase in union power such that R increases must
be balanced by an increase in F (L I + L ∗O,i ), which it can only do if L ∗O,i falls. In
this highly stylized sense, strong labor unions therefore increase unemployment.
From (10.13), we have another interesting implication. The first-order condi-
tion implies that L I + RL O,i = λU (w∗I,i ) at the optimal insider wage rate w∗I,i .
Recall that if A G prevails, then L ∗O,G will be relatively high, which in turn means
that w∗I,G < w∗I,B (since marginal utility decreases with w∗I,i ). Thus, insider wages
are countercyclical and will be higher in bad times. The intuition is that firms
and insiders reach this decision jointly in order to keep outsiders from getting
employed.
10.5 Search and matching models

Search and matching models, finally, focus on the fact that, unlike most other
factors of production, workers are extremely heterogeneous. Matching the right
person to the right job is indeed a complicated affair that frequently fails and could
be a source of unemployment, according to theory.3 Search and matching models
were pioneered in works such as Diamond (1982) and Mortensen and Pissarides
(1994). The account below relies on Acemoglu and Autor (2009).
Let us assume that new matches on the labor market M are created through a
“matching function”
M(U, V ) = ηU β V 1−β
where η > 0 is an efficiency parameter for new job creation, U = μN > M is the
number of unemployed people (the unemployment rate μ times the total size of
the labor force L), and V = v N is the number of vacancies in the economy (v is the
vacancy rate). The number of employed people in the economy is L = N − μN.
It is commonly assumed that the matching function displays constant returns to
scale, i.e. it can be described as Cobb–Douglas. Note that a marginal increase in
the number of unemployed will only increase the number of newly created jobs
by β M/U < 1.
If we divide M through by the total labor force N, we can rewrite the
expression as
M(U, V )
m(μ, v) = = ημβ v 1−β

N
The function m(μ, v) < 1 should thus be thought of as the matching rate in the
labor force as a whole, which is a function of the unemployment rate μ and the
vacancy rate v.
Let us assume that we can describe the “tightness” of the labor market as
θ = v/μ. θ relates the vacancy rate to the unemployment rate. If θ is low, this
means that there are relatively few vacancies for every person unemployed. A low
θ might thus be thought of as indicating a tight labor market where it is quite
difficult to find a job.
Let us further define the (Poisson) arrival rate of a successful matching per
vacancy as
β
m(μ, v) μ
q(θ ) = =η = ηθ −β
v v
where q (θ ) = −ηβθ −β−1 < 0. A Poisson process means that in some given short
time interval t, the probability that one vacancy will be filled with one matched
worker is t · q(θ ). This probability thus increases with the unemployment rate μ
and with the length of the time interval t.
The Poisson arrival rate of a match for an unemployed worker is similarly
m(μ, v)
= ηθ 1−β = q(θ )θ (10.14)
μ
This function has some interesting properties. Let us assume that the time interval
under consideration is t = 1. Then (10.14) shows the probability that an unem-
ployed worker will find a job during one time unit. Consider now a small increase
in the unemployment rate μ. In that case, the matching rate per vacancy q(θ )
will increase. However, since there are also more unemployed people around to
compete for the vacancies, the labor market has tightened and the net effect on the
probability of finding a job is negative: ∂(m(μ, v)/μ)∂μ = −η(1 − β)θ 1−β /μ < 0.
As in the Shapiro–Stiglitz model, the total number of jobs lost during a certain
time interval is bL = b(N − μN), where b shows the fraction of the employed who
lose their jobs. b might be thought of as the probability that an individual worker
succumbs to an exogenous shock during a short time interval t that makes her
lose her job.
In equilibrium, it should be the case that the flow of people into employment
is constant, L̇ = ηU β V 1−β − bL = 0. This also means that the number of jobs
created through matching (ηU β V 1−β ) should be equal to the number of jobs
destroyed (bL). Dividing through by N, we obtain the equilibrium condition
ημ∗,β v 1−β = q(θ )θ μ∗ = b(1 − μ∗) (10.15)
The endogenous variable in this equation is the unemployment rate μ∗ . The

unemployment rate that satisfies (10.15) is also a steady-state level. To see why,
see Figure 10.3, which depicts job creation and job destruction as a function of μ.
The m(μ, v) curve is positive and concave in μ whereas the job destruction curve
is negative and linear.
If we start from a level μ > μ∗ , then we see that more jobs are created through
the matching process than are destroyed. This means that the unemployment rate μ
should decrease. The situation is of course the reverse if μ < μ∗ . Only at the
steady-state level μ = μ∗ will the system be in equilibrium.
The schedule in Figure 10.3 can also be used for analyzing various comparative
statics. Consider, for instance, an increase in the efficiency of matching η, perhaps
as a result of a new government unemployment program. An increase in η will
shift the m(μ, v) curve upwards at any given level of μ. The new steady-state
level is now μ∗∗ < μ∗ . An increase in v would have a similar effect, whereas an
increase in b would imply an increase in μ∗ .
mnew(μ,v)
b
m(μ,v )
b(1– μ)
μ∗∗ μ∗ μ=1 μ
Figure 10.3 Equilibrium in the search and matching model.

The equilibrium condition in (10.15) can be used to draw the so-called
Beveridge curve, showing the equilibrium relationship between the number of
people unemployed U and the number of existing vacancies V (sometimes
referred to as the U–V curve). Formally, the slope of the curve can be found by
taking implicit derivatives of a function B = ημ∗,β v 1−β − b(1 − μ∗ ) = 0 created
from the equilibrium condition in (10.15):
∂B
∂μ∗ −(1 − β)ημ∗,β v −β −(1 − β)η
= − ∂∂vB = ∗,β−1 v 1−β + b
= <0
∂v ∂μ
ηβμ θ (ηβθ 1−β + b)
A negative relationship is certainly well in line with intuition, since more vacan-
cies should imply a lower rate of unemployment. This can also be deduced from
Figure 10.3, where a higher level of vacancies implies a higher level of match-
ing and a lower equilibrium unemployment rate. The negative association is well
documented empirically for many countries in the world.
Part III
Macroeconomic Policy
11 IS–MP, Aggregate Demand, and
Aggregate Supply
In this third part of the book, we will now shift our focus in order to analyze the
effects of macroeconomic policy. Most of this chapter will be based on the IS–MP
model of the goods and money markets. This model is not micro-founded since it
is not based on optimizing household behavior. Instead, it follows in the Keynesian
tradition of assuming certain behaviors of variables at the macro level. Only the
specifications of aggregate supply will rely on micro foundations. The analysis in
this chapter is therefore quite different from the analysis in most other chapters.
We start off with the traditional Keynesian framework where we discuss aggre-
gate expenditure and multipliers. We then derive the aggregate demand function
from equilibria in the goods and money markets. We also elaborate on the prope-
rties of the aggregate supply function under varying assumptions of price and
wage stability and provide an overview of the Lucas critique of traditional Keyne-
sian economic policy. After that, we present a new model that introduces financial
intermediation into the standard IS–MP framework. Finally, we also present some
of the main ideas in the so-called new Keynesian paradigm.
11.1 Aggregate expenditure and the multiplier

The traditional Keynesian model focuses to a great extent on aggregate demand.
The typical starting point is an equation describing the user side of the eco-
nomy. Let total output be Y , then total expenditure in a closed economy model
(we assume away exports X and imports M) is E t = Ct + It + G t . In equilib-
rium, we should have that total expenditure equals total output so that E t = Yt .
As noted in Chapter 8, the Keynesian consumption function is often described
as Ct = ca + cmpc Ytd = ca + cmpc (1 − τ )Yt , where ca > 0 is referred to as the
autonomous part of consumption, which is independent of income, cmpc ∈ (0, 1) is
the marginal propensity to consume (MPC), τ is the (percentage) income tax rate,
and Y d is disposable income, net of taxes.
The second component on the user side, gross investment It , is given by a func-
tion It = i 0 − i 1r . i 0 is simply the intercept of this function when the real interest
rate r is zero. The main idea is that investment decreases with the real interest
rate.1 Government expenditure G t is assumed to be independent of income. Taken
102 Macroeconomic policy
together, this means that we can write
Yt = E t = ca + cmpc (1 − τ )Yt + It + G t
If we isolate income on the left-hand side and rearrange, we end up with
1
Yt = (ca + It + G t )
1 − cmpc (1 − τ )
This well-known expression has two parts; autonomous spending (ca + It + G t )

and the multiplier 1/[1 − cmpc (1 − τ )] > 1. It shows how the equilibrium level
of income (and output) will change with a change in autonomous spending.

Of particular interest are the effects of an increase in government spending
G t . A one-euro increase in G t will according to this model increase output by
1/[1 − cmpc (1 − τ )] > 1 euro. If we assume some concrete numbers, for instance
that cmpc = 0.5 and that τ = 0.2, then the multiplier is 1/[1 − 0.5(1 − 0.2)] = 5/3,
i.e. an increase in G t by 1000 euros would increase total output by 1667 euros.
Clearly, the higher is cmpc and the lower is τ , the greater is the multiplier.
This multiplier effect is illustrated in Figure 11.1. The figure shows the classi-
cal Keynesian cross, where an equilibrium initially exists at Y ∗ . At this level, total
aggregate expenditure equals total output (given by the 45◦ line). An increase
in autonomous government spending from G t to G t shifts the intercept and
the curve upwards, keeping the slope intact. Total expenditure increases, which
implies that total output and incomes must increase in order to restore equi-
librium. When income increases, consumption rises too via the MPC, which
increases expenditure even further. A new equilibrium is eventually found at an
total expenditure, E
An increase in G shifts the E curve
ca + It + Gt’ slope: 1/[1 – cmpc(1 – τ)]
ca + It + Gt
45°
Y
Y* Y **
Figure 11.1 The Keynesian cross and the multiplier impact of an increase in government
spending.
IS–MP, aggregate demand, and aggregate supply 103
output levelY ∗∗ > Y ∗ .
Note also that the multiplier effect implies that G t − G t <
∗∗ ∗
Y − Y , i.e. equilibrium output will increase by a larger amount that the initial
increase in G t .
11.2 The IS–MP model

In this section, we will analyze short-run economic policy in the traditional Keyne-
sian manner by assuming that prices are completely fixed. Rather than showing the
more common IS-LM model, we make the modification here, as in Romer (2005),
that we treat monetary policy as endogenous. Hence, “MP” replaces “LM” in the
title, although most qualitative insights remain the same as in older models.
11.2.1 The goods market

The demand for goods is described by the well-known IS curve. It shows the
combinations of output and the real interest rate where total output equals total
expenditure, as discussed in the previous section. Real output is written as
Y = E (Y, r, G)
where we now use a general expenditure function E (Y, r, G). E increases with
output Y (through the consumption function) and with real government pur-
chases G and decreases with the real interest rate r (for instance, since investments
decrease with r ). Should Y exceed total expenditures, actual production will be
unused and inventories will accumulate.
We describe the partial derivatives with the following notation:
∂ E(·) ∂ E(·) ∂ E(·)

= E Y > 0, = Er < 0, = EG > 0 (11.1)
∂Y ∂r ∂G
We will assume that a one-dollar increase in disposable income increases planned
expenditures by less than one dollar (as with the MPC) so that E Y < 1.
As was mentioned above, the IS curve shows the combinations of Y and r where
Y = E(Y, r, G) applies. The slope of the curve can be found by differentiating with
respect to r on both sides of the equality:

dY
dY
= EY +Er
dr
IS dr
IS
which after a bit of rearranging gives us

dY
Er
=
dr
I S 1 − E Y
Hence, the slope will be steeper the greater the sensitivities of planned expendi-
tures to the real interest rate r and to changes in incomes Y are.
11.2.2 The money market
The money market means the market for the supply and demand of what is
called “high-powered money”, including currency (coins and notes) and reserves
(on highly liquid bank accounts). The supply side of “real balances” is simply
the total nominal stock of money, M, divided by the aggregate price level in
society, P, measured for instance by a consumer price index.
The demand for real balances is determined by the nominal interest rate i , which
we will here express as the real interest rate r plus expected inflation π e so that
i = r + π e , and by the real level of income Y :
M
= L(r + π e , Y ), L r+π e < 0, LY > 0 (11.2)
The demand for money decreases with the nominal interest rate since the oppor-
tunity cost of holding money increases with the nominal interest rate that people
would receive if they made more investments that are long-term.2 In order to find
the slope of the MP curve in (Y, r ) space, consider an increase in r . If the equi-
librium in (11.2) is to remain in place, the decrease in money demand must be
balanced by an increase in Y . Hence, the MP curve will have a positive slope.
The key monetary policy variable is the real interest rate r , which is set by a
central bank. This interest rate (often referred to as the “repo rate”) is perfectly
transmitted to the rest of the economy. In setting its interest rate, the central bank
takes into consideration the output level Y as well as actual inflation π.3 The
behavioral rule is that central banks increase the real interest rate when output Y
increases or when inflation rises. Hence, r (Y, π) and the derivatives are rY > 0 and
rπ > 0.
Reformulating (11.2) gives us an expression for the level of the nominal money
supply, which will be endogenous to monetary policy:
M = PL(r (Y, π) + π e , Y )
Note the distinction that monetary policy reacts to actual inflation levels π whereas
the public, in this setting, adjusts its money demand through changes in expected
inflation π e .
11.2.3 Variations in IS–MP

The two curves can be drawn in the same diagram in (r, Y ) space, as in Figure 11.2.
As we have already derived, the IS curve is negatively sloped whereas the MP
curve has a positive slope. Equilibrium in both markets occurs where the two
curves cross. In standard Keynesian analysis, the effects of exogenous changes in
policy variables are analyzed in this diagram.
As an example, we might consider an increase in government expenditure G.
Such a change increases total expenditure, which means that the IS curve must
shift outwards from the origin, as in Figure 11.2. For a given r , the level of output
r
A higher inflation shifts the MP curve
MP
An increase in G shifts the IS curve
IS IS′
Figure 11.2 The IS–MP curves.
that satisfies an equilibrium in the goods market must be higher than before. The
MP curve remains unaffected since government expenditures are assumed not to
affect the aggregate price level in the short run. The outward shift in the IS curve
implies a higher level of equilibrium output but also a higher interest rate, r .
11.3 Aggregate demand

The aggregate demand (AD) curve shows all the combinations of inflation levels
and output where equilibria in the money and goods markets prevail simultane-
ously.4 Since the AD curve is drawn with the level of inflation on the vertical axis
and output on the horizontal axis, we now assume that prices are not completely
fixed.
In order to see how the AD curve is derived from the IS–MP curves, consider an
increase in the inflation rate, for instance due to a international oil shock. A rise in
the inflation rate forces the central bank to increase the real interest rate. A higher
r in turn increases the nominal interest rate, which reduces money demand. In the
IS–MP diagram in Figure 11.2, a higher r at the same level of Y means that the
MP curve shifts leftward and the two curves now cross again at a lower level of Y .
The IS curve remains unaltered. Hence, a rise in π is associated with a fall in Y .
This is the logic behind the negative slope of the AD curve in Figure 11.3.
To be able to say something more definite about the slope of the AD curve, we
can start by differentiating the two equations behind the IS and MP curves with
respect to inflation:

dY
dY
dr
= EY + Er
dπ
AD dπ
AD dπ
AD

dr
dY

= rπ + rY
dπ AD dπ
AD
AS
AD
Figure 11.3 The AD–AS curves with sticky wages.
Inserting the second equation into the first gives us

dY
dY
dY
= EY + Er rπ + rY
dπ
AD dπ
AD dπ
AD

dY
=⇒ (1 − E Y − Er rY ) = Er rπ
dπ
AD

dY
E r rπ rπ

= = 1−E
dπ AD 1 − E Y − Er rY
Er − rY
Y
Since the numerator is positive while the denominator is negative, the expression
as a whole will be negative, as shown in Figure 11.3.
Other changes in the components of the IS and MP curves will also have an
impact on the AD curve. We discussed previously the effect of an increase in gove-
rnment expenditures G. The outward shift in the IS curve to a new equilibrium
level of Y is associated with a simultaneous shift rightwards of the AD curve,
since output increases at a given level of inflation.
11.4 Aggregate supply

11.4.1 Keynesian formulations with sticky wages
In the original Keynesian formulations, aggregate supply was primarily derived
from the labor market. We saw in the RBC model how labor supply and demand
determined the actual quantity of people employed in the economy as well as the
output produced.
Let us imagine a representative firm whose only factor of production, for
simplicity, is assumed to be labor. Its profit function is then
= P F(L) − wL
where P is the aggregate price level (equal to the firm’s own price level), F(L)
is the production function with the usual properties F (L) > 0 and F (L) < 0,
and w is the wage rate, which is fixed in the short run. The typical competitive
firm will then hire up to the point when marginal revenue equals marginal cost:
P F (L) = w. Rewriting this equilibrium condition gives
w
F (L) =
P
To see the link to the aggregate supply (AS) curve, consider an increase in
inflation π. Such an increase will involve a rise in P, which in turn means that the
real wage falls (since the nominal wage w remains constant). In order to preserve
equilibrium, the marginal product on the left-hand side must also fall, which it can
do if L rises to some L 1 > L. If firms hire more labor, they can further produce
more output so that F(L 1 ) > F(L). Hence, there emerges a positive relationship
between inflation and output. This is the reason for the positive slope of the AS
curve in Figure 11.3.
11.4.2 The Phillips curve

Macroeconomic policy from the Second World War until the early 1970s was
partly founded on the notion that there existed a permanent trade-off between
unemployment and inflation for policy-makers to exploit. In order to derive such a
relationship explicitly, let us use the same model as in the section above and make
the additional assumption that the wage rate at time t is proportional to the price
level from last year, Pt−1 . More specifically, the relationship is
wt = γ Pt−1 (11.3)
where γ > 0 is a constant describing the strength of the relationship. The

equilibrium condition for labor demand is then
wt γ Pt−1
F (L t ) = =
Pt Pt
Pt −Pt−1 Pt Pt−1
Since πt−1 = Pt−1 = Pt−1 − 1, we know that Pt = 1
1+πt−1 . Thus, we can
write
γ
F ( L̃ t ) = (11.4)
1 + πt−1
As mentioned above, an increase in inflation therefore increases the number
of people employed in the next period. If the total labor force is Nt , then unem-
ployment is Nt − L̃ t = μN. A rise in inflation in this way decreases the level
of unemployment. This stable negative relationship between unemployment and
inflation is famously referred to as the Phillips curve (Phillips 1958). The exis-
tence of exploitable Phillips curves has been a field of intense empirical research
in recent decades.
11.4.3 The natural rate of unemployment

By the 1970s, it was fairly clear that a permanently exploitable Phillips curve
did not appear to be in place. Friedman (1968) and Phelps (1968) argued, for
instance, that a shift in policy towards a higher level of inflation with the aim of
keeping unemployment low could not in the longer run keep wages from increas-
ing. Rational workers would see through the Phillips curve reasoning and would
not accept persistent decreases in their real wage. In terms of our model, this
means that it is not likely that workers would accept a wage-setting rule according
to (11.3).
The implication of this type of reasoning is that in the longer run, there should
be a normal or natural rate of unemployment – sometimes referred to as the
nonaccelerating inflation rate of unemployment (NAIRU) – from which mone-
tary or fiscal policy could not diverge. Short-run deviations might be possible,
as indicated by the short-run Phillips curve (SRPC) in Figure 11.4, but in the
longer run there should not be an exploitable relationship between inflation and
unemployment.
π
LRPC
π*
SRPC
unemployment
NAIRU
Figure 11.4 Short- and long-run Phillips curves.

The level of output associated with the natural rate of unemployment is Ȳ , often
referred to as long-run aggregate supply or the natural rate of output. As with the
LRPC, the long-run aggregate supply curve is vertical.
11.4.4 Expectations-augmented Phillips curves

In this section we discuss Keynesian reformulations of the Phillips curve based
on later empirical evidence. It should be noted that the equations in this section
are not micro-founded but should be seen as “reduced-form” equations that are
primarily designed to be tested empirically.
If we incorporate the new assumptions about a natural rate of output, we might
express an expectations-augmented Phillips curve (drawn in (π, Y ) space this

time) in the following manner:
πt = πt∗ + λ(ln Yt − ln Ȳt ) + t (11.5)
Actual inflation πt is thus a function of the “underlying” or “core” inflation πt∗

and of the log deviation between actual output Yt and the natural rate of output Ȳt .
There is also a random error term t to incorporate the possibility of supply shocks.
λ > 0 describes how strongly inflation reacts to output deviations from the long-
run level. Note that if output is at its long-run level, the level of inflation will be
πt∗ (in the absence of random shocks).
How is πt∗ determined according to this model? In the simplest case, it is just
πt∗ = πt−1 . In this situation, it is implicitly assumed that individuals only react
to changes in monetary policy with a time lag. The assumption also implies that
there is a trade-off between changes in inflation rates and output (since we can
write πt − πt∗ = λ(ln Yt − ln Ȳt ) + t ).
If individuals are fully rational and forward-looking, however, a more proper
formulation would be πt∗ = E t−1 (πt ), where E t−1 (πt ) is the expected inflation
one period back. In this setting, individuals would make their decision based on
all available information about current and past policy choices when they make an
expectation about inflation.
Sometimes the two differing views above are combined by assuming that
πt∗ = φ E t−1 (πt ) + (1 − φ)πt−1 , where φ ∈ [0, 1] is a parameter. φ < 1 implies
that although people form rational expectations about the inflation rate, there is
some degree of inertia, so that there is a connection between core inflation and
past levels which goes beyond rational expectations. The source of such an inertia
is often left unexplained.
11.4.5 Lucas supply curve

The models above simply assumed that there was some form of price or wage
stickiness, in the literature often referred to as nominal rigidities. The exact
source of such nominal rigidities is usually not discussed in the traditional macro
literature.
Lucas (1972, 1973) developed a very influential model in which producers
observe the price of their own good Pi but cannot observe the overall aggregate
price level P. If the relative price Pi /P increases, for instance, the producer can-
not therefore know if it is caused by a change in Pi , in which case a rational
producer should increase production, or a fall in the aggregate price level, which
should leave production unaffected. Due to this imperfect information scenario,
the producer in Lucas’s model will choose to increase output at least a little to stay
on the safe side. Hence, there is also here a positive relationship between inflation
and output.5
However, the key result from the model (which we do not derive here) is a
micro-founded reformulation of the expectations-augmented Phillips curve:
ln Y − ln Ȳ = b[ p − E( p)]
In this expression, ln Y is the log of actual output, ln Ȳ is the log of long-run

aggregate supply, p reflects the (log of) the aggregate price level, E( p) is the
expectation of the aggregate price level, and b > 0 shows the sensitivity of output
to deviations from the expected level of inflation. The key insight is once again
that output will only deviate from the natural level if there is a surprise in the level
of inflation. If, for instance, people expect monetary policy to be expansionary
and their expectations are fulfilled, then p − E( p) = 0 and output remains at the
natural rate.
This critique of the traditional Keynesian models has important implications
for policy. In a very similar manner to the PIH model where individuals formed
rational expectations about future incomes, the fact that people have rational
expectations here will make economic policy based on output–inflation trade-offs
very difficult. Nonsurprising monetary policy measures might even be harmful
since they will increase inflation without affecting output or unemployment. In
general, people will anticipate policy and adjust accordingly in advance. This gen-
eral critique of how the lack of micro foundations in Keynesian analysis tends to
make policy conclusions based on historical data problematic is often referred to
as the Lucas critique.
11.5 Financial intermediation

In the IS–MP models discussed above, financial intermediaries such as banks do
not play any role. Still, a lesson from the financial crisis of 2007–08 was that the
functioning of the financial system might have important consequences for the
real economy. In this section we present an extension of the IS–MP framework,
proposed by Woodford (2010), that introduces financial intermediation into the
standard model.
Let us assume that inflation expectations are now given throughout. For sim-
plicity, we set π e = 0 so that the nominal interest rate equals the real interest rate,
i = r.
LS
LSnew
i new
LD
LDnew
lending
Figure 11.5 Impact of an increase in income on household demand and supply of

lending.
As above, the IS curve is defined as the combination of the real interest rate
r (= i ) and total income Y , where total income equals total expenditure E(Y, r ),
but let us now assume that we have no government. Total expenditure will thus be
spent on consumption and investment, and savings should equal investment.
Unlike before, we derive the slope of the IS curve through the supply and
demand of household lending. In Figure 11.5, we have drawn household demand
for lending LD as a negative function of the nominal interest rate i . Similarly,
households’ supply of lending (i.e. savings) increases with the nominal inter-
est rate. Without any credit market frictions, supply will equal demand at the
rate i .6
If there is an exogenous increase in income Y , how will this affect the supply
and demand for lending? Firstly, if all households have somewhat more income,
they should be able to increase lending supply. The LS curve therefore shifts down
to the right. The demand for loans should fall as incomes increase, but by a smaller
amount. Hence, the LD curve shifts down to the left. The result is a lower equi-
librium interest rate i new < i and a higher level of lending than before. We have
thereby shown that there must be a negative relationship between Y and i for the
goods market to be in equilibrium.
However, the analysis on the basis of Figure 11.5 assumes no credit market fric-
tions, and there is no need for financial intermediaries since the supply of credit
always meets demand. Let us now introduce a situation where financial interme-
diaries are necessary for effectively channeling credit (and perhaps for pooling
risks, as shown in Chapter 7). In order to finance their activities, the financial
intermediaries (we henceforth refer to them as banks) set a higher interest rate for
borrowing i b than for lending i s , as in Figure 11.6a. The difference is referred to
as the credit spread ω ≥ 0.
(a) i
LS
ib
credit spread ω1
is
LD
lending
L1
(b) ω
XSnew
XS
ω2
ω1
XD
lending
L2 L1
Figure 11.6 (a) Introducing financial intermediaries and credit supply frictions.
(b) Supply and demand for financial intermediation as a function of the
credit spread.
The supply and demand for financial intermediation is shown in Figure 11.6b
and is a function of the credit spread ω. Clearly, if the credit spread is large, the
interest rate on savings will be low and the demand for lending will also be rel-
atively low. The reverse holds if the credit spread is low. On the other hand, a
large credit spread should increase the number of firms willing to act as banks.
The supply of financial intermediation is a positive function of ω since it is more
profitable to be a bank when the difference between savers’ and borrowers’ inter-
est rates are large. For a given level of ω, financial innovations or productivity
increases would shift the XS schedule to the right whereas, for instance, tighter
banking regulations such as a higher capital requirement (the bank’s equity as a
share of total assets including its lending) would shift the XS curve to the left. The
equilibrium credit spread is initially given by ω1 when the two curves intersect.
The level of lending is then L 1 .
Let us now consider the effects of a sudden reassessment of the total value of
the banking system’s assets, i.e. its lending to businesses and individuals.7 A nat-
ural requirement for banks when risks increase would be to try increase the credit
spread ω. For a given level of lending, this would be equivalent to shifting the XS
curve to the left, as in Figure 11.6a to XSnew . The result is an increase in the equi-
librium credit spread to ω2 and a decrease in the amount of lending in the economy
to L 2 .
The increase to ω2 implies that i s must fall and that households’ supply of
lending to the banking system falls to L 2 . In Figure 11.7, we draw the IS and
MP curves in (i s , Y ) space instead of in (r, Y ) space as above. The MP curve is
given by M/P = L(r (Y, π) + π e , Y ) = L(i s (Y, π), Y ) since we have assumed that
π e = 0. Furthermore, money demand L(i s (Y, π), Y ) is a function of the interest
rate on savings i s rather than of i b . People are more willing to convert their long-
term assets into cash or cashable deposits if the interest rate on savings is low. The
monetary policy rule by the central bank i s (Y, π) is targeted at i s and stipulates
that i s increases with both Y and π as before. If this rule does not change, the MP
curve will not shift as a result of the contraction of the XS curve.
The IS curve will, however, be affected by the change. For a given level of Y , i s
is now lower than before due to the increase in the required credit spread. The IS
curve will therefore shift down, as in Figure 11.7. In equilibrium, there is a con-
traction of GDP. Hence, even without changes in monetary policy, a disturbance
in the credit markets can have real effects via the IS curve.
In terms of the 2008 crisis, the model above seems to suggest that a downward
spiral in credit supply and economic activity ensued after the initial shock. The
is
MP
A higher ω shifts the IS curve
ISnew IS
Figure 11.7 Impact of an increase in the credit spread.

contraction of GDP that was caused by the initial reassessment of banks’ assets
probably led to an even further shift of the XS curve to the left, which put pressure
on i s to fall even further. The IS curve thus also continued to fall to the left. As a
response, the central banks kept lowering interest rates according to their policy
rule, but eventually the level i s = 0 formed a limit to effective monetary policy
based on steering the real interest rate.
11.6 New Keynesian models

The Lucas critique and, even more so, the RBC paradigm both suggest that
changes in nominal variables such as the money stock should not have real effects.
According to the new Keynesian view, there might actually be good microeco-
nomic reasons for nominal rigidities and such rigidities might matter for the real
economy.
Consider, for instance, the assumption that there is a small cost associated with
a change in prices. In line with Mankiw (1985), we might think of such a cost
as arising from having to print new menus in the case of a restaurant. Let us
assume a market structure with some degree of monopolistic competition so that
there are many price-setting firms that find their optimal production level by set-
ting marginal revenue equal to marginal cost. Actual demand is revealed after
quantities have been set.8 If, for instance, the demand for a product decreases,
a profit-maximizing producer should decrease its price.
However, if there is a menu cost for changing prices, the producer must compare
the loss in profit from keeping the price at the existing level, to the small cost of
changing the price. If the loss is smaller than the menu cost, a rational producer
should not change the price. Hence, in this case, there are good microeconomic
reasons for price rigidity.9
In macro models of the aggregate price level, such price rigidity can be modeled
by assuming that only a fraction of firms adjust their price to the profit-maximizing
level. The remaining firms simply keep the price that they charged last period.
A particularly simple formulation used by, for instance, Gali and Gertler (2007)
is that during some period t, the actual (log) price level pt is determined by an
equation
pt = θ pt−1 + (1 − θ ) pt∗
The parameter θ should here be interpreted as the probability that firms cannot
adjust their price during period t. With probability 1 − θ , firms are able to adjust
their price to the profit-maximizing level. θ can thus be regarded as a measure of
the degree of price rigidity. It follows that each firm is expected to maintain its
price level for an amount of time equal to 1/(1 − θ ) > 0. If θ = 2/3, for instance,
then firms only adjust their prices every third period. The adjusted price level pt∗
will in turn be given by a mark-up over current and future expected marginal cost
since new Keynesian models typically assume that markets are characterized by
imperfect competition.
12 Public Finance and Fiscal Policy
Fiscal policy concerns a government’s handling of its revenues and expenditures.

Typically, there is an imbalance between revenues and expenditures, which, in
the case where expenditures exceed revenues, is referred to as a budget deficit.
Budget deficits are measured on an annual basis and are usually described as a
percentage of GDP. Accumulated budget deficits make up government debt, which
is a stock variable that can be measured at any point in time and is also described
as a percentage of GDP.
12.1 The government budget identity

The government budget identity at some point in time t shows the expenditures on
the left-hand side and sources of finance on the right-hand side in the following
manner:
G t + rt Dt ≡ Tt + Dt + Mt (12.1)
In this expression, G t is government expenditures in period t, rt Dt gives the real

interest rate rt times total accumulated government debt Dt , Tt is total tax revenue,
Dt = Dt+1 − Dt is the change in the stock of government debt, and Mt is the
change in the money stock that is used for financing the government deficit.
Starting with the expenditure side, it simply consists of government outlays and
interest payments. If the debt is high, interest payments will make up a large pro-
portion of total expenditures. On the right-hand side, the main part of government
spending should be financed by taxes Tt , but since these are not always sufficiently
high to cover expenditures, governments often run budget deficits that increase its
debt, Dt > 0. Irresponsible governments sometimes also try to cover deficits by
“printing money”, i.e. ordering the central bank to print new money that is then
handed over to the government. In developed countries, this rarely happens and
we will assume for now that Mt = 0.1
Rewriting (12.1) gives us an expression for the dynamics of the government
budget deficit:
Dt ≡ (G t − Tt ) + rt Dt
The term in parentheses is often referred to as the primary deficit. Budget deficits
imply that Dt > 0, whereas budget surpluses imply a decreasing stock of debt
Dt < 0. Note that in countries with very large debts (and hence large interest
payments), even a moderate primary surplus (G t − Tt < 0) might not be enough
for the debt to shrink.
Budget deficits are typically financed by governments’ issuing of bonds that are
sold to households and that give their holder an interest rate of rt every year. If the
total debt is bond-financed, Dt will be equivalent to the value of the stock of all
outstanding bonds.
Over the long run, it is often assumed that government revenues and expen-
ditures should add up. A “sustainable” long-term fiscal policy should thus
satisfy:
∞
∞

Gt Tt
+ D 0 ≤ (12.2)
(1 + rt )t (1 + rt )t
t=1 t=1
where rt is the real interest rate as before. In other words, over an infinite time
horizon, the present value of the flow of future government expenditures, plus
some initial debt level D0 , should not exceed the present value of the flow of all
future tax revenues. We will return to this intertemporal constraint below.
12.2 Ricardian equivalence

Should budget deficits be financed by higher taxes or by bonds? This is the central
issue in the literature on Ricardian equivalence, pioneered by Barro (1974). As
we shall see, the model is similar in spirit to the permanent income hypothesis of
aggregate consumption.
The analysis of Ricardian equivalence starts off with an assumption about an
infinitely lived representative household’s intertemporal budget constraint:
∞
∞ y t − τt
ct
≤ d 0 + (12.3)
(1 + rt )t (1 + rt )t
t=1 t=1
The present value of a lifetime of consumption expenditures should not exceed the
present value of the infinite flow of disposable incomes (labor income yt minus
income taxes τt ) plus the initial stock of bonds d0 . If the the economy is made up
of L identical individuals, we can express the individual’s constraint in (12.3) in
terms of Ct = ct L, D0 = d0 L, Yt = yt L, and Tt = τt L. ∞
t=1 Tt /(1 + rt ) =
t
When
∞
(12.2) holds with equality, we know that
t=1 G t /(1 + rt ) + D0 . Inserting this expression into (12.3) gives us
t
∞
∞ ∞
Gt
Ct Yt
≤ −
(1 + rt )t (1 + rt )t (1 + rt )t
t=1 t=1 t=1
Public finance and fiscal policy 117
In this way, we can express the households’ budget constraint as a function of
the present value of government expenditures. The key thing to notice is that the
time path of taxation does not enter the equation above. The financing of deficits
with bonds or taxes should therefore not matter to households. This is indeed
the Ricardian equivalence result. Only the quantity of government expenditures
should matter for consumption over the long run.
The intuition behind this result is that households do not value bonds as net
wealth. Consider the following example: From having a balanced budget with
no debt at some time t, the government chooses to lower taxes Tt . G t remains
the same as before, but now G t > Tt , so that debt accumulates. The government
finances the deficit by selling bonds to the households. Will this increase in dis-
posable income increase consumption? According to the Ricardian equivalence

result, it will not because households realize that the long-term budget constraint
is given by (12.2). At some point in time, governments will need to restore the
satisfaction of the constraint by increasing taxes again. Hence, the increase in dis-
posable incomes from the temporary tax decrease is saved in order to pay for the
higher taxes in the future.
In this sense, the implications of Ricardian equivalence are very similar to
those of the PIH model. The traditional Keynesian view would, however, hold
that an increase in bond financing should increase consumption since it leaves
households with a temporarily higher disposable income. Once again, there
is thus a substantial difference in the predicted effects of government policy
between the traditional Keynesian view and the view of rational, forward-looking
households.
Ricardian equivalence is one of the most often discussed results in macroeco-
nomics. A number of objections have been raised against Barro’s hypothesis. For
instance, it assumes (like the PIH model) that individuals have no liquidity con-
straints and can always save and borrow without difficulty. A second objection is
that it is not likely that individual households will act in accordance to the infinite
horizon budget constraint in (12.3). Rather, it is more likely that one generation
(with a finite lifetime) will indeed react to medium-term expansionary fiscal poli-
cies since it is unrealistic to believe that they are perfectly altruistic towards future
generations. Empirical evidence also seems to suggest significant departures from
Ricardian equivalence.2
12.3 Tax smoothing

The Ricardian equivalence analysis above focused on households that reacted to
exogenously set levels of taxes and government expenditures. But what determines
the decisions of governments regarding budget deficits or surpluses? In standard
microeconomic theory, it is shown that taxes are often associated with deadweight
losses that reduce welfare. In this section, we assume in line with Barro (1979)
a benevolent social planner who wishes to minimize the distortions caused by
taxation, while still honoring the intertemporal budget constraint.
Let us assume, as before, that government expenditures G t are exogenous to the
model. Taxes give rise to distortionary costs for the economy equal to

Tt Tt Tt
Z t = Yt z , z > 0, z >0 (12.4)
Yt Yt Yt
The expressions in (12.4) assume that the distortionary costs of taxation are an
increasing and convex function of taxes as a share of GDP, Tt /Yt (the tax quotient),
scaled by the general size of the economy Yt . The z-function provides a shortcut
for the many different distortions that taxes give rise to but provides no information
about their exact sources. No taxes implies zero distortions in this setting: z(0) = 0.
The intertemporal budget constraint that governments face is still given

by (12.2). Hence, the optimization problem becomes
∞
∞
∞ Gt
Yt Tt Tt
min z subject to = + D0
T0 ,T1 ... (1 + rt ) t Yt (1 + rt ) t (1 + rt )t
t=1 t=1 t=1
As was the case in the PIH, the first-order condition for this problem gives us an
T∗ T∗
Euler equation of the type z ( Ytt ) = z ( Yt+1 ), which in turn implies that optimally
t+1
∗
Tt∗ Tt+1
=
Yt Yt+1
In other words, distortionary costs are minimized when taxes as a share of GDP are
smoothed over time. This is the tax smoothing result, with the same basic intuition
as consumption smoothing.
Extending this model to include uncertainty is straightforward. In the case of a
quadratic z-function (as in Hall’s random-walk model), we will have
∗
Tt∗ T
= E t t+1
Yt Yt+1
Taxes as a share of GDP thus follow a random walk. The result suggests that at all
points in time, the government includes all available information about expected
future incomes and sets the current level of Tt /Yt so that it is expected to be equal
to all future levels of tax quotients. Any change from this pattern must be due to
random new information.
12.4 Political economy of government debt

In the tax smoothing model above, tax policy is determined by a benevolent
government that aims to minimize the distortionary effects of taxes. It does not
have much to say about the dynamics of budget deficits. Over the infinite time
horizon, the current value of total taxes will be equal to the current value of gov-
ernment spending, but in individual time periods there might be an imbalance. The
benevolent government will, however, correct those imbalances sooner or later.
A less optimistic observer of fiscal policy around the world would probably
argue that many governments’ levels of deficits and debts cannot possibly be sus-
tainable in the long run and cannot reasonably reflect expectations about potential
future incomes. For instance, several countries around the world have levels of
accumulated government debt above 100 percent of GDP at the same time as the
future prospects for their economies appear bleak. Historically, it is well known
that some countries have even gone “bankrupt” and have had to cancel or exten-
sively reschedule their debt repayments.3 Hence, it has been argued that long-term
fiscal policy frequently is characterized by government failure.

A potential source of such failures is the political system itself. In a famous
model, Alesina and Tabellini (1990) argue that in a democracy where two parties
with different preferences over public goods compete for office, an incumbent
party might strategically “overspend” while in office in order to constrain the fiscal
choices for an opposing party should that party win office in the next period. The
result of such a strategic use of government debt is a deficit bias in fiscal policy
that is inefficient from a social point of view.
Consider a country with two political parties that we might refer to as left (L)
and right (R). The parties do not disagree on the level of taxes but only on how
to spend the taxes raised. The left-wing party wants to spend raised taxes only on
public good gtL ≥ 0, where t indicates the period in question, whereas the right-
wing party prefers to spend on a different public good gtR . One might, for instance,
think of gtL as general health care and gtR as military defense. In this simplified
setting, there are only two time periods: period 1 and period 2. The identity of the
incumbent party in period 1 is given and there is then an election at the begin-
ning of period 2. The right-wing party is expected to win this election and rule in
period 2 with an exogenously given probability of ρ ∈ [0, 1]. The probability of a
left-wing victory in the election is thus 1 − ρ.
The utility functions of the right- and left-wing parties, given that they are in
power during the first period, are
V R = u(g1R ) + β[ρu(g2R ) + (1 − ρ)u(g2L )] (12.5)

V L
= v(g1L ) + β[(1 − ρ)v(g2L ) + ρv(g2R )]
respectively. The instantaneous utility functions u(gtR ) and v(gtL ) have the usual
properties u (gtR ), v (gtL ) > 0 and u (gtR ), v (gtL ) < 0. Note also that u(gtL ) =
v(gtR ) = 0, i.e. if the other party wins the election and provides its preferred public
good, the losing party is assumed to yield no utility. β ≤ 1 is the usual time discount
factor.
The country earns an exogenous flow of per capita incomes during the two
periods equal to y1 , y2 > 0, where we assume y2 ≥ y1 . The income tax rate is
τ < 1, so that the intertemporal flow of government revenue is τ y1 and τ y2 . The
government starts off without any government debt. In the first period, the govern-
ment in power might, however, incur a debt equal to d that must be repaid in full
in period 2. These conditions imply that the budget constraints for party j = R, L
in power are
j
g 1 = τ y1 + d
(12.6)
j
g 2 = τ y2 − d
The party in power during period 1 thus maximizes (12.5) subject to (12.6).
The choice variable that we are interested in is the level of debt chosen in the first
period, d. The optimal level of debt can be found if we insert the expressions for
(12.6) into the utility functions and then maximize with respect to d.
Let us, for instance, assume that the right-wing party is in power in period 1.
The optimization problem for the government is thus
max u(τ y1 + d) + βρu(τ y2 − d)

d
The first-order condition for this problem is
∂V R
= u (τ y1 + d) − βρu (τ y2 − d) = 0
∂d
With this general utility function, we are as always unable to calculate any explicit
solutions, although it is certainly possible to carry out comparative statics with the
help of implicit differentiation.
In order to derive an explicit solution, let us assume logarithmic utilities so that
u(gtR ) = ln gtR and u(gtL ) = ln gtL . In that case, the first-order condition tells us that
1 βρ
=
τ y1 + d τ y2 − d
By manipulating this expression, we can solve for the optimal level of government
debt:
τ (y2 − βρy1 )
d∗ = (12.7)
1 + βρ
The key insight from this expression is that the optimal level of debt incurred
by an incumbent party in the first period is negatively associated with the party’s
probability of winning the election in period 2. More formally, the derivative is
∂d ∗ (y1 + y2 )βτ
=− <0
∂ρ (1 + βρ)2
Hence, if the right-wing party is the incumbent in period 1, the optimal debt will
be smallest if ρ = 1, i.e. if the incumbent party is 100 percent sure of winning
the next election. From (12.7), we can deduce that the debt will then be d ∗ =
τ (y2 − βy1 )/(1 + β) ≥ 0. Note that as long as y2 − βy1 > 0, the party in govern-
ment will always opt for a positive amount of debt. The size of the debt increases
with the growth in income, y2 − y1 . Hence, it is more rational for fast-growing
economies to have government debt than for slow-growing countries.
A straightforward corollary of the result above is that debt will be highest when
the probability of re-election ρ approaches zero. If that is the case, we can see from
the utility function that the party in power will discount the future by a factor βρ
that will be close to zero. Since the party can only spend on the public goods
that it likes in the first period, it will take on as much debt as possible in the first
period. In the limiting case of ρ = 0, we can deduce from (12.7) that the optimal
first-period debt will be τ y2 , i.e. the debt will be so high that the opposing party
that takes over has to spend all its government revenue on paying back the debt. If,
for example, the right-wing party only cared for military spending, it would spend
τ (y1 + y2 ) on the military in period 1 so that the left-wing party would not be able
to spend anything on health care while in power in period 2.
Usually, the level of government debt is discussed in terms of shares of total
GDP. If y1 = y2 = y, then optimal debt as a share of GDP is simply
d ∗ τ (1 − βρ)
=
y 1 + βρ
This expression highlights two other important factors for countries’ willingness
to accept government debt: the time discount rate β and the tax rate τ . Societies
with more impatient individuals (a low β) are more likely to have high debt than
societies with patient individuals. Similarly, it is only intuitive that high-tax coun-
tries (a high τ ) with a large public sector will have higher debt as a share of GDP
than low-tax countries.
12.5 Debt financing versus debt forgiveness

As indicated above, many countries in the world have levels of government debt
that are not sustainable. In fact, there are several historical examples of situations
where a government’s international lenders have realized that the country’s level of
debt far exceeds the expected debt repayments in the future. What should lenders
then do, in their own interest as well as in the interest of the indebted country?
Should they continue to finance the debt with new loans, hoping that the country
will be able to pay after all, or should they forgive some of their claims so that the
level of debt stabilizes at a sustainable level?
The debt crisis among developing countries during the era of structural adjust-
ment programs led to debt rescheduling and even debt forgiveness for some highly
indebted countries. Since the financial crisis of 2007–08, the issue has become a
particularly hot topic within the European Union, with countries like Greece and
Portugal having difficulties in servicing their debts.
12.5.1 Debt overhang
In this section we briefly outline the influential analysis of the problem by
Krugman (1988). The model assumes a two-period framework as above and that
the country in question has inherited a level of debt equal to d > 0. The flows
of government revenue during periods 1 and 2 are given by τ y1 and τ y2 where
τ is an exogenously given tax rate and y1 , y2 > 0 are levels of income. The debt
can be paid back during both periods. The country can at most reduce its debt
in the first period by τ y1 , so that debt in period 2 is d − τ y1 = b ≥ 0. If b > 0,
international lenders might choose to refinance this amount at the prevailing inter-
national interest rate r > 0. The debt payment in the second period will then be
(d − τ y1 )(1 + r ).
Assume now that whereas first-period revenues are known with certainty,
second-period revenues are uncertain and might take on either a high value y2H
or a low value y2L , where y2H > y2L . The differences in income could, for instance,
be due to high or low world market prices for the goods that the country is selling.
Suppose that the probability of the good outcome is ρ and the probability of the
bad outcome is 1 − ρ. On top of this, actual revenue in period 2 might depend
on government adjustment effort e. Adjustment efforts could, for instance, be an
exchange rate realignment or the abolition of harmful tariffs that serve to improve
the functioning of the economy and increase income. For simplicity, we assume
that every unit of adjustment effort results in one additional unit of output. Taken
together, the expected level of second-period government revenue in period 1 is
therefore E 1 (y2 ) = ρy2H + (1 − ρ)y2L + e whereas the actual levels are
y2 = y2i + e
where y i ∈ {y H , y L }. We will return later to what determines the government level

of effort e.
The country’s expected maximum ability to pay in the second period is thus
τ E 1 (y2 ) = τ [ρy2H + (1 − ρ)y2L + e]. The country has an unsustainable level of
debt in the second period if b(1 + r ) > τ E 1 (y2 ) . Equivalently, a country with an
unsustainable level of debt (in Krugman’s words a debt overhang) is characterized
by a situation where
τ E 1 (y2 ) τ [ρy2H + (1 − ρ)y2L + e]

d > τ y1 + = τ y1 + (12.8)
1+r 1+r
i.e. where the level of debt exceeds the expected present value of total future gov-
ernment revenue. A typical risk-neutral or risk-averse international lender would
in general be reluctant to lend to such a country in the initial period. A result of a
refusal to lend could be that the country has to default on its debt and cancel all
repayments.
An obvious general solution to the debt overhang problem above would be
to raise the proportional tax rate to a new level τ new > τ such that τ new ≥
E 1 (y2 )
d/ y1 + 1+r . In developed economies in the Western world, this is certainly a
feasible strategy that might at least partly overcome the overhang problem. How-
ever, as emphasized by Besley and Persson (2010), for instance, a country’s fiscal
capacity to tax its citizens is closely linked to its legal capacity, or more gen-
erally, to its state capacity. Some countries, for instance in Europe, developed
such capacities over several hundreds of years, and states that were not able to
tax their citizens effectively were often simply wiped off the map through mili-
tary competition. Similar processes have not been in place in other parts of the
world, particularly not in Africa where statehood is generally quite recent. Hence,
the capacity to raise the level of τ might just not be in place in many parts of the
world.
Even if this is the case, both the indebted country and the world community usu-
ally want to avoid episodes of sovereign default, although there are many instances
of such events in history. Since it was founded in 1945, the International Mone-
tary Fund (IMF) has acted as a lender of last resort to heavily indebted countries.
By 2011, Greece was one of the countries that had to turn to the IMF for help to
finance its government debt.
How should an optimal debt contract be designed? Clearly, a profit-maximizing
financial institution would only care about its own expected profit and would not
lend to a country with a debt overhang. Given that the IMF presumably gives some
weight to the indebted country in its utility function, its optimal strategy will be
different.
Let us consider the situation in period 1 and imagine that a liquidity crisis has
emerged so that remaining debt payments exceed expected future government rev-
enue, b(1 + r ) > τ E 1 (y2 ). What are the options for a lender like the IMF? To begin
with, note that even if the debt is unsustainable according to the definition above,
it might still be the case that τ (y2H + e) > b(1 + r ), i.e. the country might be able
to pay back its debt if the good outcome y2H materializes. One strategy might
therefore be to lend the full amount and hope for the good outcome to happen
(or for government effort e to be sufficiently high) so that the IMF recovers all of
its credit. The actual repayment q would then be:

τ y2 if τ y2 < b(1 + r )
q= (12.9)
b(1 + r ) if τ y2 > b(1 + r )
If q = τ y2 , the government must default on debt payments of the size b(1 + r ) −

τ y2 > 0. The amount of tax revenue left for the government to spend on public
goods other than debt repayments will in that case be g2 = τ y2 − q = τ y2 − τ y2 = 0.
If, however, τ y2 > b(1 + r ), then g2 = τ y2 − b(1 + r ) > 0.
A second strategy for the IMF might be to provide the loans but at an interest
rate set lower than the international interest rate, r low < r . One possible choice is
simply to set the interest rate at a level such that the country just can pay back in
the bad outcome:
τ yL
q = (1 + r low )b = τ y L =⇒ r low = −1
b
in which case the country’s debt would in fact be sustainable. Actual payment,
regardless of outcome, is then q = τ y L . However, this lower interest could be seen
as a reward for irresponsible behavior and since the IMF would have to make up
for the interest differential to the international interest rate (r − r low ), it would still
be close to a government default.
A third and closely related strategy might be to reschedule or “forgive” some of
the debt already in period 1 to a new level blow < b such that
τ yL
q = (1 + r )blow = τ y L =⇒ blow =
(1 + r )
This is equivalent to a partial default by the government by an amount

(b − blow )(1 + r ) and is likely to make it hard for the country to borrow money on
the international capital market in the future.
Why would the IMF ever choose the latter two alternatives when they clearly
yield repayments that are smaller than or equal to the contingent payment
in (12.9)? Note that in the two latter cases, government spending is either g2 =
τ y2 − τ y L = τ (y H + e) > 0 or g2 = τ (y L + e) ≥ 0. By setting a low level of
repayment, the IMF gives governments stronger incentives for making necessary
adjustment efforts e. In the subsection below, we formalize this idea further.
12.5.2 Endogenous adjustment efforts

What determines a country’s adjustment efforts when it has been hit by a debt
crisis? In line with the model in Krugman (1988), let us assume that the gov-
ernment has a very simple utility function such that it only has preferences over
second-period government spending g2 and its adjustment effort e (for instance,
because the indebted country must always pay back τ y1 in the first period so that
g1 = 0 anyway). Government spending is given by g2 = τ y2 − q, where q is the
actual repayment that the government makes according to its contract with the
IMF, as specified above.
The expected utility of second-period government consumption is given by a
function E 1 (u(g2 )) such that u (g2 ) > 0 and u (g2 ) < 0. The government gets dis-
utility from making adjustment efforts e. Such necessary reforms typically imply
that some groups in society lose as compared with before, perhaps due to labor
market or pension reforms. This disutility is described by a function v(e) with the
properties v (e) > 0 and v (e) > 0 so that marginal disutility is increasing in e.
Taken together, the expected utility of the government is given by
E 1 (U ) = E 1 (u(g2 )) − v(e) (12.10)

= [u(τ [ρy2H + (1 − ρ)y2L + e] − q)] − v(e)
where we have substituted in E 1 (y2 ) = ρy2H + (1 − ρ)y2L + e in the second line.

The government’s control variable in this utility function is its level of adjust-
ment effort e. It is easy to see that adjustment effort entails a trade-off. On the one
hand, a higher e increases E 1 (y2 ), which usually increases potential government
spending g2 . On the other hand, the “marginal cost” of effort is v (e) > 0. Note
also that for all levels such that τ y2 < b(1 + r ), the government has no incen-
tive to increase its efforts since all extra government revenue will disappear in
repayments to lenders.
To see this more formally, we can derive the first-order condition for an interior
maximum on the basis of (12.10) to be
∂ E 1 (U ) ∂ E 1 (g2 )
= u (g2 ) − v (e) = 0
∂e ∂e
Since we know that u (g2 ) > 0 and v (e) > 0 for all e > 0, the existence of a positive
equilibrium level e∗ > 0 will depend on the term ∂ E 1 (g2 )/∂e. If the IMF does not
forgive any debt, if it lends at the international interest rate r , and hopes for a
positive outcome so that a full repayment can be made, then for all levels such that
τ E 1 (y2 ) < b(1 + r ), the government is expected to earn τ E 1 (y2 ) and has to repay
the same amount. A rise in e increases expected revenue and expected repayments
one for one and hence ∂ E 1 (g2 )/∂e = 0. Governments therefore have no incentive
to make adjustment efforts and the optimal level is e∗ = 0.
If the debt contract is set at the more generous level q = τ y L , then τ E 1 (y2 ) > q
and extra efforts will increase E 1 (g2 ) at all e > 0. This implies that ∂ E 1 (g2 )/∂e = 1
and that the optimal level of effort is given by u (g2 ) = v (e∗ ). In this section, we
have not explicitly modeled the utility function of the IMF, but is seems quite
intuitive that an outcome where e∗ > 0 might be part of its preference structure.
13 Inflation and Monetary Policy
Inflation is the percentage increase in the aggregate price level (usually measured
by a consumer price index) from one year to the next. Inflation has been observed
throughout history for as long as money has existed. The chapter will focus on
the causes of inflation and analyze how monetary policy affects inflation and the
economy as a whole.
At a general level, we saw in Chapter 11 that inflation will be heavily influ-
enced in the short run by aggregate supply and aggregate demand. An expansion
in demand, for example, always leads to an increase in inflation (unless the AS
curve is horizontal). Negative supply shocks – for instance, due to a rise in the
price of a key production factor like oil – will lead to a supply contraction and an
increase in inflation.
In this chapter we will delve further into mainly the political economy of infla-
tion and the perceived trade-off between inflation and output growth. We will
start off with the classical quantity theory of money and then proceed with the
time-inconsistency model of monetary policy and its potential solutions, originally
associated with Kydland and Prescott (1977). After that, we move on to an exten-
sive discussion of political business cycles and analyze how inflation is affected
by the presence of political parties with different preferences for price stability.
These sections will rely to a great extent on Alesina and Stella (2010). Finally, we
include a treatment on seigniorage.
13.1 The quantity theory of money

In the long run, it is widely recognized that inflation has a very close association
with money supply. The most general formulation of the quantity theory of money
proposes the following basic relationship:1
MV = PY
In this equation, M is the nominal stock of money as before, V is the velocity of

money, or the average frequency at which a unit of money is spent in an economy,
P is the aggregate price level, and Y is real GDP. V is a parameter that does
not change much over time and that depends among other things on a country’s
Inflation and monetary policy 127
financial sector development. Y is usually also taken as given in this setting. Thus,
the aggregate price level can be formulated as P = M VY . An increase in M will
have a positive effect on P, i.e. there will be inflation. Especially over the long
run, it has been well documented empirically that large increases in money supply
tend to be followed by inflation.
13.2 Inflation and the money market

From Chapter 11, we know that equilibrium in the money market will prevail if
real money supply equals real money demand such that
M
= L(r + π e , Y )
Rewriting this expression, and assuming for now that Ȳ and r̄ are constants and
that the monetary policy rule does not apply, we get P = M/L(r̄ + π e , Ȳ ), i.e. just
like in the quantity theory, there is a close association between the aggregate price
level and the nominal money stock. Indeed, we will have that actual inflation and
expected inflation are
Ṗ Ṁ
= π = πe =
P M
where Ṗ/P and Ṁ/M are the growth rates (the time derivative divided by the cur-
rent level) of the aggregate price index and the nominal money stock, respectively.
Prices are thus completely flexible.
Let us consider the case when a monetary authority announces that it will
increase the growth rate of money Ṁ/M. This increase in turn increases actual
as well as expected inflation π = π e . Through its positive impact on the nominal
interest rate i = r + π e , the demand for money therefore shrinks immediately.2
In order to balance the fall in money demand, real money supply must also fall,
i.e. P must increase. Hence, the rise in the money growth rates both increases the
inflation rate in the longer run and leads to a one-time discontinuous jump in the
aggregate price level.
Analogously, if monetary authorities reduced money growth, there would be a
decrease in long-run inflation but also an immediate discontinuous fall in the price
level. In order to avoid such a dramatic scenario, central banks should combine the
lowering of long-run money growth rates with a temporary increase in the level of
money supply (Romer 2005).
13.3 Time inconsistency in monetary policy

Although the quantity theory and money market analysis above highlights the
close association between money supply and inflation, it does not answer the ques-
tion of why central banks or governments choose to maintain too high a level of
money growth. The remaining sections deal primarily with this issue.
From the analysis of Phillips curves in Chapter 10, we know that a plausi-
ble explanation for persistently high levels of inflation is the perceived trade-off
between inflation and unemployment. In a famous article, Kydland and Prescott
(1977) argued that even without a long-run output–inflation trade-off, policy-
makers will tend to cause inflation rates that are above the socially optimal level,
a phenomenon referred to as inflation bias.
13.3.1 Basic assumptions

Let us assume a Lucas supply function of the same kind as in Chapter 10,
y = ȳ + b(π − π e ) (13.1)
where y is actual output (in logarithmic form), ȳ is the natural rate of output (long-
run aggregate supply), π and π e are actual and expected inflation as before, and
b > 0 is a parameter capturing the sensitivity of output to deviations in the inflation
rate. Central banks (or policy-makers more generally) are assumed to be in control
of the actual level of inflation π, whereas the public forms rational expectations
about inflation π e . One might think of these expectations as being manifested in
wage contracts so that there is an element of stickiness in inflationary expectations.
The only way that central banks can increase output from its natural level ȳ is to
surprise the public by setting an inflation rate π > π e .
Let us further assume that central banks are governed in their decision-making
by a (dis)utility or social loss function
1 1
V = (y − y ∗ )2 + a(π − π ∗ )2 (13.2)
2 2
The aim of the central bank is to conduct a monetary policy that ensures that actual
output and inflation are as close as possible to some target levels y ∗ and π ∗ . a > 0
reflects the relative weight given to fighting inflation. Most central banks have
inflation fighting as their primary objective, implying a > 1. Furthermore, central
banks as well as governments will typically want to maintain levels of output
(unemployment) that are higher (lower) than the natural rate, i.e. y ∗ > ȳ. Let us
think of y ∗ as the level of output associated with full employment. We assume, for
simplicity, that the targeted rate of inflation is π ∗ = 0.3
There are at least two main monetary policy strategies that central banks can
follow: either to make binding inflation commitments that are independent of the
public’s expectations or to allow for a discretionary monetary policy that takes the
public’s expectations into account in its decision-making.
If the central bank makes binding commitments, it will not allow itself to adjust
levels of inflation as a response to how the economy performs. Assume that the
central bank makes a commitment that, no matter what, it will maintain inflation
at π = 0. If the public considers this to be a credible commitment, its inflation
expectations will be π e = 0. In this case, the central bank will simply set π = 0,
which is the socially optimal level.
13.3.2 Discretionary monetary policy
If the central bank follows a discretionary monetary policy, then its governors
allow themselves flexibility by determining the inflation rate after the public has
formed its expectations. An important component is further that central banks rec-
ognize that actual output is given by (13.1). There is thus a possibility of boosting
output by setting an inflation rate above the expected one, as implied by the Lucas
supply function. Combing these assumptions gives us the following minimization
problem for the central bank:
1 1
min V (π) = [ ȳ + b(π − π e ) − y ∗ ]2 + aπ 2 (13.3)
π 2 2
However, the public does not want to be fooled, and forms rational expecta-
tions about the central bank’s inflation policy. Formally, the objective of the public
might be described as being aimed at minimizing (π e − π)2 . In Kydland and
Prescott’s model, the interaction between the central bank and the public takes
the form of a sequential, complete information game where the public moves first
by forming π e , whereupon the central bank sets π conditional on the observed
level of π e .
As usual in sequential games, we find the solution through backward induction,
i.e. by starting in the second stage.
Second stage: The central bank chooses the π that minimizes (13.3), taking π e
as given.
In order to find the level that minimizes (13.3), we need the first-order
condition4
∂V
= [ ȳ + b(π − π e ) − y ∗ ]b + aπ = 0
∂π
After some manipulations, the combined levels of π and π e that satisfy this first-
order condition give us the central bank’s best response (or reaction) function
b ∗ b2 π e
π r (π e ) = (y − ȳ) + (13.4)
a + b2 a + b2
where π r is the level of π that minimizes (13.3).

The reaction function in (13.4), which is graphically illustrated in Figure 13.1,
shows two important things. Firstly, the chosen level of π will always be larger
than zero since the first term b(y ∗ − ȳ)/(a + b2 ) > 0. The size of this effect
depends in turn on the deviation of the socially optimal level of output (and
unemployment) y ∗ from the natural level ȳ. This is the origin of the inflation
bias in discretionary monetary policy referred to earlier. Secondly, the central
bank’s optimal level of inflation will increase linearly with the public’s inflation
expectations π e .
π
45º
πr
slope: b 2/(a + b 2) < 1
b(y * – y )/(a + b2)

πe
Π EQ
Figure 13.1 Equilibrium inflation under discretionary monetary policy.
In order to obtain the full solution to the model, we need now to move to the
first stage:
First stage: The public forms inflation expectations π e , taking into account the
known response function of the central bank.
The public knows the rules of the game and realizes that the central bank will
attempt to surprise it by setting an inflation higher than the expected one according
to (13.4). Since the people want to minimize surprises, they will increase their
inflationary expectations until π e = π r . Since the slope of the reaction function
in (π, π e ) space is smaller than unity, there must be a level where this equality
holds. As shown in Figure 13.1, this happens at a level π = π EQ . Inserting π EQ =
π e = π r into (13.4) and solving gives us the subgame perfect Nash equilibrium
level of inflation:
b
π EQ = (y ∗ − ȳ) > 0
a
The central insight from this expression is, once again, that π EQ > 0, i.e., even
though the official inflation target is π ∗ = 0, the fact that discretionary mone-
tary policy is allowed will always give us an inflation rate that is higher than the
socially optimal level. The reason is that a target of π ∗ = 0 is time-inconsistent:
when the public’s inflation expectations are set first, it is always optimal for the
central bank to set π > 0 in the next stage. Hence, although discretionary policy is
more flexible, which sounds like a good thing, it produces inflation rates that are
higher than those that would prevail under binding inflation targets.
Note also that since π EQ = π e in equilibrium, y EQ = ȳ + b(π EQ − π e ) = ȳ, i.e.
the equilibrium level of output will remain at the natural rate of output and below
the socially optimal level y ∗ . In this sense, the net result of discretionary monetary
policy is only a socially suboptimal level of inflation.
13.3.3 Random output shocks

In reality, monetary policy is conducted in a world of genuine uncertainty about
how the economy will evolve. As the real business cycle literature has shown, there
are good reasons to believe that shocks tend to appear randomly. Furthermore, it is
often the case that monetary policy is set partly in response to the realization of
random shocks and after the public has formed its expectations (Alesina and Stella
2010).
In order to illustrate this, let us assume that output at time t is now given by
yt = ȳ + b(πt − πte + εt ) (13.5)
where εt is a random output shock with an expected value E t (εt ) = 0. For simpli-
city, let us also assume that εt can assume the values γ > 0 in good times and −γ
in bad times with a probability distribution that satisfies E t (εt ) = 0.
If we set up the same loss function as before with this new assumption, it will be
1 1
min V (π) = [ ȳ + b(π − π e + εt ) − y ∗ ]2 + aπ 2 (13.6)
π 2 2
From the first-order conditions, we can deduce that the central bank’s reaction
function will now be
b b2 (π e − εt )
π r (π e ) = (y ∗ − ȳ) + (13.7)
a+b 2 a + b2
In the second stage, when the central bank determines the level of inflation by
taking the public’s expectations π e as given, it also has information on the actual
realization of εt . The public, however, will make its expectations before knowing
what actually happens to the economy. Their rational expectation about the shock
is simply E t (εt ) = 0.
When we insert π EQ = π e into the expression above as before, we can solve for
the equilibrium level of inflation under discretion and uncertainty:
b ∗ b2 (π EQ − εt )
π EQ = (y − ȳ) +
a + b2 a + b2

b2 a b b 2 εt
π EQ 1 − = π EQ = (y ∗ − ȳ) −
a+b 2 a+b 2 a+b 2 a + b2
b
π EQ = (y ∗ − ȳ) − b2 εt
a
What this means is that equilibrium inflation will be ab (y ∗ − ȳ) − b2 γ in good

times and ab (y ∗ − ȳ) + b2 γ in bad times. Inflation will thus be higher in bad times.
The logic behind this result is that output will be even further away from the target
level y ∗ in such periods, which will make an inflationary policy that boosts output
even more tempting for the central bank. The reverse logic applies for good times.
13.3.4 Solutions to time inconsistency

Two types of remedies to time inconsistency are often discussed in the literature:
reputation and delegation. Starting with reputation, this solution depends com-
pletely on whether the monetary authority is in place more than one period. If it
is not, then it is always optimal to play a nonrepeated game as above. If, however,
the policy-maker is in charge for more than one period, expectations about infla-
tion in the second period will depend on actual inflation in the first period. In this
case, there will be an intertemporal “externality” of the first-period choice of infla-

tion and policy-makers will in general be less inclined to tolerate high first-period
inflation. Reputational considerations should thus at least lead to a lower level of
first-period inflation.
A second potential solution to time inconsistency is delegation of monetary
authority to an independent central bank that is required (often by law) to focus on
fighting inflation. In terms of equation (13.2) in our sequential game setting above,
this would be equivalent to handing over authority to some “conservative” central
banker who has an a > a (Rogoff and Sibert 1985). In that case, the intercept in
Figure 13.1 would be lower and the slope flatter, which of course would imply an
equilibrium level of inflation that is lower than before. It would, however, still be
above zero as long as y ∗ − ȳ > 0.
13.4 Political business cycles

In the models above, there was no question about what were the socially optimal
levels of inflation and output. In Western democracies, economic policy is typi-
cally a key issue in elections, and parties often differ in their stated objectives. In
the literature on partisan business cycles, pioneered by Hibbs (1977), it is argued
that left-wing parties tend to be more concerned about unemployment whereas
right-wing parties care more about bringing down inflation.
Such assumptions can be easily incorporated into the basic model.5 An implicit
assumption in this section is that political parties in government can actually influ-
ence monetary policy in a discretionary manner.6 Let the left-wing and right-wing
parties have loss functions given by
1 1
V L = (y − y ∗ )2 + a L π 2
2 2
1 ∗ 2 1
V = (y − y ) + a R π 2
R
2 2
respectively, where V L is the loss function of the left-wing party and V R the
equivalent function for the right-wing party. The key difference between the two
is that a R > a L , i.e. the right-wing party attaches a relatively higher weight to
fighting inflation. Both parties want to achieve full employment, so that y = y ∗ ,
but the left-wing parties are, as will be demonstrated, more willing to pay a price
for this in terms of higher inflation. Compared with the above, let us now set b = 1
for simplicity so that output responds fully to a surprise inflation.
If we minimize the loss functions and take the first-order conditions in the usual
way, we can derive the optimal response functions
y ∗ − ȳ πe
π r,L = + (13.8)
1 + aL 1 + aL
y ∗ − ȳ πe
π r,R = +
1 + aR 1 + aR
and the equilibrium levels of inflation
(y ∗ − ȳ) (y ∗ − ȳ)
π EQ,L = > 0, π EQ,R = >0 (13.9)
aL aR
Since a R > a L , it follows that π EQ,R < π EQ,L , i.e. equilibrium inflation will be
lower when the right-wing party is in power. Note that equilibrium output levels
will be y L = ȳ + π EQ,L − π e = ȳ and y R = ȳ + π EQ,R − π e = ȳ.
In nonelection years, this is well known to the public and they adjust their
expectations accordingly. In election years, however, there is an uncertainty about
what party will be in power. In line with the discussion about random output
shocks above, let us imagine that public expectations are made in the first stage
and that the politically controlled central bank then sets the inflation rate after
having observed the election result and in accordance with the winning party’s
preferences.
Let ρ be some objective probability of an election victory for the right-wing
party at the time expectations are formed. The probability of a left-wing victory is
therefore 1 − ρ. The expected inflation is then
π e = (1 − ρ)π r,L + ρπ r,R

∗ ∗
y − ȳ + π e y − ȳ + π e
= (1 − ρ) +ρ
1 + aL 1 + aR
After some algebra, we can isolate π e so that it becomes
(y ∗ − ȳ)[1 + a R − ρ(a R − a L )]
πe = (13.10)
(1 + a R )a L + ρ(a R − a L )
This expression implies that π e will decrease with ρ, the probability of a right-
wing election victory. This is well in line with intuition since the right-wing party
is more inflation-averse than the left-wing party.
We can insert the result in (13.10) into (13.8) to solve for the actual levels of
inflation during an election year t:
EQ,L (y ∗ − ȳ)(1 + a R )
πt =
(1 + a R )a L + ρ(a R − a L )
EQ,R (y ∗ − ȳ)(1 + a L )
πt =
(1 + a R )a L + ρ(a R − a L )
Once again, since a R > a L , it will be the case that πtEQ,L > πtEQ,R . Another inter-
esting result emerges if we compare chosen inflation levels by each party during
a “mid-term” year when there is no election, and during an election year. Let us
imagine that the mid-term year is t − 1 and that an election happens the next year
EQ,L
at t. Hence, πt−1 is the optimal level of inflation chosen by a left-wing party in
EQ,R
office during a mid-term year and is given by (13.9). Let πt−1 be defined in the
same way. Then it is easily shown that the following inequalities will hold:
EQ,L EQ,L
(i) πt < πt−1 ;
EQ,R EQ,R
(ii) πt > πt−1 ;
EQ,L EQ,R
(iii) πt > πt−1 ;
(iv) πtEQ,R < πt−1
EQ,L
The direction of the sign in (i) informs us that when a left-wing party is in
power and wins the election, inflation will fall during an election year. The reason
is that voters give the right-wing party a probability ρ > 0 of winning and hence
have lower expectations for inflation during the election year. The government-led
central bank can thus set a lower level of inflation than before after a left-wing
victory. For a right-wing party in power, the reverse logic applies in (ii); in an
election year, inflation expectations will rise due to some probability of a left-
wing party win and the economy will thus get a higher level of inflation during an
election year, even when the right-wing party wins.
Inequalities (iii) and (iv) are perhaps less surprising: if a left-wing party wins
the election and takes power from the right-wing party, inflation will increase (iii),
whereas if the reverse happens, inflation will fall (iv).
The actual levels of output under different election winners during an election
year will be
ytL = ȳ + πtEQ,L − πte (13.11)

ρ(y ∗ − ȳ)(a R − a L )
= ȳ + = ȳ + ρη
(1 + a R )a L + ρ(a R − a L )
EQ,R
ytR = ȳ + πt − πte (13.12)
(1 − ρ)(y ∗ − ȳ)((a R − a L ))
= ȳ − = ȳ − (1 − ρ)η
(1 + a R )a L + ρ(a R − a L )
where η > 0 is given by the cluster of terms in the middle of (13.11) and (13.12).
Clearly, it will be the case that ytL > ytR , i.e. with a given pre-election probability of
a right-wing election victory ρ, an actual win by the left-wing party will generate
a greater level of output than a win by the right-wing party. This is of course only
what would be expected given the parties’ preferences over economic policy.
However, note that elections will give rise to political business cycles. Output
during a mid-term year with no elections and either party in office is just the natu-
ral rate ȳ since the public correctly anticipates the level of inflation in equilibrium
and the central bank therefore cannot boost output by surprising the electorate.
But the level in (13.11) is higher than the natural rate during an election year t.
The reason is that the public this time anticipated a lower rate of inflation than the
one actually chosen by the left-wing party since they gave some probability ρ to a
victory by the right-wing party. Hence, the central bank can surprise the electorate
with a high inflation and boost output. In case of a right-wing party win, the actual
level of inflation will be lower than the one anticipated by the public and output
will therefore decline, even if the right-wing party was previously in power. One
year after the election, output levels will be back at ȳ again.
In Figure 13.2, we show one example of such a political business cycle dynamic.
It shows the development of output and inflation when the economy transits from
y election year
time
t –2 t –1 t t +1 t +2
EQ,L
EQ,L
EQ,R
time
t –2 t –1 t t+1 t+2
Figure 13.2 Political business cycle effects over time on output (y) and inflation (π) of
an unexpected (ρ > 1/2) left-wing party election victory over an incumbent
right-wing party.
right-wing party rule to an unexpected election victory by the left. In period
t − 1 and earlier, output and inflation are ȳ and (y ∗ − ȳ)/a R , respectively. When
the election period comes, the probability of a right-wing victory is considered
large, ρ > 1/2, which is reflected in low inflationary expectations and a relatively
low πtEQ,L . Due to this surprise, there is a relatively large boom in output, which
increases by ρη > 0. In period t + 1, the economy has settled down at the natural
rate of output and the higher equilibrium level of inflation.
The key insight from this section is that even if the same party is in power before
and after an election and no fundamentals in the economy have changed, there will
still be a temporary cyclic effect on inflation and output that is fully explained by
the uncertainty surrounding the election itself.
13.5 The Taylor rule

In practice, central banks do not choose the inflation rate directly. They influence
inflation mainly through their determination of the nominal interest rate that other
banks in the country are charged when they borrow from the central bank. This
rate is usually referred to as the repo rate.7 The central bank repo rate, in turn,
influences interest rates in the financial sector as a whole and the rate of price
increases in the rest of the economy.
A potentially very simple solution to the time-inconsistency problem of mone-
tary policy would be for central banks to formulate simple rules for the setting
of their nominal interest rate (the repo rate) as a response to changes in out-
put and inflation levels. We saw, for instance, in Chapter 11 that central banks
were assumed to follow a monetary policy rule saying that they should increase
the real interest rate r when Y or π increased. If such a rule were consistently
followed, there would be no room for discretionary monetary policy and the
time-inconsistency problem would be reduced.
One famous suggestion for such a rule is provided by Taylor (1993). The Taylor
rule of monetary policy is defined as
i t = πt + rt∗ + aπ (πt − π ∗ ) + a y (yt − ȳ)
where i t is the nominal interest rate at time t, πt is the actual inflation rate as
before, π ∗ is the socially optimal inflation rate, rt∗ is the equilibrium real interest
rate, yt is the actual output level (in logs), and ȳ is the natural rate of output or
the level of potential output predicted by a linear trend. aπ and a y are the weights
given to inflation and output, respectively.
In Taylor’s original work, which was used for understanding United States mon-
etary policy, it was assumed that rt∗ = 2 percent, π ∗ = 2, and aπ = a y = 1/2.8 Using
these parameter values, Taylor rule becomes
πt − 2 yt − ȳ
i t = πt + 2 + +
2 2
3πt yt − ȳ
=1+ +
2 2
Hence, when both output and inflation are at their preferred levels (so that πt − 2 =
yt − ȳ = 0), then i t = 4 percent. If inflation rises by 1 percent above this level, i.e.
πt = 3, while the other values remain constant, the rule says central banks should
raise their nominal interest rate by 1.5 percent to 5.5 percent. Similarly, if output
rose so that it was 1 percent above its long-run trend (yt − ȳ = 1), then the nominal
interest rate should be raised by 0.5 percent. Taylor (1993) showed that this rule
was very close to the actual observed response of the Federal Reserve in the United
States to levels and changes in output and inflation.
13.6 Seigniorage
The time-inconsistency problem above is mainly meant to describe the situation
in advanced Western economies with benevolent monetary authorities and gov-

ernments. In countries with weak political institutions and autocratic rulers, it is
unlikely that the real objective function of the ruling elite has much in common
with the social loss function in the section above. In such countries, the ruling
elite often controls monetary authorities and demands money to be printed in
order to finance this or that expenditure. This practice, referred to as seignior-
age, frequently results in hyperinflation.9 This section will analyze the mechanics
of seigniorage.
Consider once again the money market equation
M
= L(r + π e , Y )
P
where we assume for simplicity that r and Y are unaffected by the money growth
rate and that π = π e . Also, as in Section 11.2, inflation equals the money growth
rate: π = MṀ
= gm .
Let us assume that the government now becomes involved in a very costly
endeavour for which it cannot easily get financing from the international com-
munity, such as a war.10 The only short-run solution to the financing difficulty
is to order the central bank to print money that the government can use to meet
its new expenditures. The increase in the nominal money stock from this money
printing is Ṁ. If we divide this by the aggregate price level, we get the level of
seigniorage S:
Ṁ Ṁ M M
S= = = gm (13.13)
P M P P
The reformulated expression on the right-hand side explains why seigniorage is
sometimes referred to as an inflation tax on real money balances, where the money
growth rate is the “tax” rate.
In a money market equilibrium, it must be the case that M/P = L(r + π e , Y ).
Inserting this term into (13.13) gives us
S = gm L(r + gm , Y ) (13.14)
S
gM
Figure 13.3 The inflation-tax Laffer curve.
The interesting aspect of this expression is that it shows that money printing will
have a negative impact on money demand since it will increase the nominal interest
rate i = r + gm . There is thus a trade-off: on the one hand, printing more money
increases government revenue; on the other hand, it decreases the “tax base” by
decreasing the amount of real money balances held in society. Taking derivatives
on the basis of (13.14) shows formally the positive and negative effects:
∂S
= L(r + gm , Y ) + gm L i (r + gm , Y )
∂gm (+) (−)
Note that the second term approaches zero as gm goes to zero, whereas L(r +
gm , Y ) is positive even if gm = 0. As gm rises, it is plausible that the second nega-
tive term will eventually dominate and make the whole expression negative. In that
case, the relationship between S and gm is shaped like an inverted “U”, as shown
in Figure 13.3. The curve in Figure 13.3 has been referred to as the inflation-tax
Laffer curve.11 A nonbenevolent government should thus create inflation up to the
maximum where ∂g∂ Sm = 0.12 Although this might seem like a cynical type of anal-
ysis, it is a well-known fact that hyperinflations are almost always directly caused
by governments and that they can be curbed relatively easy if there is a political
will to do so.
14 The Open Economy
The models analyzed so far have all assumed a closed economy, i.e. one where
trade does not occur. In this final chapter, we will analyze the effects of opening the
economy to trade and to capital flows across borders. In reality, total trade (exports
plus imports) often amounts to a substantial proportion of GDP, especially in small
countries.
In the sections below, we start by defining key concepts and relations such as the
current account, the balance of payments, and nominal versus real exchange rates.
We then introduce a representative agent framework with a utility-maximizing
individual who optimizes a consumption stream in a two-period model. The pur-
pose of this exercise is to show that a balanced current account might not be
optimal in all periods. We then discuss more traditional models in the Keynesian
tradition of sticky prices such as the Mundell–Fleming model, and the theory of
exchange rate overshooting. Finally, we also analyze the criteria for an optimal
currency union.
14.1 Open economy accounting

14.1.1 The current account
In the national accounts, total GDP on the user side is written as
Y =C + I +G + X − M
where X − M = N X represents net exports, or the current account balance. When

X > M, so that the country exports more than it imports, we have a current account
surplus, whereas X < M implies a current account deficit. A current account sur-
plus means that a country produces more than it uses up domestically, so that
Y > C + I + G. Such a country might be seen as being a net lender to foreigners,
whereas countries with current account deficits are net borrowers from the outside
world.
The current account therefore changes the definition of national savings. Let
total private savings in an economy be S P = Y − C − T , where T is total taxes,
and total public savings is S G = T − G; then total national savings are
S = S P + SG = Y − C − G = I + N X (14.1)
In an open economy, total savings might thus be equal to I , but can depend on net
exports.
Equation (14.1) can further be reformulated as
N X = S P − I − (G − T )
For a given level of S P − I , a large government budget deficit G > T is most

likely associated with a current account deficit N X < 0. This scenario, sometimes
referred to as twin deficits, has characterized the world’s largest economy, the
United States, during the last decade.
14.1.2 Balance of payments

The balance of payments shows all financial and trade transactions that occur
between a country (the home country) and the rest of the world. The current
account is one part of the balance of payments. The other part is the capital
account, which measures the net capital and financial flows between the home
country and the rest of the world. The capital account is affected positively if for-
eigners buy assets in the home country (capital inflow) and negatively if domestic
people buy assets abroad (capital outflow). By construction, we must have that
Balance of payments = Current account + Capital account = 0
Hence, a current account deficit, i.e. a situation where N X = X − M < 0, must

be balanced by a capital account surplus, i.e. that foreigners buy capital assets
in the home country to an extent such as the capital inflow exceeds the capital
outflow. The assets in questions are often financial assets like bonds. Hence, cur-
rent account deficits are typically financed by international net borrowing (selling
government bonds to foreigners), whereas a surplus would imply the opposite.
Let us denote a country’s net stock of accumulated foreign financial assets at
some year t to be Bt . The current account can then be described as
N X t = Bt+1 − Bt = Yt + rt Bt − Ct − It − G t (14.2)
If N X t > 0, then Bt+1 > Bt and the economy’s stock of foreign assets increases. If
we think of Bt as bonds, then the country is “paid” with foreign bonds to make up
for the difference between exports and imports with the rest of the world. If instead
imports exceed exports, so that N X t < 0, then the country needs to borrow in order
to import more than they export. In this case, the stock of foreign assets decrease.
The existing stock of foreign bonds at time t, Bt , earns an interest return of rt .1
The open economy 141
14.1.3 Exchange rates
The nominal exchange rate e shows the price of a unit of foreign currency in terms
of the home-country currency.2 A rise in the exchange rate thus means that the
foreign currency becomes more expensive, which means a weakening, or depre-
ciation, of the home currency. In an analogous manner, should e fall, that would
imply a strengthening, or appreciation, of the home currency compared with the
specific foreign currency in question.
Let the aggregate price level in the foreign country be P f and the aggregate
price level in the home country P, as before. Then the real exchange rate is
defined as
eP f
=
P
For example, consider a comparison between the Swedish krona and the euro. The
nominal exchange rate is roughly 10 (1 euro costs 10 kronor), whereas the price
level is perhaps 5 percent higher in the euro zone, on average. In that case, the real
euro/krona exchange rate will be 10 · 1.05 = 10.5.
A rising real exchange rate, i.e. a depreciation, means that foreign goods and
services are becoming more expensive relative to those produced in the home
country. Hence, consumers in the home country will likely substitute more expen-
sive foreign goods for cheaper home-produced goods. This will imply that net
exports rise, i.e. exports from the home country increase and imports decrease.
This relationship is an outcome of a more complicated theoretical exercise called
the Marshall–Lerner condition.3
14.2 A representative agent framework

So far, we have not said anything about the choices countries make regarding
levels and dynamics of the current account, etc. In order to do that, we need to
introduce micro foundations, i.e. assume something about the preferences of a
typical citizen.
Let us consider a simple two-period model where a representative individual
maximizes the utility function
U = u(c1 ) + βu(c2 )
where ct ≥ 0 is individual consumption.4 The function has the usual properties of

u (ct ) > 0, u (ct ) < 0, and β ≤ 1.
For the economy as a whole, the current account follows the basic dynamics as
described in (14.2). The investment equation is given by It = K t+1 − K t , where
K t is the capital stock at time t. We assume for simplicity that capital depreci-
ation is zero. Rearranging the capital account expression in (14.2) and using the
expressions for It , the change in total domestic wealth can be written as
Bt+1 + K t+1 − (Bt + K t ) = Yt + rt Bt − Ct − G t = St (14.3)

where St is total national savings, Yt is total output, rt Bt is net return on hold-
ing foreign assets, Ct is aggregate consumption, and G t is government spending.
Hence, total savings in this economy can be used for either domestic capital or
accumulating foreign financial assets.
Individuals start off with some capital endowment K 1 > 0. In our two-period
setting with selfish individuals who have no preference for leaving anything to
a future generation, we will assume that K 3 = 0. Hence, I2 = K 3 − K 2 = −K 2 .
Furthermore, we will have that B1 = B3 = 0, i.e. the economy starts without any
foreign assets and also ends without any. For simplicity, we imagine that the econ-
omy has only one (representative) individual, so that ct = Ct . Output is produced
by using capital according to a standard production function Yt = F(K t ) where
F (K t ) > 0 and F (K t ) < 0.
The current account equation (14.2) and the assumptions just mentioned imply
that
B2 = Y1 − C1 − I1 − G 1 = F(K 1 ) − C1 − (K 2 − K 1 ) − G 1
Inserting the assumed parameter values for t = 2 into (14.3) gives us
B3 + K 3 − B2 − K 2 = −B2 − K 2 = F(K 2 ) + r B2 − C2 − G 2
The intuition for the latter expression is simply that second-period consumption
will partly come from having sold off or “eaten” the capital and assets stocks
B2 and K 2 . If we isolate B2 in the equations above and combine them, we can
get an expression that implicitly defines the intertemporal budget constraint for a
representatitive individual in the economy:
C2 + G 2 − K 2 − F(K 2 )
F(K 1 ) − C1 − (K 2 − K 1 ) − G 1 =
(1 + r2 )
F(K 2 ) − G 2 C2 − K 2
F(K 1 ) − G 1 + = C1 + K 2 − K 1 +
(1 + r ) (1 + r )
By using this constraint, we can set up the representative agent’s Lagrange

optimization problem as
= u(C1 ) + βu(C2 )

F(K 2 ) − G 2 K 2 − C2
+ λ F(K 1 ) − G 1 + − C1 − K 2 + K 1 +
1+r 1+r
where we want to find the levels of C1 , C2 , and K 2 that maximize utility.

The first-order conditions are
∂
= u (C1∗ ) − λ = 0
∂C1
∂ λ
= βu (C2∗ ) − =0
∂C2 1+r
∗
∂ F (K 2 ) 1
=λ −1+ =0
∂ K2 1+r 1+r
The first two conditions can easily be combined to obtain the usual Euler equation
result that, optimally, u (C1∗ ) = β(1 + r )u (C2∗ ). Rearranging the last condition
gives the equally familiar result that F (K 2∗ ) = r , i.e. investments should be made
up to the point where the marginal product of capital equals the marginal cost r .
Without further simplications, the first-order conditions cannot be used for find-
ing explicit solutions. Let us therefore assume that β(1 +r ) = 1, which implies that
we get the consumption smoothing result of C1∗ = C2∗ = C ∗ . Inserting this solution
back into the intertemporal budget constraint gives us
(Y1 + K 1 − G 1 )(1 + r ) + F(K 2∗ ) − G 2 − r K 2∗

C∗ =
2+r
Optimal consumption thus increases with Y1 and K 1 and is crowded out by

government spending G 1 and G 2 .
Since we have now identified the optimal consumption path, we can also define
the optimal first-period current account level to be
N X 1∗ = B2∗ = Y1 − C ∗ − I1 − G 1 (14.4)
(Y1 + K 1 − G 1 )(1 + r ) + F(K 2∗ ) − G 2 − r K 2∗
= Y1 −
2+r
− (K 2∗ − K 1 ) − G 1
Note that we allow for the case that B2 < 0, which would imply a current account
deficit and that the country is a net lender in the first period on the international
market. In the second period, we have ruled out the possibility that the country
ends with either debts or positive assets, so N X 2∗ = B3 − B2∗ = −B2∗ = −N X 1∗ .
Hence, if there is a current account surplus in the first period (B2∗ > 0), there must
an equivalent deficit in the next (and last) period, and vice versa. What we are
analyzing is therefore only the dynamics of the current account, not its “final”
level.
To start with, a current account surplus (B2 > 0) is more likely if initial capital
and income K 1 and Y1 = F(K 1 ) are high. The derivative with respect to K 1 is
∂ N X 1∗ /∂ K 1 = [F (K 1 ) + 1]/(2 + r ) > 0, which is clearly positive. Hence, coun-
tries that are initially rich in capital are likely to “choose” a current account surplus
in the first period, i.e. to export more than they import. The intuition for this sur-
plus in national saving is really the same as in the permanent income model of
consumption: since individuals aim to smoothe consumption, they will save when
incomes are abnormally high and use up the savings when incomes are lower.
How about the pattern of government spending, G 1 and G 2 ? First, note that
if G 1 = G 2 = Ḡ, then ∂ N X 1∗ /∂ Ḡ = 0. The reason is that an increase in Ḡ would
decrease C ∗ by exactly the same amount and the two effects in (14.4) would cancel
each other out. If, however, it was known for instance that the government would
decrease only G 2 while keeping G 1 constant, then we find from the derivative that
∂ N X 1∗ /∂ G 2 = 1/(2 + r ) > 0, i.e. a decrease in G 2 should cause a current account
deficit in the first period. The intuition is that a decrease in G 2 would cause optimal
consumption C ∗ to rise in (14.4) while holding all other variables constant.
14.3 The Mundell–Fleming model

Open economy issues are also frequently analyzed in a Keynesian setting with-
out micro foundations. The Mundell–Fleming model, associated with economists
Robert Mundell and Marcus Fleming, introduces a balance of payments equation
into the standard IS–LM framework. The current account, or net exports, part of
the balance of payments equation can be written as a function
X − M = N X (, Y, Y f )
where is the real exchange rate, Y is home-country aggregate income, and Y f is

the income of the rest of the world.5 As already mentioned, N X should increase
with since foreign goods will then become relatively more expensive and boost
exports. N X should, however, fall with domestic income Y since a higher income
means that people spend more on imported goods. Analogously, if the foreign
income level increases, foreign demand for home-country goods will increase so
that exports are boosted, which leads to a rise in N X.
Assuming an open economy changes the nature of planned expenditures that
make up the IS equation. Equilibrium on the goods market now requires that
Y = E(Y, r, G, , Y f )
where and Y f are included since they influence the current account. The IS curve
will actually be flatter in the open economy scenario because national income Y
has both a positive effect through increased consumption and a negative effect
through the impact on imports.
Rather than an MP curve, let us now work with a traditional LM curve. Whereas
an MP curve includes a monetary policy rule r (Y, π) whereby the central bank’s
interest rate reacts to changes in income levels and inflation, we now simply
assume
M
= L(r, Y )
P
As will be shown below, r might respond to the balance of payments situation.
Monetary policy instead enters as changes in real money supply M/P. In this
model, we are not primarily interested in inflation and will assume π e = 0 for
simplicity so that the nominal interest rate i = r + π e equals the real one, i = r .
We will indeed assume throughout this section the aggregate price level P is sticky
in the short run.
As before, we have that L r < 0 and that L Y > 0. Hence, an increase in Y will
increase money demand. In order to preserve the money market equilibrium, we
must have that r increases so that money demand falls back again to equal M/P.
The LM curve will thus feature a positive association between r and Y .
The balance of payments equation is given by the current account plus the cap-
ital account. The level of the capital account, CA, depends primarily on the real
interest differential between the home country and the rest of the world r − r f .
There are two main scenarios for how the nominal interest rate differential will
matter. In the first scenario with perfect capital mobility between countries, there
can be no interest rate difference, i.e. r = r f . If there is imperfect capital mobility,
perhaps due to regulations in the international capital market, then r and r f need
not be exactly identical. In that case, the current account will generally be a
function of the interest rate differential C A(r − r f ), where
∂C A(r − r f )
= C Ar−r f > 0
∂(r − r f )
This means that capital will flow into the country if the interest rate domesti-
cally, r , is larger than that abroad, r f .
The current account and capital account equations taken together define the
balance of payments (BOP) curve:
N X (, Y, Y f ) + C A(r − r f ) = 0
In the case of perfect capital mobility, the BOP curve will be horizontal at r = r f .
With imperfect capital mobility, the BOP curve will be upward-sloping. To see
why, consider an increase in Y . Such an increase means that imports rise, which
leads to a fall in N X. In order to preserve the balance of payments “balanced” at
zero, this requires an offsetting rise in the capital account. Such a rise can only
come about through an increase in the domestic interest rate r . When the interest
rate rises, capital will flow into the country to finance to current account deficit.
Another implication is that there will be a positive association between r and Y .
We are now equipped with three curves: the IS curve (with net exports), the LM
curve, and the BOP curve. The curves form a flexible and useful framework for
understanding the effect of fiscal and monetary policy in an open economy setting.
Many different scenarios may be analyzed. For instance, let us consider the
case of perfect capital mobility, fixed exchange rates, and an expansive monetary
policy, as shown in Figure 14.1. The economy is initially in an equilibrium where
all curves cross. An increase in the nominal money stock (with prices staying
r
LM
LM′
IS
rf BOP
Y
Y0 = Y2 Y1
Figure 14.1 Expansionary monetary policy with fixed exchange rates and perfect
capital mobility.
constant) leads to a shift in the LM curve to the right. This will cause a fall in the
real interest rate below the international level r f . As a result, people will want to
buy foreign bonds and there will be a large capital outflow. Since the exchange
rate is supposed to be fixed, the central bank needs to purchase domestic currency
in order to keep it at its fixed level. This leads to a reduction in nominal money
supply, and the LM curve shifts back again to its initial position Y0 . In other words,
under perfect capital mobility and a fixed exchange rate, monetary policy will be
ineffective.
Let us now consider a quite different scenario with imperfect capital mobility,
fixed exchange rates, and a fiscal expansion. In this case, the BOP curve is upward-
sloping, as shown in Figure 14.2. The expansionary fiscal policy (increase in G)
causes the IS curve to shift to the right. This causes the interest rate to rise, which
in turn implies large capital inflows, i.e. foreign people buy domestic government
bonds. This puts pressure on the exchange rate to appreciate ( falls). In order to
keep the fixed exchange rate, the central bank must increase money supply, which
shifts the LM curve to the right. The end result is an increase in income from Y0
to Y2 , as shown in the figure. Fiscal policy will thus be relatively effective under
these circumstances, but it would have been even more effective if there had been
perfect capital mobility.
14.4 Exchange rate overshooting

In the Mundell–Fleming model above, prices did not adjust in the short run and
expectations about exchange rate movements did not play any role. The model by
r
LM
IS′ LM′
IS
BOP
Y
Y0 Y1 Y2
Figure 14.2 Expansionary fiscal policy with fixed exchange rates and imperfect capital
mobility.
Dornbusch (1976) shows that when expectations are taken into account under flex-
ible exchange rates, there might be overshooting in the adjustment of the exchange
rate which makes these extra volatile.
The model builds on two key premises. The first is that real interest rate differ-
entials between the home country and the rest of the world should be explained by
the following expression:
rt = r f + E t (et+1 − et ) (14.5)
The home-country nominal interest rate at time t should be equal to the interest
rate abroad r f plus the expected change in the exchange rate between t + 1 and t.
Should we, for instance, observe that rt > r f , then it must be the case that the
home currency is expected to depreciate so that E t (et+1 − et ) > 0.6 Relation (14.5)
shows the equality that must prevail if an investor is going to be indifferent about
investing a certain amount of money either in the home currency or in the foreign
currency and is often referred to as the uncovered interest rate parity.
The other basic premise is the standard expression for equilibrium in the money
market:
M
= L(rt , Yt )
P
The price level P is assumed to adjust only slowly, as in most Keynesian models.
To illustrate the main implications of the model, consider an expansionary
monetary policy from the central bank at time t that increases nominal money
supply M. With prices held constant in the short run, this means that real money
supply increases. The only way that this can be balanced in the short run is through
a fall in the real interest rate rt so that money demand rises.
Let us assume that initially there are no interest differentials between the
two countries. The expansion in nominal money then has two effects. The
lower interest rate in the home country should, according to (14.5), imply that
E t (et+1 − et ) < 0, i.e. the home currency is expected to appreciate. However, as
shown in the Mundell–Fleming framework in Figure 14.1, a monetary expansion
will lead to an outflow of capital due to the lower domestic interest rate and a
depreciation of the currency in the longer run. The only way to reconcile equilib-
ria in the two markets is if the exchange rate initially depreciates by more than
its long-run equilibrium value. This would allow it to appreciate over time so
that it eventually satisfies the uncovered interest rate parity condition. Therefore
the exchange rate overshoots its long-run equilibrium value and then gradually
appreciates.
It is important to note that this scenario will only occur if the aggregate price
level is sticky. It will not happen if prices immediately adjust to the increase in
money supply.
14.5 Currency unions

There are numerous currencies in the world. Some have been created very recently,
such as the euro, whereas others have a very long history. Are the borders of one
country always the natural boundary also for a currency? Since the newly created
euro area includes 17 countries, that question must be answered negatively. But
what, then, are the determinants of an optimally balanced currency area? Would it
be optimal if the whole world had the same currency?
In a famous early article, Mundell (1961) spelled out the necessary criteria for
a successful currency union:
• Labor mobility and an integrated labor market. If this is not the case, workers
will remain in depressed regions and expanding regions will be constrained by
lack of access to labor.
• Capital mobility and price and wage flexibility.
• A common risk sharing mechanism so that adversely affected regions are sup-
ported by the more fortunate ones. This might, for instance, take the form of
income redistribution of tax revenues.
• The regions included should have similar business cycles. If certain regions
have cycles that are very asymmetric compared with the majority of regions,
the adversely affected regions might have benefited from depreciation but are
unable to carry out any with a joint currency.
The literature in this area has recently been primarily concerned with the euro
area. Recent events during 2010 appear to confirm that certain countries (such as
Greece) clearly do not fit some of Mundell’s requirements very well. In general,
capital mobility across Europe is fairly good but labor mobility is still not substan-
tial. A common risk sharing mechanism is perhaps about to be created, but it is
doubtful whether the euro countries typically face symmetric shocks. Apart from
the economic criteria cited above, there might of course also be political reasons
why countries choose to form currency unions.
15 Mathematical Appendix
15.1 Introduction
This book requires that the student be familiar with certain tools of differential cal-
culus. Fortunately, it is sufficient to limit our analysis to functions of one variable.
The following concepts are discussed:
• derivatives of some basic functions

• simple differentiation rules
• chain differentiation
• implicit function differentiation
• applications to macroeconomics
The material of the appendix is rather standard. More detailed presentations can
be found in most first-year calculus textbooks.
15.2 Derivatives of some basic functions

Consider a real-valued function f (x) where x
∈ R. The derivative of f at some
point x 0 , which is denoted by f (x 0 ) or d df (x)
x x=x0 , is equal to the change in the

value of f if one infinitesimally changes x 0 . That is,
f (x) − f (x 0 )
f (x 0 ) = lim (15.1)
x→x0 x − x0
A function is called differentiable at x 0 if this limit exists.

Since on many occasions it could be quite troublesome to calculate such a limit,
a number of rules allow us to do so without actually having to use the formal
definition of the derivative.
Constant function: f (x) = c
The derivative of the constant function is
f (x) = 0 ∀ x ∈ R (15.2)
Mathematical appendix 151
Identity function: f (x) = x
f (x) = 1 ∀ x ∈ R (15.3)
Power: f (x) = x n
The exponent could in principle be any constant n. The derivative of this
function is given by
f (x) = nx n−1 ∀x ∈R (15.4)
Notice that the identity function is a special case of a power with n = 1. To verify
the previous rule, we substitute n = 1 in equation (15.4) and we obtain f (x) =
x 0 = 1.
Exponential function: f (x) = e x
The exponential is probably the most important function in calculus. Its base is
the irrational number e ≈ 2.7. It has the very interesting and unique1 property of
being neutral to differentiation. Namely,
f (x) = e x ∀x ∈R (15.5)
Logarithmic function: f (x) = ln x

The natural logarithm ln is the inverse of the exponential function. That is, if, for
some arbitrary x ∈ R, one obtains the value y = e x , then one could go backwards
by evaluating the logarithmic function at y, which would yield ln y = x. Notice that
the logarithmic function is defined only for positive real numbers. Its derivative is
given by the following equation:
1
f (x) = ∀ x ∈ (0, ∞) (15.6)
x
15.3 Differentiation rules

On many occasions a function has a rather complicated structure, which makes it
quite difficult, to calculate its derivative. In order to simplify the differentiation,
we apply a number of rules that break the initial function into smaller pieces that
can be handled more easily.
15.3.1 Sum of functions

Consider two functions f 1 (x) and f2 (x). The derivative of the function f (x) =
f1 (x) + f 2 (x) is given by the following equation:
f (x) = f 1 (x) + f 2 (x) (15.7)
This rule can be generalized to an arbitrary finite sum of functions. Consider,

for example, the function f (x) = x 2 + e x . Obviously this is the sum of f 1 (x) = x 2
152 Mathematical appendix
and f 2 (x) = e x . The derivatives of both f 1 and f 2 can be easily calculated.
Therefore the derivative of f would be given by f (x) = 2x + e x .
15.3.2 Multiplication with a scalar

Consider a function g and let the function f be defined as f (x) = cg(x), where
c ∈ R. The derivative of f is given by the following equation:
f (x) = cg (x) ∀x ∈ A (15.8)
Take, for example, the function f (x) = 4x 3 . It is easy to see that f (x) =
4 · 3x 3−1 = 12x 2 . Setting c = −1 and combining the two rules (sum of func-
tions and multiplication with a scalar) entails a similar rule for the difference
f = f 1 − f 2 , namely, f (x) = f 1 (x) − f 2 (x).
15.3.3 Product of functions

Consider two functions f 1 (x) and f 2 (x). The derivative of the function f (x) =
f 1 (x) f 2 (x) is given by the following equation:
f (x) = f 1 (x) f 2 (x) + f 1 (x) f 2 (x) (15.9)
Consider, for example, the function f (x) = xe x . The derivative of f is given by

f (x) = (x)e x + x(e x ) = e x + xe x .
15.3.4 Division of functions

Consider two functions f 1 (x) and f 2 (x). The derivative of the function f (x) =
f1 (x)
f (x) is given by the following equation:
2
f 1 (x) f 2 (x) − f 1 (x) f 2 (x) f1 (x) f 1 (x) f 2 (x)

f (x) = = − (15.10)
[ f 2 (x)]2 f2 (x) [ f 2 (x)]2
Notice that the function is defined only for those x that do not make f 2 equal to 0,
since otherwise the division would not be defined in the first place. An example of
x 2 +1 2 +1) x 2 −2x−1
such a function is f (x) = x−1 , which yields f (x) = 2x(x−1)−(x2 = ,
(x−1) x 2 −2x+1
for every x = 1.
15.4 Chain differentiation

All the previous rules involved functions that could be separated into simple ones.
However this is not always the case. Take, for instance, the function h(x) = e x +1 .
2
This is a compound function in the sense that h is a function of some other func-
tion g, which is a function of x. That is to say, one could set g(x) = x 2 + 1 and
f (y) = e yand by plugging g(x) into y obtain h(x) = f (g(x)) = e g(x) . Since g is
depends on x and f depends on g, it is straightforward that f depends on x.
Formally, consider two functions g(x) and f (g(x)). We define the compound
function as follows:
h(x) = f (g(x)) (15.11)
How would we then differentiate h(x) with respect to x? Intuitively, we would

like to see how much h would change due to an infinitesimal change in x. A slight
change in x would trigger a change in g(x) of the size of the derivative of g. At the
same time, changing g(x) would lead to a change in f that depends on g. Thus,
the initial change in x has triggered a chain reaction. Through the effect on the
intermediate function, it has caused a change in f . The size of this change would
be equal to the size of the change in g times the size of the change that g has
caused to f . That is,
d f (g(x)) d f (g(x)) dg(x)

h (x) = = = f (g(x))g (x) (15.12)
dx dg(x) dx
2 +1
then h (x) = 2xe x
2 +1
For example, if h(x) = e x .
15.5 Implicit function differentiation

Consider a function f (x) and assume that its exact formula is not given explicitly.
All that is available is a functional relationship of the form [ f (x)]2 + f (x) − x +
1 = 0. Since it is not that easy to first solve this equation with respect to f (x) and
then differentiate it in x, we proceed by taking the derivative on both sides of the
equation, with the use of the chain rule. This gives 2 f (x) f (x) + f (x) − 1 = 0,
which is linear in f (x). Hence f (x) = 2 f (x)+1
1
.
15.6 Applications to macroeconomics

The basic task of macroeconomic theory is to study the relationship between cer-
tain measures and how they evolve over time. For that reason, we are usually
interested in growth rates.
15.6.1 Growth rate of a multiplicative function

Quite often, we have functions of the form
Y (t) = X (t)Z (t)

154 Mathematical appendix
To find the growth rate of Y (t), we need to use both the product rule and the chain
rule:
dY (t) 1 Ẏ (t) Ẋ (t)Z (t) + X (t) Ż (t)

= =
dt Y (t) Y (t) Y (t)
Ẋ (t)Z (t) + X (t) Ż (t) Ẋ (t) Ż (t)
= = +
X (t)Z (t) X (t) Z (t)
15.6.2 Growth rate as the derivative of the logarithmic function

Measures like income growth rate or inflation rate embody the concept of relative
change. In discrete time, these are usually expressed by a percentage, but since
time is continuous we are forced to go to the limit and examine their relative
change in an infinitesimally small time interval.
Consider, for instance, a variable X (t) as a function of time. The instantaneous
relative change is given by the growth rate of X with respect to time. Denoting the
derivative2 at t by Ẋ (t), we can write the growth rate as a fraction ẊX (t)
(t) .
Remember from the previous sections that the derivative of the logarithmic
function is given by d(ln X)
d X = X . Then taking the derivative of ln X (t) with respect
1
to t and applying the chain rule, we obtain
d ln X (t) 1 d X (t)
= (15.13)
dt X (t) dt
Using the appropriate notation, we can rewrite (15.13) as follows:
d ln X (t) Ẋ (t)
= (15.14)
dt X (t)
Thus, the growth rate of X (t) is nothing but the derivative of the logarithmic
function.
15.7 Basic properties of exponents and logarithms

15.7.1 Exponents
1. x 0 = 1, x = 0
2. x 1 = x
3. x n x m = x n+m
n
4. xxm = x n−m , x = 0
5. (x n )m = x nm
6. (x y)n = x n y n
xn
7. yn = ( xy )n , y = 0
15.7.2 Logarithms
1. ln 1 = 0
2. ln e = 1
3. ln x + ln y = ln(x y)
4. ln x − ln y = ln( xy )
5. ln(x c ) = c ln x
6. x = eln x
Notes
Preface
1 Any macro model that, for instance, assumes away unemployment (like many DGE
models routinely do) seems to have missed something important about how economies
work.
1 Introduction
1 GDP does not include the production of intermediate goods or services, nor production
on the black market. The market value of public services like health care are further
very imprecisely measured in the national accounts.
2 An individual’s purchase of a car for personal use counts as consumption in the national
accounts, although it might intuitively be seen as an investment.
2 The Malthusian World

1 See Clark (2007) for an extensive discussion of these relationships.
2 It can be shown that the second-order condition for a maximum is also satisfied since
∂ 2 ln Ut /n 2t < 0.
3 Note also that this expression is qualitatively similar to the population dynamics
equation (2.4).
4 In particular, the Fertile Crescent hosted the wild progenitors of goats, pigs, cattle, and
horses among the animals, as well as wheat and barley among the plants. Other regions,
such as Australia, had no domesticable species and hence remained at Stone-Age level
before European contact.
5 In the Fertile Crescent, the transition to agriculture happened around 8500 BC.
6 The existence of such a trade-off was originally suggested by Becker and Lewis (1973).
7 Time spent on providing the basics for one child τ is supposed to be fixed.
8 The only slight complication is that the e(gt+1 ) curve is drawn in (gt , et ) space rather
than in (et , gt ) space. It still satisfies the specific partial derivatives referred to above.
3 The Solow Growth Model

1 We will refer to the long run as being a time horizon that is at least longer than the
duration of a typical business cycle of about five years.
2 An endogenous variable is one that is explained by the model itself, whereas an
exogenous variable is one that we take as given and that is not explained by the model.
3 The function usually also includes the level of technological knowledge, A(t). We will
introduce A(t) below.
4 This is sometimes expressed by saying that the production function exhibits homogene-
ity of degree 1.
5 In the overlapping generations model below, we will instead assume discrete rather than
continuous time.
Notes 157
6 To check that the system is stable, try a level of k to the left of k ∗ . At this level, s f (k)
will exceed (δ + n)k, so that k̇ > 0 and we are moved to the right to an increased level
of k. The contrary occurs to the right of k ∗ . Only at k ∗ will the system reach a stable
point.
7 For a similar discussion, see Romer (1994).
8 This feature is sometimes also referred to as “Harrod neutrality”. Other assumptions
regarding technology are sometimes made, but will not be discussed here.
9 We will discuss the nature of technology further below.
4 Endogenous Growth Theory

1 Remember that for any variable x(t), it will be the case that xẋ = ∂ ∂tln x .
2 This section on the product variety model and the next section on the Schumpeterian
growth model follow a structure that is similar to the equivalent sections in Barro and
Sala-i-Martin (2004).
3 In general, in this type of model, we might view the competitive and monopoly prices
P ∗ ∈ {1, 1/α} as the two extremes in a range of levels of competition.
4 This is clearly a great simplification. In reality, R&D is usually a very risky enterprise
that frequently produces no useful results.
5 In most endogenous growth models, T is assumed to be infinite. In reality, patents
typically last for about 20 years.
6 Aghion and Howitt (1992) assume that the arrival of new innovations follows a Poisson
process. Here we will leave that process undefined.
7 The inclusion of a human capital stock is a slight departure from the model in Acemoglu
et al (2003) but in line with their general argument.
5 The Overlapping Generations Model

1 Hence, for instance, c2,t+1 is consumption by old individuals during period t + 1.
2 This implicitly assumes the presence of perfect financial markets. We will return to this
issue below.
3 This intertemporal budget constraint often enters with a “≤” sign so that consump-
tion should not exceed lifetime wage incomes. However, since we have ruled out any
bequests, the condition will be satisfied with equality.
4 This can be shown mathematically by using l’Hôpital’s rule.
5 We could also simply have inserted the expression for c2,t+1 in (5.8) into (5.6) and
reached the same result by solving for ∂/∂c1,t = 0.
6 The famous Ramsey (1928) growth model is usually set in a continuous-time frame-
work and involves dynamic programming. See Barro and Sala-i-Martin (2004) for an
extensive treatment of this model.
7 Recall that in the Solow model, we used an expression K̇ t = sY − δ K . The equivalent
expression in discrete time would be K t+1 − K t = sY − δ K t , where sY is total savings
in the economy. Total savings in the model shown here is st L, and we implicitly assume
complete depreciation in each period so that δ = 1. Hence, K t+1 = st L.
6 Equilibrium Business Cycles

1 A “representative” firm should be thought of as characterizing the ideal type of firm
behavior, just like a representative individual’s utility function represents an ideal type
of behavior from the viewpoint of economic theory. Needless to say, by no means all
firms are perfect profit maximizers in reality.
2 The second-order condition for maximum is fulfilled if the second derivative satisfies
∂2
∂ L2
< 0, which is indeed true in this case.
t
158 Notes
(1−α) A1−α 1 (1−α) A1−α 1
3 Aggregate demand for labor will be i L i,D = ( wt t ) α i K ti = ( wt t ) α K t ,
where K t is the total capital stock in the economy.
4 We might have done formal comparative statics on (6.11) by using the implicit function
rule. However, we will stick to a more informal approach here.
7 Financial Crises
1 A famous recent example of an individual failure of a financing institution is of course
the collapse of Lehman Brothers in September 2008.
2 Recent examples include Sweden in 1991 and Iceland in 2008. With some exceptions,
the financial crisis of 2008 did not turn into a general bank run among traditional banks,
mainly because of various lender insurance schemes supported by governments.
3 One might think of the utility function as describing an individual who is impatient with
a probability ρ and impatient with a probability 1 − ρ.

4 Asterisks (∗ ) will henceforth denote socially optimal levels.
5 The calculation involves quite a bit of algebra and is left for the student as an exercise.
The full solution is available upon request.
6 As shown by Chang and Velasco (2001), this result will not apply if the debt incurred
in period 0 is long-term. Hence, the longer the duration of loans, the less fragile is the
system to depositor panics.
8 Consumption and Saving

1 Formally, this result is given by the partial derivative ∂C∂τ
t
= −cmpc Yt < 0, which implies
that a decrease in τ will increase Ct .
2 This assumption will be relaxed in a section below.
3 Note that there will be T + 1 time periods since we also count period 0.
4 Negative savings implies that the individual either borrows money with future income
as collateral, or uses up accumulated savings.
5 Such savings are often compulsory and take the form of funds for retirement.
6 For an overview of empirical regularities regarding lifetime consumption patterns, see
Attanasio (1998).
7 Another interesting prediction from the PIH is that an individual’s consumption sensi-
tivity to unexpected changes in income should decrease with T , i.e. older people (with
a smaller T ) should be more prone to immediately spend windfall gains from lotteries,
for instance.
8 In older models, it was often assumed that people formed expectations about future
incomes, or about economic phenomena in general, mainly on the basis of their
observations of past behavior, so-called adaptive expectations.
9 For instance, assuming a random variable X t , E t (X t ) = X t since X t is actually observed
at t and no expectations need to be made. Furthermore, the expectation of some constant
η simply equals the constant: E t (η) = η.
10 This first-order, autoregressive random-walk process might be compared with the
AR(1) process for technology in the RBC model. In Hall’s model, the equivalent of
ρ A is equal to 1 so that the shock is only felt one period ahead.
11 One such change in expectations might, for instance, concern y2 , which has been real-
T
ized in period 2 but which was still uncertain in period 1. Note that t=2 E 2 (yt ) =
T
y2 + t=3 E 2 (yt ).
12 Such a slowly rising level of consumption is actually well in line with empirical studies
on the life-cycle hypothesis.
13 The idea that relative consumption should influence individuals in their decision-
making goes back at least to Duesenberry (1952).
Notes 159
14 Time-independent utility from consumption means that the marginal utility of Ct is
independent of C j =t . In the setting above, the marginal utility of C1 will depend on the
level of C2 .
9 Investment and Asset Markets

1 In an economy, there are usually also human capital investments, for instance through
education, and social capital investments, which involve bonding with other people for
some economic purpose.
2 According to the “accelerator” model, we have that K t = κYt , where κ > 0 is a constant
“capital–output ratio”. From this relationship, it follows that if Yt increases by Yt , then
K t will also increase, i.e. there will be investment: It = κYt .
3 The section below is similar to the analysis in Sørensen and Whitta-Jakobsen (2005,
Chapter 15).
10 Unemployment and the Labor Market

1 Recall that labor demand is implicitly derived from the firm’s standard first-order
condition for profit maximization where the marginal product equals the wage rate.
2 An alternative profit function would be that firms were maximizing expected profits and
hence would only choose one level of L O and w I based on the expected outcome.
3 The same basic search and matching model more or less also applies to the marriage
market.
11 IS–MP, Aggregate Demand, and Aggregate Supply

1 It is sometimes assumed that investment increases with income Yt , but we do not employ
that assumption here.
2 It is assumed that people earn no interest on ordinary bank deposits, only on long-term
investments.
3 For instance, European countries in the euro area have laws stipulating that central
banks should be conducted with the primary aim of holding inflation low. In Chapter 13,
we discuss one such central bank rule – the Taylor rule – which has been empirically
observed in the United States.
4 In older textbooks, the AD curve is usually drawn in with some price index P on the
vertical axis. We will, however, alter this convention here and instead have the change
in the price level, i.e. inflation, on the vertical axis.
5 It is questionable how empirically relevant this assumption about an unobservable
aggregate price level was in the year 2012.
6 In equilibrium, investment demand (I ) will therefore be equal to savings (S), which is
the original reason for the acronym IS.
7 Such a reassessment appears to have been one of the reasons for the credit crunch in the
United States in 2008.
8 Note that we thus implicitly assume imperfect competition since prices are set above
marginal cost.
9 For an overview of this literature, see Romer (1993).
12 Public Finance and Fiscal Policy

1 In the next chapter (Section 13.6), we will specifically study money printing, or
seigniorage, and the effects that it might have on the economy.
2 See Barro (1989) for an extensive discussion.
3 See Reinhart and Rogoff (2009b) for an account of such historical episodes.
160 Notes
13 Inflation and Monetary Policy
1 Notions of a positive association between the money stock and inflation appear to date
back to ancient times and at least to Copernicus in the 1500s.
2 This is sometimes referred to as the Fisher effect.
3 π ∗ is normally around 2 percent in the Western world.
4 The second-order condition for minimum is satisfied since ∂ 2 V /∂π 2 > 0.
5 The exposition below is inspired by Alesina and Stella (2010).
6 This is formally not possible in the case of fully independent central banks.
7 This is technically speaking the interest rate that is used when central banks con-
duct open market operations with the rest of the banking system through short-term
repurchase agreements. We will not discuss how this mechanism works in this text.
8 This is actually close to what also applies in the euro area. The inflation target for the
European Central Bank is 2 percent, for instance.
9 A hyperinflation is sometimes defined as being in place if the inflation rate exceeds

50 percent per month. In most hyperinflation events, inflation is much higher.
10 Wars have been a recurring cause of hyperinflations throughout history.
11 The original Laffer curve depicts the inverted “U” relationship between tax revenue and
tax rates.
12 Empirical research has shown that the maximum level of S is found at inflation rates of
roughly 200–300 percent (Romer 2005).
14 The Open Economy

1 The sum of GDP and net international factor payments Yt + rt Bt is defined as the gross
national income.
2 For example, the USD exchange rate in terms of the Swedish krona (SEK) is roughly
6.50 at the time of writing (September 2011).
3 Formally, the Marshall–Lerner condition shows that ∂ ∂ NX
> 0 if the absolute value of the
sum of the price elasticities of exports and imports is more than one.
4 The exposition below is similar to that in Obstfeld and Rogoff (1999, Chapter 1).
5 It is convenient to assume that the world consists of only two countries.
6 Countries with weak currencies will thus generally have higher nominal interest rates.
We assume that the international interest rate r f is time-invariant.
15 Mathematical Appendix
1 Actually the whole family of exponential functions f (x) = cex , c ∈ R, has this property.
2 In macroeconomics, it is conventional to use this notation.
Bibliography
Acemoglu, D. and D. Autor (2009) Lectures in labor economics. Unpublished teaching

material, http://econ-www.mit.edu/faculty/acemoglu/courses, accessed May 10, 2011.
Acemoglu, D., P. Aghion, and F. Zilibotti (2003) Vertical integration and distance to
frontier. Journal of the European Economic Association, 1(2–3), 630–638.
Acemoglu, D., S. Johnson, and J. Robinson (2005) The rise of Europe: Atlantic trade, insti-
tutional change, and economic growth. American Economic Review, 95(3), 546–579.
Aghion, P. and P. Howitt (1992) A model of growth through creative destruction. Econo-
metrica, 60(March), 323–406.
Alesina, A. and A. Stella (2010) The politics of monetary policy. NBER Working Paper
15856, NBER.
Alesina, R. and G. Tabellini (1990) A positive theory of fiscal deficits and government debt.
Review of Economic Studies, 57, 403–414.
Ashraf, Q. and O. Galor (2010) Dynamics and stagnation in the Malthusian epoch.
American Economic Review, forthcoming.
Attanasio, O. (1998) Consumption demand. NBER Working Paper 6466, NBER.
Barro, R.J. (1974) Are government bonds net wealth? Journal of Political Economy,
82(November), 1095–1117.
Barro, R.J. (1979) On the determination of the public debt. Journal of Political Economy,
87(October), 940–971.
Barro, R.J. (1989) The Ricardian approach to budget deficits. Journal of Economic
Perspectives, 3(2), 37–54.
Barro, R.J. and X. Sala-i-Martin (2004) Economic Growth, 2nd edition. Cambridge, MA:
MIT Press.
Becker, G. and H.G. Lewis (1973) On the interaction between the quantity and quality of
children. Journal of Political Economy, 81(2), 279–288.
Besley, T. and T. Persson (2010) State capacity, conflict, and economic development.
Econometrica, 78(1), 1–34.
Blanchard, O. (1985) Debt, deficits, and finite horizons. Journal of Political Economy,
93(2), 223–247.
Bowman, D., D. Minehart, and M. Rabin (1999) Loss aversion in a consumption-savings
model. Journal of Economic Behavior and Organization, 38, 155–178.
Branson, W.H. (1989) Macroeconomic Theory and Policy, 3rd edition. New York: Harper &
Row.
Caballero, R. (1997) Aggregate investment. NBER Working Paper 6264, NBER.
Chang, R. and A. Velasco (2001) A model of financial crises in emerging markets.
Quarterly Journal of Economics, 116(2), 489–517.
Clark, G. (2007) A Farewell to Alms. Princeton, NJ: Princeton University Press.
Diamond, D.W. and P.H. Dybwig (1983) Bank runs, deposit insurance, and liquidity.
Journal of Political Economy, 91(3), 401–419.
162 Bibliography
Diamond, J. (1997) Guns, Germs and Steel: The Fates of Human Societies. New York:
Norton.
Diamond, P. (1965) National debt in a neoclassical growth model. American Economic
Review, 55(5), 1126–1150.
Diamond, P. (1982) Aggregate demand management in search equilibrium. Journal of
Political Economy, 90(5), 881–894.
Dornbusch, R. (1976) Expectations and exchange rate dynamics. Journal of Political
Economy, 84(6), 1161.
Duesenberry, J. (1952) Income, Saving, and the Theory of Consumer Behavior. Cambridge,
MA: Harvard University Press.
Frederick, S., G. Loewenstein, and T. O’Donoghue (2002) Time discounting and time
preference: A critical review. Journal of Economic Literature, 40, 351–401.
Friedman, M. (1957) A Theory of the Consumption Function. Princeton, NJ: Princeton
University Press.
Friedman, M. (1968) The role of monetary policy. American Economic Review, 58(March),
1–17.
Gali, J. and M. Gertler (2007) Macroeconomic modeling for monetary policy evaluation.
Journal of Economic Perspectives, 21(4), 25–46.
Galor, O. and D. Weil (2000) Population, technology and growth: From the Malthusian
regime to the demographic transition. American Economic Review, 110, 806–828.
Hall, R.E. (1978) Stochastic implications of the life cycle-permanent income hypothesis.
Journal of Political Economy, 86(6), 971–987.
Hibbs, D.A. (1977) Political parties and macroeconomic policy. American Political Science
Review, 71(4), 1467–1487.
Jones, C.I. (1995) R&D-based models of endogenous growth. Journal of Political Econ-
omy, 103(August), 759–784.
Krugman, P. (1988) Financing vs forgiving a debt overhang. Journal of Development
Economics, 29, 253–268.
Kydland, F. and E. Prescott (1977) Rules rather than discretion: The inconsistency of
optimal plans. Journal of Political Economy, 85(June), 473–492.
Kydland, F. and E. Prescott (1982) Time to build and aggregate fluctuations. Econometrica,
50, 1345–1370.
Laibson, D. (1997) Golden eggs and hyperbolic discounting. Quarterly Journal of Eco-
nomics, 112(2), 443–477.
Lindbeck, A. and D. Snower (1986) Wage setting, unemployment, and insider-outsider
relations. American Economic Review, 76(2), 235–239.
Long, J.B. and C.I. Plosser (1983) Real business cycles. Journal of Political Economy, 91,
39–69.
Lucas, R.E. (1972) Expectations and the neutrality of money. Journal of Economic Theory,
4(April), 103–124.
Lucas, R.E. (1973) Some international evidence of output-inflation tradeoffs. American
Economic Review, 63(3), 326–334.
Malthus, T. (1798) An Essay on the Principle of Population. London.
Mankiw, G. (1985) Small menu costs and large business cycles: A macroeconomic model
of monopoly. Quarterly Journal of Economics, 100(May), 529–539.
Mankiw, G., D. Romer, and D. Weil (1992) A contribution to the empirics of economic
growth. Quarterly Journal of Economics, 107(2), 407–437.
Mortensen, D. and C. Pissarides (1994) Job creation and job destruction in the theory of
unemployment. Review of Economic Studies, 61(3), 397–415.
Mundell, R. (1961) A theory of optimum currency areas. American Economic Review,
51(4), 657–665.
Obstfeld, M. and K. Rogoff (1999) Foundations of International Macroeconomics,
Cambridge, MA: MIT Press.
Bibliography 163
Olsson, O. and D.A. Hibbs (2005) Biogeography and long-run economic development.
European Economic Review, 49(4), 909–938.
Phelps, A.W. (1968) Money-wage dynamics and labor market equilibrium. Journal of
Political Economy, 76(July), 678–711.
Phillips, A.W. (1958) The relationship between unemployment and the rate of change of
money wages in the united kingdom, 1861–1957. Economica, 25(November), 283–299.
Ramsey, F. (1928) A mathematical theory of saving. Economic Journal, 38(152), 543–559.
Rebelo, S. (1991) Long-run policy analysis and long-run growth. Journal of Political
Economy, 96(June), 500–521.
Rebelo, S. (2005) Real business cycle models: Past, present, and future. NBER Working
Paper 11401, NBER.
Reinhart, C. and K.S. Rogoff (2009a) The aftermath of financial crises. American Economic
Review: Papers and Proceedings, 99(2), 466–472.
Reinhart, C. and K.S. Rogoff (2009b) This Time Is Different? Eight Centuries of Financial
Folly. Princeton, NJ: Princeton University Press.
Rogoff, K. and A. Sibert (1985) The optimal degree of commitment to an intermediate
monetary target. Quarterly Journal of Economics, 100(4), 1169–1189.
Romer, D. (1993) The new Keynesian synthesis. Journal of Economic Perspectives, 7(1),
5–22.
Romer, D. (2005) Advanced Macroeconomics Macroeconomics. Boston: McGraw-Hill.
Romer, P. (1986) Increasing returns and long-run growth. Journal of Political Economy,
94(October), 1002–1037.
Romer, P. (1990) Endogenous technological change. Journal of Political Economy,
98(October), S71–S102.
Romer, P. (1994) The origins of endogenous growth. Journal of Economic Perspectives,
8(Winter), 3–22.
Schumpeter, J. (1934) The Theory of Economic Development (translated from the German
version published in 1912). New York: Oxford University Press.
Shapiro, C. and J.E. Stiglitz (1984) Equilibrium unemployment as a worker discipline
device. American Economic Review, 74(3), 433–444.
Solow, R. (1956) A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70(February), 65–94.
Sørensen, P.B. and H.J. Whitta-Jakobsen (2005) Introducing Advanced Macroeconomics:
Growth and Business Cycles. Maidenhead: McGraw-Hill.
Taylor, J.B. (1993) Discretion versus policy rules in practice. Carnegie-Rochester Confer-
ence Series on Public Policy, 39, 195–214.
Woodford, M. (2010) Financial intermediation and macroeconomic analysis. Journal of
Economic Perspectives, 24(4), 21–44.
Index
Please note: page numbers in bold refer to figures.
Acemoglu, D. 95 budget constraint 61; illiquidity 58,

Acemoglu, D., et al 15, 40–1, 157n7 63, 64; total liquidation value 61
adaptive expectations 158n8 Barro, R.J. 116–7, 117, 157n2, 157n6,
adjustment costs 83–4; types 84–5 159n2
aggregate demand 101–6; and inflation Becker, G. 156n6
126; IS–MP model 105–6, 106; behavioral economics xii
Keynesian model 101–3, 102 Besley, T. 123
aggregate price level 114, 148, 159n5 Beveridge curve, the 97, 98
aggregate private consumption 2, 65 binding inflation commitments 128
aggregate production function 8 births: fall in 14; and income 11–4;
aggregate supply 101, 106–10; Malthusian model and 9–11, 10
expectations-augmented Phillips Blanchard, O. 42
curves 109; and inflation 126; bonds 116, 116–7, 140
Keynesian formulations 106–7; Lucas Botswana 25
supply curve 109–10; the Phillips Bowman, D., et al 77
curve 107–9, 108 Branson, W.H. 80
Aghion, P. 29, 30, 32, 39–40, 157n6 budget constraint: housing market 85–6;
agricultural revolution 15, 156n5 intertemporal 43–4, 46, 60, 142,
AK model 29–30 157n3; lifetime 66–7, 78–9;
Alesina, A. 126, 131, 160n5 random-walk model 72
Alesina, R. 119 budget deficit 115, 140; dynamics of
appreciation 141 115–6, 118–21; financing 116;
Ashraf, Q. 11–2, 14 primary deficit 116; Ricardian
asset market see investments equivalence analysis 116–7
Attanasio, O. 158n6 business cycles xi, 53, 58, 148 see also
autonomous spending 102 equilibrium business cycles
Autor, D. 95
Caballero, R. 80
balance of payments 140; curve 145–6, capital: accumulation of physical 19–20;
146, 147; Mundell–Fleming model depreciation cost 82; and labor rivalry
144–6 31; user cost of 80, 81–2, 83
balance of trade, deficits 2 capital account 140
bank run equilibrium 61–2 capital accumulation, Solow growth model
bank runs 58, 59, 63, 64, 158n2 19–20, 24
banking regulations 113 capital depreciation 81, 82
banks: financial crises model 59–61; capital inflow 140
financial liberalization 64; hard capital mobility 145–6, 146, 147, 148
Index 165
capital outflow 140 71; precautionary saving 73–5, 74, 75;
capital per worker 20; AK model 30 random-walk model 71–3; rational
capital stock 2 expectations 71; rational expectations
central banks: aims 128, 159n3; best paradigm 65; relative consumption 65,
response (or reaction) function 129; 77–8; and savings 65–79; time
discretionary monetary policy 129; inconsistency 65, 78–9; utility
independent 132; and political maximization 67–8
business cycles 135; random output convergence 19, 24–5
shocks 131; repo rate 136–7 creative destruction 39
certainty equivalence 72, 74 credit market, perfect 67
chain differentiation 152–3 credit spread 111–4, 112, 113
Chang, R. 58, 158n6 credit supply 111–4, 111, 112, 113
child quality 16–8, 17 CRRA utility function 45–7, 75–6
child quantity 16–8 currency, price levels 141

China xi, 25 currency union 139, 148–9
Clark, G. 7 current account balance 139
Cobb–Douglas functional form 8, 48; current accounts 1–2, 139–40, 140; deficits
investments 82; Solow growth model 139, 140, 143; representative agent
20, 22–4, 23; technology shocks and framework 141–4; surpluses 139,
53–4 143–4
colonialism 15
competition, imperfect 114 deadweight losses 117
compound function 152–3 deaths, Malthusian model and 9–11, 10
Congo, Democratic Republic of 25 debt forgiveness 124
constant function 150 debt overhang 122–4
constant relative risk aversion (CRRA) deficit bias 119
45–7 delegation of monetary authority 132
constant returns assumption 20 demand 101–6; and inflation 126; IS–MP
consumption: aggregate private 2, 65; model 105–6, 106; Keynesian model
autonomous 101–2; autonomous level 101–3, 102; price elasticity 35–6
of 65; change in expectations 158n11; depreciation 141
growth rate 46–7; individual 12, 43–4, Diamond, D. W. 58
46, 55–6, 156n2; Keynesian Diamond, J. 14–5, 95
consumption function 101–3; the Diamond, P. 42, 88
Keynesian cross 102–3; lifetime 67; differentiation rules 151–2
logarithmic utility 44–5; marginal diminishing returns, law of 7, 7–9, 20
propensity to consume (MPC) 101–2; “discrete Ramsey” result 47
mid-term analysis 69–70; the discrete time 42
multiplier effect 102–3; optimal 143; discretionary monetary policy 128,
optimal intertemporal growth rate 49; 129–31, 130
optimal paths 76, 76; R&D and 49; Dornbusch, R. 147
relative 65, 77–8, 158n13; stochastic Duesenberry 158n13
component 73; time distribution 70, Dybwig, P.H. 58
71; time-independent utility from dynamic general equilibrium (DGE)
159n14; utility maximization 67–8 tradition xi, xii
consumption sensitivity 158n7
consumption smoothing 68 economic institutions, role of 15
consumption theory: Friedman’s critique economic output, technology shocks and
66; interest rates and time discount 53–4
rates 75–6, 76; Keynesian aggregate education 9, 14; Galor–Weil model 16–8,
consumption 70–1, 71; Keynesian 17; and human capital 26; and
consumption function 65; permanent technological progress 16–8, 17
income hypothesis (PIH) 66, 66–71, effective labor 25–6
166 Index
efficiency wage theory 88, 89, 89–90; liberalization 64; foreign credit 62–3;
Shapiro–Stiglitz model 90–4, 91 short-term debt 63
elasticities 8 financial crisis, 2008 xi, 58, 113–4, 121,
elections 133–6, 135 158n1, 158n2
empirical tests xii financial intermediation, IS–MP model
employment level 93 110–4, 111, 112, 113
endogenous fertility 11–4 financial liberalization 64
endogenous growth theory 26, 29–41, 49, firms: adjustment costs 83–5; aggregate
157n3, 157n5; AK model 29–30; production function 47–8; investment
allocation of labor 32–4; endogenous behaviour 80–3; market value 38;
technological change 30–4; maximization problem 81;
innovation vs imitation 40–1; the representative 157n1; value 83
R&D sector 30–1, 157n4; Romer’s fiscal capacity to tax 123
product variety model 34–9; fiscal policy 115–25; adjustment efforts

Schumpeterian growth models 39–40, 122, 124–5; balanced budget 116–7;
157n2; steady-state equilibrium 31–2 budget deficit 115, 115–6; debt
endogenous saving rate 42, 47–8 financing 121–5; debt forgiveness
endogenous technological change 30–4; 124; debt overhang 122–4; deficit bias
allocation of labor 32–4; research and 119; expansionary 145–6, 146, 147;
development (R&D) 30, 30–1; government budget identity 115–6;
steady-state equilibrium 31–2 government debt 115; political
endogenous variables 156n2 economy 118–21; Ricardian
entrepreneurs 39 equivalence analysis 116–7;
equilibrium business cycles 53–7, 157n2; sustainable 116; tax smoothing 117–8
households and 55–7; labor demand Fisher effect, the 160n2
55, 158n3; technology shocks 53–4 fixed costs 84
equilibrium inflation 130–1, 130, 131, 133 Fleming, Marcus 144
equilibrium population size 13–4, 13 flow variables 2
equilibrium profits 37 foreign credit, financial crises model 62–3
Euler equation, the 44, 68–9, 74, 143; foreign currency, unit prices 141
logarithmic utility 44–5 foreign financial assets 140
euro area, the 148, 160n8 Frederick, S., et al 78
Europe xi, 14–5 Friedman, Milton 65, 66
European Central Bank 160n8 functions: applications 153–4; chain
European Union, debt crisis 121, 148–9 differentiation 152–3; derivatives of
exchange rate overshooting 139, 146–8 basic 150–1, 160n1; differentiable
exchange rates 1, 141; fixed 145–6, 146, 150; differentiation rules 151–2;
147 division of 152; implicit
exogenous variables 156n2 differentiation 153; multiplication
expectations-augmented Phillips curves with a scalar 152; product of 152; sum
109 of 151–2
expected utility 71–2
exponential function 151, 160n1 gain–loss function 77
exponents 154 Gali, J. 114
exports 2, 139 Galor, O. 12, 14, 16–8, 17
Galor–Weil model 16–8, 17
Fertile Crescent, the 156n5, 156n4 general equilibrium model 38
final goods sector 30–1, 32, 34, 35 generations 42
financial crises model 58–64; bank run geography, and technological progress
equilibrium 61–2; bank runs 58, 59, 14–5
63, 64, 158n2; banks 59–61; banks Gertler, M. 114
total liquidation value 61; basic goods: final price 35; quality 39–40
assumptions 58–9; financial goods market, IS–MP model 103
Index 167
government budget identity 115–6 illiquidity 58, 64
government debt 1, 115, 116–7; acceptance imitation 40–1
of 121; adjustment efforts 122, 124–5; imperfect capital mobility 145, 145–6, 147
debt forgiveness 124; debt overhang imperfect competition 114
122–4; financing 121–5; IMF implicit function differentiation 153
repayments 123–4; political economy imports 2, 139
of 118–21; sovereign default 123; income 66; future 69–71, 71; and the
unsustainable level of 122–4 housing market 86, 87; individual 43,
government spending 2, 53, 124, 143, 144; 45–6; time distribution 70, 71
and government budget identity income per capita 14; and population
115–6; Ricardian equivalence analysis growth 7
117 India xi, 25
government subsidies 1 individual choice 1
Greece 121, 123, 148 individual consumption 12, 43–4, 46, 55–6,
gross domestic private investment 2 156n2
gross domestic product (GDP) 1, 156n1; individual maximization problem 56–7
and budget deficit 115; calculating individuals 1
2–3; expenditure approach 2–3; and industrial revolution, the 14, 14–5
government debt 115; income side inflation 1, 104, 105; and aggregate
2–3; and the national accounts identity demand 126; and aggregate supply
2–3; production side 2–3; and trade 126; binding inflation commitments
139 128; central bank policy 159n3;
gross national income 160n1 delegation of monetary authority 132;
growth models, fundamental assumption discretionary monetary policy 128,
7–8 129–31, 130; equilibrium 130–1, 130,
growth rate: endogenous 49; of the 131, 133; and monetary policy
logarithmic function 154; of a 126–38; and the money market 127,
multiplicative function 153–4 160n1; and output growth 126; the
Phillips curve 107–8, 109, 128; and
habit formation 77 political business cycles 126, 132–6,
Haiti 19 135; public expectations 129, 130–1,
Hall, R.E. 71–2, 73 131; and the quantity theory of money
Harrod neutrality 157n8 126–7; random output shocks 131–2;
Hibbs, D.A. 14–5, 132 and reputation 132; and seigniorage
high-powered money 104 137–8, 138; subgame perfect Nash
home country, the 140 equilibrium 130; the Taylor rule
household optimization 45–7; basic 136–7; time-inconsistency model of
assumptions 42; intertemporal budget monetary policy 126, 127–32
constraint 43–4; logarithmic utility inflation bias 128, 129
44–5; marginal utility curve 43, 43; inflation tax 137–8; Laffer curve 138, 138
utility functions 42–7, 43 innovations 40–1
households: behavior 42; budgets 1; insider–outsider theory, wage setting 88,
equilibrium business cycles and 55–7 94–5
housing market 85–7, 86 instantaneous utility function 66–7, 72
housing stock 85 intellectual property rights 33, 38–9
Howitt, P. 29, 30, 32, 39–40, 157n6 interest cost 82
human capital 26–8, 41, 157n7 interest rates 1, 56, 75–6; and credit supply
human capital investments 159n1 111–4, 112, 113; differentials 147;
hyperinflation 137, 137–8, 160n9 housing market 85, 86, 86;
international 125, 160n5; IS–MP
Iceland 158n2 model 104; Keynesian investment
ideas 26, 31 function 80; real 57; uncovered
identity function 151 interest rate parity 147
168 Index
intermediate goods sector 34, 34–6, 38–9 production 8; production function
international credit markets, liberalization 89–90; profit maximization 55; R&D
64 32; search and matching framework
international economic policy-making 1–2 89; skills 41; and technological
International Monetary Fund (IMF) 123–4, progress 25–6
125 labor demand 55, 88, 90, 93, 158n3, 159n1
international politics 1–2 labor market 32, 55, 88–98; disequilibrium
intertemporal budget constraint 43–4, 46, 88–9, 89; efficiency wage theory 88,
60, 142, 157n3 89, 89–94, 91; insider–outsider theory
investment demand 159n6 88, 89, 94–5; integrated 148; job
investment irreversibility 84 creation 96, 97; Poisson arrival rate
investment projects 59, 60 96; search and matching framework
investment returns 60 88, 95–8, 97; tightness 96
investments 80–7; accelerator model labor supply 55–7, 88–9

159n2; adjustment costs 83–5; Laibson, D. 78
aggregate level of 1; Cobb–Douglas land, and production 8
example 82; definition 2; depreciation law, rule of 9
cost 82; firm behavior 80–3; legal capacity to tax 123
government 2; gross domestic private Lehman Brothers 158n1
2; housing market 85–7, 86; increases lending 110–3, 111, 112
159n1; interest cost 82; Keynesian l’Hôpital’s rule 157n4
model 80; maximization problem 81; life expectancy 9
Tobin’s marginal q 80, 83; user cost of lifetime budget constraint 66–7, 78–9
capital 80, 81–2, 83 Lindbeck, A. 94
IS–MP model 101, 103–5; aggregate living standards 8–9
demand 105–6, 106; curves 104–5, LM curve 144–5
105; and financial intermediation logarithmic function 151; growth rate 154
110–4, 111, 112, 113; goods market logarithmic utility 44–5
103; money market 104; variations logarithms 155
104–5 Long, J.B. 53
long-run aggregate supply 109
Jones, C.I. 32 long-run growth xi–xii, 53, 156n1
loss aversion 77
Keynesian aggregate consumption 70–1, 71 Lucas, R.E. 65, 101, 109–10
Keynesian aggregate demand model 101–3, Lucas critique, the 101, 110, 114
102 Lucas supply curve 109–10
Keynesian aggregate supply model 106–7
Keynesian consumption function 65, macro economy, the 1
101–3; Friedman’s critique 66 macroeconomic outcomes 1
Keynesian cross, the 102–3, 102 macroeconomic profession, criticism of xi
Keynesian investment function 80 macroeconomics, definition 1
Keynesian theory 53; Lucas critique 101, Malthus, Thomas 7
110 Malthusian growth model xi, 7–18, 156n2,
knowledge: growth rate 31–2; innovation 156n3; collapse of 14–8; and
vs imitation 41; output elasticity 31 endogenous fertility 11–4, 13;
Krugman, P. 122, 124 equilibrium population density 14;
Kydland, F. 53, 126, 128, 129 Galor–Weil model 16–8, 17; law of
diminishing returns 7, 7–9; the
labor: allocation of 32–4; and capital population growth trap 9–11, 10, 11,
rivalry 31; effective 25–6; 17–8, 17; technological progress and
employment level 93; growth rate 10–1, 11, 14–5
20–1; marginal product of 8, 55; Mankiw, G. 114
mobility 148; output elasticity 8; and Mankiw, G., et al 26–8
Index 169
marginal costs 84–5 nominal rigidities 109–10
marginal product of labor 8, 55 nonaccelerating rate of unemployment
marginal propensity to consume 65 (NAIRU) 108–9
marginal propensity to consume (MPC) nonrival goods 26
101–2 Norway 19
marginal q 80, 83
marginal utility 74 Obstfeld, M. 160n4
market equilibrium xi Olsson, O. 14–5
market value, firms 38 Olsson and Hibbs model 15
markets xi open economy, the 139–49; accounting
mark-up pricing 36 139–41; balance of payments 140;
marriage market 159n3 currency union 139, 148–9; current
Marshall-Lerner condition, the 141 account 139–40; current account
menu costs 114 representative agent framework

microeconomics 1 141–4; exchange rate overshooting
micro-founded models xii 139, 146–8; exchange rates 141;
Middle Ages, the 10 government spending 143, 144;
mid-term analysis, permanent income Mundell–Fleming model 139, 144–6,
hypothesis (PIH) 69–70 146; representative agent framework
modern growth regime 14 141–4
monetary policy 104; discretionary 128, optimal intertemporal growth rate 49
129–31, 130; expansionary 147–8; output elasticity of labor 8
and inflation 126–38; output elasticity of land 8
Mundell–Fleming model 145, 146, output growth, and inflation 126
147; non-benevolent government output per capita 8–9; equilibrium growth
137–8; political business cycles rate 28; and population growth 9,
132–6, 135; random output shocks 9–11; steady-state level 28;
131–2; the Taylor rule 136–7; technological progress and 10–1, 11,
time-inconsistency model 126, 14, 26
127–32, 136 output per worker, growth rate 22–3, 23,
money: demand for 104; growth rates 127; 24–5
printing 115; quantity theory of 126, overlapping generations (OLG) model
126–7; velocity of 126 42–9, 156n5, 157n1; basic
money market: and inflation 127, 160n1; assumptions 42; CRRA utility
IS–MP model 104 function 45–7; endogenous saving
money supply 104, 127, 145, 147–8 rate 42, 47–8; the Euler equation 44;
monopolies 35 household optimization 42–7;
Mortensen, D. 88, 95 intertemporal budget constraint 43–4,
multiplication with a scalar 152 46, 157n3; logarithmic utility 44–5;
multiplicative function, growth rate 153–4 optimal intertemporal growth rate 49;
multiplier effect, the 102 utility functions 42–7, 43
Mundell, R. 144, 148–9
Mundell–Fleming model 139, 144–6, 146 patents 37–9
perfect capital mobility 145–6, 146
national accounts identity, the 2–3, 156n1, perfect credit market 67
156n2 permanent income 66
national savings 139–40 permanent income hypothesis (PIH) 66,
natural rate of output 109 66–71, 117; consumption sensitivity
negative savings 158n4 158n7; example 68–9; life-cycle
Neolithic revolution, the 15 implications 70, 71; mid-term analysis
net exports 139 69–70; and savings 69; utility
new Keynesian models 101, 114 maximization 67–8; weakness 71
Niger 19 personal spending 2
170 Index
Persson, T. 123 Ramsey growth model 157n6
Phillips curve, the 107–8; random output shocks 131–2
expectations-augmented 109; and random-walk model 71–3; stochastic
inflation 128; and the natural rate of component 73
unemployment 108–9, 108 rational expectations 71, 73
physical capital, accumulation of, Solow rational expectations paradigm 65
growth model 19–20 reaction functions 129
Pissarides, C. 88, 95 real balances 104
Plosser, C.I. 53 real business cycles (RBCs) 53
policy-making, international economic 1–2 real interest rates 57
political business cycles 126, 132–6, 135 real output 103
political economy xii, 118–21 real-valued function 150
political freedom 9 Rebelo, S. 29–30, 53
political institutions, role of 15 Reinhart, C. 58, 159n3

population growth 30, 32, 156n4; relative change 154
equilibrium population size 13–4, 13; relative consumption 65, 77–8, 158n13
Galor–Weil model 16–8, 17; and repo rate, the 104, 136
income 11–4; and income per capita
representative agent framework, the open
7; law of diminishing returns and 8, 9;
economy 141–4
the Malthusian trap 9–11, 10, 11,
representative individuals 1
17–8, 17; and output per capita 9,
9–11; Solow growth model 19, 23, 25; repurchase agreements 160n7
technological progress and 10–1, 11 reputation 132
Portugal 121 research and development (R&D) 26, 29,
power 151 33, 40; consumption and 49; the
precautionary saving 73–5, 74, 75 endogenous growth theory 30–1,
pre-industrial era 8 157n4; endogenous technological
change 30, 30–1; labor 32; product
premature liquidation 58, 60, 61
variety model 34, 36–7, 37–9; the
Prescott, E. 53, 126, 128, 129
production function 30–1
price elasticity 35–6
Ricardian equivalence 116–7
price rigidity 114
prices 159n5; aggregate price level 114, Ricardo, D. 7
148, 159n5; changing 114; flexibility risk 75
148 risk sharing mechanisms 148
price-setting monopolies 35 risk-aversion 44, 45–7, 76
primary deficit 116 Rogoff, K. 58, 132, 159n3, 160n4
probability 158n3 Romer, P. 29, 30, 32, 40, 49, 80, 89, 103,
product variety model 34–9, 157n2; final 157n7; product variety model 34–9
goods sector 34, 35; intermediate
goods sector 34, 34–6, 38–9; R&D Sala-i-Martin, X. 157n2, 157n6
sector 34, 36–7, 37–9; solving 37–9; savings 157n6; and consumption theory
three sectors 34–7 65–79; mid-term analysis 69–70;
production function 8, 156n4; labor 89–90; national 139–40; negative 158n4;
the R&D sector 30–1 optimal 69; and permanent income
production theory, diminishing returns 7 hypothesis (PIH) 69; precautionary
production costs 35 73–5, 74, 75; and uncertainty 73–5, 75
profit function 35, 159n2 savings rate 30, 43, 56; endogenous 42,
profit maximization 55 47–8; optimal level 45; Solow growth
profit-maximizing price 36 model 24; steady state 48
public debt 2 scalar, multiplication with a 152
Schumpeter, J. 39
quality 39–40 Schumpeterian growth models 39–40,
quantity theory of money 126, 126–7 157n2
Index 171
search and matching framework: labor Taylor rule, the 136–7, 159n3
market 88, 89, 95–8, 97; marriage technological progress 14, 156n3;
market 159n3 allocation of labor and 33; and
seigniorage 126, 137–8, 138 education 16–8, 17; endogenous 30–4;
self-correcting markets xi Galor–Weil model 16–8, 17; and
Shapiro, C. 88 geography 14–5; growth rate 15,
short run, the xi 53–4; innovation vs imitation 40–1;
short-term debt 63 labor augmenting 25; Malthusian
Sibert, A. 132 model and 10–1, 11, 14–5; Olsson and
simulations 57 Hibbs model 15; Schumpeterian
Snower, D. 94 growth models and 40; Solow growth
social capital investments 159n1 model and 25–6, 29; technology
Solow, R. 19 shocks 53–4
Solow growth model 19–28, 156n5; technology shocks 53–4, 55

accumulation of physical capital time discount rates 75–6
19–20; basic assumptions 19–20; time framework 42
Cobb–Douglas functional form 20, time series econometrics 73
22–4, 23; constant returns assumption time-inconsistency 65, 78–9
20; convergence 19, 24–5; criticisms time-inconsistency model of monetary
of 29, 30; dynamics 20–1; equilibrium policy 126, 127–32; basic assumptions
21–2, 22; extensions 25–8; golden 128; discretionary monetary policy
rule of capital accumulation 24; 128, 129–31, 130; random output
growth rate of the labor force 20–1; shocks 131–2; solutions 132, 136
growth rate per worker 22–4, 23; and Tobin, J. 83
human capital 26–8; implications Tobin’s marginal q 80, 83
22–5, 23; k-equation 19–22, 22–4, 22; total production 8
net investment K 21; and population trade, total 139
growth 19, 23, 25; replacement trade unions, insider–outsider theory 94–5
investments 21; savings rate 24, 42; twin deficits 140
steady-state equilibrium 22, 23–4, two-period, representative agent model of
31–2; system stability 157n6; and consumer optimization xii
technological progress 25–6, 29
Sørensen, P.B. 159n3 uncertainty, and savings 73–5, 75
sovereign default 123 uncovered interest rate parity 147
spending: autonomous 102; government 2, unemployment 1, 88–9, 89, 91, 94, 156n1;
53, 115–6, 117, 124, 143, 144; lifetime utility of 93; natural rate of
personal 2 108–9; the Phillips curve 107–9, 108;
state capacity to tax 123 search and matching framework 96–8
steady-state equilibrium 22, 23–4, unified growth model 16–8, 17
31–2 United Kingdom, breakout from the
Stella, A. 126, 131, 160n5 Malthusian trap 14
Stiglitz, J.E. 88 United States of America 58; monetary
stock variables 2 policy 136–7; the Taylor rule 159n3;
subsistence level, the 9, 11 twin deficits 140
substitution effect 45 user cost of capital 80, 81–2, 83
surpluses 2 utility functions: CRRA 44–5, 75–6;
Sweden 158n2 housing market 85–6; individual
Switzerland 19 55–7; logarithmic 44–5; marginal
utility curve 43, 43; overlapping
Tabellini, G. 119 generations (OLG) model 42–7, 43;
tax revenue 115, 122–3 political economy 119–21;
tax smoothing 117–8 precautionary saving 74, 74; and
Taylor, J.B. 136–7 probability 158n3; workers 90–1
172 Index
utility maximization 60, 67–8 wealth 61
utility-maximizing relative leisure 57 Weil, D. 16–8, 17
Whitta-Jakobsen, H.J. 159n3
variables of interest 1 windfall gains 158n7
Velasco, A. 58, 158n6 Woodford, M. 110
work force, growth rate 30
wages: efficiency equilibrium 91; work horse model xii
efficiency wage theory 89, 89–94; working hours 56
flexibility 148; insider–outsider theory
88, 89, 94–5; and labor supply 88–9,
89; no-shirking condition 92; relative young generation utility function 42–3
57; Shapiro–Stiglitz model 90–4, 91;
shirkers 91–2 Zambia 25

Essentials of Advanced Macroeconomic Theory 2012

Uploaded by

Copyright:

Available Formats

Essentials of Advanced Macroeconomic Theory 2012

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Essentials of Advanced Macroeconomic Theory 2012

Uploaded by

Copyright:

Available Formats

Downloaded by [Hacettepe University] at 20:58 10 April 2017

Trying to summarize the essentials of macroeconomic theory, in the wake of the

1 Financial Econometrics 10 Financial Econometrics,

Peijie Wang second edition

All rights reserved. No part of this book may be reprinted or reproduced

Typeset in Times New Roman by Sunrise Setting Ltd, Torquay, UK.

3 The Solow Growth Model 19

4 Endogenous Growth Theory 29

5 The Overlapping Generations Model 42

6 Equilibrium Business Cycles 53

8 Consumption and Saving 65

9 Investment and Asset Markets 80

10 Unemployment and the Labor Market 88

11.3 Aggregate demand 105

12 Public Finance and Fiscal Policy 115

13 Inflation and Monetary Policy 126

14 The Open Economy 139

15 Mathematical Appendix 150

2.1 The Malthusian trap 10

Trying to summarize the essentials of macroeconomic theory in 2011 is no easy

Bo Sandelin, Joachim Smend, Olof Johansson-Stenman, David Weil, and three

1.1 The issues

In an international economy with interwoven markets, macroeconomics is also a

1.2 The national accounts identity

In this key equation of macroeconomics, Yt is total GDP as before at time t, Ct

Yt = wages + rental incomes + profits + interest . . . (1.2)

The Long Run

2.2 The law of diminishing returns

In this expression, Y is total output, A is a technology shift parameter, X is the

which confirms what we just said.

2.3 The Malthusian trap

L t = L t−1 + Bt (yt−1 ) − Dt (yt−1 ) (2.4)

Birth rates, death rates

Figure 2.1 The Malthusian trap.

Birth rates, death rates

Figure 2.2 Technological progress in the Malthusian model.

2.4 Endogenous fertility

where ct = Ct /L t is consumption per capita in the economy, n t ≥ 0 is the number

A straightforward way of solving this problem is simply to rewrite the budget

The first-order condition for maximum is2

By rearranging this condition, we obtain

β(yt − ρn t ) = (1 − β)ρn =⇒ (1 − β)ρn t + βρn t = ρn t = βyt

ct∗ + ρn ∗t = yt = ct∗ + βyt =⇒ ct∗ = (1 − β)yt

The growth of the total level of population between periods t and t + 1 is

where we have substituted yt = A(X/L t )α from (2.3). The expression in (2.7)

Figure 2.3 Equilibrium population size.

The real-world equivalent of this expression could, for instance, be population

this hypothesis in their empirical study.

2.5 The collapse of the Malthusian link

2.5.1 Geography and institutions

where γ > 0 is a parameter invariant across regions and E i is the “wealth” of

In this expression, the growth rate g is a positive function of et and L t . A higher

3.2 Basic assumptions

Yt = F(K (t), L(t)) (3.1)

• Constant returns to scale: F(λK , λ(L)) = λF(K , L).

This transformation is referred to as the intensive form of (3.1) and k is for-

• f (0) = 0; f (k) > 0; f (k) < 0 at all levels of k > 0.

Y = F(K , L) = K α L 1−α (3.3)

Dividing through by L gives us the intensive form, or output per worker y:

F = A(1 − a)L − w(1 − a)L

intermediate firm’s annual level of profit Ij from investing a sum P jA in buying a