EC2066 Vle

Microeconomics
A. Daripa
EC2066
2016
Undergraduate study in
Economics, Management,
Finance and the Social Sciences
This subject guide is for a 200 course offered as part of the University of London
International Programmes in Economics, Management, Finance and the Social Sciences.
This is equivalent to Level 5 within the Framework for Higher Education Qualifications in
England, Wales and Northern Ireland (FHEQ).
For more information about the University of London International Programmes
undergraduate study in Economics, Management, Finance and the Social Sciences, see:
www.londoninternational.ac.uk
This guide was prepared for the University of London International Programmes by:
Dr Arup Daripa, Lecturer in Financial Economics, Department of Economics, Mathematics and
Statistics, Birkbeck, University of London.
This is one of a series of subject guides published by the University. We regret that due to
pressure of work the author is unable to enter into any correspondence relating to, or arising
from, the guide. If you have any comments on this subject guide, favourable or unfavourable,
please use the form at the back of this guide.
University of London International Programmes

Publications Office
Stewart House
32 Russell Square
London WC1B 5DN
United Kingdom
www.londoninternational.ac.uk
Published by: University of London

University of London 2016
The University of London asserts copyright over all material in this subject guide except where
otherwise indicated. All rights reserved. No part of this work may be reproduced in any form,
or by any means, without permission in writing from the publisher. We make every effort to
respect copyright. If you think we have inadvertently used your copyright material, please let
us know.
Contents
Contents
1 Introduction 1
1.1 Routemap to the subject guide . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Introduction to the subject area and prior knowledge . . . . . . . . . . . 2
1.3 Syllabus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Aims of the course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Learning outcomes for the course . . . . . . . . . . . . . . . . . . . . . . 3
1.6 Overview of learning resources . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6.1 The subject guide . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6.2 Essential reading . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6.3 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.6.4 Online study resources . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6.5 The VLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6.6 Making use of the Online Library . . . . . . . . . . . . . . . . . . 7
1.7 Examination advice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.7.1 Format of the examination . . . . . . . . . . . . . . . . . . . . . . 7
1.7.2 Types of questions . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.7.3 Specific advice on approaching the questions . . . . . . . . . . . . 8
2 Consumer theory 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Aims of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Overview of the chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Preferences, utility and choice . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Preferences and utility . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2 Indifference curves . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.3 Budget constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.4 Utility maximisation . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Demand curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
i
Contents
2.4.1 The impact of income and price changes . . . . . . . . . . . . . . 20

2.4.2 Elasticities of demand . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.3 The compensated demand curve . . . . . . . . . . . . . . . . . . . 24
2.4.4 Welfare measures: CS, CV and EV . . . . . . . . . . . . . . . . 25
2.5 Labour supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6 Saving and borrowing: intertemporal choice . . . . . . . . . . . . . . . . 30
2.7 Present value calculation with many periods . . . . . . . . . . . . . . . . 33
2.7.1 Bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.8 A reminder of your learning outcomes . . . . . . . . . . . . . . . . . . . . 34
2.9 Test your knowledge and understanding . . . . . . . . . . . . . . . . . . . 35
2.9.1 Sample examination questions . . . . . . . . . . . . . . . . . . . . 35
3 Choice under uncertainty 37

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Expected utility theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Risk aversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.6 Risk aversion and demand for insurance . . . . . . . . . . . . . . . . . . 40
3.6.1 Insurance premium for full insurance . . . . . . . . . . . . . . . . 40
3.6.2 How much insurance? . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.7 Risk-neutral and risk-loving preferences . . . . . . . . . . . . . . . . . . . 43
3.8 The ArrowPratt measure of risk aversion . . . . . . . . . . . . . . . . . 44
3.9 Reducing risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Game theory 49
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
ii
Contents
4.1.4 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.3 Simultaneous-move or normal-form games . . . . . . . . . . . . . . . . . 50
4.3.1 Dominant and dominated strategies . . . . . . . . . . . . . . . . . 51
4.3.2 Dominated strategies and iterated elimination . . . . . . . . . . . 52
4.3.3 Nash equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.4 Mixed strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.5 Existence of Nash equilibrium . . . . . . . . . . . . . . . . . . . . 58
4.3.6 Games with continuous strategy sets . . . . . . . . . . . . . . . . 59
4.4 Sequential-move or extensive-form games . . . . . . . . . . . . . . . . . . 59
4.4.1 Actions and strategies . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4.2 Finding Nash equilibria using the normal form . . . . . . . . . . . 61
4.4.3 Imperfect information: information sets . . . . . . . . . . . . . . . 61
4.5 Incredible threats in Nash equilibria and subgame perfection . . . . . . . 63
4.5.1 Subgame perfection: refinement of Nash equilibrium . . . . . . . . 64
4.5.2 Perfect information: backward induction . . . . . . . . . . . . . . 65
4.5.3 Subgame perfection under imperfect information . . . . . . . . . . 66
4.6 Repeated Prisoners Dilemma . . . . . . . . . . . . . . . . . . . . . . . . 68
4.6.1 Cooperation through trigger strategies . . . . . . . . . . . . . . . 69
4.6.2 Folk theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5 Production, costs and profit maximisation 79

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 A general note on costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4 Production and factor demand . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4.1 Marginal and average product . . . . . . . . . . . . . . . . . . . . 80
5.5 The short run: one variable factor . . . . . . . . . . . . . . . . . . . . . . 81
5.6 The long run: both factors are variable . . . . . . . . . . . . . . . . . . . 83
5.6.1 Isoquants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
iii
Contents
5.6.2 Diminishing MRTS . . . . . . . . . . . . . . . . . . . . . . . . . . 83

5.6.3 Returns to scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.6.4 Optimal long-run input choice . . . . . . . . . . . . . . . . . . . . 84
5.7 Cost curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.7.1 Marginal cost and average cost . . . . . . . . . . . . . . . . . . . 84
5.7.2 Fixed costs and sunk costs . . . . . . . . . . . . . . . . . . . . . . 85
5.7.3 Short run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.7.4 Long run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.8 Profit maximisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6 Perfect competition in a single market 93

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3 A general comment on zero profit . . . . . . . . . . . . . . . . . . . . . . 94
6.4 Supply decision by a price-taking firm . . . . . . . . . . . . . . . . . . . . 95
6.4.1 Which types of firms have a supply curve? . . . . . . . . . . . . . 95
6.4.2 Short-run supply . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.4.3 Long-run supply . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.5 Market supply and market equilibrium . . . . . . . . . . . . . . . . . . . 97
6.5.1 Short run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.5.2 Long run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.6 Producer surplus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.7 Applications of the supply-demand model: partial equilibrium analysis . . 99
6.7.1 Tax: deadweight loss and incidence . . . . . . . . . . . . . . . . . 99
6.7.2 Price ceiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.7.3 Price floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.7.4 Quota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.7.5 Price support policy . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.7.6 Tariffs and quotas . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
iv
Contents

7 General equilibrium and welfare 111

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
7.3 General equilibrium in an exchange economy . . . . . . . . . . . . . . . . 112
7.4 Existence of equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.5 Uniqueness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.6 Welfare theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.6.1 The first theorem of welfare economics . . . . . . . . . . . . . . . 120
7.6.2 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.6.3 The second theorem of welfare economics . . . . . . . . . . . . . . 122
7.6.4 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.7 Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8 Monopoly 129
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.3 Properties of marginal revenue . . . . . . . . . . . . . . . . . . . . . . . . 130
8.4 Profit maximisation and deadweight loss . . . . . . . . . . . . . . . . . . 130
8.5 Price discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.6 Natural monopoly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
v
Contents
9 Oligopoly 137
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
9.3 Cournot competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
9.3.1 Collusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.3.2 Cournot with n > 2 firms . . . . . . . . . . . . . . . . . . . . . . 141
9.3.3 Stackelberg leadership . . . . . . . . . . . . . . . . . . . . . . . . 142
9.4 Bertrand competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.4.1 Collusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.5 Bertrand competition with product differentiation . . . . . . . . . . . . . 143
9.5.1 Sequential pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
10 Asymmetric information: adverse selection 149

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
10.3 The scope of economic theory: a general comment . . . . . . . . . . . . . 151
10.4 Akerlofs (1970) model of the market for lemons . . . . . . . . . . . . . . 152
10.4.1 The market for lemons: an example with two qualities . . . . . . . 152
10.5 A model of price discrimination . . . . . . . . . . . . . . . . . . . . . . . 153
10.5.1 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.5.2 The full information benchmark . . . . . . . . . . . . . . . . . . . 155
10.5.3 Contracts under asymmetric information . . . . . . . . . . . . . . 155
10.6 Spences (1973) model of job market signalling . . . . . . . . . . . . . . . 158
vi
Contents
11 Asymmetric information: moral hazard 161

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
11.3 Effort choice and incentive contracts: a formal model . . . . . . . . . . . 162
11.4 Full information: observable effort . . . . . . . . . . . . . . . . . . . . . . 164
11.4.1 Implementing high effort eH . . . . . . . . . . . . . . . . . . . . . 164
11.4.2 Implementing low effort eL . . . . . . . . . . . . . . . . . . . . . . 165
11.4.3 Which effort is optimal for the principal? . . . . . . . . . . . . . . 165
11.5 Asymmetric information: unobservable effort . . . . . . . . . . . . . . . . 166
11.5.1 Implementing low effort eL . . . . . . . . . . . . . . . . . . . . . . 166
11.5.2 Implementing high effort eH . . . . . . . . . . . . . . . . . . . . . 166
11.6 Risk-neutral agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
12 Externalities and public goods 173

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
12.1.4 References cited . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
12.3 Externalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
12.3.1 Tax and quota policies . . . . . . . . . . . . . . . . . . . . . . . . 177
12.3.2 Coase theorem: the property rights solution . . . . . . . . . . . . 178
12.4 Public goods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
12.4.1 Pareto optimum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
12.4.2 Private provision . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
12.5 The commons problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
12.5.1 Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
12.5.2 A simple model of resource extraction . . . . . . . . . . . . . . . . 186
vii
Contents

viii
Chapter 1
Introduction
1.1 Routemap to the subject guide
Welcome to this course in Microeconomics.

In this introductory chapter, we will look at the overall structure of the subject guide
(in the form of a Routemap); we will introduce you to the subject area; to the aims and
learning outcomes for the course; and to the learning resources available to you. Finally,
we will offer you some Examination advice.
We hope that you enjoy this course and we wish you every success in your studies.
We start by analysing individual choice. In Chapter 2, we analyse consumer choice. We
specify properties of preferences, how to go from preferences to utility and optimal
choice by maximising utility under a budget constraint. We show how to obtain demand
functions from optimal solutions to the consumer choice problem. We discuss what
happens to the consumers welfare when prices and/or income change, and appropriate
measures to evaluate these changes. We also consider the labour supply decision of
individuals, and the intertemporal choice problem faced by savers and borrowers. We
provide a brief exposition of some basic algebra of intertemporal choice problems with
more than two periods. In Chapter 3, we model the agents behaviour in situations
involving risk, and analyse insurance problems. We then introduce strategic interaction
in Chapter 4 and provide an exposition of the basic tools of game theory. Next, we turn
to the supply side of the economy. In Chapter 5, we describe the production technology,
structure of costs and principles of profit maximisation by firms. Chapter 5 of the
subject guide is essentially a toolbox for analysis in subsequent chapters. Using these
tools, we analyse the problem of competitive firms as well as the equilibrium in a
competitive market in Chapter 6. In Chapter 7, we then introduce the general
equilibrium across all competitive markets and study the two welfare theorems that are
fundamental to our understanding of market economies. We also consider some
applications of the supply-demand model, and analyse the impact of policies on welfare.
Next, we consider markets that are imperfectly competitive. In Chapter 8, we analyse
the problem of monopoly, the associated inefficiencies and policy prescriptions aimed at
restoring efficiency. This is followed by an analysis of the problem of oligopolistic
competition in Chapter 9. Next, we turn to the problem of asymmetric information and
analyse the problems of adverse selection (in Chapter 10) and moral hazard (in Chapter
11). Finally, in Chapter 12, we consider the problem of externalities and public goods
and consider policy prescriptions arising from associated market failures.
Given the emphasis of the EC2066 Microeconomics syllabus on using analytical
methods to solve economic problems, you are encouraged to spend a considerable
amount of time doing the problems or questions given in the textbooks. Learning by
doing is likely to be more profitable than simply reading and re-reading textbooks.
1
1. Introduction
Nevertheless, a thorough reading of, and careful note-taking from, the recommended
textbook and the subject guide is a prerequisite for successful problem solving. The
subject guide aims to indicate as clearly as possible the key elements in microeconomic
analysis that you need to learn. The subject guide also presents detailed algebraic
derivations for a variety of topics. For each topic, you should consult both the subject
guide and the suggested parts of the textbook to understand fully the economic
principles involved.
1.2 Introduction to the subject area and prior

knowledge
The syllabus and subject guide assume that you are competent in basic economic
analysis up to the level of the prerequisite courses EC1002 Introduction to
economics, ST104a Statistics 1, MT105a Mathematics 1 or MT1174 Calculus.
They build on the foundations provided in these courses by specifying how your
understanding of the microeconomic principles developed so far should be deepened and
extended. Like EC1002 Introduction to economics, EC2066 Microeconomics is
designed to equip you with the economic principles necessary to analyse a whole range
of economic problems. To maximise your benefit from the subject, you should continue
to think carefully about:
the assumptions, internal logic and predictions of economic models
how economic principles can be applied to solve particular economic problems.
The appropriate analysis will depend on the specific facts of a problem. However, you
are not expected to know the detailed facts about specific economic issues and policies
mentioned in textbook examples. Rather, you should use these examples (and the
end-of-chapter Sample examination questions) to aid your understanding of how
economic principles can be applied creatively to the analysis of economic problems.
If you are taking this course as part of a BSc degree you will also have passed ST104a
Statistics 1 and MT105a Mathematics 1 or MT1174 Calculus before beginning
this course. Every part of the syllabus can be mastered with the aid of diagrams and
relatively simple algebra. The subject guide indicates the minimum level of
mathematical knowledge that is required. Knowledge (and use in the examination) of
sophisticated mathematical techniques is not required. However, if you are
mathematically competent you are encouraged to use mathematical techniques when
these are appropriate, as long as you recognise that some verbal explanation is always
necessary.
1.3 Syllabus
The course examines how economic decisions are made by households and firms, and
how they interact to determine the quantities and prices of goods and factors of
production and the allocation of resources. Further, it examines the nature of strategic
2
1.4. Aims of the course
interaction and interaction under asymmetric information. Finally, it investigates the

role of policy as well as economic contracts in improving welfare. The topics covered are:
Consumer choice and demand, labour supply.
Choice under uncertainty: the expected utility model.
Producer theory: production and cost functions, firm and industry supply.
Game theory: normal-form and extensive form games, Nash equilibrium and
subgame perfect equilibrium, repeated games and cooperative equilibria.
Market structure: competition, monopoly and oligopoly.
General equilibrium and welfare: competitive equilibrium and efficiency.
Pricing in input markets.
Intertemporal choice: savings and investment choices.
The economics of information: moral hazard and adverse selection, resulting

market failures and the role of contracts and institutions.
Market failures arising from monopoly, externalities and public goods. The role of
policy.
1.4 Aims of the course

This subject guide enables you to fully interpret the published syllabus for EC2066
Microeconomics. It identifies what you are expected to know within each area of the
syllabus by emphasising the relevant concepts and models and by stating where in
specific textbooks that material can be found. This subject guide aims to help you make
the best use of textbooks to secure a firm understanding of the microeconomic analysis
covered by the syllabus. The subject guide also complements your textbook in certain
areas where the coverage in the textbook is deemed inadequate.
1.5 Learning outcomes for the course

At the end of this course, and having completed the Essential reading and activities,
you should:
be able to define and describe:

the determinants of consumer choice, including inter-temporal choice and
choice under uncertainty
the behaviour of firms under different market structures
how firms and households determine factor prices
behaviour of agents in static as well as dynamic strategic situations
3
1. Introduction
the nature of economic interaction under asymmetric information
be able to analyse and assess:

efficiency and welfare optimality of perfectly and imperfectly competitive
markets
the effects of externalities and public goods on efficiency
the effects of strategic behaviour and asymmetric information on efficiency
the nature of policies and contracts aimed at improving welfare
be prepared for further courses which require a knowledge of microeconomics.
Each chapter includes a list of the learning outcomes that are specific to it. However,
you also need to go beyond the learning outcomes of each single chapter by developing
the ability of linking the concepts introduced in different chapters, in order to approach
the examination well.
1.6 Overview of learning resources
1.6.1 The subject guide

Each chapter of the subject guide has the following format.
The Essential reading lists the relevant textbook chapters and sections of chapters,
even though a more detailed indication of the required reading is listed throughout
the chapter.
The sections that follow specify in detail what you are expected to know about
each topic. The relevant sections of the recommended textbooks are referred to.
Wherever necessary, the sections integrate the textbook with additional material
and explanations. Finally, they draw attention to any problems that occur in
textbook expositions and explain how these can be overcome.
The boxes that appear in some of the sections give you exercises based on the
material discussed.
The learning outcomes show you what you should be able to do by the end of the
chapter.
A final section gives you questions to test your knowledge and understanding.
1.6.2 Essential reading

This subject guide is specifically designed to be used in conjunction with the textbook:
Nicholson, W. and C. Snyder Intermediate Microeconomics and its Application.

(Cengage Learning, 2015) 12th edition [ISBN 9781133189039].
Henceforth in this subject guide this textbook is referred to as N&S.
4
1.6. Overview of learning resources
This is available as an e-book at a discounted price via the VLE. Please visit
the course page for details.
Students may use the previous edition instead:
Nicholson, W. and C. Snyder Theory and Applications of Intermediate

Microeconomics. (Cengage Learning, 2010) 11th edition, international edition
[ISBN 9780324599497].
Note that the title of the twelfth edition differs slightly from that of the eleventh. If
using this edition, students should refer to the reading supplement on the VLE for
customised references.
N&S is more adequate for some parts of the syllabus while less so for others; as a
consequence we integrate some topics more with the extra material provided in the
subject guide. The textbook employs verbal reasoning as the main method of
presentation, supplemented by diagrammatic analyses. The textbooks use of algebra is
not uniformly satisfactory. The subject guide supplements the textbook in many cases
in this regard. There are also some references to the following textbook:
Perloff, J.M. Microeconomics with Calculus. (Pearson Education, 2014) 3rd edition
[ISBN 9780273789987].
Detailed reading references in this subject guide refer to the editions of the textbooks
listed above. New editions of these textbooks may have been published by the time you
study this course. You can use a more recent edition of any of the textbooks; use the
detailed chapter and section headings and the index to identify relevant readings. Also,
check the virtual learning environment (VLE) regularly for updated guidance on
readings.
Unless otherwise stated, all websites in this subject guide were accessed in February
2016. We cannot guarantee, however, that they will stay current and you may need to
perform an internet search to find the relevant pages.
1.6.3 Further reading

Please note that as long as you read the Essential reading you are then free to read
around the subject area in any textbook, paper or online resource. You will need to
support your learning by reading as widely as possible and by thinking about how these
principles apply in the real world. To help you read extensively, you have free access to
the virtual learning environment (VLE) and University of London Online Library (see
below).
Other useful textbooks for this course include:
Besanko, D. and R. Braeutigam Microeconomics. (John Wiley & Sons, 2014) 5th
edition, international student version [ISBN 9781118716380].
Varian, H.R. Intermediate Microeconomics, a Modern Approach. (W.W. Norton,
2014) 9th edition [ISBN 9780393920772].
Pindyck, R.S. and D.L. Rubinfeld Microeconomics. (Pearson, 2014) 8th edition
[ISBN 9781292081977].
5
1. Introduction
1.6.4 Online study resources

In addition to the subject guide and the Essential reading, it is crucial that you take
advantage of the study resources that are available online for this course, including the
VLE and the Online Library.
You can access the VLE, the Online Library and your University of London email
account via the Student Portal at:
http://my.londoninternational.ac.uk
You should have received your login details for the Student Portal with your official
offer, which was emailed to the address that you gave on your application form. You
have probably already logged in to the Student Portal in order to register! As soon as
you registered, you will automatically have been granted access to the VLE, Online
Library and your fully functional University of London email account.
If you have forgotten these login details, please click on the Forgotten your password
link on the login page.
1.6.5 The VLE

The VLE, which complements this subject guide, has been designed to enhance your
learning experience, providing additional support and a sense of community. It forms an
important part of your study experience with the University of London and you should
access it regularly.
The VLE provides a range of resources for EMFSS courses:
Electronic study materials: All of the printed materials which you receive from
the University of London are available to download, to give you flexibility in how
and where you study.
Discussion forums: An open space for you to discuss interests and seek support
from your peers, working collaboratively to solve problems and discuss subject
material. Some forums are moderated by an LSE academic.
Videos: Recorded academic introductions to many subjects; interviews and
debates with academics who have designed the courses and teach similar ones at
LSE.
Recorded lectures: For a few subjects, where appropriate, various teaching
sessions of the course have been recorded and made available online via the VLE.
Audio-visual tutorials and solutions: For some of the first year and larger later
courses such as Introduction to Economics, Statistics, Mathematics and Principles
of Banking and Accounting, audio-visual tutorials are available to help you work
through key concepts and to show the standard expected in examinations.
Self-testing activities: Allowing you to test your own understanding of subject
material.
Study skills: Expert advice on getting started with your studies, preparing for
examinations and developing your digital literacy skills.
6
1.7. Examination advice
Note: Students registered for Laws courses also receive access to the dedicated Laws
VLE.
Some of these resources are available for certain courses only, but we are expanding our
provision all the time and you should check the VLE regularly for updates.
This subject guide is reproduced in colour on the VLE and you may find it easier to
understand if you access the online PDF.
1.6.6 Making use of the Online Library

The Online Library (http://onlinelibrary.london.ac.uk) contains a huge array of
journal articles and other resources to help you read widely and extensively.
To access the majority of resources via the Online Library you will either need to use
your University of London Student Portal login details, or you will be required to
register and use an Athens login.
The easiest way to locate relevant content and journal articles in the Online Library is
to use the Summon search engine.
If you are having trouble finding an article listed in a reading list, try removing any
punctuation from the title, such as single quotation marks, question marks and colons.
For further advice, please use the online help pages
(http://onlinelibrary.london.ac.uk/resources/summon) or contact the Online
Library team: [email protected]
1.7 Examination advice
1.7.1 Format of the examination

Important: the information and advice given here are based on the examination
structure used at the time this subject guide was written. Please note that subject
guides may be used for several years. Because of this we strongly advise you to always
check both the current Regulations for relevant information about the examination, and
the VLE where you should be advised of any forthcoming changes. You should also
carefully check the rubric/instructions on the paper you actually sit and follow those
instructions.
In this examination you should answer eleven of fourteen questions: all eight questions
from Section A (5 marks each) and three out of six from Section B (20 marks each).
1.7.2 Types of questions

Examples of the types of questions which will appear on the examination paper appear
not only in the Sample examination paper on the VLE, but also at the end of chapters.
However, in the examination you should not be surprised to see some questions which
are not necessarily specific to one particular topic. For example, a question may require
knowledge about markets which are oligopolistic as well as those which are monopolistic
or competitive.
7
1. Introduction
Numerical questions will sometimes require the use of a calculator. A calculator may be
used when answering questions on the examination paper for this course and it must
comply in all respects with the specification given in the Regulations.
Questions will not require knowledge of empirical studies or institutional material.
However, you will be awarded some marks for supplementary empirical or institutional
material which is directly relevant to the question.
1.7.3 Specific advice on approaching the questions

You should follow all the excellent advice to candidates which is published in the annual
Examiners commentaries. For this course, the following advice is also worth noting:
Prepare thoroughly for the examination by attempting the problems/questions in

the textbooks and in this subject guide and, in particular, past examination papers
(for which there are Examiners commentaries where you can check examiners
responses).
Occasionally, you may be unsure exactly what a question is asking. If there is some
element of doubt about the interpretation of a question, state at the beginning of
your answer how you interpret the question. If you are very uncertain of what is
required, and the question is in Section B, do another question.
Explain briefly what you are doing: an answer that is simply a list of equations or
numbers will not be credited with full marks even if it gets to the correct solution.
Moreover, by explaining what you are doing, you will be awarded some marks for
correct reasoning even if there are mistakes in some part of the procedure.
It is essential to attempt eight questions in Section A. Even if you think you do not
know the answer, at least define any terms or concepts which you think may be
relevant (including those in the question!) and, if possible, present the question in
diagrammatic or algebraic form. The same applies to a specific part of a multi-part
question in Section B. The examiners can give no marks for an unattempted
question, but they can award marks for relevant points. A single mark may make
the difference between passing and failing the examination.
Although you should attempt all the questions and parts of questions that are
required, to avoid wasting time you should make sure that you do no more than is
required. For example, if only two parts of a three-part question need to be
answered, only answer two parts.
Note the importance of key words. In some of the True or false? type questions,
the words must, always, never or necessarily usually invite you to explain why
the statement is false. Notice that this is simply a way in which you can start
approaching the problem, but there is no way to know in advance the correct
answer without analysing every specific question.
It is worth noting that in this type of question, simply writing true or false will
not earn you any marks, even if it happens to be the right answer. The examiners
are looking for reasoning, not blind guesses.
8
1.7. Examination advice
Good answers to most questions require relevant assumptions to be stated and

terms to be defined. Also, do use the term ceteris paribus (meaning other things
being equal), where appropriate. If you are asked to examine the effects of a
change in a particular exogenous variable, you should not complicate your answers
unnecessarily by positing simultaneous changes in other exogenous variables.
For many questions, good answers will require diagrammatic and/or algebraic
analysis to complement verbal reasoning. Good diagrams can often save much
explanation but free-standing diagrams, however well-drawn and labelled, do not
portray sufficient information to the examiners. Diagrams need to be explained in
the text of the answer. Similarly, symbols in algebraic expressions should be defined
and the final line of an algebraic presentation should be explained in words.
The examiners are primarily looking for analytical explanations, not descriptions.
On reading a question, your first thought should be: what is the appropriate
hypothesis, theory, concept or model to use?
Remember, it is important to check the VLE for:
up-to-date information on examination and assessment arrangements for this course
where available, past examination papers and Examiners commentaries for the
course which give advice on how each question might best be answered.
9
1. Introduction
10
Chapter 2
Consumer theory
2.1 Introduction
How do people choose what bundle of goods to consume? We cannot observe this
process directly, but can we come up with a model to capture the decision-making
process so that the predictions from the model match the way people behave? If we can
build a sensible model, we should be able to use the model to understand how choice
varies when the economic environment changes. This should also help us design
appropriate policy. This is the task in this chapter and as we go through the various
steps, you should keep this overarching goal in your mind and try to see how each piece
of analysis fits in the overall scheme.
Once we build such a model, we use it to analyse how optimal consumption choice
responds to price and income variations. We also extend the analysis to cover
labour-leisure choice as well as intertemporal consumption choice.
2.1.1 Aims of the chapter

This chapter introduces you to the theory of consumer choice. You should be familiar
with many of the ideas here from EC1002 Introduction to economics, but we aim
to investigate certain aspects at relatively greater depth. The chapter also aims to
encourage you to ask questions about the meaning of concepts and their usefulness in
understanding the world. For example, you have come across utility functions before.
But surely no-one has a utility function so where do these functions come from? Why
is this concept useful? You should not accept such concepts just because they appear in
textbooks and are taught in classes. To convey this message is an important aim here.
2.1.2 Learning outcomes

By the end of this chapter, the Essential reading and activities, you should be able to:
explain the implications of the assumptions on the consumers preferences
describe the concept of modelling preferences using a utility function
draw indifference curve diagrams starting from the utility function of a consumer
draw budget lines for different prices and income levels
solve the consumers utility maximisation problem and derive the demand for a
consumer with a given utility function and budget constraint
11
2. Consumer theory
analyse the effect of price and income changes on demand
explain the notion of a compensated demand function
explain measures of the welfare impact of a price change: change in consumer

surplus, equivalent variation and compensating variation, and use these measures
to analyse the welfare impact of a price change in specific cases
construct the market demand curve from individual demand curves
explain the notion of elasticity of demand
analyse the decision to supply labour
analyse the problem of savers and borrowers
derive the present discounted value of payment streams and explain bond pricing.

N&S Chapters 2 and 3, the Appendix to Chapter 13, from Chapter 14: Sections 14.1,
14.2, 14.5 and from Appendix 14A: Sections A143, A144.
In addition, Chapter 1 and Appendix 1A provide a review of the basics of economic
models and some basic techniques. You should be familiar with this material from
earlier courses. Nevertheless, you should read this chapter and the appendix and make
sure that you understand the content. Throughout this subject guide, we will assume
that you are familiar with this material.
2.2 Overview of the chapter

We start by analysing preferences, utility and choice. Next, we learn how to construct
demand curves, and analyse their properties. We also explain various welfare measures.
We then analyse the labour supply decision of an individual before moving on to saving
and borrowing with two periods. Finally, we learn to carry out present value
calculations with many periods and analyse bond pricing.
2.3 Preferences, utility and choice

See N&S Chapter 2. See also Perloff Sections 3.1 and 3.2 for a good discussion on the
connection between preferences and utility.
2.3.1 Preferences and utility

The theory of choice starts with rational preferences. Generally, preferences are
primitives in economics you take these as given and proceed from there. The task of
explaining why certain preferences exist in certain societies falls largely under the
domain of subjects such as anthropology or sociology. However, to be able to create a
12
2.3. Preferences, utility and choice
model of choice that has some predictive power, we do need to put some restrictions on
preferences to rule out irrational behaviour. Just a few relatively sensible restrictions
allow us to build a model of choice that has great analytical power. Indeed, this idea of
creating an analytical structure that can be manipulated to understand how changes in
the economic environment affect economic outcomes underlies the success of economics
as a tool for investigating the world around us.
Section 2.2 of N&S sets out three restrictions on preferences: completeness, transitivity
and non-satiation (more is better). These restrictions allow us to do something very
useful. Once we add some further technical requirements (a full specification must await
Masters level courses but the main extra condition we need is that preferences have
certain continuity properties), these restrictions allow us to represent preferences by a
continuous function. This is known as a utility function.
Note that a utility function is an artificial concept no-one actually has a utility
function (you knew that of course, since you surely do not have one). But because we
can represent preferences using such a function, it is as if agents have a utility function.
All subsequent analysis using utility functions and indifference curves has this as if
property.
Ordinal versus cardinal utility
Preferences generally give us rankings among bundles rather than some absolute
measure of satisfaction derived from bundles. You might prefer apple juice to orange
juice but would have difficulty saying exactly how much more satisfaction you derive
from the former compared to the latter. Preferences, therefore, typically give us an
ordinal ranking among bundles of goods. Since utility is simply a representation of
preferences, it is also an ordinal measure. This means that if your preferences can be
represented by a utility function, then a positive transformation of this function which
preserves the ordering among bundles is another function that is also a valid utility
function. In other words, there are many possible utility functions that can represent a
given set of preferences equally well.
However, there are some instances where we use cardinal utility and make absolute
comparisons among bundles. Money, for example, is a cardinal measure you know that
20 pounds is twice as good as 10 pounds. In general, though, you should understand
utility as an ordinal concept.
2.3.2 Indifference curves
Once we can represent preferences using a continuous utility function, we can draw
indifference curves. An indifference curve is the locus of different bundles of goods that
yield the same level of utility. In other words, an indifference curve for utility function
u(x, y) is given by u(x, y) = k, where k is some constant. As we vary k, we can trace out
the indifference map. Note that an indifference curve is simply a level curve of a utility
function. Just as you draw contours on a map to represent, say, a mountain, so
indifference curves drawn for two goods are contour plots of a utility function over these
two goods. You can see Figure 3.3 in Perloff for a pictorial representation. You should
read carefully the discussion in N&S (Sections 2.3 to 2.5) on indifference curves. You
13
2. Consumer theory
should know how different types of preferences generate different types of indifference
curves.
Activity 2.1 For each of the following utility functions, write down the equation for
an indifference curve and then draw some indifference curves.
(a) u(x, y) = xy.
(b) u(x, y) = x + y.
(c) u(x, y) = min{x, y}.
Previously, we put some restrictions on preferences. What do these restrictions imply

for indifference curves? We have the following properties:
1. If an indifference curve is further from the origin compared to another indifference

curve, any point on the former is preferred to any point on the latter (implied by
the assumption that more is better).
2. Indifference curves cannot slope upwards (implied by more is better).
3. Indifference curves cannot be thick (again, implied by more is better).
4. Indifference curves cannot cross (implied by transitivity).
5. Every bundle of goods lies on some indifference curve (follows from completeness).
The marginal rate of substitution
A further important property concerns the rate at which a consumer is willing to

substitute one good for another along an indifference curve. The marginal rate of
substitution of a consumer between goods x and y is the units of y the consumer is
willing to substitute (i.e. willing to give up) to obtain one more unit of x.
The slope of an indifference curve (with good y on the y-axis and good x on the x-axis)
is given by:
dy MUx
= .
dx u constant
MUy
The marginal rate of substitution is the absolute value of the slope:
MUx
MRSxy = .
MUy
Typically, preferences have the following property. Consider a point where a lot of y and
very little x is being consumed. Starting from any such point, a consumer is willing to
give up a lot of y in exchange for another unit of x while retaining the same level of
utility as before. As we keep adding units of x and reducing y while keeping utility
constant (i.e. we are moving down an indifference curve), consumers are willing to give
up less and less of y in return for a further unit of x. One way to interpret this is that
14
people typically have a taste for variety and want to avoid extremes (i.e. avoid
situations where a lot of one good and very little of the other good is being consumed).
This implies that MRS falls along an indifference curve. This property is referred to as
indifference curves being convex to the origin in some textbooks. This works as a
visual description, but you should be aware that in terms of mathematics, this is not a
meaningful description there is no mathematical concept where something is convex
relative to something else. The correct idea of convex indifference curves is as follows.
Consider a subset S of Rn . S is a convex set if the following property holds: if points s1
and s2 are in S, then a convex combination s1 + (1 )s2 is also in the set S for any
0 < < 1.
Now consider any indifference curve yielding utility level u. Consider the set of all
points that yield utility u or more. This is the set of all points on an indifference curve
plus all points above. Call this set B. Diminishing MRS implies that B is a convex set.
Figure 2.1 below shows a convex indifference curve. Note that the set B (part of which
is shaded) is a convex set. Try taking any two points in B and then making a convex
combination. You will find that the combinations are always inside B.
Figure 2.1: A convex indifference curve.
Next, Figure 2.2 shows an example of non-convex indifference curves. Note that the set
B of points on or above the indifference curve is not convex. If you combine points such
as a1 and a2 in the diagram, for some values of , the convex combinations fall outside
the set B.
Note that when two goods are perfect substitutes, you get a straight line indifference
curve. At the other extreme, the two goods are perfect complements (no
substitutability) and the indifference curve is L-shaped. Indifference curves with
diminishing MRS lie in between these two extremes.
15
2. Consumer theory
Figure 2.2: A non-convex indifference curve.
Here is an activity to get you computing the MRS for different utility functions.
Activity 2.2 Compute the MRS for the following utility functions.

(a) u(x, y) = xy.
(b) u(x, y) = ln x + ln y.
(c) u(x, y) = 20 + 3(x + y)2 .
2.3.3 Budget constraint
Once we have specified our model of preferences, we need to know the set of goods that
a consumer can afford to buy. This is captured by the budget constraint. Since
consumers are generally taken to be price-takers (i.e. what an individual consumer
purchases does not affect the market price for any good), the budget line is a straight
line. See Section 2.7 of N&S for the construction of budget sets. You should be aware
that budget lines would no longer be a straight line if a consumer buys different units at
different prices. This could happen if a consumer is a large buyer in a market or if the
consumer gets quantity discounts. See Application 2.6 in N&S for an example.
2.3.4 Utility maximisation
The consumer chooses the most preferred point in the budget set. If preferences are
such that indifference curves have the usual convex shape, the best point is where an
indifference curve is tangent to the budget line. This is shown as point A in Figure 2.3.
At A the slope of the indifference curve coincides with the slope of the budget
16
Figure 2.3: Consumer optimisation.
constraint. So we have:
MUx Px
= .
MUy Py
Multiplying both sides by 1 we can write this as the familiar condition:
Px
MRSxy = .
Py
Let us derive this condition formally using a Lagrange multiplier approach. This is the
approach you are expected to use when faced with optimisation problems of this sort.
Note that the more is better assumption ensures that a consumer spends all income (if
not, then the consumer could increase utility by buying more of either good). Therefore,
the budget constraint is satisfied with equality. It follows that the consumer maximises
u(x, y) subject to the budget constraint Px x + Py y = M . Set up the Lagrangian:
L = u(x, y) + (M Px x + Py y).
The first-order conditions for a constrained maximum are:

L u
= Px = 0
x x
L u
= Py = 0
y y
L
= M Px x + Py y = 0.

From the first two conditions, we get:
Px u/x
= = MRSxy .
Py u/y
17
2. Consumer theory
Second-order condition
The first-order conditions above are, by themselves, not sufficient to guarantee a

maximum. We also need the second-order condition to hold. It is better to derive this
formally once you have learned matrix algebra, which allows a relatively simple
exposition of the second-order condition. For our purposes here, note that the
diminishing MRS condition is sufficient to guarantee that a maximum occurs at the
point satisfying the first-order conditions. This should also be clear to you from the
graph. If indifference curves satisfy the usual convexity property, there is an interior
tangency point with the budget constraint line at which the maximum utility is
attained.
Figure 2.4 below demonstrates that if preferences are not convex, the first-order
conditions are not sufficient to guarantee optimality.
Figure 2.4: Violation of the second-order condition under non-convex preferences. Note
that MRS is not always diminishing. Point A satisfies the first-order condition MRS equal
to price ratio, but is not optimal. (Source: Schmalensee, R. and R.N. Stavins (2013) The
SO2 allowance trading system: the ironic history of a grand policy experiment, Journal
of Economic Perspectives, Vol. 27, pp.10321. Reproduced by kind permission of the
American Economic Association.)
Read Sections 2.7 to 2.9 of N&S carefully and work through all the examples therein.
Note that if two goods are perfect substitutes or complements, the tangency condition
does not apply. For perfect substitutes, there is either a corner solution or the budget
line coincides with the indifference curve. In the latter case, any point on the budget
line is optimal. For the case of perfect complements, the optimum occurs where the kink
in the indifference curve just touches the budget line. Note that this is not a tangency
point the slope of the indifference curve is undefined at the kink. N&S clarifies these
cases with appropriate diagrams.
18
Demand functions
The maximisation exercise above gives us the demand for goods x and y at given prices
and income. As we vary the price of good x, we can trace out the demand curve for
good x. See Section 3.6 of N&S for a discussion. The activities below compute demand
functions in specific examples.
Example 2.1 A consumer has the following CobbDouglas utility function:
u(x, y) = x y
where , > 0. The price of x is normalised to 1 and the price of y is p. The

consumers income is M . Derive the demand functions for x and y.
The consumers problem is as follows:
max x y subject to x + py = M.
x, y
Using the Lagrange multipliers method, we get:

MUx Px
= .
MUy Py
This implies:
x1 y 1
= .
x y 1 p
Simplifying:
y 1
= .
x p
Using this in the budget constraint and solving, we get the demand functions:
M
x(p, M ) =
+
M
y(p, M ) = .
p( + )
Example 2.2 A consumer has the following utility function:
u(x, y) = min{x, y}
where , > 0. The price of x is normalised to 1 and the price of y is p. The

consumers income is M . Derive the demand functions for x and y.
The consumer would choose the bundle at which the highest indifference curve is
reached while not exceeding the budget. This is point E in Figure 2.5 where the
indifference curve just touches the budget constraint (Figure 2.5 is drawn using
/ = 1/2). Note that this is not a tangency point as the slope of the indifference
curve is undefined at the kink.
19
2. Consumer theory
Since we are at the kink, it must be that x = y. Using this in the budget
constraint, we get the demand functions:
M
x(p, M ) =
+ p
M
y(p, M ) = .
+ p
Figure 2.5: The optimum occurs at point E. Note that this is not a tangency point. The
slope of the indifference curve at the kink is undefined.
2.4 Demand curves
2.4.1 The impact of income and price changes

See N&S Chapter 3. Now that we have derived demand curves, we can try to
understand various properties of demand by varying income and prices.
Income changes
Section 3.2 of N&S explains the classification of goods according to the response of
demand to income changes.
Normal goods: a consumer buys more of these when income increases.
Inferior goods: a consumer buys less of these when income increases.
20
2.4. Demand curves
Note that it is not possible for all goods to be inferior. This would violate the more is
better assumption. The full argument is left as an exercise.
Activity 2.3 It is not possible for all goods to be inferior. Provide a careful
explanation of this statement.
The income-consumption curve
The income-consumption curve of a consumer traces out the path of optimal bundles as
income varies (keeping all prices constant). Using this exercise, we also plot the
relationship between quantity demanded and income directly. The curve that shows this
relationship is called the Engel curve. See Perloff Section 4.2 for an exposition of the
income-consumption curve and the Engel curve.
The slope of the income-consumption curve indicates the sign of the income elasticity of
demand (explained below).
Price changes
See Sections 3.3 to 3.8 of N&S. It is very important to understand fully the
decomposition of the total price effect into income and substitution effects. This
decomposition is, of course, a purely artificial thought experiment. But this thought
experiment is extremely useful in understanding how the demand for different goods
responds to a change in price at different levels of income and given different
opportunities to substitute out of a good. You should understand how these effects
(and, therefore, the total price effect) differ across normal and inferior goods, and
understand how the effect known as Giffens paradox can arise.
The idea of income and substitution effects can help us understand the design of an
optimal tax scheme. See Section 3.4 of N&S for a discussion of this issue.
Finally, you should also study the impact on the demand for a good by changes in the
price of some other good, and how this effect differs depending on whether the other
good is a substitute or a complement.
Example 2.3 Suppose u(x, y) = x1/2 y 1/2 . Income is M = 72. The price of y is 1
and the price of x changes from 9 to 4. Calculate the income effect (IE) and the
substitution effect (SE).
Let (px , py ) denote the original prices and let p0x denote the lower price of x.
Under the original prices, the Marshallian demand functions (you should be able to
calculate these) are:
M
x =
2px
and:
M
y = .
2py
21
2. Consumer theory
The optimised utility is, therefore:

M
u0 =
2 px p y
where the subscript of u is a reminder that this is the original utility level (before
the price change).
The total price effect (PE) from a price fall is:
M M
PE = 0
.
2px 2px
Using the values supplied, this is 9 4 = 5.

In Figure 2.6, the movement from A to C is the total price effect.
To isolate the SE, we must change the price of x, but also take away income so that
the consumer is on the original indifference curve. In other words, we must keep the
utility at u0 . In Figure 2.6, the dashed budget line is the one after the compensating
reduction in income. The point B is the optimal point on this compensated budget
line. The movement from the original point A to B shows the substitution effect.
How much income should we take away to compensate for the price change? This
can be calculated as follows.
Suppose M is the amount of income we take away. We need M to be such that:
M M
p
0
= u0
2 p x py
which implies:
M M M
p = .
2 p0x py 2 px py
Using the values supplied, M = 24 so that M M = 48.
Under a reduced income of 48, and given the new price p0x = 4, the demand for x is
6. The original demand for x was 4. Under the compensated price change, the
demand is 6. Therefore, the SE is 2. It follows that the rest of the change must be
the IE. Since the total price effect is 5, the IE is 3.
In terms of algebra:
M M M
SE = 0
.
2px 2px
The IE is the remainder of the price effect, so that:

M M M M M
IE = .
2p0x 2px 2p0x 2px
Simplifying:
M
IE = .
2p0x
Looking ahead, the M we calculated here is known as the compensating
variation. We will study this concept later in this chapter.
22
2.4. Demand curves
Figure 2.6: As the price of x falls, the change from A to C shows the total price effect.
The movement from A to B (under a compensated price change) shows the substitution
effect, while the movement from B to C shows the income effect.
The market demand curve
From individual demand curves, we can construct the market demand curve by
aggregating across individuals. See N&S Section 3.10 for a discussion.
2.4.2 Elasticities of demand

The notion of elasticity of demand captures the responsiveness of demand to variables
such as prices and income. The elasticity of market demand for a good can be estimated
from data, and these elasticity estimates are important for firms in setting prices and for
formulation of policy. Throughout the course, we will come across several such examples.
Sections 3.11 to 3.16 of N&S contain a detailed analysis of elasticities, which you must
read carefully. Here, let us summarise the main concepts.
Price elasticity of demand
This is the percentage change in quantity demanded of a good in response to a given

percentage change in the price of the good, given by:
dQ/Q P dQ
= = .
dP/P Q dP
Note that < 0 since demand is typically downward-sloping. Demand is said to be

elastic if < 1, unit elastic if = 1, and inelastic if > 1.
N&S outlines a variety of uses of this concept, which you should read carefully. You
should know how to calculate demand elasticity at different points on a demand curve,
and how the elasticity varies along a linear demand curve.
23
2. Consumer theory
Price elasticity of demand is the most common measure of elasticity and often referred
to as just elasticity of demand.
Other than price elasticity, we can define income elasticity and cross-price elasticity.
Income elasticity
Denoting income by M , income elasticity of demand is given by:

P dM
M = .
M dP
This is positive for normal goods, and negative for inferior goods. When M exceeds 1,
we call the good a luxury good. Necessities like food have income elasticities much lower
than 1.
Cross-price elasticity
Let us consider the elasticity of demand for good i with respect to the price of good j.
The cross-price elasticity of demand for good i is given by:
Pj dQi
ij = .
Qi dPj
This is negative for complements and positive for substitutes.
You should take a long look at the elasticity estimates presented in Section 3.16 of
N&S. Practical knowledge of elasticities forms an important part of designing and
understanding a variety of tax and subsidy policies in different markets.
Activity 2.4 Suppose u(x, y) = x y , where + = 1. Income is M . Calculate the

price elasticity, cross-price elasticity and income elasticity of demand for x.
2.4.3 The compensated demand curve

We derived the demand function for a good above. To derive the demand function for
good x, we vary the price of good x but hold constant the prices of other goods and
income. Of course, as the price changes so that the optimal choice changes, the utility of
the consumer at the optimal point also changes. This is the usual demand curve, and is
also known as the Marshallian demand curve, or the uncompensated demand curve.
Indeed, if we simply mention a demand curve without putting a qualifier before it, it
refers to the Marshallian, or uncompensated, demand curve.
A compensated, or Hicksian, demand curve can be derived as follows. Suppose as the
price of a good changes, we keep utility constant while allowing income to vary. In other
words, if the price of x, say, falls (so that the new optimal bundle of the consumer
would be associated with a higher level of utility if income is left unchanged), we take
away enough income to leave the consumer at the original level of utility. It is clear that
this process eliminates the income effect and simply captures the substitution effect.
Below, we list some properties of compensated demand curves.
24
2.4. Demand curves
A compensated demand curve always slopes downwards.
For a normal good, the compensated demand curve is less elastic compared to the
uncompensated demand curve.
For an inferior good, the compensated demand curve is more elastic compared to
the uncompensated demand curve.
You should understand that all three properties result from the fact that only the
substitution effect matters for the change in compensated demand when the price
changes.
The next example asks you to calculate the compensated demand curve for a
CobbDouglas utility function.
Example 2.4 Suppose u(x, y) = x1/2 y 1/2 . Income is M . Calculate the compensated
demand curves for x and y.
To do this, we must first calculate the Marshallian demand curves. These are given
by (you should do the detailed calculations to show this):
M M
x= and y= .
2px 2py
The optimised value of utility is:

M
V = .
2 px py
Holding utility constant at V implies adjusting M to the value M so that:

M = 2V px py .
This is the value of income which is compensated to keep utility constant at the level
given by the original choice of x and y. It follows that the compensated demand
functions are:
M
r
py
xc = =V
2px px
and:
M px
r
yc = =V .
2py py
Note that the Marshallian demand for x does not depend on py , but the Hicksian, or
compensated, demand does. This is because changes in py require income
adjustments, which generate an income effect on the demand for x.
2.4.4 Welfare measures: CS, CV and EV

See Section 3.9 of N&S for a discussion of consumer surplus, but this does not cover the
other two measures: compensating variation (CV) and equivalent variation (EV). We
provide definitions and applications below.
25
2. Consumer theory
When drawing demand curves, we typically draw the inverse demand curve (price on
the vertical axis, quantity on the horizontal axis). In such a diagram, the consumer
surplus (CS) is the area under the (inverse) demand curve and above the market price
up to the quantity purchased at the market price. This is the most widely-used measure
of welfare. We can measure the welfare effect of a price rise by calculating the change in
CS (denoted by CS).
Much of our discussion of policy will be based on this measure. Any part of CS that
does not get translated into revenue or profits is a deadweight loss. The extent of
deadweight loss generated by any policy is a measure of inefficiency associated with that
policy.
However, CS is not an exact measure because of the presence of an income effect.
Ideally, we would use the compensated demand curve to calculate the welfare change.
CV and EV give us two such measures. You should use these measures to understand
the design of ideal policies, but when measuring welfare change in practice, use CS.
Compensating variation (CV)
CV is the amount of money that must be given to a consumer to offset the harm
from a price increase, i.e. to keep the consumer on the original indifference curve
before the price increase.
Equivalent variation (EV)
EV is the amount of money that must be taken away from a consumer to cause as
much harm as the price increase. In this case, we keep the price at its original level
(before the rise) but take away income to keep the consumer on the indifference curve
reached after the price rise.
Comparing the three measures
Consider welfare changes from a price rise. For a normal good, we have
CV > CS > EV, and for an inferior good we have CV < CS < EV. The measures
would coincide for preferences that exhibit no income effect.
The example that follows shows an application of these concepts.
Example 2.5 The government decides to give a pensioner a heating fuel subsidy of
s per unit. This results in an increase in utility from u0 before the subsidy to u1
after the subsidy. Could the government follow an alternative policy that would
result in the same increase in utility for the pensioner, but cost the government less?
Let us show that an equivalent income boost would be less costly. Essentially, the
EV of a price fall is lower than the expenditure on heating after the price fall. The
intuition is that a per-unit subsidy distorts choice in favour of consuming more
heating, raising the total cost of the subsidy. To put the same idea differently, an
equivalent income boost would raise the demand for fuel through the income effect.
26
2.4. Demand curves
But a price fall (the fuel subsidy results in a lower effective price) causes an
additional substitution effect boosting the demand for heating.
To see this, consider Figure 2.7. The initial choice is point A and after the subsidy
the pensioner moves to point B. How much money is the government spending on
the subsidy? Note that after the subsidy, H1 units of heating fuel are being
consumed. At pre-subsidy prices, buying H1 would mean the pensioner would have
E 0 of other goods. Since the price of the composite good is 1, M is the same as total
income. It follows that the amount of income that would be spent on heating to buy
H1 units of heating at pre-subsidy prices is given by M E 0 . Similarly, at the
subsidised price, the amount of income being spent on heating fuel is M B 0 . The
difference B 0 E 0 is then the amount of the subsidy. This is the same length as
segment BE.
Once we understand how to show the amount of spending on the subsidy in the
diagram, we are ready to compare this spending with an equivalent variation of
income. This is added in Figure 2.8 below.
The pensioners consumption is initially at A, and moves to B after the subsidy.
Since the composite good has a price of 1, the vertical distance between the budget
lines (segment BE) shows the extent of the expenditure on the subsidy (as explained
above). An equivalent variation in income, on the other hand, would move
consumption to C. It follows that DE is the equivalent variation in income, which is
smaller than the expenditure on the subsidy. Therefore, a direct income transfer
policy would be less costly for the government.
Figure 2.7: The segment BE shows the extent of the subsidy.
27
2. Consumer theory
Figure 2.8: Per-unit subsidy versus an equivalent variation in income.
Example 2.6 Suppose that a consumer has the utility function

u(x1 , x2 ) = x1/2 y 1/2 . He originally faces prices (1, 1) and has income 100. Then the
price of good 1 increases to 2. Calculate the compensating and equivalent variations.
Suppose income is M and the prices are p1 and p2 . You should work out that the
demand functions are:
M M
x1 = and x2 = .
2p1 2p2
Therefore, utility is:
M
u (p1 , p2 , M ) = .
2 p1 p2

At the initial prices, u = M/2. Once the price of good 1 increases, u = M/2 2.
The CV is the extra income that restores utility to the original level. Therefore, it is
given by:
M + CV M
= .
2 2 2
Solving:
CV = ( 2 1)M.
Using the value M = 100, this is 41.42.
The EV is the variation in income equivalent to the price change. This is given by:
M M EV
= .
2 2 2
Solving:
( 2 1)M
EV = .
2
Using M = 100, this is 29.29.
28
2.5. Labour supply
2.5 Labour supply

See Appendix 13A of N&S. The analysis presented here complements the somewhat
basic coverage in the textbook.
The tools developed above can also be used to analyse the labour supply decision of an
agent. Every economic agent has, in a day, 24 hours. An agent must choose how many
of these hours to spend working, and how many hours of leisure to enjoy. Working earns
the agent income, which represents all goods the agent can consume. But the agent also
enjoys leisure. If the hourly wage is w, this can be seen as the price that the agent must
pay to enjoy an hour of leisure.
The agent, therefore, faces the following problem. Let Z denote the number of hours
worked, N denote the number of leisure hours, M denote income and M denote
unearned income (inheritance, gifts etc). The utility maximisation problem is:
max u(N, M )
N, M
subject to:
Z = 24 N
M = wZ + M .
We can simplify the constraints to M = w(24 N ) + M , or M + wN = 24w + M .

Therefore, we have a familiar utility maximisation problem:
max u(N, M )
N, M
subject to the budget constraint:
M + wN = 24w + M .
At the optimum we have the slope of the indifference curve (MRS) equal to the slope
of the budget line (w). Therefore:
MUN
= w.
MUM
How does the optimal choice of labour respond to a change in w? We can analyse this
using income and substitution effects. In this case, a change in w also changes income
directly (as you can see from the budget constraint), so the exercise is a little different
compared to that under standard goods.
Suppose w rises.
Income effect
The rise in w raises income at current levels of labour and leisure. Assuming leisure
is a normal good (this should be your default assumption), this raises the demand
for leisure.
29
2. Consumer theory
Substitution effect
The rise in w makes leisure relatively more expensive, causing the agent to substitute
away from leisure. This reduces demand for leisure.
The two effects go in opposite directions, therefore the direction of the total effect is
unclear. In most cases, the substitution effect dominates, giving us an upward-sloping
labour supply function. However, it is possible, especially at high levels of income (i.e.
when wage levels are high), that the income effect might dominate. In that case we
would get a backward-bending labour supply curve which initially slopes upward but
then turns back and has a negative slope.
If leisure is, on the other hand, an inferior good, the two effects would go in the same
direction and labour supply would necessarily slope upwards.
Figure 2.9 (a) below shows a backward-bending labour supply curve while Figure 2.9
(b) shows an increasing labour supply curve.
Figure 2.9: Labour supply curves.
2.6 Saving and borrowing: intertemporal choice

See N&S Sections 14.1 and 14.2. The discussion below complements the somewhat basic
discussion in the textbook.
We focus on a two-period problem. Suppose the agents endowment is Y0 in period 0
and Y1 in period 1. Given a rate of interest r, the present value of income in period 0 is
Y0 + Y1 /(1 + r). If the individual consumes all income in period 0, C0 is equal to this
present value, and C1 = 0. If all income is saved for period 1, then income at period 1 is
(1 + r)Y0 + Y1 . In this case, C1 is this amount and C0 = 0. Therefore, we have
30
2.6. Saving and borrowing: intertemporal choice
C1 = (Y0 C0 )(1 + r) + Y1 , which gives us the intertemporal budget constraint. The

problem is then as follows:
max u(C0 , C1 )
C0 , C 1
subject to the intertemporal budget constraint:
C1 Y1
C0 + = Y0 + .
1+r 1+r
This is similar to a standard optimisation problem with two goods, C0 and C1 , where
the price of the former is 1 and the price of the latter is 1/(1 + r). Unsurprisingly, the
optimum satisfies the property that:
MUC0
= 1 + r.
MUC1
If the optimal consumption bundle is C0 = Y0 and C1 = Y1 , the agent is neither a saver

nor a borrower. If C0 > Y0 the agent is a borrower, and if C0 < Y0 the agent is a saver
(lender).
How does the intertemporal consumption bundle change when r changes? Again, we can
see this by decomposing the effect into income and substitution effects. Let us look at
the problem of borrowers and savers separately.
Throughout the following analysis, we assume that both C0 and C1 are normal goods.
This should be your default assumption.
Note that the total income available to consume in period 1 is Y1 + (Y0 C0 )(1 + r).
The problem of a borrower
Income effect
For a borrower, Y0 < C0 , so that a rise in the interest rate lowers income tomorrow.
Given consumption is normal in both periods, the agent should consume less in
period 0 (borrow less).
Substitution effect
A rise in the interest rate makes immediate consumption more costly. Therefore,
the substitution effect suggests that the individual should choose to lower C0 and,
therefore, borrow less.
Since the two effects go in the same direction, the direction of change is unambiguous: a
rise in the rate of interest lowers borrowing.
31
2. Consumer theory
The problem of a saver (lender)
Income effect
For a saver, Y0 > C0 , so that a rise in the interest rate raises income tomorrow. Given
consumption is normal in both periods, the agent should consume more in period 0
(save less).
Substitution effect
A rise in the interest rate makes immediate consumption more costly. Therefore,
the substitution effect suggests that the individual should choose to lower C0 (save
more).
Since the two effects go in opposite directions, the total effect on saving is uncertain.
Usually, the substitution effect dominates so that agents save less when the interest rate
rises, but it could go the other way.
Figure 2.10 shows the intertemporal budget constraint. The endowment point is
(Y0 , Y1 ). As the rate of interest increases, the budget constraint pivots around the
endowment point as shown.
Figure 2.10: Intertemporal budget constraint.
Note that consumers reaching an optimum in the part of the budget constraint above
the endowment point are savers, and those reaching an optimum somewhere in the
lower part are borrowers.
32
2.7. Present value calculation with many periods
Activity 2.5 Using Figure 2.10 above, explain that a saver cannot become a
borrower if the rate of interest rises.
2.7 Present value calculation with many periods

The previous section discussed the calculation of present value for a two-period stream
of payoffs. We can extend this easily to multiple (or infinite) periods. This calculation is
useful in many cases for example, in calculating the repeated game payoff in game
theory. This is also useful in understanding bond pricing.
In this course, you need to know only the basics, which we present below.
Suppose we have a stream of payoffs y0 , y1 , . . . , yn in periods 0, 1, . . . , n, respectively.
Suppose the rate of interest is given by r. The present value in period 0 of this stream
of payoffs is given by:
y1 y2 yn
PV = y0 + + + + .
1 + r (1 + r)2 (1 + r)n
If we write = 1/(1 + r), we can write this as:
PV = y0 + y1 + 2 y2 + + n yn .
Suppose y0 = y1 = = yn = y. In this case:
PV = y(1 + + 2 + + n ).
We can sum this as follows. Let:
S = 1 + + 2 + + n.
Then:
S = + 2 + + n+1 .
We have:
S S = 1 n+1 .
Therefore:
1 n+1
S= .
1
So:
1 n+1
PV = y .
1
If the payoff stream is infinite, the present value is very simple:

2 1
PV = y(1 + + + ) = y .
1
33
2. Consumer theory
2.7.1 Bonds
A bond typically pays a fixed coupon amount x each period (next period onwards) until
a maturity date T , at which point the face value F is paid. The price of the bond, P , is
simply the present value given by:
P = x + 2 x + + T F.
Note that the price declines if falls, which happens if r rises. Therefore, the price of a
bond has an inverse relationship with the rate of interest.
A special type of bond is a consol or a perpetuity that never matures. The price of a
consol has a particularly simple expression:

P = x + 2 x + = x .
1
Now = 1/(1 + r). Therefore:
1/(1 + r) 1
= = .
1 r/(1 + r) r
It follows that:
x
P = .
r
This makes the inverse relationship between P and r clear.
2.8 A reminder of your learning outcomes

Having completed this chapter, the Essential reading and activities, you should be able
to:
explain the implications of the assumptions on the consumers preferences
describe the concept of modelling preferences using a utility function
draw indifference curve diagrams starting from the utility function of a consumer
draw budget lines for different prices and income levels
solve the consumers utility maximisation problem and derive the demand for a
consumer with a given utility function and budget constraint
analyse the effect of price and income changes on demand
explain the notion of a compensated demand function
explain measures of the welfare impact of a price change: change in consumer

surplus, equivalent variation and compensating variation, and use these measures
to analyse the welfare impact of a price change in specific cases
construct the market demand curve from individual demand curves
34
2.9. Test your knowledge and understanding
explain the notion of elasticity of demand
analyse the decision to supply labour
analyse the problem of savers and borrowers
derive the present discounted value of payment streams and explain bond pricing.
2.9 Test your knowledge and understanding
2.9.1 Sample examination questions

1. Indifference curves of an agent cannot cross. Is this true or false? Explain.
2. The Hicksian demand curve for a good must be more elastic than the Marshallian
demand curve for a good. Is this true or false? Explain.
3. Savers gain more when the rate of interest rises. Is this true or false? Explain.
4. Consider the utility function u(x, y) = x2 + y 2 .

(a) Does this satisfy the property of diminishing MRS? Show algebraically, and
also show by drawing indifference curves.
(b) Show that using the tangency condition (MRS equals price ratio) would not
lead to an optimum in this case.
(c) Show (in a diagram) the possible optimal bundles.
5. Consider the quasilinear utility function u(x1 , x2 ) = ln x1 + x2 (this is linear in x2 ,

but not in x1 , hence the name quasilinear). Let p1 and p2 denote the prices of x1
and x2 , respectively. Let m denote income.
(a) Calculate the demand functions.
(b) Draw the income-consumption curve.
(c) Calculate the price elasticity of demand for each good.
(d) Calculate the income elasticity of demand for each good.
35
2. Consumer theory
36
Chapter 3
Choice under uncertainty
3.1 Introduction
In the previous chapter, we studied consumer choice in environments that had no
element of uncertainty. However, many important economic decisions are made in
situations involving some degree of risk. In this chapter, we cover a model of
decision-making under uncertainty called the expected utility model. The model
introduces the von NeumannMorgenstern (vNM) utility function. This is unlike the
ordinal utility functions we saw in the previous chapter and has special properties. In
particular, the curvature of the vNM utility function can indicate a consumers
attitude towards risk. Once we introduce the model, we use it to derive the demand for
insurance and also introduce a measure of the degree of risk aversion.

This chapter aims to introduce the expected utility model which tells us how a
consumer evaluates a risky prospect. We aim to show how this helps us understand
attitudes towards risk and analyse the demand for insurance. We also aim to set up a
measure of the degree of risk aversion.

calculate the expected value of a gamble
explain the nature of the vNM utility function and calculate the expected utility
from a gamble
explain the different risk attitudes and what they imply for the vNM utility
function
analyse the demand for insurance and show the relationship between insurance and
premium
explain the concept of diversification
calculate the ArrowPratt measure of risk aversion for different specifications of the
vNM utility function.
37
3. Choice under uncertainty

N&S Sections 4.1, 4.2 and 4.3 up to and including the discussion on diversification (up
to page 135). N&S does not cover the expected utility model or the ArrowPratt
measure of risk aversion. We provide details below. These topics are also covered in
Perloff Section 16.2 (exclude the last part on willingness to gamble).

This chapter covers expected utility theory and uses the theory to derive the demand
for insurance. It also covers the ArrowPratt measure of risk aversion.
3.3 Preliminaries
You should already be familiar with concepts such as probability and expected value.
Do familiarise yourself with these concepts if this is not the case. Section 4.1 of N&S
discusses these.
Random variable
A variable that represents the outcomes from a random event. A random variable
has many possible values, and each value occurs with a specified probability.
Expected value of a random variable
Suppose X is a random variable that has values x1 , . . . , xn . For each i = 1, 2, . . . , n,

the value xi occurs with probability pi , where p1 + p2 + + pn = 1. The expected
(or average) value of X is given by:
X
E(X) = p1 x1 + p2 x2 + + pn xn = p i xi .
i
3.4 Expected utility theory

Expected utility theory was developed by John von Neumann and Oscar Morgenstern
in their book The Theory of Games and Economic Behavior. (Princeton University
Press, 1944; expected utility appeared in an appendix in the second edition in 1947).
A proper exposition of their theory must await a Masters level course, but let us try to
give a rough idea of what is involved.
Suppose an agent faces a gamble G that yields an amount x1 with probability p1 , x2
with probability p2 , . . . , and xn with probability pn . How should the agent evaluate this
gamble? Von Neumann and Morgenstern specified certain axioms, i.e. restrictions on
38
3.5. Risk aversion
choice under uncertainty that might be deemed reasonable. They showed that under
their axioms, there exists a function u such that the gamble can be evaluated using the
following expected utility formulation:
E(U (G)) = p1 u(x1 ) + p2 u(x2 ) + + pn u(xn ).
The function u is known as the von NeumannMorgenstern (vNM) utility function.
The vNM utility function is somewhat special. It is not entirely an ordinal function
like the utility functions you saw in the last chapter. Starting from a vNM u function,
we can make transformations of the kind a + bu, with b > 0 (these are called positive
affine transformations), without changing the expected utility property but not any
other kinds of transformations (for example, u2 is not allowed). The reason is that, as
we discuss below, the curvature of the vNM utility function captures attitude towards
risk. Transformations other than positive affine ones change the curvature of this
function, and therefore the transformed u function would not represent the same
risk-preferences as the original. Thus vNM utility functions are partly cardinal.
Note that the expected utility representation is very convenient. Once we know the
vNM utility function u, we can evaluate any gamble easily by simply taking the
expectation over the vNM utility values.
3.5 Risk aversion

We can show that an agent with a concave vNM utility function over wealth is
risk-averse. Let us show this by establishing that an agent with a concave u function
would reject a fair gamble.
Recall that a function f (W ) is concave if f 00 (W ) < 0, i.e. the second derivative of the
function with respect to W is negative.
Suppose G is a gamble which yields 20 with probability 1/2, and 10 with probability
1/2. Suppose an agent has wealth 15 and is given the following choice: invest 15 in
gamble G, or do nothing. Note that the expected value of the gamble, E(G), is exactly
15, so that this is a fair gamble (the expected wealth is the same whether G is accepted
or rejected).
An agent who simply cared about expected value, and not about risk, would be
indifferent between accepting and rejecting G. However, a risk-averse individual would
reject a fair gamble.
The expected utility of an agent from G is:
1 1
E(U (G)) = u(20) + u(10).
2 2
As Figure 3.1 shows, given a concave u-function:
E(U (G)) < u(15) = u(E(G)).
Therefore, the agent would not accept a fair gamble. This shows that a concave
u-function implies risk aversion.
Note that one of the implications of a concave vNM utility function is that the
marginal utility of wealth is declining. The point is noted in N&S Section 4.2.
39
Figure 3.1: The vNM utility function for a risk-averse individual. Note that the function
is concave and u(E(G)) > E(U (G)) so that the agent does not accept a fair gamble.
3.6 Risk aversion and demand for insurance

A risk-averse individual would pay to obtain insurance. To see that, it is useful to define
the certainty equivalent (CE) of a gamble. The CE of a gamble is the certain wealth
that would make an agent indifferent between accepting the gamble and accepting the
certain wealth. As Figure 3.1 shows, the CE is lower than the expected income of 10.
Suppose an agent simply faced gamble G (i.e. did not have the choice between G and
10, but simply faced G). Clearly, since the CE is lower than the expected outcome of G,
this agent would be willing to pay a positive amount to buy insurance. How much
would the agent be willing to pay? The amount an agent pays for insurance is called the
risk premium.
We now work through an example to understand how to calculate the risk premium.
3.6.1 Insurance premium for full insurance
Kims utility depends on wealth W . Kims vNM utility function is given by:

u(W ) = W.
Kims wealth is uncertain. With probability 0.5 wealth is 100, and with probability 0.5
a loss occurs so that wealth becomes 64. In what follows, we will assume that Kim can
only buy full insurance. In other words, the insurance company offers to pay Kim 36
whenever the loss occurs and in exchange Kim pays them a premium of R in every state
(i.e. whether the loss occurs or not). In what follows, we will calculate the maximum
and minimum value of R.
40
3.6. Risk aversion and demand for insurance
Let us first calculate Kims expected utility. This is given by:

E(U ) = 0.5 100 + 0.5 64 = 9.
How do we know Kim would be prepared to pay to buy full insurance? You can draw a
diagram as above to show that Kim would be prepared to pay a positive premium if
wealth is fully insured. Alternatively, you could point out that the expected utility of
the uncertain wealth (which is 9) is lower than the utility of expected wealth since:

u(E(W )) = 0.5 100 + 0.5 64 = 82 = 9.055.
This implies that the premium that Kim is willing to pay is positive.

You could also point out that W is a concave function (check that the second
derivative is negative), implying that Kim is risk-averse. Then draw the CE point as
above and point out that since expected wealth exceeds the certainty equivalent, the
premium is positive.
Let us now calculate the maximum premium that Kim would be willing to pay to
buy full insurance.
First, calculate the certainty equivalent of the gamble Kim is facing. This is given by:
u(CE) = E(U ).
Therefore:
CE = 9
implying that CE = 81. Therefore, the maximum premium Kim is willing to pay is
100 81 = 19.
There is another way of finding this, which considers the maximum premium in terms of
expected wealth (i.e. how much expected wealth would Kim give up in order to fully
insure?). You need to understand this, since some textbooks use this way of identifying
the premium. For example, this is the approach adopted by Perloff. Unfortunately,
textbooks (including Perloff) never make clear exactly what they are doing, which can
be very confusing for students. Reading the exposition here should clarify the matter
once and for all.
The maximum premium in terms of expected wealth is calculated as follows. Note that
under full insurance the gross expected wealth Kim would receive is E(W ) = 82. We
also know that CE = 81. Therefore, the maximum amount of expected wealth Kim
would give up is 82 81 = 1. (Note that this should explain why in the diagram on risk
aversion in Perloff Section 16.2, and subsequent solved problems, the risk premium is
identified as the difference between expected wealth and the CE.)
To connect this approach to the one above, consider the actual premium and coverage.
Kim loses 36 with probability 0.5. So full insurance means a coverage of 36, which is
paid when the loss occurs. In return, Kim pays an actual premium of 19 in each state.
Therefore, the change in expected wealth for Kim is:
0.5 (19) + 0.5 (36 19) = 18 19 = 1.
In other words, Kim is giving up 1 unit of expected wealth, as shown above.
41
Once again, the purpose of writing this out in detail is to make you aware that
textbooks vary in their treatment of this. Some talk about premium in terms of
expected wealth, while others calculate the actual premium, but they do not make it
clear what it is that they are doing. In answering questions of this sort in an
examination, it is easiest (and clearest) to calculate the actual premium. You can follow
the other route and define the premium in terms of expected wealth, but in that case
you should make that clear in your answer.
Next, we calculate the minimum premium.
Assuming the insurance company is risk-neutral, it must break even. So the minimum
premium (or fair premium) is Rmin such that it equals the expected payout, which is
0.5 (100 64) = 18. (Note that this is simply 100 E(W ), where E(W ) is expected
wealth, which is 82 in this case.)
As above, the other way of answering the question is to say that in terms of the expected
wealth that Kim needs to give up, the minimum is zero. Think of this as follows. Kim
simply hands over her actual wealth to the insurance company, and in return receives
the expected wealth in all states. The insurance company is risk-neutral, and in
expected wealth terms it is giving and receiving the same amount, and breaks even.
3.6.2 How much insurance?

We calculated the premium for full insurance above. But suppose we gave a risk-averse
agent a continuous choice of levels of insurance. Can we say something general about
how much insurance an agent would choose? As it turns out, we can. If insurance is
actuarially fair (which is another way of saying that the insurance company just breaks
even, so that the premium is equal to the expected payment to the agent), we can show
that any risk-averse agent would buy full insurance. If the premium is higher than this,
less than full insurance would be bought. To relate this to the section above, note that
there the agent was given a simple choice between full insurance and no insurance, and
in that case the maximum willingness to pay for insurance is 19, even though the fair
premium is 18. However, if a more continuous choice of insurance levels was provided to
that agent, the agent would optimally buy full insurance only at a premium of 18, and
optimally buy less-than-full insurance at a premium of 19.
If insurance is fair, it does not matter what the degree of risk-aversion is. Every
risk-averse agent would buy full insurance. If, on the other hand, the insurance
premium is greater than the fair level, how much insurance an agent buys depends on
their degree of risk-aversion. No agent with a finite degree of risk-aversion would buy
full insurance anymore, but the extent of insurance purchased increases as the agents
degree of risk-aversion rises.
Let us now show that full insurance is purchased when the premium is fair.
Suppose a risk-averse agent has wealth W , but faces the prospect of a loss of L with
probability p, where 0 < p < 1. The agent can buy a coverage of X by paying the
premium rX.
The wealth if loss occurs is given by WL = W L + X rX, and the wealth when no
loss occurs is given by WN = W rX. The expected utility of the agent is:
E(U ) = pu(W L + X rX) + (1 p)u(W rX) = pu(WL ) + (1 p)u(WN ).
42
3.7. Risk-neutral and risk-loving preferences
Maximising with respect to X, we get the first-order condition:
pu0 (WL )(1 r) (1 p)u0 (WN )r = 0.
Note that the second-order condition for a maximum is satisfied since the agent is
risk-averse implying that u00 < 0 (u is concave). Therefore, we have:
u0 (WL ) (1 p)r
0
= .
u (WN ) (1 r)p
If insurance is fair, that implies the insurance company breaks even, i.e. the expected
payout pX equals the expected receipt rX. Since pX = rX, we have:
p = r.
It follows that:
u0 (WL ) = u0 (WN ).
Since u0 is a decreasing function (because u00 < 0), it is not possible to have this equality
if WL 6= WN . (Note that if u0 was a non-monotonic function and was going up and
down, it would be possible to have WL different from WN but have the same value of u0
at these two different points.)
It follows that WL = WN , i.e. we have:
W L + X rX = W rX
implying that X = L. Therefore, the agent would optimally fully insure (cover the
entire loss) at the fair premium.
Note also what would happen if r > p. Then (1 p)r > (1 r)p. Therefore:
u0 (WL ) (1 p)r
= > 1.
u0 (WN ) (1 r)p
This implies that:
u0 (WL ) > u0 (WN ).
Again, because u0 is a decreasing function (u00 < 0), this implies that WL < WN , which
in turn implies that X < L. Thus if the premium exceeds the fair premium, less than
full insurance would be purchased.
3.7 Risk-neutral and risk-loving preferences

Just as a concave vNM utility function represents risk aversion, the opposite a convex
vNM utility function (so that we have u00 > 0) represents risk-loving behaviour, and
the vNM utility function is a straight line (u00 = 0) for a risk-neutral agent.
A risk-neutral agent does not care about risk and only cares about the expected value
of a gamble. In other words, a risk-neutral agent is indifferent between accepting and
rejecting a fair gamble. For a risk-neutral agent, we can write the vNM utility function
of wealth simply as:
u(W ) = W.
43
Figure 3.2: The vNM utility function for a risk-neutral individual. Note that the function
is linear and u(E(G)) = E(U (G)) so that the agent is indifferent between a fair gamble
and the safe alternative of 15.
Figure 3.2 shows the vNM utility function for a risk-neutral agent. The figure refers to
the gamble introduced in the section on risk aversion (either keep 15 or invest in a fair
gamble yielding 20 with probability 0.5, and 10 with probability 0.5).
A risk-loving agent, on the other hand, prefers a risky bet to a safe alternative when
they have the same expected outcome. In other words, a risk-loving agent would prefer
to accept a fair gamble. As Figure 3.3 below shows, for a risk-loving agent, the CE of a
gamble is higher than the expected value of the gamble (you would have to pay a
risk-loving agent to give up a risky gamble in favour of the safe alternative of getting
the expected value of the gamble).
Figure 3.3 shows the vNM utility function for a risk-loving agent.
3.8 The ArrowPratt measure of risk aversion

We discussed different degrees of risk aversion in the section above. How do we measure
the degree of risk aversion? As you might guess, the degree of risk aversion has to do
with the curvature of the vNM utility function u. The more concave it is, the greater
the degree of risk aversion. The closer it is to a straight line, the lower the degree of risk
aversion. Since the second derivative captures the curvature, a measure of risk aversion
might be u00 . However, this would not be ideal for the following reason. We know that a
positive affine transformation of u, say ub = a + bu, where a and b are positive constants,
does not change attitude towards risk. But such a transformation would change the
second derivative and, therefore, change the risk measure. This problem could be
avoided if u00 is divided by u0 . Furthermore, since the most common risk attitude is risk
aversion, and for this case u00 < 0, putting a negative sign in front of u00 would deliver a
44
3.8. The ArrowPratt measure of risk aversion
Figure 3.3: The vNM utility function for a risk-loving individual. Note that the function
is convex and u(E(G)) < E(U (G)) so that the agent prefers a fair gamble to the safe
wealth of 15.
positive risk measure under risk aversion. These help to interpret the ArrowPratt
measure of risk aversion, which is given by:
u00
= 0.
u
That is, the ArrowPratt measure of risk aversion is 1 times the ratio of the second
derivative and the first derivative of the vNM utility function. This is the most
common measure of risk aversion. There are other measures, which you will encounter
in Masters level courses.
It can be shown that the larger the ArrowPratt measure of risk aversion, the more
small gambles an individual will take. A derivation of this result must await a Masters
level course as well.
As noted above, for a risk-averse individual, u00 < 0, so the minus sign in front makes
the measure a positive number. For a risk-neutral agent, u00 = 0 so that = 0, and for a
risk-loving agent u00 > 0 so that the measure is negative.
Example 3.1 Let us calculate the ArrowPratt measure for different specifications
of the vNM utility function.
(a) Suppose u(W ) = ln W . Then u0 (W ) = 1/W and u00 (W ) = 1/W 2 . It follows

that:
1
= .
W
This agent has risk aversion that is decreasing in wealth.
45
(b) Next, suppose u(W ) = W , where 0 < < 1. Then:
( 1)W 2 1
= 1
= .
W W
Note that as increases, the degree of risk aversion decreases. As goes to 1,
the risk aversion measure goes to 0, which is right since at = 1 the agent is
risk-neutral.
(c) Next, suppose u(W ) = e . Then u0 (W ) = e and u00 (W ) = 2 e .

Therefore, = . In this case the degree of risk aversion does not depend on the
wealth level.
3.9 Reducing risk

Insurance provides a way to reduce risk. Diversification is also another way to reduce
risk. You should read carefully the discussion on this in N&S Section 4.3 (you do not
need to study this section beyond diversification).

to:
calculate the expected value of a gamble

explain the nature of the vNM utility function and calculate the expected utility
from a gamble
explain the different risk attitudes and what they imply for the vNM utility
function
analyse the demand for insurance and show the relationship between insurance and
premium
explain the concept of diversification
calculate the ArrowPratt measure of risk aversion for different specifications of the
vNM utility function.

1. A risk-averse individual is offered a choice between a gamble that pays 1000 with a
probability of 1/4, and 100 with a probability of 3/4, or a payment of 325. Which
would they choose? What if the payment was 320?
46
2. Suppose u(W ) = 1/W . What is the risk attitude of this person? Calculate the
ArrowPratt measure of risk aversion for this preference.
3. Suppose an agent has vNM utility function u(W ). Under what conditionp would
a + bu(W ) also be a valid vNM utility function for this agent? Would u(W ) be
a valid vNM utility function for this agent?
4. Suppose u(W ) = ln W for an agent. The agent faces the following gamble: with
probability 0.5 wealth is 100, and with probability 0.5 a loss occurs so that wealth
becomes 64. The agent can buy any amount of insurance: a coverage of X can be
purchased by paying premium rX.
(a) Work out the insurance coverage X that the agent would optimally purchase
as a function of r.
(b) Plot the optimal X as a function of r.
(c) Calculate the value of r for which full insurance is purchased.
(d) Calculate the value of r for which no insurance is purchased.
47
48
Chapter 4
Game theory
4.1 Introduction
In many economic situations agents must act strategically by taking into account the
behaviour of others. Game theory provides a set of tools that enables you to analyse
such situations in a logically coherent manner. For each concept introduced, you should
try to understand why it makes sense as a tool of analysis and (this is much harder) try
to see what its shortcomings might be. This is not the right place to ask questions such
as what is the policy-relevance of this concept?. The concepts you come across here are
just tools, and that is the spirit in which you should learn them. As you will see, some
of these concepts are used later in this course (as well as in a variety of other economics
courses that you might encounter later) to analyse certain types of economic
interactions.

The chapter aims to familiarise you with a subset of basic game theory tools that are
used extensively in modern microeconomic theory. The chapter aims to cover
simultaneous-move games as well as sequential-move games, followed by an analysis of
the repeated Prisoners Dilemma game.

analyse simultaneous-move games using dominant strategies or by eliminating

dominated strategies either once or in an iterative fashion
calculate Nash equilibria in pure strategies as well as Nash equilibria in mixed

strategies in simultaneous-move games
explain why Nash equilibrium is the central solution concept and explain the
importance of proving existence
specify strategies in extensive-form games
analyse Nash equilibria in extensive-form games
explain the idea of refining Nash equilibria in extensive-form games using backward
induction and subgame perfection
49
4. Game theory
analyse the infinitely-repeated Prisoners Dilemma game with discounting and

analyse collusive equilibria using trigger strategies
explain the multiplicity of equilibria in repeated games and state the folk theorem
for the Prisoners Dilemma game.

N&S Chapter 5. However, the content of this chapter is not sufficient by itself. This
chapter of the subject guide fleshes out the basic tools that you are required to know in
some detail, and you should also read this and follow the exercises carefully. For
repeated games, the interpretation of payoffs presented here is slightly different from the
textbook. While the analyses are formally equivalent, the approach in this subject guide
follows the standard one in the literature. You are likely to find this fits better with
analyses found in other, more advanced, courses or research papers.
4.1.4 Further reading

You might find the following textbook useful for further explanations of the concepts
covered in this subject guide.
Osborne, M.J. An Introduction to Game Theory (Oxford University Press 2009)

International edition [ISBN 9780195322484].

The chapter introduces simultaneous-move games and analyses dominance criteria as
well as Nash equilibrium. Next, the chapter introduces sequential-move games and
analyses Nash equilibrium as well as subgame-perfect Nash equilibrium. Finally, the
chapter introduces repeated games and analyses the infinitely-repeated Prisoners
Dilemma.
4.3 Simultaneous-move or normal-form games

A simultaneous-move game (also known as a normal-form game) requires three
elements. First, a set of players. Second, each player must have a set of strategies. Once
each player chooses a strategy from their strategy set, we have a strategy profile. For
example, suppose there are two players, 1 and 2, and 1 can choose between Up or Down
(so 1s strategy set is {U, D}) and 2 can choose between Left or Right (so 2s strategy
set is {L, R}). Then if 1 chooses Up and 2 chooses Left, we have the strategy profile
{U, L}. There are altogether 4 such strategy profiles: the one just mentioned, plus
{D, L}, {U, R} and {D, R}.
Once you understand what a strategy profile is, we can define the third element of a
game: a payoff function for each player. A payoff function for any player is defined over
the set of strategy profiles. For each strategy profile, each player gets a payoff.
50
4.3. Simultaneous-move or normal-form games
Suppose Si denotes the set of strategies of player i. In the example above, S1 = {U, D}
and S2 = {L, R}. Let si denote a strategy of i (that is, si is in the set Si ). If there are n
players, a strategy profile is (s1 , s2 , . . . , sn ). For any such strategy profile, there is a
payoff for each player. Let ui (s1 , s2 , . . . , sn ) denote the payoff of player i.
Continuing the example above, suppose u1 ({U, L}) = u1 ({D, R}) = 1 and
u1 ({D, L}) = u1 ({U, R}) = 2. Furthermore, suppose u2 ({U, L}) = u2 ({D, R}) = 2,
u2 ({D, L}) = 3 and u2 ({U, R}) = 1.
We write this in a convenient matrix form (known as the normal form) as follows. The
first number in each cell is the payoff of player 1 and the second number is the payoff of
player 2. Note that player 1 chooses rows and player 2 chooses columns.
Player 2
L R
Player 1 U 1, 2 2, 1
D 2, 3 1, 2
It is sometimes useful to write the strategy profile (s1 , s2 , . . . , sn ) as (si , si ), where si

is the profile of strategies of all players other than player i. So:
si = (s1 , s2 , . . . , si1 , si+i . . . . , sn )
(the ith element is missing). With this notation, we can write the payoff of player i as
ui (si , si ).
4.3.1 Dominant and dominated strategies

Let us now try to understand how rational players should play a game. In some cases,
there is an obvious solution. Suppose a player has a strategy that gives a higher payoff
compared to other strategies irrespective of the strategy choices of others. Such a
strategy is called a dominant strategy. If a player has a dominant strategy, his
problem is simple: he should clearly play that strategy. If each player has a dominant
strategy, the equilibrium of the game is obvious: each player plays his own dominant
strategy.
Consider the following game. Each player has two strategies C (cooperate) and D
(defect, which means not to cooperate). There are 4 possible strategy profiles and each
profile generates a payoff for each player. The first number in each cell is the payoff of
player 1 and the second number is the payoff of player 2. Again, note that player 1
chooses the row that is being played and player 2 chooses the column that is being
played.
Player 2
C D
Player 1 C 2, 2 0, 3
D 3, 0 1, 1
Here each player has a dominant strategy, D. This is the well-known Prisoners
Dilemma game. Rational players, playing in rational self-interest, get locked into a
51
4. Game theory
dominant-strategy equilibrium that gives a lower payoff compared to the situation

where both players cooperate. However, cooperating cannot be part of any equilibrium,
since D is the dominant strategy. Later on we will see that if the game is
infinitely-repeated, then under certain conditions cooperation can emerge as an
equilibrium. But in a one-shot game (i.e. a game that is played once) the only possible
equilibrium is that each player plays their dominant strategy. In the game above, the
dominant strategy equilibrium is (D, D).
In terms of the notation introduced before, we can define a dominant strategy as follows.
Dominant strategy
Strategy si of player i is a dominant strategy if:
ui (si , si ) > ui (si , si ) for all si different from si and for all si .
That is, si performs better than any other strategy of player i no matter what others
are playing.
4.3.2 Dominated strategies and iterated elimination
Even if a player does not have a dominant strategy, he might have one or more
dominated strategies. A dominated strategy for i is a strategy of i (say si ) that yields a
lower payoff compared to another strategy (say s0i ) irrespective of what others are
playing. In other words, the payoff of i from playing si is always (i.e. under all possible
choices of other players) lower than the payoff from playing s0i . Since si is a dominated
strategy, i would never play this strategy. Thus we can eliminate dominated strategies.
Indeed, we can eliminate such strategies not just once, but in an iterative fashion.
If in some game, all strategies except one for each player can be eliminated by
iteratively eliminating dominated strategies, the game is said to be dominance solvable.
Consider the game in Section 4.3. Note that no strategy is dominant for player 1, but
for player 2 L dominates R. So we can eliminate the possibility of player 2 playing R.
Once we do this, in the remaining game, for player 1 D dominates U (2 is greater than
1). So we can eliminate U . We are then left with (D, L), which is the equilibrium by
iteratively eliminating dominated strategies. The game is dominance solvable.
Here is another example of a dominance solvable game. We find the equilibrium of this
game by iteratively eliminating dominated strategies.
Player 2
Left Middle Right
Top 4, 3 2, 7 0, 4
Player 1 Middle 5, 5 5, 1 4, 2
Bottom 3, 5 1, 5 1, 6
We can eliminate dominated strategies iteratively as follows.
52
1. For player 1, Bottom is dominated by Top. Eliminate Bottom.
2. In the remaining game, for player 2, Right is dominated by Middle. Eliminate

Right.
3. In the remaining game, for player 1, Top is dominated by Middle. Eliminate Top.
4. In the remaining game, for player 2, Middle is dominated by Left. Eliminate

Middle.
This gives us (Middle, Left) as the unique equilibrium.
4.3.3 Nash equilibrium

However, for many games the above criteria of dominance do not allow us to find an
equilibrium. Players might not have dominant strategies; moreover none of the
strategies of any player might be dominated. The following game provides an example.
Player 2
A2 B2 C2
A1 3, 1 1, 3 4, 2
Player 1 B1 1, 0 3, 1 3, 0
C1 2, 3 2, 0 3, 2
As noted above, the problem with dominance criteria is that they apply only to some
games. For games that do not have dominant or dominated strategies, the idea of
deriving an equilibrium using dominance arguments does not work. If we cannot derive
an equilibrium by using dominant strategies or by (iteratively) eliminating dominated
strategies, how do we proceed?
If we want to derive an equilibrium that does not rely on specific features such as
dominance, we need a concept of equilibrium that applies generally to all games. As we
show below, a Nash equilibrium (named after the mathematician John Nash) is
indeed such a solution concept.
Pure and mixed strategies
Before proceeding further, we need to clarify something about the nature of strategies.
In the discussion above, we identified strategies with single actions. For example, in the
Prisoners Dilemma game, we said each player has the strategies C and D. However,
this is not a full description of strategies. A player could also do the following: play C
with probability p and play D with probability (1 p), where p is some number
between 0 and 1. Such a strategy is called a mixed strategy; while a strategy that just
chooses one action (C or D) is called a pure strategy.
We start by analysing Nash equilibrium under pure strategies. Later we introduce
mixed strategies. We then note that one can prove an existence result: all games have at
least one Nash equilibrium (in either pure or mixed strategies). This is why Nash
equilibrium is the central solution concept in game theory.
53
4. Game theory
A strategy profile (s1 , s2 , . . . , sn ) is a Nash equilibrium (in pure strategies) if it is a

mutual best response. In other words, for every player i, the strategy si is a best
response to si (as explained above, this is the strategy profile of players other than i).
In yet other words, if (s1 , s2 , . . . , sn ) is a Nash equilibrium, it must satisfy the property
that given the strategy profile si of other players, player i cannot improve his payoff by
replacing si with any other strategy.
A more formal definition is as follows.
Nash equilibrium in pure strategies
A strategy profile (si , si ) is a Nash equilibrium if for each player i:
ui (si , si ) ui (si , si ) for all strategies si in the set Si .
To find out the Nash equilibrium of the game above, we must look for the mutual best
responses. Let us check the best response of each player. Player 1s best response is as
follows:
Player 2s strategy Player 1s best response

A2 A1
B2 B1
C2 A1
Player 2s best response:
Player 1s strategy Player 2s best response

A1 B2
B1 B2
C1 A2
Note from these that the only mutual best response is (B1 , B2 ). This is the only Nash
equilibrium in this game.
You could also check as follows:
If player 1 plays A1 , player 2s best response is B2 . However, if player 2 plays B2 ,

player 1 will not play A1 (B1 is a better response than A1 ). Therefore, there is no
Nash equilibrium involving A1 .
If player 1 plays B1 , player 2s best response is B2 . If player 2 plays B2 , player 1s

best response is B1 . Therefore, (B1 , B2 ) is a Nash equilibrium.
If player 1 plays C1 , player 2s best response is A2 . However, if player 2 plays A2 ,

player 1 would not play C1 (A1 is a better response). Therefore, there is no Nash
equilibrium involving C1 .
From the three steps above, we can conclude that (B1 , B2 ) is the only Nash equilibrium.
54
Let us also do a slightly different exercise. Suppose you want to check if a particular
strategy profile is a Nash equilibrium. Suppose you want to check if (A1 , C2 ) is a Nash
equilibrium. You should check as follows.
If player 2 plays C2 , player 1 cannot do any better by changing strategy from A1 (4 is
better than 3 from B1 or 3 from C1 ). However, player 2 would not want to stay at
(A1 , C2 ) since player 2 can do better by switching to B2 (3 from B2 is better than 2
from C2 ). We can therefore conclude that (A1 , C2 ) is not a Nash equilibrium.
You can similarly check that all boxes other than (B1 , B2 ) have the property that some
player has an incentive to switch to another box. However, if the players are playing
(B1 , B2 ) no player has an incentive to switch away. Neither player can do better by
switching given what the other player is playing. Since player 2 is playing B2 , player 1
gets 3 from B1 which is better than 1 from A1 or 2 from C1 . Since player 1 is playing
B1 , player 2 gets 1 from B2 which is better than 0 from A2 or C2 . Therefore (B1 , B2 ) is
a Nash equilibrium.
Nash equilibrium is not necessarily unique. Consider the following game.
Player 2
A B
A 2, 1 0, 0
Player 1 B 0, 0 1, 2
Note there are multiple pure strategy Nash equilibria. Both (A, A) and (B, B) are Nash
equilibria.
Dominance criteria and Nash equilibrium
Note that while a dominant strategy equilibrium is also a Nash equilibrium, Nash
equilibrium does not require dominance. However, the greater scope of Nash equilibrium
comes at a cost: it places greater rationality requirements on players. To play (B1 , B2 ),
player 1 must correctly anticipate that player 2 is going to play B2 . Such a requirement
is even more problematic when there are multiple Nash equilibria. On the other hand, if
players have dominant strategies, they do not need to think at all about what others are
doing. A player would simply play the dominant strategy since it is a best response no
matter what others do. This is why a dominant strategy equilibrium (or one achieved
through iterative elimination of dominated strategies) is more convincing than a Nash
equilibrium. However, as noted before, many games do not have dominant or dominated
strategies, and are therefore not dominance solvable. We need a solution concept that
applies generally to all games, and Nash equilibrium is such a concept.
4.3.4 Mixed strategies

Consider the following game.
Player 2
A2 B2
A1 3, 1 2, 3
Player 1 B1 2, 1 3, 0
55
4. Game theory
Notice that this game has no pure strategy Nash equilibrium. However, as you will see
below, the game does have a Nash equilibrium in mixed strategies.
Let us first define a mixed strategy.
Mixed strategy
A mixed strategy si is a probability distribution over the set of (pure) strategies.
In the game above, A1 and B1 are the pure strategies of player 1. A mixed strategy of
player 1 could be A1 with probability 1/3, and B1 with probability 2/3. Notice that a
pure strategy is only a special case of a mixed strategy.
A Nash equilibrium can then be defined in the usual way: a profile of mixed strategies
that constitute a mutual best response.
A mutual best response in mixed strategies has an essential property that makes it easy
to find mixed strategy Nash equilibria. Let us consider games with two players to
understand this property.
Suppose we have an equilibrium in which both players play strictly mixed strategies:
player 1 plays A1 with probability p and B1 with probability (1 p) where 0 < p < 1,
and player 2 plays A2 with probability q and B2 with probability (1 q) where
0 < q < 1.
In this case, whenever player 1 plays A1 , she gets an expected payoff of:
1 (A1 ) = 3q + 2(1 q).
Whenever player 1 plays B1 , she gets an expected payoff of:
1 (B1 ) = 2q + 3(1 q).
What must be true of these expected payoffs that player 1 gets from playing A1 and
B1 ? Suppose 1 (A1 ) > 1 (B1 ). Then clearly player 1 should simply play A1 , rather than
any strictly mixed strategy, to maximise her payoff. In other words, player 1s best
response in this case would be to choose p = 1, rather than a strictly mixed strategy
p < 1. But then we do not have a mixed strategy Nash equilibrium.
Similarly, if 1 (A1 ) < 1 (B1 ), player 1s best response would be to choose p = 0 (i.e. just
to play B1 ), and again we cannot have a mixed strategy Nash equilibrium.
So if player 1 is going to play a mixed strategy in equilibrium it must be that she is
indifferent between the two strategies. How does such indifference come about? This is
down to player 2s strategy choice. Player 2s choice of q must be such that player 1 is
indifferent between playing A1 or B1 . In other words, in equilibrium q must be such
that 1 (A1 ) = 1 (B1 ), i.e. we have:
3q + 2(1 q) = 2q + 3(1 q)
which implies q = 1/2.

But if player 2 is going to choose q = 1/2 it must be that he is indifferent between A2
and B2 (otherwise player 2 would not want to mix, but would play a pure strategy).
How can such indifference come about? Well, player 1 must choose p in such a way so as
56
to make player 2 indifferent between A2 and B2 . In other words, player 1s choice of p is

such that 2 (A2 ) = 2 (B2 ), i.e. we have:
1 = 3p
which implies p = 1/3.

Therefore, the mixed strategy Nash equilibrium is as follows. Player 1 plays A1 with
probability 1/3 and B1 with probability 2/3, while player 2 plays A2 with probability
1/2 and B2 with probability 1/2.
We can also show this in a diagram. Let us first write down the best response function
of each player.
Player 1s best response function is given by:
q > 1/2 = p = 1 is the best response.

q = 1/2 = any p in [0, 1] is a best response.
q < 1/2 = p = 0 is the best response.
Player 2s best response function is given by:
p > 1/3 = q = 0 is the best response.

p = 1/3 = any q in [0, 1] is a best response.
p < 1/3 = q = 1 is the best response.
Figure 4.1 below shows these best response functions and shows the equilibrium point
(where the two best response functions intersect).
Figure 4.1: The best-response functions. They cross only at E, which is the mixed strategy
Nash equilibrium. There are no pure strategy Nash equilibria in this case.
57
4. Game theory
To recap, the essential property of a mixed strategy Nash equilibrium in a two-player

game is that each players chosen probability distribution must make the other player
indifferent between the strategies he is randomising over. In a k-player game, the joint
distribution implied by the choices of each player in every combination of (k 1)
players must be such that the kth player receives the same expected payoff from each of
the strategies he plays with positive probability.
The game above has no Nash equilibrium in pure strategies, but has a mixed strategy
Nash equilibrium. However, in other games that do have pure strategy Nash equilibria,
there might be yet more equilibria in mixed strategies. For instance, we could find a
mixed strategy Nash equilibrium in the Battle of the sexes.
Activity 4.1 Find the mixed strategy Nash equilibrium in the following game. Also
show all Nash equilibria of the game in a diagram by drawing the best response
functions.
Player 2
A B
A 2, 1 0, 0
Player 1 B 0, 0 1, 2
4.3.5 Existence of Nash equilibrium

Once we include mixed strategies in the set of strategies, we have the following
existence theorem, proved by John Nash in 1951.
Existence theorem
Every game with a finite number of players and finite strategy sets has at least one
Nash equilibrium.
Nash proved the existence theorem for his equilibrium concept using a mathematical
result called a fixed-point theorem. Take any strategy profile and compute the best
response to it for every player. So the best response is another strategy profile. Suppose
we do this for every strategy profile. So long as certain conditions are satisfied, a
fixed-point theorem says that there is going to be at least one strategy profile which is a
best response to itself. This is, of course, a Nash equilibrium. Thus upon setting up the
strategy sets and the best response functions properly, a fixed-point theorem can be
used to prove existence. You will see a formal proof along these lines if you study game
theory at more advanced levels. Here, let us point out the importance of this result.
The importance of proving existence
The existence theorem is indeed very important. It tells us that no matter what game
we look at, we will always be able to derive at least one Nash equilibrium. If existence
did not hold for some equilibrium concept (for example, games do not necessarily have a
dominant strategy equilibrium), we could derive wonderful properties of that concept,
but we could not be sure such derivations would be of any use. The particular game
58
4.4. Sequential-move or extensive-form games
that we confront might not have an equilibrium at all. But being able to prove existence
for Nash equilibrium removes such problems. Indeed, as noted before as well, this is
precisely what makes Nash equilibrium the main solution concept for
simultaneous-move games.
4.3.6 Games with continuous strategy sets

We have so far analysed games with discrete strategy sets. However, the above analysis
can easily extend to certain classes of games with continuous strategy sets. We will
analyse a few such games (the Cournot game, the Bertrand game, and the Bertrand
game with product differentiation) in detail later in discussing oligopoly (in Chapter 9
of the subject guide). Also see N&S Section 5.7 for an example of a commons problem
game with continuous strategy sets. We will also refer to this game when discussing
externalities (in Chapter 12 of the subject guide).
4.4 Sequential-move or extensive-form games

Let us now consider sequential move games. In this case, we need to draw a game tree
to depict the sequence of actions. Games depicted in such a way are also known as
extensive-form games. In this subject guide we will consider the phrases
sequential-move game and extensive-form game as interchangeable.
To start with, we assume that each player can observe the moves of players who act
before them. First, we need to understand the difference between actions and strategies
in such games. Once we clarify this, we show how to derive Nash equilibria. Finally, we
propose a refinement of Nash equilibrium: subgame perfect Nash equilibrium. In
games where all moves of previous players can be observed, subgame perfect Nash
equilibria can be derived by backward induction. We then introduce some simple cases
where information is imperfect, and show how the notion of strategies differs, and how
to derive subgame perfect Nash equilibria.
Consider the following extensive-form game. Each player has two actions: player 1s
actions are a1 and a2 and player 2s actions are b1 and b2 . Player 1 moves before player
2. Player 2 can observe player 1s action and, therefore, can vary his action depending
on the action of player 1. For each profile of actions by player 1 and player 2 there are
payoffs at the end. As usual, the first number is the payoff of the first mover (in this
case player 1) and the second number is the payoff of the second mover (here player 2).
We can define the game as a graph: it has decision nodes and branches from decision
nodes to successor nodes. However, such formal definitions are useful only at a later
stage. If you simply look at Figure 4.2 below, the depiction of the sequence of players,
their action choices at each stage and their final payoffs should be clear to you.
4.4.1 Actions and strategies

The notion of a strategy is fairly straightforward in a normal form game. However, for
an extensive-form game, it is a little bit more complicated. A strategy in an extensive
form game is a complete plan of actions. In other words, a strategy for player i must
59
4. Game theory
Figure 4.2: An extensive-form game.
specify an action at every node at which i can possibly move.

Consider the game above. As already noted, each player has 2 actions. For player 1, the
set of actions and the set of strategies is the same. Player 1 can simply decide between
a1 and a2 . Therefore, the strategy set of player 1 is simply {a1 , a2 }.
Player 2, on the other hand, must plan for two different contingencies. He must decide
what to do if player 1 plays a1 , and what to do if player 1 plays a2 . Note that such
decisions must be made before the game is actually played. Essentially, game theory
tries to capture the process of decision-making of individuals. Faced with a game such
as the one above, player 2 must consider both contingencies. This is what we capture by
the notion of a strategy. It tells us what player 2 would do in each of the two possible
cases.
Now player 2 can choose 2 possible actions at the left node (after player 1 plays a1 ) and
2 possible actions at the right node (after a2 ). So there are 2 2 = 4 possible strategies
for player 2. These are:
1. If player 1 plays a1 , play b1 and if player 1 plays a2 , play b1 .
For the sake of brevity of notation, we write these as follows. Just as we read words
from left to right, we read strategies from left to right. So we write the strategy if
player 1 plays a1 , play b2 and if player 1 plays a2 , play b1 as b2 b1 . Reading from left to
right, this implies that the plan is to play b2 at the left node and play b1 at the right
node. This is precisely what the longer specification says.
So the strategy set of player 2 is {(b1 b1 ), (b1 b2 ), (b2 b1 ), (b2 b2 )}.
Suppose instead of 2 actions, player 2 could choose between b1 , b2 and b3 at each node.
In that case, player 2 would have 3 3 = 9 strategies.
60
4.4. Sequential-move or extensive-form games
Suppose player 1 had 3 strategies a1 , a2 and a3 , and after each of these, player 2 could
choose between b1 and b2 . Then player 2 would have 2 2 2 = 8 strategies.
4.4.2 Finding Nash equilibria using the normal form

The matrix-form we used to write simultaneous-move games in Section 4.3 is known as
the normal form. To find Nash equilibria in an extensive-form game, the most
convenient method is to transform it into its normal form. This is as follows. Note that
player 1 has two strategies and player 2 has four strategies. Therefore, we have a 2-by-4
matrix of payoffs as follows:
Player 2
b1 b1 b1 b2 b2 b1 b2 b2
Player 1 a1 3, 1 3, 1 1, 0 1, 0
a2 4, 1 0, 1 4, 1 0, 1
Note that if we pair, say, a1 with b2 b1 , only the first component of player 2s strategy is
relevant for the payoff. In other words, since player 1 plays a1 , the payoff is generated by
player 2s response to a1 , which in this case is b2 . Similarly, if we pair a2 with b2 b1 , the
second component of player 2s strategy is relevant for the payoff. Since player 1 plays
a2 , we need player 2s response to a2 , which in this case is b1 , to determine the payoff.
Once we write down the normal form, it is easy to find Nash equilibria. Here let us only
consider pure strategy Nash equilibria. There are three pure strategy Nash equilibria in
this game. Finding these is left as an exercise.
Example 4.1 Find the pure strategy Nash equilibria of the extensive-form game
above.
It is easiest to check each box. If we start at the top left-hand box, player 1 would
switch to a2 . So this is not a Nash equilibrium. From (a2 , b1 b1 ) no player can do
better by deviating. Therefore, (a2 , b1 b1 ) is a Nash equilibrium.
Next try (a1 , b1 b2 ). This is indeed a Nash equilibrium. Note that you must write
down the full strategy of player 2. It is not enough to write (a1 , b1 ). Unless we know
what player 2 would have played in the node that was not reached (in this case the
node after a2 was not reached), we cannot determine whether a strategy is part of a
Nash equilibrium. So while (a1 , b1 b2 ) is indeed a Nash equilibrium, (a1 , b1 b1 ) is not.
Finally, (a2 , b2 b1 ) is also a Nash equilibrium. These are the three pure strategy Nash
equilibria of the game.
4.4.3 Imperfect information: information sets

So far we have assumed perfect information: each player is perfectly informed of all
previous moves. Let us now see how to represent imperfect information: players may not
be perfectly informed about some of the (or all of the) previous moves.
Games of imperfect information give rise to certain types of problems that require more
sophisticated refinements of Nash equilibria than we study here. In particular, if a
61
4. Game theory
player does not observe some past moves, what he believes took place becomes
important. We do not study these problems here the analysis below simply shows you
how to represent situations of imperfect information in some simple cases.
Consider the extensive-form game introduced at the start of this section. Suppose at the
time of making a decision, player 2 does not know what strategy player 1 has chosen.
Since player 2 does not know what player 1 has chosen, at the time of taking an action
player 2 does not know whether he is at the left node or at the right node. To capture
this situation of imperfect information in our game-tree, we say that the two decision
nodes of player 2 are in an information set. We represent this information set as in
Figure 4.3: by connecting the two nodes by a dotted line (another standard way to do
this is to draw an elliptical shape around the two nodes you will see this in N&S).
Figure 4.3: An extensive-form game with an information set.
Note that player 2 knows he is at the information set (after player 1 moves), but does
not know where he is in the information set (i.e. he does not know what player 1 has
chosen). Since player 2 cannot distinguish between the two nodes inside his information
set, he cannot take different actions at the two nodes. Therefore, the strategy set of 2 is
simply {b1 , b2 }. In other words, now, for both players the strategy set coincides with the
action set. This is not surprising: since player 2 takes an action without knowing what
player 1 has done, the game is the same as a simultaneous-move game. Indeed, the
normal form of the game above coincides with a game in which the two players choose
strategies simultaneously. This is shown below. Now the Nash equilibrium is simply
(a2 , b1 ).
Player 2
b1 b2
a1 3, 1 1, 0
Player 1 a2 4, 1 0, 1
62
4.5. Incredible threats in Nash equilibria and subgame perfection
4.5 Incredible threats in Nash equilibria and subgame

perfection
Let us now consider whether Nash equilibrium is a satisfactory solution concept for
extensive-form games. As you will see, when players move sequentially, Nash
equilibrium allows for some strategies by later movers that seem like threats which are
incredible. Parents often try to control unruly children by saying things like sit quietly,
or we will never let you . . . (insert favourite activity), even though they have no
intention of carrying out the threat. Children sometimes believe their parents and
respond to the threat, but they are often clever enough to see through the ruse and
ignore incredible threats. As we will see, Nash equilibria often depend precisely on such
incredible threats by later movers. In Nash equilibrium, a player is just supposed to
take a best response to the other players strategy choices so the way Nash
equilibrium is constructed does not allow the player to ignore certain strategy choices of
others as incredible. Once we look at some examples of the problem, we will try to see
whether we can refine the set of Nash equilibria to eliminate the possibility of such
threats (i.e. come up with extra conditions that an equilibrium must satisfy so that
equilibria which depend on incredible threats will not satisfy these extra conditions).
Consider the game shown in Figure 4.4. Firm E, where E stands for entrant, is deciding
whether to enter a market. The market has an incumbent firm (Firm I). If the entrant
enters, the incumbent firm must decide whether to fight (start a price war, say) or
accommodate the entrant. The sequence of actions and payoffs is as follows.
Figure 4.4: A game with an entrant and an incumbent.
The normal form is as follows.
Firm I
A F
In 2, 1 1, 2
Firm E Out 0, 2 0, 2
Note that there are two pure strategy Nash equilibria: (In, A) and (Out, F). The latter
equilibrium involves an incredible threat. Clearly, if the entrant does decide to enter the
63
4. Game theory
market, the incumbent has no incentive to choose F. Hence the threat of F is incredible.
Yet, Nash equilibrium cannot preclude this possibility. Out is the best response to F,
and once Firm E decides to stay out, anything (and in particular F) is a best response
for Firm I.
The game in Figure 4.5 presents another example.
Activity 4.2 Consider the extensive-form game in Figure 4.5.
(a) Write down the strategies available to each player.
(b) Write down the normal form of the game.
(c) Identify the pure strategy Nash equilibria.
When you write the normal form and work out the Nash equilibria, you should see
that (R, rr) is a Nash equilibrium. Look at the game above and see that this
involves an incredible threat. Player 2s strategy involves playing r after L. Player
1s strategy is taking a best response given this, and so player 1 is playing R. Given
that player 1 is playing R, the threat is indeed a best response for player 2 (indeed,
given that player 1 plays R, anything that player 2 can choose after L is trivially a
best response since it does not change the payoff, but, of course, not every choice
would lead to an equilibrium in which player 1s best response is R).
4.5.1 Subgame perfection: refinement of Nash equilibrium

Let us now describe a solution concept that imposes extra conditions (i.e. further to the
requirement that strategies be mutual best responses) for equilibrium and leads to a
refinement of the set of Nash equilibria. Several such refinements have been proposed by
game theorists. Here, we will look at only one such refinement, namely subgame
perfection. To understand the refinement, you first need to understand the idea of a
subgame.
64
Subgame
A subgame is a part of a game that starts from a node which is a singleton (i.e. a
subgame does not start at an information set), and includes all successors of that node.
If one node in an information set belongs to a subgame, so do all other nodes in that
information set. In other words, you cannot cut an information set so that only part of
it belongs to a subgame. That would clearly alter the information structure of the
game, which is not allowed.
We can now define a subgame perfect equilibrium.
Subgame perfect Nash equilibrium
A strategy combination is a subgame perfect Nash equilibrium (SPNE) if:
it is a Nash equilibrium of the whole game

it induces a Nash equilibrium in every subgame.
It should be clear from the definition that the set of subgame perfect equilibria is a
refinement of the set of Nash equilibria.
4.5.2 Perfect information: backward induction

How do we find subgame perfect equilibria? In perfect information games (recall that,
in a game of perfect information, each player knows all past moves of other players),
this is easy. Subgame perfect Nash equilibria can be derived simply by solving
backwards, i.e. by using backward induction.
Solving backwards in the entry game, we see that Firm I would choose A if Firm E
chose In. Knowing this, Firm E would compare 0 (from Out) with 2 (from In and A),
and choose In. Therefore, the SPNE is (In, A).
Let us see that this equilibrium derived using backward induction fits with the
definition of SPNE given above. The game has a subgame starting at the node after
Firm E plays In (also, the whole game is always trivially a subgame). In the subgame
after E plays In, there is only one player (Firm I), and the Nash equilibrium in this
subgame is simply the optimal action of Firm I, which in this case is to choose A.
Therefore the Nash equilibrium in the subgame is A. It follows that any Nash
equilibrium of the whole game that involves playing A in the subgame is a subgame
perfect Nash equilibrium. Here, the only Nash equilibrium of the whole game that
satisfies this property is (In, A). Therefore, this is the only SPNE.
Next, consider the game in which player 1 chooses between L, R and player 2 moves
second and chooses between `, r. In this game there are two strict subgames, one
starting after each action of player 1. In the left subgame, player 2s optimal choice is `,
and in the right subgame player 2s optimal choice is r. Given this, player 1 would
compare 3 from R and 0 from L, and choose R. The choices obtained by backward
induction are shown in Figure 4.6.
It follows that the SPNE is (R, `r). Note that it is not sufficient to write (R, `) the
equilibrium specification is meaningless unless you specify the full strategy for player 2.
65
4. Game theory
Figure 4.6: Finding subgame perfect Nash equilibria.
What player 2 plays at the unreached node is crucial. If player 2 played ` after L, R
would not be the optimal choice. Therefore, you must specify player 2s full strategy
and identify (R, `r) as the SPNE.
In these games, the SPNE is unique, but it need not be. The next activity presents an
example.
Activity 4.3 Derive the pure strategy subgame perfect Nash equilibria of the game
in Figure 4.2.
4.5.3 Subgame perfection under imperfect information

Backward induction need not work under imperfect information: you cannot fold
backwards when you come up against an information set. Indeed, this is why the
concept of a subgame perfect Nash equilibrium is more general compared to backward
induction. If we always had perfect information, we could simply define backward
induction equilibria. However, we present below an example to show you that subgame
perfection is more general than backward induction, and works in many games in which
backward induction does not give us any result.
Before we present the example referred to above, consider the imperfect information
game introduced in Section 4.4.3. Note that this game does not have any strict
subgames (recall that you cannot start a subgame from an information set or cut an
information set), so the only subgame is the whole game. Therefore, any Nash
equilibrium of the whole game is trivially subgame perfect. As discussed above, the pure
strategy Nash equilibrium in this game is (a2 , b1 ). This is also the subgame perfect Nash
equilibrium.
Next, consider the game in Figure 4.7.
Initially player 1 decides whether to come in (and play some game with player 2) or
stay out (in which case player 2 plays no role). If the choice is to come in, player 1
66
Figure 4.7: A game tree.
decides between A and B, and player 2 decides between C and D. When player 2 makes
his decision, he knows that player 1 has decided to come in (if not then player 2 would
not have been asked to play), but without knowing what player 1 has chosen between A
and B. In other words, the situation is just as if once player 1 comes in, player 1 and
player 2 play a simultaneous-move game (as the game structure shows, they do not
move simultaneously player 1 moves before player 2, but since player 2 has no
knowledge of player 1s move, it is similar to the decision problem faced in a
simultaneous move game).
1. Pure strategy Nash equilibria. Let us first identify the pure strategy Nash
equilibria. Note that player 1 has 4 strategies: Out A, Out B, In A and In B, while
player 2 has 2 strategies: C and D. You might think strategies like Out A do not make
sense, but in game theory we try to model the thought process of players, and even if
player 1 stays out, she would do so only after thinking about what she would have done
had she entered. Strategies reflect such thinking (it is as if player 1 is saying I have
decided to finally stay out, but had I come in I would have chosen A).
Let us now write down the normal form of the game.
Player 2
C D
Out A 1, 3 1, 3
Player 1 Out B 1, 3 1, 3
In A 2, 2 2, 0
In B 0, 2 5, 5
You should be able to see from this that the pure strategy Nash equilibria are
(Out A, C), (Out B, C) and (In A, D).
67
4. Game theory
2. Pure strategy subgame perfect Nash equilibria. Note that backward induction
does not work here: we cannot fold back given the information set of player 2. However,
subgame perfection still works. Let us see how applying subgame perfection can refine
the set of Nash equilibria.
Note that apart from the whole game, there is just one strict subgame, which starts at
the node after player 1 chooses In. Below we write down the normal form of the
subgame.
Player 2
C D
Player 1 A 2, 2 2, 0
B 0, 2 5, 5
As you can see, the subgame has two pure strategy Nash equilibria: (B, C) and (A, D).
If (B, C) is played in the subgame, player 1 compares 1 (from Out) with 0 (from In
followed by (B, C)) and decides to stay out. Therefore, a SPNE of the whole game is
(Out B, C).
If, on the other hand, (A, D) is played in the subgame, player 1 compares 1 with 2 and
decides to come in. Therefore, another SPNE of the whole game is (In A, D).
It follows that the pure strategy SPNE of the whole game are (Out B, C) and (In A, D).
Another way to derive these is as follows. Since the set of SPNE is a subset of the set of
Nash equilibria, and since (B, C) and (A, D) are the Nash equilibria in the subgame, it
must be that any Nash equilibria of the whole game that involves playing either (B, C)
and (A, D) in the subgame are subgame perfect. Considering the set of Nash equilibria
derived above, we can immediately infer that (Out A, C) is Nash but not subgame
perfect, while the other two are subgame perfect.
4.6 Repeated Prisoners Dilemma

Consider the following Prisoners Dilemma game:
Player 2
C D
Player 1 C 2, 2 0, 3
D 3, 0 1, 1
In a one-shot game, rational players simply play their dominant strategies. So (D, D) is
the only possible equilibrium. Suppose the game is repeated. Can we say something
about the behaviour of players in such supergames that differs from the behaviour in
the one-shot game?
First, consider the case of a finite number of repetitions. Say the game is played twice.
Would anything change? The answer is no. In the second round, players simply face a
one-shot game and they would definitely play their dominant strategies. Given that
(D, D) will be played in the next period, playing anything other than D today makes no
sense. Therefore, in each period players would play (D, D). But this logic extends to any
68
4.6. Repeated Prisoners Dilemma
finite number of repetitions. If the game is played a 100 times, in the last period (D, D)
will be played. This implies that (D, D) will be played in the 99th period, and so on.
While the logic is inescapable, actual behaviour in laboratory settings differs from this.
Faced with a large finite number of repetitions, players do cooperate for a while at least.
Therefore, it is our modelling that is at fault. To escape from the logic of backward
induction, we can assume that when a game is repeated many times, players play them
as if the games are infinitely repeated. In that case, we must apply forward-looking logic
as there is no last period from which to fold backwards.
(A quick note: you should be aware that there are other games with multiple Nash
equilibria where some cooperation can be sustained even under a finite number of
repetitions. You will encounter these in more advanced courses. Here we only consider
the repeated Prisoners Dilemma.)
Let us now analyse an infinitely repeated Prisoners Dilemma game.
Payoffs: discounted present value
First, we need to have an appropriate notion of payoffs in the infinitely repeated game.
Each player plays an action (in this case either C or D) in each period. So in each
period, the players end up playing one of the four possible action profiles (C, C), (C, D),
(D, C) or (D, D). Let at denote the action profile played in period t. Then in period t,
player i receives the payoff ui (at ). The payoff of player i in the repeated game is simply
the discounted present value of the stream of payoffs.
Let denote the common discount factor across players, where 0 < < 1. If todays
date is 0, and a player receives x in period t, the present value of that payoff is t x. The
discount factor can reflect players time preference. This can also arise from a simple
rate of interest calculation, in which case can be interpreted as 1/(1 + r), where r is the
rate of interest. Note that higher values of indicate that players are more patient (i.e.
value future payoffs more). If is very low, the situation is almost like a one-shot game,
since players only value todays payoff, and place very little value on any future payoff.
Given such discounting, the payoff of player i in the repeated game is:
ui (a0 ) + ui (a1 ) + 2 ui (a2 ) + .
More concisely, the payoff is:

X
t ui (at ).
t=0
If the payoff is the same every period (x, say), this becomes:
x
x(1 + + 2 + ) = .
1
4.6.1 Cooperation through trigger strategies

Next, we need to consider strategies by players. The history at t is the action profile
played in every period from period 0 to t 1. A strategy of a player consists of an
initial action, and after that, an action after every history. Consider the following
trigger strategy.
69
4. Game theory
Trigger strategy
Start by playing C (that is, cooperate at the very first period, when there is no
history yet).
In period t > 1:
if (C, C) was played last period, play C
if anything else was played last period, play D.
Suppose each player follows this strategy. Note that cooperation (playing (C, C)) would
work only until someone deviates to D. After the very first deviation, each player
switches to D. Since anything other than (C, C) implies playing (D, D) next period,
once a switch to (D, D) has been made, there is no way back: the players must play
(D, D) forever afterwards. This is why this is a trigger strategy.
Another way of stating the trigger strategy is to write in terms of strategy profiles.
Start by playing (C, C).
In period t:
if (C, C) is played in t 1, play (C, C)
otherwise play (D, D).
Let us see if this will sustain cooperation. Suppose a player deviates in period t. We
only need to consider what happens from t onwards. The payoff starting at period t is
given by:

3 + + 2 + = 3 + .
1
If the player did not deviate in period t, the payoff from t onwards would be:
2
2 + 2 + 2 2 + = .
1
For deviation to be suboptimal, we need:
2
>3+
1 1
which implies:
1
> .
2
Thus if the players are patient enough, cooperation can be sustained in equilibrium. In
other words, playing (C, C) always can be the outcome of an equilibrium if the discount
factor is at least 1/2.
4.6.2 Folk theorem

We showed above that the cooperative payoff (2, 2) can be sustained in equilibrium.
However, this is not the only possible equilibrium outcome. Indeed, many different
payoffs can be sustained in equilibrium.
70
For example, note that always playing (D, D) is an equilibrium no matter what the
value of is. Each player simply adopts the strategy play D initially, and at any period
t > 1, play D irrespective of history. Note that both players adopting this strategy is a
mutual best response. Therefore, we can sustain (1, 1) in equilibrium. In fact, by using
suitable strategies, we can sustain many more in fact infinitely more payoffs as
equilibrium outcomes. For the Prisoners Dilemma game, we will describe the set of
sustainable payoffs below.
The result about the large set of payoffs that can be sustained as equilibrium outcomes
is known as the folk theorem. These types of results were known to many game
theorists from an early stage of development of non-cooperative game theory. While
formal proofs were written down later, we cannot really trace the source of the idea,
which explains the name.
Here we state a folk theorem adapted to the repeated Prisoners Dilemma game. To
state this, we will need to compare payoffs in the repeated game to payoffs in the
one-shot game that is being repeated. The easiest way to do that is to normalise the
repeated game payoff by multiplying by (1 ). Then if a player gets 2 every period,
the repeated game payoff is (1 ) 2/(1 ) = 2. As you can see, this normalisation
implies that the set of normalised repeated game payoffs now coincide with the set of
payoffs in the underlying one-shot game. So now we can just look at the set of payoffs of
the one-shot game and ask which of these are sustainable as the normalised payoff in
some equilibrium of the repeated game.
For the rest of this section, whenever we mention a repeated game payoff, we always
refer to normalised payoffs. Note that in this game, a player can always get at least 1 by
simply playing D. It follows that a player must get at least 1 as a (normalised) repeated
game payoff.
To see the whole set of payoffs that can be sustained, let us plot the payoffs from the
four different pure strategy profiles. These are shown in Figure 4.8 below. Now join the
payoffs and form a convex set as shown in the figure. We now have a set of payoffs that
can arise from pure or mixed strategies.
The Folk theorem is the following claim. Consider any pair of payoffs (1 , 2 ) such
that i > 1 for i = 1, 2. Any such payoff can be supported as an equilibrium payoff for
high enough .
As noted above, for the Prisoners Dilemma game we can also sustain the payoff (1, 1)
as an equilibrium outcome irrespective of the value of .
The set of payoffs that can be supported as equilibrium payoffs in our example is shown
as the shaded part in Figure 4.8.
Example 4.2 Consider the following game.

Player 2
C D
Player 1 C 3, 2 0, 1
D 7, 0 2, 1
(a) Find conditions on the discount factor under which cooperation can be
sustained in the repeated game in which the above game is repeated infinitely.
71
4. Game theory
Figure 4.8: The set of payoffs that can be supported as equilibrium payoffs under an
infinitely-repeated game.
(b) Under what conditions is there an equilibrium in the infinitely-repeated game in

which players alternate between (C, C) and (D, D), starting with (C, C) in the
first period?
(c) Draw the set of payoffs sustainable in a repeated game equilibrium according to
the folk theorem.
(a) First, note that player 2 has no incentive to deviate from (C, C). To ensure that
player 1 does not deviate, consider the following strategy profile. Play (C, C)
initially. At any period t > 0, if (C, C) has been played in the last period, play
(C, C). Otherwise, switch to (D, D).
Under this strategy profile, player 1 will not deviate if:
3
7+2 6
1 1
which implies > 4/5.
(b) We want to support alternating between (C, C) and (D, D), starting with
(C, C) in period 0, as an equilibrium.
Note first that player 2 has no incentive to deviate in odd or even periods.
Player 1 cannot gain by a one-shot deviation in odd periods, when (D, D) is
supposed to be played. So the only possible deviation is by player 1 in even
periods, when (C, C) is supposed to be played.
To prevent such a deviation, consider the following strategy profile.
Start by playing (C, C) in period 0.
In any odd period t (where t = 1, 3, 5, . . .) play (D, D) (irrespective of
history).
72
In any even period t (where t = 2, 4, 6, . . .):

play (C, C) if (C, C) has been played in the previous even period t 2
otherwise play (D, D).
Note that this is a version of the trigger strategy. After any deviation from
cooperation in even periods, (D, D) is triggered forever.
If player 1 does not deviate in any even period t, player 1s payoff (t onwards) is:
Vt = 3 + 2 + 3 2 + 2 3 +
= 3(1 + 2 + 4 + ) + 2(1 + 2 + 4 + )
3 + 2
= .
1 2
If player 1 does deviate in even period t, player 1s payoff (t onwards) is:

Vtdev = 7 + 2 + 2 2 + = 7 + 2 .
1
Therefore, player 1 prefers not to deviate in either period 0 or in any even
period if:
3 + 2
> 7 + 2
1 2 1
which simplifies to 5 2 4 > 0, implying:
r
4
> 0.89.
5
Note also the repeated game payoff generated in this equilibrium. To see this, first
normalise the payoff by multiplying by (1 ). The normalised payoff of player 1
starting any even period is:
3 + 2 3 + 2
(1 ) 2
= .
1 1+
Similarly, the normalised payoff of player 1 starting any odd period is
(2 + 3)/(1 + ). Note that either payoff goes to 5/2 as 1.
For player 2, the normalised payoff from the equilibrium is (2 + )/(1 + ) starting
any even period, and (1 + 2)/(1 + ) starting any odd period. Note that either
payoff goes to 3/2 as 1.
In other words, this exercise shows you an example of an equilibrium that sustains a
payoff in the interior of the set of payoffs which is sustainable according to the folk
theorem (which, in this case, is anything strictly above (2, 1), or the point (2, 1)
itself).
73
4. Game theory

to:
analyse simultaneous-move games using dominant strategies or by eliminating

dominated strategies either once or in an iterative fashion
calculate Nash equilibria in pure strategies as well as Nash equilibria in mixed

strategies in simultaneous-move games
explain why Nash equilibrium is the central solution concept and explain the
importance of proving existence
specify strategies in extensive-form games
analyse Nash equilibria in extensive-form games
explain the idea of refining Nash equilibria in extensive-form games using backward
induction and subgame perfection
analyse the infinitely-repeated Prisoners Dilemma game with discounting and

analyse collusive equilibria using trigger strategies
explain the multiplicity of equilibria in repeated games and state the folk theorem
for the Prisoners Dilemma game.

1. Consider the strategic form game below with two players, 1 and 2. Solve the game
by iteratively eliminating dominated strategies.
Player 2
A2 B2 C2
A1 3, 3 1, 4 0, 5
Player 1 B1 2, 1 3, 2 1, 0
C1 1, 0 0, 1 1, 0
2. Identify actions and strategies of each player in each of the following games (Figure
4.9 to Figure 4.11).
74
Figure 4.9: Perfect information.
Figure 4.10: Imperfect information for player 2.
Figure 4.11: Imperfect information for both players.
75
4. Game theory
3. Consider the extensive-form game of imperfect information in Figure 4.12.
(a) Write down the actions and strategies for each player.
(b) Identify the pure and mixed strategy Nash equilibria.
(c) Identify the pure and mixed strategy subgame perfect Nash equilibria.
4. Consider the extensive-form game in Figure 4.13.
(b) Identify the pure strategy Nash equilibria.
(c) Identify the pure strategy subgame perfect Nash equilibria.
76
5. Consider the extensive-form game in Figure 4.14.
(b) Identify the pure strategy Nash equilibria.
(c) Identify the pure strategy subgame perfect Nash equilibria.
6. Find the pure strategy subgame perfect Nash equilibria for the game in Figure 4.9
in Question 2.
7. Suppose the following game is repeated infinitely. The players have a common
discount factor , where 0 < < 1.
Player 2
C D
Player 1 C 3, 2 0, 3
D 5, 0 2, 1
(a) Find conditions on the discount factor under which cooperation (which implies
playing (C, C) in each period) can be sustained as a subgame perfect Nash
equilibrium of the infinitely-repeated game. Your answer must specify the
strategies of players clearly.
(b) In a diagram, show the set of payoffs that can be supported in a repeated
game equilibrium according to the folk theorem.
77
4. Game theory
78
Chapter 5
Production, costs and profit
maximisation
5.1 Introduction
Over the next few chapters we develop models to analyse different market structures. In
particular, we analyse two extremes: a competitive market and a monopoly. Later on we
also analyse oligopoly. To develop these models, we need to make use of a variety of
concepts on the nature of technology, costs and optimisation by firms. This chapter
introduces various concepts to this end, many of which you should already be familiar
with from previous courses.

This chapter aims to introduce a variety of concepts and calculations that can be used
to analyse the production decisions of firms under a variety of market structures.

explain the concepts of average and marginal product and their relationship
derive the optimal input purchasing decision of a firm in the short run
derive the optimal input mix decision in the long run
explain the concepts of isoquants and marginal rate of technical substitution
explain how the shape of isoquants depends on input substitutability
explain the concepts of marginal cost, average cost, fixed cost, sunk cost and
variable cost
explain the various cost curves faced by a firm in the short run and in the long run
derive the profit-maximising output of a firm
explain the shut-down decision of a firm.

N&S Chapters 6 and 7, Sections 8.1 and 8.2, and Sections 13.1 and 13.2.
79
5. Production, costs and profit maximisation

The chapter starts by introducing production and factor demand. It then analyses input
choice problems in the short run and in the long run. Next, it analyses cost curves in
the short and long runs. Finally, it introduces the problem of profit maximisation.
5.3 A general note on costs

The term cost in economics wherever it occurs refers to opportunity cost. In other
words, the cost of a resource is measured by how much value it would add in its next
best use. This is often different from how an accountant would view the costs of a firm.
In coming up with a figure for the annual costs of an owner-managed firm an
accountant would typically add up the expenses on various categories, but would not
include the opportunity cost of the work that the owner has put in or the opportunity
cost of own funds that the owner might have invested in the firm. However, for an
economist, these are costs just as wages to employees are costs. In sum, the notion of
opportunity cost tries to capture the true underlying cost of all factors, rather than
costs that are useful just for accounting purposes.
5.4 Production and factor demand

The production function of a firm records the maximum output from a given set of
inputs. In general, a firm might use several types of inputs. In our discussions, we will
assume that the firm uses two classes of inputs: capital and labour. We will also typically
assume that capital (such as plants and machinery) is fixed in the short run and only
variable in the long run, while labour is a variable input over any length of time.
The production function with two inputs can be written as a function:
Q = f (K, L).
5.4.1 Marginal and average product

The marginal product of labour is simply the extra output by raising labour input at
the margin while holding other inputs constant, given by:
f (K, L)
MPL = .
L
Similarly, the marginal product of capital is given by:
f (K, L)
MPK = .
K
We will assume that beyond some level of an input marginal product is diminishing. In
other words, beyond some level of labour (which could be zero), MPL is decreasing in
80
5.5. The short run: one variable factor
labour, and beyond some level of capital MPK is decreasing in capital. In the region
with diminishing marginal product, we have:
2 f (K, L)
<0
L2
and:
2 f (K, L)
< 0.
K 2
The average product of labour is given by:
Q
APL = .
L
You can define the average product of capital similarly.
What is the relationship between average and marginal product? You should
understand the general relationship between average and marginal. Suppose you play a
computer game and your average score is 60. The next time you play, you score 100.
What will happen to your average score? It will go up. Here the marginal score
exceeded the average, so the average goes up. Similarly, if the marginal were below the
average, the average would fall. The same principle applies to average and marginal
product (and to other comparisons such as between average and marginal cost that we
would make later). The average product rises if the average is below the marginal and
falls if the average is above the marginal. At low levels of output, the marginal product
is higher than the average product, and then eventually it decreases. Once it crosses
below the average product, the average product falls. It follows that the crossing point
is the highest point on the average curve.
Example 5.1 Plot the average and marginal product of labour for the following
production function:
Q = 10L + 30L2 L3 .
These are shown in Figure 5.1 below.
5.5 The short run: one variable factor

In the short run, capital is fixed, only labour is variable. How should a firm choose how
much labour to hire, and how does the level of hiring vary when the wage rate varies?
We consider the input choice by a firm assuming that the firm buys inputs from a
competitive market for inputs so that the firm takes the input prices as given in making
decisions about how much of any input to buy.
With this in mind, the choice of labour input can be derived as follows. If the firm hires
one extra unit of labour, its output rises by MPL . One extra unit of output raises
revenue by MRL . Thus the extra revenue from employing one extra unit of labour is
given by the marginal revenue product of labour:
MRPL = MRL MPL .
81
Figure 5.1: Average and marginal product of labour for Activity 5.1.
On the other hand, one extra unit of labour costs w (i.e. w is the marginal cost of
labour). Thus the profit-maximising level is given by MRPL = w.
Let us derive this formally. Let P (Q) be the (inverse) demand function facing the firm.
This is simply P if the firm is a price-taker (a competitive firm), but under imperfect
competition, market price depends on firm output.
The firm maximises profit given by:
P (Q) f (K, L) wL rK
where K is the fixed level of K in the short run, and where Q = f (K, L). Differentiating
with respect to L, we get:
Q
P 0 (Q) f + P (Q)fL w = 0.
L
Since Q/L = fL , we have:
fL P 0 (Q) f + P (Q) = w.

Again, because f (K, L) = Q, this can be written as:
fL P 0 (Q) Q + P (Q) = w.

Now, total revenue is P (Q) Q. So marginal revenue is P 0 (Q) Q + P (Q). Therefore, the
equation above is simply:
MPL MRL = w.
Note that for a competitive firm, price is given and does not vary with firm output.
Therefore, P 0 (Q) = 0, and the optimal labour hire equation becomes:
P fL = w
or price times marginal product equals the wage rate.
82
5.6. The long run: both factors are variable
How does demand for labour change as the wage rate varies in the short run? Note that
the marginal product of labour decreases with labour. Marginal revenue decreases as
output increases, and output increases as labour increases. Therefore, as labour
increases, MR falls. For a competitive firm MR is constant at price P . Thus either both
MR and MP fall as labour increases (imperfect competition) or MR is constant (at price
P ) while MP falls for competitive firms. Therefore, MRP falls as labour increases. It
follows that if the wage rate falls, more labour is hired.
Figure 5.2 below shows the effect of a fall in the market wage rate.
Figure 5.2: Marginal revenue product.
5.6 The long run: both factors are variable
5.6.1 Isoquants
With two variable factors (typically this would be the case in the long run), we can
represent the production function by its level curves. An isoquant traces out the
different combinations of K and L that produce a given level of output. Analogously to
an indifference curve, the slope of an isoquant is given by the marginal rate of
technical substitution (MRTS). The marginal rate of technical substitution of L for
K is the amount by which capital can be reduced if the firm hires one extra unit of
labour. Analogously to an indifference curve:
MPL
MRTSLK = .
MPK
5.6.2 Diminishing MRTS

Analogously to an indifference curve, we assume that as L increases, MRTSLK
decreases, giving an isoquant the familiar convex shape. If Q0 is the output along an
isoquant, then convexity means that the set of combinations of K and L producing Q0
or more is a convex set.
See Section 6.3 of N&S for a diagram showing isoquants.
83
5.6.3 Returns to scale

Suppose we scale up all inputs by a positive factor t. If output scales up by t as well, we
say that the production function exhibits constant returns to scale. If output scales up
by less than t, we have decreasing returns to scale. Why might this happen? What
prevents us from scaling up everything? It must be that there is some factor that does
not get scaled up. This could be the managers ability to manage a larger firm or some
other such factor. Therefore, in some fundamental sense we are not really scaling up
everything. However, conventionally, if we multiply all standard factors by t but output
scales up by a factor lower than t, we obtain diminishing returns to scale. Finally, if
output scales up by a factor greater than t, we obtain increasing returns to scale.
5.6.4 Optimal long-run input choice

We consider the input choice by a firm assuming that the firm buys inputs from a
competitive market for inputs so that the firm takes the input prices as given in making
decisions about how much of any input to buy.
We must minimise total costs wL + rK subject to the constraint that we produce Q0
units, i.e. f (K, L) = Q0 . First, set up the Lagrangian:
L = wL + rK + (Q0 f (K, L)).
The first-order conditions for a constrained minimum are:
L f
=w =0
L L
L f
=r =0
K K
L
= Q0 f = 0.

From the first two first-order conditions, we get:
w f /L
= = MRTSLK .
r f /K
These first-order conditions are not sufficient to guarantee a minimum. However, as
under utility maximisation, the second-order condition for a minimum is satisfied so
long as we have diminishing MRTS.
5.7 Cost curves

See N&S Sections 7.3 to 7.7.
5.7.1 Marginal cost and average cost

Suppose TC(Q) is the total cost function for a firm. Then the marginal cost is:
dTC(Q)
MC =
dQ
84
5.7. Cost curves
and the average cost is:

TC(Q)
AC = .
Q
Recall the previous discussion about the relation between the average and the marginal.
As pointed out there, when the marginal exceeds the average, the latter rises and vice
versa. It follows that if AC has a U-shape, MC must pass through the minimum point.
Before this point, MC is lower than AC so AC falls, and after this point MC is above
AC so AC rises.
5.7.2 Fixed costs and sunk costs

Fixed costs must be incurred to set up a firm. Some types of fixed costs cannot be
recovered even if the firm stops production or goes out of business. Buildings and plants
can be resold, but the money spent on advertising the firms brand name is sunk, as is
the cost incurred in printing stationery.
5.7.3 Short run

In the short run, capital is fixed, but labour is variable. Thus there are two categories of
costs: fixed costs, F , and variable costs, VC(Q). We have:
TC(Q) = F + VC(Q).
It follows that:
F VC(Q)
AC(Q) = + .
Q Q
The first component is the average fixed cost, and the second is average variable cost.
Example 5.2 Plot the AC, AVC and MC for the cost curve:
TC(Q) = 150Q 20Q2 + Q3 + 500.
Figure 5.3 below shows these cost curves.
Example 5.3 Let C(Q) = F + VC(Q) be the cost function of a firm where F is the
fixed cost and VC(Q) is the variable cost. Show algebraically that MC passes
through the lowest points of AC and AVC. Assume C 00 > 0 and V 00 > 0.
Note that dC/dQ = dVC(Q)/dQ = MC.
Let us find the minimum point of AC, which is C(Q)/Q. We have:
dC(Q)/Q 1
= 2 [Q C 0 (Q) C(Q)].
dQ Q
At the minimum point:
dC(Q)/Q
= 0.
dQ
85
Figure 5.3: Average cost, average variable cost and marginal cost curves for Activity 5.2.
Since Q > 0, this implies:

Q C 0 (Q) C(Q) = 0
which implies:
C(Q)
C 0 (Q) = .
Q
You should verify that at the point where the first-order condition holds, the second
derivative of average cost is given by C 00 /Q. Since C 00 > 0, the second derivative is
strictly positive, satisfying the second-order condition for a minimum. Thus at the
minimum point of AC, MC equals AC.
In the same way, by minimising VC(Q)/Q instead of C(Q)/Q we can show that at
the minimum point of AVC, MC equals AVC.
5.7.4 Long run
In the long run, all factors are variable, and the optimal input mix is chosen according
to the long-run optimisation rule derived previously. Since capital can be varied in the
long run, short-run costs are at least as high as the long-run costs, and if in the short
run the firm chooses anything other than the capital size associated with the lowest
long-run average costs, short-run costs are strictly higher.
See N&S Section 7.4 for an explanation using isoquants.
The long-run average cost curve (LRAC) is an envelope of the short-run average cost
curves (SRACs), as shown in Figure 5.4. Note that the long-run average cost is lower
than the minimum short-run average cost over the region where the firm experiences
increasing returns to scale. This is the region where the LRAC is falling. Similarly, over
the region where the LRAC is increasing, the firm experiences decreasing returns to
scale and, again, the LRAC is lower than the minimum of the SRAC. For the right
long-run plant size, the minimum point of the SRAC is also the minimum point of the
LRAC. Note also that the LRMC is flatter, i.e. more elastic, compared to the SRMC.
86
5.7. Cost curves
Figure 5.4: The long-run average cost curve is an envelope of the short-run average cost
curves.
As shown in Figure 5.4, if in the long run a firm experiences increasing returns to scale
at low levels of output, but decreasing returns at higher levels of output, the long-run
average cost is U-shaped, and the long-run MC cuts this from below at the lowest point.
However, if the firm experiences constant returns to scale, the long-run average cost
curve is a horizontal line at the level of the fixed long-run average cost. In this case
LRAC and LRMC are the same.
Example 5.4 Consider the production function:
Q = L0.5 K 0.5 .
In the short run, K is fixed at K = 100. The wage rate is 25 and the rental rate of
capital is 8. Derive the short-run and long-run cost functions.
To produce Q in the short run, we need L so that L0.5 K 0.5 = Q, i.e. we have:
Q2
L(Q) = .
K
Therefore, the short-run cost function is:
Q2
CSR (Q) = 8K + 25 .
K
You can easily work out the short-run AC, AVC and MC. These are shown in Figure
5.5 below for K = 100.
To derive the long-run cost, use the Lagrangian method to minimise cost for a given
level of output. This gives us the tangency condition price ratio equal to MRTS.
Using this:
w 0.5L0.5 K 0.5 K
= 0.5 0.5
= .
r 0.5L K L
87
Figure 5.5: Short-run cost curves.
So K = 25L/8. It follows that:
CLR (Q) = wL + rK = 25L + 25L = 50L.
Now:
5
Q = (25L/8)0.5 L0.5 = L
2 2
which implies:
2 2
L= Q.
5
Therefore:
CLR (Q) = 20 2 Q.

In this case LRAC = LRMC = 20 2.
Figure 5.6 below shows the short run cost and MC curves (the MC curves are the
dashed lines) for fixed short-run K levels of K = 100 and K = 200.

The minimum of SRAC with capital fixed at K occurs at 2 2K/5 (you should
derive this explicitly). The LRAC is the envelope of SRAC as K varies.
In this case we have a constant-returns-to-scale technology, so the LRAC is
horizontal and coincides with the LRMC.
Example 5.5 Consider a firm with production function Q = min{K, L}. The
wage rate is w, and the rental rate of capital is r. The firm wants to produce Q0
units of output. In the short run, K is fixed at K.
(a) Calculate the short-run optimal cost as a function of , , input prices w, r and
Q0 .
88
5.7. Cost curves
Figure 5.6: Short-run average cost curves for K = 100 (SRAC1) and K = 200 (SRAC2),
the associated marginal cost curves (shown as dashed lines), and the long-run average
(as well as marginal) cost curve.
(b) Calculate the long-run optimal cost as a function of , , input prices w, r and
Q0 .
(a) We consider the short-run optimum. Note that it is impossible to produce more
than K. Thus if Q0 exceeds this level, it cannot be produced in the short run.
Let us assume Q0 6 K. Then we need L so that L = Q0 , or:
Q0
L= .

The optimal short-run cost for Q0 6 K is, therefore, given by:
Q0
CSR = rK + w .

(b) We now consider the long-run optimum. To minimise cost, note that we must
have K = L. To produce Q0 , L and K must be such that:
K = L = Q0 .
Therefore:
Q0
K=

and:
Q0
L= .

It follows that the minimised value of cost is:

r w
C(Q0 ) = Q0 + .

89
5.8 Profit maximisation
See Chapter 8 of N&S.

The profit of a firm is given by revenue minus cost: = P (Q) Q C(Q). This is
maximised at the quantity where marginal revenue equals marginal cost:
d dC(Q)
(P (Q) Q) = 0.
dQ dQ
For a price-taking firm, the price P does not depend on output, so d(P Q)/dQ = P , and
profit is maximised at the output where P = MC.
Profit-maximising output and shut-down decision
To decide the optimal level of output, a firm must consider two things.
Profit-maximising output
The firm must set MR equal to MC to determine the profit-maximising output level
Q .
Shut-down decision
Once Q is decided, the firm must see if it is worth producing at Q .
In the short run, a firm continues to operate even if revenues do not cover
all costs. If fixed costs on equipment such as plants and machinery have been
incurred, this cannot be altered in the short run. So the short-run shut-down
decision pays no attention to fixed costs and only considers if the firm is covering
its variable costs. The firm should continue to operate so long as it covers its
variable costs. In other words, the firm shuts down production in the short run
if P Q < VC, or:
P < AVC.
In the long run all factors are variable. Therefore, the firm shuts down if it
cannot cover all its costs even at the minimum level of LRAC. In other words,
a firm shuts down in the long run if:
P < LRACmin .
You should note that if some part of the fixed costs is sunk to start with, this plays no
role in the calculation to shut down even in the long run. This part of the cost is not
recoverable by shutting down, so the shut down decision does not take it into account.
90
5.9. A reminder of your learning outcomes

to:
explain the concepts of average and marginal product and their relationship
derive the optimal input purchasing decision of a firm in the short run
derive the optimal input mix decision in the long run
explain the concepts of isoquants and marginal rate of technical substitution
explain how the shape of isoquants depends on input substitutability
explain the concepts of marginal cost, average cost, fixed cost, sunk cost and
variable cost
explain the various cost curves faced by a firm in the short run and in the long run
derive the profit-maximising output of a firm
explain the shut-down decision of a firm.

1. Derive the returns to scale associated with the following production functions:
(a) Q = min{L, K}, , > 0.
(b) Q = AL + K , , > 0 and + = 1.
2. Consider a firm with production function Q = K L . The wage rate is w, and the
rental rate of capital is r. Calculate the short-run optimal cost as a function of , ,
input prices w, r and Q.
3. Consider a firm with production function Q = min{K, L}. The wage rate is w,
and the rental rate of capital is r. Derive the long-run cost function of the firm.
4. Consider a firm with production function Q = K 1/2 L1/2 . The wage rate is w, the
rental rate of capital r. Derive the long-run cost function of the firm.
5. A firms short-run cost curve is C(Q) = 100 + 50Q bQ2 + Q3 /2. For what values
of b are cost, average cost and average variable cost positive at all levels of Q?
6. Consider a firms input choice in the long run. Suppose the production function of
a firm is given by Q = L K . The wage rate is w, and the capital rental rate is r.
(a) Calculate the firms optimal input mix for producing output Q0 .
91
(b) Suppose = = 0.5, w = 12, r = 3 and Q0 = 40. Calculate the optimal input
mix.
(c) What is the cost at the optimal input mix derived in (b)?
(d) Calculate the marginal products of labour and capital at the optimal choice.
7. Suppose Q = aP b , where b < 0.

(a) Calculate the price elasticity of demand.
(b) Calculate MR as a function of P .
(c) Show that at the limit as the price elasticity goes to , the firm is a
price-taker.
92
Chapter 6
Perfect competition in a single market
6.1 Introduction
The analysis of consumer behaviour in Chapter 2 showed how we can derive the
demand curve in a market. In the first part of this chapter, we analyse the behaviour of
competitive firms, and show how we can derive the supply curve in a market. In the
second half of this chapter, we study the partial equilibrium competitive model that
puts demand and supply together to derive the equilibrium in individual markets. Note
that this is different from a general equilibrium analysis that considers all markets in
an economy together, which we do in the next chapter. The partial equilibrium
approach allows us to understand in a simple setting how small price-taking agents
(both buyers and sellers) come together in a competitive market. Furthermore, we can
use this supply-demand model to carry out a partial equilibrium analysis of the impact
of policy on specific markets.

In this chapter we aim to study the behaviour of competitive firms and the outcome
under competition in specific markets. In other words, we analyse the partial
equilibrium competitive model here. We aim to show that even though such
competition is an idealised extreme, the model is very useful in studying a variety of
markets and policy initiatives.

explain the supply decision of a price-taking firm in the short run and in the long
run
derive the industry supply curve in the short run and in the long run
calculate the market equilibrium price and quantity given supply and demand
curves
analyse the impact of entry and exit decisions of firms and their impact on industry
supply
relate the nature of input costs to the long-run industry supply curve
explain supply elasticities
93
6. Perfect competition in a single market
explain the concept of producer surplus and use this in conjunction with consumer
surplus as a measure of welfare
use the measure of welfare to evaluate the impact of a variety of tax and subsidy
policies on the market
explain the concept of the incidence of a tax or a subsidy
use the measure of welfare to evaluate the impact of policies: examples include
price ceilings (such as rent control), price floors (such as minimum price guarantees
for agricultural products), permits and quotas, price support policies as well as
trade policies involving tariffs and quotas.

Section 8.5 and Chapter 9 of N&S. See also Perloff Chapter 9 for a variety of
applications of the competitive model. Also see the auxiliary textbooks mentioned in
the introduction for analyses of different types of policies.

The first part of the chapter introduces the supply decision of a competitive firm. The
chapter then introduces competitive market equilibrium and the notion of producer
surplus. Finally, the chapter evaluates a variety of policies using the notions of
consumer and producer surplus as welfare measures.
6.3 A general comment on zero profit

In this and other chapters, we refer to firms earning a zero profit. You might wonder
whether this assumption makes sense after all, why would a firm bother producing if
it makes no profits at all? However, the assumption of zero profits does not imply the
absence of an accounting profit. Each factor of production must cover at least their
opportunity cost, which is the amount the factor would earn in its next best use. It
follows that the owners of the firm must receive a normal level of remuneration that
would cover the opportunity cost of being the owner, so that they do have the incentive
to keep producing. In accounting terms there will be a normal level of profit that
covers such opportunity costs. The zero-profit assumption simply means that there are
no profits beyond this normal level. This is why sometimes the assumption is stated as
no super-normal profits.
94
6.4. Supply decision by a price-taking firm
6.4 Supply decision by a price-taking firm
6.4.1 Which types of firms have a supply curve?
In discussing markets, we often talk about market supply and market demand and how
the intersection of the two determines market equilibrium. But does any kind of market
structure produce a market supply curve? In fact, this is not true. A market supply
curve is an appropriate concept only when the firms in an industry are price-takers. In
this case, we can associate a specific firm and industry output to each price and,
therefore, trace out firm and industry supply curves, respectively. But if a firm has
control over price, that is, if the market price depends on firm output, the firm would
consider market demand and see at which point on market demand it should produce.
In this case, there is no unique relationship between price and output. As demand
conditions vary, so would the price-quantity pair of a firm vary. Therefore, we cannot
trace out a supply curve. So keep in mind that whenever you say supply curve you are
in fact talking about price-taking firms and, by implication, a competitive market
structure.
6.4.2 Short-run supply
A price-taking firm faces a notional perfectly-elastic demand curve. Essentially, the firm
can sell any amount at the given price, P . The idea is that a firm in a competitive
market is a very small part of the overall market so that whatever amount such a firm
produces is a very small part of the overall supply in the market. Therefore, variations
in the output of a single firm have a vanishingly small impact on the market price,
hence the notional firm demand curve is horizontal. Figure 6.1 below shows profit
maximisation by a firm in the short run.
Figure 6.1: Short-run profit maximisation by a competitive firm. The demand curve facing
the firm is perfectly elastic at price P (i.e. the firm is a price-taker). Profit is maximised
at P = MC and the maximised value of profit is (P AC(Q )) Q .
95
The (inverse) supply curve of a price-taking firm is given by:

P = MC for P > AVC.
In other words, the supply curve is the part of the short-run MC that is above the AVC
curve.
Example 6.1 Derive the short-run firm supply curve given the following cost
functions:
(a) C(q) = 100 + 3q + 5q 2 .
(b) C(q) = 100 + 10q 6q 2 + q 3 .
(a) Suppose C(q) = 100 + 3q + 5q 2 . Then:
AVC = 3 + 5q
and:
MC = 3 + 10q.
In this case, MC is always above AVC, so the (inverse) short-run supply curve
of a price-taking firm is given by:
P = 3 + 10q.
You can invert this to get the usual form of supply curve:
P 3
q= .
10
(b) Suppose C(q) = 100 + 10q 6q 2 + q 3 . Then:
AVC = 10 6q + q 2 .
The minimum point of AVC is at:
6 + 2q = 0
or q = 3. At q = 3, AVC = 1. Therefore:
AVCmin = 1.
The (inverse) supply curve is given by P = MC for P > 1. For lower values of P
the supply is zero. Therefore, the supply curve is given by:
P = 10 12q + 3q 2 for P > 1, and q = 0 for P < 1.
This is shown in Figure 6.2 below.

If you are simply deriving the supply curve in an examination question, you can
stop here. Here, let us derive quantity supplied as a function of price by
inverting the first part. To invert, solve the quadratic equation:
3q 2 12q + (10 P ) = 0.
96
6.5. Market supply and market equilibrium
Figure 6.2: Firm supply curve.
This has two solutions:

P +2
q =2 .
3
Note that at the minimum AVC, q = 3. In other words, the firm either supplies
zero (up to price 1) and then supplies 3 or more as the price rises above 1. Thus
q < 3 is not admissible. Therefore, we can discard the solution with the negative
sign after 2. Then we have the supply curve:
(
0 for P < 1
q=
2 + P + 2/ 3 for P > 1.
6.4.3 Long-run supply

In the long run, all factors are variable. Therefore, the (inverse) supply curve is given by
the LRMC curve above the minimum point of the LRAC curve:
P = LRMC for P > LRACmin .
6.5 Market supply and market equilibrium

Sections 9.1 to 9.8 of N&S provide an analysis of market supply and market equilibrium
in short and long runs. You should go through these sections carefully. Here we provide
a brief outline of the topics.
6.5.1 Short run

The short-run market supply curve can be obtained by summing individual firm
supply curves.
97
Example 6.2 Consider the firm supply functions derived in Activity 6.1. Suppose
there are n identical firms in the industry. Derive the market supply function in each
case.
(a) We have q = (P 3)/10. Since there are n firms, the market supply is Q = nq.
Therefore, the market supply curve is given by:
P 3
Q=n .
10

(b) Here, we have q = 2 + P + 2/ 3 for P > 1. Therefore, the industry supply is:
r
P +2
Q = 2n + n
3
for P > 1. As before, for P < 1, supply is zero.
The intersection of market demand and market supply determines the market
equilibrium price and quantity.
6.5.2 Long run

If there is a fixed number of firms in an industry, and no entry is possible, the long-run
firm supply curve is simply:
P = LRMC for P > LRACmin
and the long-run market supply is simply the firms supplies added together.
However, once we allow entry and exit, this is a little different. The process of entry and
exit implies that firms always end up earning a zero profit and producing at the
minimum long-run average cost. How this minimum behaves as the market output
expands determines the shape of the long-run market supply curve.
If input costs do not rise as new firms enter, the long-run supply curve is horizontal at
the minimum LRAC. In this case the long-run market supply equation is:
P = LRACmin .
If input costs rise, we have an increasing-cost industry and you should be able to show
that in this case we get a long-run market supply curve that slopes upwards. Similarly,
if we have a decreasing-cost industry, the long-run market supply curve slopes
downwards. However, whether there are any realistic cases of a downward-sloping
supply function is unclear see N&S for a discussion.
Just as we can calculate demand elasticities, we can calculate supply elasticities in the
short and long runs. The elasticity of supply is simply:
P dQS
S = .
QS dP
98
6.6. Producer surplus
This is positive in all cases other than for the long-run supply under a decreasing-cost
industry (if this is at all possible).
Furthermore, the long-run supply curves are more elastic than their short-run
counterparts (why?).
6.6 Producer surplus

Just as we derived consumer surplus using demand curves, we can describe producer
surplus using upward-sloping supply curves. See Section 9.9 of N&S for a description of
producer surplus. Understanding consumer and producer surplus thoroughly is very
important for working through the variety of applications of the competitive model of
supply and demand.
6.7 Applications of the supply-demand model: partial

equilibrium analysis
We have developed a model of competitive markets throughout this and the previous
chapter. Why is this model useful? As you will see, this model can be put to work to
analyse a whole host of policy options. Indeed, much of our understanding of tax and
subsidy policies, as well as price control policies, derives from this fairly simple model.
The simplicity of the model and yet its capacity to accommodate a large number of
variations makes it a powerful tool of analysis.
However, there is also a limitation. We are looking at the impact of policy on the specific
market on which it is aimed. For example, if the government announces a rent control
policy, we look at the impact on the market for house rentals. This type of analysis is
known as partial equilibrium analysis since we are not considering the impact of a
policy on other markets, or the feedback effect from these other markets on the original
market. For such an analysis, we need to go to a general equilibrium framework.
However, general equilibrium analysis is more complicated, so partial equilibrium
analysis is useful as a simple tool to gain an understanding of the impact of a policy.
Read N&S Section 9.10 carefully. This develops the analysis of tax incidence and trade
tariffs. You should also see other textbooks (such as Perloff, Chapter 9) for the analysis
of price floors (such as agricultural price support), price ceilings (such as rent control),
permits and quotas, as well as other applications of the competitive model.
Here we provide a brief discussion of a few applications.
6.7.1 Tax: deadweight loss and incidence

One of the most important applications of the supply-demand model is to study the
impact of taxes and subsidies. There are two aspects that you should understand. The
first is the impact on efficiency. A per-unit tax (or subsidy) creates a deadweight loss.
Second, there is the question of incidence, i.e. which side of the market actually pays
the tax or enjoys the subsidy in equilibrium. The answer to these questions depends on
99
the elasticities of supply and demand. You need to understand these issues clearly. N&S
develops this in Section 9.10. See also Perloff, Section 9.5.
The general principle is that the more inelastic end of the market bears more of any
(per-unit) tax, and benefits more from any (per-unit) subsidy.
Example 6.3 Suppose the (inverse) demand function in a market is given by

P = 20 2Q, and the (inverse) supply curve is P = 2 + Q. Suppose the government
imposes a per-unit tax of t = 6 on suppliers. Calculate the incidence of the tax, the
impact on consumer and producer surplus, the tax revenue and the deadweight loss
from the tax.
Note that before tax, the market equilibrium quantity is 6, and the equilibrium price
is 8. After tax, for any quantity, the price rises by t, i.e. the inverse supply curve
shifts up by the extent of the tax, so that it becomes P = 2 + Q + t. In this case,
P = 8 + Q. The new market equilibrium quantity is 4, and the equilibrium price is
12.
Note that consumers pay an extra 4 per unit. Therefore, the incidence on consumers
is 4 per unit. The suppliers receive 12, but pay 6 to the government per unit so they
receive a net-of-tax price of 6 per unit, 2 lower than before tax. Thus the incidence
on suppliers is 2 per unit. Note that the demand curve is less elastic than the supply
curve, hence the difference in incidence.
Let us now work out the welfare impact. In the following discussion, we denote
consumer surplus by CS, producer surplus by PS, and deadweight loss by DWL and
relate these to Figure 6.3.
Without the tax, the CS is (20 8) 6/2 = 36, the PS is (8 2) 6/2 = 18, so the
total surplus is 54. There is no DWL.
After the tax, the quantity falls to 4. The CS is the triangle ABC, with an area of
(20 12) 4/2 = 16. The PS is the triangle FGH, with an area of (6 2) 4/2 = 8.
The tax revenue is the rectangle BGHC, with an area of 24. The shaded triangle
EHC is the DWL, with an area of 6 2/2 = 6.
Another way to calculate the deadweight loss is to compare the total surplus (TS)
before and after tax. Before tax, the TS is 54. After tax, the TS is 16 + 8 + 24 = 48.
The difference, 6, is the DWL.
100
6.7. Applications of the supply-demand model: partial equilibrium analysis
Figure 6.3: The impact and incidence of a tax.
Example 6.4 Suppose the demand curve for wheat is Q = 100 10P , and the
supply curve is Q = 10P . What are the effects of a subsidy of s = 1 per unit on the
equilibrium, government subsidy costs, consumer surplus, producer surplus and total
welfare?
In the following discussion, again we denote consumer surplus by CS, producer
surplus by PS and deadweight loss by DWL.
In Figure 6.4 below, the initial equilibrium is E, where P = 5 and Q = 50. The CS is
(10 5) 50/2 = 125 and the PS is 5 50/2 = 125. The total surplus is 250 and
there is no DWL.
After the subsidy, the supply curve becomes P = Q/10 s = Q/10 1. The new
equilibrium is P = 4.5, Q = 55. The buyers pay 4.5 per unit, while sellers receive
4.5 + 1 = 5.5 per unit.
Note that the incidence is split equally between the two sides of the market. This
results from the fact that the supply and demand elasticities have the same absolute
value.
Now consider the impact of the subsidy.
The new CS is the triangle ABC which has an area of (10 4.5) 55/2 = 151.25.
The new PS is the triangle FGH which has an area of 5.5 55/2 = 151.25. The
government spending on subsidy is 1 55 = 55, which is the rectangle GBCH. Also,
there is now a DWL given by the shaded triangle EHC which has an area of
1 5/2 = 2.5.
Another way to calculate the DWL is to compare the total surplus (TS) before and
after the policy. Before the policy the TS is 250. After the policy, the TS is
151.25 + 151.25 55 = 247.5. Therefore, the DWL is 250 247.5 = 2.5.
Note that the DWL arises because after the subsidy production extends to levels
101
Figure 6.4: The impact of a subsidy.
where the marginal benefit (given by the demand curve) is lower than the marginal
cost (given by the supply curve). Note also the comparison with the DWL from a
tax. A per-unit tax gives rise to a DWL because it reduces quantity so some of the
surplus remains unexploited. For a subsidy, on the other hand, the DWL arises
because production extends beyond the efficient quantity.
6.7.2 Price ceiling

A different type of intervention involves setting a price ceiling. Rent control is an
example of such a policy. Of course, to be effective, the ceiling needs to be below the
price that would, in the absence of any control, clear the market. Figure 6.5 shows the
deadweight loss from the imposition of rent control which sets a price ceiling at PC . The
shaded areas show the deadweight loss in each case. It is clear that rent control works
well (generates a lower efficiency loss) when supply is relatively inelastic. This would
happen, for example, if suppliers do not have many alternative uses for the existing
housing stock.
The analysis above makes one implicit assumption that is unlikely to be true in
practice. Note that at the ceiling price, demand exceeds supply. In other words, more
consumers are queuing up for the object (rent-controlled apartments, say) than the
number of objects available for sale. In this situation, there is no guarantee that the
consumers with the highest value for the object will obtain the object. It may be that
some consumers who value the object just above the price ceiling get the object, while
some consumers who value it most (consumers located at the top part of the demand
curve) do not. In practice, such misallocations are bound to happen. However, the
analysis above ignored this possibility and assumed implicitly that the consumers who
value the object the most are the ones who obtain it. In other words, the analysis above
is the most optimistic case. Unless otherwise mentioned, this is the case you should
analyse in an examination. However, we present below the most pessimistic case to
102
Figure 6.5: Deadweight loss under rent control with different supply elasticities.
make the nature of the implicit assumption clear.
The most pessimistic case
Figure 6.6 shows the outcome under the most pessimistic case. The deadweight loss is
larger in this case.
6.7.3 Price floor

A minimum wage is the most well-known price floor policy. Figure 6.7 shows the impact
of a price floor. Note that some of the surplus is lost, and out of the rest, some of the
consumer surplus is shifted to suppliers. Note also that there is an implicit assumption
that the sellers with the lowest cost (those at the lowest end of the supply curve) are
the ones supplying in the market. However, this need not be the case: all sellers up to
103
Figure 6.6: Outcome under a price ceiling, the most pessimistic case. In this case
consumers between QD QC and QD obtain the good. The surplus area A is now lost,
but area B is gained, so the deadweight loss is given by area A minus area B.
production level Q2 would be happy to supply the product at the price floor, therefore a
misallocation is possible, which would raise the deadweight loss further.
Figure 6.7: Price floor at PF . The deadweight loss is given by area A plus area B.
104
6.7.4 Quota
Figure 6.8 shows the impact of a quota at Q1 . Note that, as in the case of a price floor
above, some of the surplus is lost, and out of the rest, some of the consumer surplus is
shifted to suppliers. Note also that there is an implicit assumption that the sellers with
the lowest cost (those at the lowest end of the supply curve) are the ones who obtain
the quota. Again, a misallocation would raise the deadweight loss. However, irrespective
of the initial allocation of quotas, if they are tradeable, it is likely that they would end
up in the hands of the most efficient sellers. In this respect, the implicit assumption is
perhaps not too restrictive in this case. When you study the commons problem in
Chapter 12 of the subject guide, you will come across an instance of transferable quotas:
an arrangement known as ITQs, used in managing fishing levels in several countries.
Figure 6.8: Quota at Q1 . Once the quota is imposed, the market price rises to P1 . The
deadweight loss is given by area A plus area B.
6.7.5 Price support policy
Under a price support policy for agricultural products, the government maintains a
minimum market price in order to ensure a minimum income level for farmers. To ensure
that the price does not fall below the announced support price, the government buys up
any excess supply at the support price. Figure 6.9 shows the effect of such a policy.
105
Figure 6.9: Price support.
Let us calculate the deadweight loss from the policy in Figure 6.9 above, and then try
to see whether there is a different policy which attains the same objective of raising the
income level for farmers, but incurs a smaller deadweight loss.
In this case, the change in consumer surplus is (A + B), the change in producer
surplus is +(A + B + C). Next, the cost to the government is PS (Q2 Q1 ). Adding
these up, the net loss from the policy is PS (Q2 Q1 ) minus the triangle C (area
B + D + E + F + G). This is shown in Figure 6.10 below.
More generally, this shows why income generation for farmers through price support
policies (the EU common agricultural policy is an important example) is inefficient. The
policy gets farmers to produce more (even though demand is less at a higher price) and
then spends government revenue to buy up the unsold amount. This is, of course, a
poor use of resources.
A better solution is to simply give the amount of money A + B + C to farmers directly,
and not interfere with the market. This avoids the unnecessary excess production, and
incurs no deadweight loss.
6.7.6 Tariffs and quotas
Suppose that the government wants to raise the price of a commodity on the domestic
market by reducing imports. Assume that the importing country is a price-taker in
world markets. Below, we compare the welfare effects of an import tariff with an import
quota.
The best way to proceed is to compare the two policies using a diagram. As shown in
Figure 6.11, a tariff can reduce imports and raise the domestic price as follows. Without
intervention, the domestic price is equal to the world price, PW . At this price, domestic
consumption is QD , of which QS is produced domestically, and QD QS is imported. A
106
Figure 6.10: Price support and deadweight loss. The deadweight loss is the pattern-filled
area which is PS (Q2 Q1 ) minus the triangle C.
Figure 6.11: The welfare effects of an import tariff and an import quota.
tariff of:
T = P PW
raises the price to P , increases domestic production to Q0S , and reduces exports to
Q0D Q0S .
The welfare loss from this policy is as follows. Consumer surplus changes by
107
(A + B + C + D). Producer surplus changes by A, and the government gains a tariff

income of D (imports times T ). The net welfare change is then:
W = (A + B + C + D) + A + D = B C.
Thus the deadweight loss is the area B plus the area C.

The same outcome in terms of imports and domestic price is achieved by imposing a
quota equal to Q0D Q0S . However, now the rectangle D is extra income of foreign
exporters, as they collect the higher price of P . Thus deadweight loss to the domestic
economy is the areas B and C as before, plus the area D. In other words, the
deadweight loss to the economy is higher under a quota compared to a tariff.

to:
explain the supply decision of a price-taking firm in the short run and in the long
run
derive the industry supply curve in the short run and in the long run
calculate the market equilibrium price and quantity given supply and demand
curves
analyse the impact of entry and exit decisions of firms and their impact on industry
supply
relate the nature of input costs to the long-run industry supply curve
explain supply elasticities
explain the concept of producer surplus and use this in conjunction with consumer
surplus as a measure of welfare
use the measure of welfare to evaluate the impact of a variety of tax and subsidy
policies on the market
explain the concept of the incidence of a tax or a subsidy
use the measure of welfare to evaluate the impact of policies: examples include
price ceilings (such as rent control), price floors (such as minimum price guarantees
for agricultural products), permits and quotas, price support policies as well as
trade policies involving tariffs and quotas.
108

1. An industry comprises of n identical firms. The short-run cost function of a
price-taking firm is given by C(q) = 4q 2 + 100. The market demand curve is
Q = 600 10P .
(a) Find the market equilibrium price and quantity.
(b) Will there be entry or exit from this industry in the long run?
2. In the short run, the demand for cigarettes is totally inelastic. In the long run,
suppose that it is perfectly elastic. What is the impact of a cigarette tax on the
price that consumers pay in the short run and in the long run?
3. Suppose the market demand curve is infinitely elastic. In a diagram, show the
deadweight loss from a subsidy. What is the incidence of any subsidy?
4. Suppose the government sets a price ceiling below the unregulated competitive
equilibrium price. Show the effect on the market equilibrium, consumer surplus,
producer surplus and total welfare.
5. Consider an agricultural price support policy where the government announces the
support price PS . Farmers then choose how much to produce, and sell everything to
consumers at the price at which everything can be sold. Let this price be P . The
government then pays farmers a deficiency payment equal to (PS P ) per unit on
the entire quantity produced.
Suppose the demand curve for wheat is Q = 100 10P and the supply curve is
Q = 10P . Suppose the government announces a support price of PS = 6 using a
deficiency payment programme.
(a) Calculate the quantity supplied, the price at which output is sold and the
deficiency payment.
(b) Calculate the effect on consumer surplus, producer surplus and total welfare.
109
110
Chapter 7
General equilibrium and welfare
7.1 Introduction
Previously, we have studied competitive equilibrium in specific markets. This chapter
now introduces the idea of an economy-wide general equilibrium. The analysis of
competitive markets allowed us to evaluate welfare in a specific market. A general
equilibrium allows an evaluation of welfare in the entire economy. To simplify the
exposition and make the main ideas clear, we focus extensively on an exchange
economy. We consider the conditions for a general equilibrium and discuss the two
welfare theorems in detail. Later on, we assume that there is one agent (or a
representative agent) in an economy to introduce production in a simple setting.

The previous chapter considered equilibrium in a single market. Here we aim to
consider the general equilibrium in the economy as a whole. We aim to characterise the
equilibrium in a simple exchange economy, and understand the welfare properties,
especially the two well-known welfare theorems and their implications. Next, we aim to
incorporate production in the economy and characterise the equilibrium and welfare
properties under production.

construct an Edgeworth box and depict the contract curve, the set of possible
equilibrium exchanges given an endowment point, and depict the equilibrium price
ratio and allocation
derive algebraically the contract curve, the equilibrium price ratio and allocation
explain that only relative prices matter
explain the importance of establishing existence
state and provide an informal proof outline for the first welfare theorem in the
context of an exchange economy with two agents and two goods as well as in the
context of an economy with production
discuss the policy implications of the first welfare theorem and understand the
situations in which the theorem might not apply
111
7. General equilibrium and welfare
state and provide an informal proof outline for the second welfare theorem in the
context of an exchange economy with two agents and two goods
discuss the policy implications of the second welfare theorem and understand the
derive the equilibrium conditions in an economy with production.

N&S Chapter 10, Sections 10.1 to 10.7. However, the coverage is not ideal for the case
of an exchange economy. Below we extend coverage of some of the topics, and also give
references to parts of Perloff, Chapter 10.

The chapter starts by covering general equilibrium in an exchange economy. Next, the
chapter remarks briefly on the existence of equilibrium before moving on to considering
the content and implications of the two welfare theorems. Finally, the chapter outlines
general equilibrium in an economy, including production.
7.3 General equilibrium in an exchange economy

See N&S Section 10.7 and Perloff Sections 10.2 and 10.3, which contain a better
exposition.
An exchange economy is characterised by endowments and preferences of all
individuals. We first give a general definition of the general equilibrium (or exchange
equilibrium) for such an economy, sometimes called the Walrasian equilibrium. Suppose
there are I > 1 agents and n > 1 goods. Let xij denote the demand by agent i for good
j, and let eij be agent is endowment of good j. The utility function of agent i is
ui (xi1 , . . . , xin ). The budget constraint of i is:
p1 xi1 + p2 xi2 + + pn xin = p1 ei1 + p2 ei2 + + pn ein .
General equilibrium
A general equilibrium for a pure exchange economy consists of prices (p1 , . . . , pn ) and
quantities ((x11 , . . . , x1n ), . . . , (xI1 , . . . , xIn )) such that:
at the given prices, for each i, the bundle (xi1 , . . . , xin ) maximises utility of agent
i, subject to the agents budget constraint
all markets clear, i.e. for each market j, total demand equals total supply:
I
X I
X
xij = eij .
i=1 i=1
112
7.3. General equilibrium in an exchange economy
The final consumption bundles of agents are together known as the equilibrium
allocation.
Here we study a simple case with two agents (A and B) and two goods (X and Y ).
General equilibrium in such a two-person exchange economy can be usefully studied
using an Edgeworth box diagram. You should be able to construct such a diagram
and depict the set of Pareto efficient outcomes (the contract curve), be able to identify
the possible set of Pareto-improving trades given initial endowments, be able to identify
the possible set of equilibrium exchanges, and be able to show the equilibrium price
ratio and equilibrium allocation.
Let XA and YA denote the endowments of goods X and Y , respectively, for agent A.
Similarly, let XB and YB denote the endowments of goods X and Y , respectively, for
agent B. Let:
X = XA + XB and Y = YA + YB .
In an Edgeworth box, the contract curve shows the Pareto efficient allocations. This is
shown in Figure 7.1 below.
Figure 7.1: The contract curve traces out the Pareto efficient allocations in the Edgeworth
box. These are points of tangency between the indifference curves of A and B. Once we
reach such a point of tangency, it is impossible to make one agent better off without
making the other agent worse off. Therefore, if through trading a point on the contract
curve is reached, no further trade is possible. Conversely, if trading has reached any point
not on the contract curve, further Pareto-improving trades are possible.
We can also derive the set of Pareto-efficient allocations algebraically. To do this, we

solve the following problem:
max uA (XA , YA )
XA , YA , XB , YB
113
subject to the following constraints:

uB (XB , YB ) = u
XA + XB = X
YA + YB = Y .
Now form the Lagrangian:

L = uA (XA , YA ) + uB (XB , YB ) u + 1 X XA XB + 2 Y YA YB .
The first-order conditions are:

L uA
= 1 = 0
XA XA
L uA
= 2 = 0
YA YA
L uB
= 1 = 0
XB XB
L uB
= 2 = 0.
YB YB
We omit the three other derivatives of L with respect to the Lagrange multipliers that
simply recover the three constraints.
From the above equations:
uA /XA 1 uB /XB
= = .
uA /YA 2 uB /YB
Therefore, we have:
MRSA = MRSB .
This is precisely the condition that the indifference curves of the two agents must be
tangent to each other at Pareto-efficient allocations.
Example 7.1 Suppose:
uA = X Y 1 and uB = X Y 1 .
(a) Derive the contract curve equation.
(b) Show that the contract curve is a straight line if = .
(a) The contract curve can be derived by equating the MRS of the two agents:
YA YB
= .
1 XA 1 XB
Using the constraints XA + XB = X and YA + YB = Y , we have:
YA Y YA
= .
1 XA 1 X XA
114
Figure 7.2: Contract curves for various values of .
After some algebraic manipulation (which you should do yourself), we get:
(1 ) Y XA
YA = .
XA ( ) + (1 )X
This is the equation of the contract curve.
(b) If we set = , the contract curve becomes:
Y
YA = XA
X
which is obviously a straight line (a ray from the origin).
Figure 7.2 plots the contract curve for = 0.4 and various values of assuming
X = 50 and Y = 100.
Figure 7.3 shows the set of allocations that are possible to achieve by trading starting
from the initial allocation, E. These allocations give at least as much utility as the
endowment to either agent. In other words, each point in this set of allocations
(excluding E itself and the other crossing point between the indifference curves) Pareto
dominates E.
We can now introduce prices. The slope of the price line is PY /PX . Trading can take
place along any such price line. However, as Figure 7.4 shows, any arbitrary price line
cannot guarantee an equilibrium.
Figure 7.5 shows the property that the price line must satisfy in order to produce an
equilibrium. It must be tangent to the indifference curves of both agents. This can of
course happen only on the contract curve, since this also requires the indifference curves
themselves to be tangent to each other.
115
Figure 7.3: E is the endowment point. The shaded area enclosed by the indifference curves
passing through E shows the allocations that are Pareto-improving compared to E, i.e.
this is the area of mutually-beneficial trades. The segment FG is the part of the contract
curve on which equilibrium trades lie.
Figure 7.4: The price line shown here is not consistent with equilibrium. Note that there
is excess demand for Y (the total demand for Y exceeds the total endowment of Y ) and
excess supply of X. So neither market is in equilibrium. Starting from this situation, the
price of X would fall and the price of Y would rise, so that the price line should become
steeper and steeper until the points L (where As utility is maximised given the prices)
and M (Bs utility is maximised given the prices) coincide and we get an equilibrium.
116
Figure 7.5: The equilibrium price line (the slope of the price line is the ratio of prices) is
tangent to the indifference curves of both A and B at the same point. Since the indifference
curves are therefore also tangent to each other, this point must be on the contract curve.
Here H is the equilibrium allocation.
Relative prices
Note that the price line and, therefore, the equilibrium allocation would remain
unchanged if all prices are multiplied by the same positive number. Essentially, in a
general equilibrium we can only determine relative prices. If you study economics at
Masters level, you will see that this is a consequence of Walras law which states that if
out of n markets, n 1 are in equilibrium, then so must be the remaining one. This
implies that there are only n 1 independent prices. In our context with two goods, if
the market for good X is in equilibrium, so must be the market for good Y . So there is
only one independent price. This is the relative price of good X, say. In other words, we
can only determine the ratio PX /PY and not the levels of PX and PY separately. So we
could declare PY to be equal to 1 and determine PX . (It is as if the price of good X is in
units of good Y .)
Example 7.2 Consider an economy with two goods, x and y, and two agents, A
and B. Suppose:
uA = x y 1 and uB = x y 1 .
Suppose the endowments are (Ax , Ay ) for A and (Bx , By ) for B. Derive the
equilibrium price ratio and the equilibrium allocation.
The budget constraint of A is:
px xA + py yA = px Ax + py Ay .
The budget constraint of B is:
px xB + py yB = px Bx + py By .
117
Let MA = px Ax + py Ay and MB = px Bx + py By .
You should derive explicitly the demand functions by maximising utility for each
agent subject to the budget constraint of that agent. The demand functions are:
MA
xA =
px
(1 )MA
yA =
py
MB
xB =
px
(1 )MB
yB = .
py
Now consider market-clearing conditions. We know that if one market clears, so

must the other. Therefore, we only need to solve for supply equal to demand for one
market. Let us consider market-clearing for good x.
Market clearing for good x requires:
xA + xB = Ax + Bx .
This gives us:

MA MB
+ = Ax + Bx .
px px
Let the relative price of y be p, i.e. we have:
py
= p.
px
Substituting the values of MA and MB , we have:
(Ax + pAy ) + (Bx + pBy ) = Ax + Bx .
Solving:
p(Ay + By ) = (1 )Ax + (1 )Bx
which gives us:
(1 )Ax + (1 )Bx
p= .
Ay + By
We have derived the equilibrium price ratio py /px (i.e. the relative price of y).
Substituting this in the demand functions, we can obtain the equilibrium allocation.
For example:
xA = (Ax + pAy )
where p has been derived above. You can derive the rest of the allocation in this
manner.
118
7.4. Existence of equilibrium
7.4 Existence of equilibrium
In Chapter 4 of the subject guide we discussed the importance of showing that an

equilibrium exists. For exactly the same set of reasons, it is important to establish the
existence of a general competitive equilibrium before exploring the properties of such an
equilibrium. If an equilibrium cannot be shown to exist, we cannot escape the
possibility that deriving properties of an equilibrium is an empty exercise.
We noted in Section 4.3.5 that Nash proved the existence of his equilibrium concept in
1951 using a fixed-point theorem. Using such a theorem, Kenneth Arrow, Gerard
Debreu and Lionel McKenzie proved the existence of a general competitive equilibrium
under relatively weak assumptions in 1954.
If you study economics at more advanced levels, you will see a formal proof that
constructs a mapping from a set of prices to itself, which allows a fixed-point theorem
to be used to prove existence.
7.5 Uniqueness
General equilibrium in an economy need not be unique. Figure 7.6 shows three
equilibrium allocations, F, G and H, starting from endowment E, supported by different
equilibrium price lines.
Figure 7.6: Multiple equilibria.
119
7.6 Welfare theorems
Pareto efficient allocation
An allocation of resources is Pareto efficient if it is not possible to reallocate resources

to make one person better off without making someone else worse off.
The terms Pareto optimal, Pareto efficient, and efficient are used interchangeably.
Note that the notion of efficiency in economics says nothing about equity. Contract
curves include points where one consumer gets a very high utility while another gets a
very low utility. Later, when we study monopoly, you will see that if a monopolist can
extract the entire surplus, the market outcome is efficient, while if the monopolist
cannot do so the outcome is typically inefficient even though consumers do receive some
surplus. To take equity into account, we would need stronger notions involving social
justice criteria. However, except for a small set of topics where such notions are called
for, we do not worry about equity, and simply consider efficiency as the welfare measure.
7.6.1 The first theorem of welfare economics
The first theorem of welfare economics says that the competitive equilibrium is Pareto
efficient. In other words, the process of decentralised decision-making by agents and the
equilibrating force of market prices together ensure that there are no gains from trade
left unexploited. Starting from any equilibrium allocation, if we want to make any one
agent better off, we can do that only by making someone else worse off.
The first theorem of welfare economics applies in a very general setting, and is an
absolute blockbuster. This remarkable result is fundamental to our understanding of
markets.
The first thing it tells us is about the remarkable ability of the price mechanism to
eliminate inefficiencies. Even though markets are a conglomeration of a large number of
agents trying to maximise their own utility without paying attention to what others
might be doing, the price mechanism creates the right incentives for agents to exploit
every trading opportunity. Wherever any opportunity for efficiency-enhancing trade is
left, prices will be such that agents will have the incentive to exploit these
opportunities. This is the invisible hand at work the market price mechanism ensures
that each agent acting in pure self-interest results in a Pareto optimal outcome.
You should be able to show the result easily in a two-agent two-good exchange
economy. Looking at Edgeworth box diagrams, you should be able to see that it is
impossible to make anyone better off without making the other agent worse off. Let us
try to outline an informal proof. Suppose we have an equilibrium that is not Pareto
efficient. Then there must be some other allocation inside the Edgeworth box that
makes both agents better off. Since each agent is choosing the best point on the price
line (which is the budget line for both agents), to make agent A better off we must
choose an allocation that lies to the right of the price line, and to make B better off we
need to choose an allocation to the left of the price line. But there is no point common
to the set of points to the right and left of the price line. Therefore, it is impossible to
120
7.6. Welfare theorems
make both agents better off (or make one agent better off and the other left indifferent).
It follows that the equilibrium must be Pareto efficient.
From this it follows that a set of competitive markets functions very well using this
mechanism, and there is no case for efficiency-enhancing government intervention. The
invisible hand works well.
7.6.2 Implications
While this theorem is fundamental to our understanding of economics, you should also
be aware that the argument can be taken too far. The doctrine of laissez faire
advocates that the state should not intervene in markets. The doctrine has influenced
the economic stance of many conservative political parties across the world over the last
few centuries, who typically argue for a smaller state.
To see the flaw in this doctrine, you need to understand the concept of a market
failure. A market failure occurs when the equilibrium in a market involves some
deadweight loss, i.e. the equilibrium admits some degree of inefficiency. The theory of
general equilibrium that we described above assumes away such problems, but these are
relevant in many situations, as we describe below. Indeed, from this point onwards, the
rest of the course considers topics where such failures do arise. In such cases, we can
make a case for the visible hand of the government, intervening through policies to try
to reduce or eliminate the market inefficiency. Of course, if such failures arise only in
rare cases or only in markets that have a small role to play in the overall economy, we
would still conclude that there is little role for policy and the laissez faire type
arguments are largely correct.
In fact, market failures are a central feature of large and important parts of the
economy. Let us try to understand why. For market failures to arise, there must be
some factor impeding the functioning of competitive markets or the markets are not
competitive to start with.
Consider, first, the problem of public goods. These are goods that we consume together
and consumption is neither rival nor excludable. As you should know, in this case
competitive markets work poorly. If we try to use private provision through competitive
markets, it results in under-provision. However, public goods such as national defence,
roads, basic public health, basic education, defence against natural catastrophes etc. are
important for any economy. Provision of such goods using receipts from taxes is a
crucial role for the government in the economy.
Second, in some situations one agents actions affect other agents directly. But the agent
does not take this externality into account when acting. The result is that the market
outcome is inefficient. The individually-optimal levels of negative externality-generating
activities are higher than the socially optimal level, and that for positive
externality-generating activities are lower than the socially optimal level. The invisible
hand does not work well in such cases individual optimisation does not lead to social
optimality. Are these problems important? Indeed, some of our biggest problems are
caused by negative externalities. These include global warming, pollution, congestion,
urban noise, and so on. In all these cases, government policy has a crucial role to play
to bring individual incentives in line with the social optimum.
Third, the market itself might not be made up of a large number of price-taking firms.
121
If only a few firms (or just one firm) are present, they act strategically, taking into
account their own impact on the market price as well as the reactions of other firms. In
this case the market outcome is typically inefficient. Of course, before you can think of
such situations as not conforming to the first welfare theorem, you must investigate the
source of market power enjoyed by firms. If such power is intrinsic to the nature of the
products or technology, then indeed the first welfare theorem does not apply. In these
cases, regulation plays an important role in reducing the deadweight loss. If, on the
other hand, market power results from entry barriers created exogenously by policy
itself, then it is not a fundamental violation of the theorem. In such cases, you should
argue that the policies that create a distortion in the first place should be removed so
that competition and with it efficiency is restored.
Fourth, situations involving asymmetric information often result in market failures.
Such information problems are endemic to markets for information-sensitive financial
assets, insurance markets, as well as credit markets. In these cases, the role of policy is
less clear since policymakers themselves are likely to face the same information
problems. However, some of the problems can be reduced by building the right
institutions. Examples of such institutions include rules for stock trading that seek to
minimise information problems, adding deductibles or forcing people to belong to a
large insurance pool to reduce the information problems faced by an insurer, building a
long-term relationship with a creditor to reduce information problems in lending. The
primary role of government policy in such cases is to try to facilitate such institutions.
To summarise, the first welfare theorem is one of the most (if not the most) important
results in economics. It provides a benchmark against which to judge the social welfare
from any market outcome. There are various instances where market outcomes do
admit inefficiency, in which case there is a role for policy.
Finally, note that policy itself can be subject to problems. Governments themselves
might sometimes serve special interests, and in such cases interventions need not be
efficiency-enhancing. Along with market failures, you should also keep in mind the
possibility of such government failures.
7.6.3 The second theorem of welfare economics
The first welfare theorem says that any general equilibrium allocation is efficient. The
second theorem makes a sort of reverse statement. It says that if all agents have convex
preferences, any efficient allocation can be obtained as an equilibrium if we can suitably
change the initial endowment point.
This is easy to see in a two-agent two-good economy. As you know, efficient allocations
are points on the contract curve. Let us pick an arbitrary point H on the contract curve.
This is a point of tangency between indifference curves of the two agents, i.e. the
indifference curves have the same slope at H. Consider a price line that also has this
same slope passing through H. This is shown in Figure 7.7 below. If agents are faced
with this price line, which point would they choose to consume at? So long as
preferences are convex (diminishing MRS), both agents would choose H as the
utility-maximising point. Therefore, if we can change the endowment from the point E
(as shown in Figure 7.7) to H, the fact that H is a point of exchange equilibrium implies
that the market process would not move away from H. To summarise, starting from any
122
7.6. Welfare theorems
point H on the contract curve, we have shown that there is a price line which supports
H as a competitive equilibrium if we change the initial allocation suitably.
Figure 7.7: The second welfare theorem.
7.6.4 Implications
What does this mean? To see this, suppose we want a particular distribution of welfare
among agents. Can we achieve this distribution without wasting resources? That is, can
we reallocate resources in an efficient manner? The theorem shows that the answer to
this question is yes.
We noted that a key assumption needed for the theorem is that preferences are convex.
Figure 7.8 below shows why this is the case. As the figure shows, given non-convex
preferences, it might not be possible to find a price line to support a Pareto efficient
allocation as a competitive equilibrium. H is a Pareto efficient allocation, but on the
price line that has the same slope as the common slope of the indifference curves at H,
the optimal choice of agent A is not H, but G. Therefore, H is a Pareto efficient
allocation but it cannot be supported as an equilibrium. In other words, if the
government wanted to achieve a redistribution to the point H and redistributed
endowments suitably, the economy would not stay at point H as the market process
would move it away.
Next, consider the question of the redistribution of resources. How do we do this? In an
ideal world we would use lump-sum taxes and transfers. Indeed, if we could use
lump-sum taxes quite easily in practice, this theorem would spell out the right type of
policy to achieve different welfare distributions.
However, lump-sum taxes are difficult to impose in practice. A lump-sum tax is a tax
123
Figure 7.8: The necessity of convex preferences
imposed on something that an agent cannot change. A tax on labour endowment or any
type of earned income is distortionary any such tax would lead agents to change how
much labour they supply or how much effort they invest in earning. Furthermore, even
if you could find a lump-sum tax (such as a poll tax), people might not fully trust the
government to make use of the proceeds in the right manner. This might make such a
tax politically unpalatable. These problems with lump-sum taxes limit the force of the
second welfare theorem as a tool for the real world.
7.7 Production
The discussion so far has been in the context of an exchange economy. N&S Sections
10.3 and 10.4 introduce a simple model of general equilibrium with production. You
should read this carefully, and understand the informal proof of the first welfare
theorem (in Section 10.4) in such an economy.
Consider an economy with two goods (x and y) and one agent in society. This sort of an
economy is often referred to as a Robinson Crusoe economy. In an economy with many
identical agents you can simply solve the problem of a representative agent. This is
why analysing single-agent economies makes sense. Single-agent economies are used
often by macroeconomists in building a micro-foundation for macroeconomic models.
The utility function of Robinson is given by u(x, y).
The production possibility frontier shows the combinations of x and y that can be
produced using the resources available in the economy. Figure 7.9 shows such a frontier.
Note that the slope of the frontier at any point shows how much y we need to give up to
124
7.7. Production
obtain one extra unit of x. This is called the marginal rate of transformation (MRT).
Typically, the MRT is increasing as we raise x (and lower y). One way to interpret this
is as follows. Some resources in the economy are better at producing x while other
resources are better at producing y (as an example, think of the goods as guns and
butter and the inputs as iron and milk). At low values of x, extra x can be added by
giving up very little y because we are releasing resources that are very good at
producing x but not so good at producing y. As we keep increasing the production of x,
the amount of y we need to give up grows larger and larger as we release resources that
are less and less useful in producing x.
Figure 7.9: A production possibility frontier. This shows the combinations of x and y that
can be produced using the resources available in an economy. All points on the frontier
are technically efficient in the sense that all resources are employed in production.
Let us write the societys production possibility frontier in the implicitform

T (x, y) = 0. (For example, if the production possibility frontier is y = c x2 , we can
write it in the implicit form as x2 + y 2 c = 0.) The slope of the production possibility
frontier (with good y on the y-axis and good x on the x-axis) is given by:
T /x
.
T /y
The marginal rate of transformation is the absolute value of the slope:
T /x
MRT = .
T /y
We are now ready to consider the optimisation problem faced by Robinson. Robinsons
problem is to maximise utility subject to the production constraint:
max u(x, y) subject to T (x, y) = 0.

x, y
125
Set up the Lagrangian:

L = u(x, y) T (x, y).
Derive the first-order conditions:
L u T
= =0
x x x
L u T
= =0
y y y
L
= T (x, y) = 0.

From the first two first-order conditions, we have:
u/x T /x
=
u/y T /y
i.e. the MRS is equal to the MRT along the production possibility frontier.
Both the MRS and the MRT must be equal to the price ratio px /py in equilibrium.
Example 7.3 There are two goods, X and Y , in an economy. The production
function is given by:
X 2 + Y 2 = 200.
Each individual has the utility function:
u(X, Y ) = X 0.5 Y 0.5 .
Calculate the general equilibrium price ratio and the amounts of X and Y produced
in equilibrium.
The MRT is given by X/Y . The profit-maximising condition is:
X pX
= .
Y pY
The utility-maximising condition is MRS = pX /pY or:
Y pX
= .
X pY
Since MRS = MRT, we have:
X Y
=
Y X
or X = Y . Using this in the production function, X = Y = 10. Furthermore, the
equilibrium price ratio is pX /pY = 1.

to:
126
construct an Edgeworth box and depict the contract curve, the set of possible
equilibrium exchanges given an endowment point, and depict the equilibrium price
ratio and allocation
derive algebraically the contract curve, the equilibrium price ratio and allocation
explain that only relative prices matter
explain the importance of establishing existence
state and provide an informal proof outline for the first welfare theorem in the
context of an exchange economy with two agents and two goods as well as in the
context of an economy with production
discuss the policy implications of the first welfare theorem and understand the
state and provide an informal proof outline for the second welfare theorem in the
context of an exchange economy with two agents and two goods
discuss the policy implications of the second welfare theorem and understand the
derive the equilibrium conditions in an economy with production.

1. Consider a given allocation of resources that is not Pareto efficient. Not all Pareto
efficient allocations in the economy Pareto dominate this allocation. Is this true or
false? Explain.
2. Consider a pure exchange economy with two agents, Ann and Bob, who consume
only two goods, x and y. In the economy there are 4 units of x and 2 units of y.
Ann is endowed with 1 unit of x and 2 units of y, and Bob is endowed with 3 units
of x and 0 units of y. Their utility functions are:

uA (xA , yA ) = xA yA and uB (xB , yB ) = xB + 2yB .
(a) Draw the Edgeworth box, showing the endowment point, and the indifference
curves.
(b) Derive the contract curve.
(c) Derive the demand functions of the two agents.
(d) Calculate the exchange equilibrium price ratio.
(e) Calculate the exchange equilibrium allocation.
127
3. The following questions are about the first welfare theorem.

(a) State the first welfare theorem.
(b) Provide an outline of the proof in the context of an exchange economy with
two agents and two goods.
(c) Discuss an outline of the proof in an economy with production.
(d) Discuss the implications of this theorem.
(e) In what situations might the theorem fail? Discuss.
4. The following questions are about the second welfare theorem.

(a) State the second welfare theorem.
(b) Discuss an outline of the proof in the context of an exchange economy with
two agents and two goods.
(c) Explain why the theorem requires preferences to be convex.
(d) Discuss the policy implications of this theorem.
(e) Discuss the practical problems with the implications of the theorem.
128
Chapter 8
Monopoly
8.1 Introduction
Previously, we analysed competitive markets. In these markets firms are price-takers,
and the market price is determined by the interaction of industry supply and market
demand, a process over which no single firm has any control. In this chapter we develop
the model of the opposite market structure one where the supply side consists of a
single firm. The other dimension in which the analysis differs is that a monopolist is
relatively unconstrained in choosing its pricing strategy. Instead of choosing a single
price, it can choose different prices across different identifiable groups or markets, and
also charge a fee to give a consumer access to purchasing. We analyse the consequences
of these pricing schemes. Finally, certain markets naturally lend themselves to a
monopoly structure. Such natural monopolies also present unique regulatory challenges.
We analyse the problem of natural monopolies and explain the regulatory problems.

We aim to explain how profit maximisation by a monopoly firm leads to a distortion in
the market, and try to understand the policies that try to address this problem. We also
aim to analyse different price discrimination schemes and try to understand what
impact they have on profits and efficiency loss. Overall, our main focus here is on the
design of pricing schemes as well as the design of regulatory schemes to eliminate or
limit any deadweight loss.

derive the marginal revenue curve associated with any demand curve
explain the relationship between marginal revenue and price elasticity of demand as
well as Lerners index
analyse the profit-maximising choice by the monopolist and the associated

deadweight loss
analyse the role of regulatory policies
analyse a variety of price discrimination policies that a monopolist can employ
analyse the structure of natural monopoly and the regulatory problem in this case.
129
8. Monopoly

Sections 8.3 and 8.4, and Chapter 11 of N&S. The coverage in this textbook is not
enough on its own. The subject guide develops some aspects more formally. Perloff
Sections 12.4 and 12.5 develop non-linear pricing and two-part tariffs formally.

The chapter analyses the profit-maximising outcome for a monopolist and the resulting
deadweight loss. Next, the chapter analyses a variety of price discrimination schemes.
Finally, the chapter introduces the problem of natural monopoly.
8.3 Properties of marginal revenue

You should study carefully the properties of marginal revenue (MR) detailed in N&S
Sections 8.3 and 8.4. You should understand how to derive the MR curve for a
downward-sloping demand curve (in this case, the average revenue is falling which
implies marginal revenue is below average revenue).
You should also understand the relationship between MR and price elasticity. This can
be derived as follows:

d dP Q dP 1
MR = (P (Q) Q) = Q+P =P 1+ =P 1+
dQ dQ P dQ
where < 0. To make this fact explicit, it might be useful to rewrite this using the
absolute value of price elasticity:

1
MR = P 1 .
||
We have the following:
If < 1 (i.e. || > 1), demand is elastic and MR > 0.
If = 1 (i.e. || = 1), demand is unit elastic and MR = 0.
If > 1 (i.e. || < 1), demand is inelastic and MR < 0.
Figure 8.1 below shows the relationship between elasticity and MR for the linear
demand curve P = 10 Q.
8.4 Profit maximisation and deadweight loss

You should know how to derive the profit-maximising solution for a monopolist and
how to depict this in a diagram. You should be able to calculate the deadweight loss at
the monopoly optimum.
130
8.4. Profit maximisation and deadweight loss
Figure 8.1: Elasticity and marginal revenue for a linear demand curve.
Figure 8.2: Profit maximisation and deadweight loss under monopoly.
A monopolists profit is T R T C, and this is maximised at the quantity where

MR = MC. Figure 8.2 shows the profit-maximising quantity and price. Note that the
profit at the optimum is given by (P AC(Q )) Q .
Figure 8.2 also shows the deadweight loss under monopoly. Note that QC is the
competitive output where P = MC. What is the source of the deadweight loss under
monopoly? Essentially, a monopolist tries to maximise profit by restricting output. This
causes the output to be lower than the competitive output at which there would be no
deadweight loss.
131
8. Monopoly
Profit maximisation under a linear demand curve
Consider the linear demand curve shown in Figure 8.1. Note that at MR = 0 we have
|| = 1. Therefore, for any positive MC, the optimal output-price combination occurs at
the left of the unit elasticity point. At all such points demand is elastic. It follows that
for a linear demand curve, the monopolist always produces at the elastic part of the
curve.
Regulation
Since a monopoly gives rise to an efficiency loss, the question of regulation becomes
relevant. You should be able to understand and analyse regulatory strategies to reduce
monopoly deadweight loss. One way to regulate a monopoly is to specify the price that
a monopolist can charge. By restricting price to the level P = MC(QC ), the regulator
can eliminate deadweight loss. Another policy is to encourage new entry in the industry
so that there is increased competition, which reduces the price towards marginal cost
and reduces deadweight loss.
The Lerner index
Since profit is maximised at MR = MC, at the optimum point:

1
MC = P 1+

which implies:
P MC 1 1
= = .
P ||
The left-hand side is known as the Lerner index, which is a measure of the monopoly
power of a firm. For a competitive firm, P = MC so that the Lerner index is 0. If MC is
0 or P is infinity, the Lerner index is 1, which is its maximum value. For a monopolist,
it lies between 0 and 1 and rises as demand becomes less elastic.
8.5 Price discrimination
A monopoly firm need not charge a single price to all customers. There are a variety of
ways in which it can practise price discrimination. N&S Section 11.4 discusses perfect
price discrimination, market separation, two-part tariffs, as well as a variety of
non-linear pricing schemes. A durable good monopolist faces competition with its own
future self, which affects pricing. See N&S for details.
In the activity below we develop an example of market separation as well as a two-part
tariff.
132
8.5. Price discrimination
Example 8.1 A monopolist has two customers with the following demand
functions:
Q1 = 10 P1 (Demand of customer 1)
Q2 = 12 P2 (Demand of customer 2).
Here Pi is the price charged to customer i, i {1, 2}. The monopolist has a constant
marginal cost of 1, and no fixed costs.
(a) Suppose the monopolist can differentiate between customers, and customers
cannot trade between themselves, allowing the monopolist to engage in
third-degree price discrimination. What is the price charged to each consumer?
(b) Now suppose the monopolist cannot differentiate between customers and must
charge them the same price. Calculate the monopolists optimal single price P
as well as the quantity sold to each customer.
(c) Suppose the monopolist cannot differentiate between customers. However, in
addition to a per-unit price P , the monopolist can also charge a fixed fee F . A
customer must pay this fee irrespective of the quantity purchased when a
positive amount is purchased. Derive the monopolists optimal price and fee.
(a) With third-degree price discrimination, the monopolist sets MR = MC for each
customer. You should be able to show that Q1 = 4, P1 = 6, Q2 = 6 and P2 = 8.
The monopolists profit is 52 and the total surplus is 78.
(b) In this case the monopolist sets a single price to maximise profit given total
demand, which is Q = 24 2P . You should be able to show that in this case
Q = 10 and P = 7. The profit is 50 and the total surplus is 79.
(c) At any price P < 10, if the fixed fee is set equal to the surplus of consumer 1
(given by (10 P )2 /2), the total demand is Q1 + Q2 = 24 2P , and the
revenue is:
(10 P )2
2 + (P 2)(24 2P ).
2
Maximising, the first-order condition is:
2(10 P ) + 24 2P 2(P 2) = 0
which implies P = 4. The fixed fee is then 18, and the revenue is 36 + 32 = 68.
If the fixed fee is set equal to the surplus of consumer 2, then consumer 1 does
not buy. The monopolist is just dealing with consumer 2. In this case, since the
monopolist can set a fixed fee to extract all consumer surplus, it is best to
maximise the surplus by setting price equal to marginal cost, and then
extracting the full surplus.
Therefore, the optimal price is P = 2 and the fixed fee is:
(14 2)2
F = = 72.
2
The profit of the monopolist is 72, which is higher than the previous cases. The
total surplus in this case is also 72, which is lower than the total surplus in (a)
and (b).
133
8. Monopoly
8.6 Natural monopoly

In some industries, there is a large fixed cost but a relatively low marginal cost of
production. For example, it is costly to set up a gas, electricity, water or telephone
network, but once this fixed cost (installing pipes, cables etc.) is incurred, the cost of
connecting an additional subscriber is close to zero. In such cases, the long-run average
cost curve falls over the relevant range of output. In such cases, it is efficient to have a
single firm because multiple firms would only replicate the fixed cost unnecessarily. In
other words, a single firm can produce the market output at a lower cost compared to
several firms. A monopoly with this characteristic is called a natural monopoly.
Utilities are often granted such status by governments because they are thought to be
natural monopolies. Figure 8.3 shows the situation faced by a natural monopoly.
Figure 8.3: A natural monopoly with low fixed MC and falling AC. Note that the
monopolist would incur a loss if under marginal cost pricing. If regulated price is set
at AC, the shaded area is the deadweight loss. This is, of course, much lower than the
deadweight loss without regulation.
A standard monopolist can be regulated by setting the regulated price equal to marginal
cost at the efficient quantity. However, such an approach runs into difficulty here. Since
the AC is falling, the MC is below the AC, implying that if the regulated price is set
equal to the MC, the monopolist would incur a loss and quit rather than produce. As
you can see from Figure 8.3, pricing at the MC implies a loss for the monopolist. The
government then must give a lump-sum subsidy to the monopolist to cover the losses.
However, governments typically do not want to be in a position to subsidise a
monopolist forever. In that case, the government can regulate the monopolist by setting
the price equal to average cost. This implies that the deadweight loss is not fully
eliminated, as shown in Figure 8.3. Alternatively, the monopolist might be allowed to
charge a higher price to some consumers and a lower price to others. A multi-product
monopolist might be allowed to make a profit on some products while making a loss on
134
others through marginal cost pricing, while making a zero profit overall.

to:
derive the marginal revenue curve associated with any demand curve
explain the relationship between marginal revenue and price elasticity of demand as
well as Lerners index
analyse the profit-maximising choice by the monopolist and the associated

deadweight loss
analyse the role of regulatory policies
analyse a variety of price discrimination policies that a monopolist can employ
analyse the structure of natural monopoly and the regulatory problem in this case.

1. A firm that delivers Q units of water to households has a total cost of
C(Q) = mQ + F . If any entrant would have the same cost, does this market satisfy
the conditions to be a natural monopoly?
2. Suppose the cost function of a monopolist is C(Q) = 10Q 2Q2 + 3Q3 . Derive the
supply function of the monopolist.
3. Suppose a firm has cost function ln Q + 1, where Q > 1. Suppose the government
wants to regulate the firm and set the regulated price equal to the MC. Propose a
policy that achieves this goal.
4. Suppose that a monopoly producer of widgets has a constant marginal cost of 6

and sells its products in two separate markets whose inverse demand functions are:
P1 = 24 Q1 and P2 = 12 0.5Q2 .
(a) Calculate the profit-maximising price-quantity combinations in these two

markets.
(b) Calculate the monopolists total profit.
(c) Calculate the deadweight losses in the two markets.
135
8. Monopoly
(d) Now suppose regulation forces the monopolist to charge a single price.
Calculate the profit and deadweight loss in this case. Would both markets be
served? Does the regulation reduce the deadweight loss?
5. Suppose a monopolist can sell to two different customers with demand curves:
Q1 = 24 P1
Q2 = 24 2P2 .
The monopolist has a constant marginal cost of 6. Calculate the optimal two-part
tariff.
136
Chapter 9
Oligopoly
9.1 Introduction
When an industry comprises of a few firms, the nature of competition is very different
from that under perfect competition. In trying to maximise profit, a firm must not only
take into account the impact of its own supply on the market price, but must also
account for the strategic responses of other firms. We have already analysed this type of
interaction when we studied game theory, and identified Nash equilibrium as the central
solution concept. Recall that a Nash equilibrium is a mutual best response: given what
others are doing, each firm must play a best response. We now use this solution concept
to analyse the outcome under quantity and price competition.

In this chapter we aim to analyse the market structure when there are a few firms in
the market. We aim to analyse the nature of strategic interaction among firms when
firms compete in quantities as well as under price competition. We aim to understand
the oligopoly models of Cournot competition with and without collusion, Stackelberg
leadership, Bertrand competition with and without collusion, as well as price
competition with differentiated products under simultaneous and sequential
price-setting.

derive the Nash equilibrium in a Cournot game
analyse collusion in a Cournot game
analyse Stackelberg leadership
derive the Nash equilibrium in a Bertrand game
analyse collusion in a Bertrand game
derive the Nash equilibrium in a Bertrand game with differentiated products
derive the equilibrium in the game of sequential price-setting with differentiated

products.
137
9. Oligopoly

N&S Chapter 12, Sections 12.1 to 12.5. In Section 12.2 you can ignore the subpart on
Capacity Choice and Cournot Equilibrium. In Section 12.3 you mainly need to read
the subpart on Bertrand Model with Differentiated Products. In Section 12.5 you
mainly need to read the subpart on First-mover Advantages. A cursory reading of
other subparts of Sections 12.3 and 12.5 will suffice. In the subject guide we provide a
more complete algebraic analysis. Also, the case of sequential price-setting with
differentiated products is not covered in the textbook you must rely on the subject
guide for this case.

In this chapter, we consider models of Cournot competition, Bertrand competition, and
price competition with differentiated products. In these cases the solution concept is
Nash equilibrium. We also consider two models where firms move sequentially: the
Stackelberg model of quantity competition and sequential price-setting with
differentiated products. In these cases the appropriate solution concept is subgame
perfect Nash equilibrium. Finally, we consider the possibility of repeated interaction and
use the analysis of equilibria in repeated games to understand collusion among firms.
9.3 Cournot competition

See N&S Section 12.2 for a discussion. Here we provide a more complete algebraic
analysis. For Stackelberg leadership, see the subpart on First-mover Advantages in
Section 12.5.
This is a game with continuous (rather than discrete) strategy sets. The inverse demand
function in an industry is given by p = a bQ, where p is the market price, Q is the
aggregate supply in the market, and a, b are positive constants. There are n > 1 firms in
this industry, and each firm produces the output at a marginal cost c, with c < a.
First, suppose n = 2. Let us compute the Nash equilibrium of this game, which is also
called the Cournot equilibrium, or the CournotNash equilibrium.
The profit of firm 1 is:
1 (q1 , q2 ) = (p c)q1 = (a bQ c)q1 = (a bq1 bq2 c)q1 .
The first-order condition for maximising profit is 1 /q1 = 0 which implies:
a c q2
q1 = .
2b 2
This is the best response function of firm 1 (also called the reaction function of firm
1).
After optimisation, we impose symmetry: q1 = q2 = q. This implies:
a c bq
q= .
2b
138
9.3. Cournot competition
Solving:
ac
q = .
3b
The equilibrium price is:
ac 3a 2a 2c 1 2
p = a 2b = + = a + c.
3b 3 3 3 3 3
Let D denote the profit of either firm under duopoly. This is given by (p c)q .
Therefore:
(a c)2

D 1 2 ac (a c) a c
= a+ cc = = .
3 3 3b 3 3b 9b
The best response functions and the CournotNash equilibrium are shown in Figure 9.1
below.
Figure 9.1: CournotNash equilibrium.
Here we imposed symmetry after optimisation. However, this method does not work if
the firms are not symmetric. Suppose the marginal costs are c1 and c2 . In this case, you
must derive the two best response functions, and then solve two simultaneous equations
for q1 and q2 .
The best response functions are:
a c1 q 2
q1 =
2b 2
and:
a c2 q 1
q2 = .
2b 2
Using the value of q2 in the best response of firm 1, we have:
a c1 a c2 q 1
q1 = + .
2b 4b 4
139
9. Oligopoly
Simplifying:
3q1 a 2c1 + c2
= .
4 4b
This gives us:
a 2c1 + c2
q1 = .
3b
Using this in the best response of firm 2, we have:
a c2 a 2c1 + c2 a 2c2 + c1
q2 = = .
2b 6b 3b
Note that an increase in c1 reduces the equilibrium output of firm 1, but raises that of
firm 2. Graphically, the best response function of firm 1 moves inwards, so that the
Nash equilibrium point slides up along the best response function of firm 2.
9.3.1 Collusion
We now return to the symmetric firms model. Suppose the firms could collude by some
means. Let us calculate the output, price and profit in this case.
Let QM = q1 + q2 be the collusive output. The joint profit is:
M (QM ) = (a bQM c)QM .
The first-order condition for a maximum is M /QM = 0 which implies:
ac
QM = .
2b
Note that QM is lower than total output under Cournot. Figure 9.1 in the previous
section shows the collusive output line, which joins the monopoly outputs of the two
firms.
The market price is:
ac a+c
pM = a b = .
2b 2
The total profit is:
(a c)2

M a+c ac
= c = > 2D .
2 2b 4b
The monopoly outcome cannot be sustained in a one-shot game the best response of
firm 1 to any q2 is to produce more than the collusive output. This is clear from Figure
9.1 in the previous section.
However, in an infinitely-repeated game, the collusive output can be sustained as the
outcome of an equilibrium if the players are sufficiently patient. You already know this
result from the theory of repeated games. Let us do some calculations for this specific
case to reinforce the point.
Suppose two symmetric firms engage in an infinitely-repeated game where in each stage
they play a Cournot game. Suppose each firm produces QM /2 under collusion. Each
firm has a discount factor , where 0 < < 1.
Now, suppose each firm adopts the following trigger strategy. Start by producing
QM /2 at t = 0. In any subsequent period t > 1, do as follows:
140
9.3. Cournot competition
If in period t 1 each firm produced QM /2, produce QM /2 in period t.
If in period t 1 any single firm deviated, produce the CournotNash quantity

q = (a c)/3b in period t and in all future periods.
Such a strategy is called a trigger strategy since a deviation by any single firm triggers
non-cooperation in all future periods. We denote the total duopoly CournotNash
equilibrium quantity by QD , i.e. QD = q1 + q2 .
Now, the best possible deviation by a firm is to produce the one-period best response to
QM /2. Let us call this QDev . You can easily work out the precise figure from the best
response functions. Let i (x) denote the profit of firm i from producing quantity x.
If firm i deviates in any period, it gets i (QDev ) that period, but subsequently each firm
produces the Cournot quantity so that the payoff in each subsequent period is i (QD ).
So the total payoff starting from the deviation period is:

i (QDev ) + i (QD ) + 2 i (QD ) + = i (QDev ) + i (QD ).
1
If firm i had not deviated, its payoff over the same periods would be:
1
i (QM /2) + i (QM /2) + 2 i (QM /2) + = i (QM /2).
1
Therefore, for deviation to be unprofitable, we need:
1
i (QM /2) > i (QDev ) + i (QD ).
1 1
Simplifying:
i (QM /2) > (1 )i (QDev ) + i (QD )
which gives us:
i (QDev ) i (QM /2)
> .
i (QDev ) i (QD )
Since i (QM /2) > i (QD ) (the payoff of each firm under collusion is necessarily higher
than that under Cournot competition), the expression on the right-hand side is less
than 1. Therefore, for high enough (close enough to 1), the above inequality is
satisfied for all i, and for such values of , collusion can be sustained.
9.3.2 Cournot with n > 2 firms

Let us now consider n > 2 firms. The market demand function is:
p = a b(q1 + + qn ) = a b(qi + Qi ).
The profit of firm i is:
i (qi , Qi ) = (a b(qi + Qi ) c)qi .
The first-order condition for maximising profit is /qi = 0, which implies:

a bQi c
qi = .
2b
141
9. Oligopoly
Impose symmetry after optimisation: q = q1 = = qn . This implies Qi = (n 1)q.

Therefore:
a (n 1)bq c
q=
2b
which implies:
ac
q = .
b(n + 1)
The total output is:
n(a c)
Q= .
b(n + 1)
Note that as n , we have:
ac
Q
b
and:
ac
pab = c.
b
Therefore, the Cournot outcome converges to the competitive outcome as n goes to
infinity.
9.3.3 Stackelberg leadership

Consider again two symmetric firms.
Instead of being a simultaneous-move game, if firms play an extensive-form game with
one firm choosing quantity before another, the firm choosing first is the Stackelberg
leader, while the second mover is the follower.
What is the subgame perfect Nash equilibrium in this case? The second mover will
choose a best response to the first movers quantity. Therefore, the second mover will
always be on his best response function. Knowing this, at the first stage, the first mover
simply chooses the best point on the second movers best response function.
Suppose firm 1 moves first, followed by firm 2. Then the best response of firm 2 is as in
the Cournot case:
a c q1
q2 = .
2b 2
This is the equilibrium in the subgame. Knowing this, at the initial stage, firm 1
maximises:
1 (q1 , q2 ) = (a bq1 bq2 c)q1

a c bq1
= a bq1 + c q1
2 2

a c bq1
= q1 .
2 2
Maximising:
ac
q1 = .
2b
Using this:
ac
q2 = .
4b
142
9.4. Bertrand competition
Note that the second mover produces half the output of the first mover. Compared to
the CournotNash outputs, the first mover produces more and, given this, the second
movers best response is to produce a lower output. The total output is:
3 (a c)
Q =
4 b
which is greater than the total output under Cournot competition. Consequently, the
market price:
1 3
p = a + c
4 4
is lower than that under Cournot competition. The profit of the Stackelberg leader (firm
1 in this case) is:
ac (a c)2
SL = (p c) =
2b 8b
which is higher than the profit under Cournot competition. This is, of course,
unsurprising. The Stackelberg leader could choose any point on the followers best
response function. If the leader chooses the CournotNash quantity, the followers best
response would also be to choose the CournotNash quantity. However, the leader can
do even better given the freedom to choose any point on the followers best response
function. Therefore, the Stackelberg leader must earn a greater profit compared to that
under Cournot competition. The follower, on the other hand, earns a lower profit given
by:
S (a c)2
F = .
16b
9.4 Bertrand competition

See the subparts of N&S Section 12.2 on Bertrand competition. You can ignore the
subpart on Capacity Choice and Cournot Equilibrium.
If firms compete in prices, and each firm can serve the whole market, the only
equilibrium is p = c. You should be able to go through the detailed argument to
establish this. Bertrand competition results in the competitive outcome, a result
sometimes referred to as the Bertrand paradox.
9.4.1 Collusion
If firms play an infinitely-repeated game, and are patient enough, we know from game
theory that they can sustain a high collusive price.
9.5 Bertrand competition with product differentiation

See Section 12.3 of N&S. Here we consider the following problem.
Two firms have differentiated products so that the demand functions are as follows:
q 1 = P 1 + P2
143
9. Oligopoly
and:
q 2 = P2 + P1 .
Why is this a model of differentiated products? Note that the demand for the product
of firm 1, say, does not vanish if P1 > P2 . In this case, the demand for firm 2s product
will be higher, but firm 2 will not take over the entire market. The underlying idea here
is that firms have different products that are substitutable, but not completely so. This
could be because the products serve the same basic purpose but have slightly different
features so that some people like firm 1s product better while others prefer firm 2s
product. As firm 1s price increases, some of the customers (who only consider firm 1s
product to be marginally better than firm 2s) will switch to buying firm 2s product
lowering demand for firm 1 and raising that for firm 2. Only at very high levels of P1 (in
the above example, this would require P1 > P2 + ) would q1 go to zero.
Let us now calculate the Nash equilibrium. The profit of firm 1 is given by:
1 = (P1 c)( P1 + P2 ).
The first-order condition is:
( P1 + P2 ) (P1 c) = 0.
It follows that 2P1 = + P2 + c, which implies:

+ c + P2
P1 = .
2
This is the best response function of firm 1.
Impose symmetry at this point: the solution involves P1 = P2 = P . Therefore:
P = + c.
Each firm charging a price of + c is the Nash equilibrium. The profit of each firm is 2 .
Figure 9.2 shows the shape of the reaction functions. Note that the reaction functions
slope upwards in this case because an increase in P1 , say, allows firm 2 to raise its price
optimally.
9.5.1 Sequential pricing

In the Stackelberg model, we studied the outcome under quantity competition when
firms move sequentially. Here, let us study the same market structure with pricing.
Suppose firm 1 moves first, followed by firm 2. In the second stage, firm 2 will always be
on its best response function. Therefore, firm 1 chooses the best point on firm 2s best
response function in the first stage. Let us calculate this subgame perfect Nash
equilibrium.
Firm 1 maximises:

+ c + P1
1 = (P1 c)q1 = (P1 c) P1 +
2
144
9.5. Bertrand competition with product differentiation
Figure 9.2: The figure plots the best response functions assuming the values = 3 and
c = 1. In this case the Nash equilibrium is P1 = P2 = 4.
Maximising, we get:
3
P1 = c + .
2
Using this in the best response function of firm 2, we have:
5
P2 = c + .
4
Which firm has a higher profit the leader or the follower?
92

3 3 5
1 = (P1 c)( P1 + P2 ) = + =
2 2 4 8
while:
252
2 = .
16
Note that 2 > 1 .

In the Stackelberg equilibrium, the leader gets the higher profit. Here, the follower
benefits more than the leader. The reason is that higher prices set by firm 1 at stage 1
are advantageous to firm 2, as a higher price set by firm 1 means that the environment
faced by firm 2 is less competitive. Note that the price charged by firm 1 is higher than
that under the simultaneous-move game. This means that in the sequential-move game
firm 2 faces a less competitive environment compared to the simultaneous-move game.
145
9. Oligopoly

to:
derive the Nash equilibrium in a Cournot game
analyse collusion in a Cournot game
analyse Stackelberg leadership
derive the Nash equilibrium in a Bertrand game
analyse collusion in a Bertrand game
derive the Nash equilibrium in a Bertrand game with differentiated products
derive the equilibrium in the game of sequential price-setting with differentiated

products.
1. Can a Stackelberg leader get a lower profit compared to the CournotNash

outcome?
2. Consider the Cournot duopoly in which two firms, 1 and 2, produce a homogeneous
good, and have total costs C1 (q) = 3q and C2 (q) = 2q, respectively. The market
demand is Q = 63 P . Find the equilibrium quantities, price and profits. Which
firm produces more? Which firm earns more?
3. The (inverse) market demand for a good is P = a bQ, and all firms have a
constant marginal cost c.
(a) Derive the perfect competition market quantity, price, industry profit,
consumer surplus and total surplus.
(b) Derive the monopoly market quantity, price, industry profit, consumer surplus
and total surplus.
(c) Assuming that there are two firms that compete by setting quantities
simultaneously, derive the Cournot duopoly market quantity, price, industry
profit, consumer surplus and total surplus.
(d) What can you say about the different market structures if you compare them?
146
4. Consider two firms that produce a homogeneous good and compete by setting
quantities. Suppose that firm 1 has a cost function C1 (q) = 100 + 5q, and that firm
2 has a cost function C2 (q) = 300 + 5q. The market demand is Q = 105 P .
(a) Assuming that they choose their quantities simultaneously, determine their
equilibrium quantities, price and profits.
(b) Assuming that firm 1 chooses its quantity first, determine the equilibrium
quantities, price and profits.
(c) Assuming that firm 2 chooses its quantity first, determine the equilibrium
quantities, price and profits.
5. Suppose two firms (firm 1 and firm 2) sell differentiated products and compete by
setting prices. The demand functions are:
P2
q 1 = 2 P1 +
2
and:
P1
q 2 = 2 P2 + .
2
Firms have a zero cost of production.
(a) Find the Nash equilibrium in the simultaneous-move game. Also find the
quantities sold by each firm and the profit of each firm.
(b) Find the subgame perfect Nash equilibrium if firm 1 moves before firm 2. Also
find the quantities sold by each firm and the profit of each firm.
147
9. Oligopoly
148
Chapter 10
Asymmetric information: adverse
selection
10.1 Introduction
For a wide range of products (for example, things you buy from a grocery store) the
markets are large and decentralised, and exchange is anonymous. There are, however,
important exceptions. Consider the market for used cars. It is difficult to tell how good
or bad a used car is unless you use it for some time. The seller, on the other hand,
knows how good the car is.
In such situations, you would care about the identity of the seller. In fact, in such cases
anonymous exchange is often impossible, as that would lead to a breakdown of the
market. The customers are willing to pay a price based on the average quality. However,
this price is not acceptable to the better-than-average quality car owner. This leads the
owners of better-than-average cars to leave the market. Knowing this, customers would
revise average quality downwards and consequently be willing to pay a lower price. But
the same phenomenon then applies again, leading to further falls in average quality and
willingness to pay, and so on.
The problem is called the lemons problem or adverse selection.
The same problem arises in selling insurance. Consider the problem of selling health
insurance to older people who know much more about their state of health than an
insurance company. If the insurance company sets a price based on the average medical
expenditure, this would typically lead to a market price (in this case, the insurance
premium) that is too high for the healthiest individuals and only the people who feel
that they are very likely to claim the insurance would buy insurance. This would lead,
as in the case of used cars above, the market to fail.
Banks (or other lenders) face the same problem in credit markets. If a high rate is set in
expectation of high average risk, safer borrowers will exit the market and only the
riskiest borrowers will want to borrow. Again, as in the market for lemons, this causes a
market failure.
A different sort of problem arises when you, as the manager of a company, are trying to
hire some salespeople. As the job requires a door-to-door sales campaign, you cannot
supervise them directly. And if the workers choose not to work very hard, they can
always blame it on the mood of the customers. If you pay them a fixed wage
independent of the sales they achieve, they are unlikely to work hard.
This problem is known as moral hazard.
Note the difference between adverse selection and moral hazard. In the first case, the
asymmetry in information exists before you enter into the exchange (buy a used car, or
149
10. Asymmetric information: adverse selection
sell insurance). In the latter, however, the asymmetry in information arises after the
wage contract is signed.
This is why another name for adverse selection is hidden information and another
name for moral hazard is hidden action.
In this chapter and the next, we will consider certain adverse selection and moral
hazard problems and remedies in some detail.

In analysing markets so far, we have assumed that consumers and firms either have all
relevant information about each other, or, if something is not known (the outcome of a
random event, say), all parties face the same uncertainty. However, in many important
markets, one party to a transaction is better informed compared to the other party.
This chapter and the next aim to introduce you to this class of problems. The analysis
in this chapter is concerned with one particular class of problems of this sort called
adverse selection. The chapter aims to introduce you to this class of problems and offer
some insights about potential solutions.

explain the different types of asymmetric information problems
explain how the scope of enquiry of economics is expanded by considering

information asymmetry problems
explain the types of problems that arise under adverse selection
analyse the problem of adverse selection in the market for lemons
analyse the problem of price discrimination by a monopolist under adverse selection
analyse separating and pooling equilibria in the signalling model of education.

N&S Chapter 15, Sections 15.3 to 15.5. In Section 15.4, exclude the first two topics
Moral hazard with several agents and Auctions. From Section 15.4, you only need to
study the last topic The Market for Lemons. Study also Applications 15.4 and 15.5. In
the subject guide we develop an example of a market for lemons with two qualities. We
also develop a formal model of the price discrimination problem, a variant of which is
studied informally in Section 15.3. You should read the textbook presentation of the
model in Section 15.3 to develop the intuition about adverse selection problems, but
you should study all formal details of the price discrimination model developed in the
subject guide.
150
10.2. Overview of the chapter

The chapter starts with a discussion of the scope of economic theory, then introduces
some concepts useful for the analysis in this chapter. The chapter then covers the first
model of adverse selection: Akerlofs model of the market for lemons. Next, the chapter
covers the second model of adverse selection: the problem of price discrimination when
consumer types are unknown. Finally, the chapter covers the third model of adverse
selection: the job market signalling model of Spence.
10.3 The scope of economic theory: a general

comment
As you study the topics under asymmetric information, try to see what this analysis,
combined with game theory, adds to economic theory. If you think about this question,
you should be able to see that many different types of transactions and institutional
arrangements cannot be understood by using the tools developed under the study of
competitive markets. For example, while there might be some unskilled labour markets
which look like a competitive labour market in which all parties are market wage-takers,
many other labour markets do not work this way. In many jobs, agents enter into
complicated contracts with their employers. Also, the form of remuneration often
deviates from a simple wage payment and includes things like performance bonuses and
equity options.
A model of a competitive labour market is simply not rich enough to explain these
institutional arrangements. Similarly, you could ask why insurance contracts carry
deductibles, why certain assets are traded in large liquid markets while others are
traded in illiquid markets by specialist traders, why the trading and pricing rules of an
exchange matter at all, why different firms have different organisations of management,
why and how to regulate banks.
The study of competitive markets and general equilibrium theory will not help you
much in answering these. Indeed, in a general equilibrium environment you would not
even be able to explain why banks exist, let alone why or how they need to be
regulated. However, with the tools of asymmetric information theory and game theory,
we can address all of these, and many other questions. We will address only a small
subset of these questions here, but as you read around these topics, try to see how vast
the scope of the enquiry is once you learn to wield these tools properly. Indeed, such
methods cover many types of questions across several fields of social sciences,
evolutionary biology and the design of tort law.
These tools have expanded the scope of economic enquiry so dramatically that it is
really a hopeless task to try to define it. Some textbooks still contain statements like
economics studies how scarce resources are allocated among competing uses. This is
indeed correct as a description of the study of general competitive equilibrium, but as a
definition of the whole of economic theory such statements are entirely out-of-date.
Modern tools have ensured that it is no longer possible to neatly define the scope of
economics.
151
10.4 Akerlofs (1970) model of the market for lemons
N&S briefly discusses this problem in the last part of Section 15.4. Perloff Section 18.3
contains a more detailed informal discussion of the model and policy implications. Here
we present an example with two qualities.
This is based on the famous article that gave rise to the field of asymmetric information:
George A. Akerlof (1970) The market for lemons: quality uncertainty and the market
mechanism, Quarterly Journal of Economics, volume 84, pages 488500.
Suppose there are four kinds of cars there are new cars and old cars, and in each of
these two categories there may be good cars and bad cars. Buyers of new cars purchase
them without knowing whether they are good or bad. After using a car for some time,
the owners can find out whether the car is good. This leads to an asymmetry: sellers of
used cars have more information about the quality of the specific car they offer for sale.
10.4.1 The market for lemons: an example with two qualities
Consider a market for used cars. There are some low quality cars and some high quality
cars. Potential sellers have a car each, and there are many more buyers than possible
sellers in the market. A high quality car rarely breaks down. A low quality car provides
a poorer ride quality over longer journeys and also breaks down with higher probability
compared to high quality cars.
A seller values a high quality car at 1,000 and a low quality car at 300. A buyer values a
high quality car at 1,300 and a low quality car at 400. All agents are risk-neutral.
Assume that the sellers get the entire surplus from trade.
Market failure
Suppose quality is observable to sellers but not to buyers. Buyers only know that a
fraction of 1/2 of the cars in the market are high quality and the rest are low quality.
Let us analyse the market outcome in this case.
The average value of buyers is 850. If all cars are in the market, this is the most buyers
would pay. But at this price high quality cars withdraw. So only low quality cars would
sell. Knowing this, buyers would be willing to pay at most 400. Assuming sellers get the
entire surplus, the market price is 400 and only low quality cars exchange hands. The
market outcome is not efficient since the gains from trading high quality cars are not
exploited.
Separation through refunds
Suppose low quality cars break down with probability 0.8, and high quality cars break
down with probability 0.1. Suppose the sellers have an option of promising a refund of
R if the car breaks down.
Let us see if such a policy can lead to a separating equilibrium. In a separating
equilibrium, it must be the case that sellers of high quality cars sell at 1,300 and
promise a refund of R in the event of a breakdown, while sellers of low quality cars sell
152
10.5. A model of price discrimination
at 400 without any refund promise.

For high quality sellers to sell at 1,300 with a refund promise, it must be that they like
this better than not selling, which means:
1300 0.1R > 1000
which implies R 6 3000. This constraint is called the participation constraint (PC) of
high quality sellers.
Also, the low quality sellers must prefer to sell at 400 without a guarantee rather than
imitate the high quality sellers (offer a refund of R and sell at 1,300). This implies:
400 > 1300 0.8R
which implies R > 1125. This is called the incentive compatibility constraint (IC) of low
quality sellers.
The value of R must be such that the PC of high quality sellers and the IC of low
quality sellers are both satisfied. This implies that the range of values of R for which
the separating equilibrium holds is 1125 6 R 6 3000.
Forced refunds
Suppose the government decides to force each seller to offer a full refund if the car sold
by the seller breaks down. How does this change the market outcome? Is the market
outcome efficient?
Note that the maximum possible price in the market is 1,300. But even at a price of
1,300, 0.2 1300 = 260 is below 300. It follows that low quality sellers would not
participate in the market, and only high quality sellers would remain and sell at 1,300.
But this is not efficient as the gain from trade of low quality cars is not realised.
In other words, forcing sellers to issue a refund destroys the role of a refund policy in
separating the two qualities. Since all sellers must issue a refund if the car breaks down,
low quality sellers prefer to withdraw from the market, making it impossible to exploit
gains from trading low quality cars. Policy choices that are not sensitive to the role of
an instrument in providing information to uninformed agents do not necessarily
promote efficiency.
10.5 A model of price discrimination

N&S discusses informally a problem of price discrimination with quantity-price pairs in
Section 15.3. Here we develop a formal model of price discrimination with
quality-price pairs (rather than quantity-price pairs). When reading, do keep in mind
this distinction between the discussion in the textbook and the model studied here.
10.5.1 The model

There is a seller and a buyer. The seller sells a unit of quality q at price t. The cost of
producing quality q is c(q) = q 2 .
153
The profit of the seller is given by:
(t, q) = t q 2
and the utility of the buyer is given by:
u(t, q, ) = q t.
Here is a parameter that reflects how much the buyer cares about quality. is usually
referred to as the buyers type. This is the buyers private information.
In this model we suppose can take two values, 1 and 2 , where 2 > 1 .
A contract is a quality and price pair (q, t) offered by the seller. We assume, for
simplicity, that the seller has all the bargaining power.
The buyer gets 0 utility if they do not buy. Therefore, any contract must give the buyer
at least 0. This is called the participation constraint (PC) of the buyer.
The indifference curve of a buyer of type i , i {1, 2} is given by:
i q t = constant.
The slope of an indifference curve is given by dt/dq = i .

Next, an iso-profit curve for the seller is given by:
t q 2 = constant.
The slope of the iso-profit curve at quality q is 2q. Note that the slope does not depend
on t, only q. So if we fix any q and change t, the slope does not change. In other words,
all iso-profit curves have the same slope at points along any vertical line (at all values of
t). Figure 10.1 shows the indifference map of the buyer of type 2 and that of the seller.
The arrows show the direction of improvement.
Figure 10.1: The indifference map of a buyer of type 2 and that of the seller.
154
10.5.2 The full information benchmark

Under full information, the seller offers a contract (q1 , t1 ) to type 1 and another
contract (q2 , t2 ) to type 2 .
To determine the optimal contract for each type i , i {1, 2}, the seller solves:
max ti qi2 subject to i qi ti > 0.

qi , ti
However, since the seller has all the bargaining power, there is no reason to give the
buyer any more than 0, and thus the constraint holds with equality. Thus the seller
solves:
max ti qi2 subject to i qi ti = 0
qi , ti
which can be rewritten as:

max i qi qi2 .
qi
Thus qi = i /2, and ti = i qi = i2 /2. Figure 10.2 below shows the optimal contracts
under full information. The optimal contract for type i is obtained at the point of
tangency between the iso-profit curve and the indifference curve of type i at the
reservation utility level (here the reservation utility is 0).
Figure 10.2: The optimal contracts under full information.
10.5.3 Contracts under asymmetric information

Under asymmetric information, the full information contracts are not incentive
compatible. To see this, note from Figure 10.2 that type 1 has no incentive to take the
contract meant for type 2 . Doing so would yield a strictly negative payoff for type 1 .
155
However, type 2 does want to take the contract meant for type 1 . From Figure 10.2,
you should be able to see that the indifference curve of type 2 passing through (q1 , t1 )
is associated with a higher utility compared to the original indifference curve of 2 .
You can also see this easily using algebra. Type 2 gets a 0 payoff from (q2 , t2 ). But the
utility from (q1 , t1 ) is:
1 1
2 q1 t1 = 2 1 12 = (2 1 ) > 0.
2 2
Alternatively, note that 2 q1 t1 > 1 q1 t1 = 0. Therefore, the full information
contracts are not incentive compatible for type 2 .
Indeed, given any choice of (q1 , t1 ), incentive compatibility for type 2 requires that
(q2 , t2 ) must be on the indifference curve of type 2 passing through (q1 , t1 ). This is
shown in Figure 10.3 below. Note that for any (q2 , t2 ) on the solid indifference curve for
type 2 , that type must be indifferent between accepting (q2 , t2 ) and (q1 , t1 ). If (q2 , t2 ) is
anywhere above this line, incentive compatibility does not hold, and type 2 would
rather choose (q1 , t1 ). If (q2 , t2 ) is anywhere below this line, type 2 would strictly prefer
it, but then the seller is not choosing optimally they could raise t2 (say) and still
satisfy incentive compatibility.
Figure 10.3: Incentive compatibility.
Which point should the seller choose on the solid indifference curve for type 2 ? Where
the sellers iso-profit curve is tangent to this line. The line has a slope of 2 , while the
slope of the iso-profit curve is 2q. Therefore, the optimal q2 equals 2 /2, which is the
same as q2 under full information.
So the only thing that changes is the price t2 . As Figure 10.4 below shows, given any
choice of (q1 , t1 ) on the participation line of type 1 , and given q2 = q2 = 2 /2, t2 must
be lowered to satisfy incentive compatibility for type 2 .
156
Under full information, the contracts for the two types can be chosen independently,
but this is no longer possible under asymmetric information. Now, for any choice of
(q1 , t1 ), t2 has to be adjusted down to prevent type 2 from taking the contract meant
for type 1 . In the process, type 2 gets a surplus over its participation level (which is
0). This surplus is called the information rent of type 2 , shown in Figure 10.4.
Figure 10.4: Incentive compatibility and information rent of type 2 .
In terms of algebra, the seller must offer contracts (q1 , t1 ) and (q2 , t2 ) such that:
the participation constraint of type 1 (PC1 ) is just satisfied (binds), so that:
1 q1 t1 = 0
the incentive compatibility constraint of type 2 (IC2 ) is just satisfied (binds), so

that:
2 q2 t2 = 2 q1 t1 .
Furthermore, we know that the optimal q2 is as under full information:
q2 = q2 = 2 /2.
Since q2 is known, given any choice of q1 , the two constraints (PC1 ) and (IC2 ) above
determine t1 and t2 , respectively. How does the seller choose q1 ? We do not solve this
problem here, but the seller must consider the proportion of the two types in the
population and determine an optimal value. The exact q1 can be determined as a
function of the proportion and 1 and 2 , but this is not important for us here.
The lesson that you should take away from this is the difference that asymmetric
information makes. Under full information, each contract can be determined separately
157
by satisfying the participation constraint of each type. Under asymmetric information,

incentive constraints need to be considered. In our simple model, the low type does not
have any incentive to take the contract meant for the high type, but the high type
would take the low types contract unless the high types price t2 is lowered to satisfy
the high types incentive compatibility constraint.
Sellers who sell different quality versions of a product (household appliances, say) at
different prices need to solve this type of problem all the time. If the high quality version
is very expensive, most consumers might select the lower quality but cheaper product.
The seller must take this into account when setting the price of the high quality version.
10.6 Spences (1973) model of job market signalling

See Section 15.5 of N&S for an exposition of this model. In previous applications, the
uninformed party was trying to screen the informed. A different type of solution arises
if the informed party can signal their information to the uninformed party. Spences
(1973) model of job market signalling studies precisely such a scenario.
N&S develops separating and pooling equilibria. You should study these carefully and
understand the structure of each equilibrium.
Here let us summarise the main points. There are two types of workers some with a
high productivity (the high type) and some with a low productivity (the low type). High
type workers have a cost of acquiring education of c` per year of education, whereas the
low types have a cost of acquiring education of ch per year of education, where ch > c` .
Under full information, the competitive yearly wage for a high type is wh whereas the
competitive yearly wage for a low type is w` . Suppose employment lasts for n years.
Separating equilibrium
A separating equilibrium has the following structure. There is a threshold number y of

years of education such that if years of education y of a worker is at least y , they are
considered to be a high type, and any worker with fewer than y years of education is
considered to be a low type.
Is such an equilibrium possible? For any y > y , the total wage benefit is
B = (wh w` )n. For a separating equilibrium, it must be that any high type does want
to acquire at least y years of education:
B > c` y .
Furthermore, low types should not want to send the same signal, so we must have:
B 6 ch y .
(We assume that when indifferent, types behave benignly, i.e. high types send and low
types do not send the signal. We could, of course, not assume this and require instead
the inequalities to be strict.) Thus, for a separating equilibrium, we must be able to
choose a y such that B/ch 6 y 6 B/c` .
158
Pooling equilibrium
A second type of equilibrium is a pooling equilibrium. This has some subtleties which
you will appreciate more when you study game theory at a more advanced level. But let
us give a preliminary idea here.
Consider a pooling equilibrium in which both types of workers acquire an education.
Suppose the firm requires yb years of education and suppose all types acquire this much
education. Then the firm only knows that all types are in the market and pays an
average wage w, reflecting the average productivity.
Acquiring more education does not help, so there is no such deviation possible. But is it
possible for a type to deviate to lower than yb years of education? Clearly, if anyone
would want to deviate, it is the low type for whom the cost of education is higher. But
in equilibrium the firm expects all types of workers to get yb years of education. If some
worker deviates to no education, what would the firm believe about this workers type?
This is a rather complex question: what is a reasonable belief when you see something
that you are not expecting to see in equilibrium? Game theorists have devoted a lot of
time to answering this question, and have proposed many ways to try to deal with this
subtle issue, but there is no definitive answer. Anyway, here let us suppose that any
deviation to fewer than yb years of education is met with the worst belief about the
deviator: the firm believes this worker is a low type and would pay w` . In this case,
deviation is not beneficial if:
wn ch yb > w` n.
b = (w w` )n. The above condition can be written as:
Let B
B
b > ch yb.
If this holds, we can have a pooling equilibrium where all types acquire yb years of
education.

to:
explain the different types of asymmetric information problems
explain how the scope of enquiry of economics is expanded by considering

information asymmetry problems
explain the types of problems that arise under adverse selection
analyse the problem of adverse selection in the market for lemons
analyse the problem of price discrimination by a monopolist under adverse selection
analyse separating and pooling equilibria in the signalling model of education.
159

1. The signalling model of education would break down if the costs of acquiring
education were equal for individuals with different abilities. Is this true or false?
Explain your answer.
2. Consider a second-hand car market with two kinds of cars, type A which are
completely reliable, and type B which break down with probability 1/2. There are
20 car owners, 10 with each kind of car, and 20 potential buyers. The car owners
and buyers value a car at 1,000 and 1,500, respectively, if it works and both owners
and buyers value a non-working car at zero. Both buyers and owners are assumed
to be price-takers and risk-neutral. Finally, each seller is aware of the type that her
car belongs to, but this information is not available to buyers.
(a) What values do buyers and sellers place on type A and type B cars?
(b) How many cars, and of what type(s) are supplied at each price?
(c) What is the demand for cars at each price?
(d) How many equilibria are there in the market?
3. A seller sells a unit of a good of quality q at a price t. The cost of producing at

quality level q is given by q 2 . There is a buyer who receives a utility of q t by
consuming a unit of quality q at price t. If they decide not to buy, they get a utility
of zero. can take two values, 1 = 1 and 2 = 2. Assume that the seller has all the
bargaining power.
(a) Suppose the seller can observe . Derive the profit-maximising price-quality
pairs offered when = 1 and when = 2. Show that the quality offered when
= 2 is twice the quality offered when = 1.
(b) Prove that the full information price-quality pairs are not incentive compatible
if the seller cannot observe .
(c) Suppose the seller cannot observe . Assuming q1 = 1/4, derive a set of
price-quality pairs that satisfy incentive compatibility. (This part is a little
harder.)
160
Chapter 11
Asymmetric information: moral
hazard
11.1 Introduction
Moral hazard arises when the behaviour of an individual changes as a result of entering
into a contract. The owner of a bike that is fully insured against theft might be more
likely to forget to lock it. The true level of care (or effort) is hidden from the insurer.
The asymmetry in information here is described as hidden action, as distinct from that
of hidden information or adverse selection.
The issue can be studied more generally as a principal-agent problem. Many economic
transactions have the feature that unobservable actions of one individual have direct
consequences for another, and the affected party may seek to influence behaviour
through a contract with the right incentives. In the above example, the insurance
company (the principal) is affected by the unobservable carelessness of the insured (the
agent): it then chooses a contract with only partial insurance to preserve the right
incentives. Other economic relationships of this type include shareholders and
managers, manager and salespersons, landlord and tenants etc.
In the following sections, we specify a model to capture an effort choice problem and
suggest contractual solutions.

The previous chapter introduced the problem of adverse selection or hidden
information. This chapter aims to introduce the class of problems called moral hazard
that arises when actions of an agent are hidden. The chapter considers a principal-agent
relationship and aims to show how moral hazard and risk attitudes interact in
determining the optimal form of incentive contracts offered by the principal to the agent.

explain the types of problems that arise under moral hazard
explain the problem of effort choice and analyse effort choice in a principal-agent
model
explain the trade-off between risk-sharing and effort incentives in the

principal-agent model and analyse the optimal risk-sharing arrangement under full
161
11. Asymmetric information: moral hazard
and asymmetric information
explain informally the impact of moral hazard in a variety of settings.

N&S Chapter 15, Sections 15.1 to 15.2. Read also Applications 15.1 to 15.3 for some
other applications of moral hazard. In the subject guide we develop a formal model of
the principal-agent wage-contract model studied informally in Sections 15.1 and 15.2.
You must study the textbook and the coverage in the subject guide carefully.

This chapter covers a model of effort choice by an agent and incentive contracts
between a principal and the agent. First, the model is introduced. Next, the chapter
covers the optimal contract under full information followed by that under asymmetric
information, assuming that the agent is risk-averse. Finally, the chapter covers the
optimal contract under asymmetric information in the case of a risk-neutral agent.
11.3 Effort choice and incentive contracts: a formal

model
N&S informally discusses the trade-off between risk-sharing and incentives in a
principal-agent setting in Sections 15.1 and 15.2. Here we develop a model to explore
this trade-off in a more formal manner.
A principal (P ) hires an agent (A) to carry out a particular project. Once hired, A
chooses an effort level, and his choice affects the outcome of the project in a
probabilistic sense: higher effort leads to a higher probability of success, and that
translates into a higher expected profit for P . If the choice of effort was observable, P
could stipulate, as part of the contract, the level of effort that is optimal for her (that
is, for P ). When effort cannot be monitored, P may yet be able to induce a desired
effort level, by using a wage contract with the right incentive structure. What should
these contracts look like? The formal structure can be set up as follows:
1. Let e denote the effort exerted by the agent on the project. A can either work hard
(choose e = eH ) or take it easy (choose e = eL ). An informational asymmetry arises
when the choice of e cannot be monitored by P .
2. Let the random variable denote the (observable) profit of the project. Profit is
affected by the effort level chosen, but is not fully determined by it. (If it was fully
determined, the principal can infer the true effort choices by observing the
realisation of : we would no longer be in a situation of hidden actions.) Higher
effort leads to higher profit in a probabilistic sense. Assume that can be either
high (H ) or low (L ). Let P ( | e) be the probability of under effort e.
162
11.3. Effort choice and incentive contracts: a formal model
Under high effort, the high profit is more likely. Therefore, we have:
P (H | eH ) > P (H | eL ).
Example 11.1 As an example, consider the following scenario. Profit can be 10 or

0. The numbers in the boxes are probabilities. Under high effort, 10 occurs with
probability 2/3, and 0 with 1/3. Note that under low effort, 10 is less likely than
under high effort.
Profit
10 0
Agents effort eH 2/3 1/3
eL 1/3 2/3
Moral hazard arises from the fact that the realised profit is not fully informative
about effort. If, on the other hand, the probabilities were (1, 0) under eH and (0, 1)
under eL , there would be no moral hazard, as profit would reveal effort completely.
3. The principal is risk-neutral: she maximises expected profit net of any wage
payments to the agent.
4. The agent is risk-averse in wage income, and dislikes effort. His utility function
takes the form:
v(w, e) = u(w) g(e)
where w is the wage received from the principal, g(e) is the disutility of effort and
u is the vNM utility function, where u0 > 0, u00 0 (implying risk aversion).
Finally, high effort is more costly so g(eH ) > g(eL ).
Note the central conflict of interest here. The principal would like the agent to choose
higher effort, since this leads to higher expected profits. But, other things being equal,
the agent prefers low effort. However, (to anticipate our conclusion), if the
compensation package is carefully designed, their interests could be more closely
aligned. For instance, if the agents compensation increases with profitability (by means
of a profit-related bonus, say), the agent would be tempted to work hard (because high
effort will increase the relative likelihood of more profitable outcomes). However, this
will also make the agents compensation variable (risky), and that is not so efficient
from a risk-sharing perspective. The inefficiency in risk-sharing will be an unavoidable
consequence of asymmetric information.
5. The principal pays the agent a profit-contingent wage schedule {wH , wL }. Here wH
is the wage paid if the profit is H , and wL is the wage paid if the profit is L . In
general, wH > wL . Note that if wH = wL = w , the principal is paying a fixed wage
w that does not depend on the profit level. A fixed wage gives full insurance to the
agent (income does not vary no matter what the profit is). However, if wH > wL ,
the wage schedule carries risk.
6. Let u0 be the reservation utility of the agent. This is the expected utility the agent
gets from alternative occupations. Any wage contract that the principal offers the
163
agent must give the agent a net payoff of at least u0 , otherwise the agent would not
accept the contract. This is known as the agents participation constraint (PC).
Let us now show how to determine the optimal wage schedule under full information
and under asymmetric information.
11.4 Full information: observable effort

The case of observable effort serves as a benchmark. Once we know what the outcome is
under full information, we can judge the outcome under asymmetric information against
the benchmark and see how much extra inefficiency arises from the presence of
asymmetric information.
Suppose effort is observable. The principal can then write a contract with the agent
directly specifying effort. Suppose the principal wants to specify eH . What wage must
be paid?
11.4.1 Implementing high effort eH

The agent would agree to undertake eH so long as he obtains a utility of at least u0 .
This is called the agents participation constraint (PC). This is given by:
P (H | eH ) u(wH ) + P (L | eH ) u(wL ) g(eH ) > u0 . (PC)
Optimality of a fixed wage under observable effort
Which wage schedule should the principal choose to implement eH ? If the agent chooses
effort eH , the principal gets the surplus:
P (H | eH ) (H wH ) + P (L | eH ) (L wL ).
The principal wants to maximise this surplus subject to the agents PC.
We will do this by Lagrangian optimisation. But even before doing that, the optimal
arrangement should be intuitively clear. Choose a wage schedule (wH , wL ) which just
satisfies the agents PC, i.e. PC holds with equality. Suppose the wage schedule is risky,
i.e. wH > wL . Starting from this, can the principal do better? Recall that the principal
is risk-neutral while the agent is risk-averse. If the principal gives insurance to the agent
by replacing the risky wage schedule with a fixed wage equal to the expected wage, the
agent would be strictly better off while the principal would remain indifferent. But now
the agents utility is strictly above u0 , so the principal can reduce the fixed wage down
to a level that will again make the agent indifferent between the utility from
participating and u0 . In the process, the principal is strictly better off.
In other words, when the principal gives the agent a risky wage, there is inefficient
risk-sharing. Facing a risky wage, the agent would pay to buy income insurance. But
the principal is risk-neutral, so she can give this insurance to the agent and extract that
payment (i.e. reduce the expected wage). In yet other words, starting from a fixed wage
that just satisfies PC, if the principal replaces the fixed wage with a risky wage, the
agent will not participate as his payoff will fall below u0 . The principal has to also
164
11.4. Full information: observable effort
compensate the agent for the addition of risk in order to satisfy the PC. So putting risk
on the agent is costly for the principal. Since effort is directly observable, there are no
incentives to worry about and risk serves no purpose. Therefore, when effort is
observable, it is optimal for the principal to pay a fixed wage that does not depend on
the profit level.
Let us now show this by doing a Lagrangian optimisation. However, you should really
not rely on algebra for this the intuition should be clear to you.
The principals problem is to maximise surplus (i.e. net-of-wages profit) subject to the
PC of the agent. Note that there is no reason for the principal to leave any extra
surplus for the agent, so the optimal wage schedule would be such that PC binds (holds
with equality). With this in mind, the principals problem is:
max P (H | eH ) (H wH ) + P (L | eH ) (L wL )
wH , wL
subject to:
(H | eH ) u(wH ) + (L | eH ) u(wL ) g(eH ) = u0 .
Set up the Lagrangian with as the Lagrange multiplier, and derive the first-order
conditions:
L
= P (H | eH ) + P (H | eH ) u0 (wH ) = 0
wH
L
= P (L | eH ) + P (L | eH ) u0 (wL ) = 0.
wL
We ignore the derivative with respect to which recovers the participation constraint.
Simplifying, we get:
1
= u0 (wL ).
u0 (wH ) =

00 0
Since u < 0, u is a decreasing function. Therefore, from the equalities above, we get
wH = wL .
Therefore, to implement eH , the principal optimally offers a fixed wage w such that:
u(w ) g(eH ) = u0 .
11.4.2 Implementing low effort eL

The same logic as above applies here the cheapest way to implement the low effort is
to offer the agent a fixed wage w such that the agents PC binds (holds with equality):
u(w ) g(eL ) = u0 .
11.4.3 Which effort is optimal for the principal?

The principal can implement eH by paying the agent a fixed wage w . Therefore, the
expected surplus from eH is given by:
P (H | eH ) H + P (L | eH ) L w .
165
The expected surplus from eL , which can be implemented by paying a fixed wage w , is
given by:
P (H | eL ) H + P (L | eL ) L w .
The principal should compare the two and implement the effort that yields a higher
surplus.
11.5 Asymmetric information: unobservable effort
11.5.1 Implementing low effort eL

First, consider the problem of implementing the low effort eL . Note that the principal
does not need to provide incentives not to work hard. In fact, whenever there are no
particular incentives to work hard, the agent would automatically choose low effort. We
know that ideally the principal wants to pay the agent a fixed wage as that allows the
principal to satisfy the agents PC at the lowest possible expected wage payment. But a
fixed wage provides no incentive to work hard! The agent gets the same wage no matter
what the profit level is, so he might as well exert low effort. What is the optimal fixed
wage? We already worked that out in the previous section under observable effort.
Paying that same fixed wage under asymmetric information would result in the agent
choosing eL . Therefore, as far as making the agent take low effort is concerned,
asymmetry of information does not place any extra constraints. The principal can
simply pay a fixed wage w that just satisfies the agents PC to implement eL . Here
w is given, as in the full information case, by:
u(w ) g(eL ) = u0 .
11.5.2 Implementing high effort eH

We know from above that under asymmetric information, implementing eL is a trivial
problem. However, implementing high effort is a non-trivial problem. Now the principal
not only must ensure that the agent accepts the wage contract, the agent must also be
given some incentive to ensure that he exerts high effort eH . Specifically, the wage
schedule {wH , wL } that the principal offers must satisfy the two constraints described
below.
The wage schedule must satisfy the agents PC, i.e. it must be such that the agents
payoff under eH is at least as large as u0 :
P (H | eH ) u(wH ) + P (L | eH ) u(wL ) g(eH ) > u0 . (PC)H
The compensation package must also be such that the agent prefers to exert eH
rather than eL . In other words, the wage schedule must create the right incentives
for choosing high effort over low effort. This is called the incentive compatibility
constraint IC. This is given by:
P (H | eH ) u(wH ) + (L | eH ) u(wL ) g(eH ) >

P (H | eL ) u(wH ) + P (L | eL ) u(wL ) g(eL ). (IC)
166
11.5. Asymmetric information: unobservable effort
Note that a fixed wage cannot satisfy the IC. If the wage is fixed, we have
wH = wL = w. Then the left-hand side is u(w) g(eH ), while the right-hand side is
u(w) g(eL ). Since g(eH ) > g(eL ), the IC is violated. Putting some risk on the agent is
unavoidable in providing an incentive for high effort.
To determine the optimal values of wH and wL , we need to maximise the surplus of the
principal subject to the PC and IC. However, in our simple model there are just two
wage levels and two constraints. Therefore, once we put numerical values for the various
probabilities and effort costs, the constraints determine the values of wH and wL fully
(so no maximisation is necessary). The following section contains an example.
However, in more general models that you would encounter if you pursued economics at
a Masters level, there could be (say) more than two profit levels, which would indicate
that the compensation package would consist of more than two wage levels. In such
cases, some further maximisation needs to be done.
The following activity goes through a detailed example of the theory we described
above.
Example 11.2 Consider the following principal-agent model. A principal hires an

agent to work on a project in return for wage payment w > 0. The agents utility
function is separable in the effort and wage received: we have v(w, ei ) = u(w) g(ei ),
where u() is his vonNeumann Morgenstern utility function for money, and g(ei ) is
the disutility associated with effort level ei exerted on the project.
Assume that the agent can choose one of two possible effort levels, e1 or e2 , with
associated disutility levels g(e1 ) = 5/3, and g(e2 ) = 4/3. The value of the projects
output depends on the agents chosen effort level in a probabilistic fashion. If the
agent chooses effort level e1 , the project yields profit H = 10 with probability
P (H | e1 ) = 2/3, and L = 0 with the residual probability. If the agent chooses effort
level e2 , the project yields 10 with probability 1/3, and 0 with the residual
probability.
The principal is risk-neutral: she aims to maximise the expected value of the output,

net of any wage payments to the agent. The agent is risk-averse, with u(w) = w,
and his reservation utility equals 0.
We discuss the following questions.
(a) If effort is observable, the principal optimally offers a fixed wage to implement
any effort level. Provide a brief intuition for this result.
Since the agent is risk-averse and the principal is risk-neutral, an optimal
risk-sharing result is for the principal to take all the risk. If the agent were to
carry unwanted risk, he would have to be given a higher expected wage than in
the optimal case. Since this is not necessary for incentive reasons because effort
is observable, the first best solution is for the principal to take all the risk and
completely insure the agent.
You can check this formally using the derivation presented in Section 11.4.1 of
the subject guide.
(b) If effort is observable, what wage w should the principal offer if she wants to
implement e1 ? What wage w implements e2 ? Which induced effort level
167
provides a higher expected return to the principal, net of wage costs?

Because of (a), we know that under the observable effort the wage depends on
effort ei {1, 2} and not on the profit outcome (high or low). To implement e1
at minimum cost, the principal simply needs to pay a fixed wage w to satisfy
the participation constraint (PC1 ), given by:
u(w ) g(e1 ) = u0 = 0.
Similarly, to implement e2 , the principal pays w to satisfy (PC2 ).

We get:
i. (high effort) e1 :

PC1 : w = 5/3 w = 25/9
principals profit: 20/3 25/9 (e1 ) = 35/9.
ii. (low effort) e2 :

PC2 : w = 4/3 w = 16/9
principals profit: 10/3 16/9 (e2 ) = 14/9.
Since (e1 ) > (e2 ), the principal will implement e1 and pay a fixed wage of
25/9.
(c) Suppose, next, that the agents choice of effort level is not observable. In this
circumstance, a contract consists of an output-contingent wage schedule
{wH , wL }. What wage schedule will implement e1 in this case? What expected
net return does the principal get in this case? How does this compare with the
value in (b), where effort was observable?
State-dependent contracts are {wH , wL }. Implementing e1 at minimum expected
wage cost requires minimising 2wH /3 + wL /3 subject to the participation
constraint (PC1 ) as well as the incentive constraint (IC) of the agent. Solving
the two constraints completely determines wH and wL , so that there is no
further minimisation to be done. We have:
2 1 5
(P C1 ) wH + wL 0
3 3 3
and:
2 1 5 1 2 4
(IC) wH + wL wH + wL .
3 3 3 3 3 3
Since there are two variables to be determined, and two inequalities, it is

possible to find wH and wL so that both bind. (IC) implies wH = 1 + wL .
We can use this in (P C1 ) and obtain:

2(1 + wL ) + wL = 5 (wL = 1, wH = 4)
with an associated profit of:

2 1 11
(e1 ) = (10 4) + (0 1) (e1 ) = .
3 3 3
168
11.6. Risk-neutral agent
Note that the profit was 35/9 under full information, and here it is 33/9. Why
does the principals profit fall by 2/9 under asymmetric information? This is
because the principal is now putting risk on the agent and, therefore, has to pay
a higher expected wage to compensate. Hence the profit is lower.
(d) If effort is not observable, what contract is best for the principal? Should she
implement e1 or e2 ?
Implement e2 . Any flat wage that satisfies the agents participation constraint
implements e2 . The cheapest way for the principal to do this is to offer the first
best flat wage w1 = 16/9.
As before, (e2 ) = 14/9. Since (e1 ) > (e2 ), the principals optimal choice
under asymmetric information is e1 .
11.6 Risk-neutral agent

The analysis above explained the trade-off between risk-sharing and incentives with a
risk-neutral principal and a risk-averse agent. Putting risk on the agent is good for
incentives but bad for risk-sharing, and bad risk-sharing is costly for the principal. If,
however, the agent is risk-neutral, the problem of the principal becomes a trivial one.
Now there is no trade-off: putting risk on the agent is not costly, so the principal can
generate incentives for free.
In particular, the principal can simply put all the risk on the agent and ask for a fixed
fee (akin to a franchising contract). This contract leads to the same outcome as when
effort is observable. Why? Note that the agents problem is the same as the principals
problem if the principal faced a lump-sum tax. But a lump-sum tax does not change
anything the agent would optimally choose whatever the principal would have chosen
under full information. In other words, it is exactly as if the principal has sold the
project to the agent, so that the agent has all the incentive as the owner to choose the
profit-maximising effort.
To see this, consider the example above. Suppose the agents utility function is:
v(w, ei ) = w g(ei ).
Under full information, to implement the high effort e1 , the principal pays w = 5/3 and
to implement e2 , pays w = 4/3. The principals profit from e1 is 5 and from e2 the
principals profit is 2.
Now consider the scenario under asymmetric information. We already know that the
principal can implement e2 by the same fixed wage as above, and earn 2 as before
asymmetric information has no impact on the profit from implementing the low effort.
However, what is the profit from implementing e1 ? Recall that some loss of profit was
inevitable when the agent was risk-averse. Here, suppose the principal simply asks for a
fixed fee of F = 5. The agent is the residual claimant and bears all the risk. Will this
work?
If the agent adopts e1 , he gets 20/3 5/3 F = 5 5 = 0, while the payoff from
adopting e2 is 10/3 4/3 F = 2 5 < 0. Note that the payoff from e1 satisfies both
169
the participation constraint and the incentive compatibility constraint. Therefore, the
agent would adopt e1 , and the principal gets 5, which is the same as under full
information. So, for the principal, asymmetry of information does not impose any loss
compared to full information because putting risk on the agent is not costly.

to:
explain the types of problems that arise under moral hazard
explain the problem of effort choice and analyse effort choice in a principal-agent
model
explain the trade-off between risk-sharing and effort incentives in the

principal-agent model and analyse the optimal risk-sharing arrangement under full
and asymmetric information
explain informally the impact of moral hazard in a variety of settings.
1. Car insurance and home insurance typically include a deductible. Briefly explain
the reason behind this feature.
2. Explain the trade-off between risk-sharing and incentives in a principal-agent

setting with a risk-neutral principal and a risk-averse agent. How does the trade-off
change if the agent is risk-neutral?
3. A risk-neutral principal hires an agent to work on a project at wage w. The agent

exerts effort e on the project. The agents utility function is:

v(w, e) = w g(e)
where g(e) is the disutility associated with the effort e.

The agent can choose one of two possible effort levels, e1 or e2 , with associated
disutility levels g(e1 ) = 1, and g(e2 ) = 1/2. If the agent chooses effort level e1 , the
project yields 8 with probability 1/2, and 0 with probability 1/2. If he chooses e2 ,
the project yields 8 with probability 1/4 and 0 with probability 3/4.
The reservation utility of the agent is 0.
170
(a) Suppose the effort level chosen by the agent is observable by the principal. A
wage contract then specifies an effort level (e1 or e2 ), and an output-contingent
wage schedule {wH , wL }. Here wH is the wage paid if the project yields 8, and
wL is the wage paid if the project yields 0. If effort is observable, it is optimal
for the principal to choose a fixed wage contract (that is, set wH = wL ) for
each effort level. Informally, explain the intuition for this result.
(b) If effort is observable, which effort level should the principal implement? What
is the best wage contract that implements this effort?
(c) If effort is not observable, which effort level should the principal implement?
What is the best wage contract that implements this effort?
171
172
Chapter 12
Externalities and public goods
12.1 Introduction
As noted in Chapter 7, the presence of externalities, as well as public goods, gives rise
to both qualitatively and quantitatively important departures from the first welfare
theorem. Much of economic policy is focused on dealing with these departures.
In some situations, the utility of an agent is affected by the level of activity of another
agent directly. However, the acting agent does not take this externality into account.
The result is that the market outcome is inefficient. The individually-optimal level of an
activity that generates a negative externality is higher than the socially-optimal level,
and that for an activity that generates a positive externality is lower than the
socially-optimal level. The invisible hand does not work well in such cases individual
optimisation does not lead to social optimality.
Some of the biggest problems facing the world are caused by negative externalities.
These include global warming, pollution, congestion, urban noise, and so on. In all these
cases, government policy has a crucial role to play to bring individual incentives in line
with the social optimum.
Competitive markets also work poorly in the presence of public goods. However, public
goods such as national defence, roads, basic public health, basic education and defence
against natural catastrophes are important for any economy. Provision of such goods
using receipts from taxes is a crucial role for the government in the economy.
When you study these topics, you should try to understand the nature of different
policies: their merits, demerits and limits. You should also try to understand how much
of a difference good economics can make to the world. Unfortunately, good economics is
not always appreciated or endorsed by politicians, industry leaders or the wider
population. The cap-and-trade markets in the United States are a case in point.
Economists have been urging governments to have such markets since at least the
1960s. It took another three decades for the first such system to be set up, amid great
scepticism expressed by politicians as well as industry leaders. This was the United
States acid rain programme, part of the Clean Air Act (1990). The caps on NOx and
SO2 took effect in 1995. Within a few years, it was clear that the system was a huge
success. Yet, even now, scepticism is regularly expressed about emissions trading by the
opponents of the policy in the United States Congress, whose entrenched positions
effectively eliminated any chance of the Obama administration establishing a
nationwide CO2 emissions trading system. While we highlighted the case of the United
States here, disputes about economic policy arise across the world, and, unfortunately,
bad economics often drives good economics away. As students of economics, you should
try and understand these problems well, so that through individuals like yourself, good
173
12. Externalities and public goods
economics can find more of a voice in the world.

This chapter aims to introduce a class of problems that constitute, from a policy point
of view, the most important class of departures of market outcomes from those implied
by the first welfare theorem.

explain the concept of externalities and show how externalities lead to market
failures
analyse remedial policies based on taxes or subsidies
explain the content of the Coase theorem and analyse the Coasian property rights
approach to solving the problem of externalities
explain how an overall quota can be implemented efficiently using cap-and-trade

systems, how such systems have reduced pollution from several sources and how
such systems can sometimes run into difficulties
explain the concept of public goods and analyse how private provision leads to an
inefficiently low provision of such goods
explain the concept of the tragedy of the commons and demonstrate how strategic
behaviour leads to over-exploitation of open-access resources in simple models
explain the policy responses to commons problems.

N&S Chapter 16, Sections 16.1 to 16.6. See also Section 5.7 for a simple model of the
commons problem. In the subject guide we develop a more algebraic approach to these
topics, which you should also read carefully.
12.1.4 References cited

Coase, R.H. (1960) The problem of social cost, Journal of Law and Economics, Volume
3, pages 144.
Schmalensee, R. and R.N. Stavins (2013) The SO2 allowance trading system: The ironic
history of a grand policy experiment, Journal of Economic Perspectives, Volume 27,
pages 103121.
Stavins, R.N. (2011) The problem of the commons: Still unsettled after 100 years,
American Economic Review, Volume 101, pages 81108.
174
12.2. Overview of the chapter
UK Department of Energy and Climate Change, Participating in EU ETS,

https://www.gov.uk/guidance/participating-in-the-eu-ets.

The chapter covers the problems posed by the presence of externalities, and associated
policy prescriptions. Next, the chapter presents the public goods problem and policies
promoting the optimal provision of such goods. Finally, the chapter covers the commons
problem and policies to address the problem.
12.3 Externalities
See N&S Sections 16.1 to 16.4 for basic concepts. Here we consider a simple model of a
society with two agents, 1 and 2. Agent 1 decides on the level x of an action. Assume
that x is costless. To model externalities, we assume that not just agent 1 but also
agent 2 is affected by agent 1s choice.
Agents have quasilinear utility. The utility function of agent i is i (x) + mi , where
i = 1, 2. Here you can think of good m as money. Let m be the total amount of good m.
Furthermore, i (x) denotes the utility of i from x. We assume that 01 > 0 and 001 < 0.
Also, 002 < 0. What about the sign of 02 ? If this is positive, we have a case of a positive
externality, while 02 < 0 implies a case of a negative externality. Below we consider the
case of a positive externality briefly, but most of our analysis is concerned with the case
of a negative externality.
Note that the utility function is linear in mi , but not in x. Hence the name quasilinear.
Individual optimum
Next, since x is costless, the individual optimum for agent 1 is reached at x such that:
01 (x ) = 0.
Note that the second-order condition holds since 1 is concave by assumption.
Pareto optimum
The Pareto optimal outcome can be found simply by maximising the sum
1 (x) + 2 (x). This is a consequence of the quasilinear utility function: we can simply
compare the utilities from x with money amounts. Suppose we have a negative
externality. Suppose the marginal utility of agent 1 from x is higher than the marginal
disutility of agent 2 from x. Suppose we increase x and keep agent 2s utility the same
by transferring money to agent 2 equal to the extra disutility. Since the extra utility
gained by agent 1 from the increase in x is greater than the money transferred to agent
2, agent 1 has a net gain. So whenever agent 2s marginal disutility is lower than agent
1s marginal utility, we can make a Pareto improvement. This leads to the idea that at
the optimum the sum of marginal utilities must be zero. A similar intuition applies to
175
the case of a positive externality. We will show this more formally when discussing
public goods. This gives us the following.
The Pareto optimal level of x is denoted by xo . This maximises the total 1 (x) + 2 (x).
Therefore, xo is such that:
01 (xo ) + 02 (xo ) = 0.
Note that the second-order condition holds since 1 and 2 are both concave by
assumption.
In what follows we use the terms Pareto optimal, efficient and socially optimal
interchangeably.
Positive externality
Consider first the case of a positive externality. In this case the social benefit of x
obviously exceeds the individual benefit of agent 1, and, as shown in Figure 12.1 below,
the level x undertaken by agent 1 is lower than the efficient level xo .
Figure 12.1: Positive externality.
Negative externality
From here onwards we consider the case of a negative externality. The policy solutions
analysed later can be appropriately modified (for example, a tax solution becomes a
subsidy solution) for the case of a positive externality.
In this case 02 > 0. Figure 12.2 below shows the functions 01 and 02 . At the
socially-optimal outcome these are equal. Note that now the individual optimum
exceeds the socially-optimal level. While 01 is the marginal social benefit, 02 is the
marginal social cost. In deciding the level of x, agent 1 does not take this social cost
into account and, therefore, the level of x chosen by agent 1 is too high relative to the
socially-optimal level.
176
12.3. Externalities
Figure 12.2: Negative externality.
12.3.1 Tax and quota policies

Impose a tax of t per unit on agent 1. Now agent 1s problem is to maximise 1 (x) tx.
The first-order condition is 01 (x) = t. Put t = 02 (x). Then the first-order condition
for optimisation by agent 1 is 01 (x) = 02 (x). We know that the solution to this is the
socially-optimal level xo . This is shown in Figure 12.3 below.
Figure 12.3: The optimal tax under a negative externality.
Other than tax, a quota at xo would also, of course, restore social optimality. However,
the tax and quota solutions depend on the government knowing a lot about the
preferences of the agents. Absent such information, the correct level of tax or quota
177
would be hard to gauge.
12.3.2 Coase theorem: the property rights solution
The Coase theorem is not really a theorem in the usual formal sense. It is an informal
statement that expresses an idea that Coase expounded in his famous article on the
social cost of externalities (Coase, 1960).
The idea is as follows. Consider a situation involving two parties where an externality
arises from the action of one party. The Coase theorem says that under conditions of
perfect competition (that is, under the standard assumption in the economic analysis of
competitive markets that there are no transaction costs of trading), if property rights
are well-defined, bargaining between parties would lead to the efficient outcome. This is
true irrespective of who the property rights are allocated to (to the party generating the
externality or to the party affected).
This approach says that it is unnecessary to provide incentives through taxes, subsidies
or quotas. Indeed, once property rights are well-defined, it allows agents to trade in the
externality-generating activity, and as in any other market under perfect competition,
all beneficial trading opportunities are exploited. Furthermore, taxes can therefore only
cause harm by distorting incentives. From a legal point of view, the implication is that
if we do have a situation of no transaction costs, legal institutions do not matter. If one
party generates, say, a negative externality for another, the law can hold either party
liable for the damages and it would still result in an efficient outcome.
Coases argument shows that it might be logically inconsistent to treat externalities as a
departure from otherwise perfect markets. If markets are indeed perfect, and some legal
liability regime is in place, rational agents should be able to use the same tools that
other markets use, to solve the problem of social cost from externalities. Under perfectly
competitive markets and a well-defined property rights regime, social costs therefore
coincide with private costs, and tax-based solutions are unnecessary and even
potentially harmful.
This is the more formal message in terms of economic theory. Coases intended message
was not that the problem of social cost does not exist or that legal institutions do not
matter. He intended to argue that transaction costs do matter and ignoring them
gives us predictions that are unrealistic.
However, both messages are important. As economists looking at policy, you should
carefully consider the case for a property rights based approach among other
alternatives. Indeed, such ideas have influenced the creation of various pollution rights
markets that have allowed reducing pollution in an efficient manner. Furthermore, you
should also be aware that different types of transaction costs do arise in applications,
and one must take that into account to judge the efficacy of a rights-based solution.
Let us now explore the idea of property rights formally in our model.
Suppose agent 2 has the right to an externality-free environment. If there is no
negotiation, agent 2 can then prevent agent 1 from taking any positive level of activity.
If the two agents negotiate, agent 2 can ask for a payment of T and allow agent 1 to
take some positive level of x. To determine T and x, consider agent 2s optimisation
problem. For any T and x, agent 2s payoff is 2 (x) + T . Agent 2 must maximise this
178
12.3. Externalities
subject to satisfying agent 1s participation constraint, which is 1 (x) T > 1 (0). Note
that agent 2 does not need to leave a surplus for agent 1, so she can choose T such that
agent 1s participation constraint binds. Therefore, agent 2s problem is to:
max 2 (x) + T subject to 1 (x) T = 1 (0).

T, x
Using the value of T from the constraint, the problem becomes:
max 2 (x) + 1 (x) 1 (0).

x
Since the last term is just a constant, the first-order condition is 02 (x) + 01 (x) = 0.
Therefore, the optimised value of x coincides with the socially optimal xo .
Next, suppose agent 1 has the right to choose any level of x. If there is no negotiation,
agent 1 would choose the previously-derived individually-optimal level x . If the two
agents negotiate, agent 1 can ask for a payment of T and reduce x to some level below
x . To determine T and x, consider agent 1s optimisation problem:
max 1 (x) + T subject to 2 (x) T = 2 (x ).

T, x
Using the value of T from the constraint, the problem becomes:
max 1 (x) + 2 (x) 2 (x ).

x
Since the last term is just a constant, the first-order condition is 01 (x) + 02 (x) = 0.
Therefore, the optimised value of x again coincides with the socially optimal xo .
The above analysis demonstrates that irrespective of who has the property rights, so
long as it is well-defined, costless negotiation leads to the socially-optimal outcome.
However, a variety of different transaction costs can arise in reality, limiting the force of
the solution in many cases. Even if there are two (or very few) parties, the parties might
not have all the relevant information about each other. Preferences might be private
information, a situation known formally as incomplete information. Under incomplete
information, bargaining is typically inefficient (i.e. not all gains from trade can be
exploited). Furthermore, even with just two parties, one or both could comprise of
many individuals (an upstream chemicals firm polluting a river, generating a negative
externality for the downstream community of fishermen). Negotiation then depends on
those individuals being able to speak with one voice through a leader. There might also
be free-rider problems if committing to a negotiation is costly in terms of time and
effort. Yet further, there might be many parties involved, making negotiation difficult or
even a complete non-starter.
Coases own work, coupled with his editorship of the Journal of Law and Economics
spanning a period of 17 years, helped create the field of law and economics. Indeed,
game theory and information theory allow questions about the design of legal
institutions, especially those concerned with tort law (contract law), to be very much
part of the field of enquiry of microeconomics.
As noted above, Coases ideas have influenced the creation of pollution rights markets
that play an important role in controlling harmful emissions. Below, we consider the
approach in more detail, and especially consider the United States market for permits
for sulphur dioxide emissions.
179
Note also that N&S mentions Obamas cap-and-trade policy proposals of 2009 in
Application 16.4. This is misleading in light of later events. The partisan policy stance
that has characterised American politics in recent years meant that the bill did not
stand any chance of passing and was withdrawn. Later, individual states were given the
responsibility of developing their own systems. We mention this in our discussion of
cap-and-trade below.
Example 12.1 Controlling pollution: the cap-and-trade policy.

An important policy in controlling pollution such as sulphur dioxide (SO2 ) or carbon
dioxide (CO2 ) emissions is to set an overall quota (a cap) for access rights and then
allow trade in pollution rights. Firms with a lower cost of abatement would then be
net sellers in the emissions trading market, while firms with higher abatement costs
would be net demanders. The market-equilibrium reallocation of allowances to the
higher abatement cost firms achieves any given cap at a lower cost than, say,
regulation which specifies a uniform quota across all pollution sources without the
possibility of trade.
Cap-and-trade has been used in the United States and Europe, as well as other
countries. The European Union (EU) Emissions Trading System (ETS) is the largest
multi-country, multi-sector greenhouse gas emissions trading system in the world. As
noted in the website of the UK Department of Energy and Climate Change, the EU
ETS includes more than 11,000 power stations and industrial plants across the EU
with around 1,000 of these in the UK. These include power stations, oil refineries,
offshore platforms and industries that produce iron and steel, cement and lime,
paper, glass, ceramics and chemicals. Tradable emission allowances are allocated to
participants in the market. One allowance gives the holder the right to emit one
tonne of CO2 (or its equivalent). Participants covered by the EU ETS must monitor
and report their emissions each year and surrender enough emission allowances to
cover their annual emissions.
Stavins (2011) notes that in the 1980s, leaded gasoline was phased out of the US
market with a program similar to cap-and-trade among refineries, saving about $250
million per year compared with a programme without trading. Since 1995, trade in
an SO2 allowance has reduced emissions by half, saving $1 billion per year compared
with a conventional approach.
President Obama had announced plans for setting a CO2 emissions trading scheme
in 2009, but partisan divisions among US lawmakers made that goal impossible and
the bill was withdrawn. Subsequently, the Obama administration announced that
states must develop their own policies to reduce carbon emissions from power
sources. California, the first state with a comprehensive cap-and-trade system,
started its programme in 2012. All indications are that this is working well, though a
fuller evaluation will only be possible in the future.
Other than the US (second) and the EU (third), the most important polluting
nations are China (first) and India (fourth). Both China and India have run pilot
CO2 emissions trading schemes. China has promised to introduce a nationwide
emissions trading programme by 2017. (See also the Wikipedia entry on Emissions
Trading for details of such schemes across different parts of the world.)
How would you know if a cap-and-trade policy is working well? The price of permits
180
12.3. Externalities
in the market is a good signal to judge the policy. A very high price implies that the
cap is very low, and the cost this imposes on firms might be very high. On the other
hand, a market price close to zero implies that there is little demand in the market
relative to supply, so that the system does not do much to reduce the costs of
abatement. In other words, in the latter case the cap is ineffective.
In reality, cap-and-trade has not been an unqualified success. In the EU ETS, there
were so many permits issued in the first phase (200507) that the permit price came
down to 0 in 2007. The later phases of the system have recovered from the collapse,
but there have been persistent criticisms alleging oversupply and consequently
low-price permits which, in turn, imply low incentives for reducing emissions.
Schmalensee and Stavins (2013) describe how the SO2 emissions trading system in
the US, which was very successful for the first decade of its operation starting in the
mid-1990s, collapsed later because of the wider regulatory environment.
As the authors note, beginning in 1995 and over the subsequent decade, the SO2
allowance trading programme performed exceptionally well along all relevant
dimensions. SO2 emissions from electric power plants decreased 36 per cent from
15.9 million to 10.2 million tonnes between 1990 and 2004, even though electricity
generation from coal-fired power plants increased 25 per cent over the same period.
By 2010, SO2 emissions had declined further, to 5.1 million tonnes.
The costs of achieving these environmental objectives with cap-and-trade were
significantly less than they would have been with a command-and-control regulatory
approach. Cost savings were at least 15 per cent, and perhaps as much as 90 per
cent, compared with counterfactual policies that specified the means of regulation in
various ways and for various portions of the programmes regulatory period. In
addition to efficiency achieved through trading, there is evidence that the
programme brought down abatement costs over time by providing incentives for
innovation and diffusion that were generally much stronger than those provided by
traditional command-and-control regulation.
As the authors point out, it was widely recognised by the late 1990s that the SO2
cap needed to be lowered significantly to attain environmental goals. The Bush
administration promulgated its Clean Air Interstate Rule in 2005, with the purpose
of lowering the cap on SO2 emissions significantly, and allowance prices rose in
anticipation of this lower cap. After peaking in 2005 at more than $1,200 per tonne,
SO2 allowance prices dropped just as fast as they had risen, aided by an
announcement from the US Environmental Protection Agency that it would
re-examine the Clean Air Interstate Rule and speculation about impending legal
challenges. The act was later vacated by the courts. The Obama administration
similarly tried and failed to reduce the cap through legislation. The end result is
that state-level and source-specific regulation has been imposed. As the authors
point out, the allowance market remains nominally in place, but the imposition of
prescriptive state-level and source-specific regulation has virtually eliminated the
demand for federal SO2 allowances. By the time of the Environmental Protection
Agencys 2012 auction, market-clearing prices had fallen to virtually zero.
181
Figure 12.4: SO2 allowance prices in the US, 19942012. Source: Schmalensee and Stavins
(2013).
12.4 Public goods

Public goods are goods that agents in an economy consume together and consumption
is neither rival nor excludable. Examples include national defence, roads, basic public
health, basic education, defence against natural catastrophes etc.
See N&S Sections 16.5 and 16.6 for basic concepts and an informal description of the
market failure problem: private provision leads to under-provision. Here we show this
more formally in a simple model.
Consider the following public good provision problem with n > 1 agents. Let xi denote
the contribution of agentPi, i {1, . . . , n}, and let xi denote the total contribution of
the other agents: xi = xj . Finally, let X be the total contribution of all agents, i.e.:
j6=i
X = xi + xi .
There is a constant marginal cost c > 0 of providing the public good.

We assume agent i has quasilinear utility ui (X) + mi , for i = 1, 2, . . . , n, where you can
think of the m-good as money. The total amount of the m-good is m. Furthermore,
u0i (X) > 0 and u00i (X) < 0. Note that an agent derives utility from not just his or her
own contribution but the total contribution. Indeed, this is the crucial distinction
between a private good and a public good.
12.4.1 Pareto optimum

Let us first show that the ParetoPoptimal level of X can be found simply by maximising
the sum of utilities minus cost: i ui (X) cX. This is a consequence of quasilinear
utility functions, which allow us to compare money costs and marginal utilities from X.
182
12.4. Public goods
We present the details of the argument below. However, once you have understood the
argument, you can put it at the back of your mind and simply use the result.
Under the quasilinear P
utility specified above, the Pareto optimal outcome can be found
simply by maximising i ui (X) cX.
Proof: For simplicity, let us show this assuming n = 2. The same method can be used
for any value of n. Note that having paid the cost cX, the total amount of the m-good
that is available for division between the agents is m cX.
With this in mind, and with two agents, the Pareto optimal outcome can be found by:
max u1 (X) + m1 subject to u2 (X) + m2 = u and m1 + m2 = m cX.

X, m1 , m2
Using the last constraint and substituting the value of m2 , this can be rewritten as:
max u1 (X) + m1 subject to u2 (X) + m cX m1 = u.

X, m1
Form the Lagrangian:
L = u1 (X) + m1 + (u2 (X) + m cX m1 u).
Derive the first order conditions:

L
= u01 (X) + (u02 (X) c) = 0
X
L
= 1 = 0.
m1
Thus = 1, which implies we are maximising u1 (X) + u2 (X) cX + m u. Since m u
is a constant, we are simply maximising u1 (X) + u2 (X) cX. From this, and also from
the first-order conditions, it is clear that the optimising condition is u01 (X) + u02 (X) = c.
The concavity of u1 and u2 ensure that the second-order condition for a maximum holds.
Pn
For n > 2, use the constraint mi = m cX to substitute the value of m1 and write
i=1
the problem as:
n
X
max u1 (X) + m cX mi
X, m2 ,...,mn
i=2
subject to ui (X) + mi = ui , for i = 2, 3, . . . , n. Now form the Lagrangian:

n
X n
X
L = u1 (X) + m cX mi + i (ui (X) + mi ui ) .
i=2 i=2
The first-order conditions

P with respect to mi imply i = 1 for i = 2, 3, . . . , n. Thus we
are simply maximising i ui (X) cX plus a constant independent of X, which we can
ignore.
Essentially, quasilinear utilities allow us to compare the utilities from X directly in
terms of money, which implies that one-for-one utility transfers can be made among
individuals, which leads to this simple formulation.
From the arguments above, we get the following.
183
The efficient provision of the public good is X o which satisfies the first-order
condition: X
u0i (X o ) = c.
i
12.4.2 Private provision

Let us show that private provision of the public good leads to a level of provision that is
lower than the efficient level. Suppose the public good is provided by a competitive
market. Each individual buys some amount of the public good at the market price p.
Note that any provision by an agent benefits all other agents as well.
Now, the payoff function of agent i is given by utility minus expenditure: ui (X) pxi ,
where u0i (X) > 0 and u00i (X) < 0, and p denotes price. Therefore, agent i faces the
problem:
max ui (xi + xi ) pxi .
xi
To simplify the algebra, let us assume that for each i, the level of provision is positive.
(In other words, we are ruling out the possibility that for some agents there is a corner
solution at zero since marginal utility is lower than price at any positive level of
provision. Taking this possibility into account complicates the algebra without changing
the conclusions.)
Since each persons maximum occurs at a positive x, the first-order condition holds with
equality for each i. Therefore, the first-order condition for any i is u0i (xi + xi ) = p,
which is the same as:
u0i (X ) = p.
Competitive supply implies p = MC = c. Therefore:
u0i (X ) = c.
o
or X ? Whereas efficiency requires i u0i (X o ) = c, under private
P
Which is larger, XP
provision we have i u0i (X ) = nc > c. Therefore:
X X
u0i (X ) > u0i (X o ).
i i
But
P since u00i
< 0, u0i (X)
is a decreasing function of X for each i. This implies that
0
i ui (X) decreases as X rises. It follows that:
Xo > X
which implies that private provision of a public good leads to an inefficiently low
provision.
Consider
P 0 o the simple case in which individuals are identical so that ui = u. Then
0 o
i ui (X ) = nu (X ). Then the socially-optimal provision is given by:
nu0 (X o ) = c
while the private provision level is given by:
u0 (X ) = c.
Figure 12.5 shows X o and X . Clearly, X o is larger than X .
184
12.5. The commons problem
Figure 12.5: Private provision of a public good. The public good is provided at level X
under private provision while the socially-optimal level is X 0 .
12.5 The commons problem
As noted in the last section, public goods are neither excludable nor rival. Private
goods, on the other hand, are both rival and excludable. However, many resources are
rival but not excludable. This gives rise to an over-exploitation issue known as the
commons problem or the tragedy of the commons.
Section 5.7 of N&S presents a simple game where the Nash equilibrium shows the
tragedy of the commons. In this section we develop a further simple model of resource
extraction.
12.5.1 Evidence
Stavins (2011) notes that the consequences of open access predicted by theory have
been validated repeatedly with empirical data. A study of the Pacific halibut fishery in
the Bering Sea estimated that the efficient number of ships was nine, while the actual
number was 140. An examination of the New England lobster fishery found that in 1966
the efficient number of traps set would have been about 450,000, while the actual
number was nearly one million. Likewise, an analysis of the North Atlantic stock of
minke whale found that the efficient stock size was about 67,000 adult males, whereas
the open-access stock had been depleted to 25,000. In terms of social costs, an analysis
of two lobster fisheries in eastern Canada found that losses due to unrestricted entry
amounted to about 25 per cent of market value of harvests, due mainly to excess
deployment of resources for harvest, with fishery effort exceeding the efficient level by
some 350 per cent.
185
12.5.2 A simple model of resource extraction

A common property resource is being used by n agents to produce an output Q. The
Pn
effort expended by agent i in extracting the resource is denoted by Li . Let L = Li .
i=1
The production function is Q = L. Let c be the constant marginal cost of extraction
facing any agent. Finally, suppose the output share of agent i is given by:
Li
qi = Q.
L
Socially-optimal level Lo
Let us first work out the socially-optimal level of L. This is given by L0 such that:
1
=c
2 L0
implying L0 = 1/4c2 .
Nash equilibrium level L

Here qi = Li Q/L = Li / L. We can write L as Li + Li . The problem of agent i is:
Li
max cLi .
Li Li + Li
The first-order condition is:
1 1 Li
c = 0.
L 2 L3/2
Adding over i, we have:
n 1 L
nc = 0
L 2 L3/2
which implies:
1 1
n = nc
L 2
which simplifies to: 2
1 1
L= 2 1
c 2n
which is the Nash equilibrium level. Note that L increases as n increases. The efficient
solution obtains when n = 1. Therefore, with more than one agent, we have an
inefficiently high extraction of the resource.
Privatisation solution
As we know from above, the efficient solution is obtained when n = 1. If a single agent
is assigned the right to extract, efficient extraction results. However, for certain types of
commons resources such as fishing assigning property rights to a single agent may
not be feasible.
186
Market for extraction rights: individual transferable quotas
Just as in pollution control (described above), the government could set an overall
quota and issue individual transferable quotas (ITQs) that can be traded. Trading ITQs
ensures that these go to those with the lowest cost of exploitation.
Stavins (2011) writes: ITQ systems have been used successfully in some 150 major
fisheries of 170 species in seventeen countries some with very significant fishing
industries, such as Australia, Canada, Iceland, and New Zealand. . . . In fact, New
Zealand regulates virtually its entire commercial fishery this way. Since 1986, the system
has been effective, largely eliminating overfishing, restoring stocks to sustainable levels,
and increasing fishermens profits. Several ITQ systems are in operation in the United
States, including ones for Alaskas Pacific halibut and Virginias striped-bass fisheries.
Tax
Consider a tax of t per unit of extraction. This raises the marginal cost of extraction to
c + t so that, after the tax, the equilibrium level of L is:
2
1 1
Lt = 1 .
(c + t)2 2n
Set t such that Lt = L0 , i.e. we have:
2
1 1 1
1 = 2
(c + t)2 2n 4c
which gives us:
n1
t= c.
n
This tax would restore efficiency. Note, however, that setting a tax optimally requires a
lot of information, which the policymaker might not have.
Informal cooperation
Note that the Nash equilibrium in the game above is akin to the Prisoners Dilemma
equilibrium. If all agents could agree to hold back their extraction, everyone would be
better off. However, more exploitation than this level is best for each individual, so that
no-one holds back in equilibrium.
Just as in the Prisoners Dilemma, there is scope for cooperation in equilibrium if the
game is repeated. This type of approach can explain how several rural societies,
especially in third-world countries, have been able to overcome a variety of problems
involving local commons resources.

to:
187
explain the concept of externalities and show how externalities lead to market
failures
analyse remedial policies based on taxes or subsidies
explain the content of the Coase theorem and analyse the Coasian property rights
approach to solving the problem of externalities
explain how an overall quota can be implemented efficiently using cap-and-trade
systems, how such systems have reduced pollution from several sources and how
such systems can sometimes run into difficulties
explain the concept of public goods and analyse how private provision leads to an
inefficiently low provision of such goods
explain the concept of the tragedy of the commons and demonstrate how strategic
behaviour leads to over-exploitation of open-access resources in simple models
explain the policy responses to commons problems.

1. Is a tax on cigarettes likely to induce the market for cigarettes to be more efficient?
Is a ban on smoking likely to promote greater efficiency?
2. Your neighbour has an electrical device that is costly to run, but kills insects very
effectively. The longer they keep the device on, the fewer insects appear in your
garden.
(a) Explain the externality in this scenario.
(b) Adapt the simple model developed in this chapter of the subject guide to
reflect the underlying externality in this situation and show formally how a tax
policy would work.
(c) Show formally how a Coasian property rights solution would work.
3. Explain how an emissions trading system operates and discuss its merits.
4. Explain why private provision leads to an inefficient provision of a public good.
5. Consider an environment where two firms operate with the following profit
functions:
1 = pq1 3q12 q2
2 = pq2 2q22 .
188
(a) Compute the level of production that each firm chooses when maximising its
profit individually.
(b) Compute the efficient level of production for each firm.
(c) Compute the per-unit tax that induces the efficient level of production. Do you
have to impose a tax on both firms?
6. This is a more difficult question. Consider the following two-player public goods
game. Players decide simultaneously whether to contribute to the public good. Each
player derives a benefit of 1 if at least one of them contributes and zero otherwise.
The cost of contribution for each player i = 1, 2 is given by ci and is known only to
that player. Player 1 believes that c2 is distributed uniformly on the interval [0, 2].
Similarly, player 2 believes that c1 is distributed uniformly on the interval [0, 2].
Consider the following type of strategy. Each player contributes if their cost is
below a cut-off c .
Compute the Nash equilibrium of this game. What is the likelihood of the public
good being provided in this equilibrium?
189
Notes
Notes
Notes
Comment form
We welcome any comments you may have on the materials which are sent to you as part of your
study pack. Such feedback from students helps us in our effort to improve the materials produced
for the International Programmes.
If you have any comments about this guide, either general or specific (including corrections,
non-availability of Essential readings, etc.), please take the time to complete and return this form.
Title of this subject guide:
Name
Address
Email
Student number
For which qualification are you studying?
Comments
Please continue on additional sheets if necessary.
Date:
Please send your completed form (or a photocopy of it) to:

Publishing Manager, Publications Office, University of London International Programmes,
Stewart House, 32 Russell Square, London WC1B 5DN, UK.

EC2066 Vle

Uploaded by

Copyright:

Available Formats

EC2066 Vle

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

EC2066 Vle

Uploaded by

Copyright:

Available Formats

What are the aims and learning outcomes of the course?

What are the aims and learning outcomes of the course?

What resources are available for students to use for their learning?

What resources are available for students to use for their learning?

Microeconomics

University of London International Programmes

Published by: University of London

2.4.1 The impact of income and price changes . . . . . . . . . . . . . . 20

3 Choice under uncertainty 37

4.1.4 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5 Production, costs and profit maximisation 79

5.6.2 Diminishing MRTS . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6 Perfect competition in a single market 93

6.9 Test your knowledge and understanding . . . . . . . . . . . . . . . . . . . 109

7 General equilibrium and welfare 111

10 Asymmetric information: adverse selection 149

11 Asymmetric information: moral hazard 161

12 Externalities and public goods 173

12.7 Test your knowledge and understanding . . . . . . . . . . . . . . . . . . . 188

1.1 Routemap to the subject guide

Welcome to this course in Microeconomics.

1.2 Introduction to the subject area and prior

the assumptions, internal logic and predictions of economic models

how economic principles can be applied to solve particular economic problems.

interaction and interaction under asymmetric information. Finally, it investigates the

Consumer choice and demand, labour supply.

Choice under uncertainty: the expected utility model.

Market structure: competition, monopoly and oligopoly.

General equilibrium and welfare: competitive equilibrium and efficiency.

Pricing in input markets.

Intertemporal choice: savings and investment choices.

The economics of information: moral hazard and adverse selection, resulting

1.4 Aims of the course

1.5 Learning outcomes for the course

be able to define and describe:

the nature of economic interaction under asymmetric information

be able to analyse and assess:

be prepared for further courses which require a knowledge of microeconomics.

1.6 Overview of learning resources

1.6.1 The subject guide

1.6.2 Essential reading

Nicholson, W. and C. Snyder Intermediate Microeconomics and its Application.

Henceforth in this subject guide this textbook is referred to as N&S.

Nicholson, W. and C. Snyder Theory and Applications of Intermediate

1.6.3 Further reading

1.6.4 Online study resources

1.6.5 The VLE

1.6.6 Making use of the Online Library

1.7 Examination advice

1.7.1 Format of the examination

1.7.2 Types of questions

1.7.3 Specific advice on approaching the questions

Prepare thoroughly for the examination by attempting the problems/questions in

Good answers to most questions require relevant assumptions to be stated and

Remember, it is important to check the VLE for:

up-to-date information on examination and assessment arrangements for this course

2.1.1 Aims of the chapter

2.1.2 Learning outcomes

explain the implications of the assumptions on the consumers preferences