EC3099

Undergraduate study in Economics,
Management, Finance and the Social Sciences
Industrial
economics
G. Symeonidis and
P. Schiraldi
EC3099
2023
Industrial economics
G. Symeonidis and P. Schiraldi
EC3099
2023
Undergraduate study in
Economics, Management,
Finance and the Social Sciences
This subject guide is for a 300 course offered as part of the University of London’s
undergraduate study in Economics, Management, Finance and the Social Sciences. This is
equivalent to Level 6 within the Framework for Higher Education Qualifications in England,
Wales and Northern Ireland (FHEQ).
For more information, see: london.ac.uk
This guide was prepared for the University of London by:
P. Schiraldi, Lecturer in Economics, Department of Economics, London School of
Economics and Political Science
It draws on material in the previously published editions of the guide by:
G. Symeonidis, Reader in Economics, Department of Economics, University of Essex.
This is one of a series of subject guides published by the University. We regret that
due to pressure of work the authors are 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 communicate these through the discussion
forum on the virtual learning environment.
University of London
Publications Office
Stewart House
32 Russell Square
London WC1B 5DN
United Kingdom
london.ac.uk
Published by: University of London

© University of London 2023
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
Introduction............................................................................................................. 1
Route map to the subject guide...................................................................................... 1
Introduction to the subject area...................................................................................... 1
Aims and objectives........................................................................................................ 2
Learning outcomes......................................................................................................... 2
Employability outcomes.................................................................................................. 2
Syllabus.......................................................................................................................... 2
Knowledge required for this subject................................................................................ 3
The structure of this subject guide................................................................................... 3
Overview of learning resources....................................................................................... 4
Examination advice........................................................................................................ 8
Chapter 1: Review of game theory.......................................................................... 9
Introduction................................................................................................................... 9
Learning outcomes......................................................................................................... 9
Essential reading............................................................................................................ 9
Further reading............................................................................................................... 9
Description of a game .................................................................................................... 9
Nash equilibrium.......................................................................................................... 10
Static games................................................................................................................. 11
Dynamic game with finite horizon................................................................................. 14
Dynamic games with infinite horizon............................................................................. 17
A reminder of your learning outcomes........................................................................... 19
Sample examination questions...................................................................................... 19
Chapter 2: Size and structure of firms.................................................................. 21
Introduction................................................................................................................. 21
Learning outcomes....................................................................................................... 21
Essential reading.......................................................................................................... 22
Further reading............................................................................................................. 22
Technological factors.................................................................................................... 22
Transaction costs, incomplete contracts and integration................................................. 23
Property rights.............................................................................................................. 27
Empirical evidence........................................................................................................ 30
Chapter 3: Separation of ownership and control.................................................. 33
Introduction................................................................................................................. 33
Further reading............................................................................................................. 33
Managerial incentives................................................................................................... 34
Limits to managerial discretion..................................................................................... 36
The profit maximisation hypothesis............................................................................... 37
i
EC3099 Industrial economics
Chapter 4: Short-run price competition................................................................ 39

Introduction................................................................................................................. 39
Further reading............................................................................................................. 40
The basic Bertrand model............................................................................................. 40
Bertrand competition with capacity constraints.............................................................. 41
The Cournot model....................................................................................................... 44
Chapter 5: Dynamic price competition.................................................................. 49
Introduction................................................................................................................. 49
Further reading............................................................................................................. 50
Modelling collusion...................................................................................................... 50
Price wars: theories and evidence.................................................................................. 54
Econometric analysis of market power and collusive behaviour...................................... 56
Chapter 6: Entry deterrence, entry accommodation and predation...................... 61
Introduction................................................................................................................. 61
Further reading............................................................................................................. 62
Strategic substitutes and strategic complements............................................................ 62
First-mover advantage, entry deterrence and entry accommodation............................... 63
A taxonomy of business strategies................................................................................ 66
Predation..................................................................................................................... 68
Chapter 7: Product differentiation and non-price competition ........................... 75
Introduction................................................................................................................. 75
Further reading............................................................................................................. 76
Horizontal product differentiation................................................................................. 76
Brand proliferation as an entry deterrence strategy........................................................ 81
Vertical product differentiation...................................................................................... 83
Markets with asymmetric information........................................................................... 85
Chapter 8: Price discrimination............................................................................. 91
Introduction................................................................................................................. 91
Further reading............................................................................................................. 92
First-degree price discrimination................................................................................... 92
Third-degree price discrimination.................................................................................. 93
Second-degree price discrimination .............................................................................. 94
ii
Contents
A reminder of your learning outcomes......................................................................... 100

Sample examination questions.................................................................................... 100
Chapter 9: Vertical relations................................................................................ 101
Introduction............................................................................................................... 101
Learning outcomes..................................................................................................... 101
Essential reading........................................................................................................ 102
Further reading........................................................................................................... 102
Efficiency arguments for vertical restraints................................................................... 102
Vertical restraints as instruments that restrict competition........................................... 107
Policy towards vertical restraints................................................................................. 109
Chapter 10: The determinants of market structure............................................. 115
Introduction............................................................................................................... 115
Further reading........................................................................................................... 116
Basic concepts............................................................................................................ 116
Theory of market structure in exogenous sunk cost industries...................................... 117
Theory of market structure in endogenous sunk cost industries.................................... 119
Empirical evidence...................................................................................................... 121
Chapter 11: Competition and industrial policy................................................... 125
Introduction............................................................................................................... 125
Further reading........................................................................................................... 125
Competition policy: objectives and difficulties in design and implementation................ 126
Competition policy in practice..................................................................................... 128
Policy towards research and development................................................................... 132
Chapter 12: Regulation....................................................................................... 137
Introduction............................................................................................................... 137
Further reading........................................................................................................... 137
Regulation of firms with market power under symmetric information........................... 138
Regulation under asymmetric information................................................................... 139
Dynamic issues in regulation and regulatory capture................................................... 142
Liberalisation and regulation....................................................................................... 143
Appendix 1: Sample examination paper............................................................. 147
Appendix 2: End-of-chapter Activities reading list............................................. 151
Chapter 1................................................................................................................... 151
Chapter 2................................................................................................................... 151
Chapter 3................................................................................................................... 151
Chapter 4................................................................................................................... 151
iii
Chapter 5................................................................................................................... 152

Chapter 6................................................................................................................... 152
Chapter 7................................................................................................................... 152
Chapter 8................................................................................................................... 152
Chapter 9................................................................................................................... 152
Chapter 10................................................................................................................. 153
Chapter 11................................................................................................................. 153
Chapter 12................................................................................................................. 153
iv
Introduction
Introduction
Route map to the subject guide

This subject guide has been written in such a way that you can obtain a
basic understanding of the topics in the syllabus from the guide, before
going on to read the various recommended texts to broaden and deepen
this understanding. This is necessary because there is no single text that
covers all the topics and also because some of the best texts in industrial
economics are written at a level rather more advanced than is required for
this course. The subject guide was therefore also designed to give you a
clear indication of the level of analysis that will be expected of you in the
examination.
The guide is, however, not a substitute for the careful study of the readings
listed at the beginning of each chapter. Its purpose is to help you organise
your study and give you a starting point for each topic in the syllabus.
While there is no single best way to organise your study for this course,
it may be useful, for each topic in the syllabus, to start with the relevant
chapter of the guide, then do the recommended reading for that particular
topic, then come back to the guide and attempt the Sample examination
questions at the end of the chapter.
Each chapter of the guide contains Learning outcomes to enable you to
check your progress. Each chapter also contains Activities and Sample
examination questions. There are two types of Activity. Some are self-test
questions designed to test your understanding of the material contained
in the chapter and/or the recommended reading. You are strongly
encouraged to attempt these questions as well as the Sample examination
questions. Simply reading the guide and the recommended texts is not
sufficient to maximise your benefit from the subject or to prepare for the
examination.
Other Activities ask you to study additional material (including internet
resources) and are mainly designed to deepen your understanding of
the links between economic theory and empirical evidence: case studies,
results from laboratory experiments and mini-surveys of the evidence on
particular issues. These Activities are intended to be optional. They are
there to give you the opportunity to find out more about topics of interest
to you.
Introduction to the subject area

This subject guide provides an introduction to current theory and
empirical work in industrial economics. It starts by examining game
theory and the internal structure of firms. It then moves on to the analysis
of various aspects of strategic interaction among firms in both price and
non-price dimensions and the determinants of industrial structure. Finally,
it discusses the role of policy in the context of competition and industrial
policies and regulation.
The emphasis throughout will be on understanding how the theoretical
tools can be used to analyse real-world issues. The theory will be
examined against empirical evidence, and its implications for public policy
and business strategy will be discussed.
1
Aims and objectives

This course aims to:
• provide you with the analytical skills required for understanding
problems in industrial economics, including applications of game
theory
• examine the key questions on the internal organisation of firms
• analyse various aspects of strategic interaction between firms and the
determinants of industrial structure
• provide you with the ability to apply economic models of firm
behaviour to analyse questions in business strategy, competition policy
and regulation.
Learning outcomes
At the end of the course, and having completed the Essential reading and
Activities, you should be able to:
• model game-theoretical situations involving payoff interdependence
and analyse static and dynamic games of complete information
• describe and explain the determinants of the size and structure of
firms and the implications of the separation of ownership and control
• describe and explain the pricing behaviour by firms with market power
and its welfare implications
• apply analytical models of firm behaviour and strategic interaction
to evaluate various business practices, including tacit collusion, entry
deterrence, product differentiation, price discrimination and vertical
restraints
• recognise and explain the basic determinants of market structure and
the key issues in competition policy and regulation.
Employability outcomes
Below are the three most relevant skill outcomes for students undertaking
this course which can be conveyed to future prospective employers:
1. complex problem-solving
2. decision making
3. adaptability and resilience.
Syllabus
This course provides an introduction to current theory and empirical
work in Industrial economics. It starts by examining the internal structure
of firms. It then moves on to the analysis of various aspects of strategic
interaction between firms and the determinants of industrial structure.
Finally, it discusses the role of policy in the context of competition and
industrial policies and regulation. The emphasis will be throughout on
understanding how the theoretical tools can be used to analyse real world
issues. The theory will be confronted against empirical evidence, and its
implications for public policy and business strategy will be discussed.
2
Introduction
Knowledge required for this subject

Knowledge of microeconomic analysis at an intermediate level is necessary
for students taking EC3099 Industrial economics. This subject guide
assumes that you are fully familiar with the theory of costs, the analysis of
alternative market structures such as perfect competition, monopoly and
oligopoly at an intermediate level, and concepts from consumer theory
and welfare economics.
EC3099 Industrial economics makes considerable use of game theory.
For the purposes of this subject only some knowledge of elementary game
theory is required. The emphasis is on using the game-theoretic techniques
in applications, so you do not need to worry about abstract definitions.
You must be familiar with the concepts of Nash equilibrium in static games
and subgame-perfect Nash equilibrium in dynamic games. However, you
do not need to have any knowledge of games with incomplete information
for this course.
Finally, the algebra required is simple calculus. Most of the mathematical
problems you will be faced with in this course are simple maximisation
problems. Regarding statistics, only some knowledge of elementary
probability theory is required. Some familiarity with basic econometric
techniques will help you better understand and appreciate some of the
empirical readings, although it is not essential for the examination.
The structure of this subject guide

This subject guide is divided into three parts. The first part analyses topics
in the theory of the firm.
Chapter 1 will review the key concepts in game theory used in this
guide.
Chapter 2 looks at the theory and evidence on the factors determining
the size and structure of firms. It reviews the technological view of the
firm, and then focuses on the transaction costs–property rights approach.
The links between investment specificity, contracts and vertical integration
are discussed. The chapter ends with a review of empirical evidence.
Chapter 3 is concerned with the implications of the separation of
ownership and control in modern large firms. A number of issues are
examined, including managerial incentives, the limits to managerial
discretion, and the foundations of the profit-maximisation hypothesis.
The next part examines various aspects of oligopolistic interaction.
Chapter 4 is about short-run competition among firms. Alternative
models are discussed and compared, including the Bertrand model, the
model of price competition with capacity constraints, and the Cournot
model.
Chapter 5 looks at dynamic price competition. Topics analysed include
repeated interaction, collusion and cartel stability, and theories of price
wars. The chapter concludes with a discussion of empirical analyses of
market power and collusive behaviour.
Chapter 6 focuses on situations where firms can act strategically to
influence the decisions of other firms. It examines first-mover advantages
and the value of irreversible decisions, strategies to deter entry or induce
exit, and concludes with a useful framework for classifying business
strategies.
3
Chapter 7 addresses the specific issues associated with product

differentiation and non-price competition. It analyses both horizontal and
vertical product differentiation, and shows, in the context of a particular
case study, how this analysis can help understand brand proliferation as a
strategy to deter entry. This chapter also includes an analysis of markets
with asymmetric information.
Chapter 8 reviews the various types of price discrimination by firms and
discusses applications, such as the use of tie-in sales.
Chapter 9 focuses on vertical relationships between firms and reviews
several reasons why firms may want to use vertical restraints. Explanations
of vertical restraints that emphasise efficiency gains and others that
emphasise welfare losses through the restriction of competition are
analysed. Empirical evidence on vertical restraints is discussed.
Chapter 10 deals with the determinants of market structure. Specific
topics include the links between competition and market structure, and
technology and market structure. The theory is tested against empirical
evidence.
The final part of the guide focuses on public policy towards industry.
Chapter 11 reviews competition policy in the EU, the USA and Japan,
and examines the main economic issues in the design and implementation
of competition policy. Industrial policy towards R&D is also discussed.
Chapter 12 examines topics in the regulation of firms with market
power, both under symmetric information and under asymmetric
information. It also examines competition in regulated industries, and
provides some empirical evidence on these issues.
Overview of learning resources

Esssential reading
Books
There is no single text that covers all the topics in the syllabus. The list
below contains several references. An excellent reference for many of the
topics covered is:
Tirole, J. The theory of industrial organization. (Cambridge, MA: MIT Press,
1988) [ISBN 9780262200714].
This book mainly focuses on theory. Moreover, certain parts of it contain
material which is too advanced for this course. As a general rule, you do
not have to worry about any material contained in the appendices. Finally,
not all chapters of the book are relevant for the syllabus of this course.
However, Tirole’s text does an excellent job of combining simple formal
economic analysis with a rich informal discussion of many important
issues in industrial economics. Several of the chapters in the subject guide
use or build upon examples taken from Tirole’s book.
Another very good reference which covers most topics in the syllabus at a
less advanced level than Tirole’s text is:
Church, J.R. and R. Ware Industrial organization: a strategic approach. (Boston:
Irwin McGraw-Hill, 2000) [ISBN 9780256205718].
This book emphasises strategic behaviour and covers a wide range of
topics in great detail (once again, not all of these topics are included in the
syllabus of this course). It also contains a large number of case studies to
motivate or illustrate the economic analysis.
4
Introduction
Two other books are useful for particular parts of the syllabus. For
Chapter 10, you may refer to:
Sutton, J. Sunk costs and market structure. (Cambridge, MA: MIT Press, 2007)
[ISBN 978022693585].
And for Chapter 12, you may read certain parts of:
Armstrong, M., S. Cowan and J. Vickers Regulatory reform. (Cambridge, MA:
MIT Press, 1994) [ISBN 9780262011433)].
Detailed reading references in this subject guide refer to the editions of the
set textbooks listed above. New editions of one or more 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 books; 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.
Journal articles and book chapters

A number of journal articles or book chapters are included in the category
of Essential reading because they are discussed in some detail in the guide.
They are the following:
Porter, R.H. ‘A study of cartel stability: The Joint Executive Committee, 1880–
1886’, Bell Journal of Economics 14(2) 1983, pp.301–14.
Rey, P. and J. Stiglitz ‘The role of exclusive territories in producers’ competition’,
RAND Journal of Economics 26(3) 1995, pp.431–51.
Schmalensee, R. ‘Entry deterrence in the ready-to-eat breakfast cereal industry’,
Bell Journal of Economics 9(2) 1978, pp.305–27.
Sutton, J. ‘Market structure: Theory and evidence’, in Armstrong, M. and R.
Porter (eds) Handbook of industrial organization, Volume 3. (Amsterdam:
North-Holland, 2007) [ISBN 9780444824356]. Working paper version
available at: http://personal.lse.ac.uk/sutton/market_structure_theory_
evidence.pdf
These articles are all available in the Online Library.
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 text, 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 VLE and University of London
Online Library (see below).
There are several other textbook references for this subject. Some of the
most useful are:
Cabral, L. Introduction to industrial organization. (Cambridge, MA: MIT Press,
2000) [ISBN 9780262032865].
Carlton, D.W. and J.M. Perloff Modern industrial organization. (United
States: Pearson Addison Wesley, 2005) fourth (international) edition
[ISBN 9780321223418].
Pepall L., D. Richards and G. Norman Industrial organization: contemporary
theory and empirical applications. (Chichester: Wiley-Blackwell, 2014) fifth
edition [ISBN 9781118250303].
Shy, O. Industrial organization. (Cambridge, MA: MIT Press, 1995)
[ISBN 9780262691796].
The first three of these can be used to complement Tirole’s more formal
analysis with a more descriptive, yet rigorous, treatment of the various
topics. Cabral’s book is relatively concise, but still covers a lot of ground,
5
and contains a good discussion of the modern literature on market

structure. The book by Pepall, Richards and Norman is the most recent one
and provides a very good balance of theory and empirical applications.
Shy’s text is analytically more advanced than the other three, with an
emphasis on simple formal economic models.
Two good collections of case studies in antitrust (competition) policy are:
Kwoka, J.E. and L.J. White (eds) The antitrust revolution: economics,
competition, and policy. (New York: Oxford University Press, 2013) sixth
edition [ISBN 9780199315499].
Lyons, B. (ed.) Cases in European competition policy. (Cambridge: Cambridge
University Press, 2009) [ISBN 9780521713504].
All the above books are useful for the course as a whole or as primary
references for particular topics. Additional references for particular topics
of the syllabus will be given at the beginning of each chapter. A full list of
Further reading can be found in the VLE.
Most current research in industrial economics is published in academic
journals. You have free access to a number of these via the University of
London Online Library. In addition to the numerous general economics
journals which regularly publish articles in industrial economics, there are
a number of specialist journals, including:
• Journal of Industrial Economics
• Journal of Economics & Management Strategy
• International Journal of Industrial Organization
• RAND Journal of Economics.
Information on competition issues – including news, reports and case
studies – is posted on the websites of competition authorities. For instance:
• European Union: ec.europa.eu/competition/index_en.html
• United Kingdom: www.gov.uk/government/organisations/competition-
and-markets-authority
• United States: www.ftc.gov (Federal Trade Commission) and www.
justice.gov/atr/ (Antitrust Division, Department of Justice).
Unless otherwise stated, all websites in this subject guide were accessed in
May 2023. We cannot guarantee, however, that they will stay current and
you may need to perform an internet search to find the relevant pages.
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.london.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 ‘Forgot
password’ link on the login page.
6
Introduction
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:
• Course materials: Subject guides and other course materials
available for download. In some courses, the content of the subject
guide is transferred into the VLE and additional resources and
activities are integrated with the text.
• Readings: Direct links, wherever possible, to essential readings in the
Online Library, including journal articles and ebooks.
• Video content: Including introductions to courses and topics within
courses, interviews, lessons and debates.
• Screencasts: Videos of PowerPoint presentations, animated podcasts
and on-screen worked examples.
• External material: Links out to carefully selected third-party
resources.
• Self-test activities: Multiple-choice, numerical and algebraic
quizzes to check your understanding.
• Collaborative activities: Work with fellow students to build a body of
knowledge.
• Discussion forums: A space where you can share your thoughts
and questions with fellow students. Many forums will be supported by
a ‘course moderator’, a subject expert employed by LSE to facilitate the
discussion and clarify difficult topics.
• Past examination papers: We provide up to three years of past
examinations alongside Examiners’ commentaries that provide guidance
on how to approach the questions.
• Study skills: Expert advice on preparing for examinations and
developing your digital literacy skills.
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.
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 cited 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
using the ‘Chat with us’ function.
7
Examination advice
Important: the information and advice given here are based on the
examination structure used at the time this guide was written. Please note
that subject guides may be used for several years. Because of this you are
strongly advised 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.
This subject is assessed by a three-hour examination. This currently
consists of eight questions divided into two sections, each of which has
four questions. Section A includes essay-type questions, while Section
B includes problem-type questions. You will be required to answer four
questions, two from each section. A Sample examination paper is included
at the end of this guide.
Problem-type questions are quite specific as to what you are required to
do, and a good answer generally involves some use of mathematics. Essay-
type questions are sometimes less specific, but a good answer to an essay-
type question must include some rigorous economic analysis, usually with
reference to an economic model or models.
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.
You can find guidance on examination technique in the annual Examiners’
commentaries for the course which are available on the VLE.
8
Chapter 1: Review of game theory
Introduction
Game theory is a formal approach to studying the strategic behaviour
of interacting, ‘rational’ decision-makers. The strategic behaviour arises
because of the presence of pay-off interdependence (i.e. the optimal
choice by an agent and its pay-off depends on the actions of the others).
This chapter will review and make you familiar with key aspects of
noncooperative game theory that have proved most useful in industrial
organisation.
The analysis is intentionally informal and is not meant to be a complete
overview or a substitute for a game theory textbook.
Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• describe and model game-theoretical situations involving payoff
interdependence
• identify and describe players’ strategies
• identify non-credible threats in dynamic games.
• explain the notions of Nash equilibrium and subgame-perfect Nash
equilibrium
• analyse and solve static and dynamic games of complete information.
Essential reading
Church, J.R. and R. Ware Industrial organization: a strategic approach. Chapters
7 and 9
Tirole, J. The theory of industrial organization. Chapter 11.
Further reading
Gibbons, R. Game theory for applied economics. (Princeton, NJ: Princeton
Osborne, M.J. and A. Rubinstein A course in game theory. (Cambridge, MA: MIT
Press, 1994) [ISBN: 9780262650403].
Fundenberg, D. and J. Tirole Game theory. (Cambridge, MA: MIT Press, 1994)
[ISBN 9780262061414].
Kreps, D. and R. Wilson ‘Sequential equilibrium’, Econometrica 50(4) 1982,
pp.863–94.
Description of a game
A game is described by four elements:
1. Players: the identity of those playing the game.
2. Rules: the timing of all players’ moves; the actions available to a
player at each of the moves; and the information that a player has at
each move.
3. Outcomes: the set of outcomes is determined by all of the possible
combinations of actions taken by players.
9
4. Payoffs: the payoffs of the game represent the players’ utilities over
the outcomes of the game.
Before formalising a game and discussing any solution concept, we need
to introduce some definitions and classifications. The strategy available to
the player is a complete plan of actions (what to do in every contingency).
In proceeding, we assume that the description of the game (the payoffs
and the strategies available to the player) is common knowledge. We also
assume that that it is common knowledge that each player is ‘fully rational’
(i.e. each player chooses a strategy that maximises their utility given their
subjective beliefs). Games can be classified on the basis of (i) the timing
of moves, and (ii) uncertainty about the payoffs of rivals. In complete
information games, players not only know their own payoffs, but also
the payoffs of all the other players. In a game of incomplete information,
players know their own payoffs, but some will not know the payoffs
of some of the other players. Here, we will mostly focus on complete
information games. A game is static if players move once at the same time
or dynamic if players move sequentially or repeatedly. In dynamic games
we can distinguish between games of perfect or imperfect information,
depending on whether or not all the players know the entire history of the
game when it is their turn to move.
A game can be described in a strategic or an extensive form.
The extensive form specifies the order of play, the information and action
available to a player, when it is their turn to move and the payoffs for all
players. Kreps and Wilson (1982) offer a more formal definition.
Player 1
S W
Player 2
S W S W
(2,2) (5,3) (3,5) (0,0)

Figure 1.1: Extensive form.
A simpler and more concise representation of strategic situations is the
strategic form. The strategic form specifies the players, the strategies,
the choice that each player can make and the payoffs associated with every
chosen combination of strategies.
SS Player 2
SW WS WW
Pl. 1 S 2,2 2,2 5,3 5,3
W 3,5 0,0 3,5 0,0
Figure 1.2: Strategic form of Figure 1.1.
Nash equilibrium
Is it possible to make a prediction about how the game should be played?
Are there obvious strategies that the players should adopt to maximise
their payoffs? To make an optimal decision, a player must generally form
an opinion about how their opponents will behave. A natural consistency
10
requirement is that other players’ choices coincide with the actual choices
they make. The Nash equilibrium satisfies this requirement. It is the most
common form of solution concept which is applied to analysed games of
strategic interaction.
A strategy profile s* ∈ S is a Nash equilibrium (NE) for n-person game
if, for each player, i, ui(s*)≥ ui(s*-i - si ) for all si ∈ Si where s*-i =(-s*1,…
-s*i-1, s*i+1,…,s*n )
Interpretation of Nash equilibrium (NE):
• Self-enforcing agreements: Suppose that the players in a game
can communicate prior to making their moves, but still cannot make
binding agreements. Then the only agreements that will survive
the playing of the game are agreements to play Nash equilibrium
strategies. Only at Nash equilibria will no player have an incentive to
deviate and break the agreement.
• Stable social conventions: Whether we walk on the right or left
of a pavement or an escalator is a social convention. Ignoring how
such conventions develop, we know that they are unlikely to persist if
they are not Nash equilibria, because players will have an incentive to
deviate. Once all pedestrians have decided to walk on the right side
of the pavement, anyone deviating would be knocked over, so this
convention does have the Nash equilibrium property.
• Rationality determines the obvious equilibrium: It has
been suggested that, in fact, rationality should result in the Nash
equilibrium strategies being played. Rational players who have the
same information about the game and know that each player is
rational should all agree on how the game will be played.
A special case of a NE is the one that involves dominant strategies.
A dominant strategy (if available) is a strategy that will lead to the best
outcome regardless of the strategies that the other players choose.
Static games
In this section, we analyse static games of complete information. In doing
so we will be using the normal representation exclusively.
In the following section we use the concept of this solution to predict the
outcome of a strategic interaction called the Prisoner’s Dilemma.
Prisoner’s Dilemma
Two members of a criminal gang are arrested and imprisoned. Each
prisoner is in solitary confinement with no means of speaking to or
exchanging messages with the other. The police admit they do not have
enough evidence to convict the pair on the main charge. They plan to
sentence both to two years in prison on a lesser charge. Simultaneously,
the police offer each prisoner the possibility of confessing or remaining
quiet with the following outcomes:
• If A and B both confess, each of them will serve five years in prison.
• If A confesses but B remains silent, A will serve one year and B will
serve 10 years in prison.
• If A remains silent, but B confesses, A will serve 10 years in prison and
B will serve only one year.
• If A and B both remain silent, both of them will serve two years in
prison (on the lesser charge).
11
Confess Quiet
Confess −5, −5 −1, −10

Quiet −10, −1 −2, −2
Figure 1.3: Representation in strategic form and Nash equilibrium.

The unique Nash equilibrium of the game is (Confess, Confess).
Notice that the NE is not necessarily efficient or welfare maximising. The
strategic interaction may lead to welfare loss as in the Figure 1.3 above,
where the equilibrium that maximises welfare is (Quiet, Quiet).
A NE always exists and it might involve mixed strategies. A mixed strategy
for a player is a probability distribution over the (pure) strategies.
Consider the game below, known as Matching Pennies:
Player 2
Player 1 Heads Tails
Heads +1, –1 –1, +1
Tails –1, +1 +1, –1
Figure 1.4: Matching pennies.

There are no NE in pure strategies, so we need to compute a NE in
mixed strategies. In Matching Pennies, a mixed strategy for player i is the
probability distribution (y, 1 − y), where y is the probability of playing
Heads, 1 − y is the probability of playing Tails, and 0 ≤ y ≤ 1. The mixed
strategy (0, 1) is simply the pure strategy Tails, and the mixed strategy
(1, 0) is the pure strategy Heads. In a mixed strategy NE, players must be
indifferent among all the actions that they play. If they are not equal, then
player i should play the highest-utility action always (not mixing).
Each player’s probabilities must be chosen to make the opponent
indifferent. In the Matching Pennies:
12
Player 1’s expected utility from Player 2’s mixed strategy
probability y 1–y
2
1 h t EU1:
H +1, –1 –1, +1 2y – 1
T –1, +1 +1, –1 1 – 2y
EU1(H) = y � 1 + (1 – y) � –1 = 2y – 1
EU1(T) = y � 1 + (1 – y) � –1 = 1 – 1
Player 2’s expected utility from Player 2’s mixed strategy
2
1 h t
probability
x H +1, –1 –1, +1
1–x T –1, +1 +1, –1
EU2: 1 – 2x 2x – 1
EU2(h) = x � –1 + (1 – x) � 1 = 1 – 2x
EU2(t) = x � 1 + (1 – x) � –1 = 2x – 1
In equilibrium EUi(H) = EUi(T) for both players, so this condition

determines a system of two equations in two unknowns which can be
solved to obtain x=1/2 and y=1/2.
13
Dynamic game with finite horizon

We now consider instances where players do not move simultaneously
but sequentially and/or the stage game (like the Prisoner’s Dilemma) is
repeated over time. In this section, we only consider finite horizon games,
that is, a game where the number of actions taken by the players is finite.
The central issue in dynamic games is the question of credibility. In many
games, players make threats of the form ‘If you do X (which I do not
like), then I will make you regret X by doing Y ’ (e.g. an incumbent firm
threatens to launch a price war against a new entrant, thereby rendering
entry unprofitable). The issue of credibility concerns the incentives to
carry through the threat. The key point is that if the threat is not in the
player’s interest, then it is noncredible and hence should not influence the
other players’ behaviour. This idea will be formalised using a refinement of
the NE.
In analysing dynamic games, it is more convenient to use the extensive
form of the game as it will be critical to know the information and choices
available to a player whenever it is their turn to play. As specified above,
we need to distinguish perfect or imperfect information games.
C Q
C Q C Q
(–5, –5) (–1, –10) (–10, –1) (–2, –2)

Figure 1.5: Sequential game with perfect information.
In Figure 1.5 above, Player 2 knows (observes) the move made by Player 1
and in which nodes they are in when it is their turn to move.
C Q
2
C Q C Q
(–5, –5) (–1, –10) (–10, –1) (–2, –2)

Figure 1.6: Sequential game with imperfect information.
Contrary to the scenario outlined above, Player 2 does not observe Player
1 move when it is their turn to move. Equivalently, this is also a way to
represent a simultaneous move game.
14
In a general extensive form game, there may be both nodes where players
know the exact past history of actions (singleton information sets) and
nodes that they cannot distinguish from one another (non-singleton
information sets). A strategy in general specifies an action for each
information set (singleton or non-singleton). Thus, a strategy specifies an
action for each possible observed past history of play
A subgame is a part of the game tree that is a game in its own right:
• In particular, whenever a node is part of the subgame, all players know
that it is.
• In other words, subgames are subtrees that are not cut through by
information sets.
L1 R1
2 2
L2 R2 L3 R3
1 1 1
L4 R4 L4 R4 L5 R5 L6 R6
2 3
L7 R7 L8 R8 L8 R8
Figure 1.7: The start of a subgame (highlighted with a red dot).

As discussed above, we want to predict the outcome of a strategic
interaction that does not involve a non-credible threat. Therefore we
consider subgame-perfect Nash equilibrium (SPNE). SPNE is a
refinement of the NE. A SPNE is a strategy profile that induces a Nash
equilibrium in every subgame. Thus, in an SPNE, all players find it
optimal to stick to their strategies at any point in the game (even off the
equilibrium path). This rules out non-credible threats. In order to find
SPNE, we need to solve the game from the bottom by solving for a NE in
every subgame (backward induction). Note that every SPNE is also a NE,
but not every NE is a SPNE.
Look again at Figure 1.1. It is not difficult to show (best using the normal
form, as in Figure 1.2) that there are three NE: 1
1. (S, WS) (i.e player 1 plays S and player 2 plays W if S and S if W)
2. (S, WW) (i.e player 1 plays S and player 2 plays W if S and S if W)
3. (W, SS) (i.e player 1 plays S and player 2 plays S if S and S if W).
Note the strategy of Player 2 is the plan of action in every possible
contingency, even those that will be never reached, given Player 1’s move.
Strategies for Player 2 seem ‘funny’ in (2) and (3) They involve non-
credible threats. In fact, the only SPNE is the first one. To compute the
SPNE, let us solve for the NE is every subgame starting from the bottom:
15
Step 1: Player 2 chooses W in the first subgame.
NE (1) is subgame perfect (S, [W if S, S if W])
Player 1
S W
S W S W
(2, 2) (5, 3) (3, 5) (0, 0)
Step 2: Player 2 chooses S in the second subgame.
Player 1
S W
S W S W
(2, 2) (5, 3) (3, 5) (0, 0)
Step 3: Player 1 chooses S in the third subgame.
Player 1
S W
S W S W
(2, 2) (5, 3) (3, 5) (0, 0)
16
Dynamic games with infinite horizon

Now, we consider dynamic games with infinite horizon. The infinite
horizon game – called ‘repeated game’ or ‘supergame’ – is an infinite
repetition of a stage game. For example, consider the supergame where
the stage game is the Prisoner’s Dilemma (see Figure 1.3).
Finite or infinite repetition? Nothing lasts forever, so it is not really
plausible to have a strategic interaction (a game) repeated an infinite
number of times. On the other hand, in real-world, repeated games, it is
also implausible that you will know how long the game will last (as the
finitely repeated model assumes). Therefore, the answer to this question
is to interpret an infinitely repeated game as one in which you do not
know when it finishes; thus the discount factor is then proportional to the
probability that the game will end in a given period.
Let the action chosen by player i in round t be ait. Let the payoffs to player
i, given his action ait and his opponent’s action ajt in the stage game,
be Πi(ait, ajt). Both players discount the future by the factor δ < 1 and
maximise the present discounted values of their profits:
∞
∑ δt−1 ∏i(ait, ajt)
t=1
Players’ strategies are functions specifying an action at each stage for each
possible history of play. A history here is a list of all the actions taken by
both players in all previous stages. Play in period t can be made contingent
on the history of the game. The history of the game becomes important
because players make it matter by basing their future behaviour on past
play. This raises the interesting possibility of explicitly reacting to and
punishing a ‘cheater’ which might allow the firms to sustain different
equilibrium outcomes than in static or dynamic games with finite horizon.
For example, in the Prisoner’s Dilemma, players can threaten retaliation
or promise future cooperation based on their opponent’s play today. Of
course, they will want to make sure that such threats and promises are
credible.
As for finite horizon game, we need to find SPNEs. However, since there
is no last period, backward induction cannot be applied in infinite games.
In addition, there is, in principle, an infinite number of infinitely complex
deviations to check to ensure a strategy profile is an equilibrium. How do
we find SPNEs? Fortunately, there is a convenient theorem based on the
single-deviation principle that gives us an infinite-horizon analogue
of backward induction.
A strategy profile s* in an infinitely repeated game is an SPNE if – and only
if – there is no player i and strategy si that agrees with si* except at a single
t and history ht and such that si is a better response to s-i* conditional on
history ht being reached.
Thus, we simply need to check that no player has an incentive to deviate
at any one stage game (following any given history), holding fixed all
other players’ strategies and our player’s actions at all other histories.
Is it possible to achieve cooperation in the Prisoner’s Dilemma game? We
know that the NE in the stage game is (Confess, Confess). Can repetition
lead to an outcome that differs from the stage game equilibrium? The
answer is ‘no’ if we consider a finitely repeated Prisoner’s Dilemma. By
backward induction, in the last period: there is no future, so both players
must play the stage game NE, (Confess, Confess). In the second-to-last
period, both players know that in the next period the other will confess,
17
so in this period there is no incentive to cooperate (there will be reward

next term!). Hence, both play Confess again; proceeding in this manner,
we find that in the unique SPNE both prisoners always confess, as in the
one-shot game.
Now suppose that the Prisoner’s Dilemma is repeated infinitely. It is easy
to see (by the single-deviation principle) that (Confess, Confess) is still an
SPNE. However, there may now be a more interesting outcome. Consider
the following strategy (grim trigger strategy) for both players:
• At t=1, play Quiet
• At t > 1, play Quiet if (Quiet, Quiet) has been chosen in all previous
stages; otherwise, play Confess.
To find out when these form an SPNE, we must apply the single-deviation
principle to all possible histories of play.
1. Following a history where somebody has already confessed:
• We know the opponent will always confess.
• The best response to this in any stage game is also to confess.
• Thus the player cannot profitably deviate at this history.
2. Following a history where no one has yet confessed:
• If at this history prisoner i remains quiet, cooperation will continue
indefinitely (under the proposed strategies), so i’s payoff will be:
• If at this history prisoner i confesses instead, he will get −1 in
this period and be relegated to the stage game NE profits of −5
thereafter; i’s payoff will be:
∞
−2
∏C = ∑ δt−1 (−2) = 1−δ
t=1
Remaining quiet is possible in equilibrium if:
1
δ≥
4
Thus, cooperative behaviour is sustainable in the infinitely repeated
Prisoner’s Dilemma when the prisoners are not very impatient.
This idea will be used to study tacit collusion in the Oligopoly Game.
Activities
1. Consider the following games. Identify the strategies available to each player and the relative
payoffs. Comment on their relationship to the game of Matching Pennies.
a. between the Internal Revenue Service and the accuracy of a taxpayer’s filing
b. a football player’s choice of which side to kick the ball during a penalty and the choice by
the goalkeeper of where to dive
c. between motorists and the police over speeding.
2. Until recently, medical doctors and lawyers have been prohibited from engaging in competitive
advertising. If the Prisoner’s Dilemma applies to this situation, how would the presence of this
restriction be likely to affect profits?
3. Climate change and sustainability are often at the forefront of current news reporting. Individuals
are becoming increasingly aware of the importance of protecting the environment. Every nation
and every individual benefits if others restrain their pollution, yet many people continue to pollute
and not recycle, thus harming the planet. Model this situation as a game. Identify the strategies
and the payoffs. Comment on its relation with the Prisoner’s Dilemma.
18
A reminder of your learning outcomes

• describe and model game-theoretical situations involving payoff
interdependence
• identify and describe players’ strategies
• identify non-credible threats in dynamic games.
• explain the notions of Nash equilibrium and subgame-perfect Nash
equilibrium
• analyse and solve static and dynamic games of complete information.
Sample examination questions

1. Find all Nash equilibria (pure and mixed strategy) of the games below:
a.
Mr Odd Couple
Boxing Ballet
Mrs Odd Couple Boxing (3, 1) (0, 0)
Ballet (0, 0) (2, 4)
b.
B
U D
A U (3, 3) (1, 4)
D (4, 1) (0, 0)
2. A philanthropist has to choose whether to contribute to University Z or

to University W. Each university is interested only in the contributions
it receives. The philanthropist announces that the contribution
will be decided by the following mechanism. The proposal is for a
contribution of £20,000 to University Z and £0 for University W. If
University Z accepts, the contribution is assigned. Otherwise the
philanthropist will increase the total contribution to £80,000 and
asks University W whether it wants to keep all of that sum for itself or
whether it prefers that the contribution is equally shared between the
two universities (i.e. £40,000 each).
a. Draw the extensive form of the game.
b. Find the subgame-perfect Nash equilibrium of the game.
c. Represent the game in its normal form.
d. Find the pure-strategy Nash equilibria.
3. Consider the following game:
Player 2
A B
Player 1 A (2, 2) (−2, 6)
B (6, −2) (0, 0)
a. Find all Nash equilibria of this game.
19
b. Assume that the above stage game is played infinitely many times.
After each round, players observe the moves made by the other
player. The total payoffs of the repeated game are the discounted
(with discount factor δ) sums of the payoffs obtained in each
round. For what values of the discount factor δ playing (A, A) in
every round is the Nash equilibrium?
20
Chapter 2: Size and structure of firms
Introduction
What explains the size and structure of firms? In fact, you may also ask:
why do agents group together to form firms? This chapter aims to provide
some answers to these questions and also to examine some more specific
issues, namely why some transactions take place within firms while
others are conducted through external contractual relationships; what
determines, in the case of contracts between firms, the types of contracts
used; and finally, what the implications of alternative ownership structures
are for efficiency.
This chapter focuses on efficiency explanations for the size and
structure of firms. It is important to understand what ‘efficiency’ means
in this context. Efficiency motives are those associated with minimising
costs or maximising producer surplus in a way that may also be socially
beneficial, that is increase total social welfare. The efficiency of a certain
organisational form refers then here primarily to the firm or firms
involved, not necessarily to society as a whole.
You should bear in mind that there are also market power explanations for
the size and structure of firms. Unlike efficiency motives, market power
motives induce behaviour which, while profitable for the firm or firms
involved, is definitely detrimental to social welfare. For instance, two firms
producing the same product may merge not to reduce costs but simply
to enhance their ability to exercise market power. Or a downstream firm
may purchase an upstream firm (a supplier of certain inputs) in order to
control supplies to downstream firms with which it competes. Much of
the second part of this subject guide is concerned with the behaviour of
firms that have market power. So, to make an overall assessment of the
factors that determine the size and structure of firms you should first work
through most of the guide.
There are two broad classes of efficiency explanations for the size and
structure of firms: the technological view of the firm and the transactions
costs-property rights approach.
Learning outcomes
• describe and evaluate two different approaches to explaining the size
and structure of firms
• explain the notions of ‘transaction costs’, ‘investment specificity’,
‘opportunistic behaviour’, ‘incomplete contracts’ and ‘residual rights of
control’ and their relevance for the theory of the firm
• explain the effect of investment specificity on the decision of firms to
enter into contractual relationships or to integrate
• analyse the factors that determine who should have control inside a
vertically integrated firm.
21
Essential reading
Church, J.R. and R. Ware Industrial organization: a strategic approach. Chapter 3.
Tirole, J. The theory of industrial organization. Introductory chapter.
Further reading
Books
Hart, O. Firms, contracts, and financial structure. (Oxford: Oxford University
Press, 1995) Chapters 1–4.
Perry, M.K. ‘Vertical integration: determinants and effects’, in Schmalensee,
R. and R. Willig (eds) Handbook of industrial organization, Volume 1.
(Amsterdam: North-Holland, 1989).
Journals
Aghion, P. and R. Holden. ‘Incomplete contracts and the theory of the firm:
What have we learned over the past 25 years?’, Journal of Economic
Perspectives 25(2) 2011, pp.181–97.
Hart, O. ‘An economist’s perspective on the theory of the firm’, Columbia Law
Review 89(7) 1989, pp.1757–74.
Hart, O. and J. Moore ‘Property rights and the nature of the firm’, Journal of
Political Economy 98(6) 1990, pp.1119–58.
Joskow, P.L. ‘Contract duration and relationship-specific investments: Empirical
evidence from coal markets’, American Economic Review 77(1) 1987,
pp.168–85.
Klein, B., Crawford, R.G. and Alchian, A.A. ‘Vertical integration, appropriable
rents, and the competitive contracting process’, Journal of Law and
Economics 21(2) 1978, pp.297–326.
Lafontaine and Slade ‘Vertical integration and firm boundaries: The evidence’,
Journal of Economic Literature 45(3) 2007, pp.629–85.
Monteverde, K. and D.J. Teece ‘Supplier switching costs and vertical integration in
the automobile industry’, Bell Journal of Economics 13(1) 1982, pp.206–13.
Technological factors
According to the technological view, optimal firm size and diversification
depend on the degree of economies of scale and scope.1 For instance, 1
See Tirole (1988),
a single-product firm may have an average cost curve such as the one pp.18–21 for details. On
the concepts of scale
depicted in Figure 2.1. To minimise average cost, the firm will in this
and scope economies,
case operate at a size between q1 and q2. Also, if x and y are quantities of you can also read
two different products, it is technologically more efficient for these to be Church and Ware
produced by a single firm rather by two separate firms if the cost function (2000), Chapter 3.
exhibits economies of scope: C(x, y) < C(x, 0) + C(0, y).
AC
AC(q)
q1 q2 q
Figure 2.1
22
Technological constraints are important, but they are not the whole story.
In particular, there are two problems with the technological view of the
firm:
• It is not clear why the AC curve rises at high output. If producing
quantity qA + qB were to cost more than producing qA and qB
separately, why can’t there be a single firm that consists of two
independent divisions producing qA and qB respectively?
• The technological view of the firm may explain the joint use of
facilities, but not joint ownership. Why can’t agents write contracts to
exploit economies of scale and scope without joint ownership?
Transaction costs, incomplete contracts and integration

The starting point for the transaction costs-property rights approach is
the idea that the choice between organising activity internally and using
the market (or contracts) is determined by a comparison of the costs and
benefits of these two modes of organisation. Williamson has identified
some economic factors that matter for this choice. There are three
elements in his approach:
1. investment specificity
2. opportunistic behaviour
3. bounded rationality.
Many long-run relationships between economic agents involve relation-
specific investments, in other words investments that pay off a
maximum return only if the particular relationship continues for some
time. Examples of specificity are site specificity (e.g. a firm builds a plant
next to another firm’s works to save on transport costs); physical asset
specificity (e.g. a firm designs equipment with characteristics specific to a
particular order); and human capital specificity (e.g. an employee invests
in acquiring skills which are specific to a certain job). In all these cases,
the first best use value of an investment is higher than its value in any
alternative use.
Now ex ante (i.e. before any investments are made) there is a competitive
situation. For instance, if the relationship is between a buyer and a
supplier of a certain product, there will be many suppliers and buyers
and they can select each other out of the pool of competitive suppliers
and buyers. But ex post (i.e. after investments have been made) there is a
bilateral monopoly, because if the parties trade with each other they can
make gains which will not be made otherwise. This creates the possibility
of a ‘hold-up’ or opportunistic behaviour. Each party wants to
appropriate the common surplus ex post, so there will be bargaining. This
can create several problems, in addition to any costs of haggling:
• The level of trade ex post may not be efficient if there is asymmetric
information. This can occur irrespective of whether the relation
involves ex ante investments or not.
• The level of investment ex ante will not be efficient, even under
symmetric information. The reason is that once a party has sunk the
cost of the investment, it has lost any extra bargaining power. So
even if the efficient volume of trade occurs ex post, the division of the
surplus will be such that the level of investment ex ante is not efficient.
These notions will be made more precise with the help of the formal
model below. But before that, consider the following question. If the
23
parties could write a contract ex ante specifying the terms of trade ex post,
would they be able to achieve efficiency? The answer is that they might.
So why can’t they write such a contract? This is where the bounded
rationality idea comes in: a complete contract is often impossible to
write. There are several reasons for this, such as unforeseen contingencies,
prohibitive costs of contracting over all contingencies (even if these are
foreseeable), or prohibitive monitoring costs. As a result, contracts are
incomplete. Some bargaining will have to take place ex post, and this may
lead to inefficiencies.
We can now state an important theoretical prediction of the transaction
costs approach: the more specific the investment, the larger the scope for
efficiency losses due to opportunistic behaviour. Hence the more specific
the investment, the higher the probability of integration (i.e. common
ownership) as opposed to a contractual relationship.
A model
Let us formalise some of the above ideas using a simple model of a vertical
relationship between a buyer and a seller.2 There are two periods, t = 1 2
This part follows Tirole
(ex ante) and t = 2 (ex post). We first want to focus on the issue of ex post (1988), pp.21–29.
efficiency, so assume for now that there is no investment ex ante. The two
parties can, if they wish, trade one unit of an indivisible good in period 2.
Let v denote the value of the good to the buyer (this can be the difference
between the value to the buyer in this relationship and that in an
alternative relationship), c the production cost, and p the price at which
the parties trade. If there is trade, the buyer has a gain of v – p, while the
seller has a gain of p – c. If there is no trade, both gain zero.
If v and c are known to both parties at the beginning of period 2, then the
volume of trade is efficient, which is another way of saying that trade will
occur if and only if there are gains from trade, in other words if and only if
v ≥ c. That is so because, if v ≥ c, the parties will agree to trade at a price
p such that c ≤ p ≤ v rather than not trade and make zero surplus.3 While 3
If v = c, the parties are
if v < c, at least one party would be making negative surplus if trade were indifferent between trading
at p = c = v and not
to occur, so this party will refuse to trade. More generally, it can be shown
trading, so we assume for
that under symmetric information we always obtain the ex post efficient simplicity that they choose
outcome. to trade – this is only a
technical point.
If, however, there is asymmetric information, the volume of trade may be
inefficient. Suppose that both parties know c but only the buyer knows v.
All the supplier knows is that v is distributed as a random variable with
cumulative distribution function F(v) and density function f(v) =∂F(v)/∂v
on the interval [v, v] (hence F(v) = 0, F(v) = 1). Gains from trade exist
with some probability between 0 and 1, that is v < c < v (the problem
would be trivial otherwise). To simplify the problem, suppose further that
the supplier has all the bargaining power in period 2, so he makes a take-
it-or-leave-it offer to the buyer at price p.4 The buyer will accept this offer 4
It would be more realistic
if v ≥ p, so trade will occur if and only if v ≥ p, that is with probability: to assume that both parties
v have some bargaining
prob(v ≥ p) = ∫ f(s)ds = 1−F(p) . power, but this would
p complicate the analysis
Recall that if trade does not occur the supplier ends up with zero. So the without changing the
supplier’s expected profit is given by E(Π) = (p – c)[1 – F(p)]. The supplier qualitative results.
will choose p to maximise this, so:
∂E(∏)
=1−F(p)−(p−c)f(p)= 0 .
∂p
From this equation it can be seen that in general the supplier chooses
p > c, so there are circumstances where trade does not occur even though
24
there are gains from trade. In particular, this is the case when p > v ≥ c. In
this case the buyer rejects the supplier’s offer since he would make a loss
by accepting, so trade does not occur even though v ≥ c. For efficiency, on
the other hand, we would require p = c, so that trade occurs if and only if
v ≥ c. More generally, it can be shown that when both value and cost are
private information and gains from trade are not certain, the volume of
trade is not efficient.
Activity
Show that, if the parties could sign a contract in period 1 in this simple model, they could
devise a contract such that the efficient volume of trade is obtained.
Answer
The contract should simply give the power to choose the price p in period 2 to the buyer
(i.e. the informed party). The buyer would then set p = c in order to appropriate all the
surplus.5 So trade would occur if and only if v ≥ c, which is what we require for efficiency. 5
Why is the supplier
The fact that the buyer would appropriate all the surplus is irrelevant as far as efficiency prepared to trade if
is concerned, because all that efficiency requires is the maximisation of the ‘pie’. In any p = c? Recall that to
avoid some purely
case, the parties could also specify in the contract a lump sum payment from the buyer to
technical problems
the seller to create any division of the surplus: the outcome would depend on the relative we have assumed
bargaining power of the parties in period 1, when the contract is signed. that when a party is
indifferent between
Note, however, that there may be no fully informed party, if both c and v are private
trading and not trading,
information. Furthermore, unlike our simple model, a complete contract cannot be written it chooses to trade.
in many practical situations. In those cases, asymmetric information will lead to inefficient Alternatively, you can
outcomes. imagine that the buyer
offers p = c + ε, where
So far there was no ex ante relation-specific investment in the model. Now ε is positive and very
small.
let us assume that one of the parties, say the supplier, can invest in period
1, say in cost reduction. In particular, let c be a function of investment I:
c(I), with c'(I) < 0, c''(I) > 0. Assume v ≥ c(0), so that there are always
gains from trade. We want to focus on how ex post bargaining affects the
volume of investment undertaken, so we will further assume that there is
symmetric information and hence the volume of trade ex post is efficient.
In other words, we abstract from additional complications created by
asymmetric information. Both v and c are commonly known and trade
always occurs since v ≥ c whatever the level of I. Finally, assume for
simplicity that the two parties have no ‘outside option’ ex post, that is
to say their only chance to realise a positive surplus is to trade between
themselves (there are no other buyers or sellers in period 2).
In period 2 the trading price p will be determined through bargaining. If
the two parties have equal bargaining power, the ex post surplus will be
split equally between them, namely:
v+c (I)
v – p = p – c (I ) ⇔ p =
2
Recall that v – p is the buyer’s ex post surplus (i.e. his gain over and above
his second-best alternative, which is in this case zero), while p – c(I) is the
seller’s ex post surplus. Note that the cost of investment I is not relevant
as far as the division of the ex post surplus is concerned because this
investment has already been sunk when the two parties bargain.
In period 1 the supplier chooses how much to invest. When making this
decision he anticipates what will happen in period 2, that is, he anticipates
that p = [v + c(I)]/2. He chooses I to maximise his net profit, which is
equal to the ex post surplus minus the cost of investment:
25
v+c (I)
p – c (I ) – I = – c(I )– I
2
The first-order condition is – c'(I) = 2, and this implicitly defines the
privately optimal level of investment Ip. Is this level of investment
efficient? The efficient level of investment is the value of I that maximises
the joint net profit v – c(I) – I. In other words, it is the value of I that
maximises the ‘pie’ net of the cost of investment. You can also think of it as
the value of I that would be chosen if the supplier and the buyer merged
into a single entity. The first-order condition is –c'(I) = 1 and this defines
the efficient level of investment I*. Since c"(I) > 0 (i.e. the cost function is
strictly convex) we have Ip < I* (see Figure 2.2). The supplier invests less
than what is required for efficiency.
c
slope: –2
slope: –1
c (I)
Ip I* I
Figure 2.2
The intuition is simple. Since the ex post surplus is divided between the
two parties, the investing party does not capture all the cost savings from
its investment. This ‘distortion’ of incentives leads to underinvestment.
Vertical integration restores the efficient level of investment. The model
can also be refined to analyse the effect of the degree of investment
specificity on the level of ex ante investment. It turns out that the level of
investment is correspondingly lower the higher the degree of specificity.
Thus the higher the degree of specificity, the bigger is the incentive for the
firms to merge to restore efficiency if a contractual solution is not feasible.
(Actually in our example a contractual solution is feasible, but in more
general settings it would not be.)
Activity
Could a contract between the parties restore the efficient outcome? When should this
contract be signed and what should it specify? Assume that a contract which directly
specifies the level of investment that the supplier is to undertake is not feasible because
investment levels, although observable by the parties, are not verifiable in a court.
Answer
The parties could sign a contract ex ante that specifies the process through which the
amount of trade and a lump sum transfer are determined ex post. The contract should
give to the investing party (i.e. the seller) the right to set the price p in period 2. The seller
would set p = v and appropriate all the ex post surplus. A division of this through a lump
sum could be agreed. In period 1 the seller would choose I to maximise v – c(I) – I, which
would give the efficient level of investment.
But what if both parties in a relationship can make an investment to increase the gains
from trade? Also, recall that complete contracts are difficult to write in more general
settings.
26
The above analysis has left some questions unanswered, however. Exactly
why does integration solve or reduce the hold-up problem (i.e. exactly
what changes when two firms merge)? Does it matter who acquires
whom when two firms merge – in other words, does it matter who has
control within the integrated firm? And if vertical integration improves
efficiency, then why don’t firms always merge (i.e. why are there limits to
integration)?
Property rights
Hart and others have pointed out that, given that contracts are incomplete,
one thing that greatly matters in a relationship is which party has the right
to make decisions and settle potential disputes in case of unspecified
contingencies. Obviously, it is the owner of the physical asset(s) who has
this right, the residual right of control. According to this view, a firm
is seen as a collection of physical assets that belong to it: machines,
inventories, buildings, client lists, patents, cash, etc. – excluding human
capital. ‘Ownership’ is defined as the right to specify all usages of these
assets in any way not inconsistent with a previous contract, custom or law.
Note that the possession of residual rights of control does not rule out ex
post renegotiation bargaining. What it does is determine the ‘status quo
point’ in the bargaining process, in other words it puts the party that has
these rights in a better bargaining position. In this way it affects the
division of the surplus ex post and therefore also influences the level of
investment ex ante.6 6
Hart and Moore (1990),
in their Introduction,
A model discuss a similar
example with one asset
We look at a simple model that examines how ownership changes the and three agents which
division of the ex post surplus and hence the ex ante investment. In this illustrates many of the
model there are two periods. In period 1 a buyer and a supplier contract main ideas and results
to trade in period 2. Trade will occur in any case, but there is uncertainty in the property rights
approach. Tirole (1988),
about the final specification of the good to be traded: a basic design is
pp.29–34, and Hart
specified in the contract in period 1, but an opportunity to improve its (1995), Chapters 2–4,
quality may arise in period 2 – and this cannot be described in period 1. provide a more formal
treatment as well as rich
The cost to the supplier of making an improvement in period 2 is c > 0.
informal discussion.
The value of the improvement to the buyer is initially uncertain and can
take two values: it is v > c with probability x and 0 with probability 1 – x.
This uncertainty is resolved at the beginning of period 2 (i.e. before the
decision whether to make the improvement or not is made). We assume
for simplicity that v and c are already known to both parties in period
1. Note that v and c are extra value and cost (beyond the good’s basic
design).
The buyer can invest in period 1 to increase the probability that the
improvement that comes about can be used (i.e. it has value v rather
than 0). In other words, the buyer chooses x. You may think of this as an
investment in flexibility. The cost of the investment is x2/2. The level of
investment is not verifiable by outsiders, so it cannot be specified in the
contract.
In this model both parties can take actions that increase the expected gains
from trade: the buyer can invest in flexibility in period 1 and the supplier
may pay for product improvement in period 2. We now consider three
different institutional arrangements.
27
1. Non-integration
This leads to unconstrained bargaining in period 2 (i.e. the two parties
bargain over whether the improvement must be made and, if they
disagree, the improvement is not made because it was not specified
in the contract). Since v > c > 0, the improvement will be made if
and only if its value is revealed to be v rather than 0. Gains from the
improvement, if it is made, are split equally between the two parties.
The buyer gets surplus (v – c)/2 if the improvement is made (i.e. with
probability x) and 0 otherwise (with probability 1 – x).
In period 1, when the buyer chooses x, his expected net surplus from
the improvement is:
x(v−c)/2+(1−x)0−x2/2 .
The buyer will choose x to maximise this, which gives: xNI = (v – c)/2.
The joint (buyer’s plus supplier’s) expected net surplus from the
improvement is:
WNI = xNI(v−c)+(1−xNI)0−x2NI /2=3(v−c)2/8 .
2. Integration under supplier control
The supplier decides in period 2 whether the improvement is made
– or the supplier can bargain and offer to give this authority away in
return for a transfer. This is equivalent to unconstrained bargaining,
because the status quo point is the same: if the parties cannot agree
on some other arrangement, the supplier will choose not to make the
improvement. Why? Because not making the improvement is the less
costly action for the supplier in case of disagreement, since all of v
from the improvement would then be kept by the buyer.
The parties will agree that the improvement be made if its value is
revealed to be v. Gains from the improvement will be split equally
between them.7 If the value is revealed to be 0, the improvement will 7
We assume equal
not be made. The buyer will get surplus (v – c)/2 if the improvement
bargaining power in
is made and 0 otherwise. In period 1, the buyer’s expected net surplus period 2. The party with
from the improvement is therefore given by the same expression as the residual rights of
under non-integration. control can unilaterally
decide whether the
The previous results apply: xNI and WNI. improvement is made
3. Integration under buyer control (or not) in a case of no
agreement, but this does
The buyer decides in period 2 whether to impose the improvement not imply that it has all
– or the buyer can bargain and offer to give away this authority in the bargaining power in
return for a transfer. In this case the buyer can choose to impose the the negotiations.
improvement if some other arrangement is not reached: the status quo
point is now different.
If it turns out that the value of the improvement is v, the improvement
will be imposed by the buyer and the buyer will get surplus v. The
supplier will have to pay c but will not get any part of v, since the
improvement will have been imposed by the buyer rather than agreed
by the two parties.
If it turns out that the value of the improvement is 0, the buyer gets no
benefit from imposing it, so he will prefer to bargain with the supplier.
The improvement will not be made, but the gain from not making it
– c – will be split equally between the parties. That is, the buyer will
obtain a transfer of c/2 from the supplier and make surplus c/2.
In period 1, the buyer’s expected net surplus from the improvement is:
xv + (1−x)c/2−x2/2 .
28
The buyer will choose x to maximise this, which gives:

xBC = v – c/2 > xNI. The intuition is that because the buyer can impose
the improvement on the supplier and therefore does not have to worry
about its cost, his ‘investment in flexibility’ is larger than what it would
have been under non-integration. The joint expected net surplus from the
improvement is:
WBC = xBC(v−c)+(1−xBC)0−x2BC /2=(1/2)(v−c/2)(v−3c/2) .
Which one of these three institutional arrangements is the ‘best’? The
best arrangement is the one with the highest joint expected net surplus
– because any division of the joint expected net surplus between the
parties can be specified through a lump sum transfer. Comparing WNI
with WBC, there are two cases. For c < v/2, we get WBC > WNI, so buyer
control is best. For c > v/2, we have WNI > WBC, so supplier control (on
non-integration) is best. To conclude: if the buyer makes a relatively large
investment in period 1 and the supplier a relatively small one in period
2 (c < v/2), the firms should integrate and the buyer (the party whose
investment matters more) should have control. On the other hand, if the
buyer makes a relatively small investment in period 1 and the supplier a
relatively large one in period 2 (c > v/2), then either the firms should not
integrate or, if they do, the supplier (the party whose investment matters
more) should have control.
General conclusions of the property rights approach

Several conclusions have emerged from the property rights approach.
Consider the case of two owner-managed firms that enter into a long-run
relationship and must both make a relation-specific investment. Then:
• Integration reduces opportunistic behaviour because if, say, firm
A acquires firm B, then the manager of firm B loses control of the
physical assets of firm B, so he has much less bargaining power.
• In a merger it matters which of the parties has residual rights of
control within the merged firm (i.e. it matters who acquires whom).
• A party is more likely to own, and hence have control over, an asset if
it has a large investment to make. In other words, efficiency requires
that the residual rights of control rest with the party whose ex ante
investment has the larger effect on the joint profit.
• Efficiency requires that highly complementary assets are under
common ownership. On the other hand, independent assets should
be separately owned: there are limits to integration. Why? If firm A
acquires firm B, for example, then the manager of firm B will have
much lower incentives to undertake investments since the payoff from
these will be partly appropriated by the owner of firm A. So if assets
are independent, the costs of integration (in terms of underinvestment
by the manager of firm B) will be higher than any potential benefits.
You should also bear in mind some qualifications to this. First, the above
conclusions apply more directly to the case of owner-managed firms.
Things are more complicated when there is separation of ownership from
control and delegation of authority within firms, although the general
framework should still be valid. Second, another way of solving the
problem of opportunism may be for firms to try to build a reputation for
non-opportunistic behaviour when they interact repeatedly with each 8
This point will become
other. However, this would not necessarily work because the payoff from clearer when you have
behaving opportunistically might be bigger than the payoff from adhering studied the material on
to non-opportunistic behaviour.8 repeated interaction in
Chapter 5 of this guide.
29
Lowering transaction costs and mitigating opportunistic behaviour are

not the only reasons for vertical integration (i.e. integration between a
supplier and a buyer). As we will see in later chapters of this guide, a
firm may also vertically integrate to eliminate negative externalities that
arise in buyer–seller relationships even in the absence of relation-specific
investments, or to indirectly price-discriminate, or to increase its market
power by hindering rival, non-integrated firms’ access to outlets or sources
of supply.
Empirical evidence
Much of the empirical work on the determinants of firm size and structure
that has followed the transaction costs approach has focused on the role of
investment specificity for vertical integration.9 Two examples are discussed 9
Other studies have
below. Lafontaine and Slade (2007) provide a survey of the empirical focused on the related
issue of the role of
literature on the boundaries of the firm.
investment specificity
Klein, Crawford and Alchian (1978) describe the story of the 1926 merger for the type and
between General Motors (GM) and Fisher Body. In 1919 GM, a US car duration of contracts
signed between firms.
manufacturer, entered a 10-year contract with Fisher Body for the supply
Joskow (1987) is a good
of car bodies. To minimise the scope for opportunistic behaviour, the example.
contract specified that GM should buy all their closed car bodies from
Fisher and also specified the trading price – with the additional provision
that this price could not be greater than the average market price of
similar bodies produced by firms other than Fisher.
However, demand conditions changed dramatically after 1919: there was
a large increase in demand for cars, especially cars with closed bodies –
the type manufactured by Fisher. GM thought that Fisher’s costs had gone
down because of scale economies in the production of bodies and were
unhappy with the price they had to pay for Fisher bodies. Also, Fisher
refused to locate their plants close to GM plants – a move which GM
thought would increase production efficiency but which would diminish
the bargaining power of Fisher. These tensions ended in 1926, when GM 10
This is the prevailing
acquired Fisher.10 view on the GM–Fisher
A second case study discussed by Klein, Crawford and Alchian (1978) merger. See, however,
Activity 3 below.
is the petroleum industry. Oil wells are often connected to refineries by
pipelines. The owner of the pipeline has significant market power, so there
is scope for opportunistic behaviour. This explains why it is a common
arrangement for pipelines to be jointly owned by producers (oil wells) and
refiners. In the case of tankers, on the other hand, there is no relation-
specific investment, so no scope for hold-up. This helps to explain why
tankers are generally under independent ownership.
Monteverde and Teece (1982) have examined why firms in the automobile
industry produce some components in-house while they buy others
from independent suppliers. One of their main hypotheses was that car
manufacturers will vertically integrate (produce in-house) when the
production process for components generates transaction-specific know-
how. That is so because it is then more difficult for car manufacturers
to switch to other suppliers, so there is more scope for opportunistic
behaviour by suppliers.
Monteverde and Teece (1982) tested this hypothesis econometrically using
data on 127 different components used by two big US car manufacturers.
The dependent variable in their regressions was a binary variable for
in-house production versus production by an external supplier. Their
independent variables included the cost of developing a component (a
30
proxy for the know-how generated in the production of a component), a

dummy variable for firm-specific components versus generic components,
a firm dummy to control for company effects, and other variables. Their
results confirmed the predictions of the transaction costs approach:
• The higher the development cost of a component, the more likely that
production was in-house.
• Firm-specific components were more likely to be produced in-house
than generic components.
Activities
1. A student’s university education normally involves a contract between the student and
the university and separate contracts between the university and each of the student’s
instructors. Why is this the case? Why not a series of contracts between the student
and each of his or her instructors? Would it matter if contracts were complete?
2. Since the 1980s there has been a trend towards de-integration across many industries
as well as a trend towards more flexible technologies. Could the two trends be
related, and in what way?
3. Alchian and Demsetz (1972)11 suggested that team production provides a rationale 11
Note that full
for the existence and structure of firms. They argued that factors of production are references for all
readings cited in the
often more productive when they are members of a team rather than when used on
Activities can be found
their own. However, it is difficult to measure the contribution of each member to the in Appendix 2.
team’s output or to monitor the effort of each member of a team. This creates an
incentive for each team member to shirk and free ride on the efforts of others.
Contracting among themselves is difficult for team members. Firms therefore exist in
order to solve or mitigate this problem. In particular, the role of the firm’s owner is to
be the central locus with which all the other factors of production contract; to monitor
the employees; and to determine appropriate compensation. The owner has incentives
to exert effort efficiently because they are the residual claimant to the firm’s output.
Is this argument convincing? Why or why not? In particular, to what extent can
the Alchian and Demsetz view explain why the problem of joint production and
monitoring must be solved by vertical integration rather than a series of contracts
between each team member and the monitor?
4. The Klein, Crawford and Alchian (1978) interpretation of the 1926 merger between
General Motors and Fisher Body has been criticised by some economists, including the
Nobel prize winner Ronald Coase, who have argued that the reason for the merger
had nothing to do with opportunistic behaviour and hold-up problems. This debate
is not just about a particular event in economic history. It is about one of the most
frequently cited examples of market failure. If the critics are right, then market failure
in vertical relations between firms may be less prevalent than the theory would lead
us to believe.
What do you think? Make up your own mind after reading different views on the
GM–Fisher Body case. A good collection of articles – representing the different views
and including contributions by some of the protagonists of the debate – has been
published in the April 2000 issue of the Journal of Law and Economics. You can also
search the internet for more information.
31

Having completed this chapter, as well as the Essential reading and
• describe and evaluate two different approaches to explaining the size
and structure of firms
• explain the notions of ‘transaction costs’, ‘investment specificity’,
‘opportunistic behaviour’, ‘incomplete contracts’ and ‘residual rights of
control’ and their relevance for the theory of the firm
• explain the effect of investment specificity on the decision of firms to
enter into contractual relationships or to integrate
• analyse the factors that determine who should have control inside a
vertically integrated firm.

1. Consider the following model of a vertical relationship between a
buyer and a seller. There are two periods and the two parties can,
if they wish, trade one unit of an indivisible good in period 2. Let v
denote the value of the good to the buyer, c the production cost, and
p the trading price. Assume that c < 1/2. Both c and v are commonly
known at the beginning of period 2. The seller can invest in period 1 to
increase the value of the good to the buyer (for instance, he can spend
on R&D to increase the quality of the product). In particular, v(I) =
3I – I2/2. The level of investment I cannot be specified in a contract
because it is not verifiable and therefore such a contract would not be
enforceable in court.
a. What is the efficient level of investment?
b. In the absence of any contract, what is the level of investment
chosen by the seller if the ex post surplus is to be divided equally
between the two parties? Explain why this level is not efficient.
c. Suppose that the parties sign a contract which gives to the seller
the right to choose the trading price in period 2 (i.e. after the
investment has been made). What will be the level of I chosen by
the seller?
d. Now suppose that the parties sign a contract which gives to the
buyer the right to choose the trading price in period 2. What will
be the level of I chosen by the seller? What is your conclusion
about who should have the power to decide the price in period 2?
Explain the intuition for your results.
2. Analyse how investment specificity affects the ex ante incentives for
investment when there is ex post bargaining over the surplus. Then
explain how investment specificity and the incompleteness of contracts
may affect the decision of a firm to vertically integrate and discuss
briefly any relevant empirical evidence.
32
Chapter 3: Separation of ownership and control
Chapter 3: Separation of ownership and

control
Introduction
A common assumption in most economic theory is that firms maximise
(expected) profits. This is probably what the owners of a firm would like
to do. However, in modern large companies, it is not the shareholders who
run the firm, but the managers, who are likely to have other objectives
than profit maximisation.
This separation of ownership and control gives rise to several important
issues. First, given that the owners typically have less information than the
managers and cannot perfectly monitor the behaviour of the latter, how
can they design incentive schemes that induce the managers to behave as
much as possible according to their (the owners’) interests? Second, given
that such contracts are generally not perfect solutions to the problem of
managerial discretion, what other mechanisms are there that may limit
the ability of managers to pursue their own objectives rather than those of
the owners? Third, is profit maximisation a reasonable description of firm
behaviour?
Learning outcomes
• explain the implications of the separation of ownership and control in
modern large companies
• analyse optimal incentive mechanisms offered by the owners of a firm
to the firm’s manager
• describe different mechanisms that may restrict managerial discretion
and discuss their limitations
• assess the validity of the profit-maximisation hypothesis.
Essential reading
Church, J.R. and R. Ware Industrial organization: a strategic approach.
Chapter 3.
Tirole, J. The theory of industrial organization. Introductory chapter.
Further reading
Book
Holmstrom, B.R. and J. Tirole ‘The theory of the firm’, in Schmalensee, R. and
R. Willig (eds) Handbook of industrial organization, Volume 1. (Amsterdam:
North–Holland, 1989).
Journals
Abowd, J.M. and D.S. Kaplan ‘Executive compensation: six questions that need
answering’, Journal of Economic Perspectives 13(4) 1999, pp.145–68.
Bloom, N. and J. Van Reenen. ‘Why do management practices differ across firms
and countries?’, Journal of Economic Perspectives 24(1) 2010, pp.203–24.
33
Nickell, S. ‘Competition and corporate performance’, Journal of Political

Economy 104(4) 1996, pp.724–46.
Nickell, S., D. Nicolitsas and N. Dryden ‘What makes firms perform well?’,
European Economic Review 41 1996, pp.783–96.
Symeonidis, G. ‘The effect of competition on wages and productivity: evidence
from the UK’, Review of Economics and Statistics 90 2008, pp.134–46.
Managerial incentives
An obvious way for the owners to restrict managerial discretion or to
induce managers to put more effort into the job is to offer monetary or
other incentives to managers. Essentially this involves linking the
compensation of managers to firm performance. Some important insights
on the use of incentives can be drawn from a simple model of a firm run
by a single manager.1 The profit of the firm can take one of two values, Π1
1
(1988), pp.36–39.
and Π2, with Π1 < Π2. The manager chooses between two levels of effort,
high and low (for simplicity: zero). His utility is U = u(w − Φ) if he makes
high effort and U = u(w) if he makes zero effort, where u is an increasing
and strictly concave function, w is the manager’s wage, and Φ > 0 is the
monetary disutility of high effort. Note that the strict concavity of u
implies that the manager is risk averse.2 Whether the firm makes Π1 or Π2 2
The strict concavity of u
depends on the manager’s effort as well as on the firm’s environment, implies that u’’(w) < 0.
which is uncertain. In particular, if the manager makes high effort, the
profit is Π2 with probability x and Π1 with probability 1 − x. If the
manager makes no effort, the profit is Π2 with probability y and Π1 with
probability 1 − y. We have 0 < y < x < 1.
Now consider the following set-up. First, the owners of the firm choose
a contract (an incentive scheme) for the manager. At this stage, they do
not yet know what the profit of the firm will turn out to be. The contract
therefore specifies the wage of the manager for each of the two possible
values of Π. The objective of the owners is to maximise expected net profit
E(Π − w). Note that this objective function implies that the owners are
risk neutral.
Given the incentive scheme chosen by the owners, the manager decides
whether to accept the job or not and, if he accepts, chooses the level of
effort that maximises his expected utility E(U). We assume that he can
always obtain a reservation wage w0, and hence utility U0 = u(w0), by
working outside the firm, so he will never accept to work for the firm if his
expected utility from doing so is less than U0. After the manager has made
his choice, the profit is observed and the manager gets paid. The question
is what incentive scheme the owners should choose to maximise E(Π − w).
If the owners could observe the manager’s effort level, there would be no
need for an incentive mechanism, since the owners could then impose an
effort level on the manager.3 All that they would need to do is ensure that 3
The contract would
the manager accepts the job. This implies ensuring that the manager provide for a severe
punishment if the
obtains utility exactly U0: any payment giving him a higher utility would
manager fails to exert
be unnecessary and would reduce the expected profit of the firm. If the the level of effort
owners wanted no effort, they should pay the manager the reservation prescribed.
wage w0 whatever the profit turned out to be. Faced with this contract, the
manager would accept the job and make zero effort. Net expected profit
would be yΠ2 + (1 − y)Π1 − w0. If the owners wanted high effort, they
should pay the manager w0 + Φ whatever the profit. The manager would
then accept the job and make high effort. Net expected profit would be
xΠ2 + (1 − x)Π1 − (w0 + Φ). Obviously, the owners would choose to
impose high effort if and only if:
34
Let us assume that this holds, so the owners would prefer high effort, if
effort were observable – otherwise the problem under unobservable effort
level would be trivial.
Activity
Prove that, when effort is observable, the owners offer w0 if they want no effort and
w0 + Φ if they want high effort. Conclude that, when effort is observable, the risk averse
party bears no risk.
Answer
Solve the following trivial maximisation problem for the owners: choose w to maximise
E(Π – w) subject to the manager getting utility at least equal to U0. Since the manager
gets the same wage whatever the realisation of profit turns out to be, he bears no risk. All
the risk is borne by the owners (the risk neutral party).
The need for an incentive scheme arises when the effort level of the
manager cannot be observed by the owners and hence cannot be
prescribed in the contract. If they want to induce the manager to exert
high effort, the owners must reward the manager with a higher wage in
the event that profit turns out to be Π2 rather than Π1. More specifically,
the owners must design a wage structure wi(Πi), i = 1, 2, that maximises
their expected net profit:
x(∏2 – w2)+(1 – x)(∏1 – w1)
subject to ensuring that the manager accepts the job and chooses to exert
high effort, that is subject to a ‘participation constraint’:
xu(w2 – Φ) + (1 – x)u(w1 – Φ) ≥ u(w0)
and an ‘incentive-compatibility constraint’:
xu(w2 − Φ) + (1 – x)u(w1 − Φ) ≥ yu(w2) + (1 − y)u(w1).
The first constraint says that, under the incentive scheme, the expected
utility of the manager if he exerts high effort is at least U0, so the manager
will accept the job. The second constraint says that the expected utility of
the manager if he makes high effort (the left-hand side of the inequality) is
at least as large as his expected utility if he makes zero effort (the right-
hand side of the inequality), so the manager chooses to make high effort.
In this maximisation problem both constraints are satisfied with equality.
The incentive scheme chosen by the owners, if they want to induce high
effort, will have the following properties:
• the manager will be rewarded if profit is high: w2 > w1. This can be
derived from the incentive-compatibility constraint
• the expected wage xw2 + (1 − x)w1 will be higher than w0 + Φ, the
wage under observability of effort. This is a result of the concavity of
u. Hence, the owners’ net profit will be lower.
Note that if the owners want to induce no effort, all they need to do is
offer w0 whatever the profit; this will ensure that the manager participates
and chooses to make no effort. Net profit will be the same as under
observability of effort. What will the owners choose to do: offer a contract
that induces high effort or one that induces no effort? It depends on
whether their maximised net profit is higher under high effort or under
zero effort. We have assumed, of course, all along that
(x – y)(Π2 − Π1) > Φ, that is to say the owners would prefer high effort
to zero effort, if effort were observable. But this does not ensure that the
35
same is true when effort is unobservable, because unobservability reduces

the owners’ net profit under high effort but not under no effort. In other
words, an additional effect of unobservability is that the owners are more
likely to tolerate managerial slack.
Limits to managerial discretion

There are several other mechanisms, apart from direct monetary incentives
of the kind examined above, that can limit managerial discretion and
reduce slack. Some of the most important are as follows:
• The threat of takeovers. If managers fail to maximise profits,
the stock market value of the firm will be lower, and this may
induce outsiders to take over the firm and replace the managers.
The effectiveness of this mechanism is reduced by the fact that the
collection of information on the firm by outsiders may be costly,
takeovers may be subject to free-rider problems, and managers may
resist takeovers. Takeovers may also have perverse incentive effects,
for instance they may cause managers to put too much emphasis on
short-term profits to the detriment of long-term profits.
• Reputation effects. Managers care for their careers and are eager
to acquire good reputations. This may reduce slack, and may even
cause managers to work too hard (i.e. harder than the socially optimal
level) early in their career.
• Supervision. Monitoring the managers’ (more generally, the
employees’) performance in order to obtain better information on their
‘effort’ may be costly but feasible. The effectiveness of this mechanism
may be reduced by the difficulty of measuring individual effort
when team work is important, and also by the possibility of collusion
between supervisors and supervisees.
• Competition in the product market. The effects of product
market competition on managerial incentives can take several forms
and may sometimes be ambiguous.4 One idea is that if a firm does not 4
A review of the
maximise profits, there is a higher probability that it will not be able to literature is given in
Nickell (1996).
compete with more efficient firms and will therefore go bankrupt.
Managers wishing to avoid this will work hard to maximise profits.
• Organisational form. The internal organisation of a firm can help
mitigate managerial slack, especially by lower managers. The ‘unitary
form’ firm allows greater specialisation of labour, but supervision by
the top management becomes more difficult as the firm grows. In the
‘multi-divisional form’ firm, on the other hand, it is possible for the top
management to measure the performance of the different divisions
within the firm and compare them with one another. One reason for
the gradual decline of the U-form and the emergence of the M-form
may have been the need to limit managerial discretion.
Empirical evidence
Empirical work on the performance of firms (for instance, Nickell, 1996;
Nickell et al., 1996) has looked at a number of factors external to the
firm that are associated with improved productivity growth. Three such
factors have been identified: product market competition, financial
market pressure (i.e. a high level of debt) and shareholder control (i.e.
the existence of a dominant external shareholder from the financial
36
sector). Using industry-level data, Symeonidis (2008) has found clear

evidence of a negative effect of cartels on productivity. Other studies have
established a positive effect of trade liberalisation on the productivity
of firms in various countries. Bloom and Van Reenen (2010) argue
that productivity differences across firms and countries largely reflect
variations in management practices and that product market competition
has an important influence in increasing aggregate management quality by
helping to eliminate badly managed firms.
The profit maximisation hypothesis

Although there are many ways in which managerial discretion can be
restricted, none of them is perfect, so we should expect deviations from
profit maximisation due to the separation of ownership and control. In
addition, performing complex calculations is a time-consuming, effort-
demanding, and sometimes impossible task, especially under conditions of
uncertainty, so members of a firm may often follow simple ‘rules of thumb’.
Ultimately, the question is how significant any deviations from profit
maximisation are likely to be. If we accept, as many economists do, that
large deviations from profit maximisation will probably not allow a firm to
survive in the long run, then the hypothesis that firms maximise expected
profits seems a reasonable approximation of firm behaviour for most
purposes, and in particular for the analysis of the interaction between
firms in the market, which is the main focus of industrial economics.
Activities
1. Is it a good idea to have a junior clerk in a bank on an incentive contract? Is it a good
idea to have the chief executive officer of a bank on an incentive contract? Why or
why not?
2. Profit maximisation is usually taken to be the goal of the firm in standard industrial
economics models. To what extent is this a realistic view of firm behaviour? What are
plausible alternatives to profit maximisation?
3. What is the empirical evidence on the effects of monetary incentives and
performance-related pay? Do they generally increase productivity as the
theory predicts? Can they possibly backfire, and in what ways and under what
circumstances?
Make up your own mind after reading some of the extensive empirical literature.
References include Lazear (2000); Gneezy, Meier and Rey-Biel (2011); and Ederer and
Manso (2013). You can also search the internet for more.
4. There is one category of firms where ownership and control are not separated, or
not as separated as in most large firms: family businesses. Why are family firms so
prevalent? Why are some successful and others struggling? What are the implications
of family control for the governance and overall performance of firms? Are family
firms an efficient response to certain institutional and market environments, or
are they a result of cultural norms that might be costly for corporate decisions and
economic outcomes?
Try to answer these questions after reading some of the literature on family firms.
Good references are: Bertrand and Schoar (2006); Bennedsen, Nielsen, Perez-
Gonzalez and Wolfenzon (2007); and Villalonga and Amit (2006). And there is much
more on the internet.
37

• explain the implications of the separation of ownership and control in
modern large companies
• analyse optimal incentive mechanisms offered by the owners of a firm
to the firm’s manager
• describe different mechanisms that may restrict managerial discretion
and discuss their limitations
• assess the validity of the profit-maximisation hypothesis.

1. The profit of a firm can take one of two values, Π1 and Π2, where
Π2 − Π1 > 10. The firm is run by a manager who chooses between
two levels of effort, e = 1 (high) and e = 0 (low). The manager’s
utility function is U = w1/2 – e, where w is her wage. Whether the
firm makes Π1 or Π2 depends on the manager’s effort and on the
firm’s environment, which is uncertain. In particular, if the manager
makes high effort, the profit is Π2 with probability 0.8 and Π1 with
probability 0.2. If the manager makes no effort, the profit is Π2 with
probability 0.3 and Π1 with probability 0.7. Before the realisation
of Π is observed, the owners of the firm choose a contract for the
manager which specifies the value of w for each of the two possible
values of Π. The owners’ objective is to maximise expected net profit
E(Π − w). Given the incentive scheme chosen by the owners, the
manager decides whether to take the job and, if she accepts, chooses e
to maximise her expected utility E(U). Her reservation wage is w0 = 4.
After the manager has made her choice, the profit is observed and the
manager gets paid.
a. What is the optimal contract if the owners can observe the
manager’s effort?
b. What is the optimal contract if the owners cannot observe the
manager’s effort?
c. Show that the net profit of the owners is lower if the manager’s
effort is unobservable than if it is observable.
2. ‘An optimal incentive scheme offered by the owners of a firm to its
manager should reward the manager when profits are high and
penalise him when profits are low.’ Discuss this statement with
reference to an economic analysis of the relationship between the
owners and the manager that takes into account the fact that the
manager’s effort level may not be observable by the owners.
38
Chapter 4: Short-run price competition
Introduction
This chapter begins our analysis of firms’ conduct in oligopolistic markets.
We will start with the simplest competitive situations. In particular, we
will assume that the only decision firms have to make is to set a price for
their product or a level of output. In fact, of course, firms can use many
instruments to compete in a market, and subsequent chapters will analyse
several examples of more complex strategic situations. The main reason
why it makes sense to abstract from these additional considerations in this
chapter is that price and quantity are instruments that firms can change
relatively easily in the short run. On the other hand, other decisions
are more difficult to change. These include product design, the level
of capacity, an advertising-based brand image, product quality or cost
determined by research and development (R&D), and so on. Ultimately
there is also the decision of whether or not to enter or stay in a market.
Since these long-run decisions are relatively difficult to change, they are
taken as given when making shorter-term decisions. We can therefore
proceed to analyse short-run competition among firms in the context
of fixed cost structures and product characteristics, and with a fixed
number of firms in the market.
In this chapter we focus on static models of oligopoly: the firms
interact only once in the market and their actions are simultaneous – each
firm chooses an action without knowing the actions taken by its rivals. We
will examine several models, including the basic Bertrand model, a model
of price competition among capacity-constrained firms and the Cournot
model. Some implications for competition policy will also be discussed.
Repeated interaction is the subject of the next chapter.
How realistic are static models of oligopoly, given that in reality firms
normally interact repeatedly rather than once? As we will see, the
equilibria of static models are valid equilibria of many dynamic models.
Moreover, static models are useful for deriving the full range of equilibria
of dynamic models. Finally, the equilibria of static models are often good
approximations to real-world firm behaviour.
Throughout the second part of the guide, we will be using the theory of
non-cooperative games to model firms’ interaction. The equilibrium
market outcomes are therefore the equilibria of these games. To identify
them, we will be making use of some elementary solution concepts from
non-cooperative game theory: the Nash equilibrium and the subgame-
perfect equilibrium.1 1
See Tirole (1988),
pp.423–32, or Shy
(1995), Chapter 2, for
Learning outcomes an introduction to these
game-theoretic concepts
By the end of this chapter, and having completed the Essential reading and and discussion of
activities, you should be able to: applications.
• describe and derive the Bertrand paradox
• analyse how the introduction of capacity constraints in the Bertrand
model leads to equilibrium outcomes with price greater than marginal
cost and positive profits
• explain the theoretical foundations of the Cournot model
• analyse the Cournot model for various assumptions regarding the
demand, the number of firms, and the cost structures.
39
Essential reading
Church, J.R. and R. Ware Industrial organization: a strategic approach. Chapter 8.
Further reading
Books
2000) Chapter 7.
Carlton, D.W. and J.M. Perloff Modern industrial organization. (United States:
Pearson Addison Wesley, 2005) Chapter 6.
Pepall L., D. Richards and G. Norman, Industrial organization: contemporary
theory and empirical applications. (Chichester: Wiley-Blackwell, 2014)
Chapters 9–10.
Shy, O. Industrial organization. (Cambridge, MA: MIT Press, 1995) Chapter 6.
The basic Bertrand model

Consider a very simple set up as follows. There are two firms, 1 and 2,
producing a homogeneous product (the result easily generalises to N
firms). The two firms interact only once and they simultaneously and
independently set prices p1 and p2 respectively. The market demand for
the product is given by q = D(p), and both firms have the same constant
marginal cost c. The firm with the lowest price gets all the market demand
at that price; if the two prices are the same, each firm gets half the market
demand at that price.
The Nash equilibrium outcome of this game is p1* = p2* = c. In other
words, firms price at marginal cost and make zero profit.
Activity
Prove this result.
Answer
The proof consists in distinguishing cases and showing that in all of them except the case
p1 = p2 = c there exists a profitable deviation by at least one firm. In particular:
• pi > pj > c : this cannot be a Nash equilibrium (NE) because firm i will want to reduce
its price slightly below pj , capture the whole market and make a positive profit rather
than zero
• pi = pj > c : this cannot be a NE because either firm will want to reduce its price
slightly and almost double its profit by serving the whole market
• pi > pj = c : this cannot be a NE because firm j will want to increase its price slightly,
maintain the whole market and make positive profit rather zero
• pi < c and/or pj < c : this cannot be a NE because one or both firms will want to set
price equal to c and stop making losses
• pi = pj = c : yes, this is a Nash equilibrium as none of the firms can increase its profit
by deviating – if a firm increases its price it still makes zero profit, if it reduces its
price it makes a loss and this is worse than zero profit.
The intuition is that unless prices are the same and equal to the common
marginal cost, each firm has an incentive to undercut the other. Note that
there is a qualification to this result for the case of asymmetric marginal
40
costs: in that case, at equilibrium the low-cost firm sets a price marginally
lower than the cost of the high-cost firm2 and makes positive profit. Still, 2
Provided this is
this profit is small if the cost difference is small, and the high-cost firm not higher than
the monopoly price
makes no sales and no profit. The outcome of the simple Bertrand game
corresponding to its own
has therefore justifiably been called the ‘Bertrand paradox’. cost; otherwise it sets
There are three main resolutions to this paradox: repeated interaction, the monopoly price.
product differentiation and capacity constraints.
Bertrand competition with capacity constraints

To understand why capacity constraints matter, take the simple model of
the previous section and assume that both firms have production capacity
smaller than D(c): that is to say, no firm can cover the entire demand at
a price equal to the common marginal cost. Then p1 = p2 = c is no longer
a Nash equilibrium. Why? If firm i raises its price slightly above c, given
that pj = c, all consumers will want to buy from firm j. However, firm j will
not be able to satisfy the whole demand, so some consumers will end up
buying from firm i. Hence firm i will make positive profit rather than zero.
Since pi = c is not a profit-maximising response to pj = c, p1 = p2 = c is not
a Nash equilibrium.
The exact equilibrium outcome in the above model depends on how small
the capacities are and on what specific assumption we make about the
way consumers are rationed. In general, however, models with capacity
constraints have Nash equilibria with price greater than marginal cost and
positive profits. Note that rigid capacity constraints are a special case of
decreasing returns to scale (i.e. a technology such that the marginal cost
increases with output). Such models also have equilibria with price greater
than marginal cost.
We now look at a simple model to fix these ideas.3 An additional important 3
implication of this model is that, in certain cases, a game where capacity- (1988), pp.212–216.
constrained firms compete in prices is formally equivalent to a game where

firms set quantities and an auctioneer determines the market-clearing
price.
Consider a market where two firms, 1 and 2, produce a homogeneous
product. Demand is q = D(p) = 1 − p, or equivalently p = 1 − q1 – q2. The
firms have capacity constraints q1 and q2, and we assume that qi < 1/3,
i = 1, 2. The marginal cost of production is zero for qi ≤ qi and infinite for
qi > qi. Finally, we assume that consumers are rationed according to the
‘efficient’ rationing rule. The rationing of consumers results from the fact
that the low-price firm cannot serve the entire market. The question then
arises as to which consumers end up buying from the high-price firm; this
is important because it determines the shape of the residual demand of
the high-price firm. For instance, suppose that firm 1 is the low-price firm.
Then under the efficient rationing rule, the residual demand of firm 2 is
given by D(p2) – q1 if D(p2) > q1, and zero otherwise. This is illustrated in
Figure 4.1, which depicts both the market demand D(p) and the residual
demand of firm 2 (note that with price p1 the low-price firm 1 sells up
to capacity). One interpretation of the efficient rationing rule is that, if
consumers have unit demands, those with the highest willingness to pay
buy from the low-price firm until its capacity is exhausted, and the rest
buy from the high-price firm.
41
residual demand for firm 2
market demand
p1
_ q
q1
Figure 4.1
We will now show that the unique Nash equilibrium outcome of this game
is for both firms to set the price p* = 1 – q1 – q2. At this price both firms
sell up to their respective capacities and the market clears. Note that this
price is higher than marginal cost (which is zero), and therefore implies
positive profits for both firms.
To show that this is a Nash equilibrium, we need to show that none of the
firms has an incentive to unilaterally deviate from this equilibrium. Is it
profitable for firm i to set a price lower than p*, given that firm j sets price
p*? The answer is no. By charging p* firm i sells exactly q1. Firm i cannot
produce more than q1 anyway, so by reducing its price below p* it would
simply sell the same quantity at a lower price and would therefore make
less profit.
Is it profitable for firm i to set a price higher than p*, given that firm j sets
price p*? The answer is again no, but the argument now is more subtle.
Suppose that firm i sets a price p ≥ p*. Then it has residual demand 1 –
p – qj, because at price p total market demand is given by 1 – p and firm
j sells qj. Firm i makes profit Π = p(1 – p – qj). Using the inverse demand
function, the expression for profit can be written as (1 – q – qj )q, where
q is the quantity sold by firm i at price p. Note that the profit function
Π = (1 – q – qj)q is exactly the same as the profit function of a firm that
chooses output q given that the rival firm chooses output qj. This profit
function is concave in q, that is Π''(q) < 0. Also, ∂Π/∂q = 1 – 2q – qj.
Evaluated at q = qi, this derivative is equal to 1 – 2qi – qj, which is positive
because both qi and qj are less than 1/3. In other words, if firm i starts
from qi and marginally reduces its quantity, its profit will fall. This result
and the concavity of the profit function ensure that any reduction of q
below qi will reduce profit. Another way of saying this is that if firm i starts
from p* and increases its price, its profit will fall.
We have therefore shown that it is not profitable for firm i either to set a
price lower than p* or to set a price higher than p*, given that firm j sets
price p*. Hence p1 = p2 = p* is the unique Nash equilibrium of the game.
Activity
Consider the model of price competition with capacity-constrained firms analysed above.
Assume, however, that one of the firms, say firm 1, has capacity higher than 1/3. In
particular assume that q1 = 2/5. Under what condition is p1 = p2 = p* = 1 – q1 – q2 still
the unique Nash equilibrium of the game? What do you think will happen if this condition
is not satisfied?
42
Answer
Much of the analysis is the same as above. For the equilibrium p1 = p2 = p* = 1 – q1
– q2 to be sustained, we need 1 – 2q1 – q2 > 0 ⇔ q2 < 1/5 and 1 – q1 – 2q2 > 0 ⇔
q2 < 3/10. The former condition ensures that the latter one is also satisfied. Hence we
need q2 < 1/5. If this condition is not satisfied, then p1 = p2 = p* = 1 – q1 – q2 is not a
Nash equilibrium. And since D(c) > qi, i = 1, 2, it follows that p1 = p2 = c is not a Nash
equilibrium either. There exists, however, a Nash equilibrium in mixed strategies with
prices greater than marginal cost.
There are two main conclusions from the above analysis. First, we have
seen how a static pricing game between symmetric firms can lead to Π > 0
if the firms are capacity-constrained. Second, everything is as if the firms
put quantities equal to their capacities on the market and an auctioneer
determined the price that clears the market. The equilibrium of this game
is the Cournot equilibrium (see the analysis of the Cournot model below).
Choice of capacities
One issue which was swept under the carpet in the above discussion is the
choice of capacities. We assumed that the firms were capacity-constrained,
and significantly so (qi < 1/3, i = 1, 2). But can’t firms build capacities
that would allow them to cut price down to marginal cost and supply the
whole market if they so choose?
To examine this question, we would need to construct a more complex
game than the one we have analysed, namely a two-stage game with choice
of capacities in the first stage and price competition in the second stage,
when capacities are taken as fixed. Now, intuitively one would expect firms
to strategically refrain from building too much capacity, because too much
capacity would destroy their profits in the price competition stage. This is
exactly what the formal analysis of such games predicts. In fact there is a
much stronger result, due to Kreps and Scheinkman: if demand is concave
and the rationing rule is the efficient one, then the outcome of this two-
stage game is the same as the outcome of the one-stage Cournot game
(which involves Π > 0, as we will see below).
One difficulty in oligopoly theory has been that the widely used Cournot
model, which assumes that firms compete by setting quantities, may
lack strong foundations, since firms typically compete by setting prices,
not quantities. We have seen, however, that the Cournot model can be
interpreted in either of the following ways:
• as a one-stage pricing game between capacity-constrained firms
• as a reduced-form game for the two-stage game with choice of
capacities in the first stage and price setting in the second stage.
Both these results rest on particular assumptions concerning the rationing
rule, so in more general settings we would not get equilibrium outcomes
that look exactly like the Cournot outcome. However, it is generally valid
to think of quantity competition as a choice of capacity or scale that
determines the firms’ cost functions and hence the conditions of price
competition. It is therefore valid to interpret the distinction between price
competition (Bertrand) and quantity competition (Cournot) as a difference
in the flexibility of production: if costs rise steeply with output in the short
run in a particular industry, then the Cournot model is more appropriate
for this industry; if not, then the Bertrand model is more appropriate.
Furthermore, the Bertrand model may become more appropriate than the
Cournot model for the same industry in an economic downturn, when
capacity constaints may not be binding.
43
The Cournot model

Consider a model of competition between two firms, 1 and 2, producing a
homogeneous product. The inverse demand function has the general form
p = P(q1 + q2), where q1 and q2 are quantities produced by firm 1 and firm
2 respectively and p is the market price. The demand curve is downward
sloping, so P'(q1 + q2) < 0. The total cost of firm i is given by Ci(qi), i = 1,
2. The two firms interact only once and they simultaneously set quantities.
The Nash equilibrium of this game is computed as follows. Firm 1 chooses
q1 to maximise its profit ∏1 = q1P(q1 + q2) – C1(q1), taking q2 as given. The
first-order condition for this maximisation problem is:
∂∏1
= 0 ⇔ P(q1 + q2)+ q1P´(q1 + q2) – C´1(q1) = 0.
∂q1
This equation implicitly defines the optimal choice of q1 for any given level
of q2. It is called the ‘reaction function’ of firm 1. Note that the equation
essentially says that firm 1’s marginal revenue equals its marginal cost
at the optimal value of q1. The second-order condition for a maximum is
satisfied provided the profit function is concave.
Similarly, firm 2 chooses q2 to maximise ∏2 = q2P(q1 + q2) – C2(q2), taking
q1 as given. The first-order condition, or the reaction function of firm 2, is:
∂∏2
= 0 ⇔ P(q1 + q2)+ q2P΄(q1 + q2) – C´2(q2)=0.
∂q2
The Nash equilibrium is the solution of the system of the two first-order
conditions.
We can identify two important properties of this equilibrium just by
looking at the first order conditions:
• The price is greater than marginal cost, namely:
p – Ci′ = –qiP′(q1 + q2) > 0, i = 1, 2.
• The first-order condition for firm i can be written as (p – Ci′)/p =
si/ε, where si = qi/(qi + qj) is the market share of firm i and ε is the
absolute value of the elasticity of demand: 1/ε = –(1/p)(q1 + q2)[P′(q1
+ q2)]. Thus the price-cost margin (p – C1′)/p (also called the ‘Lerner
index’) increases with the market share of firm i and decreases with
the (absolute value of the) elasticity of demand.
Activity
Derive the result (p – C1′ )/p = s1/ε from the first-order condition for firm 1.
Answer
Straightforward algebraic manipulations yield the result.
A third important property is that the Cournot price is lower than the
monopoly price. This can be seen for the special case of constant marginal
cost and isoelastic demand from the Lerner index equation above. It also
holds more generally. The intuition is that setting quantity in the Cournot
model implies a negative externality. When a firm produces more, on the
one hand, its own profit increases because of the increase in sales, and on
the other hand, both its own and rival profit fall because the market price
falls. But the firm only considers the adverse effect of the price fall on its
own profit, not on the profit of the other firm. In contrast, a monopoly
consisting of two divisions or two firms that maximise joint profit would
internalise this externality. Therefore a Cournot firm produces ‘too much’
44
output compared to the monopoly output. As a result, he market price is

lower than the monopoly price and aggregate profit lower than monopoly
profit.
To derive the equilibrium explicitly, let us further assume a linear demand
function q = a – p and constant marginal costs c1, c2. To ensure positive
equilibrium quantities we also assume that the exogenous demand shift
parameter a is sufficiently greater than c1 and c2. Then:
∂∏1 a – q2 – c1
= a – 2q1 – q2 – c1 = 0 ⇔ q1 = R1(q2) =
∂q1 2
and:
∂∏2 a – q1 – c2
= a – 2q2 – q1 – c2 = 0 ⇔ q2 = R2(q1) = .
∂q2 2
The functions R1 and R2 are the reaction functions of firm 1 and firm 2
respectively. Note that when a firm increases its quantity, the optimal
reaction of the other firm is to reduce its quantity. Solving the system of
the two equations we obtain the equilibrium quantities:
a – 2c1 + c2 a – 2c2 + c1
q1* = q2* = .
3 , 3
It can be easily checked that the second-order conditions are satisfied.
Activity
Derive the equilibrium profits.
Answer
Use the inverse demand function p = a – q1 – q2 to derive the equilibrium price p* and
then substitute p*, q1* and q2* into the profit functions ∏1 = q1(p – c1) and ∏2 = q2(p –
c2). You should obtain ∏1* = (a – 2c1 + c2)2/9, ∏2* = (a – 2c2 + c1)2/9.
We can draw several conclusions from these results:

• a firm’s output and profit decrease with own marginal cost and
increase with rival marginal cost
• the more efficient firm has a higher market share and profit than the
less efficient one: if c1 < c2, then q1* > q2* and ∏1* > ∏2*
• a firm’s output and profit increase in the demand parameter a
• profits are positive even when c1 = c2, a result which is in sharp
contrast with the Bertrand model.
The Cournot model and competition policy

The analysis proceeds along similar lines when the number of firms is
greater than two. An interesting property of the Cournot model with N
firms is that, under certain conditions regarding demand and costs, market
power and total industry profits increase as concentration in an industry
rises.4 This may be taken as one justification for the view that higher 4
See Tirole (1988),
concentration increases prices and profits because firms have more market pp.218–223.
power. Another justification is that higher concentration may facilitate

collusion among firms.
This potential link between concentration and market power has been
influential in the practice of competition policy. For instance, as discussed
in Chapter 11 of this guide, competition laws typically specify that a
merger would be contestable if the level of market concentration and/
or market shares after the merger or the change in concentration due to
45
the merger are higher than specified thresholds. Note that this does not
mean that any merger that satisfies the criteria for contestability is also
necessarily (or even likely) to be detrimental for competition or reduces
social welfare.
Concentration measures can be useful, but it is important to bear in mind
that concentration is itself endogenous, and that both concentration and
profitability are ultimately determined by basic industry characteristics,
such as technology, demand, and even the competitive regime (Cournot,
Bertrand, collusion or whatever). For instance, an increase in marginal
cost asymmetry across firms in a Cournot oligopoly raises both industry
concentration and industry profits – but the positive association between
the two is not causal. Moreover, as discussed in Chapter 10 of this guide,
a positive association between concentration and industry profits may not
exist at all once the endogeneity of concentration is taken into account.
Because profits are higher under quantity-setting (Cournot) than under
price-setting (Bertrand), the two models are often interpreted as
representing different degrees of competition, with Bertrand competition
being tougher than Cournot competition. This interpretation is valid for
games where firms do not make any long-run decisions, except perhaps
the decision to enter or not the market. But it is not appropriate when
firms make long-run choices such as investment, advertising, R&D, etc. 5
The reason will become
before setting prices or quantities.5 clear when you have
read Chapter 6 of this
Activities guide.
1. Which model, Cournot or Bertrand, do you think provides a better approximation

to each of the following industries: steel manufacturing, internet access, hotels,
insurance. Why?
2. Three common criticisisms against the use of the Cournot oligopoly model are (i) that
firms normally choose prices, not quantities, (ii) that firms usually do not take their
decisions simultaneously, and (iii) that firms often do not know their rivals’ costs. How
would you respond to these criticisms?
3. Consider a market where N symmetric firms produce a homogeneous product and
compete by simultaneously setting quantities. The inverse demand function is given
by p = a – Q, where Q is total quantity produced, that is Q = q1 + q2 + … + qN. The
marginal cost is constant and equal to c for all firms.
a. Derive the Cournot-Nash equilibrium. (Hint: After deriving the first-order
condition for firm i, use the symmetry of the model to significantly simplify the
computations.)
b. Derive the equilibrium price, profit for firm i and industry profit, and show that all
three are decreasing in the number of firms N.
4. One way to test the predictions of oligopoly models is to conduct laboratory
experiments. Of course, the conditions in the laboratory are very different from the
conditions that firms face in the real world. For one thing, players in laboratory
experiments are not very experienced and the stakes are much lower. Nevertheless,
there is a large literature on laboratory experiments in industrial economics. What
have we learned from it? Do agents behave as economic theory predicts? A useful
reference is the survey by Holt ‘Industrial organization: a survey of laboratory
research’, published in Kagel and Roth (1995). It is available online at:
http://people.virginia.edu/~cah2k/iosurvtr.pdf
Other useful references are the January 2000 and the January 2011 special issues
of the International Journal of Industrial Organization. All these references cover a
range of topics, from simple static games to more complicated dynamic games with
commitment, product differentiation, asymmetric information, and so on.
46
5. The digital economy has exploded in the last 20 years. It has lowered the cost of
creating and distributing certain types of products and services as well as the cost of
acquiring information about these goods and the cost of collecting and using data
on consumer tastes and behaviour. What are the main features of online markets and
how have they affected firms’ competitive strategies and traditional retail markets?
Two recent surveys of the economics of the internet address these and several other
issues: Ellison and Ellison (2005); and Levin (2011). And you can search the internet
for more.

• describe and derive the Bertrand paradox
• analyse how the introduction of capacity constraints in the Bertrand
model leads to equilibrium outcomes with price greater than marginal
cost and positive profits
• explain the theoretical foundations of the Cournot model
• analyse the Cournot model for various assumptions regarding the
demand, the number of firms, and the cost structures.

1. a. Describe and prove the Bertrand paradox.
b. Consider a market with two price-setting firms producing a
homogeneous product. The demand function is q = D(p) = 1 – p,
which implies the inverse demand p = 1 – q1 – q2. The two firms
have capacity constraints q1 and q2 , where q1 + q2 = 3/5. The
marginal cost of production is zero for qi ≤ q1 and infinite for qi >
q1 . Finally, assume that consumers are rationed according to the
efficient rationing rule.
i. Show that if q1 = q2 , there is a unique Bertrand-Nash
equilibrium where p1 = p2 = p* = 1 – q1 – q2
ii. Show that when q1 ≠ q2 , the equilibrium under part (i) breaks
down when the firms’ capacities are too dissimilar.
2. a. Consider a market with two quantity-setting firms producing a
homogeneous product. The inverse demand function is given by
p = 1 – q1 – q2 and the two firms have constant marginal costs c1
and c2 such that c1 + c2 = 2c, where c is a constant.
i. Compute the Cournot-Nash equilibrium.
ii. Show that as the two firms become more asymmetric (i.e. ci
moves away from c), total industry profit increases.
iii. Compute an index of concentration in this market and show
that it increases as the two firms become more asymmetric.
b. A researcher has estimated a model of industry profitability
using cross-industry data and has found a positive coefficient on
the concentration variable. He claims that the results show that
higher concentration leads to higher industry profit. What is the
theoretical basis for this claim? Do you agree with this conclusion?
What would your advice be to a policy-maker worried about the
high level of concentration in many industries?
47
Notes
48
Chapter 5: Dynamic price competition
Introduction
Attempts by firms to establish arrangements among themselves with a
view to increasing prices and profits are a recurrent theme in business
history and a matter of great concern for public policy. Such behaviour
is generally referred to as collusion and can be either explicit or tacit.
Explicit collusion to fix prices or market shares or to allocate customers or
geographical areas is illegal under most competition laws, including US
and EU law. The legal treatment of tacit collusion is more ambiguous.
Collusive behaviour poses a whole range of questions. Are such
arrangements stable? Under what conditions is collusion more likely to
occur? When do price wars occur and why? How can economists help in
identifying the existence of collusion in particular industries?
In the first part of this chapter, we will examine models where a given
set of firms interact repeatedly for a finite or infinite number of periods.
As mentioned in the previous chapter, repeated interaction is one of the
resolutions to the Bertrand paradox. In particular, this approach is very
useful in explaining how firms can set prices at a higher level than the
static Bertrand or even Cournot prices, even though they behave non-
cooperatively (i.e. they do not sign any binding agreements). In the second
part of the chapter, we will discuss an econometric methodology which can
be used – with variations, depending on the kind of data available and the
type of industry in question – to measure the degree of market power of
firms and detect collusive behaviour in any given industry.
Learning outcomes
• explain how firms that act non-cooperatively can collude
• use simple models of repeated games to discuss factors that facilitate
or hinder the stability of collusion and interpret any relevant empirical
findings
• describe and evaluate different theories of price wars
• describe and evaluate an econometric methodology for measuring
market power in any particular industry.
Essential reading
Books
10 and 12.
Journal
Porter, R.H. ‘A study of cartel stability: The Joint Executive Committee, 1880–
1886’, Bell Journal of Economics 14(2) 1983, pp.301–14.
49
Further reading
Books
Bresnahan, T.F. ‘Empirical studies of industries with market power’, in
Schmalensee, R. and R. Willig (eds) Handbook of industrial organization,
Volume 2. (Amsterdam: North-Holland, 1989).
2000) Chapter 8.
Pearson Addison Wesley, 2005) Chapters 5–6.
Jacquemin, A. and M.E. Slade ‘Cartels, collusion, and horizontal merger’, in
Schmalensee, R. and R. Willig (eds) Handbook of industrial organization,
Volume 1. (Amsterdam: North-Holland, 1989).
Pepall L., D. Richards and G. Norman, Industrial organization: contemporary
theory and empirical applications (Wiley-Blackwell, 2014) Chapter 14.
Symeonidis, G. The effects of competition. (Cambridge, MA: MIT Press, 2002).
Journals
Einav L. and J. Levin ‘Empirical industrial organization: a progress report’,
Journal of Economic Perspectives 24(2) 2010, pp.145–62.
Ellison, G. ‘Theories of cartel stability and the Joint Executive Committee’,
Fershtman, C. and A. Pakes ‘A dynamic oligopoly with collusion and price wars’,
Levenstein, M.C. ‘Price wars and the stability of collusion: a study of the pre-
World War I bromine industry’, Journal of Industrial Economics 45(2) 1997,
pp.117–37.
Levenstein, M.C. and V.Y. Suslow ‘What determines cartel success?’, Journal of
Economic Literature 44(1) 2006, pp.43–95.
Slade, M. ‘Strategic pricing models and interpretation of price-war data’,
European Economic Review 34 1990, pp.524–37.
Symeonidis, G. ‘In which industries is collusion more likely? Evidence from the
UK’, Journal of Industrial Economics 51(1) 2003, pp.45–74.
Symeonidis, G. ‘The effect of competition on wages and productivity: evidence
from the UK’, Review of Economics and Statistics 90 2008, pp.134–46.
1
We focus here on
Modelling collusion behaviour to increase
prices, or restrict
Firms that collude have generally three problems to solve. They must
quantities sold. Of
coordinate a price or a set of prices.1 They must ensure that collusion is course, firms can also
enforceable (i.e. that none of the parties has an incentive to cheat). And collude in other ways
they must sustain high profits against potential entry by other firms. (e.g. by regulating
capacity expansion or
Economic theory has focused much more on the second of these issues
by allocating territories).
than on the other two.2 There are several ways of modelling collusion Such arrangements
using formal games where firms interact repeatedly. All of them are seem, however, to
formalisations of the basic idea that firms may be able to sustain collusion be less frequent than
by threatening to retaliate in future periods in the event of a price cut collusion in price.
today. You should see these approaches as complementary, since there 2
Church and Ware
seems to be a variety of patterns of oligopoly behaviour in practice. (2000) discuss the
coordination issue.
The simplest and most popular approach is one which is based on the
theory of repeated games with complete information.3 Recall the one-shot 3
Tirole (1988),
Bertrand game of the previous chapter: two firms that produce a pp.253–61, presents
homogeneous product and have the same constant marginal cost c a not-too-technical
discussion of alternative
simultaneously set prices. We have seen that the Nash equilibrium
approaches.
50
involved p1 = p2 = c and zero profit for both firms. Now suppose that this
game is repeated T times, where T is finite or infinite, with the outcomes
of all preceding plays being observed by both firms before the next play
begins. The resulting game is a ‘repeated game’. At each period the firms
choose their prices simultaneously. Each firm seeks to maximise the
present discounted value of its payoffs from the T stage games.
What is the subgame-perfect equilibrium of this repeated game? Two cases
should be distinguished. If T is finite, then the unique subgame-perfect
equilibrium is for the firms to set p1 = p2 = c in every period.4 If, on the 4
See Tirole (1988),
other hand, T is infinite, then playing p1 = p2 = c in every period is still a p. 432.
subgame-perfect equilibrium, but it is no longer unique. In particular,
consider the following symmetric ‘trigger’ strategies:
• each firm sets the monopoly price pM in the first period (period 0).
Then it sets pM in any period t if in every period preceding period t
both firms have charged pM; otherwise it sets p = c forever.
These strategies constitute a subgame-perfect equilibrium of the repeated
game if the discount factor δ that firms apply to future profits is high
enough, in particular if δ ≥ 1/2.5 5
δ lies between 0 and 1.
δ close to 1 means that
To prove this result, we start by showing that these strategies constitute firms care a lot about
a Nash equilibrium. Suppose firm j adopts the above trigger strategy. We future profits and/or that
need to show that it is a best response for firm i to adopt this strategy as they interact frequently.
well. Take the first period. If firm i adopts the trigger strategy, then it sets A third interpretation
∏M of δ allows for the fact
price pM, obtains half the monopoly profit (assuming that the firms that the infinite horizon
2
share the monopoly profit equally), and faces the same choice in the next assumption need not be
taken literally. All that is
period. If playing the trigger strategy is optimal in the first period, it will
needed is some positive
also be optimal in the next and subsequent periods, so cooperation never probability at each
breaks down and the present discounted value of firm i’s payoffs is: period that the firms
∞ will interact again next
∏M ∏M
∑ δ (∏t
/2)=
M
2
(1 + δ + δ2 + - - -) =
2(1–δ)
. period; the higher this
t=0 probability, the higher
the value of δ.
If, on the other hand, the firm ‘defects’ in period 0, it gets maximum profit
ΠM in that period (less an arbitrarily small amount ε). Why? When firm
i defects it seeks to maximise its profit for that period given that firm j
sets the collusive price. Hence firm i will set a price marginally below pM
and will get the whole market. The firm gets zero profit thereafter, since
cooperation breaks down forever from the next period onwards. The
present discounted value of its payoffs is simply ΠM (ignoring ε).
Firm i will adopt the trigger strategy if the present discounted value of its
profits is higher when adopting the trigger strategy than when deviating
from collusion, namely if:
∏M
≥ ∏M ⇔ δ ≥ 1/2.
2(1 – δ)
The same argument can be made for every period in which all the
preceding outcomes have been (pM, pM). Finally, we also need to show that
if firm i finds itself in a period where all the preceding outcomes have not
been (pM, pM), its best response is to set pi = c forever. This is so, because
firm j will also set pj = c forever.
We have shown that the trigger strategies form a Nash equilibrium of the
game. To show it is also subgame-perfect, we need to show that it is a Nash
equilibrium in every subgame of the game. This is so, since every subgame
of this infinitely repeated game is identical to the game as a whole.
Thus collusion can be sustained in an infinitely repeated game if firms play
trigger strategies. Note that there are many equilibria of the game just
51
analysed. Replace pM with any price between c and pM. Then the trigger
strategies constitute a subgame-perfect equilibrium for δ ≥ 1/2. This is
part of a general result, which is referred to as the ‘Folk theorem’. For the
game analysed above, the Folk theorem says that there exist strategies that
form a subgame-perfect equilibrium such that firm i makes average per
period payoff Πi where Πi ≥ 0 (so the strategies are individually rational)
and Πi + Π j ≤ ΠM (so the strategies are feasible), for δ sufficiently close
to 1. In other words, anything between the static Bertrand profit (Πi = 0)
and the monopoly profit (Π i = ΠM) can be an average per period payoff
for firm i. These results can be generalised to N firms.
Activity
Consider a game similar to the one analysed above, but with N, rather than with two,
firms. Show that the specified trigger strategies constitute a subgame-perfect equilibrium
for δ ≥ 1 – 1/N.
Answer
Note that the per period payoff of each firm under collusion is ΠM/N. Then apply the same
argument as for the two-firm case.
More generally, if πC is the per period collusive payoff, πD the defection

payoff and πP the per period punishment payoff, collusion is sustainable
using trigger strategies if:
πC δ πP ⇔ δ ≥ δ* = π D – πC
(1+δ+δ2+…)πC ≥ πD + (δ+δ2+…)πP ⇔ ≥ πD + .
1–δ 1–δ πD – πP
This result implies that collusion is easier to sustain the lower the value of
δ*. Why? Because the lower the value of δ*, the larger the range of δ’s for
which collusion is feasible. Since ∂δ*/∂πC < 0, ∂δ*/∂πP > 0 and ∂δ*/∂πD
> 0, it follows that the higher the value of πC and the lower the values of
πP and πD, the more likely is collusion to occur in an industry.
Several points should be emphasised with respect to the above analysis.
• The game has an infinite number of equilibria. Collusion is sustainable,
but not necessary, and its ‘degree’ can vary. This theory is silent as to
how firms coordinate or which equilibrium will be chosen.
• Although the trigger strategies specified above may seem too harsh in
the sense that punishments last forever, similar results are obtained for
strategies that involve punishments that last only for a finite number
of periods, after which firms revert to collusion.
• An objection to the above model is that firms will have an incentive
to renegotiate the agreement once a defection occurs so as to avoid
punishment. But if they know they can do this, will they be able to
sustain collusion in the first place? The response to this objection has
been to construct more complicated models of repeated games which
have the property that collusion is sustainable even if renegotiation is
feasible.
• In the game we discussed, if firms play the trigger strategies, then
deviations do not occur along the equilibrium path. In other words,
this analysis explains collusion, but not price wars. As we will see,
price wars can nevertheless occur in more complicated models of
repeated games.
Repeated games of complete information are not the only way of
modelling collusion. Another approach, partly motivated by results of
52
experiments where agents interact for a finite number of periods, is based

on repeated games of incomplete information. Recall that cooperation
cannot be sustained as a subgame-perfect equilibrium in a finitely repeated
game of complete information. And yet in many experiments cooperation
between players does occur in long but finite repeated games – at least for
several periods.
To model this formally, we assume that players can be of two different
types, and that each player does not know the type of her opponent.
One type behaves in the standard way and never cooperates in a finite
repeated game, but the other type always cooperates unless she has been
cheated upon. There are multiple equilibria in such a game and one of
these involves cooperation for several periods and then reversion to non-
cooperation in the last few periods of the game – which is consistent with
the experimental results. We will say more on this type of analysis when
we discuss predation in Chapter 6 of this guide.
Factors that facilitate collusion

Can this analysis provide any insights as to the factors that facilitate or
hinder collusion? We have seen that collusion is feasible if the discount
factor δ that firms apply to future profits is at least as high as a certain
critical value δ*. Therefore the higher this δ*, the smaller is the range of δ’s
for which collusion is feasible, and hence the less likely is collusion to
occur. Now δ* was equal to 1/2 in the above example with two firms, but
in more general settings it will depend on various structural characteristics
of the industry in question. To see how this works, consider the following
6
See Tirole (1988),
application.6
pp.247–251, for more
Suppose that in our two-firm model it takes two periods, rather than one, examples.
to react to a defection. Then the present discounted value of the gain from
cooperation is still ΠM/2(1 – δ), but the present discounted value of the
gain from defection is ΠM + δΠM because the firm that defects can obtain
the monopoly profit for two periods before punishment begins. Collusion
is sustainable if ΠM/2(1 – δ) ≥ ΠM + δΠM ⇔ δ ≥ 1/21/2≈ 0.71. Note that
δ* is higher when the reaction lag is two periods than when it is one
period. In other words, for values of δ between 1/2 and 1/21/2 collusion
is not sustainable when the reaction lag is two periods, although it is
sustainable when the reaction lag is one period.
There seems in fact to be some evidence that collusion is more difficult
in industries where firms observe the prices set by their rivals with a
considerable time lag, as when manufacturers sell through specific deals
with a small number of large buyers. There is also evidence that firms
rarely collude on ‘long-run’ variables like capacity or R&D. Because these
long-term decisions take time to implement, reaction to rivals’ behaviour
is slower than in the case of price. In both these instances we observe that
retaliation lags hinder collusion. This is consistent with the predictions of
repeated game theory.
The theory of repeated games has provided an explanation for some of
the ‘stylised facts’ regarding collusion. Other facts, however, are more
difficult to explain. For instance, collusion occurs more often in capital-
intensive and in homogeneous good industries than in labour-intensive
or differentiated good industries. You should bear in mind that the theory
of repeated games focuses on the enforcement of collusion. This theory
predicts that enforcement is easier under certain circumstances: when
demand is stable or growing, there is low uncertainty about demand,
interaction is frequent and reaction lags are short, monitoring of rivals is
easy, market concentration is high, there is multimarket contact between
53
firms, there is little innovation in the industry, and there are no large cost
or size differences across firms.
On the other hand, being able to establish a collusive agreement and
to sustain profits in the face of potential entry are also necessary for
successful collusion in practice. Coordination is easier when there are few
firms in the industry, the product is homogeneous, product characteristics,
costs and demand are stable over time, uncertainty is low, and there are
no large cost or size differences across firms. Avoiding entry is easier in
industries with high capital intensity and high entry costs.
Symeonidis (2003) provides an empirical analysis of industry
characteristics that facilitate collusion using 1950s data from the UK.
Collusion was not illegal in the UK at the time, so there is relatively good
information on which industries were collusive and which were not.
Collusion was more likely to occur in capital-intensive, low-advertising,
low-R&D, moderately growing industries than in labour-intensive, high-
advertising, high-R&D, slowly or very fast growing industries.
Firms can also take steps to facilitate collusion. These are called
‘facilitating practices’ and include things such as giving advance notice
of a price change, exchanging information on prices and costs, meeting-
competition clauses (‘We will meet any price that a rival charges’) and
most-favoured-customer clauses (‘We will offer to all our customers the
lowest price we offer to any customer’). These practices usually work
either by increasing the observability of prices and hence facilitating
monitoring of rivals or by allowing firms to commit to punishing price
cuts. On the other hand, information exchanges, meeting-competition
clauses, etc. may also improve the knowledge of market conditions, help
reduce prices and promote competition.
Collusion is generally regarded as an anti-competitive practice that reduces
social welfare. This is obvious when entry is restricted and the only choice
variable firms have is price, or output: collusion raises prices and reduces
output below the socially optimal level. It is less obvious when entry is
free, and especially when firms also make long-run decisions on things like
product quality or product variety and cannot collude on these long-run
decisions as they can collude on price. Fershtman and Pakes (2000) and
others have shown that product quality and variety can increase under
price collusion, and the benefits for consumers could in principle even
compensate for the negative effect of collusive prices. However, normally
these effects are probably too small in practice to reverse the standard
negative welfare effect of collusive prices.
Symeonidis (2002, 2008) presents evidence that, under free entry, cartels
may reduce social welfare not so much through higher price-cost margins
but by slowing down productivity growth and by allowing for excessive
entry of firms into the collusive industries.
Price wars: theories and evidence

There are several theories of how and why price wars occur. They should
be seen as complementary because each seems to be relevant for some
industries and not for others. We briefly discuss three theories.7 7
Slade (1990)
provides an
In the Green–Porter model of price wars, there is uncertainty about the overview of the
level of demand at the time when firms choose their prices, or output three theories and
levels, in each period. Moreover, firms cannot observe the decisions of their describes some
rivals, past or present. So a cartel firm that has low sales in a given period empirical evidence
on price wars.
does not know whether this occurred because demand was low or because
54
some other firm deviated from the collusive agreement. In this model a
subgame-perfect equilibrium exists with the following properties: firms
collude, but every now and then a negative demand shock occurs which
may trigger a price war lasting for a certain number of periods before firms
revert to collusion. A distinctive feature of this model is that along the
equilibrium path of the game no firm ever cheats. Thus price wars are
accidents.8 Another feature of certain versions of the model is that firms do 8
The firms know this,
not achieve the joint monopoly payoff during a collusive period. That is, but collusion breaks
down nevertheless. If it
each of N (symmetric) firms makes a profit which is lower than ΠM/N. The
didn’t, then it would not
main testable prediction of the model is that price wars are triggered by be sustainable in the
negative demand shocks (i.e. they occur during recessions). first place.
An alternative theory of price wars is the one by Rotemberg and Saloner.
In their model demand again fluctuates randomly over time, but now
firms can observe the current state of demand before setting their price
each period. They can also observe their rivals’ past (but not their current)
actions. In this context the gain from collusion is not affected by the
fluctuations in demand since in the long run they level out. However, the
gain from cheating is bigger when current demand is high than when it
is low. Thus to ensure the stability of collusion the firms in a cartel may
need to reduce the collusive price and profit below the joint monopoly
level during a boom in order to also reduce the defection payoff and hence
the incentive to defect. Although this is not actually a price war (collusion
does not break down), the main prediction of the model is that prices may
move countercyclically. This is the opposite of the Green–Porter result.
However, this prediction is quite sensitive to the assumption that demand
shocks are not serially correlated.
Finally, according to a third theory of price wars, proposed by Slade, firms
have imperfect information about demand or about their rivals’ costs.
In particular, permanent changes in demand or costs of one firm are
observed imperfectly. So, for instance, a firm whose cost has decreased
may initiate a price war to gain market share; this also serves as a signal
to its rivals. Ultimately collusion is re-established. More generally, price
wars in this theory are devices whereby firms obtain information about
the new conditions in their industry. Unlike Green–Porter-type price wars,
which are punishment phases in an ongoing collusive agreement, the price
wars caused by permanent changes in demand or relative costs are actual
breakdowns in collusion and form part of a process of renegotiation of a
new agreement and a redistribution of the collusive profits.
The evidence suggests that each of the three theories is a good description
of collusive behaviour in some industries. Support for the Green–Porter
model seems to come from detailed studies of a US railroad cartel of
the 1880s (see next section). Using annual data from 1947 to 1981,
Rotemberg and Saloner have found that prices in the US cement industry
have moved countercyclically during that period. And Levenstein (1997)
has found that some at least of the price wars in the Pre-World War I US
bromine industry were breakdowns of collusion following a change in
relative costs of firms in the industry. More generally, some studies find
that unanticipated downturns in demand lead to price wars, others find
that both positive and negative large demand shocks destabilise collusion
and still others find no significant effect of the business cycle on cartel
stability one way or the other.
The evidence also points to a set of important factors for cartel stability
which have not been much emphasised in formal theoretical models – such
as entry barriers and internal cartel organisation. Levenstein and Suslow
(2006) provide a comprehensive review of the empirical literature on
55
the effects and the stability of cartels. They conclude that, while cartels
sometimes break down because of cheating, the biggest problem cartels
face is entry and the need to adjust the collusive agreement in response to
changing economic conditions.
Econometric analysis of market power and collusive

behaviour
The literature offers two types of empirical analyses of collusive behaviour
in individual industries: descriptive industry case studies and econometric
studies. We will examine a particular example of the econometric
approach: Porter’s study of a nineteenth century US railroad cartel, the
Joint Executive Committee (JEC). This study takes the case of a known
cartel and focuses on the issue of whether price wars occured and why.
However, the methodology used has much wider application. This type
of analysis can be used to examine oligopolistic behaviour in an industry
where there is no prior information about the occurrence of collusion. In
that case the objective is to measure the degree of market power of firms
in the industry. The question is whether prices and profits are closer to
those in a static Nash equilibrium or to those under the joint monopoly
outcome.9
9
Bresnahan (1989)
and Church and
This has significant policy implications. Competition policy authorities are Ware (2000),
very much interested in being able to infer the existence of collusion from pp.440–52, provide
observed market behaviour. This is not at all easy. Econometric analysis detailed discussions
of econometric
may help, although more direct methods – such as raids into company
approaches to
premises and search for relevant documents in cases of suspected collusion estimating market
– are also typically used. Note that if the competition authorities could power.
observe the demand and cost parameters in an industry, they could easily
say whether prices are ‘competitive’ or ‘collusive’. The problem is that
they do not observe these parameters; they often only observe prices and
quantities sold. The big advantage of the econometric approach outlined
here is that the econometrician is in the same position as the competition
authorities.
The JEC operated in the 1880s at a time when cartels were not illegal.
Between 1880 and 1886 there were several periods that look like price
wars. Were there in fact price wars, and why did they occur? What do they
tell us about how these firms colluded?
Porter (1983) builds a supply and demand model. The product is
homogeneous. Now a general problem in estimating a demand and supply
system is the ‘identification problem’: to identify the two functions, we
need at least one exogenous variable that affects demand but not supply
and at least one that affects supply but not demand. Porter’s econometric
specification of the demand function is:
ln Qt = α0 + α1 ln pt + α2Lt + u1t
where α1 is the elasticity of demand (a negative number), L is a dummy
variable taking the value 1 for periods where lakes were navigable (and
therefore an alternative transport mechanism existed) and 0 otherwise,
and u1 is an i.i.d. random variable.
To derive the supply function, Porter starts from the very basic conditions
that characterise firms’ behaviour. A general form of the cost function for
firm i is Ci(qit) = Aiqitλ + Fi. Profit maximisation implies setting marginal
cost equal to marginal revenue, that is:
56

 θit
λAiqitλ–1 = pt1+

α1 

where θ is a ‘degree of collusion’ parameter. To understand what θ is,
consider three behavioural regimes:
• in the Bertrand model, θit = 0, because marginal cost equals price
• in the Cournot model, we have seen in the previous chapter that
(p – Ci′)/p = si /|α1|, where si is the market share of firm i. Rearranging
we get Ci′ = p(1 – si /|α1|). Hence in the Cournot case θit = sit
• under joint monopoly (perfect collusion), θit = 1, because for a
monopolist the first order condition for profit maximisation is Ci′ = p(1
– 1/|α1|).
Thus θ∈[0,1], and the higher the value of θ the larger the ‘degree’ of
collusion.
Porter has no firm-level data, only industry data, so he aggregates. After
some manipulation, he ends up with an industry supply curve:

 θt
pt1+ = DQtλ–1

α1 
 ,
θ
where t = ∑ s θ
i it it
, Qt
=∑ q
i it and D is a constant.
Taking logarithms of both sides and writing this as an econometric

equation, he obtains:
ln pt = ß0 + ß1 ln Qt + ß2St + ß3It +u2t ,
where S are dummies to account for entry into and exit from the cartel,
I is a dummy taking the value 1 during a collusive period and 0 during a
reversion to ‘competitive’ (Bertrand) behaviour, and u2 is an i.i.d. random
variable. Now the coefficients of this econometric equation are functions
of the structural parameters of the theoretical model. In particular, ß0 =
ln D, ß1 = λ – 1, and ß3 = – 1n (1 + θt/α1). To calculate θt, the degree
of collusion during a collusive period, we need to obtain estimates of
ß3 and α1 from the above two-equation econometric model. Note that θt
= 0 would imply Bertrand behaviour, θt = ∑i sit2 would imply Cournot
behaviour, and θt = 1 would imply perfect collusion.
Two things are particularly important with the procedure outlined above:
• The econometric specification of the supply function was derived
directly from the first-order conditions of the firms’ maximisation
problem. So there is a direct link between theory and econometrics.
• All the data one needs to estimate the supply and demand model are
prices, quantities and some information on the various dummy
variables.10 Demand and cost parameters are not known and they are 10
Porter obtained
in fact coefficients to be estimated. information on It
from trade journals.
Estimation of the model using two-stage least squares and weekly data Alternatively, he used
over four years gave α^1 = –0.74 and ß^3 = 0.38. Hence θ^ = 0.34. Estimation ‘switching regression’
using switching regression techniques and maximum likelihood gave techniques to infer from
θ^ = 0.58. In both cases this is significantly lower than 1, which is consistent the data when It = 0
and when It = 1.
with the Green–Porter model (or at least some versions of it): collusive
behaviour is somewhere between Bertrand and joint monopoly.
However, the other prediction of the Green–Porter model, namely that
price wars are triggered by negative demand shocks, was not confirmed.
Porter looked at the residuals of the demand function. According to
the Green–Porter model, these should be negative during the periods
preceding a price war, but Porter did not find this.
57
Ellison (1994) has re-examined the JEC using a different econometric

technique to model the transition between collusive and competitive
regimes. He also assumed that the error term of the demand equation is
serially correlated and follows an AR(1) process. He estimated the same
two equations as Porter and obtained θ^ = 0.85, which is much closer to
the joint monopoly outcome than Porter’s estimate. He also found some
evidence that price wars were in fact triggered by low demand (consistent
with the Green–Porter model) and that occasional secret price-cutting
occurred in normal demand states as well (not consistent with the Green–
Porter model).
We now summarise the basic features of econometric analyses of
market power in general:
• They typically involve the estimation of demand and supply equations.
• Data at the firm level may be available, and this allows for a richer
model than the one which only uses industry data.
• Demand and cost parameters, as well as behavioural parameters
associated with the degree of collusion, are not known and must be
estimated.
• Typically we need data on prices and quantities over a considerable
time period. This is because we need a lot of variation in p and q to
estimate the model efficiently. This is made easier when there are
shocks in demand, costs or firms’ conduct.
• Several variants of the basic model exist. For example one may
distinguish between firms that are ‘leaders’ and firms that are
‘followers’ by assuming that the two groups of firms have different θ’s.
There are also some limitations of this methodology:
• Some assumptions must always be made, otherwise there will be
too many parameters to estimate – for instance that the elasticity of
demand is constant or that θ is the same for all firms within a group
and over time. The results may depend on the assumptions made.
• The methodology can be extended to differentiated good industries,
but more restrictions need to be imposed for the model to be
estimated.
• The results may depend on the econometric estimation method.
Activities
1. It is often claimed that collusion is harder to sustain in industries where firms face
large and infrequent orders from buyers because infrequency of interaction between
firms hinders collusion. What is the theoretical basis for this claim?
2. Two firms set prices simultaneously, either high (H) or low (L), in each of an infinite
number of periods. The prices are observed by both firms before the next play begins.
The discount factor is δ < 1. The per period payoffs are as follows. If both firms play
H, each gets 10. If both firms play L, each gets 7. If one firm plays H and the other
plays L, their respective payoffs are 5 and 12. Identify the Nash equilibrium of the
one-period game. Then describe a ‘trigger strategy’ that can enable the firms to reach
a more profitable outcome non-cooperatively. How high must the discount factor be
for this outcome to be sustainable?
3. Two firms produce a homogeneous good and compete by setting prices each period
for an infinite number of periods. Each of the two firms owns a minority share k of its
rival. This share is small so that each firm keeps full control of its own activities and
decisions; the rival just receives share k of the firm’s profits. Analyse how this pattern
of cross-ownership affects the likelihood of collusion between the two firms.
58
4. Levenstein and Suslow (2006) write: ‘Cartels break up occasionally because of

cheating or lack of effective monitoring, but the biggest challenges cartels face
are entry and adjustment of the collusive agreement in response to changing
economic conditions. Cartels that develop organizational structures that allow
them the flexibility to respond to these changing conditions are more likely to
survive. Price wars that erupt are often the result of bargaining issues that arise
in such circumstances. Sophisticated cartel organizations are also able to develop
multipronged strategies to monitor one another to deter cheating and a variety of
interventions to increase barriers to entry.’
Assess the empirical evidence on which this statement is based. The Levenstein and
Suslow article is a good starting point, but you are also encouraged to check for more,
including information from antitrust cases. Once you have assessed the evidence,
think about its implications for economic theories of collusion and price wars.
5. Do cartels matter much? Assess the empirical evidence on their impact on prices and
profits as well as on non-price variables such as advertising, innovation, investment,
productivity, and market concentration. Once again, the Levenstein and Suslow survey
is a good starting point, but you are also encouraged to check for more, including
econometric studies of cartels and information from antitrust cases. What are the
implications of this evidence for the welfare effects of cartels and for competition
policy?
6. Since it is often difficult to detect collusion, competition authorities in the US, EU
and elsewhere operate so-called ‘leniency programs’. These differ in their details,
but they all involve lenient treatment of the first firm that provides evidence that
leads to successful prosecution of a cartel, while everybody else is subject to heavy
fines. The effect of these programs has been to increase significantly the number
of detected and successfully prosecuted cartels. On the other hand, a leniency
program could increase the expected net gain of forming a cartel and could lead
to more cartels being formed. What is then the overall welfare effect of leniency
programmes? And what is the optimal design of such a program? This is currently an
active area of research. Good starting points are Motta and Polo (2003); and Brenner
(2009). Search also the internet for recent theoretical, empirical and experimental
contributions.

• explain how firms that act non-cooperatively can collude
• use simple models of repeated games to discuss factors that facilitate
or hinder the stability of collusion and interpret any relevant empirical
findings
• describe and evaluate different theories of price wars
• describe and evaluate an econometric methodology for measuring
market power in any particular industry.

1. Consider two identical firms with constant marginal cost c which
compete in quantities in each of an infinite number of periods. The
quantities chosen are observed by both firms before the next play
begins. Inverse demand is given by p = 1 – q1 – q2. The firms use
‘trigger strategies’ and they revert to static Cournot behaviour if
cooperation breaks down.
59
a. What is the lowest value of the discount factor δ such that the
firms can sustain the monopoly output level?
b. Suppose δ is too small to sustain the monopoly output. In
particular, suppose δ = 1/2. What is the most profitable subgame-
perfect equilibrium that can be sustained using trigger strategies?
(Assume c = 0 for simplicity.)
2. Describe an econometric study of oligopolistic behaviour, including
a discussion of the link between theory and econometric modelling.
What are the limitations of this type of empirical work?
60
Chapter 6: Entry deterrence, entry accommodation and predation
Chapter 6: Entry deterrence, entry

accommodation and predation
Introduction
One of the most interesting aspects of competition among firms is the use
of irreversible actions that affect the future behaviour both of the firm that
takes the action and of its rivals. In this chapter we examine a variety of
strategic situations involving first-mover advantages, strategies to deter
entry, and strategies to increase future profit by undertaking a costly and
not easily reversible action today. A common theme in all these situations
is that it may be profitable for a firm to limit its own flexibility by making
a ‘commitment’, that is a long-run (difficult to change) decision, because in
this way it influences the behaviour of its rivals. Of course, if a firm cannot
influence the behaviour of its rivals, it will never find it profitable to limit
its own flexibility. This would be the case if the commitment cannot be
observed by the rivals.
We start with some preliminaries: the notions of strategic substitutes and
strategic complements. Then we examine a simple model that highlights
the value of irreversible decisions, and we use this model to analyse entry
deterrence and entry accommodation strategies. Finally, we
develop a more general framework that can be used to analyse in a unified
way a wide range of strategic situations. A game-theoretic device that is
very useful in tackling these issues is the multi-stage (usually two-stage or
three-stage) game. Unlike the static games of Chapter 4, these games are
dynamic. Unlike the repeated games of Chapter 5, they do not consist in
the simple repetition of a stage-game. The solution concept for these
games is the subgame-perfect equilibrium and the procedure for solving 1
See Tirole (1988),
them is backward induction.1
pp.428–432, or Shy
In the final part of the chapter we discuss predation, a strategy that (1995), pp.22–27.
involves a reduction in profit or even a loss in the short run in order to
drive competitors out of the market so that larger profits can be made in
the long run.
Learning outcomes
• explain the notions of strategic substitutes and strategic complements
• analyse simple models with sequential actions and describe first-mover
advantages in these models
• explain under what conditions a firm may be able and/or willing to
deter the entry of a potential rival
• use reaction functions to analyse strategic investment decisions by
firms in situations where the investment influences the firms’ future
profit functions
• analyse firms’ predatory tactics (i.e. behaviour that aims to induce exit
of competitors).
61
Essential reading
Books
13–16, 21.
Tirole, J. The theory of industrial organization. Chapter 8 and part of Chapter 9.
Journal
Schmalensee, R. ‘On the use of economic models in antitrust: the Realemon
case’, University of Pennsylvania Law Review 1979, pp.994–1050.
Further reading
Books
2000) Chapter 15.
Pearson Addison Wesley, 2005) Chapters 11 and 13.
Gilbert, R.J. ‘Mobility barriers and the value of incumbency’, in Schmalensee,
Ordover, J.A. and G. Saloner ‘Predation, monopolization, and antitrust’, in
Schmalensee and Willig (eds) (1989).
Chapters 11–13.
Journals
Bulow, J.I., J.D. Geanakoplos and P.D. Klemperer ‘Multimarket oligopoly:
Strategic substitutes and complements’, Journal of Political Economy 93(3)
1985, pp.488–511.
Fudenberg, D. and J. Tirole ‘The fat cat effect, the puppy dog ploy and the lean
and hungry look’, American Economic Review, Papers and Proceedings 74(2)
1984, pp.361–68.
Lieberman, M.B. ‘Excess capacity as a barrier to entry: An empirical appraisal’,
Journal of Industrial Economics 35(4) 1987, pp.607–27.
Singh, S., M. Utton and M. Waterson ‘Strategic behaviour of incumbent firms
in the UK’, International Journal of Industrial Organization 16 1998,
pp.229–52.
Strategic substitutes and strategic complements

Recall the Cournot model with two firms. With linear demand D(p) = a –
p and constant marginal costs c1, c2 the reaction functions are:
a – q2 – c1 a – q1 – c2
q1 = R1(q2)= , q2 = R2(q1)= .
2 2
Note that dqi/dqj < 0 (i.e. the reaction functions are downward-sloping)
(see Figure 6.1). When the reaction functions are downward-sloping,
we say that the actions of the firms are ‘strategic substitutes’. The
interpretation is that when firm j increases its quantity qj, the optimal
reaction of firm i is to decrease its quantity qi. In other words, if firm j
becomes more aggressive, it causes firm i to become less aggressive. More
generally, it can be shown that sign(dRi /dqj) = sign(∂2∏i/∂qi∂qj). Under
strategic substitutability, this cross-partial derivative is negative: if firm
62
j increases qj, the profit that firm i makes from a unit increase in qi (i.e.
∂∏i/∂qi) falls.
q2 p2
R1(q2) R1(p2)
Nash equilibrium
R2(p1)
R2(q1)
Nash equilibrium
q1 p1
Figures 6.1 and 6.2
In the Cournot model the firms’ action is setting quantity. It turns out that
when the strategic variable is quantity, reaction functions are typically
downward-sloping and we have strategic substitutability. However,
this may also be the case for several other decision variables – strategic
substitutability is not tied to any specific strategic variable.
If, on the other hand, the reaction functions are upward-sloping (Figure
6.2), we say that the actions of the firms are ‘strategic complements’. This
is typically the case when the products are differentiated and substitutes in
demand and firms compete in prices. Suppose, for example, that demand
for good 1 is given by q1 = a – bp1 + dp2 and demand for good 2 is given
by q2 = a – bp2 + dp1, where a, b, d are all positive parameters.2 The 2
Note that d > 0
products are differentiated because, even if pi > pj , qi ≠ 0 – unlike the implies that the two
goods are substitutes
Bertrand model for homogeneous products. Firm i, i = 1, 2, chooses pi to
in demand. Demand
maximise its profit ∏i = (pi – ci)qi = (pi – ci)(a – bpi + dpj), taking pj as substitutability or
given. The first-order condition is: complementarity should
not be confused with
∂∏i a + dpj + bci
= a – 2bpi +dpj + bci = 0 ⇔ pi = Ri(pj) = , strategic substitutability
∂p i
2b or complementarity.
which defines the reaction function of firm i. Note that dpi/dpj > 0.
The interpretation is that when firm j decreases its price pj , the optimal
reaction of firm i is to also decrease its price pi. In other words, if firm j
becomes more aggressive, it causes firm i to also become more aggressive.
Now sign(dRi /dpj) = sign(∂2∏i/∂pi∂pj), and this expression is positive
under strategic complementarity. That is, the profit that firm i makes from
a unit increase in pi (i.e. ∂∏i/∂pi) rises if firm j increases pj. As was the case
with strategic substitutability, strategic complementarity is not tied to any
specific strategic variable.
First-mover advantage, entry deterrence and entry

accommodation
We begin with the simplest possible model that analyses first-mover
advantages and highlights the value of long-run commitments. There are
two firms in an industry and they play a multi-stage game. At stage 1, firm
1 chooses a level of capacity K1 which then remains fixed. At stage 2, firm
2 observes K1 and then chooses a level of capacity K2. You may think of this
as a three-stage game, with the two firms simultaneously setting prices at
stage 3, given their respective capacities. However, we will abstract from
price competition at stage 3 in what follows to focus on the sequential
63
choice of capacities. Note that this is a version of the so-called Stackelberg

model of duopoly, with capacity rather than quantity as the choice variable.
Let the reduced-form profit functions (after solving for short-run price
competition with given capacities) be ∏1(K1,K2) = K1(1 – K1 – K2) and
∏2(K1, K2) = K2(1 – K1 – K2).3 Important properties of these profit
functions are ∂∏i/∂Ki < 0 and ∂2∏i/∂Ki∂Kj < 0 – the second of these says 3
If we were solving for the
that capacities are strategic substitutes. To compute the subgame-perfect subgame-perfect equilibrium
of the three-stage game, we
equilibrium of the game we use backward induction. At stage 2, firm 2
would start from stage 3
chooses K2 to maximise ∏2 taking K1 as given. Setting ∂∏2/∂K2 = 0, we get and solve for the equilibrium
K2 = (1 – K1)/2, the reaction function of firm 2. At stage 1, firm 1 chooses prices, taking capacities
K1 to maximise ∏1 anticipating the behaviour of firm 2 at stage 2 (i.e. as given. Then we would
anticipating its reaction function). That is, firm 1 chooses K1 to maximise substitute these equilibrium
∏1 = K1(1 – K1 – K2) subject to K2 = (1 – K1)/2. Essentially, firm 1 can prices into the firms’ profit
functions and we would
‘choose’ any point on firm 2’s reaction function as an equilibrium of the
thus obtain profit functions
game. Naturally, it chooses the point where its own profit is largest. in terms of K1 and K2 only.
Substituting the reaction function of firm 2 into ∏1 we obtain ∏1 = K1 Then we would proceed
[1 – K1 – (1 – K1)/2]. Setting ∂∏1/∂K1 = 0 we obtain the equilibrium backwards to stage 2.
value K1* = 1/2. Finally, substituting this into the reaction function of
firm 2 we obtain K2* = 1/4. Equilibrium profits are ∏1* = 1/8 and ∏2* =
1/16 < ∏1*.
Now compare these profits with those in a game where firms choose
capacities simultaneously (the equivalent of the Cournot model). In that
case K^1 = K^2 = 1/3 and ∏ ^1 ^2
= ∏ = 1/9 < 1/8 = ∏1* (check these results). We
conclude that by exploiting its first-mover advantage, firm 1 makes higher
profit than firm 2 in the sequential game: ∏1* > ∏2*. Firm 1 also makes
higher profit than the profit it would make in the simultaneous-move
game: ∏1* > ∏ ^ 1. The reverse is the case for firm 2. The reason is that, by
committing itself to a given level of capacity, firm 1 influences the choice of
firm 2: firm 1 builds more capacity than it would build in a simultaneous-
move game (K1* > K^1) and, because capacities are strategic substitutes,
firm 2 builds less capacity (K2* < K^2) as a result. Note also that total
industry capacity is higher and industry profit is lower in the Stackelberg
game than in the Cournot game.
What is crucial for this strategy to work is that the first mover’s decision
is irreversible. If firm 1 could change K1 at stage 2 of the game, then the
overall outcome would be the Nash equilibrium of the simultaneous-move
game. So if capital can be easily resold in the second-hand market, there is
no scope for a credible commitment. In other words, for the commitment
to be credible the investment must be sunk.
Does the first mover always have an advantage in more general settings?
No – it depends on the slope of the reactions functions. If the actions of
the firms are strategic complements, the ‘leader’ makes less profit than the
‘follower’. See below for more on how the slope of the reaction functions
influences firm strategy and profitability.
Strategic entry deterrence and entry accommodation

So far in this model, firm 1 cannot strategically deter firm 2 from entering
the industry and building some capacity. Let us now extend the model to
allow for this possibility. Assume that there is a fixed cost of entry f < 1/16.
The timing of the game is the same. At stage 1, firm 1 enters the industry
and chooses a level of capacity by comparing its profit under two possible
courses of action (to be specified below). At stage 2, firm 2 observes what
firm 1 has done and decides whether or not to enter and with how much
capacity.
64
If firm 1 decides to ‘accommodate’ the entry of firm 2, then we have the

case already examined above: K1* = 1/2, ∏1* = 1/8 and ∏2* = 1/16.
Note that these profits are now profits gross of the cost of entry. Net
equilibrium profits are ∏i* – f, for i = 1, 2.
What if firm 1 tries to deter the entry of firm 2? Firm 2 will not enter if
the gross profit it expects to make by entering does not exceed the entry
cost, that is if ∏2(K1,K2) = K2(1 – K1 – K2) ≤ f.4 When will this be the case? 4
If Π2(K1, K2) = f,
Suppose that firm 1 installs capacity K1. If firm 2 were to enter, it would firm 2 is indifferent
then choose K2 to maximise ∏2 taking K1 as given. The optimal capacity between entering and
not entering. Let us
level of firm 2 would be given by its reaction function K2 = (1 – K1)/2.
assume, to simplify the
Hence the maximum gross profit of firm 2, conditional on K1, would be: exposition, that in this
  
 1 – K1 1 – K – 1 – K1 =  1 – K1
2 case it does not enter.
  
∏2*(K1) = K2(1 – K1 – K2)=  
  2 .
 2
 
 1
2 
Note that ∂∏2*/∂K1 < 0 for all relevant values of K1: the higher the level of
K1, the lower the value of ∏2*. So there is a level of K1, say K1b, that just
deters the entry of firm 2. K1b is defined by a zero net profit condition for
firm 2, namely ∏2*(K1b) = f ⇔ (1 – K1b)2/4 = f ⇔ K1b = 1 – 2f1/2 > 1/2
(since f < 1/16). In other words, any K1 ≥ 1 – 2f1/2 will deter the entry of
firm 2 because it will result in firm 2 not making enough gross profit to
cover its entry cost in case of entry.
We still need to find out under what conditions the entry deterrence
strategy is more profitable than the entry accommodation strategy for firm
1. The gross profit of firm 1 when entry of firm 2 is deterred is:
∏1 = K1b (1 – K1b – K2) = K1b(1 – K1b) = (1 – 2 √ f )[1 – (1 – 2√ f )] = 2√ f – 4f .
Note that we have set K2 = 0 and K1b = 1 – 2f1/2.5 Hence entry deterrence is 5
Firm 1 will never want
more profitable than entry accommodation if and only if 2f1/2(1 – 2f1/2) > to install more capacity
than K1b. This is easy
1/8 ⇔ f > 0.0053.
to see from the profit
So the choice of firm 1 depends on the magnitude of the entry cost f. This function of firm 1 when
makes sense. If f is small, then firm 1 must install such a large capacity if K2 = 0: ∏1 is decreasing
it wants to deter the entry of firm 2 that entry accommodation is in fact in K1 for values of K1
higher than 1/2 and
preferable. If f is large enough, then a smaller capacity will do the job and
hence for any K1> K1b
entry deterrence becomes preferable to entry accommodation. (recall that K1b > 1/2).
Note that in our model, if firm 1 builds capacity K1b and deters the entry of
firm 2, it must then use all this capacity – even though the optimal level
of K for a monopolist is lower than K1b (it is actually equal to ½). This is
necessary for entry deterrence to work in our model because we have
assumed, for simplicity, that investment in capacity is costless and that
only entry is costly – hence the stage 1 decision would not be irreversible
if firm 1 were allowed not to use in stage 2 some of the installed capacity.
Assuming costly investment in capacity would not change the qualitative
results, it would only change the critical value of f for which entry
deterrence is just as profitable as entry accommodation for firm 1.
An interesting question in more complex models of entry deterrence than
the one presented here is whether a firm may have an incentive, as part
of an entry-deterring strategy, to make a costly investment in capacity that
will not be used for production and just serves to deter entry of potential
rivals. There is some theoretical support for this strategy, but the empirical
evidence is mixed. In particular, Lieberman (1987) found that in only
three out of 38 chemical products in his sample did incumbent firms hold
‘excess capacity’ to deter entry. More common was a strategy of capacity
expansion by incumbent firms once entry had occurred in an attempt to
deter expansion by the entrants.
65
A final remark: investment in capacity is only one of a number of possible

entry deterrence strategies available to firms. Others include product
proliferation (see Chapter 7 of this guide), tying (see Activity 3 below),
binding contracts between buyers and sellers, and investment in product
quality, cost reduction or brand image.
Contestable markets
We make here a brief reference to a theory of market structure proposed
by Baumol, Panzar and Willig in the early 1980s. The main idea is that,
if potential entrants can costlessly enter and exit an industry, incumbent
firms must set price equal to average (and in some cases even marginal)
cost, and cannot make positive profits. The requirements for this
mechanism to work is that entry does not involve any sunk costs and the
incumbents are rather slow in responding to entry by cutting price. These
are strong assumptions, and they are generally not satisfied in practice. It
is therefore not surprising that the theory of contestable markets has had
little empirical support. It has rather helped focus attention on the role of
sunk costs for entry deterrence and the determination of market structure.
A taxonomy of business strategies

Consider now a more general setup as follows. There are two firms and
two periods. In period 1 one or both firms take an action that influences
their own and the rival’s payoff function in period 2. This can be an
investment in capacity or R&D which affects marginal cost in period 2.
Or it can be an investment in creating a brand image which increases the
consumers’ willingness to pay for the product, or a distribution network,
etc. In period 2 the firms compete in prices or in quantities. This kind of
setup can be used to analyse a wide range of strategic situations. What is
interesting is that all these can be analysed on the basis of a small number
of basic properties.
To fix ideas, suppose that at stage 1 there is no production and one of the
firms, say firm 1, chooses a level of investment K1. Production takes place
at stage 2. Assume a fixed cost of production f and constant marginal
costs c1 and c2. Let the marginal cost of firm 1 depend on its choice of
investment: c1 = c1(K1), with dc1/dK1 < 0.
In determining its choice of K1 firm 1 will compare its overall profit
from two possible courses of action: deterring entry (i.e. production)
of firm 2 or accommodating entry of firm 2. Take first the case of entry
deterrence. This is the relevant case if f is large. The question is: should
firm 1 ‘overinvest’ or ‘underinvest’? Let us make precise these notions of
overinvestment and underinvestment. Relative to what? Relative to
6
In other contexts, an
increase in the strategic
the level of investment that would be optimal for firm 1 if firm 2 could not
variable by a firm may
observe K1. In that case there would be no first-mover advantage, so the cause that firm to
game would be equivalent to a simultaneous-move game: firm 1 would behave less aggressively
not be able to influence the behaviour of firm 2 by making a commitment. in subsequent periods.
Let us call this the ‘benchmark’ level of K1. We would then say that
investment makes the
If competition at stage 2 is in quantities, the reaction functions at stage 2 firm ‘soft’. See Tirole
are downward-sloping. By increasing K1 firm 1 reduces c1, hence it shifts (1988), pp.323–328, for
its stage 2 reaction function outwards. That is, the optimal reply for any more precise definitions
of ‘toughness’ and
given q2 is a higher q1. Thus a higher K1 leads to more aggressive
‘softness’ and, more
behaviour by firm 1 at stage 2. We describe this property of our model by generally, for a more
saying that investment makes firm 1 ‘tough’.6 Now as K1 increases, q1* rises analytical exposition
and q2* falls. See Figure 6.3, in which point O represents the stage 2 of the issues discussed
equilibrium when firm 1 chooses the benchmark level of capacity. Also, as here.
66
K1 increases, the stage 2 equilibrium profit of firm 2 gross of the fixed cost
f decreases. At some point it becomes less than f, so firm 2 will prefer not
to produce at all. We conclude that to deter the entry of firm 2, firm 1
must overinvest.
q2
R1 R1‘
increase in K1
O‘
R2
q1
Figure 6.3
Activity
Using reaction functions, show that if stage 2 competition is in prices, we again need
overinvestment to deter entry.
Answer
Under price competition, the reaction functions at stage 2 are upward-sloping. An
increase in K1 shifts the stage 2 reaction function of firm 1 to the left. Equilibrium prices
fall, and profits also fall – at least initially, when starting from the benchmark level of K1.
So, unless f is very small, the equilibrium profit of firm 2 will at some point become less
than f and the entry of firm 2 will be deterred.
Take now the case of entry accommodation. That is, firm 1 chooses K1
given that it will have to compete with firm 2 at stage 2. This is the
relevant case if f is small. Should firm 1 overinvest or underinvest, in the
same sense as above?
If stage 2 competition is in quantities, an increase in K1 shifts the reaction
function of firm 1 outwards (see again Figure 6.3). An increase in K1
above the benchmark level causes the market share of firm 1 to rise at the
expense of firm 2, so up to a certain point the stage 2 equilibrium profit of
firm 1 also rises. In short: because the reaction functions are downward-
sloping, more aggressive behaviour by firm 1 (a rise in q1) leads to a softer
action by firm 2 (a fall in q2). In this situation, firm 1 has an incentive to
overinvest, given that investment makes it ‘tough’.
p increase in K1
2
R1 ‘ R1
R2
O
O‘
p1
Figure 6.4
67
If stage 2 competition is in prices, an increase in K1 shifts the reaction

function of firm 1 to the left (see Figure 6.4). This is because an increase
in K1 causes c1 to fall, so for any price p2 firm 1’s optimal reply is a lower
p1. Again a higher K1 leads to more aggressive behaviour by firm 1 at stage
2, that is investment makes firm 1 ‘tough’. An increase in K1 above the
benchmark level causes a fall in both p1* and p2*, so up to a certain point
the stage 2 equilibrium profit of firm 1 falls. In short: because the reaction
functions are upward-sloping, more aggressive behaviour by firm 1 (a fall
in p1) leads to a more aggressive response by firm 2 (a fall in p2). In this
situation, firm 1 has an incentive to underinvest, given that investment
makes it ‘tough’.
The framework presented above can be used to analyse a variety of
strategic situations, including product differentiation, learning by doing,
the imposition of price floors or quotas, multimarket contact, etc.7 In all
7
See Tirole, pp.328–
336, and Church and
these cases, the investment decision, and in particular the decision to
Ware, Chapter 16, for
strategically overinvest or underinvest, depends on three things: discussion of these and
• Does a firm want to deter or accommodate entry? other examples.
• Does investment make a firm ‘tough’ or ‘soft’?

• Are the stage 2 actions strategic substitutes (downward-sloping
reaction functions) or strategic complements (upward-sloping reaction
functions)?
Activity
It is sometimes claimed that ‘learning-by-doing’ through overproduction is a profitable
strategy since it allows firms to reduce their future costs and even deter the entry of rivals.
Do you agree?
Answer
Learning by doing may or may not be a profitable strategy for a firm. It may not be
profitable when reducing future costs results in tougher competition. In particular, when
there is initially only one firm in the market, learning by doing works if the objective is to
deter entry or the firms’ actions are strategic substitutes. However, it is not profitable if
the objective is to accommodate entry and the firms’ actions are strategic complements.
When there are initially two firms in the market, both of which can use the learning-by-
doing strategy, the results are more ambiguous.
A final remark: It should be clear now why the quantity-setting model and
the price-setting model cannot be interpreted as representing different
degrees of competition in games where firms make long-run choices
such as investment in capacity, R&D, and so on before setting prices or
quantities. The differences between quantity-setting and price-setting in
these games are driven by the distinction between strategic substitutability
and strategic complementarity, not by the degree of competition.
Predation
A predatory strategy is one that that involves a reduction in profit or even
a loss in the short run – through a price cut, an increase in advertising
expenditure, and so on – in order to drive rival firms out of the market so
that larger profits can be made in the longer term. A variant of this is when a
firm attacks (preys upon) a rival in market X so as to deter this or other rivals
from entering into other markets Y, Z, etc. where the firm also operates.
According to the ‘Chicago School view’, cutting price is costly and more
so for a dominant firm with a large market share – since the price cut
68
applies to a larger base of output for a dominant firm than for a smaller
firm. Furthermore, the threat to continue predatory pricing forever is not
credible, since the firm will keep losing money. Finally, the notion that
once competitors are driven out of the market the price will be raised and
larger profits will be made is not credible either, since the high price will
attract entry in the future. For these reasons, the Chicago School view
maintains that predation cannot be a rational behaviour by firms and will
be very rare in practice.
Since predation is generally difficult both to define (what price is low
enough to qualify for predation?) and to detect (what is predation and
what is competitive price cutting?), the Chicago School view has been
influential. And yet there have been several cases where firms did appear
to use predatory tactics. How can this be justified as rational behaviour by
firms?
There are two main economic theories of predation as rational firm
behaviour: the long purse theory and the reputation model.
Economic theories of predation

According to the long purse (or ‘deep pockets’) theory, a firm may have
larger financial reserves and may therefore be able to sustain losses for
longer than a rival. Even though it cannot sustain losses indefinitely, it may
be able to do so for the period needed to drive the rival bankrupt and out
of the market. In fact, the rival may anticipate this and exit immediately
rather than sustain losses over many periods.
A necessary condition for this to work is that there are imperfections
in financial markets, so that the rival is financially constrained. If, for
instance, there is asymmetry of information between lenders and firms,
then a lender will not know whether the rival firm is losing money because
it is the victim of predation or simply because it is inefficient. A lender may
therefore be reluctant to provide financing. Predation can be successful in
driving out the rival in this case.
An alternative theory of predation relies on the notion that firms often
have incomplete information about their competitors’ payoff functions
– for instance, they may not know their competitors’ costs. The motivation
for such models has come not only from antitrust cases but also from a
well-known result in game theory: the ‘chain-store paradox’.
Suppose an incumbent monopolist operates in a finite number of distinct
markets, say N. It faces N potential entrants, one in each market, who
must make their entry decisions sequentially. In each market the following
stage game is played: first the potential entrant decides whether to enter
or not; then the monopolist either preys (starts a price war inflicting losses
on both himself and the entrant) or accommodates. The overall game
consists of this stage game being played N times. The paradox is that at
the subgame-perfect equilibrium of this game predation never occurs and
the potential entrants enter in all markets. Why? In the last period, the
monopolist is better off accommodating and hence the potential entrant
enters. In the period before the last one, the monopolist realises that his
action will have no effect on the outcome of the last period, so again he is
better off accommodating and therefore again the potential entrant enters.
And so on by backward induction to the first period.
And yet there is evidence that in practice predation does sometimes
occur in such situations. The way to resolve the paradox is to assume
that the incumbent monopolist can be of two different types, and that
the potential entrants do not know the monopolist’s type and they can
69
only guess, on the basis of prior probabilities as well as the monopolist’s

actions during the game. One type always preys, even if there is only
one round of play – perhaps because he is ‘crazy’ and enjoys preying or
because he has considerable capacity and can benefit from a low average
cost when selling a lot of output. The other type is ‘sane’ and would always
accommodate entry if there were only one round of play. However, in
the chain-store context the sane type may try to convince the potential
entrants that he is crazy with sufficiently high probability. Why would he
do this? To deter future entry. How might he convince them? By acting
crazy for a while.
In particular, there is an equilibrium in this game of incomplete
information (called ‘perfect Bayesian equilibrium’) with the following
properties:
• Entry does not occur in the first few markets. With a lot of markets still
at stake, the sane monopolist would prey to build up his reputation if
entry occurred, and the potential entrants realise this – so they don’t
enter.
• Since the number of markets to defend decreases over time, the
monopolist’s incentive to prey is reduced after the first few markets,
so potential entrants are more encouraged to enter. Actually, the
equilibrium involves randomising in this intermediate range of
markets: the potential entrants may or may not enter and the
monopolist may or may not prey in case of entry.
• Entry may or may not occur in the last few markets, but if it occurs it
is always accommodated.
The central idea of this story is that in the presence of incomplete
information a firm may have an incentive to strategically manipulate
its rivals’ beliefs (i.e. to create a reputation for itself – in the case of
predation, a reputation for toughness).
Although the chain-store story is more about predation to deter entry
than predation to induce exit, it can be easily modified to account for
rational predation in a single market where more than one firm operate. A
reputation model of predation leading to the exit of a rival is a multi-stage
game where exit can occur for two different reasons: either because one
of the firms initially in the market is the ‘crazy’ type, and therefore always
preys; or, more interestingly, because one of the firms initially in the
market preys to signal to a rival that it may be crazy even though in fact
it is not. Either way, the rival is sufficiently uncertain about its opponent’s
type that it prefers to exit to avoid possible further losses.
Policy implications
Predation is welfare-reducing in the long run and is regarded as an anti-
competitive practice, so it is illegal in most competition laws. However,
it is hard both to define and to detect. Most price reductions are made
by firms that perceive competitive conditions in their market as having
become tougher, or firms that are more efficient than others, or firms that
legitimately try to increase their market share – and in all these cases,
price cuts are beneficial for consumers.
Having a simple rule with which to judge predatory pricing allegations
can therefore be useful. Areeda and Turner have suggested such a rule:
predation exists whenever a firm sets a price lower than its average
variable cost (a proxy for short-run marginal cost). The idea is that a firm
would never price below marginal cost, so any price below that must be
predatory.
70
But there are many problems with the Areeda-Turner rule. On the
one hand, there are circumstances where a firm may set P < AVC
without being involved in predation – for instance, in markets with
significant ‘learning by doing’, where selling a lot today significantly
reduces production costs in future periods. On the other hand, there are
circumstances when a firm that uses a predatory strategy does not even
need to set P < AVC – for instance, when the firm enjoys a cost or brand
image advantage over a rival. The next secion presents a case study of
alleged predation where P < AVC. So predatory pricing is generally judged
under a ‘rule of reason’.
A case study
Schmalensee (1979) provides an extensive economic analysis of a
particular case of alleged predation.8 Realemon had 90 per cent of the 8
See Church and Ware
US processed lemon juice market throughout the 1960s. A regional firm, (2000), Chapter 21,
and Pepall, Richards
Golden Crown, began to expand in the early 1970s and entered many
and Norman (2014),
regional US markets. Since Realemon enjoyed a premium brand position, Chapter 31, for brief
the price of Golden Crown was considerably lower than Realemon’s price. reviews of several more
By 1974, Golden Crown had 15 per cent of the national market while the recent cases.
share of Realemon had fallen to 75 per cent. Realemon then responded
with price cutting and increased advertising in geographical areas where
Golden Crown had entered, although it raised prices in areas not much
affected by Golden Crown.
At this point, Golden Crown complained to the antitrust authorities.
Realemon was found guilty of predatory pricing and price discrimination.
It was ordered to stop selling at ‘unreasonably low’ prices and to license
the Realemon trademark to all potential entrants for ten years.
To assess the Court decision and to draw general conclusions from this
case for antitrust policy, we need to consider three questions:
• Did Realemon have significant short-run market power; in other
words, was it able to price significantly higher than marginal cost?
• Did Realemon practise predatory pricing?
• Did Realemon have long-run market power; in other words, how
difficult was entry in this market?
The first question seems straightforward to answer: With a market share
of 75 per cent or higher, Realemon did have significant short-run market
power. So it could potentially abuse this power by predation or otherwise.
This brings us to the second question, which is trickier. Under the Areeda-
Turner rule, there was no predation since P > AVC. However, because of
its premium brand status, Realemon was able to price above its average
variable cost and still cause Golden Crown to price below its average cost
to effectively meet the Realemon prices. But pricing below average cost
would mean losses for Golden Crown, therefore exit from some markets or
at least an end to its expansion programme. Was this Realemon’s intent? If
so, then the Realemon tactics might well have amounted to predation even
though the Areeda-Turner rule was not satisfied.
How can predatory intent be established? One question here is how much
weight arguments about intent should carry, given that they are usually
based on what some regional managers may have said or written. And
then what if Realemon had intent but failed to deter the expansion of
Golden Crown? Should they be found guilty of ‘attempted’ predation?
These are not easy questions to answer, but on the whole the Court may
have been right in this case to conclude that Realemon had practised
predation with the intent of hindering the expansion of Golden Crown.
71
The final point to discuss when assessing the Court decision relates to
Realemon’s long-run market power. The main advantage of Realemon
in the lemon juice market seems to have been a first-mover advantage.
Consumers had greater experience with the Realemon brand and would
not switch to rival brands unless there was a considerable price difference.
This was reinforced by advertising. So entry was not easy. The Court’s
decision was an attempt to reduce or break the long-run market power
of Realemon: trademark licensing could provide entrants with the
opportunity to overcome the entry barrier created by the trademark. One
may ask whether the decision was perhaps too severe. After all, first-mover
advantages and brand loyalty exist in a large number of industries and
no one ever suggested an antitrust intervention. On the other hand, such
unusually tough measures may be justified in a case where market power
has been abused.
As for general conclusions, one thing which is clear is how difficult it is for
antitrust authorities to deal with predation, partly because firms that cut
their prices are likely to argue that they should not be penalised for being
efficient and that consumers benefit as a result. Furthermore, a rule of
reason approach seems best, while a simple rule such as the Areeda-Turner
rule does not always work. Under a rule of reason approach one would
need to examine (i) intent, (ii) actions, and (iii) effects to evaluate a case
of alleged predation.
Activities
1. Consider a duopoly where output decisions are made by the firms’ managers rather
than the owners. Could the owners of one firm design managerial incentives such
that the manager’s behaviour results in more profit for the firm than would be
obtained under standard profit maximising behaviour? Can you think of a way to
achieve this? Explain.
2. A domestic and a foreign firm produce a homogeneous product and compete in
quantities. The government decides to impose a quota on the quantity sold by the
foreign firm. Using reaction curves show that, if the quota results in only a small
reduction of the foreign firm’s volume of sales, it increases the profit of the domestic
firm and reduces the profit of the foreign firm.
3. Two price-setting firms are competing in a market for a homogeneous product. There
are 10,000 people in the population, each of whom is willing to pay at most 10 for
one unit of the good. Initially, both firms have a marginal cost of 5. Assume that the
firms are not capacity constrained and cannot collude. What is the equilibrium in this
market and what are the firms’ profits?
Suppose now that a new technology becomes available that lowers the marginal cost
to 3. The cost to a firm of purchasing this technology is 10,000. The firms must now
simultaneously decide whether to adopt the new technology or not, and following
this decision, they simultaneously set prices. Each firm can observe whether its
rival acquired the new technology or not before setting its price. What is (are) the
equilibrium (equilibria) in the market now?
4. Describe the ‘chain-store paradox’ as a game involving an incumbent monopolist and
N potential entrants. First derive the subgame-perfect equilibrium under complete
information. Why is this result a ‘paradox’? Then explain how incomplete information
about the incumbent’s ‘type’ can lead to entry deterrence as a subgame-perfect
equilibrium and thereby resolve the paradox. Discuss the economic intuition behind
your answer.
5. The most famous antitrust case of the last 30 years was probably Microsoft versus the
US competition authorities. This was a complex case. Microsoft was alleged to have
used its market power in the market for personal computer operating systems to
72
exclude rival firms from the market for web browsers by ‘bundling’ its operating
system together with its web browser. Moreover, the US competition authorities were
concerned that Microsoft’s actions might stifle potential competition in the market for
operating systems itself. Microsoft argued that its practice of bundling led to efficiency
gains and benefited consumers. Microsoft was later involved in an antitrust case
brought by the European Commission. In both instances, the firm was found in breach
of the competition laws.
The case is included in this chapter of the guide, despite the complex pricing tactics
and the fact that the background was an industry subject to rapid technological
change, because the core issue in the case is the alleged strategic behaviour of a firm
to exclude rivals from a market.
Who was right, Microsoft or the US competition authorities? What lessons can be
learned from this case? Make up your mind after reading different views of the
Microsoft case. Useful collections of articles have been published in the Spring 2001
issue of the Journal of Economic Perspectives and the March 2001 issue of the
Journal of Industry, Competition and Trade. Some of these articles, and many other
papers on the case, including some more recent ones, are available on the internet.

• explain the notions of strategic substitutes and strategic complements
• analyse simple models with sequential actions and describe first-mover
advantages in these models
• explain under what conditions a firm may be able and/or willing to
deter the entry of a potential rival
• use reaction functions to analyse strategic investment decisions by
firms in situations where the investment influences the firms’ future
profit functions
• analyse firms’ predatory tactics (i.e. behaviour that aims to induce exit
of competitors).

1. Consider an industry with two firms that produce a homogeneous
product and compete in quantities. Unit costs of the two firms are c1
and c2 and the industry inverse demand function is given by p = a –
(q1 + q2).
a. Find the Nash equilibrium of the single-stage game in which firms
simultaneously set quantities. Show how the equilibrium profit of
firm 1 depends on c1 and c2 and provide an intuition for this result
using reaction curves.
b. Now consider a two-stage game as follows. The second stage is as
described in part a above. In the first stage firm 1 can undertake
R&D to lower its second stage unit cost. In particular, spending R
yields c1 = c – R1/2, where c is a constant. Discuss the incentives
that firm 1 has to invest in R&D. What is the optimal choice of R
for firm 1? Provide some intuition using reaction curves.
2. Consider a market where there are two differentiated goods. The
demand for good 1 is given by q1 = a – bp1 + dp2 and that for good
2 is given by q2 = a – bp2 + dp1, where a > 0 and 0 < d < b. The
production cost of each good is zero.
73
a. Suppose that both goods are produced by the same firm (a

monopolist). Compute the prices set by the monopolist.
b. Suppose now that each good is produced by a different firm and
the firms choose prices simultaneously. Compute the Bertrand–
Nash equilibrium prices and confirm that they are lower than the
monopoly prices.
c. Show how an increase in the size of the market (i.e. an increase
in the parameter a affects the equilibrium prices under part b.
Illustrate using reaction curves.
d. Now assume that each good is produced by a different firm but
the firms set prices sequentially; in particular, firm 2 can observe
the price set by firm 1 before setting its own price. Compute the
subgame-perfect equilibrium price of firm 1 in this two-stage
game.
74
Chapter 7: Product differentiation and non-price competition
Chapter 7: Product differentiation and

non-price competition
Introduction
In many markets products are differentiated. Consumers care not only
about product price, but also about product characteristics, quality, the
location of the seller, pre-sale or post-sale services, and so on. Thus firms
compete in these dimensions as well as on price. An important feature of
differentiated goods markets is that the cross-price elasticity of demand is
not infinite at equal prices. That is, a firm can set its price above the price
of another firm and still have some demand for its product – even in the
absence of capacity constraints.
Industrial economists identify two types of product differentiation. Under
horizontal differentiation, different consumers prefer different
products when the prices of these products are the same – for instance,
orange cakes and chocolate cakes. Another example is location: if you live
in the centre of a town you might prefer to buy a product from a shop near
your home rather than buy the same product at the same price from a shop
in a suburb. So ‘spatial differentiation’ is a form of horizontal differentiation.
Under vertical differentiation, all consumers would buy the same
product if all product prices were the same. Of course, since at equilibrium
products have different prices and consumers have different incomes, they
do not all buy the same product: the point is that consumers agree over
the preference ordering. A typical example is the case where products are
differentiated by quality – for instance, a Volkswagen versus a Mercedes.
The distinction between horizontal and vertical differentiation is
analytical. In many markets, products are differentiated along both
dimensions. In this chapter we discuss both types of differentiation and
their implications for competition and market structure.
For the most part we will assume that consumers have perfect information
about product quality or other product characteristics. However, when
products are differentiated, buyers often have less product information
than sellers. In the last part of the chapter we discuss the implications of
asymmetric information for market outcomes. As we will see, asymmetric
information leads to inefficiency, sometimes even the collapse of the
market; and while several solutions to this problem exist, they are often
imperfect.
Learning outcomes
• analyse how product differentiation relaxes price competition
• analyse how firms choose product characteristics in the context of
simple models of product differentiation
• describe the theory and evidence on product proliferation as an entry
deterrence strategy
• compare horizontal and vertical product differentiation and explain
their different implications for market structure and profitability
• explain the notions of moral hazard and adverse selection and describe
how asymmetric information affects the working of product markets. 75
Essential reading
Books
11 and 6.
Tirole, J. The theory of industrial organization. Chapter 7 and part of Chapter 2.
Journal
Schmalensee, R. ‘Entry deterrence in the ready-to-eat breakfast cereal industry’,
Bell Journal of Economics 9(2) 1978, pp.305–27.
Further reading
Books
2000) Chapter 12.
Pearson Addison Wesley, 2005) Chapters 7 and 17.
Chapters 7 and 19.
Shy, O. Industrial organization. (Cambridge, MA: MIT Press, 1995) Chapters 7
and 12.
Journals
Berry S. and J. Waldfogel ‘Product quality and market size’, Journal of
Industrial Economics 58(1) 2010, pp.1–31.
George, L. and J. Waldfogel ‘Who affects whom in daily newspaper markets’,
Journal of Political Economy 111(4) 2003, pp.765–84.
Genesove, D. ‘Adverse selection in the wholesale used car market’, Journal of
Political Economy (1993), pp.644–65.
Judd, K.L. ‘Credible spatial preemption’, Rand Journal of Economics 16(2) 1985,
pp.153–66.
Salop, S.C. ‘Monopolistic competition with outside goods’, Bell Journal of
Economics 10(1) 1979, pp.141–56.
Shaked, A. and J. Sutton ‘Relaxing price competition through product
differentiation’, Review of Economic Studies 49(1) 1982, pp.3–13.
Waldfogel, J. ‘Preference externalities: An empirical study of who benefits
whom in differentiated product markets’, Rand Journal of Economics, 34(3)
2003, pp.557–68.
Horizontal product differentiation

We first focus on horizontal product differentiation. We will examine:
• how horizontal product differentiation ‘relaxes’ price competition, that
is to say, it leads to equilibrium prices higher than marginal costs even
in the absence of capacity constraints or repeated interaction
• how firms choose the characteristics of their products
• entry and the determinants of market structure in horizontally
differentiated markets.
76
The linear location model

Consider the following stylised model, a variant of which was first
analysed by Hotelling. Consumers are uniformly distributed along a ‘linear
city’ of length L, and at each point on the line lies a single consumer. All
consumers are similar and each of them consumes exactly one unit of the
good – their reservation price is sufficiently high so that they always
choose to buy, but they never buy more than one unit. There are two firms
or stores, which sell the same physical good. Suppose, to begin with, that
the firms’ locations are given. Firm 1 is located at distance a from the west
end of the city and firm 2 is located at distance b from the east end of the
city (see Figure 7.1).1 The unit cost of the good is c, same for both firms. 1
We assume, without
Consumers incur a quadratic transportation cost, that is a consumer loss of generality, that
situated at distance x away from the location of a firm incurs a cost tx2 to firm 1 is never to the
right of firm 2
go to that firm and return back. The firms compete by setting prices
(i.e. a + b ≤ L).
simultaneously. We now compute the Nash equilibrium of this pricing
game.
p1 p2
West end firm 1 firm 2 East end
marginal consumer
a x b
Figure 7.1
Suppose prices are p1 and p2. Obviously, consumers situated close to firm
1 will prefer to buy from this firm, and those situated close to firm 2 will
prefer to buy from firm 2. To be more precise, we define the ‘marginal
consumer’, that is the consumer who is just indifferent between buying
from firm 1 or firm 2. He or she is located at distance x to the right of firm
1, where x is defined by:
p + tx 2 = p + t ( L − a −b − x ) 2
1 2
The expression on the left-hand side is the generalised cost (the sum of the
product price and the transportation cost) of the marginal consumer if he
buys from firm 1. The expression on the right-hand side is the generalised
cost of the marginal consumer if he buys from firm 2. These generalised
costs are depicted in Figure 7.1 by the two ‘umbrellas’.
Solving this equation for x we obtain:
L−a −b p1 − p 2 .
x= −
2 2t ( L − a − b)
All consumers to the left of the marginal consumer buy from firm 1, while
all consumers to the right of the marginal consumer buy from firm 2.
Hence demand for firm 1 is:
L−a −b p1 − p 2
D1 ( p1 , p 2 ) = a + x = a + −
2 2t ( L − a − b)
and demand for firm 2 is:
77
L−a−b p1 − p 2 .
D 2 ( p1 , p 2 ) = L − a − x = L − a − +
2 2t ( L − a − b)
Note that the negative effect of pi on Di and the positive effect of pj on

Di are less strong when the firms have more local market power, that is,
the higher the value of t and the lower the value of a + b (the greater
the distance between the firms). Profits are given by ∏1(p1,p2) = (p1
– c)D1(p1,p2) and ∏2(p1,p2) = (p2 – c)D2(p1,p2) respectively. The Nash
equilibrium is the solution to the system of first-order conditions ∂∏1/∂p1
= 0 and ∂∏2/∂p2 = 0. Solving the system we obtain (check):
a −b b−a .
p 1 = c + t ( L − a − b)( L + ), p 2 = c + t ( L − a − b)( L + )
3 3
Hence equilibrium prices are higher than marginal cost c – unless a + b =

L, that is to say, unless firms are at the same location, in which case there
is no differentiation. We conclude that product differentiation by location
allows firms to make positive profits even in a one-shot simultaneous-
move pricing game. Note that prices increase in t , and are equal to
marginal cost for t = 0: higher t makes demand less sensitive to price,
increases the local market power of firms and softens the competition for
customers. To find the equilibrium profits, substitute p and p into the
1 2
profit functions ∏1 and ∏2. It turns out (check):
1 1 2 1
∏ = t ( L − a − b)(3L + a − b) 2 , ∏ = t ( L − a − b)(3L − a + b) 2 ,
18 18
which are positive for a + b ≠ L.

Consider now a two-stage game as follows. At stage 1 the firms
simultaneously choose their locations: firm 1 chooses a and firm 2 chooses
b. Then at stage 2, given their locations, the firms simultaneously set
prices. The timing of the game captures the important distinction between
long-run and short-run decisions, in particular the fact that prices are
easier to change than locations. We are looking for the subgame-perfect
equilibrium of this game.
We use backward induction. At stage 2, a and b are given, so our previous
1
analysis applies. That is, the second stage equilibrium profits ∏ ( a, b)
2
and ∏ ( a, b) are given by the expressions above. At stage 1, firms 1 and
2 choose a and b, respectively, anticipating how the stage 2 subgame will
be played. In other words, they anticipate that their choice of location
will affect their choice of price in1 the next stage and, in particular, they
2
anticipate the expressions for ∏ and ∏ .
1
At stage 1, then, firm 1 chooses a to maximise ∏ and firm 2 chooses b to
2
maximise ∏ . After some manipulation, we obtain:
1
d∏ 1
= − t (3L + a − b)( L + 3a + b),
da 18
which is always negative. So to maximise its profit in the two-stage game
firm 1 will choose the lowest possible value for a, namely a = 0: it will
locate at the west end of the city. Similarly for firm 2: it will locate at
the east end of the city. Hence we obtain ‘maximal differentiation’ in this
particular model. Note that the result would be the same if the location
decisions were taken sequentially in a three-stage game with choice of
location by firm 1, then by firm 2 (which observes the choice of location
by firm 1), then price competition.
78
The intuition for this result is as follows. There are two effects, working in
opposite directions:
• the demand effect: a firm wants to move towards its rival because in
this way it increases its market share, and therefore its profit (for given
prices)
• the competition effect: the firms want to move away from each other
because in this way products become more differentiated, price
competition is ‘relaxed’, and profits increase.
In our model the competition effect dominates, so we get maximal
differentiation.
Note that, although the same two effects operate in more general settings,
the competition effect need not dominate. For instance, if in our example
we assume a transportation cost tx γ, where γ∈[1,2], then:
• for γ < 1.26, there is no equilibrium in pure strategies
• for 1.26 < γ < 1.67, firms differentiate but not maximally
• for γ > 1.67, firms differentiate maximally.
Activity
Suppose the price is exogenously fixed (by the government or by a manufacturer if the
two firms in the model are retailers) and the two firms only compete by simultaneously
choosing locations. Show that the Nash equilibrium of this game involves ‘minimal
differentiation’: both firms locate at the centre. Interpret the result.
Answer
Start by distinguishing cases:
• the firms are not located at the same point
• the firms are located at the same point but not the centre, and
• both firms are located at the centre.
Then show that in all but the third case there exists a profitable deviation by at least one
firm. The main thing to remember is that consumers buy from the nearest firm, so each
firm seeks to maximise its market share given the location of its rival. The intuition for the
result is that in this case there is no competition effect, since the choice of location does
not affect the prices. There is only a demand effect, which works against differentiation.
More generally, the choice of product specification – and therefore the

degree of differentiation at equilibrium – will be determined by a trade-off
between forces that promote and forces that oppose differentiation. The
force promoting differentiation is the fact that it relaxes price competition.
Note, however, that this works only when the second stage competition is
in prices (i.e. the stage 2 actions are strategic complements). The forces
opposing differentiation are two. The first is the demand effect. This effect
is even clearer in markets where demand is concentrated around a few
poles (i.e. where consumers are not uniformly distributed). The second is
the presence of positive externalities that induce firms to locate close to
one another. For instance, when consumers do not know the characteristics
or prices of products sold by all firms in a market, lowering consumer
search costs by locating close to one another may lead to higher overall
demand for the product.
One final remark: The model presented here can be easily interpreted
as a product differentiation rather than a location model. In particular,
a firm’s ‘location’ can be interpreted as a choice of product specification.
79
A consumer’s ‘location’ would represent her most preferred product

specification (variety). And finally, the ‘transportation cost’ would
represent the utility loss from not consuming the preferred variety.
Consumers generally benefit from product diversity. However, when
consumers are not uniformly distributed in the product characteristics
space and fixed costs are significant, some consumers may not get the
variety they want. In particular, the market will provide products which
are preferred by numerous consumers but not necessarily those which
are preferred by small groups of consumers with less prevalent tastes. Or
perhaps small groups will end up having a very restricted choice. Because
of these ‘preference externalities’, an increase in the size of the larger
group may even hurt the smaller group, since firms may then be even less
likely to target the smaller group (George and Waldfogel, 2003; Waldfogel,
2003).
The circular location model

The linear location model is very useful in analysing the location decision.
In this section we want to study the entry decision in differentiated
product markets. To this end we introduce a slightly different setup –
which is also a more appropriate representation for products where there
are no clear ‘end points’ in the market. Consumers are now uniformly
distributed with density S (a measure of market size) on a circle with
perimeter 1, and each consumer buys exactly one unit of the good. This
implies that for any distance x along the circle, total demand is equal to
Sx. Transport costs are linear: a consumer situated at distance x away from
a firm incurs a cost tx to go to that firm and return. (The qualitative results
also hold for quadratic transport cost.) Firms locate around the circle, each
in one location, and produce the same physical good at the same unit cost
c. There is a large number of potential entrants. Entry is free, in the sense
that it is not restricted or regulated, but there is a sunk cost of entry f.
Firm i
1/ 1/
Firm i–1 N N Firm i+1
Figure 7.2
We look for a subgame-perfect equilibrium of the following two-stage
game. At stage 1, firms decide whether or not to enter at sunk cost f. They
do not choose location, but are automatically located at equal distance
(1/N) from one another on the circle. At stage 2, those firms that have
entered simultaneously set prices.
We proceed through backward induction. Assume that N firms have
entered and we are at stage 2. Since firms are symmetric, we look for a
symmetric equilibrium where all firms set price p. Now in practice, firm i
competes directly only with the two neighbouring firms, so we can ignore
the remaining product space (see Figure 7.2).
Let us focus on the decision problem of firm i, taking as given the prices
of the other firms. As a matter of fact, let us assume that all the other
firms set price p. This will be true in the symmetric equilibrium, so we can
anticipate it. Suppose then that firm i sets price pi, while each of the two
neighbouring firms sets price p. A consumer located at distance x away
from firm i is indifferent between buying from firm i or the neighbouring
firm if:
80
t
p+ − pi
1  N .
p i + tx = p + t  − x  ⇒ x =
 N  2t
Demand for firm i’s product comes from consumers to its left and to its
right, so the total distance covered is 2x. Hence the demand faced by
firm i is given by Di ( pi, p ) = 2 Sx . The firm chooses pi to maximise
∏i= ( p i – c)D i ( pi , p )= ( pi – c) S ( p + t/N – p i )/t , taking p as given. The first-
order condition is:
S t
d ∏ i / dpi = 0 ⇔ ( p + − 2 pi + c) = 0 .
t N
This defines the optimal choice of pi given p for any firm i. We can now
use the fact that at the symmetric equilibrium we will have pi = p, so the
above equation gives p=c+t/N. This is the second stage equilibrium price,
which is of course a function of N. To find the equilibrium gross profit of
each firm as a function of N, substitute p into the profit function ∏i, and
also set pi = p. We obtain ∏ = S( p – c)/N = St/N2.
At stage 1 there is free entry. This means that firms will be entering the
industry as long as they can make a non-negative net profit. Note that ∏
decreases as N rises (i.e. as more firms enter). Initially ∏ is higher than f
(assuming f is not too large), so firms keep entering, thereby causing ∏
to fall. There will be some point where ∏ = f, and at that point entry will
stop because any further entry is unprofitable. In particular, one additional
entrant would cause ∏ to fall below f and the firms will anticipate that
they would not be making enough gross profit at stage 2 to cover the entry
cost at stage 1. From the free-entry zero-profit condition ∏ = f, we get
St/N2 = f ⇔ N* = (St/f)1/2. This is the equilibrium number of firms in the
two-stage game. Substituting into the expression for p, we obtain p* = c +
(tf/S)1/2.
To summarise the two main results of the model:
• the equilibrium price is higher than marginal cost
• the equilibrium number of firms increases in market size S and unit
transport cost t and falls with the entry cost f.
Brand proliferation as an entry deterrence strategy

Can incumbent firms in an industry ‘crowd’ the product space by
producing many different brands as a way to deter entry and maintain
profits? What is required for this strategy to work? The theoretical
discussion on brand proliferation was triggered by an antitrust case. In the
1970s the four biggest US producers of ready-to-eat breakfast cereals were
alleged to have successfully deterred entry by other firms in that market
through brand proliferation.
Some facts about this particular industry: concentration was very high
throughout the 1950s and 1960s, market growth was rapid, profitability
was high, advertising was heavy, there were no significant scale
economies, the six leading producers introduced 80 new brands between
1950 and 1972, and there was no significant entry by new firms. How can
the lack of entry be explained?
The brand proliferation model relies on three assumptions.
1. There are increasing returns to scale at the brand level. For example,
the cost of producing a brand may be given by f + cq, where q is
quantity produced. If there were no fixed cost f, then it would be
profitable to produce a brand for each point in the product space and
81
satisfy everyone’s exact preferences. A firm could enter and make

positive profit by selling to a single consumer, so there would be no
scope for entry deterrence.
2. There is localised competition between brands. If each brand were
to compete for customers with all other brands and not just with
neighbouring brands, then there would be no scope for deterring entry
by brand proliferation.
3. There are substantial costs of repositioning a brand in the product
space or of withdrawing a brand. If there were no such costs, the
incumbent firms would not be committed to their chosen locations –
so entry would occur and the incumbents would relocate or withdraw
brands until every firm made zero net profit.
The circular model of horizontal differentiation presented above satisfies
all three assumptions. Consider the following setup. There is a group of
firms that produce N brands along the circle. The equilibrium price of each
brand is p(N), with dp/dN < 0. Gross profit per brand is given by ∏[p(N),
N], with ∂∏/∂p > 0 and ∂∏/∂N < 0. Hence the overall effect of N on ∏ is
given by the total derivative d∏/dN = ∂∏/∂N + (∂∏/∂p)(dp/dN) < 0.
The critical number of brands N that yields zero net profit per brand
is defined by ∏[p(N), N] = f, where f is the fixed and sunk cost of
introducing a brand. For any N < N, ∏[p(N), N] = f and therefore the
incumbents make positive profits.
What profit will an entrant make by establishing a new product? An
entrant will locate midway between two existing brands on the circle.
Because competition is localised and repositioning or withdrawal of a
brand is prohibitively costly, the entrant’s market share will be roughly
1/2N and its gross profit will be approximately ∏[p(2N),2N], that is to say
the profit that corresponds to a situation with 2N symmetric brands on the
circle.2 Hence entry will not occur as long as ∏[p(2N),2N] < f. A sufficient 2
This is strictly true if the
condition for this to hold is 2N > N ⇔ N > N /2. two neighbouring firms
can charge different
Hence incumbents can make positive net profits and deter entry through prices to consumers on
brand proliferation: they simply need to ensure that N/2<N<N. their left than to those
Intuitively, this says that the number of brands must be sufficiently on their right – so that
low so that they can all enjoy positive net profits, and at the same time the entry of a single
firm between two
sufficiently high so that a potential entrant cannot make positive net profit
existing ones has no
by introducing a brand. repercussions elsewhere
Schmalensee has used an extended version of this model to explain the on the circle. If this
is not the case, then
lack of entry into the US ready-to-eat breakfast cereal industry. In
the entrant’s price will
particular, he has also included advertising and argued that high levels of be somewhat higher
advertising were an additional entry-deterring mechanism. Schmalensee than p(2N) and the
has also discussed additional complications stemming from the fact that entrant’s gross profit
the market was growing and that entry did occur in the 1970s when a new somewhat higher than
∏[p(2N),2N] – but
market segment (‘natural cereals’) emerged. He has pointed out that the
still much lower than
minimum efficient firm size was three times the minimum efficient brand ∏[p(N),N], so the
size in this industry, and that this was an additional factor that helped qualitative results will
prevent entry during the 1950s and 1960s, despite the high demand apply.
growth: any entrant would have to spot three different profitable niches at
the same time, while any incumbent could profitably introduce just one
additional brand. In the 1970s, on the other hand, a substantial new
market segment emerged within a few years, attracting new entry. All in
all, it seems that the lack of entry into the US ready-to-eat breakfast cereal
industry can be explained to a large extent by the joint effect of two
different strategies of incumbent firms: a brand proliferation strategy and
an advertising escalation strategy. The latter refers to a situation where, as
82
market size increases, firms spend more on advertising (which is a fixed

cost) and this prevents concentration from falling.3
3
See Chapter 10 of this
guide for more on the
Although the model of brand proliferation as an entry deterrence strategy advertising escalation
has been much discussed because of the welfare issues involved, you mechanism.
should bear in mind that product proliferation can also simply be a search
for successful products. In fact, some of the assumptions of the model
may not hold in general across industries. Recall that repositioning costs
and exit costs must be substantial for the brand proliferation strategy to
work. It probably makes sense to assume high repositioning costs, but why
would it be costly to withdraw a brand in the ready-to-eat cereal industry
or many other industries? This is a valid objection. One possible answer
might be that the cost is in terms of reputation: if an incumbent withdraws
one brand in the face of entry, then he will face more entry, while if he
doesn’t, he may convince other potential entrants that he is not likely to
be withdrawing brands and thereby discourage them from attempting to
enter. For a proper analysis of this point, however, we need to assume that
potential entrants have incomplete information about the incumbent’s
‘type’.
Activity
Suppose that in the model analysed above exit costs are zero. Show that in this case
brand proliferation does not deter entry.
Answer
Suppose a new firm has entered at exactly the location of an existing brand, say brand.
Then price competition drives the price of that brand down to marginal cost and the gross
profit of the entrant as well as of the incumbent producing brand i down to zero. Also, the
increased competition will lead to price reductions for the neighbouring brands, i – 1 and
i +1, and this will in turn affect brands i – 2 and i + 2, and so on. So if the incumbent firm
that produces brand i also produces some other brands, it will see its profits from those
other brands fall as a result of entry at location i.
In this situation, the incumbent firm will have an incentive to withdraw brand i since
withdrawal (exit) is costless and will raise its overall profit by relaxing competition overall
(the firm is making zero profit from brand i anyway). On the other hand, the entrant, once
in the market, has nothing to gain from withdrawing its brand. All this will be anticipated
by the new firm before it enters, so entry will occur. We conclude that brand proliferation
cannot deter entry in the absence of exit costs.
Vertical product differentiation

Under vertical differentiation, products differ in some characteristic which
we will denote by u and generally call ‘quality’ in what follows. This
can be related to a physical characteristic of the product or to its brand
image. Essentially, as we will see below, u is a measure of the consumers’
willingness to pay a higher price for a product. Furthermore, all consumers
have the same ranking of the different varieties of the product: if ui < uj
and pi = pj, then all consumers would buy product j.
Underlying most of the literature on vertical product differentiation is
a two-stage game as follows. At stage 1, each firm i chooses ui. In many
models there is a fixed and sunk cost F(ui), for instance advertising or
R&D, which is incurred at this stage and increases the value of
u: dF/dui > 0. At stage 2, firms compete in prices or in quantities. The
choice of ui affects stage 2 competition through its effect on the firms’
demand functions: given the prices and the u’s of all the rival firms, an
83
increase in ui shifts the demand curve of firm i outwards (i.e. it increases

the consumers’ willingness to pay for product i). In some models, the
choice of ui also affects the marginal cost of production c(ui): dc/dui > 0.
We now briefly sketch a simple model of vertical product differentiation,
focusing on the structure of the model and the basic results.4 Firms offer 4
A detailed discussion
products which vary in quality u. Consumers are uniformly distributed of the model can be
found in Shaked and
with density S (a measure of market size) according to their income t,
Sutton (1982).
which lies between a and b. Each consumer buys one unit of a single
product or none. A consumer has utility U(t,0) = u0t if she doesn’t buy and
U(t,k) = uk(t – pk) if she buys good k at price pk, where t is income prior to
any purchase. The marginal cost of production is independent of quality
– assume, for simplicity, that it is zero. Finally, the game has three stages:
at stage 1 firms decide whether or not to enter the industry at a small sunk
cost ε, at stage 2 each firm chooses a level of u, and at stage 3 firms set
prices.
The main results of this model are the following:
• the ‘finiteness property’: there exists a lower bound to the number of
firms that can survive (i.e. make non-negative profit) at equilibrium,
which is independent of market size S and the entry cost ε.
• the equilibrium number of firms is increasing in the degree of income
heterogeneity. For instance, if b/a ≤ 2, only one firm can survive; if
2 < b/a ≤ 4, there is room in the market for exactly two firms; and so
on
• suppose that 2 < b/a ≤ 4, so exactly two firms enter. Then the firms
choose different qualities, u1 and u2 > u1, set prices p1 and p2 > p1 and
they both make positive sales and positive net profits.
You should compare these results with those obtained under horizontal
product differentiation. In the circular location model, for example, we
have found that, as market size S → ∞ or the entry cost ε → 0, we get
N* → ∞. Also, free (and simultaneous) entry at a positive sunk entry
cost implies zero net profit at equilibrium. In our vertical differentiation
model, in contrast, as S → ∞ or ε → 0, N* is finite and does not change.
Also, firms make positive net profits in the long run even under free and
simultaneous entry.
How robust are the results of our vertical differentiation model? It turns
out that one crucial feature of the model is the assumption that marginal
cost is independent of quality. The finiteness property in particular holds if
marginal cost does not increase too fast with quality, but it breaks down if
c increases fast with u. In practice, then, the finiteness property is likely to
hold when most of the cost of quality is a fixed cost, such as advertising or
R&D, rather than a variable cost.
Actually the finiteness property still holds under general conditions in
markets with both horizontal and vertical differentiation, provided that
c(u) is relatively flat so that c does not increase too fast with u. The
intuition is that if c(u) is relatively flat, then price differences across
qualities will be relatively small. So a firm that produces a quality slightly
lower than another firm will not generate sufficient demand to make entry
worthwhile in the presence of fixed costs associated with entry and quality
choice – no matter how large the market size. If, on the other hand, c(u) is
steep, price differences across qualities will be more substantial. So a firm
that enters, say, with a quality between two existing qualities, and with a
price between the prices of these two qualities, will have a better chance
of capturing a significant share of the market.
84
Berry and Walfogel (2010) provide some empirical evidence which is

consistent with these theoretical predictions. In the restaurant industry,
where quality is produced largely with variable costs (i.e. c(u) is relatively
steep), the range of qualities on offer increases with market size. In
daily newspapers, where quality is produced with fixed costs (i.e. c(u) is
relatively flat), an increase in market size raises the average quality but
has little effect on the range of qualities: the market offers little additional
variety as it grows larger.
Markets with asymmetric information

Agents in a market may have private information regarding their own
characteristics (e.g. their willingness to pay, their ability or skills, their
costs, the quality of a product they sell, the expected returns from an
investment…) or actions (for instance, their effort on a job, their pricing
behaviour, the quality of a product they produce…). The presence of
hidden characteristics often gives rise to ‘adverse selection’, while
hidden actions may result in ‘moral hazard’. In both cases, the presence
of asymmetric information leads to inefficiency and possibly even the
collapse of the market.
We will focus on the effects of asymmetric information in markets for
experience goods (i.e. products whose quality is not observable by
consumers before a purchase and is learned only after the product is
bought).
There are a number of possible solutions to the problems caused by
asymmetric information, but they are often imperfect. In the case of
experience goods, the seller may offer a warranty to the buyer to signal
that the good is of high quality. However, if the performance of a good
(e.g. a used car) depends also on the way the buyer consumes it as well as
on its intrinsic quality, or if the quality of a good is difficult to measure in
an objective way, the seller may give a limited rather than a full warranty,
so the inefficiency will be reduced but not eliminated. Other solutions
involve repeat purchases by consumers or building a reputation for quality
by a firm. Finally, government regulation or intervention is often desirable:
minimum quality standards, laws against misleading advertising, laws on
disclosure of information, safety regulations, certification (as for doctors or
accountants), and so on.
Moral hazard
A simple model will highlight how moral hazard leads to inefficiency.
A monopolist sells a product to consumers who are all identical, each
with utility U = θs – p if they buy one unit of the product at price p and
the product has quality s, and U = 0 if they do not buy. The monopolist
chooses p and s to maximise profit. The level of quality can be either sL
(low) or sH (high), where sL < sH. The unit production cost is cH for high
quality and cL < cH for low quality. Assume that θsH – cH > 0 and θsH – cH
> θsL – cL, so that producing the high quality is socially efficient. The
consumers do not observe the quality of the product before purchasing.
Let us also, without loss of generality, normalise the number of consumers
to one.
Clearly, an equilibrium where the monopolist produces high quality cannot
exist, since the monopolist would save cH – cL by cutting quality, and this
would not reduce demand. If θsL ≥ cL, an equilibrium exists where the
monopolist produces low quality at price p = θsL, appropriates the entire
consumer surplus and makes profit Π = θsL – cL ≥ 0. If θsL < cL, there is no
85
equilibrium and the market disappears, since there is no price such that
Π ≥ 0 and U ≥ 0. Both these outcomes are inefficient.
In the absence of warranties, repeat purchases, reputation effects or
government intervention, is there any other mechanism that could restore
efficiency in this setup? Suppose that a fraction λ of consumers, where
0 < λ < 1, can observe the quality of the product before purchasing. In
this case, an equilibrium where the monopolist produces the high quality
may exist, provided that λ is sufficiently high. Intuitively, the informed
consumers will be prepared to buy the high quality provided it is sold at
price p ≤ θsH. Also, the monopolist will be prepared to produce the high
quality if his profit from doing so is greater than the profit he makes by
producing the low quality – and this will be the case if p is sufficiently
high. We can conclude that the monopolist’s high price serves as a signal
of high quality to uninformed consumers, a result which seems consistent
with empirical evidence from markets for experience goods.
Adverse selection
We now look at a model of adverse selection. An object of quality s is
offered for sale by its owner (agent 1) to a potential buyer (agent 2). The
quality is known by the owner, but the potential buyer only knows that s
is uniformly distributed on the interval [0, smax]. The seller has utility θ1s
if he keeps the object and p if he sells it at price p. The buyer has utility
θ2s – p if he buys at price p and zero otherwise. Furthermore, the buyer
is risk neutral, so his expected utility is given by θ2sα – p, where sα is the
expected quality of the object conditional on its being offered for sale.
Finally, assume that θ1 and θ2 are known to both parties and θ1 < θ2, so
that trading is efficient irrespective of s: for any price p such that θ1s < p <
θ2s, both parties gain from trade.
Suppose there is a price p at which the two parties agree to trade. The
buyer should then infer that if the seller is willing to sell at price p, it must
be the case that p ≥ θ1s ⇔ s ≤ p/θ1 ⇔ s∈[0, p/θ1]. Since s is uniformly
distributed, it follows that the expected quality of the object for the buyer,
given that it is offered for sale, is sα = p/2θ1. In other words, the expected
quality is biased downwards by the decision of the owner to offer the
object for sale.
The potential buyer will buy if and only if θ2sα – p ≥ 0 ⇔ θ2 ≥ 2θ1, that
is, if the tastes of the owner and the buyer differ a lot. If they don’t differ
too much (i.e. if θ2 < 2θ1,) trade does not occur: whatever s, there is no
price at which the owner is willing to sell and the potential buyer willing
to buy. Note that in more general settings the market may not break down
completely, it may only shrink because of adverse selection.
One of the best known applications of adverse selection is the following
model of the used car market. There are two qualities of used cars:
‘lemons’ are used cars of bad quality, while ‘plums’ are used cars of good
quality. Owners of used cars know the quality of their cars, but potential
buyers do not – although they do know that a proportion π of the cars in
the market are of good quality. There are many owners and many potential
buyers. Since quality is not observed by buyers, all used cars must be sold
at the same price, irrespective of their quality.
Now a buyer is willing to pay an average price for what is, from his point
of view, a car of average expected quality. But the owner of a good quality
car might not be willing to sell at this average price – in which case, only
cars of low quality remain for sale: low quality drives high quality out of
the market.
86
To see this, suppose that the owner of a lemon wants to sell if the price in
the market of used cars is p > vS(L), where vS(L) is the value of a lemon to
the owner. The owner of a plum is willing to sell if p > vS(P), where vS(P)
is the value of a plum to the owner. A potential buyer is willing to buy if
the expected utility from doing so is at least as large as p, in other words,
if p ≤ πvB(P) + (1 – π)vB(L), where vB(L) and vB(P) are, respectively, the
value of a lemon and of a plum to the buyer. The parameters vS(P), vB(P),
vS(L) and vB(L) are common knowledge. We can assume vS(P) > vS(L), so
owners of lemons will always want to sell at any given price p if owners of
plums are willing to sell at that price. For owners of plums to want to sell
and buyers to be willing to buy we need:
vS(P) < p ≤ πvB(P) + (1 – π)vB(L).
Is this condition always satisfied? No. For example, let vB(P) = 2000,
vS(P) = 1500, vB(L) = 1000, vS(L) = 500. Note that vS(P) < vB(P) and
vS(L) < vB(L), so trade in both plums and lemons is socially efficient. The
above condition becomes 1500 < p ≤ 2000π + 1000(1 – π), which is not
satisfied if π is small. In this case, trade in plums will not take place even
though it is efficient. Only lemons will be offered for sale, buyers will
realise this and trade in lemons will take place at a price between vB(L)
and vS(L).
Remark: The cause of the inefficiency in these models is the asymmetry of
information, not the lack of it. In other words, if all parties were equally
uninformed, there would be no loss in efficiency. For instance, if owners
of cars were equally uninformed about their quality as potential buyers,
they would be willing to sell if πvS(P) + (1 – π)vS(L) ≤ p. Potential buyers
would be willing to buy if p ≤ πvB(P) + (1 – π)vB(L). Assume vS(P) < vB(P)
and vS(L) < vB(L), so that trade in both plums and lemons is efficient. Then
it follows that πvS(P) + (1 – π) vS(L) < πvB(P) + (1 – π)vB(L), irrespective
of the value of π, so a price p could be found such that trade occurs.
Empirical evidence
Genesove (1993) presents an empirical test for adverse selection in the
wholesale used car market. To better understand his approach, consider
the following example. Two individuals are selling apples at the same
exogenously given price. Seller A hates apples and has an orchard filled
with apple trees. Seller B loves apples and has an orchard with few apple
trees. All apples look identical, but in fact some are of good quality and
some are of bad quality. From which seller will you buy? You may think
that A will be selling all his apples, both good and bad, while B will be
selling only bad apples (and eating the rest). So the average quality will
be higher from seller A, and you will therefore buy from him. Now at
equilibrium, the price is endogenous and will reflect average quality.
Therefore seller A – the seller with the greater propensity to sell – will be
selling his apples at a higher price than seller B.
There are two types of sellers in the wholesale used car market: new car
dealers (NCDs), who sell used cars in addition to their main business; and
used car dealers (UCDs), who sell used cars only. Car dealers use auctions to
adjust their portfolio of used cars. The sellers in these auctions have private
information about vehicle quality, which the buyers (who are NCDs and
UCDs themselves) do not have. A survey of car dealers shows that NCDs sell
a larger fraction of their older used cars than UCDs at auctions – rather than
keeping them for retail sale to final consumers. The reason is that NCDs,
unlike UCDs, want a limited stock of used cars, especially older ones, since
their main business is new cars. Since NCDs have a greater propensity than
UCDs to sell older cars at auctions, NCDs are the equivalent of seller A and
87
UCDs are the equivalent of seller B in the motivational example above. So

if adverse selection is prevalent in the wholesale used car market, prices
(winning bids) for used cars at auctions should be higher for older cars
sold by NCDs than for older cars sold by UCDs; on the other hand, the NCD
premium should disappear for newer cars.
Is this the case? Using data on 893 used cars made between 1984 and
1988 and sold in a 1989 auction, Genesove performs a regression of the
winning bid on various vehicle characteristics as well as a dummy variable
for NCDs interacted with model year – so the NCD premium is allowed to
vary with the age of the car (UCDs is the reference category). He finds a
statistically significant 14–17 per cent NCD premium for 1984 cars, but no
statistically significant NCD premium for 1985 or later models. Thus the
results provide weak evidence for adverse selection.
Activities
1. Consider the linear location model and assume that the location decisions are taken
sequentially. This is now a three-stage game: choice of location by firm 1, then by firm
2 (which observes the choice of location by firm 1), then price competition. Compute
the subgame-perfect equilibrium and provide intuition for the result.
2. Consider the circular location model and assume that the transportation cost is
quadratic rather than linear. Compute the equilibrium price, profit and number of
firms under free entry and interpret the result.
3. Consider the linear location model where the price is exogenously fixed and the firms
compete by simultaneously choosing locations. However, assume there are four firms
in the market rather than just two. Derive the Nash equilibrium of this game and
interpret the result.
4. What are the most important elements of product differentiation in practice? How
important is geography as opposed to brand characteristics? Clearly, the anwer to
these questions depends to a large extent on the particular industry one examines.
For instance, geography is not important for competition among online retailers. For
two or three industries of your choice, make a list of the main sources of product
differentiation and assess their importance.
5. Consumers believe that a certain product they are buying is of high quality with
probability p and low quality with probability 1 – p, where 0 < p < 1. A high-quality
good is valued by consumers at vH and costs cH to produce, while a low-quality good
is valued at vL ≤ vH and costs cL ≤ cH to produce.
i. If high-quality firms do nothing to signal their quality, what price would
consumers be willing to pay for the product?
ii. Let cH = 0.9vH and vH = 2vL. For what values of p would consumer beliefs about
quality be consistent? That is, for what values of p is there actually a positive
probability that some firms are producing high-quality goods?
6. Advertising could in principle help a firm establish a reputation for quality (see, for
instance, Nelson, 1974). The idea is that a high-quality producer may spend a lot
on advertising a product so that the investment can only be recouped from future,
not current, business, when the true quality of the product is known. Low-quality
producers cannot imitate this strategy, and therefore advertising will play a role as a
signal of quality.
This argument sounds convincing, but there is not much empirical evidence to support
it. The correlation between quality and advertising of products seems very weak
overall, although in some product categories it is positive and in others negative
(Caves and Greene, 1996). How can this be explained? Is the theory problematic or is
the evidence unclear because other factors may blur the link between advertising and
quality? Read some of the relevant literature and make up your mind.
88

• analyse how product differentiation relaxes price competition
• analyse how firms choose product characteristics in the context of
simple models of product differentiation
• describe the theory and evidence on product proliferation as an entry
deterrence strategy
• compare horizontal and vertical product differentiation and explain
their different implications for market structure and profitability
• explain the notions of moral hazard and adverse selection and describe
how asymmetric information affects the working of product markets.

1. Discuss the view that incumbent firms in an industry can use a strategy
of product proliferation to deter entry. Present the basic argument
using a model of horizontal product differentiation, examine critically
the assumptions necessary for this strategy to work, and briefly discuss
any relevant empirical evidence.
2. Describe, with reference to any appropriate economic models, at least
two different reasons why firms can make considerable net profits in
the long run even under conditions of free entry.
89
Notes
90
Chapter 8: Price discrimination
Introduction
Firms often sell the same product at different prices to different
consumers. This practice is called price discrimination. There is also
a more subtle form of price discrimination, namely using different
qualities of a good or different packages in order to separate consumers
into different groups. In all these cases, the objective of the firm is to
increase its profit by capturing a larger fraction of consumer surplus than
it would capture under uniform pricing or by offering a single quality or
package. The scope for price discrimination is higher in those markets
where the possibilities for arbitrage are lower. Arbitrage is the practice
whereby consumers with low reservation price buy the good to resell it to
consumers with high reservation price. Note, however, that not all price
differences for the same good are due to price discrimination.
We distinguish three types of price discrimination.
1. First-degree (or perfect) price discrimination occurs when a firm
manages to capture all the consumer surplus by selling each unit of a
good at its reservation price. This is a useful benchmark, although it
is unlikely to occur in practice, because a firm typically does not have
exact information on each consumer’s preferences and also because of
extensive possibilities for arbitrage.
2. Second-degree price discrimination occurs when a firm cannot
distinguish between consumers with different preferences on the
basis of an exogenous characteristic. However, it extracts part of the
consumer surplus by using self-selecting devices, such as different
packages or options, that separate consumers into different groups
according to their preferences. That is, the firm designs options in
such a way that different groups of consumers select different options.
Examples include business versus economy class fares, hardback versus
paperback editions of books, end-of-season sales, and last-minute
bargain holidays.
3. Finally, under third-degree price discrimination a firm extracts
part of the consumer surplus by using direct signals about demand of
different groups of consumers, such as age, occupation, location, etc.
That is, the firm has some information on differences in preferences
between these groups and uses this information to price-discriminate.
Examples include special prices for students and senior citizens,
different journal subscription prices for libraries and individuals,
and different prices across countries for a variety of products
(pharmaceuticals, DVDs, cars, and so on).
Many of the interesting issues regarding price discrimination can be
analysed in the context of a single firm, abstracting from oligopolistic
interaction between firms. We follow this approach below. The firm is
formally a ‘monopolist’, but you should think of this as meaning, more
generally, ‘firm with significant market power’ (possibly operating in an
oligopolistic industry). In fact, price discrimination is more difficult in
markets where competition is fierce and firms have little market power.
91
Learning outcomes
• explain the notions of first-degree, second-degree and third-degree
price discrimination
• analyse simple models of third-degree price discrimination, derive the
inverse elasticity rule and discuss welfare implications and applications
of the analysis
• describe the use of a two-part tariff and of fully nonlinear pricing by a
monopolist that practices second-degree price discrimination
• describe the use of tie-in sales as a form of second-degree price
discrimination and explain other possible reasons for tying.
Essential reading
Church, J.R. and R. Ware Industrial organization: a strategic approach.
Chapter 5.
Further reading
Books
2000) Chapter 10.
Pearson Addison Wesley, 2005) Chapters 9–10.
Pepall L., D. Richards and G. Norman, Industrial organization: Contemporary
Chapters 5–6 and 8.
Stole, L.A. ‘Price discrimination and competition’, in Armstrong, M. and R.
North-Holland, 2007).
First-degree price discrimination

We briefly sketch a simple model of first-degree price discrimination. A
monopolist offers a single good and there are N identical consumers, each
with the demand function q = D(p)/N. This demand function is known to
the monopolist. Under standard linear pricing, the firm maximises profit
by charging the monopoly price for each unit sold and makes monopoly
profit, but it cannot capture any of the consumer surplus at that price.
However, the firm can make a higher profit by using a two-part tariff, in
other words a pricing schedule of the form T = A + pq: this implies that in
order to have the right to consume the good, each consumer must pay a
fixed fee A.1 Suppose the firm charges price p per unit sold; then each
1
Under linear pricing,
A=0.
consumer would be willing to pay an amount up to her net consumer
surplus at that price. What is the combination of p and A that yields
maximum profit to the firm?
It turns out that the firm maximises profit by setting p equal to the
‘competitive’ price (i.e. the price given by the intersection of the aggregate
demand curve D(p) and the marginal cost curve) and A equal to the
net consumer surplus of each consumer at that price. Thus the firm
appropriates the entire social surplus (producer surplus plus consumer
surplus) at the ‘competitive’ price.
92
What is the deadweight loss under first-degree price discrimination? It

is zero, since the price per unit is equal to marginal cost. There is a large
redistribution from consumer to producer surplus, but the total social
welfare is maximised (assuming there is no rent-seeking for maintaining a
monopoly position that leads to producer surplus being spent on activities
without social value).
If consumers have heterogeneous demands, the optimal policy is to set the
per unit price equal to marginal cost and to have a personalised fixed fee
equal to each consumer’s net surplus at that price. This is very difficult to
implement because the firm must know each consumer’s demand function,
and consumers with high demands have no incentive to reveal them
truthfully. We now turn to a form of price discrimination that does not
require knowledge of the demand function of each individual consumer.
Third-degree price discrimination

The main feature of third-degree price discrimination is that the
monopolist can separate the aggregate demand into m ‘groups’ or ‘markets’
on the basis of some exogenous and observable buyer characteristic. These
m groups have m distinct demand functions for the product, all of which
are known to the monopolist. Assume that there is no arbitrage between
groups, but the monopolist cannot price-discriminate between consumers
within a given group. In this situation, the monopolist’s problem is to
choose a linear tariff (a per unit price) for each of the m markets. That is,
the monopolist chooses prices p1, p2, …, pm to maximise profit:
m
∑i=1 pi Di(pi) – C(q) ,
where Di(pi) denotes the quantity demanded in market i and:

m
q =∑i=1 Di(pi) .
The optimal set of prices is then given by the inverse elasticity rule:
pi – C'(q) 1
=
pi εi
for each market i, where εi is the absolute value of the price elasticity of
demand in market i. Thus the monopolist should charge a higher price
in markets with lower elasticity of demand – that is, in markets where
consumers are less responsive to changes in price. This may well explain
why profit-maximising firms offer discounts to students and senior
citizens, why many utilities charge different business and household rates,
or why differences in the price of a product across regions or countries
are sometimes hard to explain only in terms of cost differences and trade
restrictions. Note, however, that before attributing any such cases to
price discrimination, one has to check that arbitrage between markets is
impossible or costly.
We now prove the inverse elasticity rule for the case where the monopolist
serves two different markets (the proof is similar for the general case of m
markets). The firm chooses prices p1 and p2 to maximise profit
Π = p1D1(p1) + p2D2(p2) – C[D1(p1) + D2(p2)]. The first-order conditions
are given by ∂∏/∂pi = Di(pi) + pi Di′(pi) – C′[D1(p1) + D2(p2)]Di′(pi) = 0,
for i = 1,2. Now move the term Di(pi) to the right-hand side, divide both
sides by piDi′(pi) and set q = D1(p1) + D2(p2). We obtain:
pi – C'(q) 1
=
pi −pi D'i / Di ,
for i = 1, 2, which is the inverse elasticity rule since –piDi′(pi)/Di is the
absolute value of the price elasticity of demand in market i.
93
The welfare implications of third-degree price discrimination are as follows.

Consumers in low-elasticity markets are hurt by price discrimination since
they pay a higher price, buy a smaller quantity and enjoy less consumer
surplus than under uniform pricing. The reverse is true, however, for
consumers in high-elasticity markets – they gain from price discrimination.
Moreover, producer surplus increases. So the overall effect on social welfare
is ambiguous and depends on the specifics of demand functions.
Note that these results hold when the monopolist serves both markets in
the absence of price discrimination. However, if price discrimination were
prohibited, say, then the firm might choose not to serve the high-elasticity
market at all, and that would definitely result in lower welfare than under
price discrimination. This is an additional reason for a cautious public
policy towards third-degree price discrimination.
What if a monopolist that sells a product to different groups of buyers and
wants to price-discriminate is prevented from doing so by arbitrage or
for some other reason? An interesting case arises when the monopolist is
selling an intermediate product to other producers. Suppose, for instance,
that the monopolist sells to two different groups of buyers, the demands
of which are independent, with elasticities ε1 < ε2. It turns out that, if
price discrimination is not feasible, the monopolist can still achieve the
same profit as under price discrimination by vertically integrating into the
market with the high elasticity of demand (i.e. market 2).
Activity
Prove this result.
Answer
If price discrimination were feasible, the monopolist would set prices p1 and p2 < p1,
according to the inverse elasticity rule for markets 1 and 2. If price discrimination is not
feasible, the monopolist can buy a firm in market 2 and sell the final product in that
market at a price that reflects the low selling price for the intermediate good, namely p2.
The monopolist can also announce that it will sell its intermediate good to any firm in
either market at price p1. Then there is no apparent price discrimination. However, this will
drive all the other firms in market 2 out of business, since they will not able to effectively
compete with the monopolist’s subsidiary. Thus the monopolist will achieve the same
profit as under price discrimination.2 2
See Tirole (1988),
p.141, for details.
Second-degree price discrimination

Suppose now that there is no observable characteristic on the basis of
which the monopolist can separate consumers into different groups, so
third-degree price discrimination is not feasible. In this case the firm
can still discriminate indirectly by offering various packages or options
which are designed so that different groups buy different packages (i.e.
self-select). A simple form of second-degree price discrimination is a
single two-part tariff T(q) = A + pq. This essentially offers a continuum
of bundles (q,T) for consumers to choose from. Two-part tariffs are quite
common, although more complex non-linear pricing schemes are optimal
for the firm, at least if implementation and monitoring costs are not too
high.
Two-part tariff
Let us first consider a model where the firm can only use a two-part tariff
to price-discriminate.3 There are two types of consumers, with utility
3
This part and the next
follow Tirole (1988),
function U = θV(q) – T if they consume q units of the good and pay T, and
pp.143–148.
94
U = 0 if they do not buy. We assume that V(0) = 0, V′(q) > 0, and V″(q) <
0. θ is a taste parameter that differs across the two consumer types, with θ2
> θ1. The firm produces the good at the constant marginal cost c < θ1. It
knows the values of the θ’s, but it cannot tell the consumers apart.
A type θi consumer chooses qi to maximise Ui. The first-order condition,
setting T = A + pqi , is given by θi V′(qi) – p = 0 ⇔ p = θi V′(qi). This is the
inverse demand function of a type θi consumer. Since V″(q) < 0 and
θ2 > θ1, it immediately follows that, for any given per unit price p, q2 > q1.
That is, the demand curve of θ2 types is always above the demand curve of
θ1 types (see Figure 8.1). It can also be seen from Figure 8.1 that the net
consumer surplus is larger for θ2 types than for θ1 types, for any given price
p. Let Si(p) denote the net consumer surplus of type θi.
p
area ABC: net consumer surplus for θ1 types
B
area ADE: net consumer surplus for θ2 types
demand for θ2 types

A E C
demand for θ1 types
q
Figure 8.1
In these circumstances, the monopolist will compare its profits from two
possible courses of action – and choose the one that gives the highest
profit. The first is to serve only the type θ2 consumers. In this case the
problem reduces to one of choosing the optimal two-part tariff for first-
degree price discrimination applied to type θ2 consumers.4 The second 4
Recall that all
consumers of a certain
course of action is to serve both types of consumers. The problem then is
type are identical, so
to choose the two-part tariff that maximises total profit subject to giving arbitrage within a group
an incentive to both types to buy. Suppose the per unit price is p. Then the is not an issue.
highest fixed fee that a type θ1 is willing to pay for the right to consume
the good is A = S1(p), his consumer surplus. A type θ2 is also willing to pay
this fixed fee because S2(p) > S1(p). On the other hand, the firm has no
reason to set a lower fixed fee for any given p – it would be throwing away
profits. So the firm will set A = S1(p) and its problem then will be to
choose the per unit price p that maximises its overall profit:
∏ = NS1( p) + ( p – c)D( p) ,
where N is the number of consumers and D(p) is the aggregate demand of
the N consumers at price p.5 Setting ∂Π/∂p = 0 implicitly defines the 5
In Tirole’s exposition
optimal per unit price p*, and hence also the optimal fixed fee A* = the number of
consumers is normalised
S1(p*). The monopolist will therefore offer the two-part tariff A* + p*q,
to 1, so the aggregate
and type θ2 consumers will choose to consume a higher quantity than type demand as defined
θ1 consumers. here is N times the
aggregate demand as
To conclude, let pM, pFD and pSD denote the profit-maximising per unit
defined in Tirole. You can
prices under uniform monopoly pricing, first-degree price discrimination see that N is simply a
and second-degree price discrimination respectively in the above model, multiplication factor in
on the assumption that both consumer types are served. Use similar this model with constant
superscripts for profits Π and total social welfare W (the sum of producer marginal cost and its
and consumer surplus). It can be shown that pM > pSD > pFD = c, ΠM < ΠSD value does not affect the
results.
< ΠFD, and WM < WSD < WFD. Note that in these comparisons the higher
the per unit price, the lower the welfare, because the fixed fee is just a
95
redistribution from consumers to the firm and hence does not affect W.
So when both consumer types are served, the two-part tariff results in
higher overall welfare compared to uniform monopoly pricing, although
it redistributes surplus from consumers to the firm. This result does not
necessarily hold if the firm decides to serve only the high-demand group.
Tie-in sales as price discrimination

The above framework can be used to analyse how a firm may use tie-
in sales as a price discrimination device. Suppose that a firm is the only
producer of a ‘basic’ good that is consumed in fixed quantity and jointly
with a ‘complementary’ good which is consumed in variable amounts
and can also be supplied by a competitive industry at a price equal to its
marginal cost. For instance: photocopying machines and copying paper,
durable good and maintenance. Suppose, as above, that there are two
types of consumers, with utility function U = θV(q) – T if they consume
q units of the complementary good and pay T, and U = 0 if they do not
buy. Consumers buy only one unit of the basic good, or none. θ is a taste
parameter that differs across types, with θ1 < θ2. The unit cost of the basic
good is c0, while the unit cost of the complementary good is c. The reason
why the above analysis of two-part pricing is relevant in the present case
is that the basic good can be seen as a prerequisite for consuming the
complementary good. Hence the maximum amount a consumer would be
willing to pay for the basic good is the net surplus he or she derives from
the consumption of the complementary good.
In this context, the manufacturer of the basic good can increase his profit
if he makes the purchase of the complementary (‘tied’) good a condition
for the sale of the basic (‘tying’) good to any buyer and can also prevent
arbitrage. How? Say the firm sets a price p for the complementary good
and a price A for the basic good. On the assumption that the firm serves
both types of consumers, it will set A = S1(p) and choose p to maximise
Π = NS1(p) + (p – c)D(p) – Nc0, so the result will be the same as in our
model of two-part pricing above.6 In particular, p* > c. 6
The cost of production
of the basic good is a
If the firm could not use tie-in sales, the price of the complementary good fixed cost, so it does
would be equal to its marginal cost c, and the producer of the basic good not affect the firm’s
would set A = S1(c) > S1(p*). In other words, the result of tie-in sales is a maximisation problem
higher price for the complementary good and a lower price for the basic over p. However, it
good. This amounts to price discrimination hurting type θ2 consumers. will be relevant for the
decision to sell to both
Note that p* > c implies that welfare is unambiguously lower under tie-in
groups of consumers or
sales if both consumer types are served. That’s because the use of tie-in to only one group.
sales results in a lower than socially efficient level of consumption of the
complementary good – the fact that the price of the basic good is lower is
irrelevant since this is a redistribution from the manufacturer of the basic
good to consumers. On the other hand, you must bear in mind that, if tie-
in sales are not allowed, the chances increase that the firm will choose to
serve the θ2 types only, so welfare could in fact increase under tie-in sales.
Important note: You should bear in mind is that price discrimination
is not the only reason for tie-in sales. In fact, several common reasons for
tying are not related to price discrimination:
• Tying is sometimes a profitable strategy for a firm when there are
at least two consumer types whose demands for the tied goods are
negatively correlated (i.e. some consumers value good 1 more than
good 2 and others value good 2 more than good 1). Selling the
goods as a bundle may allow the firm to extract more surplus from
consumers if some consumers who would buy only one of the goods if
they were sold separately now buy the bundle.
96
• Tie-in sales may also be used by a firm with market power in one
market to eliminate competitors and gain market share in another
market (the one for the tied good) where it initially has little
market power. Note that we do not need to have consumers with
heterogeneous preferences for this to work. This was the main charge
against Microsoft in what was perhaps the most widely publicised
antitrust case of the last 30 years.
• Nor are tie-in sales necessarily welfare reducing: they may lower
transaction costs and increase efficiency, as in the case of goods that
are usually consumed together and/or their joint production and
marketing are subject to significant economies of scope, thus resulting
in lower prices that benefit consumers. This was one of the main lines
of defence in the Microsoft case. In fact, much tying occurs in fairly
competitive markets by firms without significant market power.
Optimal nonlinear pricing

As mentioned above, a two-part tariff is just one form of second-degree
price discrimination. In many situations, a monopolist can use a more
complex and more profitable pricing policy. This consists in offering
different packages or options, each directed to one consumer type.
These packages are chosen to maximise the firm’s profit subject to
participation constraints (i.e. each type must be willing to buy) and
incentive-compatibility constraints (i.e. no type must have an incentive
to buy a package intended for another type). For example, there may be
two options for membership of a gym, one that involves a fixed monthly
fee and unlimited visits at no extra cost and one with no fixed fee but
a relatively high price per visit: the first option would be intended for
frequent users, the second for occasional visitors. Or the options may be
differentiated by quality, if different consumers have different willingness
to pay for quality: higher quality products at higher prices for high-
valuation consumers and lower quality products at lower prices for low-
valuation consumers. Or the same product may be sold at a higher price
initially, when availability is guaranteed, and a lower price after a certain
time, and possibly also subject to availability – in an attempt to separate
less patient or more committed customers from more patient or less
committed ones.
When there are two consumer types and therefore two packages, it turns
out that one important feature of the optimal pricing policy is that low-
demand consumers consume an inefficiently small amount of the good
while high-demand consumers consume efficiently. Furthermore, low-
demand consumers derive no net surplus, while high-demand consumers
derive a positive net surplus, just as in our model of two-part pricing.
To see this, let us go back to the model we used above to analyse second-
degree price discrimination with a two-part tariff. Recall that there are two
types of consumers, with utility function U = θV(q) – T if they consume q
units of the good and pay T. θ is a taste parameter that differs across the
two types, with θ2 > θ1, so the θ2 types are high-demand and the θ1 types
are low-demand consumers. Furthermore, we assume that the proportion
of θ1 types in the population is λ, so the proportion of θ2 types is 1 – λ. The
firm knows the values of the θ’s and of λ, but it cannot tell the consumers
apart.
To optimally price-discriminate, the firm offers two packages, where each
package consists in q units of the product sold for an overall charge of T.
The package (q1, T1) is directed at θ1 types, while the package (q2, T2) is
directed at θ2 types. Assume that both types are served. Finally, to simplify
97
the analysis, we set without loss of generality the number of consumers

to 1.
The firm seeks to maximise its profit:
∏ = λ(T1 – cq1)+(1 – λ)(T2 – cq2)
subject to four constraints – a participation constraint for the low-demand
consumers:
θ 1V (q1)– T1 ≥ 0, (1)
a participation constraint for the high-demand consumers:
θ 2V (q2)– T2 ≥ 0, (2)
an incentive-compatibility constraint for the low-demand consumers:
θ 1V (q1)– T1 ≥ θ 1V (q2)– T2, (3)
and, finally, an incentive-compatibility constraint for the high-demand
consumers:
θ 2V (q2) – T2 ≥ θ2V (q1)– T1, (4)
Solving this problem confirms that the firm’s profit-maximising choice of
the bundles (q1, T1) and (q2, T2) implies appropriating the entire surplus of
the low-demand types but leaving some surplus to the high-demand types.
Activity
Prove this result.
Answer
This is not as complicated a problem as it might look at first sight. To begin with, note
that if (1) holds, then high-demand consumers will always choose to buy – they can buy
the bundle (q1, T1) and obtain net surplus θ2V(q1) – T1 > 0 (recall that θ2 > θ1). So we can
ignore (2), the participation constraint for high-demand consumers and proceed.
Furthermore, let us also ignore (3), the incentive-compatibility constraint for the low-
demand consumers – we will check later that it is satisfied.
The firm’s problem now is to maximise its profit subject to (1) and (4). Since the firm
always benefits from high fees, (1) and (4) will hold with equality. From (1) we obtain T1
= θ1V(q1): the firm appropriates the entire surplus of the low-demand types. From (4),
also using (1), we get T2 = θ2V(q2) – (θ2 – θ1)V(q1). Since (θ2 – θ1)V(q1) > 0, we conclude
that T2 < θ2 V(q2): the firm leaves some surplus to the high-demand types. Why? Because
otherwise they would buy the bundle targeted for the low-demand types and obtain
positive surplus – so price discrimination would not work (the types would not self-select
as the firm wants them to).
Next, we substitute T1 and T2 from the above equations into the profit function and
solve for the profit-maximising values of q1 and q2. From the first-order conditions we
get θ2V′(q2) = c and θ1V′(q1) > c, which imply that low-demand consumers consume an
inefficiently small amount of the good while high-demand consumers consume efficiently
and also that q1 < q2.
Finally, it can be easily checked that when (1) and (4) hold with equality, (3) is also
satisfied.
98
Activities
1. Starworks Ltd. sells to two types of consumers, with respective demand functions q =
1 – p (in proportion λ) and q = 2 – p (in proportion 1 – λ). Its unit production cost is
zero. The firm faces a fringe of competitive firms selling the same good at price p0 < 1
(presumably these firms have unit cost equal to p0).
a. What is the optimal pricing policy of Starworks Ltd under standard linear pricing
(i.e. in the absence of any price discrimination)?
b. If the firm can tell consumers apart and prevent arbitrage, what is the optimal
pricing policy of Starworks Ltd?
2. Consider the same setup as in the previous question, but now assume that the firm
cannot tell consumers apart. What is the optimal two-part tariff under second-degree
price discrimination?
3. ‘Caffè’ is a small shop located next to the Department of Economics where
generations of economics students have bought their coffee in the breaks. Mr George,
the owner of ‘Caffè’, is a keen user of price discrimination. In particular, he sells
either one cup of coffee for £1 or ten cups of coffee for £8. As most people can only
drink one cup of coffee at a time, those who buy ten cups are allowed to spread the
consumption over time using a system with a card. What type of price discrimination
is this and how could it work in this particular example? Mr George often hears
students saying: ‘I don’t want to buy the card for ten cups of coffee, because then I
drink too much coffee’. Based on your knowledge about price discrimination, do you
think that students with a card drink too much coffee? Explain.
4. A competition authority has hired you to evaluate the market for rental cars. A survey
of customers of the top five car rental firms (which account for approximately 80 per
cent of all rentals) at five large airports reveals substantial variation in the rental rates
charged to different customers of the same firm. Rates vary considerably across a
large number of dimensions: across different airports, across days of the week (with
weekend rates substantially lower than Monday to Thursday rates), over rental periods
(one-day versus weekend, week or month), and across car models. In addition, there
appear to be a large number of promotional rates used by different customers (AAA
discounts, corporate discounts, advertised specials, advance reservation rates etc.), so
that rentals that appear to have identical characteristics (day of week, location, length
of rental, model of car) often entail different prices.
What are possible explanations for non-uniform pricing in a market? Assess the
plausibility of each explanation for the pattern of pricing observed in the rental car
market. Include in your assessment any preconditions that attach to each explanation
and whether they are likely to be satisfied in this market.
5. Is there a link between the degree of competition in a market and the extent of price
discrimination by firms? Although some degree of market power is necessary for price
discrimination, it is not obvious whether price discrimination would rise or fall with
the degree of competition or the level of concentration in an oligopolistic industry.
Assess the theoretical arguments and the empirical evidence on this question. Good
starting poins are Stole (2007); Stavins (2001); and Gerardi and Shapiro (2009).
99

• explain the notions of first-degree, second-degree and third-degree
price discrimination
• analyse simple models of third-degree price discrimination, derive the
inverse elasticity rule and discuss welfare implications and applications
of the analysis
• describe the use of a two-part tariff and of fully nonlinear pricing by a
monopolist that practices second-degree price discrimination
• describe the use of tie-in sales as a form of second-degree price
discrimination and explain other possible reasons for tying.

1. There are two types of consumers of a certain good, with utility
function U = θV(q) – T if they consume q units of the good and pay
T, and U = 0 if they do not buy. Assume that V(q) = [1 – (1 – q)2]/2,
the same for all consumers. θ is a taste parameter that differs across
types: it takes the value θ1 for a fraction λ of the total population of
consumers and the value θ2 > θ1 for the remaining fraction 1 – λ.
The total number of consumers is normalised to 1 for simplicity. A
monopolist produces the good at unit cost c < θ1 and knows the values
of θ1, θ2 and λ.
a. Compute the demand function and the net consumer surplus for
the two types and show that the net consumer surplus is higher for
θ2 types.
b. Derive the aggregate demand function. (Hint: define θ = 1/[λ/θ1
+ (1 – λ)/θ2] to simplify the algebra.)
c. Suppose that, because of full arbitrage between consumers, the
monopolist cannot engage in any form of price discrimination.
Derive the optimal linear tariff on the assumption that the firm
chooses to serve both types of consumers.
d. Now suppose that the firm can observe each consumer’s type and
there is no arbitrage between types, so it can practice first-degree
price discrimination. Derive the optimal two-part tariff for each
type.
e. Finally, suppose that the firm cannot tell the consumers apart
and there is no full arbitrage, so it practices second-degree price
discrimination by setting a two-part tariff. Derive the optimal two-
part tariff on the assumption that the firm chooses to serve both
types of consumers.
2. Analyse how tie-in sales can be used as a price discrimination device
and discuss the welfare implications of this practice.
100
Chapter 9: Vertical relations
Introduction
In nearly all the topics discussed so far in the second part of this guide we
have made the simplifying assumption that firms produce goods which
they then sell directly to the final consumers. In this chapter, in contrast,
we focus on situations where ‘upstream’ firms possessing market power
sell to ‘downstream’ firms which, in turn, sell to the final consumers.
Examples include manufacturers selling to retailers, or manufacturers of
an intermediate good selling to manufacturers of the final good. We will
be using the terms ‘manufacturer’ and ‘retailer’ to denote respectively the
upstream and the downstream firm in what follows.
In these situations, manufacturers sometimes impose contractual restraints
on retailers, such as:
• resale price maintenance (RPM): the imposition by the
manufacturer of a fixed or minimum retail price
• franchise fee or two-part pricing: the payment by the retailer of
a fixed fee in order to carry the manufacturer’s product – so the total
payment from retailer to manufacturer is T(q) = A + pwq, where the
manufacturer chooses the fixed fee A and the wholesale price pw and
the retailer chooses quantity q or the retail price p
• exclusive territories: the allocation to each retailer of exclusive
rights to sell in a territory
• exclusive dealing: the requirement that a retailer does not sell
products of rival manufacturers.
What is the reason for imposing these vertical restraints? Are they
used to restrict competition, in which case they unambiguously reduce
social welfare? Or are they used to increase the efficiency of the vertical
relationship, in which case they may be socially beneficial as well? In this
chapter we examine efficiency arguments for vertical restraints as well as
theories that view them as restrictions on competition. Throughout our
analysis the efficiency of a vertical arrangement refers primarily to the
optimal (joint profit-maximising) outcome for the firms involved, and
may or may not imply welfare maximisation for society as a whole. Some
empirical evidence is also discussed and policy implications are drawn.
A common feature of the models described in this chapter is that upstream
firms unilaterally determine the price and any vertical restraints that
downstream firms face. While this is a realistic description of many
upstream-downstream relationships, it is not a general feature. In some
markets where downstream firms have significant market power, prices
and terms of sale are determined through bargaining between buyers and
sellers.
Learning outcomes
• analyse models of a vertical relationship between a manufacturer and
a retailer and show how vertical restraints can be used to eliminate
inefficiencies in such a relationship
101
• evaluate various efficiency arguments for vertical restraints in the light

of any relevant empirical evidence
• explain at least three different reasons why vertical restraints can
restrict competition
• use the theory on vertical restraints to discuss any relevant empirical
evidence and evaluate public policy in this area.
Essential reading
Books
Church, J.R. and R. Ware Industrial organization: a strategic approach. Chapter
22.
Journal
Rey, P. and J. Stiglitz. ‘The role of exclusive territories in producers’
competition’, Rand Journal of Economics 26(3) 1995, pp.431–51.
Further reading
Books
2000) Chapter 11.
Competition Commission New cars: a report on the supply of new motor cars
within the UK, Cm 4660 (London: TSO, 2000).
Monopolies and Mergers Commission (MMC) New motor cars: a report on the
supply of new motor cars within the UK, Cm 1808 (London: HMSO, 1992).
Katz, M.L. ‘Vertical contractual relations’, in Schmalensee, R. and R. Willig
(eds) Handbook of industrial organization, Volume 1. (Amsterdam: North-
Holland, 1989).
Chapters 17–18.
Journals
Cooper, J.C., L.M. Froeb, D. O’Brien and M.G. Vita ‘Vertical antitrust policy as
a problem of inference’, International Journal of Industrial Organization 23
2005, pp.639–64.
Mathewson, G.F. and R.A. Winter ‘An economic theory of vertical restraints’,
Rand Journal of Economics 15(1) 1984, pp.27–38.
Efficiency arguments for vertical restraints

The basic vertical externality
We start with the simplest setup involving one manufacturer and one
retailer.1 The manufacturer sells to the retailer at wholesale price pw > c, 1
where c is the unit production cost. Assume that c < 1. The retailer, who (1988), pp.174–178.
resells the product to the final consumers, takes pw as given and chooses
the retail price p. Demand by consumers is given by q = D(p) = 1 – p (so
the choice of p also determines q).
The first-best or ‘efficient’ outcome is the one which maximises the
aggregate profit (i.e. manufacturer’s profit plus retailer’s profit). This is
102
found by assuming that the two firms merge into an ‘integrated structure’.
In this case the problem is to choose p to maximise (p – c)(1 – p). The first-
order condition is –(p – c) + (1 – p) = 0 ⇔ pI = (1 + c)/2. Substituting
into the profit function, we obtain total profit ΠI = (1 – c)2/4.
What happens if the two firms make separate decisions? First the
manufacturer chooses pw , then the retailer observes pw and chooses p.
This is a two-stage game, so we use backward induction. At stage 2 the
retailer chooses p to maximise his profit (p – pw) (1 – p). The first-order
condition gives pNI = (1 + pw)/2. Substituting into the retailer’s profit
function, we obtain the retailer’s profit ΠR = (1 – pw)2/4. At stage 1 the
manufacturer chooses pw to maximise his profit (pw – c)(1 – p) anticipating
the retailer’s reaction, in other words anticipating that p = (1 + pw)/2.
The manufacturer’s problem is therefore to choose pw to maximise (pw – c)
[1 – (1 + pw)/2]. The first-order condition gives pw = (1 + c)/2. Note
that this also implies pNI = (3 + c)/4 and ΠR = (1 – c)2/16. Substituting
the value for pw into the manufacturer’s profit function, we obtain the
manufacturer’s profit ΠM = (1 – c)2/8. Therefore the total profit ΠM + ΠR
= 3(1 – c)2/16 < (1 – c)2/4 = ΠI. Moreover, pNI = (3 + c)/4 > pI.
The conclusion is that the non-integrated industry makes lower total profit
than the integrated structure. Also, the price to consumers is higher if the
industry is not integrated – and hence the quantity consumed is lower.
This is the result of double marginalisation. Our example shows that
double marginalisation is bad both for firms and for consumers. Therefore
eliminating it causes both profit and consumer surplus to increase. This
is a general result: it holds also for a general demand function. The
intuition is that the retailer could, by reducing the retail price below
pNI, increase demand for the product and generate incremental profit
pw – c for the manufacturer which would more than compensate for the
loss of own profit; however, the retailer cares only about own profit and
therefore ignores this ‘vertical externality’. Note that if either the upstream
or the downstream industry is perfectly competitive, there is no double
marginalisation and no vertical externality (since pw = c or p = pw or
both).
Obviously, vertical integration would eliminate double marginalisation and
restore efficiency. How can this be achieved without vertical integration
but through the use of appropriate vertical restraints? One solution is
resale price maintenance in the form of a fixed price or a price ceiling
(maximum price). The manufacturer sets pw = p* and imposes a constraint
on the final price: p ≤ p*. The retailer then sets p = p* and makes zero
profit. Note that a minimum price would not work in this case. A potential
disadvantage of resale price maintenance is that the retailer bears all the
risk when we allow for demand uncertainty.
Another solution is a franchise fee (two-part tariff). The manufacturer sets
pw = c, and extracts the retailer’s profit through a franchise fee A = (1 –
c)2/4. Given that pw = c, the retailer will choose p to maximise (p – c)(1
– p). As we have seen above, this gives p = (1 + c)/2 = pI and profit equal
to (1 – c)2/4 (i.e. the same as for the integrated industry).
Note, however, that there are also some problems with franchise fees in
more general settings. In particular, the retailer bears all the risk if demand
or cost is uncertain, unless the franchise fee is adjusted; the retailer may
have private information about demand or cost, so it may be hard for the
manufacturer to compute and extract the retailer’s profit by a franchise fee
– so the manufacturer may have to share the rents with the retailer; and in
a model with many retailers, franchise fees are not sufficient to extract ΠR.
103
Provision of services
Let us now extend and generalise this model. Suppose the manufacturer
sells to the retailer at price pw > c, where c is the unit production cost.
The retailer takes pw as given and chooses the retail price p. However,
the retailer can also supply pre-sale services – information, free delivery,
advertising in the local press, etc. – that increase demand for the product.
The retailer then chooses the level of services s along with the retail price.
Assume that the manufacturer cannot specify the level of s in the contract
with the retailer, since s is difficult to observe and/or verify in court in case
of dispute. Demand by consumers is given by q = D(p,s), with Dp = ∂q/∂p
< 0 and Ds = ∂q/∂s > 0. Finally, assume that supplying a level s of services
costs the retailer Φ(s) per unit of output, with Φ′ = dΦ/ds > 0. So the total
cost of services is qΦ(s).
The first-best outcome is the one which maximises the aggregate profit
(i.e. manufacturer’s plus retailer’s):
max∏J =[ p – c – Φ( s)]D( p, s)=( pw – c)D( p, s)+[p – pw – Φ(s)]D( p, s)
p,s
This maximisation problem gives first-order conditions:

( pw – c)Dp( p, s)+[p – pw – Φ(s)]Dp( p, s)+D( p, s)= 0 (1)
and
( pw – c)Ds( p, s)+[p – pw – Φ(s)]Ds( p, s) – Φ'(s)D( p, s)= 0 (2)
Denote by pm, sm the values of p, s that maximise the aggregate profit.
These are the efficient choices (from the point of view of the integrated
structure).
Now what happens when the two firms make separate decisions? First
the manufacturer chooses pw, then the retailer chooses p and s. Clearly,
the manufacturer’s choice of pw to maximise his profit (pw – c)D(p,s)
involves setting pw > c. More interesting for our analysis is the retailer’s
maximisation problem. The profit of the retailer is [p – pw – Φ(s)]D(p,s).
The retailer chooses p and s to maximise this, which gives first-order
conditions:
[p – pw – Φ(s)]Dp( p, s) + D( p, s)= 0 (3)
and:
[p – pw – Φ(s)]Ds( p, s) – Φ'(s)D( p, s)= 0 (4)
Now compare equations (1) and (3). The term (pw – c)Dp(p,s) < 0 is
missing in (3). In other words, when choosing p the retailer does not
take into account the effect of this choice on the manufacturer’s profit. In
particular, the retailer sets p > pm: the price is ‘too high’. Compare also
equations (2) and (4). The term (pw – c)Ds(p,s) > 0 is missing in (4). That
is, when choosing s the retailer does not take into account the effect on
the manufacturer’s profit. In particular, the retailer sets s < sm: services are
underprovided. In both cases there is a vertical externality which occurs
because both the manufacturer and the retailer have market power. The
presence of these externalities hurts both consumers (through higher price
and lower level of services) and firms (since joint profit is not maximised).
To see more clearly why the retailer sets the price too high relative to the
efficient (joint profit-maximising) level, substitute the retailer’s first-order
condition with respect to p into the vertically integrated firm’s first-order
condition with respect to p. At the price chosen by the retailer, ∂ΠJ/∂p =
(pw – c)Dp(p,s) < 0, so a fall in p below the level chosen by the retailer is
needed for ΠJ to be maximised. Similarly, if we substitute the retailer’s
first-order condition with respect to s into the vertically integrated firm’s
first-order condition with respect to s, we can see that at the level of
104
service chosen by the retailer, ∂ΠJ/ϑs = (pw – c)Ds(p,s) > 0. Therefore

a rise in s above the level chosen by the retailer is needed for ΠJ to be
maximised; in other words, services are underprovided by the retailer
relative to the efficient level.
One solution is vertical integration. Another, through the use of vertical
restraints, is for the manufacturer to set pw = c, and extract the retailer’s
profit through a franchise fee A = [pm – c – Φ(sm)]D(pm,sm). Given that
pw = c, the retailer will choose p and s to maximise [p – c – Φ(s)]D(p,s),
which is the same as the maximisation problem of the integrated structure.
So the retailer chooses pm and sm and makes profit A. Hence the efficient
outcome is attained. Finally, a third solution is ‘quantity forcing’ (i.e.
the requirement that the retailer buys from the manufacturer a fixed
quantity) together with a linear wholesale price pw. Note that resale price
maintenance is not sufficient here because it cannot take care of the
service externality.
Horizontal externalities and free riding

A more complicated, and more realistic, situation involves one
manufacturer and many retailers. In addition to the vertical externalities
identified above, there are in this case horizontal externalities, that
is, externalities among retailers. An example is the provision of pre-sale
services. These create a public good problem, because each retailer cannot
fully appropriate the returns from supplying such services; for instance, a
consumer may obtain information about the product from one shop, then
buy it from another. The result is that retailers free-ride on each other and
services are undersupplied.
To fix ideas, while keeping the analysis simple, let us consider an extreme
case: marginal cost pricing downstream. Suppose there are two retailers
selling a homogeneous good and competing in advertising (service) and
price. The firm with the lowest price gets all the demand at that price,
while if p1 = p2 the firms share the market. Furthermore, total demand
increases in A1 and A2, where A1, A2 are advertising levels of the two
firms. Each retailer has a unit cost of pw and there is an additional cost
of advertising. Competition in this industry can be modelled as a three-
stage game: at stage 1, an upstream monopolist chooses pw; at stage 2,
the retailers simultaneously set advertising levels and incur advertising
costs; at stage 3, the retailers buy the good from the manufacturer and
simultaneously set retail prices. The timing of the game reflects the fact
that advertising decisions are more long-term than pricing decisions –
since the brand image created or the product information provided by
advertising respond relatively slowly to changes in advertising levels.
We use backward induction and start from the last stage. Here the retailers
take pw, A1 and A2 as given, since these have already been set. Bertrand
competition at this stage leads to marginal cost pricing: p1 = p2 = pw.
Now go back to stage 2. Each retailer anticipates the stage 3 outcome
and chooses advertising to maximise profit. It is then easy to see that
since stage 3 profit is zero anyway and advertising is costly, each retailer
will choose an advertising level of zero, irrespective of the pw set by the
upstream monopolist at stage 1.
The above simple model presented a case of extreme free riding. In
settings with imperfect competition among retailers (for instance, if the
good is not perceived by consumers as homogeneous at different retailers),
there will be some retailer profit and a positive level of advertising, but
less than the efficient level set by a vertically integrated structure.
105
Mathewson and Winter (1984) have constructed a model to examine free

riding and the use of vertical restraints when there is imperfect competition
in retailing. There are N retailers, differentiated by location. Retailers can
boost demand by advertising the product. In particular, there are two classes
of consumers, informed and uninformed, and a retailer can advertise to 2
Advertising that boosts
increase the number of informed consumers.2 However, a proportion of demand but not the
advertising messages ‘spill over’, that is, some consumers who get a message willingness to pay a
higher price is called
from one retailer will ultimately buy from another retailer. In this setup
‘informative’. Retailers’
there are both vertical and horizontal externalities that cause the amount of advertising is often of
advertising to be inefficiently low. Total demand and profit will be less than this kind. Advertising that
what a vertically integrated structure would achieve. increases the willingness
to pay, as in the vertical
The efficient outcome (from the point of view of the integrated structure) product differentiation
can be attained through various combinations of vertical restraints which models discussed in
internalise these externalities. An important point here is that retailers Chapter 7 of this guide,
must be given incentives by being allowed to capture the returns to their is called ‘persuasive’.
Manufacturers’
investments in service provision. So, for example, a franchise fee plus
advertising is often of
resale price maintenance in the form of a minimum price (or exclusive this kind: it aims to create
territories) does the trick. RPM (or exclusive territories) encourages a brand image for the
consumers to buy where the services are provided since they cannot find a product. Note, however,
better price elsewhere, so it deals with the horizontal externality. The fixed that these are basically
fee takes care of the vertical externality. analytical categories and
sometimes advertising
Exclusive dealing may have implications for efficiency similar to those is both ‘informative’ and
of RPM and exclusive territories: if retailers concentrate on selling the ‘persuasive’.
product of a particular manufacturer, this could lead to greater provision
of services. Exclusive dealing may also induce the manufacturer to supply
promotional services efficiently. That’s because the manufacturer will be
certain that retailers will not encourage the consumers to buy a competing
brand (which may carry a higher profit margin for retailers if the
competing manufacturer does not incur the same promotional expenses).
It should be emphasised again at this point that, in all the above
discussion, ‘efficiency’ refers to the optimal outcome for the integrated
structure (and – through the ultimate appropriation of the retailer profit
by the manufacturer via a franchise fee – for the manufacturer). This may
or may not be welfare-enhancing for society as a whole. For instance,
vertical restraints such as RPM or exclusive territories may well increase
the level of services provided, but they may also lead to higher prices. The
welfare implications are not clear.
Quality certification
The horizontal externality argument is often invoked as an explanation,
and justification, for resale price maintenance. However, RPM has often
been imposed on goods that involve few inappropriable pre-sale services.
What explains its use in such cases?
One story, a generalisation of the free-riding argument, is that the retailer
offers ‘quality certification’. The idea is that pre-sale services are not
tangible; rather the simple fact that a certain retailer carries the product
may signal that the product is of high quality. Now such a retailer may
be unwilling to store the product or the manufacturer may believe that
his own reputation will be damaged if the product can also be found in a
discount store at a low price. Hence the manufacturer imposes RPM or/
and refuses to supply discount stores.
A well-known case where the use of RPM may have been linked to an
attempt to offer quality certification is the case of Levi Strauss jeans. Until
the mid-1970s Levi Strauss set a minimum retail price for their jeans. They
106
distributed jeans through two sorts of outlets: traditional outlets, such as

department stores, and specialty stores (i.e. chains of jeans shops). They
refused, however, to sell through discount stores. Quality certification is
one likely reason for RPM and for refusing to sell to discount stores. In
any case, RPM reduced competition among retailers. Following a Federal
Trade Commission complaint, Levi Strauss abandoned RPM in 1977 – but
still kept their jeans out of discount stores. Price-cutting started from jeans
stores and spread to department stores and to other brands of jeans as
well. The result was a large increase in sales and profit of Levi Strauss.
Also, since prices were lower, consumer surplus increased.
One might conclude that setting a high minimum retail price was not
profitable for Levi Strauss, at least under the demand conditions of the
mid-1970s. Demand proved to be much more elastic than Levi Strauss had
thought, so a low-price strategy was in fact better not only for consumers
but also for Levi Strauss. Or was it? In 1980 the boom in jeans sales faded
and profits of Levi Strauss and other jeans makers fell. It is not clear
whether this had something to do with the entry in 1979 of ‘designer
jeans’ into the market. These filled the gap left by the move of Levi Strauss
to the low-price segment of the market, but may have happened anyway.
Overall, it is not clear whether setting a high minimum retail price was a
sensible strategy for Levi Strauss in the long term or not.
Vertical restraints as instruments that restrict

competition
There are various stories in the literature, three of which will be
mentioned here. Perhaps the most common argument against resale price
maintenance is that it may be used to restrict competition among retailers.
Suppose that unrestricted competition among retailers would result in a
retail price p*. If the manufacturers impose a fixed or minimum retail price
p > p*, competition among retailers will become less intense.3 This will 3
Think of this in terms of
increase retailers’ profits (provided p is not too high) and will reduce our analysis of reaction
functions in Chapter 6 of
welfare, since output will be further away from the socially optimal level
this guide.
than in the no-RPM case. Things may be even worse in the presence of
efficiency differences across retailers. RPM may then allow the survival of
the less efficient by preventing the more efficient from cutting price and
hence gaining market share. You should bear in mind that exclusive
territories may have similar effects on retailer competition as RPM.
A second story focuses on the effect of exclusive dealing or long-term
contracts with retailers on competition among manufacturers. The basic
idea is that exclusive dealing and long-term contracts can be used as a
barrier to entry into the upstream market. The reason is that they may
force new manufacturers to set up their own distribution networks, which
is costly. As a result, new manufacturers will be less inclined to enter.
Exclusive dealing and long-term contracts can also limit competition
among existing manufacturers and retailers. If, for instance, there is spatial
differentiation among retailers, exclusive dealing softens inter-brand
competition by introducing an element of spatial differentiation between
brands.
Finally, a third story argues that vertical restraints such as RPM or
exclusive territories may also be used to restrict competition among
manufacturers. RPM, for instance, could help competing manufacturers
sustain collusion by facilitating the detection of price cuts. Even in the
absence of collusion, such practices can soften competition among
manufacturers. Let us examine this story in more detail.
107
Exclusive territories and competition among manufacturers

We analyse a simple model, based on Rey and Stiglitz (1995), which
involves the use of exclusive territories. There are two manufacturers,
producing a differentiated product. Final demand for good 1 within any
given territory is given by q1 = 1 − p1 + bp2, while final demand for good
2 is q2 = 1 − p2 + bp1, where 0 < b < 1 − so the own-price effect on
demand is stronger than the effect of the rival’s price. The products are
distributed through retailers. Assume, for simplicity, zero production and
retail costs.
Consider first the case where there are no exclusive territories. In any
given territory, good 1 is distributed by a large number of retailers who
all compete with each other. Assume, for simplicity, that there is no
spatial differentiation among retailers in any given territory (consumers
have zero travel costs), so the retail price of good 1, p1, is driven down
to w1, which is the price at which the manufacturer of good 1 sells it
to retailers. Similarly for good 2: p2 = w2. Thus the retailers make no
profits. The manufacturers’ profits in any given territory are ∏1 = w1(1
– w1 + bw2) and ∏2 = w2 (1 − w2 + bw1). The two manufacturers set
their prices simultaneously. Manufacturer i chooses wi to maximise ∏i
taking wj as given, where i = 1, 2, j ≠ i. Solving the system of the two
first-order conditions and also substituting into the profit functions, we
obtain equilibrium prices and profits w1* = w2* = p1* = p2* = 1/(2 −
b), ∏1* = ∏2* = 1/(2 − b)2 (check these results). Everything is as if the
manufacturers sold directly to consumers.
Consider now the case where each manufacturer grants exclusive
territories to the retailers carrying his product. Hence intra-brand
competition is eliminated. The only competition now is between good
1 and good 2 in any given territory. The profit functions of the two
competing retailers – one selling good 1, the other selling good 2 – in any
given territory are (p1– w1)(1 – p1 + bp2) and (p2 − w2)(1 – p2 + bp1),
respectively, where w1 and w2 have already been set by the manufacturers
and are taken as given by the retailers. We assume that each retailer can
observe both w1 and w2. The two retailers set retail prices simultaneously.
Retailer i chooses pi to maximise his profit taking pj as given, where i =
1, 2, j ≠ i. Solving the system of the two first-order conditions, we obtain
equilibrium prices
2 + b +2wi + bwj
piR=
4 – b2
with corresponding quantities
R R R
2 + b – 2wi + b2wi + bwj
qi = 1 – pi + bpj =
4 – b2
where i = 1, 2, j ≠ i.
Competition among the retailers takes place at stage 2 of a two-stage
game. We assume that at stage 1 the manufacturers simultaneously choose
two-part tariffs (since this allows them to make higher profits than a linear
tariff), each of them anticipating the way the retailers will act at stage 2.
Manufacturer i, i = 1, 2, chooses a wholesale price wi and a fixed fee Fi
that the retailer of good i must pay if he wants to carry the good.
What do manufacturers do at stage 1? For one thing, each of them can use
the fixed fee to extract all the profit made by his retailer at stage 2. Also,
each of them may make some additional profit by charging a wholesale
price above marginal cost (which is zero). The total profit of manufacturer
108
1 in any territory is therefore given by w1q1R + F1 = w1q1R + (p1R − w1) q1R

= p1Rq1R, an expression which is a function of w1, w2 and b (see above).
Similarly for manufacturer 2. Manufacturer 1 then chooses w1 to maximise
p1Rq1R taking w2 as given, and similarly manufacturer 2 chooses w2 to
maximise p2Rq2R taking w1 as given. Solving the system of the two first-
order conditions, we get the equilibrium wholesale prices
b2
^
w 1
=w
^
2
= .
4 – 2b – b2
Substituting into the respective profit functions p1Rq1R and p2Rq2R, we obtain
the equilibrium profits
^ ^ 4 – 2b2
∏1 = ∏2 = .
(4 – 2b – b2)2
Activity
Go through the computations in the above model. In particular, derive the expressions for
^i
^ i and ∏
piR, qiR, piRqiR, w where i = 1, 2.
Answer
Straightforward.
^ ^
Are the profits ∏1 and ∏2 higher than those made by the manufacturers
in the absence of exclusive territories, that is ∏1* and ∏2*? Yes. It can be
^
easily verified that ∏1> ∏i* ⇔ b3(4 − 3b) > 0, i = 1, 2, which is always
true for b∈(0,1) (check). To find the quantities produced, substitute w ^ and
i
w^ into the expression for q R above. It can be checked that these quantities
j 1
are lower than those produced without exclusive territories – which were
already lower than the socially optimal ones, given for w1 = w2 = 0. So
total welfare decreases.
The intuition for these results is that distribution by retailers with
exclusive territories allows manufacturers to obtain a retail price which
is higher, and therefore closer to the joint-profit maximising level, than
if retailers faced intra-brand competition. Quantities sold, and therefore
welfare, are lower than in the absence of exclusive territories. By suitable
choice of franchise fee, the manufacturers can appropriate the resulting
extra profit for themselves.
In this model, then, exclusive territories restrict competition among
manufacturers. The model can be extended in two ways: (1) retailer i may
not be able to observe wj, or (2) fixed fees cannot be used, say because of
arbitrage. In both cases the basic results still apply if b is high enough, i.e.
if the products are close substitutes.
Policy towards vertical restraints

The economic analysis of vertical restraints suggests that they can be used
for different purposes and they can have ambiguous welfare effects. The
empirical evidence supports this: there are case studies to support each of
the theoretical stories. In fact, one recent survey of empirical evidence on
vertical restraints (Cooper et al., 2005) has concluded that, on the whole, 4
The only exception
they appear to reduce prices and/or increase output. Most competition is resale price
laws acknowledge these ambiguities. As a result, most vertical restraints maintenance, which
are not subject to an outright prohibition, but are judged on a case by case is prohibited in many
basis: this is the ‘rule of reason’ approach.4 See Chapter 11 of this subject competition laws.
guide for more on policy towards vertical restraints.
109
A case study
To gain further insight into the implications of vertical restraints and the
complexity of the issues involved in trying to assess their effect on welfare,
we focus now on a case study of a particular industry. The UK market for
new cars was investigated by the UK Monopolies and Mergers Commission
(MMC) in 1992 and again by its successor, the Competition Commission,
in 2000. The industry was relatively concentrated in 1992 (the combined
market share of the three largest producers was 55 per cent), but market
shares tended to fluctuate. The distribution system of the industry had,
as in many other countries, three important features, namely exclusive
dealing, exclusive territories and selective distribution. In particular:
• Dealers of any make of cars were allocated territories in which they
had primary responsibility for selling and servicing cars of that make.
Manufacturers could refuse to supply retailers other than authorised
dealers, and authorised dealers were not permitted to resell cars
except to other authorised dealers and final customers.
• A dealer was allowed to sell outside his territory but there were
restrictions regarding (among other things) advertising outside the
designated territory, the number of a manufacturer’s dealerships held
by a single dealer, and the volume of a manufacturer’s cars a dealer
could sell.
• Dealers were not allowed to sell competing makes at the same site,
and usually also within a designated territory. Sometimes there were
restrictions on selling competing makes even outside the designated
territory.
• Dealers had to satisfy minimum service and quality standards imposed
by manufacturers.
The MMC concluded that the vertical restraints were likely to have both an
efficiency effect and a market power effect. Overall, the MMC concluded
that profits of manufacturers and dealers did not appear excessive – and
that while prices were sometimes higher than in other European countries,
this could not be attributed to the vertical restraints.
The main arguments for and against vertical restraints in the new car
market can be summarised as follows:
• The system of exclusive territories and the refusal to supply retailers
other than authorised dealers, combined with the restrictions on sales
outside the designated territory, restrict intra-brand competition and
also prevent efficient dealers from expanding.
• On the other hand, the system of exclusive territories and the refusal
to supply retailers other than authorised dealers, combined with
the imposition of service and quality standards, encourage dealer
investment and provision of services.
• Exclusive dealing prevents consumers from directly comparing
competing makes, so it reduces inter-brand competition.
• On the other hand, since search costs are low relative to the cost of a
new car, consumers may visit many showrooms to make comparisons.
Also, economies of scope from retailing different makes may be limited
by the fact that the dealers also provide after-sale services. Finally,
there is an efficiency gain from exclusive dealing in that dealers
concentrate more effectively on selling the particular make; this
indirectly promotes inter-brand competition.
110
Note that the MMC seemed to be less concerned with the effects of
vertical restraints on manufacturers’ incentives and competition among
manufacturers than with the effects on dealers’ incentives and competition.
Accordingly, the main objective of the MMC’s recommendations was to
increase competition among dealers. In particular, the MMC recommended
that the system of exclusive territories and that of exclusive dealing
both stay in place, but a number of ancillary restrictions be eliminated,
namely the restriction on advertising outside the designated territories,
the restriction on the selling of competing makes within or outside the
territory, and the restriction on the volume of cars a dealer may sell.
Nearly 10 years later, the UK new car market was again subject to an
investigation by the Competition Commission, the successor of the MMC.
The Commission found that new car prices in the UK had been about 10
per cent higher than in Germany, France and Italy for similar makes and
models. It concluded that the main cause was the restrictive distribution
system operated by car manufacturers through their dealer networks. It
therefore recommended several changes in the system of car distribution,
most of which were accepted by the UK government. These included:
• prohibiting suppliers from discriminating in pricing between fleet
buyers and private buyers
• allowing dealers to advertise car prices other than recommended retail
prices for new cars
• prohibiting car manufacturers from refusing to sell cars to dealers who
sell below recommended prices
• ensuring that dealers can take advantage of cheaper prices of cars by
importing from other European countries without restrictions.
The report did not, however, recommend abolition of the selective and
exclusive distribution system operated by car manufacturers, since this
was at the time allowed by the European Union. Nevertheless, UK car
prices fell by more than 10 per cent in 2000 only and have fallen further
since then. In 2002 the EU system was modified: manufacturers may now
impose either selectivity or exclusivity, but not both, and multi-brand
dealerships are allowed. See also Activity 4 below.
Activities
1. A manufacturer produces a homogeneous good at constant unit cost c and sells to a
single retailer at price w. The retailer resells the good to final consumers at price p. No
services are provided. Explain why in this setup ‘double marginalisation’ will create
an inefficiency for the firms and will also be welfare-reducing for consumers. Describe
how double marginalisation can be eliminated.
2. In some industries, manufacturers operate their own distribution networks, marketing
their products directly to retail outlets. In others, manufacturers use independently
owned wholesalers or manufacturing representatives to market their products.
What factors are likely to influence the choice of distribution method for a particular
product? Explain.
3. ‘Resale price maintenance should be illegal.’ Do you agree? Justify your answer with
reference to economic theory and any relevant empirical evidence. (You may also
want to read some of the material in Chapter 11 of this guide before answering.)
4. Compare and contrast the two official reports on the UK car market and their
conclusions. How can one explain the differences in the conclusions of the reports?
If you were trying to defend the practices of the car manufacturers at the time of the
2000 investigation, what would your arguments be? Perhaps you would highlight
111
some consumer benefits from the vertical restraints and the potential harm in case
they were abolished. Or you might put forward alternative explanations for the price
differentials across European countries.
In order to ‘decide’ the case, you will need to know more about the European car
market and about EU policy toward the car industry. There are several studies on price
discrimination in the European car market, including some available on the internet.
Moreover, until 2002 the car industry in the EU was allowed to operate a selective
and exclusive distribution system: each manufacturer could select authorised dealers
who should satisfy minimum quality or service standards and grant them territorial
exclusivity. The system was modified in 2002: manufacturers may now impose
either selectivity or exclusivity, but not both; multi-brand dealerships with brand-
specific areas in the showroom are allowed; and various other restrictions have been
abolished. Almost all car manufacturers currently operate a selective system: they
can choose their authorised dealers but these are permitted to actively sell into other
territories. The UK, as part of the EU, has been affected by this legislation. The website
of the European Commission has a page on competition issues in the car industry,
with numerous links to various sources of information, including regulations, news,
reports and economic research: ec.europa.eu/competition/sectors/motor_vehicles/
overview_en.html

• analyse models of a vertical relationship between a manufacturer and
a retailer and show how vertical restraints can be used to eliminate
inefficiencies in such a relationship
• evaluate various efficiency arguments for vertical restraints in the light
of any relevant empirical evidence
• explain at least three different reasons why vertical restraints can
restrict competition
• use the theory on vertical restraints to discuss any relevant empirical
evidence and evaluate public policy in this area.

1. a. A monopoly manufacturer of a good sells to a monopoly retailer. The
consumers’ demand for the good is q = 1 – p, where q is quantity
sold and p is the final price. The retailer has zero cost and the
manufacturer’s cost function is C(q) = q2/2. The timing is as follows:
first the manufacturer chooses a tariff, and then the retailer chooses
the final price.
i. What are the manufacturer’s and the retailer’s profit under the
optimal linear tariff, T(q) = pwq, for the manufacturer?
ii. What are these profits under the optimal two-part tariff, T(q) =
A + pwq, for the manufacturer?
b. With the same setup as in part a, suppose now that before the
manufacturer chooses the tariff, the retailer can choose to invest
to increase demand. In particular, the retailer has two possible
choices: either spending amount ε (where ε is small and positive),
in which case the demand is increased from q = 1 – p to q = 2 – p,
or spending amount zero, in which case the demand remains q = 1
– p. The timing is as follows: first the retailer makes the investment
112
choice, then the manufacturer observes this choice and chooses a

tariff, and then the retailer chooses the final price.
iii. What is the retailer’s level of investment and the
manufacturer’s profit under a linear tariff?
iv. What is the retailer’s level of investment and the
manufacturer’s profit under a two-part tariff? Compare the
manufacturer’s profit under (i)–(iv) and provide intuition.
2. Discuss the view that vertical restraints can have ambiguous welfare
effects. Include in your answer an analysis of at least two different
arguments for the use of vertical restraints and a brief discussion of
any relevant empirical evidence.
113
Notes
114
Chapter 10: The determinants of market structure
Chapter 10: The determinants of market

structure
Introduction
This chapter examines the theory and empirical evidence on the
determinants of market structure. It attempts to shed some light on
questions like:
• Why are some industries more concentrated than others?
• What is the link between competition and concentration, and between
technology and concentration?
• Should we expect that economic integration will lead to lower
concentration in the international economy than in national
economies?
• What is the role and the limitations of policies that influence the level
of concentration, such as merger policy?
This chapter describes a comprehensive framework for the analysis
of market structure, which is associated with the so-called bounds
approach to market structure. This approach has set out to identify a
number of economic mechanisms that hold across industries or for a class
of industries. It aims therefore to provide results with broad application.
However, market structure in any particular industry also depends on
some factors that are either not observable or not systematic. This means
that for any given set of observable and systematic industry characteristics,
there is a multiplicity of possible outcomes. The bounds approach
essentially consists of distinguishing between those outcomes that are
possible or likely and those that are impossible or unlikely. Thus it turns
out that we cannot generally predict the actual level of concentration in
an industry on the basis of its observable and systematic characteristics,
because this will also depend on unobservable features of the industry as
well as its history, which is to some extent the product of chance. However,
we can predict a ‘lower bound’ to concentration (i.e. its minimum possible
level). The economic mechanisms that determine this lower bound turn
out to be quite general.
The analysis of the determinants of concentration in this chapter will
also clarify the way market structure interacts with firm conduct. This
has always been an issue of great interest for economists and a major
concern for policy. An old idea in industrial economics, which is at the
heart of the ‘Structure–Conduct–Performance (S–C–P) paradigm’, is
that higher concentration causes profitability to increase in an industry
because it increases the market power of firms and may facilitate
collusion. Moreover, profits are sustained through ‘barriers to entry’ such
as economies of scale, product differentiation through advertising or R&D,
and so on. The empirical evidence on this view has been mixed. Also, the
theory itself has been subject to several criticisms.
One such criticism – the so-called ‘Demsetz critique’ – is that high
profitability in an industry may be due to large efficiency differences
across firms: as the more efficient firms get bigger and make high profits,
concentration increases at the same time – but it is itself enogenous and
it is not the cause of high profitability, even though the two variables are
positively correlated. A policy implication follows: high concentration and
115
high profitability need not be a cause of concern if only efficient firms have
high market shares and make positive net profits.
Another criticism is that not only is concentration endogenous, but it is
even affected by firm conduct, for instance, the intensity of competition
in a market and firms’ strategies regarding advertising or R&D. While
some versions of the S–P–C paradigm have recognised the existence of
these feedback effects, the interaction between firm conduct and market
structure cannot be properly understood and modelled without using
multi-stage games that emphasise the distinction between firms’ ‘long-run’
and ‘short-run’ decisions. This distinction is an important element in the
analysis of the determinants of market structure, as you will see below.
Learning outcomes
• explain the notions of exogenous and endogenous sunk cost industries
• analyse the determinants of market structure in exogenous and
endogenous sunk cost industries
• compare the theory of market structure with the S–C–P paradigm and
explain why concentration may not be a good predictor of the state of
competition in an industry
• describe the empirical evidence on the determinants of market structure.
Essential reading
Sutton, J. Sunk costs and market structure. Various chapters.
Sutton, J. ‘Market structure: theory and evidence’, in Armstrong, M. and R.
North-Holland, 2007) Working paper version available at: http://personal.
lse.ac.uk/sutton/market_structure_theory_evidence.pdf
Further reading
Books
2000) Chapters 9 and 14.
Journals
Sutton, J. ‘Technology and market structure’, European Economic Review 40
1996, pp.511–30.
Symeonidis, G. ‘Price competition and market structure: the impact of cartel
policy on concentration in the UK’, Journal of Industrial Economics 48(1)
2000, pp.1–26.
Basic concepts
Sutton (1991, 1996, 2007) has proposed and developed the bounds
approach to market structure and has produced many of the theoretical
and empirical results discussed in this chapter. A basic distinction in his
analysis of market structure is between exogenous sunk cost industries and
endogenous sunk cost industries.
116
In exogenous sunk cost industries, the only significant sunk costs

are the costs of setting up a plant of minimum efficient scale, and these
are exogenously determined by technology. You should see the notion
of an exogenous sunk cost industry as a useful approximation which
captures two things. First, the fact that firms must operate at a level
close to minimum efficient scale in order to be able to compete – it is for
this reason that setup costs are said to be exogenously determined by
technology. Second, the fact that in many industries there is limited scope
for advertising or research and development (R&D), that is for investment
that enhances brand image or product quality or reduces marginal cost.
Thus the exogenous sunk cost category comprises homogeneous good
industries as well as industries characterised by horizontal product
differentiation (including spatial differentiation).
In endogenous sunk cost industries, on the other hand, there
are, in addition to the exogenously determined sunk cost of entry,
endogenous sunk costs, such as advertising or R&D. These costs are
fixed, they are sunk before any price decisions are taken by the firms,
and they are endogenous because they can vary depending on each firm’s
own decisions. The endogenous sunk cost category comprises industries
characterised by vertical product differentiation as well as homogeneous
good industries in which there is scope for cost-reducing R&D.
The notion that certain costs are sunk is closely linked to the distinction
between long-run and short-run decision variables. Entry, advertising and
R&D are long-run choices, while pricing behaviour is a short-run decision.
This provides a way of unravelling the two-way link between firm conduct
and market structure through the use of a multi-stage game. In the short
run, market structure and long-run investments are given and we analyse
short-run conduct (i.e. pricing behaviour). In the long run, we analyse
long-run choices and the determination of market structure.
Theory of market structure in exogenous sunk cost

industries
Let us start with the benchmark case of a homogeneous good and assume
further that each firm operates a single plant. We model competition in
such an industry as a two-stage game. At stage 1 firms simultaneously
decide whether or not to enter at an exogenously fixed setup cost f. At
stage 2 they simultaneously set prices or quantities. Assume also that the
firms are symmetric.
At stage 2, we solve for the Nash equilibrium, given the number of firms N.
Denote the stage 2 equilibrium (gross) profit of firm i ∏i(N,S,t), where N is
the number of firms that have entered at stage 1, S is market size
(assumed to be exogenous), and t is the ‘intensity of price competition’:
this depends on firms’ pricing strategies,1 which in turn partly depend on 1
See the Folk theorem
exogenous institutional factors, such as the climate of competition policy mentioned in Chapter 5.
and the degree of economic integration.2 Standard assumptions are: 2
‘Price competition’
∂∏i/∂S > 0, ∂∏i/∂N < 0, and ∂∏i/∂t < 0. The second of these is the main is here equivalent to
‘short-run competition’,
prediction of the S–C–P approach; in other words the present analysis
whether firms set prices
embodies the insights of that approach. The third inequality says that or quantities.
more intense price competition lowers gross profit for given N.
At stage 1, N is determined by the free entry zero-profit condition
∏i(N,S,t) = f. In other words, firms will enter up to the point where the
cost of entry is equal to the profit they expect to make at stage 2. Taking
the total differential of the zero-profit condition, we obtain d∏i =
(∂∏i/∂S)dS + (∂∏i/∂N)dN + (∂∏i/∂t)dt = df. From this we obtain:
117
• dN*/dS = – (∂∏i/∂S)/(∂∏i/∂N) > 0

• dN*/dt = – (∂∏i/∂t)/(∂∏i/∂N) < 0
• dN*/df = 1/(∂∏i/∂N) < 0,
where N* is the long-run equilibrium number of firms. These relationships
are illustrated in Figure 10.1.
1
N
increase in t
S
f
Figure 10.1
The intuition for these results is straightforward. Take, for instance dN*/
dS > 0. As S increases, given N, ∏i rises. So starting from an equilibrium of
the two-stage game where each firm makes zero net profit, we now have
∏i > f. Hence entry will occur, which will cause ∏i to fall (since ∂∏i/∂N <
0), until we have again ∏i = f. At the new equilibrium N* will be larger.
The intuition is similar for dN*/dt < 0 and dN*/df < 0.
Activity
As mentioned in Chapter 4 of this guide, the Cournot model can be interpreted as
representing less intense competition than the Bertrand model in games where firms do
not make any long-run decisions other than the decision to enter the market or not. The
present analysis predicts therefore that the equilibrium number of firms N* in a two-stage
game with free entry at cost f > 0 at stage 1 will be smaller under price-setting than
under quantity-setting at stage 2. Prove this for the case where demand is given by q =
a – p and all firms have the same constant marginal cost c. Conclude that tougher short-
run competition raises concentration: dN*/dt < 0.
Answer
If stage 2 competition is in quantities, you already know from Activity 2 in Chapter 4 that
stage 2 equilibrium profit is ∏ = (a – c)/(N + 1)2, where N is the number of firms that
have entered at stage 1. With free entry at stage 1, we get ∏ = (a – c)2/(N + 1)2 = f,
which can be solved for N to give N* = [(a – c)/f1/2] – 1. So, in general, we have N* > 1
when firms set quantities.3
3
N* cannot take any
values less than 1, since
If stage 2 competition is in prices, then there are two possibilities for the second stage. this does not make
Either only one firm has entered at stage 1, in which case it makes monopoly profit ∏M sense. And it is equal to
at stage 2, or more than one firms have entered, in which case each of them makes zero 1 only in a very special
case.
profit at stage 2. Now go back to stage 1. If only one firm enters at stage 1, it will be
making overall (net) profit ∏M – f, while if more firms enter, each of them will be making
negative overall profit (a loss equal to the entry cost f). So there exists no subgame-
perfect equilibrium of the two-stage price-setting game with more than one firms
entering; the only subgame-perfect equilibrium outcome is N* = 1.
Next, consider the case of a horizontally differentiated product. In this

case, each firm can in general produce one or more varieties of the
118
product. This includes the case of spatial differentiation where each firm
can operate in one or more locations. There exists now a multiplicity of
equilibria: the most fragmented equilibrium is the one where each firm
produces one variety (operates in one location), while various more
concentrated equilibria, symmetric or asymmetric, can occur if firms
produce more than one variety (operate in more than one location). This
is illustrated in Figure 10.2.
Thus there is a lower bound to concentration which declines indefinitely
as the ratio of market size to setup cost S/f increases and shifts up as the
intensity of price competition t rises. The actual level of concentration can
be anywhere above the bound.
concentration
increase in t
S
f
Figure 10.2
A multiplicity of equilibria and therefore a lower bound to concentration is
possible (or even likely) not only in a horizontally differentiated but also
in a homogeneous good industry, because of the potential for asymmetries
(in costs, number of plants, or capacity) across firms. Whether we will
get a more fragmented or a more concentrated structure for given values
of S, f and t depends on various factors, including chance. For instance,
economies of scope in setup costs favour more concentrated equilibria.
Also, strategic asymmetries such as first-mover advantages favour more
concentrated equilibria – recall our discussion of brand proliferation.
The positive effect of the intensity of price competition on market
concentration that emerges from the theory of market structure has
important implications. Traditionally, economists have regarded
concentrated markets as markets where firms make high profits at the
expense of consumers: high concentration was thought to imply a low
degree of competition and high profitability. This view is not incorrect, but
it relies on assuming that incumbent firms can deter the entry of potential
rivals. With free (although not costless) entry, net profits will be relatively
low and tougher competition among firms will tend to increase the level of
concentration in a market.
Theory of market structure in endogenous sunk cost

industries
Competition in an endogenous sunk cost industry can be modelled as a
three-stage game. At stage 1 firms simultaneously decide whether or not to
enter at an exogenously fixed setup cost f. At stage 2, given the number of
firms N that have entered, each firm chooses to incur an endogenous sunk
cost Ai (advertising or R&D), which increases the consumers’ willingness to
pay for the firm’s product or reduces the firm’s marginal cost. At stage 3,
taking N and the Ai as given, firms simultaneously set prices or quantities.
119
A central assumption of this model is that Ai is a fixed cost which is

incurred prior to the price or quantity competition stage. In particular, a
higher level of advertising or R&D is not associated with any significant
increase in the marginal cost of the firm.
An equilibrium in this three-stage game (for the symmetric case) is a pair
(N*, A*) defined by two necessary conditions:
• the free entry condition ∏i(N,S,t,Ai,A-i) = f + Ai
• the first order condition for the choice of Ai: ∂∏i/∂Ai = 1 for all i,
where Πi denotes the equilibrium stage 3 profit, for given N and Ai. The
second condition says that each firm spends on advertising/R&D up to the
point where the cost of an extra unit of advertising/R&D is equal to the
gross profit it creates at stage 3 of the game.
It can be shown that for a broad class of models – robust to strategic
asymmetries, price-setting versus quantity-setting, single-product versus
multi-product firms – the following results emerge (see Figure 10.3):
concentration
S
Figure 10.3
• Non-convergence: the lower bound to concentration does not converge
to zero as market size S increases. That is, there is a minimum level of
concentration no matter how large the market becomes.
• Non-monotonicity: minimum concentration may rise or fall as S
increases.
The intuition is as follows. If S is below a threshold market size,
advertising/R&D is not profitable since the gross profit generated is not
enough to cover the fixed cost Ai. So for S below the threshold, there is
no advertising/R&D and minimum concentration declines as S increases.
For S above the threshold, advertising/R&D begins and the above results
(non-convergence and non-monotonicity) apply. The reason we get these
results is that as S increases, firms have an incentive to spend more on
advertising/R&D. Consider then the free-entry condition ∏i = f + Ai. As S
increases, ∏i rises, but at the same time Ai rises as well. If the right-hand
side increases by more (less) than the left-hand side, given the initial
number of firms N, then N must fall (rise) to re-establish the equality ∏i =
f + Ai. In any case, the escalation of advertising/R&D spending prevents
the emergence of a fragmented industry structure.
There are some further results for R&D-intensive industries in particular.
In these industries there is a trade-off between spending on R&D to
enhance the quality of an existing product and spending to develop a new
product. Whether firms opt for one or the other of these strategies depends
to a large extent on:
120
• The cost of achieving quality improvements on an existing product

(a technological characteristic that is related to the notion of
‘technological opportunity’) and the consumers’ willingness to
pay for these improvements (a demand characteristic). These
together determine the returns from spending to enhance technical
performance or quality.
• The degree of horizontal product differentiation (a demand
characteristic) – that is, the degree to which demand is fragmented
due to the existence of a variety of customer tastes or requirements.
This is a measure of the returns from introducing a new product.
In industries where there are relatively high returns from product
improvement and demand is not fragmented, firms will choose a product
enhancement strategy. These industries will therefore have both high R&D
intensity and high concentration. Firms will invest heavily on R&D in just
a few products. To cover their large R&D costs, they will need to produce a
lot of output, so there will be room for only a few firms.
On the other hand, in industries where there are relatively low returns
from product improvement and demand is fragmented, firms will
choose a product proliferation strategy. These industries will have
high R&D intensity but relatively low concentration. The possibility to
develop varieties that match the needs of particular buyers will lead to a
proliferation of varieties and will attract entry.
Empirical evidence
The empirical evidence comes both from cross-industry studies and from
case studies of particular industries. We focus here on cross-industry
results.
Symeonidis (2000, 2002) provides evidence on the price competition
effect on concentration. There was a major shift in UK competition
policy in the late 1950s, which led to the abandonment of price-fixing
agreements among firms. This ‘natural experiment’ makes it possible to
compare the evolution of concentration in the group of industries affected
by the legal changes, which experienced an increase in the intensity of
price competition, with the evolution of concentration in a ‘control’ group
of industries in which no restrictive agreements had existed.
The econometric analysis using panel data for about 130 industries
over 20 years finds a strong positive effect of price competition on
concentration in both exogenous and endogenous sunk cost industries.
In particular, the change in the competition regime caused the five-firm
concentration ratio (the sum of the market shares of the five largest firms)
to increase, on average, by about six percentage points in previously
collusive industries.
What happened is that as the intensity of price competition (the parameter
t) increased following the abolition of cartels, prices and therefore Πi fell
in the short run. Advertising fell slightly, while innovative activity was not
significantly affected. So firms, especially less efficient firms, could no
longer cover their fixed costs. There was exit and mergers until the fall
in N* caused prices to rise and restored Πi so that fixed costs could again
be covered. In fact, the abolition of cartels in the UK did not have any
significant effect on industry profits in the long run.
High market concentration has traditionally been seen by policy makers
as a cause of concern, since it has been thought to facilitate the abuse
121
of market power by firms. An important policy implication of the results

in Symeonidis (2000, 2002) is that, under free entry and exit, higher
concentration need not be associated with less price competition and
lower welfare – the opposite may sometimes be the case, as more intense
competition reduces profit margins and the number of firms that can
survive in an industry. So competition policy should perhaps be less
concerned with concentration than with ensuring that competition among
firms is ‘effective’, that is firms do not collude or otherwise abuse their
market power and there are no barriers to entry.
Sutton (1991) provides evidence on the market size–concentration
relationship using data on 20 food and drink industries for six countries.
Six of the industries were exogenous sunk cost industries (H) and 14 were
advertising-intensive (A). Sutton applies two types of econometric tests.
The first test is to estimate the lower bound to concentration separately
for the two groups of industries, using non-standard econometrics. The
estimates imply that as S/f → ∞, the four-firm concentration ratio C4
approaches 19 per cent for the A group (non-convergence) and six per
cent for the H group (convergence). Also, the estimated lower bound
for the A group is not just an upward shift of the bound for the H group,
which supports the view that advertising is an endogenous sunk cost, in
contrast with the older view of advertising as an exogenous barrier to
entry.
The second test is to run a standard regression of C4 on ln(S/f ) (and
industry and country dummy variables) separately for the two groups of
industries. The theory predicts a negative relationship for the H group and
no clear relationship for the A group. This is confirmed by the empirical
results.4
4
See Sutton (1991),
Chapter 5, for details.
Sutton (1996) presents empirical results on R&D-intensive industries.
He tests the following theoretical prediction: as the number of ‘submarkets’
or ‘technological trajectories’ within an R&D-intensive industry rises,
suggesting that product proliferation increasingly dominates product
enhancement, concentration is expected to fall. But in low-R&D industries
there should be no link between the number of submarkets and
concentration. Using data on two groups of US industries, those with R&D-
sales ratio higher than four per cent and those with R&D-sales ratio lower
than one per cent, Sutton obtains results that confirm the theory.5 5
Sutton’s more recent
work has addressed
The bounds approach to market structure has implications for competition issues such as the size
policy. It suggests that there are constraints in the use of merger and distribution of firms in
antitrust policies. In particular, such policies cannot be used to impose a an industry and market
market structure below the lower bound, as this is not sustainable. On share dynamics.
the other hand, competition policy can be used above the lower bound to
influence concentration in an industry. Merger and antitrust policies are
discussed in more detail in the next chapter of this guide.
Activities
1. The competition law in the country of Wonderland prohibits all mergers that result
in a single firm having more than 75 per cent of any market. Do you think the law
makes sense? Why or why not? What would your advice be to a new government
that proposes to make changes to the law?
2. The aircraft industry and the instrument industry are both technologically progressive,
with high levels of R&D intensity. However, the former is much more concentrated
than the latter. Do you think that this difference can be explained on the basis of
technological and demand characteristics of these industries?
122
3. Suppose there are three identical firms thinking about entering a market in which
there is no incumbent firm. There is a small but positive cost of entry F. The product
is homogeneous, with inverse demand p = 1 – Q, where Q is aggregate quantity.
Unit cost is zero. How many firms will enter if (i) competition post-entry is in prices,
(ii) competition post-entry is in quantities and F = 1/20, and (iii) competition post-
entry is in quantities and F = 1/10? Provide intuition for your results.
4. ‘When collusion among firms breaks down in an industry, concentration is expected
to rise’. Under what conditions (if any) is this statement true? Justify your answer with
reference to economic theory and any relevant empirical evidence.
5. Sutton (2007) writes: ‘One reason for [the] continuing interest in market structure is
that this is one of the few areas in economics where we encounter strong and sharp
empirical regularities arising over a cross-section of industries. That such regularities
appear in spite of the fact than every industry has many idiosyncratic features
suggests that they are moulded by some highly robust competitive mechanisms.’
What are the empirical regularities and the robust mechanisms Sutton refers to? Do
you agree with his statement above? What are, on the other hand, the factors behind
the idiosyncratic features of every industry and how important are they? You will need
to read both the theory on the determinants of market structure and (especially) some
of the empirical evidence, which is of two kinds: econometric studies that focus on
the general mechanisms, and case studies that sometimes provide additional evidence
on the working of these mechanisms and sometimes shed light on particular features
of specific industries.
Studies by Sutton himself and by Symeonidis are good starting points, as is the work
of B. Lyons, S. Davies, C. Matraves and others on economic integration and industrial
market structure in the European Union. Recent empirical papers on market structure
include Ellickson (2007) and Berry and Waldfogel (2010). You can also find reviews
of Sutton’s two books Sunk costs and market structure and Technology and market
structure by well-known economists such as T. Bresnahan, R. Schmalensee, F.M.
Scherer and others.

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• explain the notions of exogenous and endogenous sunk cost industries
• analyse the determinants of market structure in exogenous and
endogenous sunk cost industries
• compare the theory of market structure with the S–C–P paradigm and
explain why concentration may not be a good predictor of the state of
competition in an industry
• describe the empirical evidence on the determinants of market
structure.

1. Consider the following two-stage game. There is a large number of
potential entrants in a market for a homogeneous product. At stage
1 firms simultaneously decide whether or not to enter at a cost of
entry f. At stage 2 those firms that have entered simultaneously set
quantities. The demand function is given by Q = S/p, where Q is total
quantity produced and S is total expenditure on the product (here a
measure of market size, and hence exogenously fixed). The marginal
cost is constant and equal to c for all firms.
123
a. Derive the stage 2 Cournot–Nash equilibrium.

b. Derive the stage 2 equilibrium price, profit for firm i and industry
profit, and show that all three are decreasing in the number of
firms N.
c. Compute the long-run equilibrium number of firms N* and show
that it increases in S and decreases in f. Provide some intuition for
these results.
d. Now suppose that at stage 2 firms collude to raise prices above the
Cournot–Nash price that you computed in part b. Do you expect
the long-run equilibrium number of firms NCOLL to be higher or
lower than N*? Explain.
2. Describe the theory of market structure in ‘endogenous sunk cost
industries’. Then describe alternative approaches to testing this theory
and discuss the empirical evidence on the theory.
124
Chapter 11: Competition and industrial policy
Chapter 11: Competition and industrial

policy
Introduction
Government policy towards industry is necessary because firms may
behave in ways that reduce social welfare or may be prevented from
behaving in ways that would benefit society. Government policies fall into
various categories. Broadly, the role of competition policy – or antitrust
policy, as it is called in the USA – is to design and enforce a set of rules
which ensure that competition is ‘effective’; in other words, competing
firms behave in a way that results in the best possible allocation of
resources. The objective is to improve efficiency, which of course here
refers to society as a whole – so the appropriate measure is the sum of
producers’ and consumers’ surplus. Much of industrial policy is concerned
with promoting the best possible intertemporal allocation of resources,
through the provision of incentives to firms to undertake appropriate
investments in circumstances where market imperfections prevent them
from doing so. Regulation is the subject of the next chapter.
This chapter is an introduction to competition policy and industrial policy
towards R&D, with an emphasis on the economic issues rather than on
the legislative and institutional frameworks in different countries. You
should consult the readings given above for more details on how these
policies operate in various countries. In the case of competition policy,
these readings also contain extensive discussion of particular cases; going
through these cases is essential for appreciating the complexities involved
in the design and implementation of competition policy.1 1
Especially good in this
respect are Carlton and
Perloff (2005), Church
Learning outcomes and Ware (2000), Kwoka
and White (2013) and
By the end of this chapter, and having completed the Essential reading and Lyons (2009).
Activities, you should be able to:
• explain the rationale for competition policy and for industrial policy
towards R&D
• describe the difficulties and dilemmas in competition policy with
respect to mergers, restrictive agreements, and abuses of market
power
• discuss the practice of competition policy
• describe and evaluate the main forms of government policy towards
R&D.
Essential reading
Church, J.R. and R. Ware, Industrial organization: a strategic approach. Various
chapters.
Further reading
Books
Pearson Addison Wesley, 2005).
125
Geroski, P. ‘Markets for technology: Knowledge, innovation and appropriability’,

in Stoneman, P. (ed.) Handbook of the economics of innovation and
technological change. (Oxford: Blackwell, 1995).
Kwoka, J.E. and L.J. White (eds) The antitrust revolution: economics,
competition, and policy. (Oxford: Oxford University Press, 2013).
University Press, 2009).
Mowery, D. ‘The practice of technology policy’, in Stoneman, P. (ed.) Handbook
of the economics of innovation and technological change. (Oxford: Blackwell,
1995).
Whinston, M. ‘Antitrust policy toward horizontal mergers’, in Armstrong,
M. and R. Porter (eds) Handbook of industrial organization, Volume 3.
Journals
Carlton, D.W. ‘Does antitrust need to be modernized?’, Journal of Economic
Perspectives 21(3) 2007, pp.155–76.
Giovannetti E. and L. Magazzini ‘Resale price maintenance: an empirical
analysis of UK firms’ compliance’, Economic Journal 123(572) 2013,
pp.F582–F585.
Hall, B.H. ‘The assessment: Technology policy’, Oxford Review of Economic Policy
18(1), 2002, pp.1–9. (Special issue: Technology policy).
Hall, B.H., C. Helmers, M. Rogers and V. Sena ‘The importance (or not) of
patents to UK firms’, NBER Working Paper 19089 (2013).
Competition policy: objectives and difficulties in design

and implementation
The principal goal of competition policy is to set rules for competition
among firms so as to achieve improved market efficiency. There are,
however, three distinct aspects of efficiency. The first is static allocative
efficiency. Thus a central concern of competition policy authorities is
the extent to which firms can exercise market power by pricing above
marginal cost; this is because, as is well known, pricing above marginal
cost results in a ‘deadweight loss’ for society. The second aspect is static
productive efficiency (i.e. the internal efficiency of firms). And the third
aspect is dynamic efficiency, allocative or productive.
There are several reasons why the design and implementation of
competition policy are far from straightforward.
• Firm conduct which may improve one aspect of efficiency may also
worsen another. For example, a horizontal merger between two
firms may result in higher price-cost margins, which is bad for static
allocative efficiency, but it may also lead to lower current costs,
which is good for productive efficiency, or lower future costs or better
products through increased investment in R&D, which is good for
dynamic efficiency.
• In many cases, the effect of a particular business practice on welfare
may not be clear, even when only one aspect of efficiency, say static
allocative efficiency, is considered. Recall, for instance, our discussion
of the welfare effects of price discrimination and vertical restraints in
previous chapters of this guide.
• Some practices, such as tacit collusion, are difficult to detect, thus
making the implementation of policy difficult even in cases where
the welfare implications of firm conduct are thought to be relatively
unambiguous.
126
• It is not always easy to identify the degree of market power which is

acceptable in a particular industry, given the technological and other
constraints faced by firms in the industry. Some degree of market
power is necessary under conditions of increasing returns to scale:
price has to be higher than marginal cost, so that firms can cover their
fixed costs, including investment and research costs. Moreover, a firm
may obtain market power because it is efficient or innovative, which
surely cannot be bad in itself. Recall our discussion in Chapter 9 of
the constraints on competition policy imposed by the endogeneity of
market structure.
• The efficiency goal has not been the only objective of competition
policy in practice. In the United States, for instance, there has been
a debate between proponents of the efficiency approach and those
arguing for a competition policy that focuses more specifically on
consumer welfare or even a policy with wider social and political
objectives, including the protection of small firms and the dispersion
of economic power. Competition policy in the European Union has
traditionally put specific emphasis on consumer welfare as well
as on overall efficiency, and has also served to promote economic
integration.
• Conflicts may also arise between competition policy and other
government policies, such as industrial or trade policy. In Japan, for
instance, competition policy has traditionally been subsidiary to other
concerns, such as the creation of strong domestic firms that are able
to compete with foreign firms in some sectors, or the rationalisation
of production through government-sanctioned cartel arrangements in
other sectors. The wish to create dominant firms able to compete with
non-EU firms has often influenced EU policy as well.
A central issue in most competition cases is the assessment of the
market power of the firms involved, or the assessment of the change in
market power which has been or may be brought about by a certain action
of the firms. This may be complicated because of difficulties in defining
the relevant product or geographical market, the absence of precise
information about firms’ costs and perhaps also about demand parameters,
and the need to take into account market dynamics that may influence the
extent to which market power is likely to persist in time. These difficulties
are particularly severe in research-intensive industries which are subject
to rapid and constant change. Thus when assessing the degree of market
power, the competition authorities are likely to consider several direct or
indirect indices, none of which is conclusive in itself. For instance:
• the market share of the firm or firms involved in the relevant market
• the volatility of market shares in the industry (since high volatility may
indicate low persistence of market power)
• any evidence on price–cost margins and/or the profitability of the
firms
• estimates of the elasticity of demand for the product in question (since
high elasticity suggests low market power)
• estimates of the degree of substitution between the product in question
and other products
• the extent to which competition operates at a local, national or
international level
• the degree of actual or potential competition from imports
127
• the significance of entry barriers, including those created by sunk

investments in capacity, by technological superiority, by brand image,
or by vertical relationships
• the bargaining power of buyers.
Competition policy in practice

Merger policy
With merger policy the competition authorities can influence market
structure in particular industries by preventing certain mergers between
firms from taking place. The rationale for merger policy is that a
consequence of many mergers is the creation of additional market power
– which reduces welfare, at least in the absence of significant fixed costs.
It is easy to verify this in the context of a symmetric Cournot model: a fall
in the number of firms in an industry increases price and reduces output
and consumer surplus. Even though industry profit rises after the merger,
total welfare falls. Note, however, that this is not necessarily true in the
presence of fixed costs.
Activity
Consider a Cournot homogeneous good oligopoly with N firms and inverse demand
function P = a − Q, where Q is the industry output. Each firm has marginal cost c < a
and zero fixed cost. Derive the Cournot-Nash equilibrium and show that industry profit
decreases in N, but consumer surplus and total social welfare increase in N. Conclude
that a horizontal merger in this market (a fall in N) raises industry profit but reduces
welfare.
Answer
The derivation is straightforward. At the Cournot-Nash equilibrium each firm produces
quantity q* = (a – c)/(N + 1). Each firm makes profit (a – c)2/(N + 1)2. Industry profit
is given by N(a – c)2/(N + 1)2 and decreases in N. Industry output N(a – c)/(N + 1),
consumer surplus N2(a – c)2/2(N + 1)2 and total welfare all increase in N.
Note that a merger in a symmetric Cournot model results in lower profit

for the merged firm than the joint profit of the two separate firms before
the merger. The reason is that the reduction in output of the merged firm
induces all other firms in the industry to raise their output. This fall in
profit is clearly a paradox: why would then firms merge? The paradox can
be resolved in models where firms are capacity constrained, or compete in
prices rather than quantities, or benefit from cost reductions following the
merger.
Despite its limitations, the symmetric Cournot model illustrates the
‘unilateral effects’ of horizontal mergers (i.e. mergers of firms in the
same industry): the change in market shares affects prices and welfare.
Horizontal mergers can also have ‘coordinated effects’ – these refer to
how a fall in firm numbers influences the ability of firms to collude. In
the case of vertical mergers (i.e. mergers of firms in vertically related
industries), the primary concern is the possibility that vertical integration
hinders the access of non-integrated firms to outlets or sources of supply.
However, a large number of mergers do not have any significant effect on
competition. To ensure that competition authorities are not occupied with
such mergers, competition laws typically specify criteria that a merger
must satisfy in order for it to be contestable. In particular, a horizontal
128
merger would not be contestable if the level of concentration and/or the

market shares after the merger or the change in concentration due to
the merger are not higher than specified thresholds. Note that applying
these criteria requires defining the relevant market – which may not be
straightforward.
For mergers that are in principle contestable, the next step is to examine
what are the likely effects on competition. This involves assessing the
market power of the firms involved as well as changes in market power
brought about by the merger. In the case of horizontal mergers, the
primary concern is the rise in market share and in industry concentration.
Barriers to entry are an important consideration, because, if these are low,
then a merger by two firms may lead to the entry of another firm into the
industry – and this would (partially) counteract the competition effects of
the merger.
Moreover, even when a merger raises significant market power issues,
the competition authorities must examine whether the merger should be
allowed, despite its effect on competition. This may sometimes be justified
because of countervailing benefits, such as current or future efficiency
gains, that more than compensate for the static welfare losses. These
countervailing benefits can take several different forms: a lower marginal
or average cost for the merged firm because of ‘synergies’ or economies of
scale, increased ‘buyer power’ of the merged firm that allows a reduction
in purchasing costs at the expense of suppliers, improved R&D capabilities,
elimination of a vertical externality such as double marginalisation or the
underprovision of services (in the case of a vertical merger), and so on.
Activity
Consider a Cournot homogeneous good oligopoly with three firms and inverse demand
function P = 1 − Q, where Q is the total quantity produced. Marginal costs are c1 =
0.1, c2 = 0.35 and c3 = 0.4. There are no fixed costs. A merger between firms 2 and 3 is
proposed. The merged firm will have marginal cost c2 = 0.35. Show that this merger will
reduce industry output but will also increase total welfare. Explain.
Answer
The derivation is straightforward. At the pre-merger Cournot-Nash equilibrium, q1 = 0.36,
q2 = 0.11 and q3 = 0.06. Industry output is Q = 0.54 and total welfare is W = 0.29. After
the merger, q1 = 0.38, q2 = 0.13, Q = 0.51 and W = 0.30. The intuition is that the more
efficient firm (firm 1) expands output after the merger, while the combined output of the
less efficient firms falls. Since c1 is considerably lower than c2, the average cost of the
industry falls significantly and this causes industry profit to rise sufficiently to more than
compensate for the fall in consumer surplus.
In practice, both in the US and in the EU, there has been a certain reluctance
to clear mergers on efficiency grounds, and several European countries
which had favoured the creation of ‘national champions’ during the 1960s
and 1970s have since adopted a tougher attitude towards mergers.
Finally, exogenous changes in market conditions may sometimes result in
a necessary adjustment of market structure. For instance, some horizontal
mergers that lead to a more concentrated industry structure are made
necessary by a decline in demand or the intensification of competition
caused by stricter competition laws or economic integration. Thus under
the US and EU legislations a merger can be allowed if it can be shown
that one of the firms will otherwise go out of business (the ‘failing firm
defence’).
129
Econometric techniques are sometimes used to simulate merger effects.

This involves an analysis in some ways similar to the one presented in
Chapter 5 of this guide for measuring the degree of market power of firms
in an industry – and is therefore subject to the same limitations. Demand,
cost and behavioural parameters are estimated from pre-merger prices and
quantities and are used – together with various assumptions about how
some of these may change following the merger – to predict the post-
merger equilibrium and evaluate the welfare effect of the merger.
Policy towards restrictive practices

Horizontal agreements among firms to fix prices or output levels or to
allocate customers or geographical areas are typically detrimental for
competition and not defensible on efficiency grounds. Thus they are
generally prohibited in most competition laws, including those of the EU
and the USA.
There are also some exceptions, the most important of which concerns
agreements to cooperate in research and development.2 In the EU, some 2
See the section on
agreements for cooperative joint ventures in the production and policy toward R&D for
a discussion of research
distribution of goods have also been allowed on efficiency grounds. In
joint ventures.
some countries, such as Germany or Japan, cartels that have the purpose
of reducing capacity in declining industries have been allowed.
The main difficulty regarding horizontal restrictive practices is detection.
As was explained in Chapter 5 of this guide, it is not easy to distinguish
between collusive and non-collusive behaviour on the basis of the data
typically available to competition authorities. For instance, parallel
pricing – the similarity of prices and price changes – is not sufficient for
concluding that collusion exists. As a response to the detection problem,
competition authorities have also focused on secondary agreements that
may facilitate collusion, such as agreements among firms to exchange
information on prices, costs, etc. However, information exchanges may
also improve the knowledge of market conditions and thus promote
competition. A distinction sometimes applied is between exchanges of
individual firm data and dissemination of aggregate industry data; only
the former are suspected of being used to sustain collusion.
Policy towards the abuse of a dominant position

Competition authorities are not so much worried by the mere existence
of market power as by the possibility that firms which possess market
power abuse it, that is, use it in a way that eliminates, restricts or distorts
competition. Therefore the authorities will intervene to control firm
conduct whenever this is thought to be abusive. There are a variety of
business practices that may constitute abuses under different legislatures,
including ‘excessive’ prices, strategies that deter entry or expansion of
rivals, price discrimination, tying, predatory pricing, and vertical restraints.
Now some of these practices, such as price discrimination or vertical
restraints, can have ambiguous welfare effects. Others, such as introducing
new products or building capacity in anticipation of a rise in demand,
can be legitimate competitive actions even if they also deter entry. And
still others, such as predatory pricing, are definitely welfare-reducing and
hence illegal, but are also difficult to detect. Most legislatures therefore
recognise the need for a detailed investigation when it comes to assessing
whether a firm or group of firms have abused their market power. This is
the so called rule of reason approach. Such an investigation involves
establishing the existence of market power and then examining whether
there has been an abuse.
130
Nevertheless, some legislatures take a tough stance against some

particular practices. One such case is resale price maintenance (RPM).
This is illegal in many countries, even though its effect can be similar to
the effect of granting exclusive territories, a practice often judged under
a rule of reason or even permitted under conditions. The ban on RPM
may be due to the presumption that it will usually restrict competition
and may be welfare reducing even when the reason for its use has to do
with horizontal externalities. By banning RPM, the gain from reduced
administration costs and from creating a more certain environment for
firms is likely to more than compensate for the loss from cases where RPM
could be welfare enhancing.3 3
Giovannetti and
Magazzini (2013)
Another case is price discrimination in EU as well as in US competition analyse recent empirical
policy. In the European Union, the effective prohibition of geographic price evidence on the use of
discrimination is due to the concern for promoting economic integration RPM by UK firms. They
in addition to economic efficiency. The European Commission has been report that 37.5 per
tough on producers that have made agreements with distributors barring cent of firms in their
sample explained that
the distributor from selling the goods to customers in other than its own
they had used RPM
country. In the United States, the hostility against price discrimination to protect themselves
stems from social and political in addition to economic objectives (in against free-riding; 20
particular, the wish to protect small distributors). However, US competition per cent argued that
authorities have often been reluctant to attack price discrimination. their aim had been to
support brand image,
Among those practices that are typically subject to a rule of reason which is consistent with
approach since they are potentially anti-competitive, some of the most the quality certification
important ones are those which may create barriers to entry or expansion argument for RPM; and
the remaining 42.5
by rivals, as is shown in the following examples.
per cent indicated that
• It may be abusive conduct for a firm which controls an input that is they had used RPM
essential for the supply of the final product to consumers to refuse following complaints by
to sell that input to other firms. This input can be a raw material, downstream firms, which
suggests that RPM
technological know-how, access to a network, etc. However, firms also
in those cases could
have the right to profit from their investments, otherwise they will have helped reduce
not undertake investments in the first place. So there is sometimes a downstream competition
dilemma for competition authorities. and new entry.
• Exclusive dealing can be used by a dominant firm to hinder or prevent

entry by rivals. But since it can also lead to efficiency gains, the
competition authorities must assess its effect in each particular case.
In some competition laws, there is a presumption that both exclusive
dealing arrangements and exclusive territories are normally not anti-
competitive, although they may be in certain cases. Thus EU law is
generally more lenient than US law towards such practices, provided
they do not raise concerns about market foreclosure or geographical
price discrimination.
A disadvantage of the rule of reason approach is high administrative costs,
and this is why economists have also been trying to come up with some
simple rules to guide policy on abuses by dominant firms. In the case
of predatory pricing, for instance, a simple rule is that predation exists
whenever a firm sets a price lower than its average variable cost (a proxy
for short-run marginal cost). But as mentioned in Chapter 6 of this guide,
there are circumstances where a firm may set a price below marginal cost
without being involved in predation as well as circumstances where a firm
that uses a predatory strategy does not need to price below marginal cost.
So predatory pricing is currently judged under a rule of reason.
131
Policy towards research and development

Government policy towards R&D is necessary because the market
mechanism cannot generally be relied upon to produce the socially
optimal amount of R&D. There are several reasons for this, relating
to market imperfections and the nature of the R&D process itself. In
particular:
• Research involves significant sunk costs. On the other hand,
innovations, in the form of new or improved products or processes,
can be imitated, some more easily than others. Hence there is an
incentive problem for potential innovators. If they cannot appropriate
a significant part of the returns from their innovations, they will not be
able to cover their research costs, and hence they will not undertake
any R&D in the first place.4 4
In Chapter 2 of this
guide we discussed, in
• Some R&D projects are so expensive that it is difficult even for large a different context, the
firms to undertake them. Furthermore, research activities often involve inefficiencies generated
a high level of uncertainty, so firms may have difficulty in obtaining when a firm cannot
external finance for some R&D projects that are nevertheless worth appropriate the full
returns from a sunk
undertaking. Such market imperfections may be due to asymmetric
investment.
information between firms and financial institutions.
• The private return to R&D is lower than the social return because
of positive externalities – again leading to underprovision of R&D
investment relative to the social optimum.
Policy towards R&D can take many forms, including the patent system, the
treatment of R&D joint ventures, and government funding of private firms’
R&D.5
5
In addition to promoting
the production of
innovations, governments
Patents
can implement policies to
Patents provide an innovator with exclusive rights to a new product achieve a faster rate of
or process. The patent system is therefore designed to alleviate the adoption of innovations.
appropriability problem and thus directly promote innovation. Moreover, Mowery (1995) discusses
both types of policies.
it is meant to encourage disclosure of new discoveries and thus indirectly
increase the rate of technological progress. On the other hand, since
patents also reduce competition after an innovation is made by granting
a legal monopoly, too much patent protection would not be welfare-
enhancing. So a trade-off exists between incentives to create and innovate,
and access to innovations (the diffusion of the results obtained).
Patent law addresses this problem by offering an exclusive right for a
limited period only. The discussion on optimal patent design has therefore
mostly focused on how long patents should apply and how broad their
scope should be. In many countries, including some in the EU and
USA, patent protection is given for a maximum of 20 years from the
date of application. In exchange for the exclusionary rights, the patent
holder must disclose the invention as part of a publicly available patent
document. When the patent expires, the invention enters the public
domain (i.e. the owner no longer holds exclusive rights to the invention,
which becomes available to commercial exploitation by others).
As for the effectiveness of patents in protecting innovations, the empirical
6
However, firms also
use other means of
evidence suggests that it is limited and varies considerably across
appropriation, such
industries. The main reason is that competitors can legally ‘invent around’ as secrecy, lead time,
patents. Other reasons may include stringent legal requirements for investment in marketing
proving that a patent is valid or that it is being infringed, and the fact that and customer service,
a firm may want to avoid disclosing information through patents.6 and learning by doing.
132
In the USA during the 1980s there was a sharp increase in legal protection
of patents against infringement. This was followed by a sharp rise in
patent applications and patent grants. Some of this was due to a rise in
patent applications by information technology firms. However, there is
little evidence that increased patent protection has led to more innovation.
The degree of protection was reduced in more recent court decisions.
According to Hall et al. (2013), in the UK the share of firms patenting
among those reporting that they have innovated is only about 4 per cent.
(In the USA it is 5.5 per cent.) Survey data from these firms suggest that
they do not consider patents or other forms of formal intellectual property
(such as copyright and trademarks) as important as informal means of
appropriation for protecting innovations. In particular, about 25 per cent
of all innovating firms consider patents to be medium or highly important,
while about 45 per cent consider informal intellectual protection to be
medium or highly important. The use of patents is also so limited because
most firms are small and medium entreprises, many innovations are new
to the firm but not to the market, and many sectors are not patent active.
Research joint ventures

Research joint ventures are agreements between firms to cooperate by
sharing expenditures as well as benefits associated with given research
projects. They are generally allowed in many competition laws, including
US and EU law. Research joint ventures may be welfare-enhancing
because:
• They may prevent needless duplication of R&D expenditures and/or
allow firms to exploit complementarities in their research capabilities.
• They may allow firms to pool resources for R&D projects that are too
expensive for a single firm to undertake.
• They may improve incentives by reducing the appropriability problem.
Note, however, that research joint ventures may also reduce incentives
to perform R&D because they soften R&D competition and might lead
participants to free ride on each other if monitoring or incentives are
not effective. Also, they may facilitate price collusion in the production/
distribution stage. Notwithstanding these deficiencies, R&D cooperation
has not only been allowed but also in many cases specifically subsidised in
the EU, the USA and Japan. The empirical evidence on the effectiveness of
these policies is mixed.
Asymmetric information, R&D funding and government subsidies

Uncertainty and sunk costs are typical of research investments, especially
for radical, as opposed to incremental, innovations. Furthermore,
information regarding an R&D project is often different between the
prospective innovator and other parties, including banks and other potential
external funding sources. The problems of adverse selection and moral
hazard, which we discussed in Chapter 7 of this guide, also apply to R&D
and are expected to result in under-financing of innovation in the private
market.
Adverse selection is a problem because if some potential innovators
know that they have very profitable or not very risky projects and others
know that they have less profitable or riskier projects, the latter may
prefer external funding while the former may seek external funding only
if internal financing is not available. Lenders will know this, and since
they are likely to have less information than the innovators about the
prospects or riskiness of any given project, they may be reluctant to invest.
133
Even good projects may have difficulty getting funding, since potential
innovators may be reluctant to reveal sensitive information that might
jeopardise the value of the project or simply because such signalling is
difficult for technically complex projects.
Moral hazard is a problem because once an innovator has secured
external funding, he or she may have an incentive to lower the quality
of the investment or to divert time and effort to another project. Lenders
may fear this, and since they know they will have difficulty monitoring
the innovator’s effort, they may be reluctant to invest in the first place.
Obviously, the underprovision of R&D because of adverse selection and
moral hazard problems is not just bad for innovators, it is also inefficient
for society as a whole.
Activity
Suppose that two innovation opportunities exist, each costing £100 in initial outlay,
but generating different possible returns. Investment A generates a return of £150 with
probability 0.5 and £100 with probability 0.5, while investment B generates a return of
£180 with probability 0.5 and £10 with probability 0.5. Define and calculate the expected
return for each investment. Then compare and contrast the investment choice of a self-
financed investor and an investor financed by debt. Finally, explain what your analysis
implies for the financing of new technological investment.
Answer
The net expected value of investment A is 0.5 × (150 – 100) + 0.5 × (100 – 100) = 25.
The net expected value of investment B is 0.5 × (180 – 100) + 0.5 × (10 – 100) = –5.
A self-financed investor would choose A since it has the largest net expected value. An
investor financed by debt would choose B since in case the return is low (£10), he/she
would default and not return the full loan of £100. In particular, assuming that the most
that can be returned is the payoff of the investment, the expected return of project B for
an investor financed by debt is 0.5 × (180 – 100) + 0.5 × (10 – 10) = 40 > 25.
To analyse the implications of this adverse selection problem for the financing of
innovations, one needs also to examine the point of view of the lender: he/she would
never want to finance project B. However, only innovators with projects such as B will
seek debt financing. Note that adverse selection is caused not by the uncertainty per se
but by the asymmetry of information: the lender cannot know the payoffs of investments
but the investor can.
Government financing of R&D performed by individual firms is an obvious

way to attempt to increase total R&D expenditure. Government subsidies
can be horizontal, for instance implemented as tax allowances based on
a firm’s total R&D spending. The USA, Japan and other countries have
such schemes. Or they can be more selective, for instance being given as
grants and targeted towards particular sectors, technologies or projects.
Most countries as well as the EU use such grants. Horizontal subsidies
have some drawbacks, for example they may fail to have much effect on
firms’ R&D decisions or they may encourage firms to undertake projects
with limited value. On the other hand, a problem with selective financing
of R&D is that governments may have less good information than firms on
the likelihood of making a discovery or on its value. Note that to assess the
effectiveness of R&D subsidies and grants one must examine whether any
additional R&D performed as a result of these policies (taking into account
the possibility that government research subsidies may, to some extent,
crowd out private R&D investment) more than compensates for the public
revenues lost. As with cooperative research agreements, the empirical
evidence on this issue is not conclusive.
134
Activities
1. Suppose you are asked to assess the likely effects of a proposed horizontal merger in
a particular industry. What are the things you would look at and how would each of
these help you make your assessment?
2. Suppose you are asked to examine whether a firm that has exclusive dealing
agreements with its distributors is guilty of abusing market power. What are the
things you would look at and how would each of these inform your investigation?
3. Some economists argue that competition policy raises special issues in innovative
industries: monopoly power may need to be tolerated and horizontal agreements
among firms in such industries should receive special treatment. Do you agree? Justify
your opinion with reference to economic theory and any relevant empirical evidence.
4. Consider a homogeneous good Cournot duopoly with inverse demand function given
by p = 1 – Q. The two firms have identical marginal costs equal to 0.4 and propose
a merger. The firms claim that the merger will result in a decrease of the marginal
cost of the merged firm by x per cent. How large would x need to be for welfare to
increase rather than decrease as a result of the merger?
5. The case studies included in Kwoka and White (2013) and Lyons (2009) are strongly
recommended reading if you wish to obtain a better understanding of competition
policy issues. These books contain analyses of several US and EU antitrust policy
cases, many of them written by well-known economists. They cover horizontal
mergers, collusion and horizontal agreements, abuse of dominant position, vertical
relations and more. The authors provide not only the facts and the outcome for each
case, but also an economic analysis and an assessment. Read critically the arguments
for and against, and assess the decision of the court and the author’s conclusions.
6. Are patents largely worthless? Would there be the same amount of innovation with or
without them? These sound like extreme statements, but some researchers do actually
argue that except for some industries, such as pharmaceuticals, where patents may
play a significant incentive role, the main function of patents is to facilitate trade
in technology through licensing agreements rather than to provide incentives to
innovate. Others maintain that the benefits of the patent system are too small to
outweigh the welfare losses caused by lobbying and rent seeking.
Assess the evidence on the private and social value of the patent system in light of
the empirical literature on firms’ patenting behaviour; the effectiveness of patent
protection relative to other forms of intellectual property rights protection across
different industries; the relationship between strength of patent laws and innovation;
and the political economy of patent systems. A good starting point is Hall et al. (2014)
as well as a collection of articles on patents published in the Winter 2013 issue of the
Journal of Economic Perspectives.

• explain the rationale for competition policy and for industrial policy
towards R&D
• describe the difficulties and dilemmas in competition policy with
respect to mergers, restrictive agreements, and abuses of market
power
• discuss the practice of competition policy
• describe and evaluate the main forms of government policy towards
R&D.
135

1. Describe the difficulties faced by competition authorities in the
design and implementation of merger policy. Include in your answer
a discussion of the possible economic causes and consequences of
mergers and any evidence from the operation of merger policy in
practice.
2. Explain the rationale for government policy towards research and
development in the form of (a) a patent system, and (b) R&D
subsidies. Then discuss the advantages and disadvantages of these
policies, with reference to any relevant empirical evidence.
136
Chapter 12: Regulation
Introduction
In some industries effective competition is difficult or impossible since
firms possess a lot of market power and are likely to abuse it. In such
cases the direct regulation of firms becomes necessary. This is typically
the case in industries with ‘natural monopoly’ elements, that is industries
whose technological characteristics are such that average industry cost is
minimised when a single firm serves the whole market. The utilities are
examples of industries which combine naturally monopolistic activities,
such as transmission networks, with potentially competitive activities, such
as the provision of services over the networks. These industries pose two
sets of questions, namely: what is the best industry structure and what is
the best way to regulate firm behaviour.
In many cases these questions have been posed during the process of
privatisation of previously state-owned monopolies. While privatisation
may often improve the productive efficiency of firms, it may also worsen
allocative efficiency; hence the need for public policy following the
privatisation of a monopoly. This can take the form of liberalisation: the
removal of restrictions on competition such as restrictions on entry. Or it can
involve the restructuring of the industry: the breaking up of the monopoly.
Or it can involve the regulation of firm conduct. Or, finally, it can be a
combination of these. This chapter examines some of the theory and practice
of regulation as well as the links between liberalisation and regulation.
Learning outcomes
• describe different ways of regulating a firm under symmetric
information
• analyse the links between type of regulation, incentives, productive
efficiency and allocative efficiency under asymmetric information
• explain the implications of a dynamic analysis of regulation
• describe the links between liberalisation and regulation.
Essential reading
Armstrong, M., S. Cowan and J. Vickers Regulatory reform. Chapters 1–6
and 11.
24 and 26.
Further reading
Books
Armstrong, M., S. Cowan and J. Vickers Regulatory reform. (Cambridge, MA:
MIT Press, 1994) Chapters 7–10.
Braeutigam, R.R. ‘Optimal policies for natural monopolies’, in Schmalensee,
137
Church, J.R. and R. Ware Industrial organization: a strategic approach. (Irwin
McGraw-Hill, 2000) Chapter 25.
Joskow P.L. and N.L. Rose ‘The effects of economic regulation’, in Schmalensee,
R. and R. Willig (eds) Handbook of industrial organization, volume 2.
Journals
Armstrong, M. and D.E.M. Sappington ‘Regulation, competition, and
liberalization’, Journal of Economic Literature 44(2) 2006, pp.325–66.
Helm, D. and T. Jenkinson ‘The assessment: introducing competition into
regulated industries’, Oxford Review of Economic Policy 13 1997, pp.1–14
(Special issue: Competition in Regulated Industries).
Vickers, J. ‘Regulation, competition, and the structure of prices’ Oxford Review
of Economic Policy 13 1997, pp.15–26 (Special issue: Competition in
Regulated Industries).
Regulation of firms with market power under symmetric

information
The simplest possible setup involves a monopolist producer of a single
good and a regulator whose only objective is to maximise the sum of
producer and consumer surplus. Moreover, there is symmetric information
about market demand and the monopolist’s costs, and the analysis is static.
In this framework, a simple rule for the regulator would be to require that
price equals marginal cost. However, if the firm operates under increasing
returns to scale – which is a sufficient, though not necessary, condition
for being a natural monopoly – it will then make a loss, since price will be
lower than average cost.
One response to this problem is to set a uniform price that maximises
welfare – or any weighted sum of consumer and producer surplus with
more weight put on the former – subject to the condition that the firm
breaks even. It turns out that, in the absence of direct subsidies to the firm,
this is the price pA given by the intersection of the demand curve with the
average cost curve (Figure 12.1).
p
pA
AC
demand
qA q
Figure 12.1
Since the AC curve is everywhere declining under increasing returns to
scale, the firm will produce exactly quantity qA when the regulator sets the
price pA: if it produces less it will make losses, while if it produces more
it will not be able to sell the extra quantity. Why is this optimal under the
circumstances? For one thing, the price cannot be set lower than pA, since
the firm will then not be able to sell a sufficient quantity to achieve p =
138
AC and will therefore make a loss. If, on the other hand, the price is set
by the regulator higher than pA, the (possibly weighted) sum of producer
and consumer surplus clearly decreases. Note that, since AC > MC at price
pA, we have p > MC; so average cost pricing results in some allocative
inefficiency.
An even better solution is a two-part tariff – a pricing schedule of the form
T(q) = A + pq. This is in fact fully efficient when all consumers are
identical, so that they can be regarded as a single consumer with a surplus
function (gross of any fixed fee) V(p). Suppose that the regulator chooses
a price p and a fixed fee A to maximise the sum of consumer surplus and
profit, namely W = [V(p) – A] + [A + pq(p) – C(q(p))], subject to a
consumer participation constraint V(p) – A ≥ 0 and a firm breakeven
constraint A + pq(p) – C(q(p)) ≥ 0. It turns out that price should be set at
the level given by the intersection of the demand curve with the marginal
cost curve, where p = MC and the market clears. In addition, the fixed fee
A should be set at a level such that both constraints are satisfied.1 1
When the objective
function of the regulator
Activity is a weighted sum of
producer and consumer
Prove this result. surplus with more
weight put on consumer
Answer surplus, A should be
set to just allow the
Ignore for the moment the two constraints and maximise W with respect to p, noting that firm to break even and
∂V(p)/∂p = –q(p). The first-order condition implies p = ∂C(q)/∂q. Provided that V(p) ≥ price should again be
pq(p) – C(q(p)), A can then be chosen so that both constraints are satisfied. equal to marginal cost.
The reason is that the
problem then reduces
When consumers are not identical, a two-part tariff can again be used to to simply choosing price
improve welfare relative to a uniform price, but it will not achieve full to maximise consumer
surplus subject to the
efficiency since it will involve a price higher than marginal cost.2 However,
firm making zero profit.
welfare could be improved through more general nonlinear pricing
schemes, such as different packages directed to different consumer types.
We have assumed so far that the monopolist produces a single product. 2
Compare with
What about the case of a multiproduct monopolist? Note that the analysis of
second degree price
‘multiproduct’ may also refer here to a single product sold to different
discrimination in
classes of consumers, or in different areas, on in different periods. Chapter 8 of this guide.
Marginal cost pricing will again result in the firm making losses if it is a
natural monopoly.3 Therefore one solution is ‘Ramsey pricing’, that is the 3
The definition of
set of linear prices that maximise the weighted sum of consumer surplus a natural monopoly
and profit subject to the firm breaking even. In the special case of has to do with
the technological
independent demands for the products, Ramsey pricing gives (pi – ∂C/∂qi)/
characteristics of an
pi = λ/εi, ∀i, where εi is the elasticity of demand for product i and λ is industry, and is distinct
chosen so that the firm makes zero profit. An even better solution is to use from the question of
nonlinear pricing. In the case of a single product sold during different whether an actual
periods, some with a higher demand than others, an additional problem is monopoly exists in the
that capacity may be costly to adjust over time, so there is a trade-off industry. See Armstrong
et al. (1994), pp.49–50,
between excess capacity during off-peak times and rationing during peak
for a formal definition.
times.4 In practice, Ramsey pricing is not often used, for a number of
4
See Armstrong et
reasons, including the regulators’ imperfect information about demand
al. (1994), pp.51–58,
elasticities and the political unacceptability of price discrimination in the and Church and Ware
utility industries driven by differences in demand or service costs. (2000), for details on
these schemes.
Regulation under asymmetric information

Firms are typically better informed than regulators about cost and demand
conditions in their industry. Moreover, cost usually partly depends on the
firm’s cost-reducing effort, and the regulator is less well informed than the
firm about that effort as well. So regulatory schemes must be devised that: 139
• take into account the existence of asymmetric information about

industry conditions, and
• provide appropriate incentives to the firm to reduce costs.
The theoretical literature suggests that optimal regulation under
asymmetric information must involve lump-sum transfers from the
government to the firm, in addition to setting a price for the product. In
practice, however, regulators do not typically have the power to make
lump-sum transfers, so the problem becomes one of regulating price (and
possibly quality as well). The regulatory authority is then faced with the
following dilemma. If it fixes a price in advance, it gives good incentives
to the firm to reduce its cost, but the price may end up being much higher
than cost; so productive efficiency is served, but allocative efficiency is
not. If, on the other hand, the regulator makes the price dependent on the
actual cost realisation, the firm makes no excessive profits, but it has little
incentive to cut costs; so the scheme improves allocative efficiency, but it
worsens productive efficiency.
Examining this trade-off can provide some insights to an important policy
issue, namely the choice between ‘price-cap regulation’, which refers to
specifying a price independently of the realised cost, and ‘rate of return
regulation’, where price closely reflects realised cost. A form of price-cap
regulation has been used in the UK following the privatisation of utilities
and other industries, while rate of return regulation has traditionally been
used in the US.5 5
Carlton and Perloff
(2005), Chapter 20,
The following simple model of regulation of a single-product firm captures
and Church and Ware
some of the links between regulation, incentives and efficiency.6 The firm’s (2000), Chapter 26,
unit cost is given by c = θ – e, where θ is a random variable and e ≥ 0 is provide detailed
the effort level chosen by the firm. Effort has a cost for the firm, given by discussions of rate of
the function ψ(e) = e2/2. The regulator does not observe θ, although he return regulation.
knows that it is distributed according to the density function f(θ), with 6
This part follows
mean µ and variance s2. Neither does the regulator observe the level of e Armstrong et al. (1994),
chosen by the firm. However, he observes the realised marginal cost c. We pp.39–42.
also assume that when the firm makes its decision about effort, it does not
know the value of θ, and so it faces some uncertainty about its profit. The
firm is also assumed to be risk averse. In particular, its objective function is
given by E(Π) – (γ/2)var(Π), where E(Π) is expected profit, var(Π) is the
variance of profit and γ is a measure of risk aversion. This is not an
unrealistic assumption, since it is probably the manager of the firm who
will take the decision on effort rather than the shareholders. Note that the
case of a risk-neutral firm (γ = 0) and the case of no cost uncertainty (s2
= 0) are special cases within this model. To simplify, we abstract from the
question of output choice by assuming that demand is perfectly inelastic 7
This implies that total
and equal to 1 at all prices.7 expenditure on the
product is equal to the
The timing of the game between the regulator and the firm is as follows.
product’s price.
At stage 1, the regulator sets a price rule of the form p(c) = p + (1 – ρ)c,
where ρ∈[0,1] is a parameter which determines the sensitivity of price to
unit cost: ρ = 1 represents a pure price cap, while ρ = 0 is a form of rate of
return regulation since it implies that the price-cost differential is fixed. At
stage 2, the firm observes this rule and chooses e to maximise its objective
function.
Suppose the objective of the regulator is to choose p and ρ to minimise
expected total expenditure on the product,
E[p(c)] = p + (1 – ρ)E(c) = p + (1 – ρ)(µ – e) ,
subject to the participation constraint E(Π) – (γ/2)var(Π) ≥ Π0, where
Π0 is the (non-random) reservation ‘utility’ level of the firm (if Π0 = 0,
140
this reduces to a form of breakeven constraint). Note that minimising total

expenditure is here equivalent to maximising consumer surplus, since
quantity is fixed and equal to 1.
We start with the firm’s action at stage 2 of the game. Its profit (i.e. revenue
minus production cost minus the cost of effort) is
p + (1 – ρ)(θ – e) – (θ – e) – e2/2 = p – ρ(θ – e) – e2/2 .
Hence its expected utility is
E(∏) – (γ/2)var(∏) = p – ρ(μ – e) – e2/2 – γρ2σ2/2 .
Note that var(Π) = var[ p – ρ(θ – e) – e2/2] = var(–ρθ) = ρ2σ2. The
firm chooses e to maximise this, taking p and ρ as given. The first-order
condition gives e* = ρ.
Now we go back to the regulator’s decision at stage 1. Since the regulator
anticipates that e* = ρ, he will choose ρ and p to minimise p + (1 – ρ)(µ –
ρ) subject to the participation constraint p – ρ(µ – ρ) – ρ2/2 – gr2s2/2 ≥ Π0.
Clearly, whatever the optimal value of ρ, total expenditure minimisation
implies that p will be set at the lowest level that allows the firm to make at
least utility Π0. In other words, it will be set so that the firm makes exactly
Π0 and hence the participation constraint holds with equality. Substituting
p from the participation constraint (with equality) into the regulator’s
objective function, the problem simplifies to choosing the value of ρ that
minimises
∏0 + ρ(μ – ρ) + ρ2/2 + γρ2σ2/2 + (1 – ρ)(μ – ρ).
You can check that the first-order condition gives ρ* = 1/(1 + gs2).
To summarise the main results of this model:
• The lower the value of ρ, the lower the effort level of the firm. This result
is independent of the values of γ and σ2. Thus price-cap regulation is
better for productive efficiency than rate of return regulation.
• If the firm is risk-neutral (γ = 0) or both the firm and the regulator have
perfect information about cost (s2 = 0), then ρ* = 1, that is the optimal
scheme is pure price-cap regulation. The more risk averse the firm or
the higher the cost uncertainty, the more sensitive should the regulated
price be to cost. Thus price-cap regulation may be problematic if there is
significant cost uncertainty.
Finally, what does this model say about the trade-off between allocative and
productive efficiency? Note that the assumption of perfectly inelastic demand
does not allow us to directly introduce the allocative efficiency issue in this
model. Nevertheless, this is introduced somewhat indirectly through the risk
aversion of the firm. For instance, think what would happen if the regulator
chose ρ* = 1 in the interest of productive efficiency even if γ and σ2 were
both positive. Then total expenditure would not be minimised because p
would have to be higher to ensure that the firm participates. By reducing ρ*
the regulator sacrifices some cost savings to improve the firm’s insurance,
and hence also improve allocative efficiency. When γ = 0 or s2 = 0, this
effect no longer operates, and so ρ* = 1 is optimal. In a more general setting,
demand would be elastic and the allocative efficiency issue would be directly
introduced. In this case, a value of ρ* between 0 and 1 would normally be
optimal even if γ or s2 were equal to 0.
When deciding on the form of price regulation for a multiproduct
monopolist, an important question is whether to regulate the price of each
individual product or some measure of the average of the prices set by the
firm. The regulator is likely to have less good information than the firm
141
about costs, so an advantage of having a single constraint is that it allows

the firm to adjust relative prices according to relative costs. The
disadvantage is that the firm is likely to practice price discrimination even
when this is not socially beneficial. In general, however, some form of
8
Armstrong et al.
average price regulation is desirable.8
(1994), pp.66–74 and
A form of price regulation that has been applied to many privatised 79–83, give details
utilities in the UK is the so-called RPI-X regulation. A firm subject on possible schemes,
both in a static and in a
to RPI-X regulation is constrained to have a weighted average of price
dynamic framework.
increases during one year not greater than the percentage increase in the
Retail Price Index (RPI) minus a factor X. In essence, the regulator sets
targets for cost reduction over time, adjusted to allow for inflation. If the
firm reduces cost by more than the target, it can keep the profits. X is fixed
by the regulator, does not change between regulatory reviews, and the
number of years between regulatory reviews is normally also fixed. There
is some flexibility in the system to allow for price adjustments when there
are unanticipated changes in the cost of certain inputs. Also, the price cap
applies to an average of prices and firms have some freedom with respect
to relative prices, although this freedom is not complete. Because of the
danger of underprovision of quality or distortion of investment incentives
under price-cap regulation, quality is also regulated and capital investment
is monitored. Finally, the factors that are taken into account by the
regulators when setting the factor X include the value of the firm’s existing
assets and the cost of capital (in order to ensure that the firm will be
financially viable), expected rates of demand and productivity growth, and
the progress of competition in activities which are potentially competitive.
Privatisation and RPI-X regulation in the UK have resulted in lower prices
and higher productivity in most privatised industries.
Dynamic issues in regulation and regulatory capture

Regulation is a dynamic process that involves interaction between
the firm and the regulator, whether this takes place continuously, as
in rate of return regulation, or with a ‘regulatory lag’, as in price-cap
regulation. Hence present actions are influenced both by past actions
and by anticipated future actions. In particular, the regulator can
exploit information revealed by the firm or the sunk nature of the firm’s
investments and impose a tougher regime in the future. Since the firm
realises this, it behaves strategically to obtain a more favourable regime.
One problem, which occurs under asymmetric information, is that the firm
may not want to reveal that its costs are low or make considerable effort to
reduce them, because then the regulator will have better information and
may set a price to reduce the firm’s profits in the next regulatory review. It
can be shown, in fact, that social welfare might be higher if the regulator
could commit not to use the information revealed by the firm today when
setting the price tomorrow. In practice, this results in a trade-off regarding
the length of the regulatory lag. If price is kept fixed for a long period, the
firm has greater incentives to reduce cost (especially in the beginning of
the period), but consumers may pay excessive prices for a long time. If, on
the other hand, the regulatory lag is short, prices reflect costs more closely,
but the incentives for cost reduction are lower.
Another problem is that the firm may underinvest because, when the
regulator next reviews the price, the investment will be sunk and the firm 9
Compare with a similar
may be prevented from capturing the returns from its investment.9 A situation in Chapter 2 of
this guide.
solution to this problem may be to give the firm a ‘fair’ return on all
‘useful’ investments. This can be explicitly specified as a right of the firm
142
or left to be implicitly enforced by the fact that the regulator will want to
build a reputation of being fair.
Finally, you should bear in mind that regulators are not necessarily the
benevolent agents we have so far assumed them to be. Rather than
maximising social welfare, they may act in the interest of incumbent
firms in the regulated industry (which, in turn, may engage in wasteful
lobbying). To reduce the scope for ‘regulatory capture’, it is therefore
desirable to limit the regulators’ discretion.
Liberalisation and regulation

Starting from a situation where there is a protected monopolist in a
potentially competitive activity, there are three questions that can be
asked:
• Should there be completely free entry into the industry or should entry
be restricted?
• Should the monopolist be broken up or left intact?
• Given that liberalisation cannot always fully substitute for regulation,
how should the two be related?
There are some advantages of liberalisation relative to regulation. It may
promote productive and allocative efficiency by overcoming asymmetric
information problems and it can help avoid problems with policy
credibility and regulatory capture. On the other hand, liberalisation also
has problems. The most important is that it may simply not be sufficient
to generate effective competition, because of significant sunk costs, first-
mover advantages, or for other reasons. Or, conversely, it may lead to
inefficient entry, such as ‘cream-skimming’ – although there seems to be
little evidence that technological conditions in the utility industries favour
cream-skimming.
It has also been argued that franchising – selling the right to serve a
market for a certain period to the lowest bidder – could be an alternative
to regulation in some cases. The idea is that in circumstances where
competition in a market is not feasible, competition among firms to
obtain the contract (‘competition for the market’) may improve efficiency
by dissipating excessive profits. Franchising has been used for relatively
simple products and services that involve low sunk costs (for instance,
rubbish collection or train services), but it is impractical for most activities
in the utility industries because:
• simple and complete contracts cannot be specified, so monitoring and
enforcement are difficult and costly
• a firm which obtains the contract once may have a significant
advantage in subsequent rounds of bidding
• a firm which fails to get its contract renewed must receive considerable
compensation for its sunk costs.
In some cases where franchising has been used, such as in rail services in
the UK, the need for price and quality regulation has remained.
Rather than being substitutes, liberalisation and regulation can be
complements. For instance, some degree of competition in regulated
industries may facilitate regulation by improving the information available
to the regulator. Moreover, the utility industries combine naturally
monopolistic activities with potentially competitive ones. Unless there
are significant economies of scope between the two types of activities,
143
there is no reason why competition should not be introduced where it

works. Thus, although transmission networks (for instance, electricity
distribution) are run by natural monopolies that must be regulated, there
is usually a case for competition in the provision of services over the
networks (electricity generation or supply of electricity to customers).
There are then several issues that must be addressed:
• If there is room for only a few firms to be given the right to use the
network, should firms compete for that right, by periodic allocation of
franchises, and if so how should that be organised?
• Should the firm which runs the network also be allowed to compete in
providing services over the network?
• How should the government regulate the terms on which this firm
gives network access to other firms?
The choice of industry structure involves assessing efficiency effects
against market power effects, as in many other policy issues. Several of
the topics discussed in previous chapters of this guide are relevant here.
For instance, the larger the economies of scope and the lower the chance
of a ‘hold-up’ problem, the higher the likelihood of a welfare gain from
allowing the monopoly to operate in the competitive market segment.
The larger the chance that vertical integration raises the cost of access of
non-integrated firms to the network, the higher the likelihood of a welfare
loss. Of course, the access price can be regulated too. On the one hand,
the socially optimal access price should not be too high, since this will
discourage efficient entry and provide incentives for entrants to duplicate
the network. On the other hand, it should not be too low, since this will
encourage inefficient entry and reduce the incentive of the incumbent to
invest in the network.
Activities
1. Consider the model of regulation of a single-product firm under asymmetric
information analysed above, but assume that the regulator’s objective is to maximise
profit minus total expenditure. Show that this does not affect the optimal choice of
ρ and that p can be set anywhere between a lower bound (so that a firm’s expected
utility is at least Π0) and an upper bound (so that consumers have non-negative
expected utility).
2. Lump-sum transfers to firms are not generally used in regulation. Can you think of
any reasons why?
3. A natural monopolist has known total costs C(Q) = 300 + 15Q and faces market
demand Q = 200 – 2P. Derive the monopolist’s output and profit and the consumer
surplus when: (i) price is set equal to marginal cost; (ii) price is set equal to average
cost; (iii) there is two-part pricing and the monopolist chooses the tariff to maximise
profit; (iv) there is two-part pricing and a regulator chooses the tariff to maximise
consumer surplus, subject to the monopolist breaking even. For parts (iii) and (iv), you
can assume that there are N identical consumers. Comment on the optimal regulation
scheme in this market.
4. The deregulation of the Californian electricity sector in the 1990s has been the
subject of considerable criticism. How can we explain the experience of this sector,
including the much publicised crisis of 2000–2001? What are the lessons to be
learned? There is a large literature on this topic, including review articles by Joskow
in the Journal of Economic Perspectives (1997) and the Oxford Review of Economic
Policy (2001), and Borenstein in the Journal of Economic Perspectives (2002). You can
also search the internet for more.
144
5. What is best for social welfare: regulated monopoly or unregulated competition?

This issue is particularly relevant in many industries (such as gas, electricity, water,
telecommunications and transport) where significant scale economies make
production by many firms inefficient, but where some competition may be useful to
help discipline incumbent firms. It turns out that the answer depends on a number
of factors. Review the economic arguments for one or the other of these two forms
of industrial organisation. A good starting point is Armstrong and Sappington
(2006). They also discuss the optimal design of liberalisation policy in settings where
competition is preferable to regulated monopoly. Read their analysis and assess the
advantages and disadvantages of various types of liberalisation policy.
6. Not all regulated industries are natural monopolies. In fact, many are not, and some
would even appear to come close to the textbook example of perfect competition!
Take, for instance, taxi services. There are many buyers, many sellers, and the cost of
entry is low. Yet taxi services are regulated throughout the world. Entry is restricted by
limiting the number of licenses, and often prices are regulated too. What could be the
justification for this? Does regulation of taxi services increase social welfare?
You may want to read a recent official report about taxi service regulation in the UK
to find out more: The Regulation of Licensed Taxi and PHV Services in the UK (2003),
available online.
A lot more can be found on the internet, both on entry regulations in general and
on the taxi industry in particular, including a recent comprehensive OECD study, Taxi
Services Regulation and Competition (2007), and numerous references to articles in
academic journals.

• describe different ways of regulating a firm under symmetric
information
• analyse the links between type of regulation, incentives, productive
efficiency and allocative efficiency under asymmetric information
• explain the implications of a dynamic analysis of regulation
• describe the links between liberalisation and regulation.

1. ‘Price-cap regulation is more efficient than rate of return regulation’.
Discuss this statement, with reference to an economic analysis of
the links between type of regulation, incentives of regulated firms to
reduce costs, allocative efficiency, and productive efficiency.
2. Describe, with reference to theory as well as empirical evidence, the
ways in which competition and regulation may interact. Include in
your answer a discussion of situations where the two are viewed
as substitutes, as well as of situations where they are viewed as
complements.
145
Notes
146
Appendix 1: Sample examination paper
Important note: This Sample examination paper reflects the

examination and assessment arrangements for this course in the academic
year 2023–2024. The format and structure of the examination may have
changed since the publication of this subject guide. You can find the most
recent examination papers on the VLE where all changes to the format of
the examination are posted.
Time allowed: three hours and 15 minutes

Candidates should answer FOUR of the following EIGHT questions:
TWO from Section A and TWO from Section B.
If more than four questions are answered, only the first questions
attempted will be counted
Section A
1. Describe the transaction costs–property rights approach to the theory
of the firm, including a careful examination of the main elements of
this approach and discussion of the basic theoretical predictions. What
is the empirical support for this approach?
2. Discuss how tacit collusion among firms may be sustainable as an
equilibrium when firms interact repeatedly. How does this analysis
help clarify some of the conditions facilitating or hindering collusion?
What do you think are the policy implications of the analysis?
3. Answer all parts:
a. Consider a domestic and a foreign firm that produce differentiated
products and compete in prices in the domestic market. Examine
how a government-imposed quota on the sales of the foreign firm
may affect equilibrium prices and profits.
b. Consider a market with two firms that compete in quantities.
Prior to the quantity competition, one of the firms can reduce its
marginal cost by investing in R&D. Examine its incentive to do so.
4. Describe the theory on the determinants of market structure in
advertising-intensive industries. In what ways are the theoretical
predictions different from those for ‘exogenous sunk cost industries’?
How can this theory best be tested? Discuss briefly the empirical
evidence on the theory.
Section B
5. Answer all parts:
a. Describe and prove the Bertrand paradox.
b. Consider a duopoly producing a homogeneous product. Firm 1
produces one unit of output with one unit of labour and one unit
of capital, while firm 2 produces one unit of output with two units
of labour and one unit of capital. The unit costs of labour and
capital are w and r respectively. The demand is p = a – q1 – q2 and
the firms compete in quantities.
147
i. Compute the reaction functions of the two firms. How are

these affected by changes in the demand shift parameter a and
in the input prices w and r?
ii. Compute the Cournot–Nash equilibrium.
iii. Show that firm 1’s equilibrium profit is not affected by the
price of labour.
6. A monopoly supplier of an intermediate good sells to two different
competitive downstream industries. Total demand for the good by
industry A is given by qA = 1 – pA and total demand by industry B is
given by qB = 1/2 – pB. The unit production cost of the monopolist is
zero.
a. Suppose that the monopolist must charge a uniform linear price to
the two industries. What is the profit-maximising price?
b. Now suppose that the monopolist can practice third-degree price
discrimination. What is the profit-maximising price for each
industry?
c. If third-degree price discrimination is not possible, how can the
monopolist achieve the same result by integrating forward into
one of the downstream markets? Which one will he integrate into
and why?
d. Finally, suppose that the monopolist faces a fringe of competitive
firms selling the same intermediate good. The unit production cost
of each of these firms is 1/3. How does this affect the monopolist’s
decisions under parts a–c?
7. Consider a vertically differentiated market with two firms competing
in prices and each supplying one quality level of the good. The
quality levels are fixed at s1 = 1 and s2 = 2. Consumers are uniformly
distributed with density S according to a taste parameter θ over the
interval [0, 1]. Each consumer buys one unit of the good or none.
Consumer utility is ui(s, p) = θis – p from purchasing a good of quality
s at price p, and zero if no good is bought.
a. Derive the demand functions for firms 1 and 2. (Hint: define
the marginal consumer who is just indifferent between buying
from firm 1 and buying from firm 2; furthermore, define another
marginal consumer who is just indifferent between not buying and
buying from firm 1.)
b. Assuming zero marginal cost for both firms, compute the Nash
equilibrium prices and profits. Are there any consumers who do
not buy the good?
c. Suppose now that firm 2’s marginal cost rises to 1. Compute the
new equilibrium prices and profits.
d. Compare the equilibrium profits in parts b and c and provide
economic intuition for why they differ in the two cases.
8. Two firms producing a homogeneous good compete in a two-stage
game. In stage 1, firm 1 can purchase cost-reducing capital equipment
k. In stage 2, firms compete by simultaneously choosing quantities.
Firm 1’s total cost (including the cost of the capital equipment) is q1(2
– k/4) + k2/18, where q1 is firm 1’s output. Firm 2’s cost is 2q2, where
q2 is firm 2’s output. Market (inverse) demand is given by P = 50 – 2Q,
where Q = q1 + q2.
148
a. Derive the firms’ reaction functions for the game in stage 2. What
is the effect of k on the reaction functions? Illustrate your answer
with a plot of the reaction functions for different levels of k.
b. Find the subgame perfect equilibrium quantities. What is the level
of investment, k, that firm 1 makes at stage 1?
c. Now suppose the firms play a three-stage game. In stage 1, the
incumbent firm 1 purchases cost-reducing capital equipment k. In
stage 2, firm 2 decides to enter or not. If firm 2 enters, the firms
compete by simultaneously choosing quantities in stage 3. If firm
2 does not enter, firm 1 is a monopolist in stage 3. Market demand
and cost functions are given as before. There is no entry cost for
firm 2. What level of k would firm 1 have to choose in order to
deter firm 2 from entering? Should (or can) firm 1 deter entry?
d. How might your answer in part c change if firm 2 had to incur a
positive entry cost F to enter the industry at stage 2 of the game?
Explain.
END OF PAPER
149
Notes
150
Appendix 2: End-of-chapter Activities reading list
Appendix 2: End-of-chapter Activities

reading list
To start your research for the end-of-chapter Activities, see if you can find
and read some of the following suggested articles or book chapters.
Chapter 1
7 and 9.
Gibbons, R. Game theory for applied economics. (Princeton, NJ: Princeton
Osborne, M.J. and A. Rubinstein A course in game theory. (Cambridge, MA: MIT
Press, 1994) [ISBN 9780262650403].
Fundenberg, D. and J. Tirole Game theory. (Cambridge, MA: MIT Press, 1994)
[ISBN 9780262061414].
Kreps, D. and R. Wilson ‘Sequential equilibrium’, Econometrica 50(4) 1982,
pp.863–94.
Chapter 2
Alchian, A.A. and H. Demsetz ‘Production, information costs, and economic
organization’, American Economic Review 62(5) 1972, pp.777–95.
Klein, B., Crawford, R.G. and A.A. Alchian ‘Vertical integration, appropriable
rents, and the competitive contracting process’, Journal of Law and
Economics 21(2) 1978, pp.297–326.
Various articles in the Journal of Law and Economics 43(1) 2000.
Chapter 3
Bennedsen, M., K.M. Nielsen, F. Perez-Gonzalez and D. Wolfenzon ‘Inside the
family firm: the role of families in succession decisions and performance’,
Quarterly Journal of Economics 122(2) 2007, pp.647–91.
Bertrand, M. and A. Schoar ‘The role of family in family firms’, Journal of
Economic Perspectives 20(2) 2006, pp.73–96.
Ederer, F. and G. Manso ‘Is pay for performance detrimental to innovation?’,
Management Science 59(7) 2013, pp.1496–513.
Gneezy, U., S. Meier and P. Rey-Biel ‘When and why incentives (don’t) work to
modify behavior’, Journal of Economic Perspectives 25(4) 2011, pp.191–210.
Lazear, E.P. ‘Performance pay and productivity’, American Economic Review
90(5) 2000, pp. 1346–361.
Villalonga, B. and R. Amit ‘How do family ownership, control and management
affect firm value?’, Journal of Financial Economics 80(2) 2006, pp.385–417.
Chapter 4
Ellison, G. and S.F. Ellison ‘Lessons about markets from the internet’, Journal of
Economic Perspectives 19(2) 2005, pp.139–58.
Holt, C.A. ‘Industrial organization: A survey of laboratory research’, in Kagel,
J. and A. Roth (eds) Handbook of experimental economics. (Princeton:
Princeton University Press, 1995).
151
Levin, J. ‘The economics of internet markets’, NBER Working Paper No. 16852
(2011).
Various articles in the Journal of Industrial Organization 18(1) 2000 (Special
issue: Experimental economics and industrial organization).
Various articles in the Journal of Industrial Organization 29(1) 2011 (Special
issue: Experiments in industrial organization).
Chapter 5
Brenner, S. ‘An empirical study of the European corporate leniency program’,
International Journal of Industrial Organization 27(6) 2009, pp.639–45.
Levenstein, M.C. and V.Y. Suslow ‘What determines cartel success?’, Journal of
Economic Literature 44(1) 2006, pp.43–95.
Motta M. and M. Polo ‘Leniency programs and cartel prosecution’, International
Journal of Industrial Organization 21(3) 2003, pp.347–79.
Chapter 6
Various articles in the Journal of Economic Perspectives 15(3) 2001; available
online at: www.aeaweb.org/jep/issues/phh
Various articles in the Journal of Industry, Competition and Trade 1(1) 2001.
Chapter 7
Caves R.E. and D.P. Greene ‘Brands’ quality levels, prices, and advertising
outlays’, International Journal of Industrial Organization 14(1) 1996,
pp.29–52.
Nelson, P. ‘Advertising as information’, Journal of Political Economy 82(4) 1974,
pp.729–54.
Chapter 8
Gerardi K. and A.H. Shapiro ‘Does competition reduce price dispersion? New
evidence from the airline industry’, Journal of Political Economy 117(1)
2009, pp.1–37.
Stavins, J. ‘Price discrimination in the airline market: The effect of market
concentration’, Review of Economics and Statistics 83(1) 2001, pp.200–02.
Stole, L.A. ‘Price discrimination and competition’, in Armstrong, M. and R.
Chapter 9
Competition Commission New cars: a report on the supply of new motor cars
within the UK, Cm 4660 (London: TSO, 2000).
Monopolies and Mergers Commission (MMC) New motor cars, Cm 1808
(London: HMSO, 1992).
Various readings on competition in the European car industry available online:
ec.europa.eu/competition/sectors/motor_vehicles/overview_en.html
152
Appendix 2: End-of-chapter Activities reading list
Chapter 10
Berry, S. and J. Waldfogel ‘Product quality and market size’, Journal of
Industrial Economics 58(1) 2010, pp.1–31.
Ellickson, P.B. ‘Does Sutton apply to supermarkets?’, Rand Journal of Economics
38(1) 2007, pp.43–59.
Sutton, J. Sunk costs and market structure. (Cambridge, MA: MIT Press, 1991).
Various chapters.
Sutton, J. ‘Market structure: Theory and evidence’, in Armstrong, M. and R.
Symeonidis, G. ‘Price competition and market structure: The impact of cartel
policy on concentration in the UK’, Journal of Industrial Economics 48(1)
2000, pp.1–26.
Various reviews of Sutton’s books: Sunk costs and market structure (Cambridge,
MA: MIT Press, 1991) and Technology and market structure (Cambridge,
MA: MIT Press, 1998).
Chapter 11
Hall B., C. Helmers, M. Rogers and V. Sena ‘The choice between formal and
informal intellectual property: A literature review’, Journal of Economic
Literature 52(2) 2014, pp.375–423.
Kwoka, J.E. and L.J. White (eds) The antitrust revolution: Economics,
competition, and policy. (Oxford: Oxford University Press, 2013).
University Press, 2009).
Various articles in the Journal of Economic Perspectives 27(1) 2013.
Chapter 12
Armstrong, M. and D.E.M. Sappington ‘Regulation, competition, and
liberalization’, Journal of Economic Literature 44(2) 2006, pp.325–66.
Borenstein, S. ‘The trouble with electricity markets: Understanding California’s
restructuring disaster’, Journal of Economic Perspectives 16(1) 2002,
pp.191–211.
Joskow. P.L. ‘Restructuring, competition and regulatory reform in the U.S.
electricity sector’, Journal of Economic Perspectives 11(3) 1997, pp.119–38.
Joskow, P.L. ‘California’s electricity crisis’, Oxford Review of Economic Policy
17(3) 2001, pp.365–88.
House of Commons Transport Committee The regulation of licensed taxi and
PHV services in the UK. (London: Stationery Office, 2003).
Organisation of Economic Co-operation and Development (OECD) Taxi services
regulation and competition. (OECD, 2007),
153
Notes
154

EC3099

Uploaded by

Copyright:

Available Formats

EC3099

Uploaded by

Document Information

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

EC3099

Uploaded by

Copyright:

Available Formats

Undergraduate study in Economics,

Management, Finance and the Social Sciences

Published by: University of London

Chapter 4: Short-run price competition................................................................ 39

A reminder of your learning outcomes......................................................................... 100

Chapter 5................................................................................................................... 152

Route map to the subject guide

Introduction to the subject area

Aims and objectives

Knowledge required for this subject

The structure of this subject guide

Chapter 7 addresses the specific issues associated with product

Overview of learning resources

Journal articles and book chapters

and contains a good discussion of the modern literature on market

Online study resources

Making use of the Online Library

Chapter 1: Review of game theory

(2,2) (5,3) (3,5) (0,0)

Figure 1.2: Strategic form of Figure 1.1.

Confess −5, −5 −1, −10

Figure 1.3: Representation in strategic form and Nash equilibrium.

Heads +1, –1 –1, +1

Tails –1, +1 +1, –1

Figure 1.4: Matching pennies.

Player 1’s expected utility from Player 2’s mixed strategy

Player 2’s expected utility from Player 2’s mixed strategy

1–x T –1, +1 +1, –1

In equilibrium EUi(H) = EUi(T) for both players, so this condition

Dynamic game with finite horizon

(–5, –5) (–1, –10) (–10, –1) (–2, –2)

(–5, –5) (–1, –10) (–10, –1) (–2, –2)

Figure 1.7: The start of a subgame (highlighted with a red dot).

Step 1: Player 2 chooses W in the first subgame.

NE (1) is subgame perfect (S, [W if S, S if W])

(2, 2) (5, 3) (3, 5) (0, 0)

Step 2: Player 2 chooses S in the second subgame.

NE (1) is subgame perfect (S, [W if S, S if W])

(2, 2) (5, 3) (3, 5) (0, 0)

Step 3: Player 1 chooses S in the third subgame.

NE (1) is subgame perfect (S, [W if S, S if W])

(2, 2) (5, 3) (3, 5) (0, 0)

Dynamic games with infinite horizon

so in this period there is no incentive to cooperate (there will be reward

Remaining quiet is possible in equilibrium if:

A reminder of your learning outcomes

Sample examination questions

2. A philanthropist has to choose whether to contribute to University Z or

Chapter 2: Size and structure of firms

Transaction costs, incomplete contracts and integration

The buyer will choose x to maximise this, which gives:

General conclusions of the property rights approach

Lowering transaction costs and mitigating opportunistic behaviour are

proxy for the know-how generated in the production of a component), a

A reminder of your learning outcomes

Sample examination questions

Chapter 3: Separation of ownership and

Nickell, S. ‘Competition and corporate performance’, Journal of Political

same is true when effort is unobservable, because unobservability reduces

Limits to managerial discretion