Managerial economics

D.J. Reyniers
MN3028, 2790028
2011

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
International Programmes 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 about the University of London International Programmes
undergraduate study in Economics, Management, Finance and the Social Sciences, see:
www.londoninternational.ac.uk
This guide was prepared for the University of London International Programmes by:
Prof D.J. Reyniers, Director, Interdisciplinary Institute of Management, and Professor of
Management, London School of Economics and Political Science, University of London.
This is one of a series of subject guides published by the University. We regret that due to
pressure of work the author is unable to enter into any correspondence relating to, or arising
from, the guide. If you have any comments on this subject guide, favourable or unfavourable,
please use the form at the back of this guide.

University of London International Programmes

Publications Office
Stewart House
32 Russell Square
London WC1B 5DN
United Kingdom
Website: www.londoninternational.ac.uk

Published by: University of London

© University of London 2006
Reprinted with minor revisions 2012
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 contact copyright holders. If you think we have inadvertently used
your copyright material, please let us know.
 Contents

Contents

Introduction ............................................................................................................ 1
Aims and objectives ....................................................................................................... 1
Subject guide breakdown .............................................................................................. 1
Learning outcomes ........................................................................................................ 2
Reading ........................................................................................................................ 2
Mathematics for managerial economics ......................................................................... 3
Online study resources ................................................................................................... 3
Examination advice........................................................................................................ 4
Some advice and ideas on how to study ......................................................................... 5
Glossary of abbreviations ............................................................................................... 5
Chapter 1: Decision analysis................................................................................... 7
Aims ............................................................................................................................. 7
Learning outcomes ........................................................................................................ 7
Essential reading ........................................................................................................... 7
Introduction .................................................................................................................. 7
Decision trees ................................................................................................................ 8
Attitude towards risk ................................................................................................... 10
Some applications ....................................................................................................... 12
The expected value of perfect information .................................................................... 14
A reminder of your learning outcomes.......................................................................... 17
Sample exercises ......................................................................................................... 17
Chapter 2: Game theory ....................................................................................... 19
Aims ........................................................................................................................... 19
Learning outcomes ...................................................................................................... 19
Essential readings ........................................................................................................ 19
Further reading............................................................................................................ 19
Introduction ................................................................................................................ 20
Extensive form games .................................................................................................. 21
Normal form games ..................................................................................................... 23
Nash equilibrium ......................................................................................................... 25
Prisoners’ dilemma ...................................................................................................... 29
Perfect equilibrium....................................................................................................... 30
Perfect Bayesian equilibrium ........................................................................................ 32
A reminder of your learning outcomes.......................................................................... 34
Sample exercises ......................................................................................................... 34
Chapter 3: Bargaining........................................................................................... 37
Aims ........................................................................................................................... 37
Learning outcomes ...................................................................................................... 37
Essential reading ......................................................................................................... 37
Further reading............................................................................................................ 37
Introduction ................................................................................................................ 37
The alternating offers bargaining game ........................................................................ 38
Incomplete information bargaining .............................................................................. 39
A reminder of your learning outcomes.......................................................................... 40
Sample exercise ........................................................................................................... 40

i
28 Managerial economics

Chapter 4: Asymmetric information ..................................................................... 41

Aims ........................................................................................................................... 41
Learning outcomes ...................................................................................................... 41
Essential reading ......................................................................................................... 41
Further reading ........................................................................................................... 41
Introduction ................................................................................................................ 42
Adverse selection ........................................................................................................ 42
Moral hazard ............................................................................................................... 45
Signalling and screening .............................................................................................. 47
Principal-agent problems ............................................................................................. 50
Effort cannot be observed ............................................................................................ 52
A reminder of your learning outcomes.......................................................................... 54
Sample exercises ......................................................................................................... 54
Chapter 5: Auction and bidding............................................................................ 57
Aims ........................................................................................................................... 57
Learning outcomes ...................................................................................................... 57
Essential reading ......................................................................................................... 57
Further reading............................................................................................................ 57
Introduction ................................................................................................................ 58
Private and common value auctions ............................................................................. 59
Private value auctions and their ‘optimal’ bidding strategies ........................................ 60
Auction revenue .......................................................................................................... 65
Common value auctions .............................................................................................. 65
Complications and concluding remarks ........................................................................ 67
Conclusion .................................................................................................................. 71
A reminder of your learning outcomes.......................................................................... 71
Sample exercises ......................................................................................................... 71
Chapter 6: Topics in consumer theory .................................................................. 73
Aims ........................................................................................................................... 73
Learning outcomes ...................................................................................................... 73
Essential reading ......................................................................................................... 73
Further reading............................................................................................................ 73
Introduction ................................................................................................................ 74
Reviewing consumer choice ......................................................................................... 74
Consumer welfare effects of a price change ................................................................. 78
Elasticity ...................................................................................................................... 79
State-contingent commodities model ........................................................................... 81
Intertemporal choice .................................................................................................... 83
Labour supply .............................................................................................................. 86
Risk and return ............................................................................................................ 90
A reminder of your learning outcomes.......................................................................... 93
Sample exercises ......................................................................................................... 94
Chapter 7: Production, factor demands and costs ............................................... 97
Aims ........................................................................................................................... 97
Learning outcomes ...................................................................................................... 97
Essential reading ......................................................................................................... 97
Further reading............................................................................................................ 97
Introduction ................................................................................................................ 97
Production functions and isoquants ............................................................................. 98
Firm demand for inputs.............................................................................................. 101

ii
 Contents

Case: monopsony and minimum wages...................................................................... 105

Industry demand for inputs ........................................................................................ 106
From production function to cost function .................................................................. 109
Division of output among plants ................................................................................ 110
Estimation of cost functions ....................................................................................... 112
A reminder of your learning outcomes........................................................................ 113
Sample exercises ....................................................................................................... 113
Chapter 8: Topics in labour economics ............................................................... 115
Aims ......................................................................................................................... 115
Learning outcomes .................................................................................................... 115
Further reading.......................................................................................................... 115
Introduction .............................................................................................................. 116
Efficiency wages ........................................................................................................ 116
Case: Politicians, sleaze and efficiency wages ............................................................. 119
Firm demand for labour ............................................................................................. 121
Internal labour markets .............................................................................................. 123
Why do wages rise over a career path? ...................................................................... 123
Managerial and executive pay.................................................................................... 127
A reminder of your learning outcomes........................................................................ 128
Sample exercises ....................................................................................................... 129
Chapter 9: Market structure ............................................................................... 131
Aims ......................................................................................................................... 131
Learning outcomes .................................................................................................... 131
Essential reading ....................................................................................................... 131
References cited ........................................................................................................ 131
Introduction .............................................................................................................. 132
Determinants of market structure ............................................................................... 132
Strategy of incumbents .............................................................................................. 133
Measures of market structure..................................................................................... 135
Perfect competition.................................................................................................... 136
Case: Competition in the insurance industry ............................................................... 137
Monopoly.................................................................................................................. 139
Monopolistic competition .......................................................................................... 141
A reminder of your learning outcomes........................................................................ 144
Sample exercises ....................................................................................................... 144
Chapter 10: Monopolistic pricing practices ....................................................... 147
Aims ......................................................................................................................... 147
Learning outcomes .................................................................................................... 147
Essential reading ....................................................................................................... 148
Further reading.......................................................................................................... 148
Introduction .............................................................................................................. 149
Price discrimination ................................................................................................... 149
Second degree price discrimination ............................................................................ 152
Third degree price discrimination................................................................................ 154
Case: The European car market .................................................................................. 157
Case: Yield management............................................................................................ 161
Commodity bundling ................................................................................................. 163
Multiproduct firms ..................................................................................................... 167
Transfer pricing .......................................................................................................... 171
Case: Transfer pricing and taxation ............................................................................. 176

iii
28 Managerial economics

A reminder of your learning outcomes........................................................................ 177

Sample exercises ....................................................................................................... 178
Chapter 11: Oligopoly ........................................................................................ 181
Aims ......................................................................................................................... 181
Learning outcomes .................................................................................................... 181
Essential reading ....................................................................................................... 181
Further reading.......................................................................................................... 181
Introduction .............................................................................................................. 182
Strategic asymmetry .................................................................................................. 185
Stackelberg ............................................................................................................... 185
Symmetric models ..................................................................................................... 188
Bertrand .................................................................................................................... 190
Collusion ................................................................................................................... 192
Case: OPEC ............................................................................................................... 194
Case: Diamonds ........................................................................................................ 195
Dynamic interaction ................................................................................................... 195
Case: Aluminium ....................................................................................................... 197
Conclusion and extensions......................................................................................... 198
A reminder of your learning outcomes........................................................................ 199
Sample exercises ....................................................................................................... 199
Appendix 1: Sample examination paper ............................................................ 201
Appendix 2 References for older versions of textbooks .................................... 207
Chapter and page references for older books ............................................................. 207
Appendix 3: Maths checkpoints ......................................................................... 211
1. Functions – a few general remarks ......................................................................... 211
2. Differentiation ....................................................................................................... 212
3. Logarithmic functions / properties of In and exp...................................................... 214
4. Integration ............................................................................................................ 214
5. Systems of equations / manipulating equations ...................................................... 215
6. Uniform distribution............................................................................................... 216
7. Probabilities ......................................................................................................... 216
8. The discount factor ‘δ’............................................................................................ 218
9. The company’s profit function................................................................................. 219
General suggestions for studying and revising ............................................................ 220
Overview and definitions of some important functions................................................ 220
10. A few hints on solving questions .......................................................................... 221

iv
 Introduction

Introduction

Aims and objectives

This course is intended as an intermediate economics paper for BSc
(Management) and BSc (Economics) students. As such, it is less theoretical
than a microeconomic principles course and more attention is given to
topics which are relevant to managerial decision-making. For instance,
business practices such as transfer pricing, tied sales, resale price
maintenance and exclusive dealing are analysed. Topics are presented
using equations and numerical examples, that is, an analytical approach
is used. The theories which are presented are not practical recipes; they are
meant to give you insight and train your mind to think like an economist.
Specification of the course are to:
• enable you to approach managerial decision problems using economic
reasoning
• present business practice topics using an analytical approach, using
equations and numerical insight.
Why should economics and management students study economics?
The environment in which modem managers operate is an increasingly
complex one. It cannot be navigated without a thorough understanding
of how business decisions are and should be taken. Intuition and factual
knowledge are not sufficient. Managers need to be able to analyse, to put
their observations into perspective and to organise their thoughts in a
rigorous, logical way. The main objective of this paper is to enable you to
approach managerial decision problems using economic reasoning. At the
end of studying this subject, you should have acquired a sufficient level
of model-building skills to analyse microeconomic situations of relevance
to managers. The emphasis is therefore on ‘learning by doing’ rather than
reading and essay writing. You are strongly advised to practise
problem-solving to complement and clarify your thinking.

Subject guide breakdown

The coverage in this subject guide is close to what I teach my second year
BSc (Management) students at the London School of Economics:
• basic microeconomics (i.e. consumer theory, labor supply, neoclassical
theory of the firm, factor demands, competitive structure, government
intervention, etc)
• some newer material regarding efficiency wages, incentive structures,
human resource management, etc
• individual (one person) decision-making under uncertainty and the
value of information; the theory of games. The latter considers strategic
decision-making with more than one player and has applications to
bargaining, bidding and auctions, oligopoly and collusion
• the effects of asymmetric information (one decision-maker has more
information than the other)
• Akerlof’s ‘lemon’ model which explains the disappearance of markets
due to asymmetric information
• situations of moral hazard (postcontractual opportunism) and adverse
selection (precontractual opportunism).
1
28 Managerial economics

Learning outcomes
At the end of the course, and having completed the Essential reading and
exercises, you should be able to:
• prepare for Marketing and Strategy courses by being able to analyse
and discuss consumer behaviour and markets in general
• analyse business practices with respect to pricing and competition
• define and apply key concepts in decision analysis and game theory.

Reading
Essential reading
This guide is intended for intermediate level courses on economics for
management. It is more self-contained than other subject guides might
be. Having said this, I do want to encourage you to read widely from the
recommended reading list. Seeing things explained in more than one way
should help your understanding. In addition to studying the material,
it is essential that you practise problem-solving. Each chapter contains
some sample questions and working through these and problems in the
recommended texts is excellent preparation for success in your studies.
You should also attempt past examination questions from recent years;
these are available online. Throughout this guide, I will recommend
appropriate chapters in the following books.
Tirole, J. The theory of Industrial Organisation. (Cambridge, Mass: The MIT
Press, 1988) [ISBN 9780262200714] this is excellent for the second part of
the guide dealing with pricing practices and strategic interactions.
Varian, H.R. Intermediate Microeconomics. (New York: W.W. Norton and Co.,
2006) seventh edition [ISBN 9780393928624] this can be useful mainly
for the review part of the course and those who prefer a less mathematical
treatment
Note on older editions: most of the relevant material from Varian
(2006) can also be found in the sixth edition of this book. Relevant
chapters for older edition are listed in Appendix 2.
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.

Further reading
We list a number of journals throughout the subject guide which you may
find it interesting to read.
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).

2
 Introduction

Mathematics for managerial economics

The mathematical appendix gives an indication of the level of mathematics
required for the course. Review it! If you are having difficult with the
mathematics in this course, you might find the following book useful.
Ivan, P.N.G. Managerial Economics. (London: Blackwell Publishers, 2001/02)
second edition [ISBN 9780631225164 (hbk 2001), 97801405160476
(pbk 2007)].
It does not assume a prior knowledge of economics and offers a less
mathematical introduction to managerial economics. However, it is
recommended as preliminary reading only and should not be used as
a substitute for the subject guide. The level of mathematics it uses is
much lower than that required for this course and it does not cover the
topics in detail. The different models are treated in a very basic way. It
can, however, be a useful back-up reference if you don’t have the basic
knowledge required to understand a topic. In particular, if you use
this book, it is recommended that you work through the mathematical
appendices which are provided at the end of each chapter. It is not
necessary to study the entire book.

Online study resources

In addition to the subject guide and the Essential reading, it is crucial that
you take advantage of the study resources that are available online for this
course, including the VLE and the Online Library.
You can access the VLE, the Online Library and your University of London
email account via the Student Portal at:
http://my.londoninternational.ac.uk
You should receive your login details in your study pack. If you have not,
or you have forgotten your login details, please email uolia.support@
london.ac.uk quoting your student number.

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:
• Self-testing activities: Doing these allows you to test your own
understanding of subject material.
• Electronic study materials: The printed materials that you receive from
the University of London are available to download, including updated
reading lists and references.
• Past examination papers and Examiners’ commentaries: These provide
advice on how each examination question might best be answered.
• A student discussion forum: This is an open space for you to discuss
interests and experiences, seek support from your peers, work
collaboratively to solve problems and discuss subject material.
• Videos: There are recorded academic introductions to the subject,
interviews and debates and, for some courses, audio-visual tutorials
and conclusions.

3
28 Managerial economics

• Recorded lectures: For some courses, where appropriate, the sessions

from previous years’ Study Weekends have been recorded and made
available.
• Study skills: Expert advice on preparing for examinations and
developing your digital literacy skills.
• Feedback forms.
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 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: http://tinyurl.com/
ollathens
The easiest way to locate relevant content and journal articles in the
Online Library is to use the Summon search engine.
If you are having trouble finding an article listed in a reading list, try
removing any punctuation from the title, such as single quotation marks,
question marks and colons.
For further advice, please see the online help pages: www.external.shl.lon.
ac.uk/summon/about.php

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
we strongly advise you to always check both the current Regulations for
relevant information about the examination, and the VLE where you
should be advised of any forthcoming changes. You should also carefully
check the rubric/instructions on the paper you actually sit and follow
those instructions.
I want to emphasise again that there is no substitute for practising
problem solving throughout the year. It is impossible to acquire a
reasonable level of problem solving skills while revising for the exam. The
examination lasts for three hours and you may use a calculator. Detailed
instructions are given on the examination paper. Read them carefully! The
questions within each part carry equal weight and the amount of time
you should spend on a question is proportional to the marks it carries.
Part A consists of compulsory, relatively short problems of the type that
accompanies each of the chapters in the guide. Part B is a mixture of some
essay type questions and longer analytical questions. In part B a wider
choice is usually available. Appendix 1 contains the exam I set in 1995 for
my second year undergraduates at LSE and indicates the level and type of
questions you should expect.
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
4 answered.
 Introduction

Some advice and ideas on how to study

(This information was originally written by Tell Fallrath as an
Appendix to the guide in 2000.)
• Check your understanding of each model in three ways:
i. can you explain its main arguments in a few simple words?
ii. can you solve a basic model analytically?
iii. can you draw the corresponding graphs?
• Do you understand the models and solutions intuitively?
• What is the overall context of a particular model that you study? How
does it relate to what you already know?
• Try to summarise each topic yourself, perhaps by using mind-maps.
Remind yourself of the examples that you have studied and how they
relate to the theory of each chapter.
• Don’t memorise, but understand! There is hardly anything that you
will need to memorise for this course. Instead make sure you feel
comfortable with the exercises.
• Instead of reading the chapter over and over again, practise problem
solving!
• Often students complain that there are not enough practise questions.
You can create an infinite number of questions yourself by changing
given examples slightly, i.e. change a Marginal Cost function from
MC = 10 to MC = 20. Check how the results change and that you
understand why they change in this particular way. For graphical
questions, change an assumption and see how the graph changes.
• Don’t blame any deficits in mathematics on the economics course!
Address any difficulties you may have with the mathematics throughout
the year as soon as they arise.
• Be curious and get help if you don’t understand something. If you are
studying at an institution, for example, ask for help from your fellow
students, teachers or past-year students.
• Relate the models to your day-to-day experience, which is much
more rewarding and fun. Examples give powerful illustrations of how
economic theory works. After all, economics is not an abstract science,
but models the world around you.

Glossary of abbreviations
Following is a list of abbreviations which are used throughout this subject
guide:
CE certainty equivalent
CEO Chief Executive Officer
CR concentration ratios
CRS computer reservation system
CV compensating variation
DGFT Director-General of Fair Trading
DM downstream monopolist
EU expected utility, European Union
EV equivalent variation, expected value
5
28 Managerial economics

EVPI expected value of perfect information

MFCG fast-moving consumer goods
HHI Herfindahl-Hirschman-Index
LRAC long-run average cost
MC marginal cost
ME marginal expectation
MES minimal effect scale
MFC most favoured customer
MMC Monopolies and Mergers Commission
MP marginal product
MR marginal revenue
MRP marginal revenue product
MRS marginal rate substitution
PC personal computer
RD residual demand
RMS revenue management system
RPM resale price maintenance
SMC sum of marginal cost
TT profit
U utility
UM upstream monopolist

6
 Chapter 1: Decision analysis

Chapter 1: Decision analysis

Aims
The aim of this chaper is to consider:
• the concept of EVPI and how it can be used
• why we may want to use expected utility rather than expected value
maximisation
• the concept of certainty equivalent and how it relates to expected
value for a risk loving, risk neutral and risk hating decision-maker
• the application of decision analysis in insurance and finance.

Learning outcomes
By the end of this chapter, and having completed the Essential readings
and exercises, you should be able to:
• structure simple decision problems in decision tree format and
derive optimal decision
• calculate risk aversion coefficients
• calculate EVPI for risk neutral and non risk neutral decision-makers.

Essential reading
Varian, H.R. Intermediate Microeconomics. [ISBN 0393928024] Chapter 12.

Introduction
It seems appropriate to start a course on economics for management with
decision analysis. Managers make decisions daily regarding selection of
suppliers, budgets for research and development, whether to buy a certain
component or produce it in-house and so on. Economics is in some sense
the science of decision-making. It analyses consumers’ decisions on which
goods to consume, in which quantities and when, firms’ decisions on the
allocation of production over several plants, how much to produce and
how to select the best technology to produce a given product. The bulk
of economic analysis however considers these decision problems in an
environment of certainty. That is, all necessary information is available to
the agents making decisions. Although this is a justifiable simplification,
in reality of course most decisions are made in a climate of (sometimes
extreme) uncertainty. For example, a firm may know how many employees
to hire to produce a given quantity of output but the decision of whether
or how many employees to lay off during a recession involves some
estimate of the length of the recession. Oil companies take enormous
gambles when they decide to develop a new field. The cost of drilling for
oil, especially in deep water, can be over US$1 billion and the pay-offs in
terms of future oil price are very uncertain. Investment decisions would
definitely be very much easier if uncertainty could be eliminated. Imagine
what would happen if you could forecast interest rates and exchange rates
with 100 per cent accuracy.
Clearly we need to understand how decisions are made when at least some
of the important factors influencing the decision are not known for sure.
The field of decision analysis offers a framework for studying how these
7
28 Managerial economics

types of decisions are made or should be made. It also provides insight

into the cost of uncertainty or, in other words, how much a decision-maker
is or should be prepared to pay to reduce or eliminate the uncertainty. To
illustrate the concept of value of information, consider the problem of a
company bidding for a road maintenance contract. The costs of the project
are unknown and the company does not know how low to bid to get the
job. An important question to be answered in preparing the bid is whether
to gather more information about the nature of the project and/or the
competitors’ bids. These efforts only pay-off if, as a result, better decisions
are taken.
Decision analysis is mainly used for situations in which there is one
decision-maker whereas game theory deals with problems in which
there are several decision-makers, each pursuing their own objectives.
In decision analysis any form of uncertainty can be modelled including
that arising from unknown features of competitors’ behaviour (as in the
bidding example). However, when decision analysis models are used to
solve problems with several decision-makers, the competitors are not
modelled as rational agents (i.e. it is not recognised that they are also
trying to achieve certain objectives, taking the actions of other players
into account). Instead, decision theory takes the view that, as long as
probabilities can be attached to other decision-makers’ actions, optimal
decisions can be calculated. An obvious objection to this approach is that it
is not clear how these probabilities become known to the decision-maker.
Game theory avoids this problem as it takes a symmetric, simultaneous
view. The reason decision analysis is used in these situations despite
these shortcomings is that it is much simpler than game theory. For this
reason and because some of the techniques of decision analysis (such as
representing sequential decision problems on graphs or decision trees, and
solving them backwards) can be used in game theory, we study decision
analysis first.

Decision trees
A decision tree is a convenient representation of a decision problem. It
contains all the ingredients of the problem:
• the decisions
• the sources of uncertainty
• the pay-offs which are the results, in terms of the decision-maker’s
objective, for each possible combination of probabilistic outcomes and
decisions.
Drawing a decision tree forces the decision-maker to think through
the structure of the problem s/he faces and often makes the process of
determining optimal decisions easier. A decision tree consists of two
kinds of nodes: decision or action nodes which are drawn as squares and
probability or chance nodes drawn as circles. The arcs leading from a
decision node represent the choices available to the decision-maker at
this point whereas the arcs leading from a probability node correspond
to the set of possible outcomes when some uncertainty is resolved. When
the structure of the decision problem is captured in a decision tree, the
pay-offs are written at the end of the final branches and (conditional)
probabilities are written next to each arc leading from a probability node.
The algorithm for finding the optimal decisions is not difficult. Starting
at the end of the tree, work backwards and label nodes as follows. At a
probability node calculate the expected value of the labels of its successor
nodes, using the probabilities given on the arcs leading from the node.
8
 Chapter 1: Decision analysis

This expected value becomes the label for the probability node. At a
decision node x (assuming a maximisation problem), select the maximum
value of the labels of successor nodes. This maximum becomes the label
for the decision node. The decision which generates this maximum value
is the optimal decision at this node. Repeat this procedure until you reach
the starting node. The label you get at the starting node is the expected
pay-off obtained when the optimal decisions are taken. The construction
and solution of a decision tree is most easily explained through examples.

Example 1.1

Cussoft Ltd., a firm which supplies customised software, must decide

between two mutually exclusive contracts, one for the government and
the other for a private firm. It is hard to estimate the costs Cussoft will
incur under either contract but, from experience, it estimates that, if it
contracts with a private firm, its profit will be £2 million, £0.7 million,
or –£0.5 million with probabilities 0.25, 0.41 and 0.34 respectively. If
it contracts with the government, its profit will be £4 million or –£2.5
million with respective probabilities 0.45 and 0.55. Which contract
offers the greater expected profit?
In this very simple example, Cussoft has a choice of two decisions – to
contract with the private firm or to contract with the government. In
either case its pay-off is uncertain. The decision tree with the pay-offs
and probabilities is drawn in Figure 1.1. The expected profit if the
contract with the private firm is chosen equals (0.25)(2) + (0.41)(0.7) +
(0.34)(–0.5) = 0.617 (£ million) whereas the contract with the government
delivers an expected profit of (0.45)(4) + (0.55)(–2.5) = 0.425 (£ million) so
that the optimal decision is to go for the contract with the private firm.
Optimal decisions are indicated by thick lines.

0.617 0.25 2
0.41
0.7
private firm
0.34

-0.5

4
0.617
0.45
government 0.425
0.55
-2.5

Figure 1.1: Decision tree for example 1.1

Example 1.2

Suppose the Chief Executive of an oil company must decide whether to

drill a site and, if so, how deep. It costs £160,000 to drill the first 3,000
feet and there is a 0.4 chance of striking oil. If oil is struck, the profit (net of
drilling expenses) is £600,000. If she doesn’t strike oil, the executive can
drill 2,000 feet deeper at an additional cost of £90,000. Her chance of
finding oil between 3,000 and 5,000 feet is 0.2 and her net profit (after
all drilling costs) from a strike at this depth is £400,000. What action

9
28 Managerial economics

should the executive take to maximise her expected profit? Try writing
down and solving the decision tree yourself without peeking! You should
get the following result.

600
168 oil
400
0.4 oil
drill
3000 ft -120
0.6
no oil drill
5000 ft no oil
168
-250

don’t drill -120 don’t drill

0
-160

Figure 1.2: Decision tree for example 1.2

Attitude towards risk

In the examples considered so far we have used the expected monetary
value (EMV) criterion (i.e. we assumed that the decision-maker is
interested in maximising the expected value of profits or minimising
the expected value of costs). In many circumstances this is a reasonable
assumption to make, especially if the decision-maker is a large company.
To appreciate that it may not always be appropriate to use EMV consider
the following story, known as the St. Petersburg paradox. I will toss
a coin and, if it comes up heads, you will get £2. If it comes up tails, I
will toss it again and, if it comes up heads this time, you will get £4; if it
comes up tails, I will toss it again and, this time, you will get £8 if it comes
up heads etc. How much would you be willing to pay for this gamble?
I predict that you would not want to pay your week’s pocket money or
salary to play this game. However, if you calculate the EMV you find:
EMV = 2(1/2) + 4(1/4) + 8(l/8) + ... + 2n(1/2n) +... = 1 + l + 1 + .. . = ∞!
Even when faced with potentially large gains, most people do not like to
risk a substantial fraction of their financial resources. Although this implies
that we cannot always use EMV, it is still possible to give a general analysis
of how people make decisions even if they do not like taking risks. As a
first step we have to find out the decision-maker’s attitude towards risk. A
useful concept here is the certainty equivalent (CE) of a risky prospect
defined as the amount of money which makes the individual indifferent
between it and the risky prospect. To clarify this, imagine you are offered
a lottery ticket which has a 50–50 chance of winning £0 or £200. Would
you prefer £100 for sure to the lottery ticket? What about £50 for sure?
The amount £x so that you are indifferent between x and the lottery ticket
is your certainty equivalent of the lottery ticket. If your x is less than £100,
the EMV of the lottery, you are ‘risk averse’. In general a decision-maker
is risk averse if CE < EMV, risk neutral if CE = EMV and risk loving if
CE > EMV. Suppose you have an opportunity to invest £1,000 in a business
venture which will gross £1,100 or £1,200 with equal probability next year.
Alternatively you could deposit the £1,000 in bank which will give you a
riskless return. How large does the interest rate have to be for you to be
indifferent between the business venture and the deposit account (i.e. what

10
 Chapter 1: Decision analysis

is your certainty equivalent? Are you a risk lover?)? Note that it is possible to
be a risk lover for some lotteries and a risk hater for others.

By asking these types of questions, we can determine a decision-maker’s

degree of risk aversion summarised in his/her utility of money
function. This enables us to still use expected value calculations but
with monetary outcomes replaced by utility values (i.e we can use the
expected utility criterion). It is possible to show that, if a decision-
maker satisfies certain relatively plausible axioms, he can be predicted
to behave as if he maximises expected utility. Furthermore, since a utility
function U*(x) = aU(x) + b, a > 0, leads to the same choices as U(x) we can
arbitrarily fix the utility of the worst outcome w at 0(U(w) = 0) and the
utility of the best outcome b at l(U(b) = l) for a given decision problem.
To find the utility corresponding to an outcome x we ask the decision-
maker for the value of p, the probability of winning b in a lottery with
prizes b and w, which makes x the CE for the lottery. For example, if the
worst outcome in a decision problem is £0(U(0) = 0) and the best outcome
is £200(U(200) = 1), how do we determine U(40)? We offer the decision-
maker the choice represented in Figure 1.3 and keep varying p until he is
indifferent between 40 and the lottery.

40
200
p

1-p
0

Figure 1.3: Determining utility

When the decision-maker is indifferent, say for p = 0.4, we have:
U(40) = U(lottery) = pU(200) + (1 – p)U(0) = p = 0.4.

Utility values can be obtained in a similar way for the other possible
outcomes of the decision problem. Replacing the monetary values by the
utility values and proceeding as before will lead to the expected utility
maximising decisions.
The definition of risk aversion can be rephrased in terms of the utility
function:
• a risk averse decision-maker has a concave utility function
• a risk lover has a convex utility function
• a risk neutral decision-maker has a linear utility function.

risk averse neutral risk loving

money

Figure 1.4: Attitudes towards risk

11
28 Managerial economics

It is possible for a decision-maker to be risk averse over a range of

outcomes and risk loving over another range. Indeed, this is how we
can explain that the same people who take out home contents insurance
buy a national lottery ticket every week. An example of a utility function
corresponding to risk loving behaviour for small bets and risk averse
behaviour for large bets is drawn in Figure 1.5.

money

Figure 1.5: Risk loving and risk averse behaviour

For continuous differentiable functions, there are two measures of risk
aversion, the Pratt-Arrow coefficient of absolute risk aversion
determined as – U'' (x)/U' (x) and the Pratt-Arrow coefficient of relative
risk aversion determined as – U′′ (x)x/ U'(x). If U is concave U′′< 0 and
hence both these coefficients are positive for risk averse individuals.1 1
Note that the
coefficient of relative
risk aversion is the
Some applications negative of the elasticity
of marginal utility of
The demand for insurance income and does not
depend on the units
People take out insurance policies because they do not like certain types in which income is
of risk. Let us see how this fits into the expected utility model. Assume an measured.
individual has initial wealth W and will suffer a loss L with probability p.
How much would she be willing to pay to insure against this loss? Clearly,
the maximum premium R she will pay makes her just indifferent between
taking out insurance and not taking out insurance. Without insurance she
gets expected utility EU0 = pU(W – L) + (1 – p) U(W ) and, if she insures
at premium R, her utility is U(W – R). Therefore the maximum premium
satisfies U(W – R) = pU(W – L) + (1 – p) U(W ). This is illustrated for a risk
averse individual in Figure 1.6. The expected utility without insurance is a
convex combination of U(W ) and U(W – L) and therefore lies on the straight
line between U(W ) and U(W – L); the exact position is determined by p
so that EU0 can be read off the graph just above W – pL. This utility level
corresponds to a certain prospect W – R which, as can be seen from the
figure, has to be less than W – pL, so that R > pL. This shows that, if a risk
averse individual is offered actuarially fair insurance (premium R equals
expected loss pL), he will insure.

12
 Chapter 1: Decision analysis

EU0

W-L W-R W-pL W

Figure 1.6: Insurance

Example 1.3

Jamie studies at Cambridge University and uses a bicycle to get around.

He is worried about having his bike stolen and considers taking out
insurance against theft. If the bike gets stolen he would have to replace
it which would cost him £200. He finds out that 10 per cent of bicycles
in Cambridge are stolen every year. His total savings are £400 and his
utility of money function is given by U(x) = x1/2. Under what conditions
would Jamie take out insurance for a year? What if he has utility of
money U(x) = ln(x)?
If he takes out insurance he obtains utility U(400 – R) where R is
the premium. Without insurance he gets (0.1)U(200) + (0.9)U(400).
Equalising these expected utilities and substituting U(x) = x1/2, gives:
(0.1)√200 + (0.9) √400 = √400 – R
or
R = 23.09
which means that insuring is the best decision as long as the premium
does not exceed £23.09. Similarly, if U(x) = ln(x), the maximum
premium can be calculated (you should check this!) as £26.79.

The demand for financial assets

Consider the problem of an investor with initial wealth W who wants to
decide on her investment plans for the coming year. For simplicity, let us
assume that there are only two options: a riskless asset which delivers a
gross return of R at the end of the year, and a risky asset which delivers a
high return H with probability p and a low return L with probability 1 – p.
It is not difficult to allow for borrowing so that the investor can invest
more than W but, to keep things simple, let us restrict the investor’s budget
to W. The decision problem then consists of finding the optimal amount of
money A(<W ) to be invested in the risky asset. Given A, the investor gets
an expected return of:
EU(A) = pU(R(W – A) + HA) + (1 – p)U(R(W – A) + LA) = pU(RW + (H – R)A)
+ (1 – p)U(RW + (L – R)A)

13
28 Managerial economics

Maximising EU(A) and assuming an interior solution (i.e. 0 < A < W) leads
to the following (first order) condition:
EU'(A) = pU'(RW + (H – R)A)(H – R) + (1 – p)U'(RW + (L – R)A)(L – R) = 0
Note that, if the risky asset always yields a lower return than the riskless
asset (H, L < R), there can be no solution to this condition since U' > 0.
In this scenario the investor would not invest in the risky asset (A = 0).
Similarly, if the risky asset always yields a higher return than the riskless
asset (H, L > R) there can be no solution to this condition and under these
circumstances the investor would invest all of her wealth in the risky asset
(A = W). For the other scenarios (L < R < H) the first order condition above
allows us, for a specific utility function, to calculate the optimal portfolio.

The expected value of perfect information

In most situations of uncertainty a decision-maker has the possibility of
reducing, if not eliminating, the uncertainty regarding some relevant factor.
Typically this process of finding more information is costly in financial
terms (e.g. when a market research agency is contracted to provide details
of the potential market for a new product) or in terms of effort (e.g. when
you test drive several cars before deciding on a purchase). A crucial question
in many decision-making contexts is therefore, ‘How much money and/
or effort should the decision-maker allocate to reducing the uncertainty?’
In this section, we study a procedure which gives an upper bound to this
question. The upper bound which is the expected value of perfect
information (EVPI) is derived as follows for a risk neutral decision-maker
(we will go back to the expected utility model later).
Assuming you know the outcomes of all probabilistic events in a decision
tree, determine the optimal decisions and corresponding pay-off for each
possible scenario (combination of outcomes at each probability node). Given
that you know the probabilities of each scenario materialising, calculate the
expected pay-off under certainty using the optimal pay-off under each
scenario and the scenario’s probability. This expected pay-off is precisely the
pay-off you would expect if you were given exact information about what
will happen at each probability node. The difference between this expected
pay-off under perfect information and the original optimal pay-off is the
EVPI. Since, in reality, it is almost never possible to get perfect information
and eliminate the uncertainty completely, the EVPI is an upperbound on how
much the decision-maker is willing to pay for any (imperfect) information.
Generally better decisions are made when there is no uncertainty and
therefore the EVPI is positive. However, it is possible that having more
information does not change the optimal decisions and in those cases the
EVPI is zero. While the concept of EVPI is extremely useful, it is really quite
abstract and difficult to grasp. So let us look at an example.

Example 1.2 (continued)

The notion of perfect information in this problem translates into the

existence of a perfect seismic test which could tell you with certainty
whether there is oil at 3000ft., at 5000ft. or not at all. Assuming such
a test exists, how much would the Chief Executive Officer (CEO) be
willing to pay to know the test result? The tree in Figure 1.7 represents
the (easy) decision problem if all uncertainty is resolved before any
decisions are made. There are three scenarios: ‘oil at 3000ft.’, ‘oil at
5000ft’ and ‘no oil’ whose probabilities can be derived from the original
tree as 0.40, 0.12, and 0.48 respectively. If the CEO is told which

14
 Chapter 1: Decision analysis

scenario will occur, her decision will be straightforward. Given the

optimal decision corresponding to each scenario and the probabilities
of the various scenarios, the optimal pay-off with perfect information
is £288,000. Recall that the original problem had an EMV of £168,000
and hence EVPI = £120,000. If a perfect seismic test were available, the
CEO would be willing to pay up to £120,000 for it.

600
drill

oil at 3000 ft 600 don’t 0

0.40
400
oil at
5000 ft drill

0.12
288 400 don’t 0

no oil
0.48

0 don’t drill

Figure 1.7: Calculating EVPI

If the decision-maker is not risk neutral, a similar method to the one we

have just discussed can be used to evaluate the EVPI in utility terms
(i.e. we calculate EU under the assumption of perfect information and
compare this with the EU in the original problem). However, this does not
tell us how much the decision-maker is willing to pay to face the riskless
problem rather than the risky one! It is in fact quite tricky to determine
the EVPI for an EU maximiser. Consider the simple decision tree in Figure
1.8 where an individual with money utility U(x) = x 1/2 chooses between a
safe and a risky strategy, say investing in a particular stock or not. In either
case there are two outcomes – O1 and O2 (e.g. the company is targeted for
takeover or not) – resulting in the monetary pay-offs indicated on the tree.
The probabilities of the outcomes are independent of the decision taken.
Using the EU criterion, the decision-maker chooses the safe strategy so
that EU = 4.

Why?

15
28 Managerial economics

1/3
16
O1
2/3
safe
O2
16

1/3 81

risky O1

2/3

O2 1

Figure 1.8: EU maximiser chooses safe

With perfect information however, the decision-maker chooses the ‘risky’
strategy if O1 is predicted and the ‘safe’ strategy when O2 is predicted, as is
indicated in Figure 1.9. This gives her:
EU = (1/3) (9) + (2/3) (4) = 17/3.

81
risky

O1 safe
1/3 16

O2 1

2/3 risky

16

Figure 1.9: Optimal choice under perfect information

How much is the decision-maker willing to pay for this increase in EU from
4 to 17/3? Suppose she pays an amount R for the perfect information. She
will be indifferent between getting the information and not getting it if
the EU in both cases is equal, or 4 = (1/3)(81 – R)1/2 + (2/3)(16 – R)1/2. Solving
this for R (numerically) gives R approximately 12.5; hence, the EVPI for
this problem is about 12.5. Note that this last equation can be solved
algebraically.

16
 Chapter 1: Decision analysis

A reminder of your learning outcomes

Having completed this chapter, and the Essential readings and exercises, you
should be able to:
• structure simple decision problems in decision tree format and derive
optimal decision
• calculate risk aversion coefficients
• calculate EVPI for risk neutral and non risk neutral decision-makers.

Sample exercises
1. London Underground (LU) is facing a courtcase by legal firm Snook&Co,
representing the family of Mr Addams who was killed in the Kings Cross
fire. LU has estimated the damages it will have to pay if the case goes to
court as follows: £1,000,000, £600,000 or £0 with probabilities 0.2, 0.5 and
0.3 respectively. Its legal expenses are estimated at £100,000 in addition to
these awards. The alternative to allowing the case to go to court is for LU
to enter into out-of-court settlement negotiations. It is uncertain about the
amount of money Snook&Co. are prepared to settle for. They may only
wish to settle for a high amount (£900,000) or they may be willing to
settle for a reasonable amount (£400,000). Each scenario is equally likely. If
they are willing to settle for £400,000 they will of course accept an offer
of £900,000. On the other hand, if they will only settle for £900,000 they
will reject an offer of £400,000. LU, if it decides to enter into negotiations,
will offer £400,000 or £900,000 to Snook & Co. who will either accept (and
waive any future right to sue) or reject and take the case to court. The legal
cost of pursuing a settlement whether or not one is reached is £50,000.
Determine the strategy which minimises LU’s expected total cost.
2. Rickie is considering setting up a business in the field of entertainment
at children’s parties. He estimates that he would earn a gross revenue of
£9,000 or £4,000 with a 50–50 chance. His initial wealth is zero. What is
the largest value of the cost which would make him start this business:

a. if his utility of money function is U(x) = ax + b where a > 0

b. if U(x) = x1/2; for x > 0 and U(x) = –(–x)1/2 for x < 0
c. if U(x) = x2, for x > 0 and U(x) = –x2 for x < 0.
3. Find the coefficient of absolute risk aversion for
U(x) = a – b.exp(–cx) and the coefficient of relative risk aversion for
U(x) = a + b.ln(x).
4. Find a volunteer (preferably someone who doesn’t know expected utility
theory) and estimate their utility of money function to predict their choice
between the two lotteries below. Pay-offs are given in monetary value (£).
Check your prediction.

17
28 Managerial economics

100
2/3

1/3
200

50
1/3

5/9
150

1/9

300

Figure 1.10: Choice between two lotteries

5. An expected utility maximiser spends £10 on a lottery ticket, with a
chance of 1 in 1 million of £1 million. He takes out home contents
insurance at a premium of £100. His probability of an insurance claim
of £1,000 is 1%. Draw his utility of money function.
6. A decision-maker must choose between (1) a sure payment of £200;
(2) a gamble with prizes £0, £200, £450 and £1,000 with respective
probabilities 0.5, 0.3, 0.1 and 0.1; (3) a gamble with prizes £0, £100,
£200 and £520, each with probability 0.25.
a. Which choice will be made if the decision-maker is risk neutral?
b. Assume the decision-maker has a CARA (constant absolute risk
aversion) utility of money function U(x) = –a exp(–cx) + b and her
certainty equivalent for a gamble with prizes £1,000 and £0 equally
likely is £470. Which choice will be made?2 2
Hint: show that c
has to equal 0.00024
7. Henrika has utility function U = M 1/2 for M ≥ 0 and U = –(–M )1/2 for
approximately for
M < 0, over money pay-offs M. U(0) = 0 and
U(1,000) = 1
a. Given a lottery with outcomes £0 and £36 with respective
probabilities 2/3 and 1/3, how much is she willing to pay to replace
the lottery with its expected value?
b. Given the table of money pay-offs below, which action maximises
her expected utility?
c. How much would Henrika be willing to pay for perfect information
regarding the state of nature?
S1 S2
A1 4 4
A2 9 1
1/3 2/3

18
 Chapter 2: Game theory

Chapter 2: Game theory

Aims
The aim of this chaper is to consider:
• the concept of information set and why it is not needed in decision
analysis
• why it is useful to have both extensive form and normal form
representations of a game
• the importance of the prisoners’ dilemma as a paradigm for many social
interactions
• the concept of dominated strategies and the rationale for eliminating
them in analysis of a game
• the concept of Nash equilibrium (this is absolutely essential!)
• the concept of non-credible threats and its application in entry
deterrence.

Learning outcomes
By the end of this chapter, and having completed the Essential readings
and exercises, you should be able to:
• represent a simple multi-person decision problem using a game tree
• translate from an extensive form representation to the normal form
representation
• find Nash equilibria in pure and mixed strategies
• explain why in a finitely repeated prisoners’ dilemma game
cheating is a Nash equilibrium
• explain the chainstore paradox.

Essential readings
Tirole, J. The Theory of Industrial Organization. Chapter 11.
Varian, H.R. Intermediate Microeconomics. Chapters 28 and 29.

Further reading
Axelrod, R. The evolution of cooperation. (New York, Basic Books 1984)
[ISBN 0465021212].
Kreps, D. and R. Wilson ‘Reputation and imperfect information’, Journal of
economic theory (1982) 27, pp.253–79.
Milgrom, P. and J. Roberts ‘Predation, reputation and entry deterrence’, Journal
of economic theory (1982) 27, pp.280–312.
‘Mumbo-jumbo, Super-jumbo’, The Economist, 12 June 1993, p.83
Nash, J. ‘Noncooperative games’, Annals of Mathematics 54 1951, pp.289–95.
‘Now for the really big one’, The Economist, 9 January 1993, p.83.
‘Plane wars’, The Economist, 11 June 1994, pp.61–62.
Selton, R. ‘ The chain store paradox’, Theory and Decision (9) 1978, pp.127–59.
‘The flying monopolists’, The Economist, 19 June 1993, p.18.

19
28 Managerial economics

Introduction
Game theory extends the theory of individual decision-making to
situations of strategic interdependence: that is, situations where players
(decision-makers) take other players’ behaviour into account when making
their decisions. The pay-offs resulting from any decision (and possibly
random events) are generally dependent on others’ actions.
A distinction is made between cooperative game theory and
noncooperative game theory. In cooperative games, coalitions or
groups of players are analysed. Players can communicate and make binding
agreements. The theory of noncooperative games assumes that no such
agreements are possible. Each player in choosing his or her actions, subject
to the rules of the game, is motivated by self-interest. Because of the larger
scope for application of noncooperative games to managerial economics, we
will limit our discussion to noncooperative games.
To model an economic situation as a game involves translating the essential
characteristics of the situation into rules of a game. The following must be
determined:
• the number of players
• their possible actions at every point in time
• the pay-offs for all possible combinations of moves by the players
• the information structure (what do players know when they have to
make their decisions?).
All this information can be presented in a game tree which is the game
theory equivalent of the decision tree. This way of describing the game is
called the extensive form representation.
It is often convenient to think of players’ behaviour in a game in terms of
strategies. A strategy tells you what the player will do each time s/he has
to make a decision. So, if you know the player’s strategy, you can predict
his behaviour in all possible scenarios with respect to the other players’
behaviour. When you list or describe the strategies available to each player
and attach pay-offs to all possible combinations of strategies by the players,
the resulting ‘summary’ of the game is called a normal form or strategic
form representation.
In games of complete information all players know the rules of the game.
In incomplete information games at least one player only has probabilistic
information about some elements of the game (e.g. the other players’ precise
characteristics). An example of the latter category is a game involving an
insurer – who only has probabilistic information about the carelessness of an
individual who insures his car against theft – and the insured individual who
knows how careless he is. A firm is also likely to know more about its own
costs than about its competitors’ costs. In games of perfect information all
players know the earlier moves made by themselves and by the other players.
In games of perfect recall players remember their own moves and do not
forget any information which they obtained in the course of game play. They
do not necessarily learn about other players’ moves.
Game theory, as decision theory, assumes rational decision-makers. This
means that players are assumed to make decisions or choose strategies which
will give them the highest possible expected pay-off (or utility). Each player
also knows that other players are rational and that they know that he knows
they are rational and so on. In a strategic situation the question arises whether
it could not be in an individual player’s interest to convince the other players
that he is irrational. (This is a complicated issue which we will consider in
20
 Chapter 2: Game theory

the later sections of this chapter. All I want to say for now is that ultimately
the creation of an impression of irrationality may be a rational decision.)
Before we start our study of game theory, a ‘health warning’ may be
appropriate. It is not realistic to expect that you will be able to use game
theory as a technique for solving real problems. Most realistic situations are
too complex to analyse from a game theoretical perspective. Furthermore,
game theory does not offer any optimal solutions or solution procedures for
most practical problems. However, through a study of game theory, insights
can be obtained which would be difficult to obtain in another way and game
theoretic modelling helps decision-makers think through all aspects of the
strategic problems they are facing. As is true of mathematical models in
general it allows you to check intuitive answers for logical consistency.

Extensive form games

As mentioned above, the extensive form representation of a game is similar
to a decision tree. The order of play and the possible decisions at each
decision point for each player are indicated as well as the information
structure, the outcomes or pay-offs and probabilities. As in decision analysis
the pay-offs are not always financial. They may reflect the player’s utility
of reaching a given outcome. A major difference with decision analysis is
that in analysing games and in constructing the game tree, the notion of
information set is important. When there is only one decision-maker, the
decision-maker has perfect knowledge of her own earlier choices. In a game,
the players often have to make choices not knowing which decisions have
been taken or are taken at the same time by the other players. To indicate
that a player does not know her position in the tree exactly, the possible
locations are grouped or linked in an information set, Since a player should
not be able to deduce from the nature or number of alternative choices
available to her where she is in the information set, her set of possible
actions has to be identical at every node in the information set. For the
same reason, if two nodes are in the same information set, the same player
has to make a decision at these nodes. In games of perfect information the
players know all the moves made at any stage of the game and therefore all
information sets consist of single nodes.
Example 2.1 presents the game tree for a dynamic game in which Player 2
can observe the action taken by Player 1. Example 2.2 presents the game
tree for a static game in which players take decisions simultaneously.

Example 2.1

t 3,0
T 2
3,-2
b
1

t 1,-1
B
2

b 4,2

Figure 2.1: Game tree for a dynamic game

In this game tree Player 1 makes the first move, Player 2 observes the
choice made by Player 1 (perfect information game) and then chooses
from his two alternative actions. The pay-off pairs are listed at the

21
28 Managerial economics

endpoints of the tree. For example, when Player 1 chooses B and Player
2 chooses t, they receive pay-offs of 1 and –1 respectively. Games of
perfect information are easy to analyse. As in decision analysis, we can
just start at the end of the tree and work backwards (Kuhn’s algorithm).
When Player 2 is about to move and he is at the top node, he chooses t
since this gives him a pay-off of 0 rather than –2 corresponding to b. When
he is at the bottom node, he gets a pay-off of 2 by choosing b. Player 1
knows the game tree and can anticipate these choices of Player 2. He
therefore anticipates a pay-off of 3 if he chooses T and 4 if he chooses B.
We can conclude that Player 1 will take action B and Player 2 will take
action b.
Let us use this example to explain what is meant by a strategy. Player
1 has two strategies: T and B. (Remember that a strategy should state
what the player will do in each eventuality.) For Player 2 therefore, each
strategy consists of a pair of actions, one to take if he ends up at the top
node and one to take if he ends up at the bottom node. Player 2 has four
possible strategies, namely:
• (t if T, t if B)
• (t if T, b if B)
• (b if T, t if B)
• (b if T, b if B)
• or {(t, t), (t, b), (b, t),(b, b)} for short.

Example 2.2

The game tree below is almost the same as in Example 2.1 but here
Player 2 does not observe the action taken by Player 1. In other words,
it is as if the players have to decide on their actions simultaneously.
This can be seen on the game tree by the dashed line linking the two
decision nodes of Player 2: Player 2 has an information set consisting of
these two nodes. This game (of imperfect information) cannot be solved
backwards in the same way as the game of Example 2.1.

t 3,0
T 2

b 3,-2
1

1,-1

B t
2 4,2
b

Figure 2.2: Game tree for a simultaneous move game

Note that, although the game trees in the two examples are very
similar, Player 2 has different strategy sets in the two games. In the
second game his strategy set is just (t, b) whereas in the first game there
are four possible strategies.

22
 Chapter 2: Game theory

Normal form games

A two-person game in normal form with a finite number of strategies
for each player is easy to analyse using a pay-off matrix. The pay-off
matrix consists of r rows and c columns where r and c are the number
of strategies for the row and the column players respectively. The matrix
elements are pairs of pay-offs (pr , pc ) resulting from the row player’s
strategy r and the column player’s strategy c, with the pay-off to the
row player listed first. The normal form representations of the games in
Examples 2.1 and 2.2 are given below.
Player 2
(t, t) (t, b) (b, t) (b, b)
Player 1 T 3,0 3,0 3,–2 3,–2
B 1,–1 4,2 *1,–1 4,2
Normal form for example 2.1

Player 2
t b
Player 1 T 3,0 3,–2
B 1,–1 4,2
Normal form for example 2.2

Example 2.3

Two competing firms are considering whether to buy television time

to advertise their products during the Olympic Games. If only one of
them advertises, the other one loses a significant fraction of its sales.
The anticipated net revenues for all strategy combinations are given in
the table below. We assume that the firms have to make their decisions
simultaneously.

Firm B

Advertise Don’t

Firm A Advertise 10,5 13,2

Don’t 6,7 11,9

If firm A decides to advertise, it gets a pay-off of 10 or 13 depending on

whether B advertises or not. When it doesn’t advertise it gets a pay-off
of 6 or 1 I depending on whether B advertises or not. So, irrespective of
B’s decision, A is better off advertising. In other words, ‘Don’t advertise’
is a dominated strategy for firm A. Given that we are assuming that
players behave rationally, dominated strategies can be eliminated.
Firm B can safely assume that A will advertise. Given this fact, B now
only has to consider the top row of the matrix and hence it will also
advertise. Note that both A and B are worse off than if they could sign a
binding agreement not to advertise.

23
28 Managerial economics

Example 2.4

In the pay-off matrix below, only one pay-off, the pay-off to the row player,
is given for each pair of strategies. This is the convention for zero-sum
games (i.e. games for which the pay-offs to the players sum to zero for
all possible strategy combinations). Hence, the entry 10 in the (1, 1)
position is interpreted as a gain of 10 to the row player and a loss of 10 to
the column player. An application of this type of game is where duopolists
compete over market share. Then one firm’s gain (increase in market
share) is by definition the other’s loss (decrease in market share). In
zero sum games one player (the row player here) tries to maximise his
pay-off and the other player (the column player here) tries to minimise
the pay-off.
If we consider the row player first, we see that the middle row weakly
dominates the bottom row. For a strategy A to strictly dominate a strategy
B we need the pay-offs of A to be strictly larger than those of B against all
of the opponent’s strategies. For weak dominance it is sufficient that the
pay-offs are at least as large as those of the weakly dominated strategy.
In our case, the pay-off of M is not always strictly larger than that of B
(it is the same if Player 2 plays R). If we eliminate the dominated strategy
B, we are left with a 2 × 3 game in which Player I has no dominated
strategies. If we now consider Player 2, we see that C is weakly
dominated by R (remembering that Player 2’s pay-offs are the negative
of the values in the table!) and hence we can eliminate the second
column. In the resulting 2 × 2 game, T is dominated by M and hence we
can predict that (M, R) will be played.

Player 2

L C R

Player 1 T 10 20 –30

M 20 10 10

B 0 –20 10

In general, we may delete dominated strategies from the pay-off matrix

and, in the process of deleting one player’s dominated strategies, we
generate a new pay-off matrix which contains dominated strategies for the
other player which in turn can be deleted and so on. This process is called
successive elimination of dominated strategies.

Activity
Verify that, in the normal form of Example 2.1, this process leads to the outcome we
predicted earlier, namely (B,(t,b)) but that, in the normal form of Example 2.2, there are
no dominated strategies.

Sometimes, as in the example above, we will be left with one strategy pair,
which would be the predicted outcome of the game but the usual scenario
is that only a small fraction of the strategies can be eliminated.

24
 Chapter 2: Game theory

Nash equilibrium
A Nash equilibrium is a combination of strategies, one for each player,
with the property that no player would unilaterally want to change his
strategy given that the other players play their Nash Equilibrium strategies.
So a Nash equilibrium strategy is the best response to the strategies that a
player assumes the other players are using.
Pure strategies are the strategies as they are listed in the normal form of
a game. We have to distinguish these from mixed strategies (which will
be referred to later). The game below has one Nash equilibrium in pure
strategies, namely (T, L). This can be seen as follows. If the row player plays
his strategy T, the best the column player can do is to play his strategy L
which gives him a pay-off of 6 (rather than 2 if he played R). Vice versa, if
the column player plays L, the best response of the row player is T. (T, L) is
the only pair of strategies which are best responses to each other.

L R
T 5, 6 1, 2
B 4, 3 0, 4

In some cases we can find Nash equilibria by successive elimination

of strictly dominated strategies. If weakly dominated strategies are
also eliminated we may not find all Nash equilibria. Another danger of
successively eliminating weakly dominated strategies is that the final
normal form (which may contain only one entry) after elimination may
depend on the order in which dominated strategies are eliminated.

Example 2.5: ‘Battle of the sexes’

The story corresponding to this game is that of a husband and wife who
enjoy the pleasure of each other’s company but have different tastes in
leisure activities. The husband likes to watch football whereas the wife
prefers a night out on the town. On a given night the couple have to
decide whether they will stay in and watch the football match or go out.
The pay-off matrix could look like this.

Wife

In Out

In 10, 5 2, 4
Husband
Out 0, 1 4, 8

This game has two Nash equilibria: (in, in) and (out, out). Only when
both players choose the same strategy is it in neither’s interest to switch
strategies. The battle of the sexes game is a paradigm for bargaining
over common standards.
When electronics manufacturers choose incompatible technologies they
are generally worse off than when they can agree on a standard. For
example, Japanese, US and European firms were developing their own
versions of high definition television whereas they would have received
greater pay-offs if they had coordinated. The computer industry, in
particular in the area of operating system development, has had its
share of battles over standards. This type of game clearly has a first
mover advantage and, if firms succeed in making early announcements
which commit them to a strategy, they will do better.

25
28 Managerial economics

Example 2.6

This example is typical of market entry battles. Suppose two

pharmaceutical companies are deciding on developing a drug for
Alzheimer’s disease or for osteoporosis. If they end up developing the
same drug, they have to share the market and, since development is
very costly, they will make a loss. If they develop different drugs they
make monopoly profits which will more than cover the development
cost. The pay-off matrix could then look like this:

A O

A –2, 2 20, 10

O 10, 20 –1, –1

There are two Nash equilibria in this game: (A, O) and (O, A). Note that
there is a ‘first mover advantage’ in this game. If Firm 1 can announce
that it will develop the drug for Alzheimer’s it can gain 20 if the
announcement is believed (and therefore Firm 2 chooses strategy 0).
Firms in this situation would find it in their interest to give up flexibility
strategically by, for example, signing a contract which commits them
to delivery of a certain product. In our scenario a firm could, with a
lot of publicity, hire the services of a university research lab famous for
research on Alzheimer’s disease.

The type of first mover advantage illustrated in this example is

prevalent in the development and marketing of new products with
large development costs such as wordprocessing or spreadsheet
software packages. The firm which can move fastest can design the
most commercially viable product in terms of product attributes and
the slower firms will then have to take this product definition as given.
Other sources of first mover advantage in a new product introduction
context include brand loyalty (first mover retains large market share),
lower costs than the second mover due to economies of scale and
learning curve effects.

Case: Airbus versus Boeing

Europe’s Airbus Industrie consortium and Boeing are both capable of

developing and manufacturing a large passenger aircraft. The rationale
for pursuing such a project is clear. Airports are getting very crowded
and, given the high volume of traffic forecast for the next decades, it
will become increasingly difficult to find take-off and landing slots; an
aircraft carrying, say, 700 or 800 passengers rather than the current
maximum of 500 is likely to increase efficiency. The world market has
room for only one entrant (predicted sales are about 500) and if both
firms start development, they will incur severe losses (to bring a super-
jumbo to market could cost US$15 billion). Assume the pay-off matrix
with strategies ‘develop’ (D) and ‘don’t develop’ (DD) is similar to the
one given below.

A\B D DD

D –3, –3 10, –1
DD –1, 10 0, 0

26
 Chapter 2: Game theory

There are two Nash equilibria: one in which Airbus builds the aircraft
(and Boeing doesn’t) and one in which Boeing builds it (and Airbus
doesn’t). In this game there is a significant ‘first-mover advantage’: if we
allow Boeing to make a decision before its rival has a chance to make
a decision it will develop the aircraft. (In reality the game is of course
more complicated and there are more strategies available to the players.
For example, Boeing could decide to make a larger version of its existing
450-seat 747, which would not be very big but could be developed
relatively quickly and at lower cost. Or it could decide to collaborate
with Airbus.)
The role of government regulation is not clear cut here. On the one
hand, governments may want to prevent the inefficient outcome of
both firms going ahead with development but, on the other hand, the
prospect of monopoly is not attractive either. Particularly if Boeing and
Airbus collaborate, the consequences of what would be an effective
cartel would be disastrous for struggling airlines. Not only would they
have to pay a high price for the super-jumbo but, if they want to buy a
smaller aircraft, they would have to rum to Boeing or Airbus who might
increase the prices of these smaller aircrafts to promote the super-
jumbo.
One way to avoid the problem of both companies starting development
is for the EU to announce that it will give a large subsidy to Airbus if
it develops the aircraft, regardless of whether Boeing also develops it.
Then ‘develop’ may become a dominant strategy for Airbus and one
Nash equilibrium would be eliminated. (To see this add 5 to A’s pay-off
if A chooses strategy D.)
The United States has persistently complained about Airbus subsidies and,
in 1992, an agreement limiting further financial support was reached.
As of July 1993 both firms had plans to go ahead independently with
development of a large aircraft despite discussing a partnership to build
a super-jumbo jet (the VLCT or very large civil transport project). In
June 1994 Airbus unveiled a design of the A3xx- 100, a double decker super
jumbo which would cost US$8 billion to develop.1 1
Case based on ‘Now
for the really big one’;
‘Mumbo jumbo, super
Example 2.7
jumbo’; ‘The flying
In the pay-off matrix below there is no Nash equilibrium ‘in pure monopolists’; ‘Plane
Wars’
strategies’ (i.e. none of the pairs (T, L), (T, R), (B, L) or (B, R) are stable
outcomes). Consider, for example, (B, L). If the row player picks
strategy B then the best response of the column player is L but, against
L, the best response of the row player is T, not B. A similar analysis
applies to the other strategy pairs.

L R
T 10, 5 2, 10
B 8, 4 4, 2

Nash (1951) showed that games in which each player has a finite number
of strategies always have an equilibrium. However, players may have to
use mixed strategies at the equilibrium. A mixed strategy is a rule
which attaches a probability to each pure strategy. To see why it makes
sense to use mixed strategies think of the game of poker. The strategies
are whether to bluff or not. Clearly, players who always bluff and players
who never bluff will do worse than a player who sometimes bluffs. Players

27
28 Managerial economics

using mixed strategies are less predictable and leaving your opponent
guessing may pay off. To see how to find a Nash equilibrium in mixed
strategies for a two-player game in which each of the players has two pure
strategies, consider the pay-off matrix of Example 2.7 again.
Example 2.7 (con’d)
Suppose the row player uses a mixed strategy (x, 1 – x) (i.e. he plays
strategy T with probability x and B with probability 1 – x) and the
column player uses a mixed strategy (y, 1 – y) (i.e. he plays strategy L
with probability y and R with probability 1 – y). Then the expected pay-
offs to the row and column player are respectively:
πr = 10xy + 2x(1 – y) + 8(1 – x)y + 4(1 – x)(1 – y)
and
πc = 5xy + 10x(1 – y) + 4(1 – x)y + 2(1 – x)(1 – y).
The row player chooses x so as to maximise her expected pay-off and the
column player chooses y so as to maximise his expected pay-off. Given y,
the expected pay-off to the row player is increasing in x as long as
y >1/2 and decreasing in x for y < 1/2 (check this by differentiating πr
with respect to x) and therefore the best response to y is x = 0 for y < 1/2,
x = l for y>1/2 and any x is optimal against y = 1/2. Following a similar
derivation for the column player we find that his best response to a
given x is to set y = 0 for x > 2/7, y = 1 for x < 2/7 and any y is optimal
against x > 2/7. These best response functions are pictured in bold in
Figure 2.3.
L

1/2

R
0 2/7 1
B X T

Figure 2.3: Mixed strategy Nash equilibrium

At a Nash equilibrium the strategies have to be best responses to
each other. Therefore, the point (x = 2/7, y = 1/2), where the response
functions intersect, forms a Nash equilibrium. We can then calculate the
players’ expected pay-offs by substituting x and y in the expressions for
πr and πc above.

The notion of Nash equilibrium is very helpful in a negative sense:

any combination of strategies which does not form a Nash equilibrium
is inherently unstable. However, when there is more than one Nash
equilibrium, game theory does not offer a prediction as to which Nash
equilibrium will be played. A topic of current research is to try to find
justifications for selecting particular types of Nash equilibria, say the
equilibria which are not Pareto dominated, over others. This type of
research could help eliminate some Nash equilibria when there are
multiple equilibria. Nevertheless game theorists have not succeeded in
28
 Chapter 2: Game theory

agreeing on an algorithm which will select the equilibrium that is or

should be played in reality.

Prisoners’ dilemma
A class of two-person noncooperative games which has received much
attention not only in economics but in social science in general, is the
class of prisoners’ dilemma games. The story the game is meant to model
concerns two prisoners who are questioned separately, without the
possibility of communicating, about their involvement in a crime. They are
offered the following deal. If one prisoner confesses and the other does
not, the confessor goes free and the other prisoner serves 10 years; if both
confess, they each spend seven years in prison; if neither confesses, they
each serve a two-year term. This is summarised in the pay-off matrix below.

Confess Don’t

Confess 7,7 0,10

Don’t 10,0 2,2

This game is very easy to analyse: for both players the strategy ‘confess’
dominates. (Remember that you want to minimise the pay-off here!)
There is one Nash equilibrium in which both prisoners serve seven-year
sentences. What is interesting about this game is that, if the prisoners
could set up a binding agreement, they would agree not to confess
and serve two years. (This type of model is used to explain difficulties
encountered in arms control for example.)
The typical application of the prisoners’ dilemma to managerial economics
translates the prisoners’ plight into the situation of duopolists deciding
on their pricing policy. If both set a high price they achieve high profits; if
both set a low price they achieve low profits; if one firm sets a low price
and its rival sets a high price the discounter captures the whole market
and makes very high profits whereas the expensive seller makes a loss. At
the Nash equilibrium both firms set low prices.
Of course in reality firms do not interact only once but they interact in
the market over many years and the question arises whether collusive
behaviour could be rationalised in a repeated prisoners’ dilemma game.
When the game is played over several periods rather than as a one-shot
game, players might be able to cooperate (set a high price) as long as their
rival is willing to cooperate and punish when the rival cheats (deviates
from cooperation). This possibility of punishment should give players
more of an incentive to cooperate in the ion- term. Axelrod (1984) ran
a contest in which he asked game theorists to submit a strategy or the
repeated version of the prisoners’ dilemma. He then paired the given
strategies (some of which were very complicated and required significant
computer programming) and ran a tournament. The ‘tit-for-tat’ strategy,
which consistently outperformed most of the others, is very simple. This
strategy prescribes cooperation as long as the other player cooperates
but deviation as soon as and as long as the other player deviates from
cooperation. It never initiates cheating and it is forgiving in that it only
punishes for one period. If two players use the tit-for-tat strategy they will
always cooperate.
Let’s think about what game theory can contribute to understanding
players’ behaviour in the repeated prisoners’ dilemma. If the game is
repeated a finite number of times then collusive behaviour cannot be
rationalised. To see this, remember that the only reason to cooperate is to
29
28 Managerial economics

avoid retaliation in the future. This means that, in the last period, there
is no incentive to cooperate. However, if both players are going to cheat
in the last period, the next-to-last period can be analysed as if it were the
last period and we can expect cheating then etc. so that we end up with
the paradox that, even in the repeated prisoners’ dilemma game, cheating
is the unique equilibrium. (Of course we are assuming, as always, that
players are intelligent, can analyse the game and come to this conclusion.
If a player is known to be irrational, an optimal response could be to
cooperate.) However, if the game is repeated over an infinite horizon or if
there is uncertainty about the horizon (i.e. there is a positive probability
(< 1) of reaching the horizon), then cooperation can be generated. What
is needed is that the strategies are such that the gain from cheating in one
period is less than the expected gain from cooperation. For example both
players could use trigger strategies (i.e. cooperate until the other player
cheats and then cheat until the horizon is reached). This will be a Nash
equilibrium if the gain from cheating for one period is smaller than the
expected loss from a switch to both players cheating from then onwards.

Perfect equilibrium
So far, with the exception of the section ‘Extensive form games’, we
have considered games in normal form. In this section we return to the
extensive form representation. Consider Example 2.1 and its normal form
representation at the beginning of the section on ‘Normal form games’.
From the pay-off matrix it is clear that there are three Nash equilibria
(T,(t, t)), (B,(t, b)) and (B,(b, b)). Two of these, the ones which have Player 2
playing b(t) regardless of what Player 1 plays, do not make much sense in
this dynamic game. For example, (B,(b, b)) implies that Player 2 – if Player
1 plays T – would rather play b and get a pay-off of –2 than t which gives
pay-off 0. The reason this strategy is a Nash equilibrium is that Player 1
will not play T.
The notion of perfect equilibrium was developed as a refinement of Nash
equilibrium to weed out this type of unreasonable equilibria. Basically, the
requirement for a perfect equilibrium is that the strategies of the players
have to form an equilibrium in any subgame. A subgame is a game starting
at any node (with the exception of nodes which belong to information
sets containing two or more nodes) in the game tree such that no node
which follows this starting node is in an information set with a node which
does not follow the starting node. In the game tree in Figure 2.4, a is the
starting node of a subgame but b is not since c, which follows b, is in an
information set with d which does not follow b.

1
b 2

1
c
1 2
1
d 1

3 1
a

Figure 2.4: a starts a subgame, b does not

30
 Chapter 2: Game theory

So (B,(b, b)) is not perfect since it is not an equilibrium in the (trivial)

subgame starting at Player 2’s decision node corresponding to Player 1’s
choice of T.
Example 2.8

t 10,7
T 2

b 3,0
1

7,10,-1
t
B
2
b 1,2

Figure 2.5: Perfect equilibrium

Using backward induction it is easy to see that the perfect equilibrium
is (T,(t, t) as indicated on the game tree. If we analyse this game in the
normal form, we find three Nash equilibria (marked with an asterisk
in the pay-off table). One of these, namely (B,(b,t)) can be interpreted
as based on a threat by Player 2 to play b unless Player 1 plays B. Of
course if such a threat were credible, Player 1 would play B. However,
given the dynamic nature of the game, the threat by Player 2 is not
credible since in executing it he would hurt not only his opponent but
himself too (he would get a pay-off of 0 rather than 7 which he could
get from playing t). By restricting our attention to perfect equilibria we
eliminate equilibria based on non-credible threats.

Player 2
(t,t) (t,b) (b,t) (b,b)
Player 1 T 10,7* 10,7* 3,0 3,0
B 7,10 1,2 7,10* 1,2
Normal form

Example 2.9 (‘entry deterrence’)

The industrial organisation literature contains many game theoretic

contributions to the analysis of entry deterrence. The simplest scenario
is where the industry consists of a monopolist who has to determine his
strategy vis-a-vis an entrant. The monopolist can decide to allow entry
and share the market with the new entrant, or (threaten to) undercut
the new entrant so he cannot make positive profits. The question arises
whether the incumbent firm’s threat to fight the entrant is credible and
deters the entrant from entering. We can analyse this game using the
notion of perfect equilibrium. On the game tree in Figure 2.6, E stands
for entrant and I for incumbent. The entrant’s pay-off is listed first.

31
28 Managerial economics

don’t
3,4
enter I
-2,0
E fight

don’t
0,10

Figure 2.6: Entry deterrence

It is easy to see that, at the perfect equilibrium, the entrant enters and
the incumbent does not fight. It is in the interest of the incumbent firm
to accommodate the entrant. There are versions of the entry deterrence
game which result in the incumbent fighting entry at equilibrium. In
the ‘deeper pockets’ version, the incumbent has access to more funds
since it is likely to be a better risk than the entrant, and can therefore
outlast the entrant in a price war. In the incomplete information
version, the entrant is not sure about the characteristics or pay-offs
of the incumbent (see Example 2.10). In the ‘excess capacity’ version,
incumbent firms make large investments in equipment which affects
their pay-off if they decide to fight entry. If firms have a large capacity
already in place, their cost of increasing output (to drive the price
down) is relatively low and therefore their threat of a price war is
credible. To see this in Example 2.9, find the perfect equilibrium if the
incumbent’s pay-off of fighting increases to 6 when the entrant enters.

This example could be extended to allow for several potential entrants

who move in sequence and can observe whether the incumbent (or
incumbents if earlier entry was successful) allows entry or not. It would
seem that, for an incumbent faced with repeated entry, it is rational to
always undercut entrants in order to build a reputation for toughness
which deters further entry. Unfortunately, as in the repeated prisoners’
dilemma, this intuition fails. Selten (1978) coined the term ‘chainstore
paradox’ to capture this phenomenon. The story is about a chain-store
which has branches in several towns. In each of these towns there is a
potential competitor. One after the other of these potential competitors
must decide whether to set up business or not. The chain-store, if there
is entry, decides whether to be cooperative or aggressive. If we consider
the last potential entrant, we have the one-shot game discussed above in
which the entrant enters and the chain store cooperates. Now consider the
next-to-last potential entrant. The chain-store knows that being aggressive
will not deter the last competitor so the cooperative response is again
best. We can go on in this way and conclude that all entrants should enter
and the chain store should accommodate them all!2 We should remember 2
For an analysis of
that, as for the repeated prisoners’ dilemma, the paradox arises because of different versions of
the chain store game
the finite horizon (finite number of potential entrants). If we assume an
in which the paradox
infinite horizon, predatory behaviour to establish a reputation can be an is avoided see Kreps
equilibrium strategy. and Wilson (1982) and
Milgroni and Roberts
(1982).
Perfect Bayesian equilibrium
In games of imperfect or incomplete information, the perfect equilibrium
concept is not very helpful since there are often no subgames to analyse.
Players have information sets containing several nodes. In these games an
appropriate solution concept is perfect Bayesian equilibrium. Consider the

32
 Chapter 2: Game theory

imperfect information three-person game in extensive form represented in

Figure 2.7. At the time they have to make a move, Players 2 and 3 do not
know precisely which moves were made earlier in the game.

T2 2,0,0
T1 2 0,2,0
T3
B2
1 M1 3 B3
2 0,2,2
T2
T3 2,8,0
B1 3
B2

4,0,0 6,4,0 B3 0,2,6

Figure 2.7: Bayesian equilibrium

This game looks very complicated but is in fact easy to analyse. When
Player 3 gets to move, she has a dominant strategy, B3: at each node in
her information set, making this choice delivers her the highest pay-off.
Hence, whatever probability Player 3 attaches to being at the top or the
bottom node in her information set, she should take action B3. Similarly,
given Player 3’s choice, Player 2 chooses B2. The choice of B2 leads to 2
or 4 depending on whether Player 2 is at the top or bottom node in his
information set, whereas T2 leads to pay-offs of 0 and 2 respectively. So
again, independent of Player 2’s probability assessment over the nodes in
his information set, he chooses B2. Player 1, anticipating the other players’
actions, chooses strategy M1.
In general for a perfect Bayesian equilibrium we require (a) that the
equilibrium is perfect given players’ assessment of the probabilities of
being at the various nodes in their information sets and (b) that these
probabilities should be updated using Bayes’ rule and according to the
equilibrium strategies. In other words, strategies should be optimal given
players’ beliefs and beliefs should be obtained from strategies. For the
game represented in Figure 2.7, these requirements are satisfied if we
set the probability of Player 2 being at his bottom node equal to 1 and the
probability of Player 3 being at her bottom node equal to any number in
[0,1].

Example 2.10

Let us return to the entry game of Example 2.9 and introduce

incomplete information by assuming that the incumbent firm could be
one of two types – ‘crazy’ or ‘sane’ – and that, while it knows its type,
the entrant does not. The entrant subjectively estimates the probability
that the incumbent is sane as x. This scenario is depicted in the game
tree in Figure 2.8 with the pay-offs to the entrant listed first. The
important difference with the game of Example 2.9 is that here there
is a possibility that the entrant faces a ‘crazy’ firm which always fights
since its pay-off of fighting (5) is higher than that of not fighting (4).
The ‘sane’ firm always accommodates the entrant. The entrant’s decision
to enter or not will therefore depend on his belief about the incumbent’s
type. If the entrant believes that the incumbent firm is ‘sane’ with
probability x then its expected pay-off if it enters is 3x – 2(1 – x) = 5x – 2.
Since the entrant has a pay-off of zero if he doesn’t enter, he will decide
to enter as long as x > 2/5.

33
28 Managerial economics

0,10
don’t enter -2,0
E fight

‘sane’ x enter
I don’t fight
3,4
don’t enter 0,10
1-x -2,5
‘crazy’ E
fight
enter I
don’t fight 3,4

Figure 2.8: Entry deterrence

A reminder of your learning outcomes

Having completed this chapter, and the Essential readings and exercises,
you should be able to:
• represent a simple multi-person decision problem using a game tree
• translate from an extensive form representation to the normal form
representation
• find Nash equilibria in pure and mixed strategies
• explain why in a finitely repeated prisoners’ dilemma game
cheating is a Nash equilibrium
• explain the chainstore paradox.

Sample exercises
1. Consider the following matrix game.

L R
T 2, 5 1, 4
B 5, –1 3, 1

Are there any dominated strategies? Draw the pay-off region. Find the
pure strategy Nash equilibrium and equilibrium pay-offs. Is the Nash
equilibrium Pareto efficient? Which strategies would be used if the
players could make binding agreements?
2. Find the Nash equilibrium for the following zero sum game. The
tabulated pay-offs are the pay-offs to Player I. Player II’s pay-offs are
the negative of Player I’s. How much would you be willing to pay to
play this game?

Player II
–1 2 –3 0
Player I 0 1 2 3
–2 –3 4 –1

3. At price 50, quantity demanded is 1000 annually; at price 60 quantity

demanded is 900 annually. There are two firms in the market. Both
have constant average costs of 40. Construct a pay-off matrix and find
the Nash equilibrium. Assume that, if both firms charge the same price,
they divide the market equally but, if one charges a lower price than
the other, it captures the whole market. Suppose the two firms agree

34
 Chapter 2: Game theory

to collude in the first year and both offer a most favoured customer
clause. What is the pay-off matrix for the second year if they colluded
the first year?
4. Find the pure and mixed strategy equilibria in the following pay-off
tables. How might the –100 pay-off affect the players’ actions?

L R L R
T 12, 10 4, 4 T 12, 10 4, 4
B 4, 4 9, 6 B 4, –100 9, 6

5. Students don’t enjoy doing homework and teachers don’t like grading
it. However, it is considered to be in the students’ long-term interest
that they do their homework. One way to encourage students to do their
homework is by continuous assessment (i.e. mark all homework), but
this is very costly in terms of the teachers’ time and the students do not
like it either. Suppose the utility levels of students and teachers are as
in the pay-off matrix below.
teacher
check don’t check
student work 0, -3 0, 0
no work -4, 4 1,- 2

a. What is the teacher’s optimal strategy? Will the students do any

work?
b. Suppose the teacher tells the students at the beginning of the year
that all homework will be checked and the students believe her. Will
they do the work? Is the teacher likely to stick to this policy?
c. Suppose the teacher could commit to checking the homework part
of the time but students will not know exactly when. What is the
minimal degree of checking so that students are encouraged to do
the work? (i.e. what percentage of homework should be checked?
6. Consider the two player simultaneous move game below where pay-
offs are in £. Find the pure strategy Nash equilibria. How would you
play this game if you were the row player? How would you play this
game if you were the column player?

a b c d
u 100, 3 2, 2 2, 1 0, –500
d 0, –500 3, –500 3, 2 1, 10

7. Two neighbours had their house broken into on the same night and
from each house an identical rare print went missing. The insurance
company with whom both neighbours had taken out home contents
insurance assures them of adequate compensation. However, the
insurance company does not know the value of the prints and offers
the following scheme. Each of the neighbours has to write down the
cost of the print, which can be any (integer) value between £10 and
£10,000. Denote the value written down by Individual 1 as x1 and the
one written down by Individual 2 as x2. If x1 = x2 then the insurance
company believes that it is likely that the individuals are telling the
truth and so each will be paid x1. If Individual 1 writes down a larger
number than Individual 2, it is assumed that 1 is lying and the lower
number is accepted to be the real cost. In this case Individual 1 gets
x2 – 2 (he is punished for lying) and Individual 2 gets x2 + 2 (he is

35
28 Managerial economics

rewarded for being honest). What outcome do you expect? What is the
Nash equilibrium?
8. Consider the extensive form game below. Find all Nash equilibria. Find
the subgame perfect equilibrium.
u
2 0,0
U
1 d

3,1
D
1,3

9. Consider the extensive form game below. What are the players’
information sets? Write this game in normal form and analyse it using
the normal form and the game tree.
t 3,2
2
T b 2,1

M t 1,2
1 2 b
1,4
B
t 2,1
2
2,4
b

10. A firm may decide to illegally pollute or not. Polluting gives it an extra
pay-off of g > 0. The Department of the Environment can decide to
check for pollution or not. The cost of this inspection is c > 0. If the
firm has polluted and it is inspected, it has to pay a penalty p > g; in
this case, the pay-off to the department of the environment is s – c > 0.
If the firm has not polluted and it is checked, no penalty is paid. If the
department of the environment does not inspect, its pay-off is 0.
a. Suppose that the Department of the Environment can observe
whether the firm has polluted or not before it formally decides
whether or not to check. Draw the game tree. What is the
equilibrium?
b. Now suppose the pollution cannot be observed before checking.
Draw the game tree. Is there a pure strategy equilibrium? Compute
the mixed strategy equilibrium, How does a change in the penalty
affect the equilibrium?

36
 Chapter 3: Bargaining

Chapter 3: Bargaining

Aims
The aim of this chaper is to consider:
• the cooperative and conflictual aspects of bargaining
• the nature of the inefficiencies associated with incomplete information
bargaining.

Learning outcomes
By the end of this chapter, and having completed the Essential readings
and exercises, you should be able to:
• analyse alternating offer bargaining games with finite or infinite
number of rounds.

Essential reading
Tirole, J. The Theory of Industrial Organisation. Section 11.5.2.

Further reading
Fundenberg, D. and J. Tirole Game Theory. (Cambridge Mass: The MIT Press,
1991) [ISBN 0262061414].
Rubenstein, A. ‘Perfect equilibrium in a bargaining model’, Econometrica (1982).

Introduction
The literature on bargaining has developed dramatically in the last
decade or so due to advances in noncooperative game theory. Bargaining
is an interesting topic of study because it has both cooperative and
conflictual elements. For example, when a seller has a lot, reservation
price for an object and a buyer has a high reservation price then, clearly,
if the two parties can agree to trade, they will both be better off. On the
other hand, conflict exists regarding the divisions of the gains of trade. The
seller will naturally prefer a high price and the buyer will prefer a low price.
Game theory helps us to model bargaining situations carefully and allows
us to check our intuition regarding, for example, how the outcome of
the bargaining will depend on the parties’ bargaining power and so on.
Questions economists are interested in include:
• under which conditions will bargaining lead to an efficient outcome
• what are good bargaining strategies?
Bargaining problems arise whenever pay-offs have to be shared among
several players. When firms succeed in running a cartel,1 for example, 1
See Chapter 11,
agreeing on how to divide cartel profits is a major problem. Managers ‘Oligopoly’
are interested in bargaining models for their predictions in the context of
management (e.g. for labour (union) negotiations and strikes). However,
most game theoretic models of bargaining are either very simplistic (and
that is certainly true for the ones I discuss in this chapter) or extremely
complex and unrealistic in their assumptions about players’ ability to
reason and calculate.

In the hope that this does not discourage you too much, let us proceed.

37
28 Managerial economincs

The alternating offers bargaining game

The alternating offers bargaining model was formulated and solved fairly
recently by Rubinstein (1982). In this model two players have to decide
on how to divide an amount of money between them. They alternate
in making suggestions about this division. However, as time goes by,
the amount of money available shrinks. It turns out that, at the perfect
equilibrium of this game, the players agree immediately. To see how
this conclusion is arrived at, suppose two players, Bert and Ernie, have
a total of £100 to divide between them. Say Bert makes an offer first.
He might decide, for example, to keep £70 to himself and offer £30 to
Ernie. Ernie will then agree or make a counteroffer. If he agrees, he will
get the £30; if he refuses the offer and makes a counteroffer, the £100
‘cake’ will shrink to £100δ. The discount factor δ (0 < δ < 1) represents
the cost of a delay in reaching an agreement (it represents the cost of a
strike, for example, where potential output and profit is lost while parties
disagree). If the game ends here (i.e. by Bert accepting or rejecting Ernie’s
counteroffer) and we assume that if no agreement has been reached
both players get zero, it is not hard to see what will happen. (You may
want to draw a game tree at this point.) Assuming Bert and Ernie are like
the usual self-interested non-altruistic players then, if Ernie decides to
make a counteroffer, he will offer Bert one penny which Bert will accept.
Ernie in this case ends up with about £100δ. If Bert wants to avoid this
predicament he will have to offer Ernie this pay-off from the start so that
he keeps £100 (1 – δ) for himself. So if there is only one round of offers,
at the perfect equilibrium, Bert will offer £100δ to Ernie and Ernie will
accept.
Now consider the same game but Bert will get to make a second offer (i.e.
the sequence of moves is Bert, Ernie, Bert). By the time Bert makes his
second move, the cake will have shrunk to £100δ2. Now Bert will have the
‘last mover advantage’ and will offer a penny to Ernie. Ernie will anticipate
this and offer Bert £100δ2 when he has the opportunity so that he keeps
£100δ – £100δ2 = £100δ (1 – δ) for himself. However, Bert can improve on
this by offering Ernie £100δ (1 – δ) in the first move and by keeping £100
(1 – δ + δ2) for himself. You should be convinced by now that the subgame
perfect equilibrium strategy for each player tells him to make an offer
which leaves his opponent very close to indifferent between accepting
the offer and continuing the game. As a consequence, the first offer is
always accepted.
What happens if we don’t give the players a deadline? Suppose they can
keep making offers and counteroffers for an infinite number of periods
but, as before, the cake shrinks in each period. Of course this eliminates
the ‘last mover advantage’ and we can look for a symmetric equilibrium
(players using the same strategy). Also, the equilibrium strategy must be
stationary (i.e. it should give the same prescription in each period because,
in the infinite version, when an offer has been rejected, the game is exactly
as it was before the offer was made except for the shrinking). So, let us
assume that Bert offers Ernie £100x and thus keeps £100(1 – x) to himself.
Ernie will consider accepting £100x or making an offer of £100x to Bert
and keeping £100 (1 – x). Since he should be indifferent between these
two options we find δ(1 – x) = x or x = δ/(l + δ). At the equilibrium Bert
gets £100/(1 + δ) and Ernie gets £100δ/(1 + δ). In the infinite version of this
game it is an advantage to be able to make the first move. Again, as in the
finite version, the first offer is always accepted.

38
 Chapter 3: Bargaining

As the time between offers and counteroffers shrinks, the discount factor
approaches 1 and the asymmetry, caused by who moves first, disappears.
It is not very difficult to extend this analysis to allow for different discount
factors of the two bargaining parties. The conclusion of this modified
bargaining game is that the more patient bargaining partner will get a
larger slice of the cake.

Experimental work
Although the alternating offers bargaining game has a simple ‘solution’
and a stark prediction about the duration of the bargaining (one
offer only), the game theoretic findings are not always replicated
in experiments. About 10 years ago, when I was a student at LSE, I
(and many others) participated in a series of bargaining experiments.
Subjects were paired to an unknown bargaining partner and could only
communicate via a formal computer program with their partner who was
sitting in a different room. Experiments of this type generally show that
‘real’ players have a tendency to propose and accept what they consider
a fair offer while rejecting what they consider a mean offer even if this
rejection means they will be in a worse position. If you were offered one
penny in the last round of the game in this section, would you accept?

Incomplete information bargaining

The alternating offers bargaining game does not look very appealing as
a paradigm for real life bargaining situations in which disagreement is
common and costly negotiations take place for several weeks or months.
It turns out that to generate delayed agreement you have to assume
that the players do not know all the information there is to know about
their opponent (i.e. players have private information). In particular,
players may have private information about their reservation price when
bargaining over the sale of an item. Models of incomplete information
bargaining can be extremely complex and I will not discuss them in
general or even give an overview.2 Instead, I will show you in a simple 2
For an advanced
example that inefficiencies can occur. This means that, if it were possible treatment of this
material see, for
to get both players to reveal their valuations truthfully, they could both be
example, Fudenberg
made better off. It is precisely because players are hiding their valuations and Tirole (1991)
that there are costly delays before an agreement is reached. Chapter 10
Two players, a seller and a buyer, are trying to come to an agreement
about the price at which a good will be sold. The seller has a valuation (or
reservation price) of 1 or 3, equally likely. The buyer has valuation 2 or 4,
equally likely. I will refer to a player as being of type i if he has valuation
i. The seller moves first and offers to sell for a price of 2 or 4. The buyer
always accepts an offer of 2 but may reject a price of 4. If the buyer
rejects, the seller can offer a price of 2 or 4 and the buyer has another
chance to accept or reject. If there is delay any pay-offs in the second
period are discounted using a discount factor ds for the seller and db for the
buyer. Table 1 contains the possible strategies for each type of player and
the pay-offs corresponding to each possible strategy pair. The first pay-off
listed is the buyer’s. Note that a seller of type 3 will never set the price
equal to 2 and that a buyer of type 2 is never willing to buy at price 4.
This restricts the number of strategies we have to consider.

39
28 Managerial economincs

Seller type 1 Seller type 3

ask 2 ask 4, then 2 ask 4, then 4 always ask 4
buyer type 2
(0, 1) (0, ds) (0, 0) (0, 0)
always rejects 4
buyer type 4
(2, 1) (2db, ds) (0, 3ds) (0, ds)
reject 4 once

do not reject (2, 1) (0, 3) (0, 3) (0, 1)

For a buyer of type 4, rejecting a price of 4 in the first period weakly

dominates accepting price 4 immediately. If we eliminate the last row in
the table then, for a seller of type 1, asking a price of 2 dominates asking
4 first and then asking 2 so that we can eliminate the second column in
the table. Given that the seller of type 1 thinks that the buyer he faces is
equally likely to be of type 2 as of type 4, asking 2 gives him an expected
pay-off of 1 whereas asking 4 twice gives him an expected pay-off of
0(1/2) + (3ds)(1/2). Thus the seller of type 1 asks 2 if 1 > (3ds/2) or ds < 2/3.
It follows that, if ds > 2/3, there is no trade, at the equilibrium, between a
seller of type 1 and a buyer of type 2 which is clearly inefficient. If ds < 2/3
there is an inefficient delay in the agreement between a seller of type 1
and a buyer of type 4 at the equilibrium. There is also an inefficient delay
of the agreement between a seller of type 3 and a buyer of type 4.

A reminder of your learning outcomes

By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• analyse alternating offer bargaining games with finite or infinite
number of rounds.

Sample exercise
Consider the alternating offers bargaining game over £100, with Bert
making the first offer and Ernie making a counter offer if he wants to.
Suppose Bert has an outside option of £50, that is, at any point during the
game, Bert can stop bargaining and get £50 (discounted if he gets it after
the first period). If Bert takes his outside option, Ernie gets zero. How does
this effect the equilibrium strategies and pay-offs?

40
 Chapter 4: Asymmetric information

Chapter 4: Asymmetric information

Aims
The aim of this chapter is to consider:
• why asymmetric information represents a problem if we are concerned
with efficiency the ‘lemon’ problem
• how using a ‘deductible’ can overcome moral hazard in the insurance
context
• the difference between signalling and screening
• the relevance of the participation and incentive compatibility
constraints in the principal-agent problem.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• give examples (preferably your own) of adverse selection and
moral hazard
• discuss how adverse selection in insurance can lead to low risk
individuals being priced out of the market and explain the use of a
partial cover insurance policy in this context
• explain Spence’s education as a signal model and work out an
example
• analyse a simple principal-agent problem.

Essential reading
Varian, H.R. Intermediation Microeconomics. Chapter 37.

Further reading
Akerlof, G.A. ‘The market for “lemons”: quality uncertainty and the market
mechanism’, Quarterly Journal of Economics (85) 1970, pp.488–500.
‘Flattened, sort of’, The Economist, 5 November 1994, p.100.
‘Generations X-onomics’, The Economist, 19 March 1994, pp.55–56.
‘Keep taking the tablets’, The Economist, 4 December 1993, p.81.
Milgrom, P. and J. Roberts ‘Price and advertising as signals of product quality’,
Journal of Political Economy 94(4) 1986, pp.796–821.
Rothschild, M. and J. Stiglitz, J. ‘Equilibrium in competitive insurance markets:
an essay on the economics of imperfect information’, Quarterly Journal of
Economics (91) 1976, pp.629–49.
Spence, M. Market signalling. (Cambridge: Harvard University Press, 1972).
‘Still money in that franchise’, International Banking Survey, The Economist, 30
April 1994, pp.43–46.
‘The consequences of kindness’, The Nordic Countries Survey, The Economist, 5
November 1994, pp.13–16.

41
28 Managerial economics

Introduction
Situations of asymmetric information arise when one party involved
in a transaction knows more than the other about relevant variables.
For example, a seller may have more information about the quality of
the product he is trying to sell than the buyer does. With asymmetric
information decisions are taken which would be inefficient under
symmetric information. In extreme cases, markets which would show
vigorous trade under perfect information may collapse completely. It
becomes very difficult under asymmetric information to draw inferences
from people’s behaviour, as they may be trying to fool you. For example,
when bargaining, a buyer may make a low bid in the hope that the seller
will believe he has a low reservation price. By bluffing in this way he
may increase his expected gain at the expense of an inefficient delay in
reaching an agreement.

Adverse selection
Asymmetry of information may lead to adverse selection where
less desirable agents are more likely to voluntarily participate or ‘self-
select’ in trade. In adverse selection problems, the type of agent is
not observable and has to be guessed. This is why adverse selection
problems are also known as hidden information problems. For
example, people who suffer from terminal diseases are more likely than
others to want to buy medical insurance or life insurance. The individuals
considering buying insurance have better information about their risks
than an insurance company does.
As a consequence of this adverse selection problem, the insurance
company cannot use risk estimates from the general population to set
its premiums. The premiums will have to be higher than those based on
the general population. Similarly, if a bank charges the same interest to
low risk and high risk borrowers, the high risk ones will be more likely to
borrow. The bank faces an adverse selection of borrowers. If you advertise
a job vacancy at a given wage, you only get people applying who are
willing to work at that wage. From a pool of potential applicants, only the
less desirable workers might apply. Adverse selection can be responsible
for market failure (i.e. there may be no market for a good whereas
profitable transactions between buyers and sellers would be possible if
everyone had full information).

The ‘lemons’ problem

The ‘lemons’ problem, first studied in a seminal paper by Akerlof (1970),
provides an excellent example of an adverse selection problem. The
terminology of this problem has its origins in American slang: used cars
are called ‘lemons’ if they are of bad quality whereas ‘plums’, ‘peaches’
or ‘cherries’ are high-quality cars. The starting point of the analysis
is a question you may have wondered about: ‘Why are one-year old cars
with low mileage so much cheaper than new cars?’ The answer lies in the
quality of one-year old cars which are offered for sale.

42
 Chapter 4: Asymmetric information

Example 4.1

Suppose one-year old cars of a particular make and model have a

quality or value (to the original buyer) uniformly distributed on the
interval from £10,000 to £20,000. The buyer of a new car only finds
out the value the car will have at age 1 after he has bought the car. He
is now considering selling the car on the used car market. Suppose a
used car buyer (there are many of them) is willing to pay £500 more
than the seller’s valuation of the car. (The reason for this could be that
new car buyers do not like driving cars over one-year old whereas used
car buyers do not care much about how new the car is.) If buyers and
sellers had perfect information about quality then all cars would be sold
(for a price between the seller’s valuation and the buyer’s valuation)
and everyone would be better off. Here it is assumed that the seller
knows the quality but the buyer does not. What will the price be?
Because of buyer competition, the price in the market equals the
(expected) value to the buyers. If the price is p, then all sellers who
value their car at p and below will try to sell their car. So all cars offered
for sale have values (to the seller) between £10,000 and p with an
average quality of (£10,000 + p)/2. Buyers are willing to pay £500 more
than the value to the seller i.e. p = (£10,000 + p)/2 + £500 which gives an
equilibrium price p = £11,000. This means that only the lowest quality
cars (those with value below £11,000) are offered for sale. Only 10 per
cent of the cars are sold. You should verify that, if the buyer premium
(the excess the buyer is willing to pay over the value of the car to the
seller) is less than £500, even fewer cars get sold.

Although Example 4.1 was phrased in terms of the used car market, the
‘lemons’ problem can occur in any market where buyers and sellers have
different information about the quality of the good being sold. In such
markets there is a tendency for low quality to crowd out high quality.
You might think that buyers and sellers should be able to write a contract
stipulating the quality sold, with appropriate penalties if the delivery is
later found to be of low quality. However, writing such contracts is costly
and only makes sense if they can be enforced. It is often very difficult for
a buyer, let alone a court, to assess quality accurately and to determine
whether any defect existed at the time of the sale or was caused by
negligent use. The adverse selection problem can be attenuated if
reputation is important, as is the case when a seller aims to repeatedly sell
to a given buyer, and by the use of warranties.1 1
Discussed in ‘Signalling
and screening’
Example 4.2

Consider the example of home contents insurance. There is a large

variance geographically in burglary rates, even within cities. For
concreteness assume that a fraction ph of potential buyers of insurance
expect a loss (which becomes a claim if they are insured) of £400 per
year and the remaining potential buyers (p1) expect a loss of £100 per
year. Since potential buyers of insurance are risk averse, they are willing
to pay a premium R which is higher than their expected claims. Suppose
the high claim individuals are willing to pay up to £500 per year and the
low claim individuals’ reservation price is £130 per year. The insurance
company does not know who the high and low claim individuals are
and we assume it has to set a uniform premium for all customers.
Ignoring any administration costs and assuming everyone is insured, the
company has to set the premium to cover its expected payout:

43
28 Managerial economics

R ≥ 400ph + 100p1
As long as there are many low claim individuals this premium does not
exceed the low claim individuals’ reservation price:

400ph + 100p1 = 400 (1 – p1) + 100p1 < 130 for p1 > 90%.

However, when there are too few low claim individuals, the premium
required by the insurance company to cover its cost would have to
increase to above £130. This implies that only high claim individuals
buy insurance at a high premium R > 400. Note that under perfect,
symmetric information everyone would buy insurance at the
appropriate rate for their risk category. However, because of the
asymmetry of information, if the insurance company continues to set
the premium based on the average claim, it will make losses since only
the high risk properties will be insured. So it must set the premium
based on the high risk clients and only these will be insured.

Problems of adverse selection can sometimes be overcome by compulsory

insurance regulation requiring all homeowners to insure their homes
or requiring everyone to take out medical insurance for example. Some
employers insure all their employees as one package to avoid adverse
selection problems. If everyone is insured, the high risk individuals will
be better off since they pay a lower premium. The low risk individuals
are not necessarily better off but they will get the insurance at a lower
premium than when the premium was based on only high risk individuals.
Whether low risk individuals are better off under this scheme depends on
how risk averse they are as the insurance they are offered is not actuarially
fair (premium > expected loss).
As another example of eliminating the adverse selection problem in this
way, consider extended five year or 50,000 miles warranties for cars.
These used to be purchased only by people who were likely to make
extensive use of the warranty. Now some manufacturers give this kind of
warranty as a standard feature and include its cost in the price of the car.

Example 4.3

The value of Target Ltd. under its current management is known only
to the current management team. EFM Ltd. is interested in taking
over Target and estimates that Target’s current value is 50, 60 or 75
(£ million) – all equally likely – whereas, after takeover and under
EFM’s management, the value would be 50, 70 or 85 respectively.
The procedure for the takeover bid is that EFM makes an offer which
Target can accept or reject. EFM only gets one chance to make an offer.
How much should they bid? Adverse selection occurs because, for any
bid EFM makes, Target will only be interested if it has a low value
(below the level of the bid). In fact we have an extreme case of adverse
selection here in that it prohibits trade altogether:
• if EFM offer 50 it cannot make a profit since Taret would only sell
if its value was 50
• if EFM offers 60, Target sells if its value is 50 or 60 which gives an
expected gain to EFM of (1/3)(–10) + (1/3)(10) = 0
• if EFM offers 75 then Target will accept the offer for sure but EFM
expects a loss: (1/3)(–25) + (1/3)(–5) + (1/3)(10) < 0.

44
 Chapter 4: Asymmetric information

In Example 4.3, a takeover bid is only accepted if the target’s value under
its current management is below it. If information were symmetric (i.e. if
the bidder had equal access to information about the target’s value as the
target’s current management, the inefficiency associated with the absence
of a takeover could be avoided). A successful bid could be made in any
case, whatever the target’s value is, and trade would take place at some
price between the target’s value under its current management and its
value after the takeover under a new management team.

Moral hazard
‘Moral hazard’ is about incentive problems where one agent cannot
observe the actions of the other agent. This is why moral hazard
problems are also known as problems of hidden action. In the typical
example, being insured changes an individual’s behaviour. If someone has
taken out insurance for theft or collision he is likely to be more careless
than when he was not insured. Whereas the term ‘adverse selection’
generally applies to characteristics or qualities of one of the parties
before a contract is entered into (pre-contractual opportunism),
‘moral hazard’ describes situations in which one of the parties misbehaves
after a contract is signed (post-contractual opportunism). A
restaurant which offers an all-you-can eat deal gets an adverse selection of
big eaters as its clientele. If a group of people eat out and decide to share
the bill equally, moral hazard implies that they are likely to overindulge.
If a company rewards employees based on seniority rather than
performance, it may get an adverse selection of low-achieving applicants.
When compensation is based on team performance rather than individual
performance, moral hazard leads to reduced efforts.
In the 1980s the American government paid out US$300 billion when
hundreds of savings and loan associations failed. This debacle has
been explained in terms of moral hazard on the part of the banks who
(are required by law to) take out deposit insurance provided by the
government. With the deposit insurance, banks do not have the proper
incentive to avoid excessive risk taking.2 For life insurance policies in the 2
See ‘Still money in that
US, the insurance company pays the beneficiary after a suicide only if the franchise’
suicide took place a year or two years from the time the policy was issued.
Life insurance statistics show that the suicide rate is lowest in the 12th and
24th month and highest in the 13th and 25th month of the policy!
Of course, if the insurer can observe the behaviour of individuals (e.g.
where they park their cars or whether they lock them – in the case of
motor insurance; whether they smoke or take drugs in the case of life
insurance), it could group them in different risk classes and set a different
premium for each class. The problem is that it is often impossible or very
costly to monitor behaviour. To counteract the moral hazard problem
in insurance, companies often do not give complete insurance but use
deductible provisions.

Example 4.4

Elmo wants to insure his car, of value V, against theft. Elmo who has
wealth W (this includes the value of his car) is risk averse and has a
strictly concave utility of money function U. Elmo knows he should
really always lock his car and fix the anti-theft device he keeps in his
trunk. However, everything else equal, he prefers to be lazy and leave
the car unlocked. If he is careful his car gets stolen with probability pf;

45
28 Managerial economics

if he is careless his probability of ending up carless is p1( p1 > pf ). If Elmo

does not have insurance he will prefer to be careful since:

(1 – pf )U(W ) + pfU(W – V ) > (1 – p1)U(W ) + p1U(W – V ).

The insurance company is risk neutral and the insurance industry is
assumed to be perfectly competitive. This implies that profits are zero
(the premium is determined as the expected claim) and the insurance
company offers Elmo his utility maximising contract (R, D) where R is
the premium and D is the payment Elmo receives if his car gets stolen. If
the insurer could observe Elmo’s behaviour then it would maximise:
(1 – pi)U(W – R) + piU(W – V + D – R) subject to pi D = R, i = 1, f.
Substituting the constraint for R and setting the derivative with respect
to D equal to zero leads to:
(1 – pi)U′(W = pi D)(–pi) + piU′(W – V + D – pi D)(1 – pi) = 0
so that D = V or, in words, the insurer offers full insurance. (No
surprises here – we’ve seen this in the chapter on decision analysis!)
Elmo ends up with utility U(W – R) and will choose to be careful and pay
a low premium.
Assume now that the insurer cannot observe its client’s behaviour. If it
offers full insurance, Elmo is going to be careless given any premium R
he is asked to pay since, once he is insured, his expected utility is
U(W – R) whether he is careless or not. The insurance company
anticipates this behaviour and therefore has to set the premium
accordingly: R = p1V. Will Elmo accept this contract? It depends on
whether his expected utility without insurance (and with carefulness)
exceeds the expected utility of the insurance contract:
(1 – pf )U(W ) + pfU(W – V ) > = < U(W – p1V )
For example if U(x) = ln(x) then the equation above reduces to:
(1 – pf )ln(W ) + pf ln(W – V ) > = < ln(W – p1V )
For W = 10,000; V = 7,000; pf = 0.01 and pi = 0.015 Elmo would insure.
If p1 = 0.05, Elmo would not insure.
This type of market failure is however easily remedied through the use
of a deductible. If the insurance company does not offer full insurance
but sets D < V, then if Elmo buys insurance his expected utility:

(1 – pi)U(W – R) + piU(W – V + D – R)
is decreasing in pi and Elmo therefore has an incentive to be careful. The
insurance company can now afford to base its premium on anticipated
careful behaviour (i.e. R = pf D). Note that the size of the deductible is
unimportant; in fact, for utility maximisation, we want D as close to V as
possible but not equal to it.

In this insurance example, inefficiency results from the policyholder’s

inability to commit to careful behaviour after he has purchased a full cover
policy. If the insurance company offers a policy with a deductible then the
client is in some sense worse off because he is risk averse and would rather
ensure 100 per cent. However, the inability to commit to careful behaviour
ceases to be a problem. By altering the insurance contract slightly from the
full cover format, the client’s incentives have changed dramatically and he
is ultimately better off than under the full cover policy!

46
 Chapter 4: Asymmetric information

Signalling and screening

In situations of adverse selection or moral hazard, the informed party
may have an incentive to reveal his information to the uninformed party.
A seller of high quality used cars could certainly do better if he could tell
potential buyers that her cars are indeed of high quality. However, just
posting a sign saying ‘I sell good cars’ will not achieve very much. For the
customers to believe the seller, the seller has to prove in some way that her
cars are good. She could do this by offering a warranty. Clearly, if offering
a warranty is only profitable when the cars are of good quality (i.e. a seller
of ‘lemons’ would make a loss if he offered a warranty), then the warranty
is a credible signal of good quality.
Similarly, firms which have low costs might want their potential
competitors to know this for entry deterrence purposes. Entry may only be
profitable if the incumbent has high costs. A low cost incumbent can signal
through limit pricing, which would be unprofitable for a high cost
incumbent. Advertising can have a pure signalling purpose. For some
goods, advertisements do not tell customers anything other than that the
good is for sale. The explanation might be that the seller feels so confident
about his product’s quality that he predicts that the customer who tries it
will become a regular buyer. Indeed, if customers would not purchase
repeatedly, the seller could not recuperate the advertising expenditure.3 3
See, for example, Milgrom and
Generally, the cost of signalling should be lower for the higher quality Roberts (1986)
parties. It should not pay for the lower quality parties to pretend to be
high quality by imitating the behaviour of high quality parties.

The uninformed party usually has an incentive to obtain information.

There are certain measures the uninformed party can take to alleviate the
asymmetric information problem. When it engages in activities to find out
the other party’s private information it is said to be screening. So the
distinction between screening and signalling is that screening is done by
the uninformed party and signalling is done by the informed party. Banks
and insurance companies can try to identify risk classes. Car insurers,
for example, look at a driver’s past record, age and sex, whether the car
owner has a garage etc. Business premises with fire sprinkler systems can
get a fire insurance policy with a lower premium. Similarly, smokers pay
more for medical insurance. For some policies, potential clients have to
undergo a medical examination and insurance can be refused based on
the outcome of this examination. Banks use sophisticated multivariate
statistics to assess borrowers’ credit histories and predict their probability
of defaulting on a loan. Used car buyers in the UK get an AA (Automobile
Association, not Alcoholics Anonymous!) inspector to check a car before
purchase.
If characteristics of customers are unobservable, firms can use self-
selection constraints as an aid in screening to reveal private
information. For example, consider the phenomenon of rising wage
profiles where workers get paid an increasing wage over their careers. An
explanation may be that firms are interested in hiring workers who will
stay for a long time. Especially if workers get training or experience which
is valuable elsewhere, this is a valid concern. A firm will then pay workers
below the market level initially so that only ‘loyal’ workers will self-select
to work for the firm.

47
28 Managerial economics

Education as a signal
The classic example of signalling, first analysed by Spence (1974), is
one in which high productivity individuals try to differentiate themselves
from low productivity individuals by investing in education. The firm
cannot distinguish between high and low quality workers but the workers
themselves know their abilities. As I will show in the following example,
good workers can use education as a signal of productivity since only
the most productive workers invest in education. In other words, the
signalling cost to the good workers is lower than to the low productivity
workers and therefore the two types of workers can be identified because
they make different choices.

Example 4.5

When workers enter their first job, it is difficult to estimate their

productivity. Assume that workers know their own productivity but
firms do not and that a high productivity worker is worth £30,000 per
year and a low productivity worker is worth £15,000 per year to the
firm. There are as many high productivity workers as there are low
productivity workers. If the labour market is competitive, competition
between firms drives up the wage to:
(0.5) £15,000 + (0.5) £30,000 = £22,500.
High productivity workers would like to convince the firm that they
are indeed worth more than average. An opportunity to do exactly that
may exist if workers can invest in education before they apply for a job.
The employer, if he is convinced that workers who are educated are
also highly productive, will pay £30,000 if the worker is educated and
£15,000 otherwise. The cost of a university education to an individual
consists of £15,000 tuition fees and living expenses (in excess of the
government subsidy) plus an opportunity cost of £45,000 (i.e. three
years of wages foregone while studying) plus a cost of effort. (This last
cost varies for high and low productivity individuals.) Let’s say that
high productivity workers are intelligent and actually enjoy studying
so that their cost of effort is zero whereas low productivity workers find
studying extremely strenuous and they value the cost of effort involved in
getting a university degree at £50,000. Hence the total cost of education
is £60,000 for a high productivity worker and £110,000 for a low
productivity worker.
For simplicity we assume that university education has no effect on
productivity and we ignore what happens after a worker leaves his first
job. If workers expect to stay in their first job for five years, signalling is
effective. If firms pay university graduates £30,000 and others £15,000,
all high productivity workers get a university degree and none of the
low productivity workers go to university. To check this, we have to show
that a high productivity worker gains from getting a degree (see (a)
below) and a low productivity worker does not (see (b) below):
(30,000) (5) – 60,000 = 90,000 > (15,000) (5) = 75,000 (a)
(30,000) (5) – 110,000 = 40,000 < (15,000) (5) = 75,000 (b)
We have found a separating equilibrium in which the two types of
workers are identified through signalling. In this example, education is
beneficial to the high productivity workers but wasteful for society as
a whole in the sense that, under perfect, symmetric information its cost
could be avoided. It is useful only as a signal here.

48
 Chapter 4: Asymmetric information

You should be able to show that, if the expected tenure for the first
job is less than four years, signalling does not pay and no worker gets
a degree. In this case we find a pooling equilibrium in which the
two types of workers use the same strategy and are paid £22,500.
Similarly, if the expected tenure is long enough, say eight years or more,
signalling will not work because everyone would prefer to get a degree
and earn £30,000. Firms which want to break even cannot pay the high
salary to graduates in this case and as a result nobody gets a degree.
Note that the high productivity workers are negatively affected by the
existence of low productivity workers. Either they have to invest in
signalling (in a separating equilibrium) or they get a wage below their
productivity (in a pooling equilibrium). This type of externality is an
important aspect of the adverse selection problem.

In the US, where labour markets are less regulated than in Europe and the
minimum wage is only US$4.25 per hour, the wage differential between
college graduates and others is large and increasing. In 1994, college
graduates earned an average of 77 per cent more than high-school
graduates.4 Of course, people increase their productivity in college by 4
See ‘Generation
learning computer skills and other skills valued by employers. At the same X-onomics’
time, the sectors of the economy which employ unskilled workers have
seen fiercer foreign competition which drives down wages. In addition it
seems plausible that employers use college education as a proxy or a signal
of ability and high productivity and that this explains part of the wage gap.
Contrast this with Sweden’s labour market where workers with a
university education are paid 3.5 per cent more than workers who did not
get a university education (few Swedes go to university).5 5
See ‘The consequences
of kindness’
An insurance menu
Uninformed parties can screen by offering a menu of choices or possible
contracts to prospective (informed) trading partners who ‘self-select’ one
of these offerings. This type of screening was developed in an insurance
context by Rothschild and Stiglitz (1976). They show that, if the insurer
offers a menu of insurance policies with different premiums and amounts
of cover, the high risk clients self-select into a policy with high cover.
Example 4.6 illustrates how, in an insurance market, inefficiency due to
adverse selection can be ameliorated through the insurance company
offering clients two contracts: a low premium, partial cover policy and a
high premium full cover policy.

Example 4.6

Assume there are no moral hazard problems (i.e. individuals do not

have any control over their probability of a claim). There are low risk
and high risk individuals (all risk averse) with claim probability pl and
ph respectively. (The insurer knows pl and ph but he does not know who
the low risk and high risk individuals are.) For simplicity I assume that
low risk and high risk individuals are identical with respect to initial
wealth W, size of the potential loss V and utility function U.
If the insurance company offers a policy (R, D) where R is the premium
and D is the payment when the policyholder suffers a loss, then it is
possible that only the high risk individuals are insured (see Example
4.2). Suppose now that the insurance company offers a menu of two
policies, a low premium policy with a deductible (Rl, Dl) and a high
premium policy with no deductible (Rh , V). The purpose of these two

49
28 Managerial economics

policies is to get clients to self select such that the low risk group
chooses the first policy and the high risk group the second. Assuming,
as before, that the insurance industry is competitive and therefore profit
on any policy is zero, the premiums are set according to Rl = plDl and
Rh = ph V.
Clearly, the low risk group prefers the fair (premium equal to expected
claim) insurance (Rl, Dl) to no insurance. Similarly, high risk individuals
prefer (Rh, V) to no insurance. What is most important though is that
the high risk group prefers (Rh, V) to (Rl, Dl). If it prefers the contract
designed for the low risk group then the insurance company cannot
break even. So we need:

U(W – Rh) > (1 – ph) U(W – Rl) + ph U(W – Rl + Dl – V). (*)

The left hand side of this inequality is the expected utility to a high
risk individual if he chooses policy (Rh, V) and the right hand side is his
expected utility if he chooses policy (Rl, Dl). Given that Rl = pl Dl and
Rh = ph V, this puts a constraint on the design of the menu of policies
in the form of an upper bound on Dl. The existence of the high risk
group imposes a policy with a deductible on the low risk group. For the
parameter values corresponding to the car theft insurance of Example
4.4 (W = 10,000, V = 7,000), U(x) = ln(x), pl = 0.01 and ph = 0.1, the
premiums are Rh = (0.1)(7000) = 700 and Rl = (0.01)Dl. The inequality (*)
for these values reduces to:

ln(10,000 – 700) > (0.9) ln(10,000 – 0.01 Dl) + (0.1) ln(10,000 – 0.01 Dl + Dl – 7,000)
or
Dl < 1945.
The low risk group cannot be offered more than a 28 per cent
(1,945/7,000) partial cover.

As we have seen in Example 4.6, insurance companies can overcome the

situation in which the bad risks crowd out the good ones, by designing
separate insurance policies for the different risk groups. Although the
insurer cannot prevent a high risk client from buying a policy designed for
a low risk client, it is possible to deter him by choosing the parameters of
the policies carefully, and in particular by using a deductible in the policy
designed for the low risks. Offering partial cover to the low risk group is
only a partial(!) solution to the adverse selection problem since the low risk
group would prefer a full cover policy. The low risk group still suffers from
the existence of the high risk group but at least it is not priced out of the
market anymore

Principal-agent problems
Principal-agent problems are an important class of moral hazard problems
in which one party, the principal, hires another party, the agent, to take
certain actions. The agent generally has more information than the
principal. In particular, the agent knows how much or how little effort
he made in pursuing the principal’s objectives. Since the principal and
the agent have different objectives the agent has to be given appropriate
incentives to act in the principal’s interest in order to avoid or limit
the moral hazard problem. In corporate governance, for example, the
principals are the shareholders and the agents are the managers.
Designing appropriate incentive systems is of great practical importance.
In the bank deposits insurance case, mentioned above, the government
50
 Chapter 4: Asymmetric information

complements the provision of this insurance with regulation on minimum

capital-adequacy standards. The European Union’s Capital Adequacy
Directive was passed in March 1993 and will be implemented at the end
of 1995. This regulation gives the bank’s owners incentives to avoid
excessively risky investments since more of their own capital is at stake.6 6
See ‘Still money in that
In Japan, doctors are known to over-prescribe drugs because their franchise’; ‘Flattened,
income is partly dependent on prescription. However, the health ministry, sort of’
after several scandals involving deaths as a consequence of over-
prescription, is rethinking the doctors’ compensation system.7

In the standard context in which the principal-agent problem is studied,

an employer hires a worker to do a particular job. Generally, the employer
7
benefits from high effort levels but the worker dislikes providing much See ‘Keep taking the

effort. A moral hazard problem arises: if the worker is not monitored or tablets’

given appropriate incentives, he will shirk.

Effort can be observed

When the employer can observe the worker’s efforts, or when there is
a deterministic relationship between effort and performance, it is not
difficult to find an incentive scheme which motivates the worker to
provide the ‘optimal’ effort. Suppose the worker generates a profit P(e) for
the employer if he works for e hours. The cost (monetary equivalent of
his disutility) to the worker of working e hours is C(e) and, if the worker
does not work for this employer, he can get an alternative job which gives
him a utility (in money terms) of u. The employer chooses the effort he
wants the worker to exert to maximise his profit minus the payment to the
worker, taking into account the worker’s reservation utility u and the fact
that it should be in the worker’s interest to make the effort chosen by the
employer. The following simple payment schemes motivate the worker to
provide the efficient amount of effort.

Payment based on effort

If the worker is paid based on his effort e according to we + K (with w
and K as constants), the employer’s problem is: max P(e) – (we + K). The
employer has to take into account the participation constraint or
the individual rationality constraint (i.e. the worker only works if
he gets at least his reservation utility: we + K – C (e) ≥ u). The employer
has no reason to give the worker more than his reservation utility and
so his objective becomes, after substituting the participation constraint,
max P(e) – (C(e) + u). The employer chooses the optimal effort e* such that
the marginal profit of effort equals its marginal cost: P′(e*) = C′(e*) . The
worker has to be encouraged to provide the optimal effort level which
leads to the incentive compatibility constraint: the worker’s net pay-
off we + K – C (e) should be maximised at the optimal effort or
w = C′ (e*). We can conclude that the worker is paid a wage per hour equal
to his marginal disutility of effort and a lump sum K which leaves him with
his reservation utility.

Forcing contract
The employer could propose to pay the worker a lump sum L which gives
him his reservation utility if he makes effort e* i.e. L = u + C(e*) and zero
otherwise. Clearly, the participation and incentive compatibility constraints
are satisfied under this simple payment scheme. This arrangement is called
a forcing contract because the employee is forced to make effort e*. In
‘payment based on effort’ and ‘franchise’, the worker can choose his effort
level.

51
28 Managerial economics

Franchise
Suppose the worker keeps the profit of his efforts in return for a certain
payment to the principal. This can be interpreted as a franchise structure.
How should the ‘employer’ set the franchise fee F? The worker now
maximises P(e) – C(e) – F and therefore chooses the same optimal effort as
before: e* such that P'(e*) = C'(e*). The principal can charge a franchise fee
which leaves the worker with his reservation utility: F = P(e*) – C(e*) – u.

Effort cannot be observed

When the effort can be observed or when output or profit can be observed
and there is a deterministic relationship between effort and output or
profit, the moral hazard problem is easily solved as we have seen above.
However, when the employer cannot observe effort the problem becomes
more complicated. Suppose the employer can observe output but output
is only stochastically related to effort. Payment based on effort is out of
the question, so what about payment based on output? The employer
could, for example, use the franchise solution. However, if the worker is
risk averse, he will need to be compensated for taking on risk. Even when
he makes a large effort, his profit could be low if he is a franchisee. The
employer, on the other hand, is more likely to be risk neutral and willing to
carry this risk. The franchise solution is inefficient here. I hope an example
will clarify this.

Example 4.7

Assume the agent’s (worker’s) utility, as before, depends on his pay w

and his effort level e. For simplicity let U(w,e) = √w – e. The principal
(employer) is risk neutral. The agent can decide to shirk or not;
shirking corresponds to e = 0 and not shirking corresponds to e = 8.
The agent generates a revenue of £0 or £10,000 for the principal,
with probabilities given in Table 4.1. You could think of the agent as a
salesperson whose effort is important in getting a sale worth £10,000.
Revenue for principal
0 10,000
e = 0 (shirk) 1/2 1/2
e = 8 (work) 1/3 2/3

Table 4.1: Problems of generating low and high revenue

Of course, when the agent is not shirking, the principal’s chance of
earning the £10,000 revenue is higher. Suppose the agent’s reservation
utility is u = 8 then, if the principal could observe the agent’s effort level,
how much would he have to pay to elicit e = 0? For e = 0, the agent’s
utility in this job is √w and for this to exceed u = 8 we need to pay him
w ≥ 64. If the principal wants to elicit effort level e = 8 which gives utility
√w – 8 to the agent he needs to pay w ≥ 162 = 256. You should be able to
show that a risk neutral principal, who can observe effort, will elicit the
high effort.
Suppose now that the principal cannot observe effort. An obvious
method to reward the agent is to pay him based on his ‘performance’
(i.e. pay him x if the revenue to the principal is 0 and y if he secures
a revenue of £10,000). Assume the principal cannot pay a negative
amount (x, y ≥ 0). How should the principal determine the appropriate
payment scheme (x, y) if he wants to make the agent work? First of all,

52
 Chapter 4: Asymmetric information

he wants to prevent the agent quitting his job or, in other words, the
participation constraint should be satisfied:
(1/3)U(x, 8) + (2/3) U(y, 8) ≥ 8, or
(1/3)(√x – 8) + (2/3)(√y – 8) ≥ 8. (1)
Also, given the compensation scheme (x, y) the agent should prefer
working to shirking (i.e. the incentive compatibility constraint should be
satisfied):
(1/3) U(x, 8)(2/3) U(y, 8) ≥ (1/2) U(x, 0) + (1/2)U(y, 0), or
(1/3)(√x – 8) + (2/3)(√y – 8) ≥ (1/2) √x + (1/2) √y. (2)
The constraints (1) and (2) can be rewritten as:
√x + 2√y ≥ 48 and – √x + √y – 48 ≥ 0.
respectively. Therefore the area above the upwards sloping line in
Figure 4. 1 represents the set of feasible compensation schemes.

1/2
y
(2)
48

24

(1)

1/2
48 x
Figure 4.1: Feasible compensation schemes

The risk neutral principal will want to maximise his expected revenue
which, given that the agent works, equals:
(1/3)(0 – x) + (2/3)(10,000 – y).
Maximising this is equivalent to minimising 2y + x which is achieved
(over the region of feasible compensation schemes) at √x = 0 and
√y = 48. This indicates that the principal should pay the agent only when
he secures the £10,000 revenue and then the payment should be 482 =
2,304. The principal thus expects a net revenue of:
(2/3)(10,000 – 2,304) = 5,130.67
which exceeds his expected net revenue of inducing no effort:
(1/2)(0) + (1/2)(10,000) – 64 = 4,936.
The principal will therefore motivate the agent to work.
The agent receives expected utility equal to his reservation utility of
8 when his effort can be observed. When effort is unobservable, his
expected utility is (1/3)(0 – 8) + (2/3)(48 – 8) = 24. His expected wage is
also higher when effort cannot be observed: (1/3)(0) + (2/3)(2304) = 1536
> 256. This higher expected wage is necessary to compensate the agent
for carrying risk. Ideally the principal should carry all risk since he is
risk neutral but, if the agent is paid a fixed wage, inefficiency due to
moral hazard results.

53
28 Managerial economics

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to
• give examples (preferably your own) of adverse selection and
moral hazard
• discuss how adverse selection in insurance can lead to low risk
individuals being priced out of the market and explain the use of a
partial cover insurance policy in this context
• explain Spence’s education as a signal model and work out an
example
• analyse a simple principal-agent problem.

Sample exercises
1. The value of a four year old Honda Accord to the owner is uniformly
distributed between £3,800 and £5,800. Whatever the value to the
owner is, the car is worth £200 more to a potential buyer. There are
many potential buyers for a four year old Honda Accord. Find the
equilibrium price.8 8
Hint: At any price p,
which cars are offered
2. A travel agent sells holidays to the Cayman Islands. Consumers are for sale? What are they
willing to pay £1,400 if the holiday is excellent and £800 if the holiday worth to a buyer?
is mediocre. The travel agent knows the quality of the holidays he sells
but the consumer does not.
a. If a fraction x of travel agents sell excellent holidays, how much is a
risk neutral consumer willing to pay for a holiday?
b. Suppose it costs a travel agent £1,000 to book a mediocre holiday
and travel agents are perfect competitors. Is there a market
equilibrium in which all holidays are mediocre?
c. Suppose it costs a travel agent £1,000 to book a mediocre holiday
and £1,100 to book an excellent holiday. Is there an equilibrium in
which all holidays are excellent?
d. Suppose it costs the travel agent the same amount, £1,000, to book
a mediocre or an excellent holiday. Find the market equilibrium
(equilibria). How much consumer surplus is generated at the
equilibrium (equilibria)?
e. For the scenarios of (c) and (d), if it were possible, would it be good
competition policy to ban the sale of mediocre holidays?
f. How can the inefficiencies resulting from asymmetric information
be avoided in this context?
3. McClean & Co. and McDirty & Co. are in the entertainment business.
McClean is interested in taking over McDirty. It does not know
McDirty’s value v under its current management but believes it to be
between £3 million and £4 million, all values equally likely. The value
v is known only to McDirty’s management team who will decide on any
takeover bid. Whatever the value v is, the company is worth 1.5v – £1.5
million to McClean.
a. Draw the value to McClean as a function of v and conclude that
McDirty is always worth more in McClean’s hands than under its
current management.

54
 Chapter 4: Asymmetric information

b. McClean has one chance to make a takeover bid p which McDirty

will acceptor reject, What is the expected gain to McClean of
offering £3.5 million?9 9
Hint: for which values
of v will McDirty accept?
c. What is the optimal takeover bid?
4. Consider Spence’s signalling model and allow workers to choose their
number of years of education before they start working. Eighty per
cent of the workforce is intelligent and highly productive whereas
the remaining 20 per cent is not intelligent and unproductive. The
intelligent workers are worth £50,000 per year to the employer and the
others are worth £20,000 per year, which is what their salary would be
if the employer could tell them apart. The cost of education is £10,000
per year for an intelligent student and £20,000 per year for a less
intelligent student. You should add the opportunity cost of not working
while studying to these numbers. Assume that the less intelligent
workers decide to get zero education.
a. Ignoring discounting, and assuming the expected length of the job is
four years, how much education should the intelligent workers get
for their education to be a credible signal of their productivity?
b. Show that everyone would be better off without signalling.
5. Martha has a disutility of effort function C(e) = e2/2 (e is the number
of hours she works) and reservation utility u = 0. I am a risk neutral
distributor of Finnish vodka and want Martha to work for me. For every
hour she works, I make a profit of m on average. I can observe how
many hours Martha works and thus pay her based on her effort.
a. What is the optimal wage labour contract w(e) = we + K, where w is
a per hour wage and K is a lump sum.
b. Martha wants to buy the franchise from me. How much could I get
her to pay for it?
c. Next year Martha graduates and she will be able to get a better job.
This will increase her reservation utility to u = 2. How does that
affect the answer to (a) and (b)?
6. A risk neutral principal hires a risk averse agent to do a job. The agent’s
utility function is U(w,e) = √w – (e – 1) where e is the agent’s effort and
w is his compensation and the agent has a reservation utility u = 1.
The agent decides to work hard (e = 2) or shirk (e = 1). The principal’s
revenue depends on the agent’s efforts but also on random factors
outside the agent’s control. The probabilities of obtaining revenue 10 or
30 are as indicated in the table below.

Revenue
R = 10 R = 30
e=1 2/3 1/3
e=2 1/3 2/3
a. Calculate the expected revenue if the agent works hard and if he
shirks.
b. If the principal could observe effort, how much would she have to
pay to get the agent to work hard? To get the agent to make a low
effort? What is her net revenue in either case? What is the optimal
forcing contract?
c. Suppose the principal can only observe revenue, not effort, and
so has to pay the agent more for the high revenue than for the
low revenue if she wants the agent to work hard. Write down the
55
28 Managerial economics

participation and incentive compatibility constraints. Draw these

constraints and indicate feasible combinations of pay levels for high
and low revenue.
d. Write the principal’s objective function as a function of the pay
levels for high and low revenue. What is the optimal compensation
scheme? Calculate the agent’s expected utility and his expected
wage and make a comparison with (b).
e. Check that motivating the high effort is worthwhile for the
principal.

56
 Chapter 5: Auction and bidding

Chapter 5: Auction and bidding

Aims
The aim of this chapter is to consider:
• the difference between private and common value auctions
• the structure of the auctions discussed and the optimal bidding
strategies
• the IID assumption on buyers’ valuations
• the effect of the number of bidders on the auction revenue and the
bids
• how a bidders’ ring works
• the revenue equivalence result
• the winner’s curse
• the effect of buyer risk aversion in a first price sealed bid auction
• the effect of seller risk aversion on his preference between first and
second price sealed bids.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• explain why bidding your true valuation is a dominant strategy in
English and second price sealed bid auctions
• derive the optimal bidding strategy in first price sealed bid auctions
with uniformly distributed private values
• analyse simple auctions of the type discussed.

Essential reading
Varian, H.R. Intermediate Microeconomics. Chapter 17.

Further reading
Boyes, W.J. and S.K. Happel ‘Auctions as an allocation mechanism in academia:
the case of faculty offices’, Journal of Economic Perspectives 3(3) 1989,
pp.37–40.
Capen, E., R. Clapp and W. Campbell ‘Competitive bidding in high-risk
situations’, Journal of Petroleum Technology 23(1) 1971, pp.641–53.
Cassady, R., Jr. Auctions and Auctioneering. (Berkeley: University of California
Press, 1967).
‘Inside the stockade’, The Economist, 2 April 1994, pp.69–70.
McAfee, R. and J. McMillan ‘Auctions and bidding’, Journal of Economic
Literature 25(2) 1987, pp.699–754.
Milgrom, P ‘Auction theory’, in Bewley, T.(ed) Advances in Economic Theory,
Fifth World Congress. (Cambridge: Cambridge University Press, 1987).
[ISBN 0521340446].
Milgrom, P. and R. Weber ‘A theory of auctions and competitive bidding’,
Econometrica 50(5) 1982, pp.1089–1122.
‘Speeding allowed’, The Economist, 16 April 1994, p.69.
‘Teetering’, The Economist, 27 August 1994, 7.
57
28 Managerial economics

Wilson, R ‘Auctions of share’, Quarterly Journal of Economics (93) 1979,

pp.675–89.
‘You say you want a revolution’, The Economist, 8 January 1994, pp.27–28.

Introduction
While the model of perfect competition involves many sellers competing to
sell, auction theory analyses situations in which many buyers compete to
buy from one seller. Auctions are a widely used alternative to the market
mechanism for transferring goods from sellers to buyers as the following
list of examples illustrates.
• Around 500 BC Babylonian women were auctioned off annually to
prospective husbands. The most attractive ones were offered first and
fetched high prices, while others were auctioned with large dowries
made up from the sale of the attractive women.1 1
Cassady (1967)

• The US Forest Service auctions timber rights.

• The US government regularly auctions off land which may contain oil.
• In sales of securities and bonds auction methods are used. The Treasury
auctions off government debt issues to financial institutions such as
banks, stock brokerages and insurance companies.
• Companies are sometimes sold through a formal bidding process which
may involve an investment bank acting as the auctioneer.
• Boyes and Happel (1989) report that, in the department of economics
at the College of Business, Arizona State University, office space
was auctioned off to the faculty. The proceeds were used to set up
a graduate scholarship fund. (The authors mention that the bidders
didn’t really care where the money went as long as the chairman didn’t
keep it!) Every interested faculty member submitted a sealed envelope
with a bid before an announced deadline. The highest bidder then got
first choice of offices, the second highest bidder could choose from the
remaining offices, etc. Anyone who bid above $75 received a window
office. The highest bid was $500, the next highest $250; a total of
$3,200 was raised.
• In Singapore prospective buyers of new cars have to bid for a
permit or Certificate of Entitlement (COE) before they make their
purchases. Successful bidders pay the lowest accepted bid price. For
example, if the quota for a particular category of cars (say 1,001 to
1,600 cc.) is 1,000 and the 1000th highest bid is S$600, then every
successful bidder (with bid in top 1000) in this category pays S$600.
The exceptions are company vehicles and heavy goods vehicles for
which double the amount in their categories must be paid. There are
exceptions for motorcycles as well. Their owners pay one-third of the
‘open’ category premium if they choose to enter this category (there is a
special motorcycle category). Diplomatic vehicles are unaffected by the
quota system.
In some auctions buyers bid to purchase items but sellers can also bid to
sell a product or service. Many governments and large companies purchase
almost exclusively by procurement through competitive bidding. Therefore
you may have to quote a fee as a consultant for particular projects. The
analysis of an auction is the same whether bidders bid to buy or to sell
(except that in the latter case the lowest bidder wins). You should bear
in mind that the purpose of bidding is not to win but to maximise your
expected gain (i.e. to win only when you are better off by winning).

58
 Chapter 5: Auction and bidding

Given that auctions are so prevalent, we should consider the reasons for
their use and the circumstances under which they are more appropriate
than other methods of sale. Auctions are never used to sell standardised
products for which there are competitive markets. The time and expense
needed to organise an auction and gather all the interested parties in the
same place would be wasted since the outcome would be the same as in
the market. However, auctions are virtually the only mechanism used to
sell special, unique items such as antiques, real estate, art, rare wine, oil
drilling sites and mineral leases. It is very difficult to post a price for such
items because of the uncertainty about demand for them; there are no
historical data a seller can use to form an estimate about potential buyers’
willingness to pay. Auctions are therefore used to determine prices for
these special objects.

Private and common value auctions

In the theory of auctions a distinction is made between private
and common value auctions. It is crucial that you understand this
categorisation! The distinction rests on how the value of the auctioned
object is modelled (i.e. which assumptions are made regarding how the
potential buyers value the object). Take the (unrealistic) extreme case
in which an item to be auctioned has a fixed value which is known to all
bidders (e.g. a £50 note); the item will be sold to the highest bidder who
pays his bid (we will see later that in some auctions the winning bidder
does not pay his bid) and is allocated randomly if there is a tie. Then of
course the Nash equilibrium2 is for everyone to bid the known value (make 2
See Chapter 2, ‘Game
sure you know why) – not very interesting! In all real-life auctions there is Theory’
some uncertainty either about the value other bidders attach to the item or
about the value of the item itself.
In a private value auction you know the value to yourself of winning
the item (or the contract in a procurement situation) but you do not know
its value to other competing bidders. Different bidders attach different
values to the item. For example, you know that your reservation price
for an antique clock is £12,000; another bidder may value it at £14,000.
When you tender for a contract to build a tunnel you may know from past
experience how much it would cost you to build it but you are unlikely to
know exactly what your competitors’ costs will be. In models of private
value auctions it is often assumed that the valuations of bidders are
drawn independently from the same distribution: the IID (independently
identically distributed) assumption. In the procurement context this
implies that costs for each company are taken as drawn from a common
distribution. Auctions for artwork are considered private value auctions as
long as bidders do not intend to buy for investment purposes. If they buy
as an investment with the intention of resale, their private value is not
independent of the other bidders’ values.
In a common value auction you are uncertain about the value of the
object to be auctioned but, whatever the value is, it is the same for all
bidders. This would be the case for an auction for an unknown amount of
money in a sealed envelope. Each bidder forms an estimate of the single
uncertain value of the object. Typical examples of common value auctions
are auctions in which a government sells the mineral rights to a plot of
land. When oil field leases or gas drilling rights are auctioned, a bidder
may not know how deep he will have to drill, how much oil or gas there
is, what the future oil and gas prices are etc. but all these factors will
affect the value of the oil field equally for all bidders. US Treasury bills are
sold through common value auctions because the participating financial
59
28 Managerial economics

institutions resell the government debt. In a procurement context, the

common value assumption is valid if all the bidders would incur the same
costs to do the job but no bidder knows the precise cost. Some bidders
may however have more information about the object than others: they
may have more experience (for example, they have drilled in an adjacent
site to the oil field on auction and can make assumptions about the site’s
characteristics). The important assumption in models of common value
auctions is that the object has the same value to all bidders but each
bidder forms his own estimate of this value.
In reality auctions can be a mixture of private value and common value
auctions: the auctioned object will have a common value but it also has
a buyer-specific value. For example, there may be several candidates
interested in buying a particular company. The actual value of the
company’s assets is likely to be the same for all potential buyers. The
company’s activities, however, may offer varying degrees of synergy,
depending on who the buyer is. Current auction theory studies this type
of hybrid scenario with common and private values but the techniques
used are quite advanced and beyond the scope of this subject guide. In
discussing optimal bidding strategies, I will only refer to strictly private
value and strictly common value auctions.

Private value auctions and their ‘optimal’ bidding

strategies
In this section I present some types of auctions which are frequently
used and discuss good ways of bidding in them. I will show what the
profit-maximising bidding strategies are and how optimal bids depend on
bidders’ estimates of the valuations and on the number of bidders. Table
5.1 summarises the characteristics of the auction forms discussed in this
section. In all auctions bidders try to choose the best strategy knowing that
their competitors are also rationally trying to optimise. This of course leads
to game theoretic analysis; auctions are very good examples of games of
incomplete information: each bidder has private information about the
value of the object being sold.
Private value How bids are What does the Optimal strategy
auctions made? winner pay?
English auctioneer stars with the highest bid stay in the auction
low bid, bidders make until the bidding
increasingly higher reaches your
bids valuation

Standard (or first bids are submitted the highest bid bid below your
price) sealed bid secretly and valuation
simultaneously
Dutch auctioneer starts with the price at which bid below your
high price and reduces the auctioneer valuation
it gradually until a stopped
bidder stops him

Second price sealed bids are submitted the second highest bid your true
bid secretly and bid valuation
simultaneously

Table 5.1: Summary of private value auctions

60
 Chapter 5: Auction and bidding

Private value English auctions

English auctions, or oral ascending bid auctions, are the most commonly
used auction format, accounting for more than 75 per cent of auctions
in the world.3 Although the term English auction may conjure up images 3
See Cassady (1967)
of Christie’s or Sotheby’s salesrooms filled with art connoisseurs bidding
millions of pounds for a Monet, use of this type of auctions is certainly not
restricted to these glamorous settings. Many agricultural products, fish,
repossessed houses and cars are also sold in oral auctions. These auctions
follow a specific format: the auctioneer starts by setting a low price and
asking for bids. As long as interested bidders remain, the price rises through
competitive and increasingly higher (verbal) bids until there is only one
bidder left. This highest bidder wins the object. It is important to realise
that this highest bidder pays an amount which is only slightly higher than
the last bid made by the second highest bidder. A Japanese version of this
auction involves an electronic system which displays a rising price. All
bidders who remain interested at the displayed price keep a button pressed.
When the price gets too high for a bidder, he releases the button and, when
there is only one bidder left, the price at which the second to last bidder
drops out is displayed with the identity of the winner.
How should you bid in a private value English auction? Imagine you
are actually in an auction room, bidding for a desired object. You have a
reservation value v: this is your private valuation of the object. You are
indifferent between acquiring the object at a price equal to your reservation
value and not acquiring it. The bidding process starts at a bid below v. Will
you make a bid? Suppose the bidding is quite competitive and the bid price
rises to v. Will you make another bid? The optimal strategy, which you may
have arrived at yourself, is given in the box below.

The optimal strategy in a private value English auction is to stay in the auction until
your valuation is reached.

In fact, this optimal strategy is dominant: whatever strategies the

other bidders use, you cannot do better than follow the above strategy.
The explanation is as follows. If you win the auction, your pay-off is the
difference between your valuation and your bid; if you do not win, your
pay-off is zero. Your only decision is when to drop out. If you drop out
before the bid price reaches your valuation, you forego the opportunity
of a positive pay-off. If you stay in the auction after the bid price has
exceeded your valuation, you risk obtaining a negative pay-off (if you
win the auction), and the highest pay-off you can get is zero. Therefore
under no circumstances can you improve on dropping out at
your valuation. This points to a possible explanation for the popularity
of English auctions: very little information gathering and preparation costs
are involved since bidders’ optimal strategies do not depend on how their
competitors bid.
From the nature of the optimal strategy it follows that, in an English
auction, the object goes to the person who values it most. He will pay a
price approximately equal to the second highest reservation price since at
this price his only remaining competitor drops out. The price is in reality
slightly above the second highest valuation because bids are raised in
discrete steps. In some real-life auctions the auctioneer determines the next
possible bid on the basis of his perception of the competitiveness of the
bidding. The bids may be raised by pennies or £100,000, depending on the
context. (However, the issue of discrete bid raises is ignored in virtually all
auction models.)

61
28 Managerial economics

Private value standard seated bid auctions

In a standard sealed bid auction, bids are invited from interested bidders
and remain secret until a preannounced date on which they are revealed
simultaneously. In the usual first price version of the sealed bid auction,
the winner is the bidder who bids the highest amount and he pays his bid.
Of course in the case of procurement or government contracts, where this
type of auction is used almost exclusively, the winner is the one who offers
the lowest price.
How should you bid in a first price sealed bid auction? Obviously, you
shouldn’t bid above your valuation (because you could get a negative
pay-off). If you bid your valuation, your pay-off is zero. It is easy to see
that you can do better if you bid below your valuation. When you bid
below your valuation you are in fact trading off larger potential gains
(valuation minus bid) against a reduced probability of winning the
auction. This implies that there is no dominant strategy here because
the reduction in the winning probability depends on your competitors’
bids. Your equilibrium bid should take into account your estimate of your
competitors’ valuations and of their bids. The optimal bid is the one which
maximises your expected gain given your probabilistic information about
equilibrium competing bids. At the equilibrium each bidder shades his bid
below his valuation. It turns out that the general (Nash equilibrium) rule
for private value standard sealed bid auctions is as follows.

The optimal strategy in an IID private value first price sealed bid auction is to bid the
expected value of the second highest valuation among the bidders assuming your
valuation is the highest.

I will not give a proof of this result but I will show you how it works in the
special case of uniformly distributed valuations. Suppose there are two
risk-neutral bidders: you and an opponent. Each of you has a valuation
which is known only to you and which is drawn independently from a
uniform distribution on [0.1]. Your valuation is v1 and you bid b1. Suppose
your opponent’s strategy tells him to bid a fraction f of his valuation v2 (i.e.
b2 = f v2). Would you ever bid above f ? If you bid above f you are sure to
win since b2 < f for v2 < l. If you bid below f your probability of winning is
the probability that the other bid is below yours:
P (win) = P(b2 < b1) = P(f v2< b1) = P(v2 < b1 / f ) = b1 / f.
Hence your expected gain from a bid b1 equals:
(v1 – b1) P(win) = (v1 – b1) (b1 / f ).
Note that there is a tradeoff between the probability and the profitability
of winning: if the bid increases, the profit (v1 – b1) decreases but the
probability (b1 / f ) of winning increases. Optimising the expected gain
with respect to b1 gives b1 = v1/2. This optimal bid is independent off: an
optimal response to any linear bid is to bid half your valuation. It follows
that the strategy pair in which bidders bid half their valuations forms a
Nash equilibrium. How does this relate to the general rule above? Well, if
you assume your value v1 is the highest, then your expectation of the other
bidder’s value is equal to v1/2.
Let us extend the analysis above to a sealed bid auction with n bidders,
each with a valuation drawn from the uniform distribution on [0,1].
Assume you are Bidder 1 and all your opponents use the same linear
bidding strategy: bi = fvi, i = 2,..., n. You win the auction if all of your rivals’
bids are below yours, that is, when bi = fvi < b1 for all i:
P (win) = P(v2< b1 / f and v3 < b1 / f and ...vn < b1 / f ) = (b1 / f )n-1.

62
 Chapter 5: Auction and bidding

Hence your expected gain from a bid b1 equals:

(v1 – b1)P(win) = (v1 – b1) (b1 / f )n-1.
Optimising the expected gain with respect to b1 leads to b1 = ((n – 1)/n)v1.
The equilibrium strategies are therefore to bid (n – 1)/n of your value. You
should be able to derive the equilibrium bidding strategy for valuations
from a uniform distribution on [L,U]. The answer is b1 = (1/n) L + ((n – 1)/n)
vi. We have of course assumed that each bidder knows how many bidders
there are. This is an unrealistic assumption if the number of bidders is
very large but it is precisely in this case that uncertainty about the number
of bidders is not a serious problem: the optimal bidding strategy is to bid
close to your valuation.
It is very important that you realise what is meant by an ‘equilibrium
strategy’. As we have seen in the game theory chapter, an equilibrium
strategy is only optimal against equilibrium strategies. It is the best you
can do if the other bidders play their equilibrium strategies. If you have
reason to believe that your rivals are not using an equilibrium strategy,
your optimal response is likely to differ from the equilibrium strategy. In
general, to find an equilibrium bidding strategy, look for best responses
(i.e. strategies which tell the players what to bid given their valuations and
the other players’ bidding strategies). With IID valuations there is always
a symmetric equilibrium, in which each player bids an increasing function
of his value. As a consequence, the bidder with the highest valuation will
make the highest bid and win the object.
As you can see from the results above, as the number of bidders increases,
each bids a higher fraction of his value and the final price goes up. It
follows from this analysis that it is in the seller’s interest to get as many
bidders as possible since the expected maximum value is increasing in n
and each bidder bids close to his value if n is large.
The effect of the number of bidders on the expected revenue has become
a relevant consideration in determining appropriate antitrust policy
with respect to takeovers and restructuring in the defence industry. The
government cannot expect to fulfil its military requirements at a low cost
if there are only one or two potential suppliers for any type of weapon.4 4
See ‘Inside the
However, this lack of competition among suppliers could be counteracted stockade’
by the government’s monopsony power.

Private value Dutch auctions

In a Dutch auction the auctioneer starts at a very high price and reduces
it gradually until one of the bidders shouts ‘mine!’. The winning bidder
then pays that price. This is called a Dutch auction because it is the
form of auctioning used in the Netherlands for the wholesale of fresh
flowers. A mechanical version of the Dutch auction involves a giant clock
and a switch for each bidder. As the clock ticks, the price goes down
continuously until one of the bidders uses his switch to stop it. The price
and the identity of the winning bidder are then displayed. This version of
the Dutch auction is used in Ontario tobacco auctions.5 5
See Cassady (1967)

The optimal strategy in a private value Dutch auction is same as for a first price sealed
bid auction.

In essence, the Dutch auction and the first price sealed bid auction
describe the same game. Each player’s strategy consists of a function of his
valuation. The only (immaterial) difference is that, in a sealed bid auction,
the bid is submitted in writing and a Dutch auction is an oral auction. In a
Dutch auction, the bidding strategy can be interpreted as the plan to claim

63
28 Managerial economics

the item if the bid level goes down to the number predetermined in the
bidding strategy. A bidder in a private value Dutch auction does not gain
anything from his presence; he could send an agent to do the bidding.
Although Dutch and English auctions are both oral auctions, they are
very different from the bidder’s perspective. As mentioned before, in an
English auction, as the price increases, the bidder can assess at any time
whether he can gain by increasing the bid level. In particular, when the
bid level reaches the bidder’s valuation it is clear that he cannot make a
gain from the auction. In a Dutch auction, the bidder is not certain at any
price (below his valuation) whether he can increase his gain by waiting or
bidding. He knows that he will win the auction if he bids but he could get
a larger surplus (valuation minus bid) if he waits.

Private value second price sealed bid auctions

The second price sealed bid auction is the same as the standard sealed
bid auction where the winner is the highest bidder but now the winning
(highest) bidder will pay the second highest price bid. For example, if A
bids £4,000, B bids £5,000 and C bids £6,000, C would be the winner
of both a first price auction and a second price auction. In the first price
auction C would have to pay £6,000 whereas in the second price auction
he would have to pay £5,000. Surprisingly, the second price sealed bid
auction is easier to analyse than the first price sealed bid auction.

The optimal strategy in a private value second price sealed bid auction is to bid your
true valuation.

Note that we found before that this is also the optimal strategy in an
English auction. Bidding your valuation is a dominant strategy, as it is for
the English auction. Why? Suppose you bid below your valuation. Then
you will gain zero if you don’t win. If you win, then you would have won
bidding your true valuation and paid the same amount (the second highest
bid). So, bidding below your valuation is dominated by bidding your
valuation since bidding lower only decreases your chances of winning and
obtaining a positive pay-off. Now suppose you bid above your valuation.
If you lose, you would certainly have lost bidding your valuation (your
bid would have been lower). Hence you did not do better by bidding
higher than your valuation in this case. If you win and the second bid is
above your valuation you will get a negative pay-off. You do worse than
by bidding your valuation. If you win and the second bid is below your
valuation, you would have won bidding your valuation as well. Bidding
above your value is dominated by bidding your value since bidding above
only increases your probability of winning the auction when you don’t
want to win it.
Note that the optimality or dominance of the strategy ‘bid your
valuation’ does not depend on knowing other bidders’ valuations or their
distributions. In other words, whatever the other bidders’ strategies are,
you cannot do better than bidding your valuation. As was mentioned
for the English auction, the dominance of the optimal bidding strategy
makes second price sealed bidding an attractive auction form. Bidders
do not have to gather any information; they do not in fact have to think
strategically. They just have to be rational enough to realise that they
should bid their valuation.

64
 Chapter 5: Auction and bidding

Auction revenue
An important issue for anyone considering the sale of an object at an auction
is how much revenue can be expected from such a sale and which type of
auction generates the highest revenue. Let us first look at the revenue to
the seller from first and second price sealed bid auctions using the example
of independent, uniformly distributed valuations. We will use the following
property of the uniform distribution. For a uniform distribution on [0,1],
the expected value is 1/2; if you make two independent draws from the
distribution, the expected value of the largest is 2/3 and of the smallest
1/3; in general, if you make n independent draws, the expected value of the
largest is n/(n + l), the second largest has expected value (n – 1)/(n + l) and so
on; the minimum has expected value 1/(n + 1).

First price sealed bid revenue

Recall that, in the two-bidder uniform valuation case, the expected winning
bid is half the expectation of the highest value (i.e. (1/2)(2/3) = 1/3). In the
n bidder case, the expected revenue is a fraction (n – 1)/n of the winner’s
valuation or (n – 1)/n of the expected value of the maximum: that is,
((n – 1)/n)(n/(n + 1)) = (n – 1)/(n + 1).

Second price sealed bid revenue

What is the expected revenue to the seller if the two bidders bid their
valuation (which they are supposed to do as their optimal strategy) in a
second price sealed bid auction? The seller gets the lowest bid, so if there
are two bidders with valuations drawn independently from a uniform
distribution on [0,1] he will expect a revenue of 1/3, the same as in a first
price sealed bid auction. For n bidders, he will expect to get (n – 1)/(n + l),
the expected second highest valuation. So the expected revenue is again the
same as for the first price sealed bid auction. In fact, this conclusion holds
quite generally:

The revenue equivalence result: Irrespective of the distribution of values, in a private

values IID auction for the sale of one itme and risk neutral bidders, a first price sealed
bid auction (or a Dutch auction) and a second sealed bid auction (or an English auction)
yield the same expected revenue. The expected revenue in these auctions is the expected
second highest valuation. Each of these auctions is an optimal auction (i.e. it produces
the maximum revenue of all possible auction methods).

Common value auctions

Common value English auctions
Private value auctions are relatively easy to analyse because the bidders
have all the necessary information at the start. They know their valuation
of the object for sale and there is nothing to learn. In a common value
auction, in contrast, where bidders are uncertain about the value of the
item for sale, and where the item has the same value independent of who
acquires it, bidders in English auctions can learn by observing each other’s
behaviour.
In particular, a bidder can observe how many active bidders there are at any
price and when they drop out. This gives some indication of their estimates.
Sometimes the identity of the bidders conveys useful information. This is the
reason why an expert buyer, whose presence indicates that a particular item
is desirable, may want to hire an agent to do the bidding for him. If he were
to bid himself, other bidders might revise their value estimates upwards and
bid more competitively.
65
28 Managerial economics

There is no dominant strategy for a common value English auction and the
equilibrium bidding strategy depends on the structure of the information
among the bidders (who knows what?) and the possibilities for obtaining
additional information about the value during the auction.

Common value first price sealed bid auctions

In a common value first price sealed bid auction, as in all common
value auctions, each bidder has his own estimate of the value. If bidders
bid according to their estimates, then the most optimistic bidder (the one
with the highest estimate) wins the object. However, he is very likely to
have overestimated the value of the object. This is the winner’s curse:
‘If you lose, you lose and if you win you also lose!’

To avoid the winner’s curse you should bid below your estimate.

The winner’s curse phenomenon is illustrated in Figure 5.1. Assume that

the probability density function of bidders’ estimates has the actual value
of the object as its expected value. This means that bidders’ valuations are
on average correct but half of the bidders overestimate the value. Given
this large probability of overestimation, bidders will be cautious and
reduce their estimates when determining the amount to bid. In the figure
they all reduce their estimate by an amount d. This gives the distribution
of bids on the left in the figure. Depending on the discount d the most
optimistic bidder may still suffer from the winner’s curse. If the discount
is too small the highest bid still exceeds v, as it does in the figure. As d
increases, the bid curve moves further to the left and the chances that the
winner has bid too much become smaller.

bids estimates

v-d v

Figure 5.1
In competitive procurements there are frequent occurrences of the
winner’s curse. If you get the consultancy job you tendered for, you are
likely to have underestimated the costs. In common value competitive
bidding for, say, an offshore oil lease, you have to correct for the possibility
of overestimation. Capen, Clapp & Campbell (1971) analyse bids for
offshore oil tracts, auctioned by the US government in the late 1960s. They
present evidence that the winners may have been cursed! For example,
the winning bid for offshore Texas Tract 506 was US$43.5 million, nearly
three times the second highest bid of US$15.5 million. Failure to properly
discount their estimates may partly explain why some independent 6
See ‘You say you want
television firms are in trouble, having paid too much for their franchises a revolution’
which were sold by the UK government in a bidding process in 1991.6
If there are many bidders, you have to shade your bid even more in order
to account for the winner’s curse phenomenon. To see this, suppose

66
 Chapter 5: Auction and bidding

each bidder estimates the value as its real value plus some (positive or
negative) error and bids his estimate minus some discount. Then the
probability of the bid of the winner exceeding the value is larger when
there is a large number of bidders since the probability that the maximum
of n IID variables exceeds any number increases in n. This is an interesting
observation because there is a natural tendency to bid more aggressively
when there are many bidders.
Note that I haven’t told you how to bid in a common value first price
sealed bid auction other than that you should discount your valuation.
This is because no simple optimal bidding strategy exists. How you
should bid depends on the nature of uncertainty about the value and the
information available to the bidders. As for the private value versions, the
optimal bidding strategy in a common value Dutch auction is the same as
in a common value first price sealed bid auction.

Common value second price sealed bid auctions

For the common value version, the optimal strategies in the second
price sealed bid auction are different from the English auction. In the
English auction there is an opportunity to learn from observing other
bidders which does not arise in a sealed bid auction. You should not bid
your expected value based on your estimate alone (for winner’s curse
reasons) and you should take into account that you win if your estimate
is highest. You can expect higher bids in a second price sealed bid auction
because the price which is paid is lower than the winning bid.

Complications and concluding remarks

The theory of auctions is a fascinating field which has grown dramatically
over the last two decades, in parallel with the advance of game theoretic
modelling in economic theory. I have tried to give you an introduction
to the more accessible material. In this section I briefly mention some
additional issues which are of relevance to real-life auctions.

Phantom bids
As mentioned above, bidders in common value English auctions obtain
information by observing who drops out and when. If the number of
active bidders decreases rapidly, bidders are likely to think they have
overestimated the value and revise their estimate downwards. This
explains the temptation by sellers to use agents who keep bidding against
the last bidder. The disadvantages of using agents in this way are that
(a) the seller incurs a risk of the high bidder dropping out and one of
his agents ‘winning’ the auction and (b) bidders may account for the
possibility that they are bidding against an agent and shade their bids
accordingly. Clearly, using phantom bids skillfully is an art!
Can the auctioneer use phantom bids profitably in a private value English
auction? Yes, he can. Recall that the winner pays the second highest
valuation. If the auctioneer takes a phantom bid after all but one ‘real’
bidder have dropped out, he increases the seller’s revenue. The larger the
difference between the highest and the second highest valuation, the more
profitable the use of phantom bids will be.
In a second price (private or common value) sealed bid auction, an
unscrupulous seller could ‘cheat’ by inserting a phantom bid just below the
winning bid, thus increasing the price paid by the winner. This may partly
explain the unpopularity of the second price sealed bid auction. Of course,
the seller could only profitably cheat in this way if the bidders are naive. If

67
28 Managerial economics

the bidders account for the possibility of a dishonest seller, they will shade
their bids and they will tend to bid as in a first price sealed bid auction.

Do you know why? Think about what the winner pays if the seller uses a high
phantom bid.

Collusion
In some auctions it is possible for bidders to collude. For example, a
‘ring’ of bidders in an English auction could agree not to bid against each
other (i.e. not to bid when the current bid was made by one of the ring
members). In this way the ring can buy the object for less than would be
possible without the ring. The object is then sold in the ring to the highest
bidder and the profit is divided among the ring members. Rings are more
likely when bidders know each other. In 1994, property developers in
Hong Kong were accused of collusion in government land auctions.7 7
See ‘Speeding allowed’

Although bidders could also collude in sealed bid auctions, the anonymity
of the bidding process makes cheating (i.e. breaking the collusive
agreement) more likely. This may be the reason why sealed bid auctions
are preferred to English auctions in procurement. Especially in auctions
which are not one-off events but are repeated very regularly with more or
less the same bidder population (for example Treasury auctions), collusion
is likely. To counteract collusion it is advisable for the seller to withhold
information such as the identity of the winner and the winning bid. This
eliminates the possibility of colluders punishing the cheaters which in turn
increases the probability of cheating.

Risk aversion
In deriving optimal strategies we have implicitly assumed that bidders are
risk-neutral. Would we find different results if bidders are assumed to be
risk averse? The answer to this question depends on the type of auction
under consideration. In a private value English or second price sealed bid
auction the dominance of bidding your true valuation still applies. The
argument given for this dominance does not assume anything about the
bidders’ attitude to risk.
However, risk averse and risk neutral bidders do bid differently in a
private value first price sealed bid auction and a Dutch auction. When the
clock in a Dutch auction is ticking away, indicating lower and lower prices,
a risk averse bidder will become increasingly nervous as the price moves
below his value. He is likely to stop the clock earlier than a risk neutral
person with the same value. A risk averse person bids higher because he
is willing to pay more to avoid the zero pay-off associated with losing. If
this does not make sense to you intuitively, think of the auction as a lottery
in which there are two possible pay-offs: a surplus equal to the value of
the item minus the bid if you win and zero if you lose. The probability
of winning depends on your bid. If you are risk averse you may prefer a
low positive pay-off with a high probability to a high positive pay-off with
a low probability. It follows that a (risk neutral) seller can make more
revenue in a first price sealed bid auction than in a second price sealed bid
auction if the bidders are risk averse, hence the qualification referring to
risk neutrality in the revenue equivalence result.

68
 Chapter 5: Auction and bidding

Example 5.1

Consider a private value first price sealed bid auction with two risk
averse bidders who have identical utility functions U(x) = x1/2 and values
drawn independently from the uniform distribution on [0,1]. Recall that
the Nash equilibrium strategy for risk neutral bidders is to bid half their
valuation. As before, assume Bidder 2 uses a linear bidding strategy: b2
= fv2. The probability that Bidder 1 wins with a bid b1 , remains
b1 /f. Bidder 1 maximises his expected utility U = (b1 / f )(b1 – v1)1/2 which
results in b1 = 2/3v1. The best response against any linear bid function is
b1 = 2/3v1. We have therefore shown that, for risk averse bidders with a
square root utility of money function, bidding 2/3 of their valuation is a
Nash equilibrium strategy. The seller gets 2/3 of the expected maximum
valuation (2/3), that is 4/9, which is higher than the expected revenue of
1/3 in the risk neutral version.

A risk averse seller will also prefer first price over second price sealed
bid auctions. The reason is that the variance of the winning bid is larger
in the latter auction form. Example 5.2 (which can be skipped if you are
frightened of integrals!) illustrates this.

Example 5.2

Consider an auction with uniform valuations and risk-neutral bidders.

We know that in a first price auction the winning bid is half the
maximum value. To determine the variance of the winning bid we
therefore need to look at the distribution of the highest value. For any
number x in [0, 1]:
P(highest value < x) = P(both values < x) = x2.
This implies that the distribution function of the maximum is F(x)
= x2 and its density function is f (x) = 2x. You should check that the
expectation of half the highest value is:
∫ 10(x/2) f (x)dx = 1/3
To find the variance calculate the second moment:
E(x2) = ∫10 x2f (x)dx = 1/2
The variance of the maximum value equals E(x2) – (E(x))2 = 1/18 (since
E(x) = 2/3) and so the variance of the winning bid is Var(x/2) = Var (x)14
= 1/72.
In a second price auction with two bidders, the winning bid is the
minimum (second highest) value. We therefore have to find the
distribution of the minimum valuation:
P (lowest valuation > x) = P (both values > x) = (1 – x)2
so that F(x) = 1 – (l – x)2 = –x2 + 2x.
The density of the minimum valuation is then f (x) = 2 – 2x and you can
check that the expected minimum valuation is 1/3. This confirms that
the seller expects the same revenue in the private value second price
sealed bid auction as in the private value first price sealed bid auction.
The variance of the minimum value can be calculated as:
E(x2) – (E(x))2 = ∫10 x2 f (x)dx – (1/3)2 = 1/6 – 1/9 = 1/18
The variance of the selling price in the second price auction is four times
higher than in the first price auction.

69
28 Managerial economics

Sequential auctions
In most auction models it assumed that the sale of an item can be
considered in isolation. In reality however, many auctions are of a
sequential or repeated nature (e.g. the government may have an annual
round of procurement tenders for stationery; several art objects may be
auctioned on the same day). In some procurement auctions the quantity
auctioned is not fixed but can be determined by the buyer (i.e. the buyer
solicits bids and his quantity demanded is given by a demand function q(p)
where p is fixed by the auction). In some share auctions a bidder bids a
price and a quantity he wants to buy. The winner will be the bidder who
offered the highest price who will get the quantity he bid. If there are any
shares left, the bidder who offered the second highest price will get the
quantity he bid at the price he offered and so on. Bidders in share auctions
may be allowed to submit several price-quantity bids. Conceptually this is
equivalent to submitting a demand curve.
I have already mentioned that repeated auctions are more likely to lead
to bidder collusion but there are other complications. The value of a good
to a potential buyer may depend on whether he has been able to acquire
complementary goods in an earlier auction. Imagine a property developer
interested in purchasing a substantial chunk of land to build a new
entertainment complex. Suppose the land is sold in relatively small land to
build a new entertainment complex. Suppose the land is sold in relatively
small plots. If the bidder is not successful in the early bidding rounds, his
valuations for plots coming up for sale later on may decrease dramatically.
In particular, the value of any plot may depend on whether he is able to
acquire the adjoining plots. Similarly, a construction company, with limited
capacity, values winning a road maintenance contract according to its
available capacity which depends on the number of contracts it has won
recently.
Another consideration in sequential auctions is the effect of winning in
an early round on your ability to bid later on. If you have a fixed budget
or are not able to borrow easily, winning the auction for the modem
painting may leave you resourceless and hence without a chance to win
the silverware auction later on. Other bidders may realise this and may be
able to use this fact to their advantage. By bidding against you to ensure
you pay a large amount of money for the painting, they hope to ensure
themselves a good deal in the silverware auction.

Other factors
Procurement contracts are not always awarded on the basis of price alone.
In many instances a buyer will consider quality issues as well. He could
do this either when the bids are opened (if he has asked bidders to specify
quality variables) or by preselecting the suppliers who are allowed to
tender. For reasons of public accountability, governments and other public
institutions are more likely to use the latter option, especially if quality is
not easily assessed objectively.
Time may be an important consideration in building contracts. When the 8
See ‘Speeding allowed’
Santa Monica freeway was damaged in the 1994 Los Angeles earthquake,
the California Department of Transportation (Caltrans) scrapped its
normal bidding procedures and construction firms were required to bid on
both the cost and the time needed to get the repairs done.8

70
 Chapter 5: Auction and bidding

Conclusion
I have only discussed a few types of auctions but many more exist. For
the more advanced and interested reader I recommend Wilson (1979),
Milgrom and Weber (1982), Milgrom (1987) and McAfee and McMillan
(1987) as further reading. Even for the auctions described in this chapter
there are many variants and the assumptions which are made to enable
us to analyse these auctions would have to be modified to deal with each
variant. For example, it is quite common for the seller to set a reserve price
below which he is not willing to sell. This reserve could be several millions
of pounds for an Impressionist painting for example. An astute auctioneer
or seller may be able to generate a higher expected revenue by carefully
selecting a reserve price. In the models considered here, the number of
buyers is fixed but in reality it is endogenous because potential suppliers
trade off the cost of preparing a bid (which may include an ‘entry fee’)
against their expected pay-off from participating in the tender. It is fair to
say that game theory has offered significant insights into the mechanisms
of auctions and strategies for rational bidders. At the same time, many
practical issues are ignored in current auction theory which limits its value
for real life bidders.

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• explain why bidding your true valuation is a dominant strategy in
English and second price sealed bid auctions
• derive the optimal bidding strategy in first price sealed bid auctions
with uniformly distributed private values
• analyse simple auctions of the type discussed.

Sample exercises
1. A house which is currently rented is going to be sold. There are two
potential buyers: the current tenant and an outsider. The owner solicits
a price from the outside buyer and allows the tenant to buy the house
if he matches this offer. The buyers and the owner know that the two
buyers’ values for the house are independent of each other and that
for each buyer the value is drawn from a uniform distribution on
[£100,000; £160,000].
a. How should the tenant decide whether or not to match the outside
offer?
b. What offer should the outside buyer make if her value is £120,000?
What offer should she make as a function of her value v?
c. (Only attempt this part if you feel brave!) What is the seller’s
expected revenue? Assuming there are two bidders, what would the
seller’s expected revenue be under an English auction?
2. In a private value first price sealed bid auction there are two bidders
with values v1 and v2 drawn independently from a uniform distribution
on [0, 1]. Assume bidder 2’s strategy is to bid b2 = v21/2. Show that the
optimal response of bidder 1 is to bid b1 = 2/3v1. Do these strategies
form a Nash equilibrium?

71
28 Managerial economics

3. I have a beautiful picture of the famous ballerina Belle Taper which

I will auction in class. I charge a £2.50 auction entry fee. You cannot
bid unless you pay this fee. You can only see the picture after you
have decided to participate in the auction (and paid!) You should not
assume that you will derive any pleasure from seeing the picture, only
from possessing it. After you have seen the picture you will know your
valuation of it. Everyone’s prior distribution of their and any other
bidder’s valuation is uniform on [£0, £15]. The auction will be of the
first price sealed bid type. Would you participate in the auction if one
other bidder participates? What if two other bidders participate? What
if the auction is of the second price sealed bid type?
4. a. ‘A second price sealed bid auction is better for the seller than a first
price auction because in a second price auction the buyer pays less
than his bid and will therefore bid higher.’ True or false?
b. ‘A first price sealed bid auction is better for the seller than a second
price sealed auction because in a first price auction your optimal bid
depends on what your opponent bids. This makes the bidding more
competitive.’ True or false?
5. Suppose there are four bidders, each with a private value uniformly
distributed on [0,1]. What is the expected price in an English auction?
in a first price sealed bid auction?
6. Steven and Robert are two risk neutral potential buyers of a
dingelhopper. Each buyer values the dingelhopper at an amount v1
known only to him. This valuation is considered by everyone, including
the seller, to have been drawn from a given distribution, uniform on
[0,10] for Steven and uniform on [10,30] for Robert. The draws for the
two individuals are independent.
a. Who will win in an English auction? What is the expected selling
price?
b. Suppose the seller can impose a minimum bid level of 10. What is
his expected revenue in this case? What if he sets the minimum bid
level at 15? What is the optimal minimum bid level?

72
 Chapter 6: Topics in consumer theory

Chapter 6: Topics in consumer theory

Aims
The aim of this chapter is to consider:
• the decomposition of the price effect into income and substitution
effects (Hicks and Slutsky method)
• the concepts of EV and CV
• the relationship between price elasticity and the effect of a price change
on revenue
• the relationship between price elasticity and the optimal markup
• the role of the income effect in generating backward bending
labour supply
• the tradeoff between risk and return in investment
• the importance of correlation between asset returns for risk reduction
• why everyone chooses the same portfolio of risky assets when a risk-
free asset is available

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• derive demand functions and labour supply functions given a utility
function
• derive the Slutsky equation
• derive Slutsky and compensated demand functions
• calculate elasticities
• explain the state-contingent commodities model
• set up the intertemporal choice problem and analyse the effect of
changes in the interest rate
• set up labour supply model
• derive the shape of the feasible set of risk-return combinations for
portfolios of two risky assets.

Essential reading
Varian, H.R. Intermediate Microeconomics. Chapter 17.

Further reading
Battalio, R.C., L Green and J.H. Kagel ‘Income-leisure tradeoffs of animal
workers’, American Economics Review 71(4) 1981, pp.621–32.
Biddle, J.E. and D.S. Hamermesh ‘Sleep and the allocation of time’, Journal of
Political Economy 98(5) 1990, pp.922-43.
Gravelle, H. and R. Rees Microeconomics. (Harlow: FT Prentice Hall, 2004)
third edition [ISBN 0582404878] Chapter 3 and 19. You may find this
treatment too theoretical and I do not expect you to study the material there.
Solberg, E.J. Microeconomics for Business Decisions (Lexington: D.C. Heath and
Co., 1992) [ISBN 0669167053] Chapters 4, 5, 6 and 7.

73
28 Managerial economics

Markowitz, H. Portfolio selection (New York: John Wiley, 1959).

‘Buyers and builders’, Asia Survey. The Economist, 30 October 1993a, pp.17–20
‘Giving the economy a fix’, The Economist, 10 April 1993b, p.90.
‘Hard Labour’, The Economist, 11 September 1993c, p.30.
‘Sharing the burden’, The Economist, 13 November 1993d, pp.18–20.
‘The home front’, The Economist, 11 December 1993e, pp.31–32.
‘New-found plans’, The Economist, 7 January 1994a, p.79.
‘Risk and return’, The Economist, 19 February 1994b, p.137.

Introduction
In this chapter my aim is to guide you through some fundamental concepts
in the theory of the consumer. You have studied many of these concepts in
your introductory course so I will leave it up to you to review some of the
material there.
Consumer theory deals with the individual consumer’s choice of how to
spend his income. The standard model of consumer choice assumes that
consumers maximise a utility function subject to a budget constraint.
In addition to a brief review of the theory of the consumer, I will show
you some of its important applications to finance, welfare economics,
uncertainty, intertemporal decision-making and labour supply.

Reviewing consumer choice

Utility, budget constraint and demand1 1
Read Solberg (1992)
Chapter 5; Varian (2006)
Make sure you know what is meant by the following terms: Chapters 2, 3, 4, 5, 6

• perfect complements (L-shaped indifference curves)

• perfect substitutes (linear indifference curves)
• bads.

You should know how the budget constraint is affected by taxes, quantity
discounts and vouchers. A voucher is an income transfer which can be
spent on a particular good.2 2
I recommend Varian
(2006) pp.29–31 on the
You should remember how demand for a good can be derived from the
Food Stamp programme
utility maximisation problem subject to a budget constraint. For the as an example.
two goods case, at an interior solution, the optimality condition for
consumer choice is that the marginal rate of substitution equals the price
ratio. In a corner solution one of the goods is not consumed. If the
goods are perfect substitutes you will always find a corner solution unless
the price ratio equals the marginal rate of substitution. You should be able
to derive consumer demand algebraically. Study the derivation of demand
for Cobb-Douglas utility U(x1, x2) = x c1x d2 in Varian (2006) p.93. The
resulting demand functions
x1 = cm and x2 = dm where m is income.
(c + d )p1 (c + d )p2
Note that for Cobb-Douglas utility the expenditure shares of each good are
constant (i.e. p1x1/m = c/(c + d) and p2x2/m = d/(c + d ) do not vary with prices
or income). Empirically this implies that we may be justified in assuming
Cobb-Douglas utility if the fraction of their income which consumers spend
on various products is relatively stable. Once we know a consumer’s utility
function we can assess how he is affected by price and income changes;
this is useful in evaluating policy decisions regarding taxation for example.

74
 Chapter 6: Topics in consumer theory

Once you know how to derive individual demand curves, it is relatively

easy to determine market demand. For any given price, market demand is
the sum of the individual demands. In other words, to find market demand
add individual demands horizontally.

Slutsky and Hicks3 3

Read Varian (2006)
Chapter 8
Make sure you know what is meant by:
• normal good (note that a good can be normal at some income levels
and inferior at others)
• inferior good
• Giffen good.

When the price of a good changes relative to prices of other goods, there
are two effects. The first one is the substitution effect which represents
the fact that the rate at which you can trade off the good under study
with another good has changed. The second effect is the income effect
which represents the idea that your purchasing power increases after a
price decrease and decreases after a price increase. There are two versions
of the substitution effect. In the Slutsky version the budget line rotates
through the original consumption bundle until it has the same slope as the
new budgetline. The tangency of this rotated budgetline and the highest
possible indifference curve determines the change in consumption due to
the Slutsky substitution effect. Since the rotated budgetline passes through
the original consumption bundle, the latter is still affordable and in that
sense purchasing power has remained constant. In the Hicks version
the budgetline rolls under the original optimal indifference curve until it
has the same slope as the new budgetline and the new tangency (with the
same indifference curve so that utility is now held constant) determines
the change in consumption due to the Hicks substitution effect.
The Slutsky demand for a good depends on the initial consumption
bundle since we are holding the consumer’s purchasing power constant in
the sense that he can still (just) afford to buy his original bundle. Figure
6.1 illustrates the Slutsky substitution effect and demand.

x2

A
C B

x1
a c b

Figure 6.1: Slutsky substitution effect

The original budget line leads to a consumer optimum at A. After the
decrease in price of good 1, the optimum shifts to B. The increase in
Slutsky demand due to the price decrease is given by c – a. Point C is the
tangency of a new (higher) indifference and a budgetline corresponding
to the new relative prices drawn through the original bundle A. We will
have to introduce some notation now to derive the Slutsky equation. The

75
28 Managerial economics

original prices for which A is the optimum are p10 and p20 and the quantities
consumed at A are x10 and x20. From the figure it is easy to see that the
Slutsky demand (x1s) which is obtained by rotating the budgetline through
A as the price of good I decreases and drawing the relationship between
price and quantity, is in fact ordinary demand (x1) where income is held
such that A is just affordable or:
x1s ( p1, p2, x10, x20) = x1(p1, p2, m = p1x10 + p2x20) (1)
To derive the Slutsky equation, we take the derivative with respect to p1:
∂x1s ( p1, p2, x10, x20)/ ∂p1 = ∂x1 ( p1, p2, m = p1x10 + p2x20)/ ∂p1
+ (∂x1( p1, p2, m = p1x10 + p2x20)/ ∂m)x10
which can be rearranged to give:
∂x1 ( p1, p2, m = p1x10 + p2x20)/ ∂p1 = ∂x1s( p1, p2, x10, x20)/ ∂p1
– ∂x1( p1, p2, m = p1x10 + p2x20)/ ∂m)x10 (2)
This last equation decomposes the price effect into the Slutsky substitution
effect and the income effect.

Example 6.1

Assume Cobb-Douglas utility U(x1, x2) = (x1, x2)1/2 and original prices
p10 = p20 = 1 and income m = 100. From the section on ‘Utility, budget
constraint and demand’, we know that the ordinary demands are given
by x1 = m/(2p1) and x2 = m/(2p2) and hence x10 = x20 = 50. From (1) we know
that Slutsky demand is given by:
x1s( p1, p2 , x10 = 50, x20 = 50) = x1( p1, p2, m = p1x10 + p2x20 = 50( p1+p2)) =
25( p1 + p2)/p1
This implies that, when the price of good 1 decreases from 1 to 1/2,
ordinary demand for good 1 increases from 50 to 100 whereas Slutsky
demand increases from 50 to 75. The difference between 75 and 100 is
due to the income effect. You should check the Slutsky equation (2) for
this example:
–75/(2p12) = – 25p2 /p12 – 1/(2p1)50 = –150 for p1 = 1/2 and p2 = 1.

You should know how to derive Hicksian or compensated demand

for the two goods case. Remember that the compensated demand function
shows the response of quantity demanded to price changes if income is
varied so as to keep the consumer on the same indifference curve (keeping
utility constant). In Figure 6.2 the derivation of compensated demand
is illustrated. For a high initial price of good x1 the optimal consumption
bundle is at A. If the price of good 1 decreases, so that budget line C
applies, the consumer will choose a point on C. However, we are not
interested in the ordinary demand now; we want to find compensated
demand. We therefore find the optimal consumption bundle if the price
ratio is as after the price decrease but the consumer stays on the original
indifference curve. In the bottom figure Hicksian demand can be drawn
simply by plotting the optimal quantity of x1 as the price ratio changes
and drawing the relationship between price and quantity. You can visualise
the budget line rolling around the bottom of the indifference curve. As
you can see on the graph, the optimal consumption bundle is found at
the tangency of the budgetline and the indifference curve, as for ordinary
demand. What we are doing mathematically when we derive the Hicksian
demand is we minimise expenditure (push the budget line down) subject
to the consumer achieving a given utility level. The Hicksian demand
function is thus a function of prices and utility.

76
 Chapter 6: Topics in consumer theory

x2

A C
B p1
1
x1
p 10

p0
1

p1
1

x1

Figure 6.2: Hicksian demand

Example 6.2

Suppose utility is given by U(x1, x2) = (x1, x2)1/2. What do the compensated
demands look like? The optimisation problem is:
min p1x1 + p2 x2 s.t. U(x1, x2) = u0.
From the tangency condition we know that the price ratio has to equal
the marginal rate of substitution or:

p1 ∂U / ∂x1 1/2√ x2 /x1 x2 p2x2

= =
p2 ∂U / ∂x 1/2 x /x = or x =
x1 1 p1
2 √ 1 2

Substitution in the constraint results in the Hicksian demand functions:

p p2
x2 = 1 u0 and x = u
√ p2 1
√ p1 0

You should understand why the substitution effect of a price change

(Slutsky and Hicks) is always negative so that Slutsky and Hicks demand
curves are always downward sloping. The income effect can be positive
or negative so that ordinary demand is not necessarily although almost
always downward sloping. In practice, the income effect is small unless
consumers spend a significant fraction of their income on the good.
Most students find the material in this section quite abstract and hard to
digest. I hope that the applications of substitution and income effects to
intertemporal choice and labour supply will convince you that studying
these concepts is worthwhile.

77
28 Managerial economics

Consumer welfare effects of a price change4 4

Read Varian (2006)
Chapter 14
Public policy frequently has an effect on relative prices of goods. This is
most obvious in the case of tax policy. If tax on particular commodities
increases or if there is a shift towards excise tax to allow a reduction in
income tax, consumer welfare is clearly affected. One way to measure
this change in the consumer’s well-being would be to compare his utility
level before and after the change. This is problematic because the actual
utility number has no significance. This may not matter if we only want to
know whether the consumer is better or worse off. However, if we want a
more precise measurement or if we want to aggregate the effect of a policy
change over all consumers we need something more precise. For instance,
we could determine how much income the consumer would be willing
to give up to avoid a price increase or how much money the government
would need to give the consumer to compensate him for a policy change
which leads to a price increase (e.g. raising VAT on fuel). Using consumers’
own monetary valuations of price changes allows aggregation.
There are several measures which could be used to assess a change in
consumer welfare. Compensating variation (CV) is the amount
of money which the individual requires to compensate him for a price
increase or the amount of money which could be taken away from the
individual after a price decrease to make him as well off as before
the price decrease. Equivalent variation (EV) is the amount of money
the consumer is willing to pay to avoid a price increase or the amount of
money the individual should get to achieve the same utility as after a
price decrease. You should remember that CV refers to a change in income
which brings you back to the original indifference curve whereas EV refers
to a change in income which brings you to the new indifference curve.
CV indicates the change in income after the price change has taken place
whereas EV indicates the change in income before the price change takes
place.
You will probably need to think about this for a while. Making some graphs
might help. Remember that you can read CV and EV on the vertical axis if x2 5
You could read the
has unit price.5 analysis in Gravelle
and Rees (2004)
The three measures of changes in consumer welfare due to price changes pp.58–65 where the
(EV, CV and the change in consumer surplus CS) are equal when the income relationship between
effect of a price change is zero. This is the case for quasilinear utility EV, CV and Hicksian
demand is explained
or U(x1, x2) = v(x1) + x2. Note that, on the indifference map corresponding to
but I do not expect this
this utility function, the vertical distance between two indifference curves level of mathematical
is constant (i.e. it does not vary with x1). CV, EV and the change in CS sophistication from you.
are all different except for quasilinear utility where all three coincide. (It
would be a good exercise for you to check this.) The change in CS is a good
approximation when the income effect is small which in turn is likely when
expenditures on the commodity as a share of total income is small.

Example 6.3

Using the same utility function U(x1, x2) = (x1, x2)1/2 and initial prices
and income p1 = p2 = 1, m =100, we can calculate the values of the three
measures of changes in consumer welfare when p1 decreases to 1/2. CV
is the amount of money we can take away from the consumer after the
price decrease to bring him back to his original utility level. To calculate
CV we need to know how optimal utility is affected by a change in price.
If we substitute the demand functions x1 = m/(2p1) and x2 = m/(2p2) into U
we find the indirect utility function:

78
 Chapter 6: Topics in consumer theory

V = m/(2( p1 p2)1/2)
so that utility after the price change is:
V1= m/(2(1/2.1)1/2) = m/21/2
and utility before the price change is:
V0= m/(2(1.1)1/2) = m/2
CV is the amount of money we can take away from the consumer to
leave him as well off as initially. Hence (m – CV)/21/2 = m/2 or
(100 – CV)/21/2 = 50 which gives CV = 29.3. EV is the amount of money
you have to give the consumer to make him as well off as after the price
decrease. This means that EV is such that:
(m + EV)/2 = m/21/2 or (100 + EV)/2 = 100/21/2
which gives EV = 41.4. The change in consumer surplus due to the price
decrease is the area to the left of demand between the original and the
new price, hence:
1
m
CS = ∫ dp = 50(log(l) – log(1/2)) = 34.7
½
2p1 1

Elasticity6 6
Read Solberg (1992)
Chapter 4; Varian (2006)
Elasticity measures tell us how responsive demand is to changes in price, Chapter 15
income, prices of substitutes or complements, etc. Whether we look at
price elasticity or income elasticity or cross price elasticity, the measure is
always defined in proportional terms. It tells us what percentage change in
demand we can expect if the other variable changes by one per cent.
You should be aware that elasticities are time-dependent. For most
products it is unrealistic to assume that the price and income elasticities
will be unchanged over a long period of time. The simplicity, at least in
theory, of calculating values for, say, income elasticity is misleading. For
marketing purposes this figure has to be interpreted carefully. Let us take
the example of Malaysian car sales. A 40 per cent rise in incomes between
1987 and 1991 was accompanied by a 290 per cent rise in car sales, giving
a rough estimate of income elasticity of 7. Should the car industry attach
much importance to this number? A closer look at the car market and the
market for many goods in fast growing economies reveals that households
do not consume a good when their income is below a threshold level
but as soon as their income reaches the threshold level they do buy. This
means that, even with small increases in per capita income, large increases
in purchases could occur. At the same time, a large increase in per capita
income may not have a significant effect on sales if the households
whose income increased were already above the threshold level. Clearly,
it is more useful to have a good forecast of the growth in percentage of
households above the threshold than of growth in per capita income.7 7
See ‘Buyers and
builders’
Whether demand for a good is price elastic depends on several factors
including:
• the number and closeness of substitutes. The elasticity of
demand for a particular brand of cigarettes is large whereas the
demand for cigarettes as a whole is inelastic. The demand for petrol
and salt is inelastic
• the period of time over which the response to a price
change is assessed. For many goods if you study the response after

79
28 Managerial economics

a longer time period the elasticity will be larger because consumers

need some time to look for substitutes after a price increase. For
consumer durables such as electrical appliances or cars however, the
opposite is true. It appears that if the price of these goods increases
people postpone their purchases and do not buy in the short-
term; hence there is a high short-term response. In the longer term
consumers cannot postpone replacement of their appliance any further
which leads to a lower long-term response to a price increase
• luxuries versus necessities. The demand for luxury items (income
elasticity > 1) is more elastic than the demand for necessities (income
elasticity <1) because of their large income effect.
The notion of price elasticity is crucial in pricing decisions. Let us ignore
costs for the time being and just determine how price p should be set if
we want to maximise revenue. Revenue R( p) is the product of price and
quantity demanded at that price: R( p) = pq( p). To see how revenue varies
with price, take the derivative:

∂R(p)
∂p
=p
∂q(p)
∂p
+ q(p) =
( p ∂q(p)
q(p) ∂p
+1
( q(p) = (1- η) q(p)

where  is the price elasticity. Therefore revenue is increasing in p

(positive derivative) as long as demand is inelastic (< l) and it decreases
in price when demand is elastic (> l). A revenue maximiser sets price
so that demand has unit elasticity. An important consequence is that
a profit-maximising firm will never set price such that demand is
inelastic. Why? If demand is inelastic, an increase in price generates more
revenue. At the same time, because of the increase in price, less is sold
and therefore costs decrease. Revenue increases and costs decrease after
a price increase so profit can be increased by increasing price as long as
demand is inelastic.
Price elasticity determines the optimal markup over costs. As you would
expect intuitively, if demand is inelastic the markup is higher than when
demand is elastic. To see why this is true, consider the case of constant
marginal cost c and express revenue as a function of quantity rather than
price: R(q) = p(q)q. For profit maximisation we set marginal revenue equal
to marginal cost. Marginal revenue is simply the derivative of revenue with
respect to quantity and so:

∂R
∂q
= p(q) + q
∂p
∂q ( q ∂p
= p 1+ p
∂q
( (
= c or p 1– 1η
( =c

which can be rewritten as:

p=c
( η
η–1
(
η
The optimal markup over cost η – 1 (η >1) is decreasing in the price

elasticity (you should check this) as mentioned above. In practice

constant elasticity demand is often assumed when estimating
demand functions:
q = ap-η mγ tβ (3)

80
 Chapter 6: Topics in consumer theory

where q is the quantity demanded, p is the price, m is income and t is the

price of a substitute or a complement. Other factors such as demographic
variables are often included. The parameters η, γ and β are the price
elasticity, income elasticity and cross-price elasticity respectively.8 The 8
You should check this
multiplicative functional form (3) assumes that the elasticities are constant using the formula for
elasticity
(i.e. they do not vary with values of price, income, etc.). The popularity of
this model can be explained by its ease of use in econometric work (linear
regression can be used after taking logarithms on both sides).9 9
If you are curious
about how demand
Regression analysis and time series analysis are not the only methods you functions are estimated
can use to get information about demand for your product. In fact in some you should read the
cases regression analysis based on historical data is out of the question examples in Chapter
because such data are not available (e.g. for a new product). Marketing 8 and Chapter 9 of
managers then resort to consumer labs or market experiments. In a Solberg (1992)

consumer lab or clinic, simulations of purchasing decisions take place. The

‘consumers’ are given a budget which they can spend on the product under
study as well as substitute products (other brands) and other products.
The experiments then consist of varying the consumers’ budget and the
relative prices. The resulting information is analysed and elasticities are
calculated. In the case of test marketing, the experiment goes on in the
real world. The market for the product is divided into geographical areas
and a different marketing mix (price, advertising, etc) is used in each area.
Consumers’ responses are measured and demand information results from
analysis of these responses.

State-contingent commodities model10 10

Read Varian (2006)
Chapter 12
When we analysed decision-making under uncertainty, we used the
expected utility model. There is an alternative approach to studying
uncertainty: the state-contingent commodities model. This
approach uses standard consumer theory terminology which is why it is
discussed here. Whereas under the expected utility hypothesis any number
of outcomes can be considered, the state-contingent commodities model,
as it relies heavily on graphical analysis, is most useful when there are
only two outcomes.
The ‘products’ a consumer derives utility from are income or wealth in
two states of the world. The states of the world could be ‘car is stolen’ and
‘car is not stolen’ for example. Of these states of the world one and only
one will actually occur. This points to a major departure from standard
consumer theory. In the state-contingent commodities model, only one of
the goods is consumed by definition whereas in the standard consumer
theory model, both goods can be (and usually are) consumed in positive
quantities.
Individuals derive utility from income bundles consisting of one income
level for each possible state of the world. If I do not have unemployment
insurance, my income bundle could be (0, m) indicating a zero income
level in the state ‘I lose my job’ and income m in the state ‘I do not lose
my job’. If I have unemployment insurance, my income bundle could be
(b – p, m – p), where b is the unemployment benefit the insurance company
pays me when I lose my job and p is the insurance premium. Depending
on the values of b and p and my chance of becoming unemployed, I may
prefer the second bundle to the first. Figure 6.3 shows an indifference map
corresponding to this situation.

81
28 Managerial economics

employed

450 line
m

m-p

unemployed
0 b-p m-p

Figure 6.3: State-contingent commodities model

The 45 degree line or the certainty line contains income bundles for
which income in both states is equal and hence there is no uncertainty.
The indifference curves are convex for a risk averse individual because,
loosely speaking, such an individual prefers the same income in both
states. Consider an individual at point A on the 45 degree line in Figure
6.4. If this individual were more risk averse she would be willing to trade
her secure position A for fewer income bundles. The income bundles she
would trade for are of course the ones above her indifference curve through
A. So higher risk aversion corresponds to more convex indifference curves.

income in state 2
certainty line

A income in state 1

Figure 6.4: Risk aversion

In general, utility (and thus the position of the indifference curves)
depends on income in both states and the probability (qi) of each state of
the world occurring:
U(m1, m2, q1, q2).
An example of such a utility function is the expected utility function we
studied before:
U(m1, m2, q1, q2) = q1u(m1) + q2u(m2) (4)
but the state-contingent income approach allows for more general utility
functions. Using the expected utility formulation (4) let’s see how we can
determine the certainty equivalent and the risk premium graphically. On
an indifference curve dU(m1, m2, q1, q2) = 0 or, from (4),
q1u′(m1)dm1 + q2u′(m2)dm2 = 0.
The slope of the indifference curve is thus:
dm2 /dm1 = –q1u′(m1)/q2u′(m2) (5)

82
 Chapter 6: Topics in consumer theory

which is the marginal rate of substitution between income in the two

states, adjusted by the probabilities of the states occurring. At the
intersection of an indifference curve with the certainty line (where
m1 = m2), the slope is dm2 /dm1 = – q1 /q2.

The iso-expected income line is the locus of income bundles (m1, m2)
which have the same expected value E(m). The equation for the iso-
expected income line for given probabilities (q1, q2) is q1m1 + q2m2 = E(m).
The slope of this line is also – q1 /q2 which implies that the tangency of an
indifference curve and an iso-expected income line occurs at the
intersection of the indifference curve with the certainty line. This is
illustrated in Figure 6.5.

m2
certainty line
B

E
e

c
C iso-expected income line

m1
b c e

Figure 6.5: Certainty equivalent and risk premium

We now want to situate the certainty equivalent of income bundle B
graphically. By definition the individual is indifferent between all points on
his indifference curve and, in particular, he is indifferent between B and C.
At point C the individual gets equal income c in both states. The certainty
equivalent of B is therefore c: the individual is indifferent between getting
c for sure or getting state-contingent income B.
It is straightforward to determine the risk premium graphically. The risk
premium is the difference between the expected value and the certainty
equivalent. Where can we read off the expected value of B? Income bundle
E is on the iso-expected income line through B and therefore has the same
expected value as B, namely q1e + q2e = e. The risk premium therefore
equals e – c.
The state-contingent commodities model has applications in the analysis of
insurance demand.11 11
See Gravelle and Rees
(2004) pp.507–14 if you
want to study this but I
Intertemporal choice12 do not expect you to.

I strongly recommend that you read Varian’s excellent treatment 12

Read Solberg (1992)
of intertemporal choice for revision purposes or if you need some Chapter 6; Varian (2006)
clarification. You should know how to set up the intertemporal Chapter 9.
consumption choice problem and derive how saving and borrowing
respond to changes in the interest rate r.
Suppose the consumer has income m1 in period 1 and income m2 in period
2 (the endowment point E in the figure 6.6 below). He can borrow and

83
28 Managerial economics

lend at interest rate r. If he does not consume in period 1 then the amount
available for consumption in period 2 equals m2 + (1 + r)m1. Similarly, if
he plans not to consume in period 2, he can borrow an amount m2 /(1 + r)
in period 1 and hence pay back m2 in period 2. In general, consumption in
period 2 will equal income in period 2 plus any savings from period 1 or
minus any repayments of loans from period 1:
c2 = m2 + (1 + r) (m1 – c1). (6)
Equation (6) is the equivalent of the budget constraint in the standard
consumer choice problem. Note that, as the interest rate r changes, the
budget line rotates through the endowment point E, becoming steeper
with an increase in r and flatter with a decrease in r.
If the individual can borrow but not lend, he has to choose a point on the
line segment to the right of E and, if he can lend but not borrow, he has to
choose a point on the segment to the left of E. If the borrowing and
lending interest rates are different, the budget constraint is piecewise
linear (and concave if the borrowing rate is higher than the lending rate)
with a kink at E.

c2

m2+ lending
(1+r)m1
E
m2

borrowing

c1
m1 m1 + m2/(1+r)

Figure 6.6: Intertemporal choice

The consumer’s preferences over combinations of current and future
consumption can be represented by a utility function U(c1, c2). At the
optimum bundle of current and future consumption, the budget line is
tangent to the highest possible indifference curve. The MRS in this context
is defined as –(1 + p) where p (> 0) measures subjective impatience.

Why?

At the optimum the MRS equals the slope of the ‘budget line’ –(1 + r)
which implies that each individual saves or borrows such that the rate r at
which he can trade current for future consumption in the market coincides
with the rate at which he wishes to make these trades.
Using this model, we can derive the amount of saving or borrowing every
individual in the market plans to do, given the interest rate. The market
interest rate can be determined by plotting the total amount of saving and
the total amount of borrowing against the interest rate. You read off the
equilibrium interest rate where the two curves intersect.

84
 Chapter 6: Topics in consumer theory

Example 6.4

Suppose the consumer has utility function U(c1, c2) = ln(c1) + a ln(c2),
where a (< 1) is a constant. The return on loans to and from the bank
is r and income is m1 in period 1 and m2 in period 2. The consumer’s
optimisation problem is:
Max ln(c1) + a ln(c2) s.t. c2 = m2 + (m1 – c1)(1 + r).
If we substitute the constraint into the objective function we obtain:
ln(c1) + a ln(m2 + (m1 – c1)(1 + r)).
Setting the derivative with respect to c1 equal to zero, we find:
1/c1 + (a/(m2 + (m1 – c1)(1 + r))(–1)(1 + r) = 0
which can be solved for c1:
c1* = (m1 + m2/(1 + r))/(1 + a)
If the consumer can lend but not borrow then the solution remains
unchanged as long as c1*< m1. If c1* > m1 i.e. m2 > a m1(1 + r) then the
consumer would like to borrow but, if he is prevented from doing so, he
will consume his endowment:
(c1, c2) = (m1, m2).

The effect of an interest rate change can be decomposed into a

substitution and an income effect. Using this decomposition it is possible
to show, for example, that an increase in r causes a decrease in c1, for a
borrower whereas it could cause a increase in c1, for a lender.

c2 c2

b c
b
c

a aE
E

c1 c1
(a) borrower (b) saver

Figure 6.7: Effect of an interest rate increase

Figure 6.7(a) shows what happens to a borrower’s consumption plans
when the interest rate increases. Initially he is planning to consume at
point a, to the right of his endowment E. After the interest rate increase he
modifies his plans to point c. The dashed line through a has the same slope
as the new budget line (I am using Slutsky substitution here). The move
from a to b is caused by the substitution effect, which always works in the
direction of less c1 if r increases. The move from b to c is due to the income
effect which leads to less c, if c1 is a normal good. Figure 6.7(b) shows the
effect of the same interest rate increase on the saver’s consumption plans.
The substitution effect works in the same direction as for the borrower
but the income effect now works in the opposite direction. If the income
effect is large enough, c1 could increase. Therefore, it is possible to get
a backward bending supply of savings (i.e. as the interest rate
increases to a high level, savers may reduce their savings!).

85
28 Managerial economics

As an exercise, show that a saver can be made better or worse off by an interest rate
decrease. Note that a saver’s initial consumption plan is not feasible after the interest
rate decrease

Case: Britain benefits from lower interest rates

When you analyse the effect of interest rate changes on the economy
in different countries you have to take into account the amount of
debt at floating rates. Fixed rate debt is by definition not affected by a
change in the interest rate. In an economy with mainly fixed rate debt
and variable rate savings, interest payments on loans will not change
much if the interest rate falls but interest receipts from savings could go
down dramatically. American households paid $6 billion less in interest
payments between 1992 and 1992 but their interest receipts fell by $73
billion.
Britain is an exception among the big economies in the world in that
it is a net debtor if we consider assets and debts held at floating rates
by firms and households. A large majority of British homeowners hold
variable rate mortgages (90 per cent versus less than 20 per cent in the
US and 10 per cent or less in Germany, France and Japan). Also, less
than half of British company debt is at fixed rates whereas in the U.S.,
France and Japan it makes up 60 per cent or more. 13 13
See ‘Giving the
economy a fix’
Since Britain is a debtor, it is better off when the interest rate decreases
whereas the rest of the world could be worse off. Why is that? For a
borrower, when the interest rate decreases, the substitution and income
effects reinforce each other and cause him to borrow more. (Make a
graph to see this!) He ends up on the new budget line to the right of
his original position. The original consumption bundle is still feasible
which means that the borrower is at least as well off as before the fall in
interest rate.

Labour supply14 14
Read Solberg (1992)
Chapter 5; Varian (2006)
The standard consumer choice problem is easily rephrased in the context Chapter 9
of the labour supply decision. The ‘goods’ are leisure l and a composite
good c (with price = 1) which can be interpreted as ‘income available to
spend on consumer goods’. The budget constraint indicates how much
consumption is possible for each choice of leisure and its implied choice of
work time given a wage rate w. The worker may have non-labour income
M which is taken into account in the formulation of the budget constraint.
In addition to the monetary constraint, the consumer also faces a time
constraint: work time and leisure time have to add up to 24 hours per day
or, more generally, H hours per time period. The worker’s optimisation
problem, represented graphically in Figure 6.8, is: Max U(c,l) s.t. c =
M + w(H – l) and at the optimum the MRS equals the price ratio-w.

86
 Chapter 6: Topics in consumer theory

C
M+w(H-1)

C*

1
1* H

Figure 6.8: Labour supply

You should know how to modify the model to allow for progressive
income tax. The budget constraint becomes piecewise linear and concave.
Investment in education can be modeled as a decrease in M (fees,
opportunity cost, etc) and an increase in w (higher wages for the more
educated) so that the budget constraint becomes steeper.
Of course the basic labour supply model assumes that the worker has
a free choice of working hours or length of working week. In reality
the employer will want to place some restrictions on that choice, often
in the form of a lower bound on the number of hours worked. We can
nevertheless still use the model to study the employee’s reaction to these
restrictions. Statistics show a large variance of number of hours worked
per week among workers. In Britain more than five million people, or 20
per cent of the workforce, work more than 48 hours per week.15 There 15
See ‘Hard Labour’
is some evidence suggesting that people are working harder than they
want to in the sense that they would rather earn less and work less than
they currently do. Job satisfaction is higher for women part-timers than
for full-time workers.16 The Economist17 quotes a survey in which 80 per 16
See ‘The home front’
cent of British part-timers say they prefer working part-time to working 17
See ‘Sharing the
full-time and a third of full-timers in the EU said they would rather work burden’
fewer hours than have a pay rise. If workers have a strong preference for
working part-time, then firms which are willing to be flexible rather than
insisting on, say, a 40-hour work week can obtain labour at a lower cost.
We can show this by drawing the indifference curve through H – l (hours
worked) = 40 and the corresponding c (point A in Figure 6.9). We can
now determine how much lower the wage rate could be (while keeping
the worker as happy as before) if the worker is allowed to choose the
length of the working week. Rotate the budget line down until you find
a tangency (point B) with the indifference curve. At B the worker is as
well off as at A although his income has decreased compared to at A. An
interesting statistic in this context is that women part-timers earned only
72 per cent of the rate per hour of full-time women workers in 1992.18 18
See ‘The home front’

87
28 Managerial economics

Figure 6.9: Flexible versus fixed hours

As the wage increases, different choices between leisure and work
time are made. This relationship between wage and hours worked is
the labour supply curve. The effect of an increase in wage can be
decomposed in a substitution and income effect. The substitution effect
pushes in the direction of more work whereas the income effect pushes
for more leisure if you assume that leisure is a normal good. If the income
effect is large enough, we get a backward bending labour supply curve
where wage increases result in fewer hours of work.
A tax on labour earnings has the same effect as a decrease in wage: the
income and substitution effect counteract each other so that the worker
could end up working more or less whereas a lump sum tax only has an
income effect making the worker work more. As an exercise you could
make a graph illustrating the effect of a fixed rate income tax when the
labour supply curve is backward bending and show that introduction of
the tax reduces labour supply for low wage rates and increases labour
supply for high wage rates.
In line with the mainstream empirical work in this area, Battalio, Green
and Kagel (1981) find evidence of a backward bending labour supply
curve at high ‘wages’ for pigeons and rats! In their experimental work
they discovered that laboratory animals trade off income for leisure and
that leisure is a normal good. ‘Work’ for pigeons is typically key pecking
or treadle running whereas rats are conditioned to do ‘work’ tasks such as
lever pressing and wheel running. The animals have to perform the task
a certain number of times to get a reward. The authors claim that rats
and pigeons behave as if they were maximising a utility function U(c, l)
and they conclude, ‘Substitutability of income and leisure appears to be a
fundamental, biologically-based, law of behaviour.’

Example 6.5

A consumer has a Cobb-Douglas utility function, U(c, l ) = cαl1–α and

budget constraint c = M + w(H – l). To derive theindividual’s labour
supply curve, set the ratio of marginal utilities equal to the slope of the
budget constraint:

∂U/∂c αcα–1l1–α 1 αl 1
= α = w or =
∂U/∂l c (1–α)l–α (1–α)c w

88
 Chapter 6: Topics in consumer theory

Solving this for l gives (*)

(1–α)c
l=
αw
and substituting this expression into the budget constraints gives:
(1–α)c
c = M + w (H– αw ) or c = α(M + wH )
Substituting this expression for c into (*) leads to:

(
l = (1-α) H +
M
w
(
and the labour supply curve is therefore:
H – l = αH – (1–α) M .
w
The number of hours worked H – l, is increasing in w and hence, for
this particular utility function, the labour supply curve is not backward
bending.

The simple model discussed above can be used to analyse many interesting
questions. Suppose the government pays a fixed amount in welfare benefit
for unemployed workers (i.e. if they work zero hours they get the benefit
but if they work at all, they do not).

You should be able to show that this provides an incentive to be idle especially if the
wage rate is low

In many countries, welfare or unemployment benefit is withdrawn

completely as soon as the recipient enters the labour market. This ‘welfare
trap’ has detrimental effects on work incentives. Contrast this with the
Canadian province of Newfoundland which has recently introduced a
plan to avoid the problem of sudden cut-off of benefit. Every individual
is guaranteed a basic income if he does not work. Rather than losing this
welfare payment when people start working they in fact receive a higher
payment in the form of a work subsidy. The income supplement then
levels off and eventually decreases to zero at an income of C$42,500 for
a family of four.19You should be able to show graphically that this plan 19
See ‘New found plans’
reduces voluntary unemployment and that it may be possible to reduce
total welfare payments (by substituting wage subsidies for unemployment
benefit) without making the recipients worse off.
In a recent paper, Biddle and Hamermesh (1990) tackle a subject which
has long been ignored not only by economists but by social scientists in
general: the demand for sleep. Given that most survey respondents when
asked state that they sleep more hours than they work, understanding the
demand for sleep seems crucial for developing a model of allocation of
time by consumers. Biddle and Hamermesh suggest that there are three
approaches, or hypotheses, to incorporating the demand for sleep in the
standard leisure-consumption tradeoff model we have worked with in this
chapter.
1. The minimum amount of sleep an individual needs is assumed to be
biologically determined and consumers do not derive utility from sleep.
Under this hypothesis, an individual sleeps according to his biological
minimum and we can set the amount of time an individual has
available for work and leisure, H, equal to 24 hours per day minus the
amount of sleep.
89
28 Managerial economics

2. A more plausible hypothesis is that the individual derives utility from

hours spent sleeping as he derives utility from leisure hours in general.
In terms of the labour supply model, the question arises whether we
make a mistake by bunching sleeping with other leisure activities.
3. There is some evidence that sleep affects productivity and hence
wage. Clearly, if we are going to take this hypothesis seriously, our
labour supply model has to be revised rather dramatically. The wage
rate becomes endogenous as it is affected by the consumer’s choice of
sleeping time: w = we+ wsts where we is the exogenous part of the wage
rate and ws measures the sensitivity of productivity or wage to time
spent asleep ts. Total time H is made up of working, waking leisure and
sleeping: H = tw + tl + ts The consumer’s problem can thus be written as:

Max U(c, ts, tl )s.t. c = M + (we + wsts) (H – ts tl ).

Biddle and Hamermesh come up with the following empirical findings:
• people with higher wages sleep less
• if the wage rate increases, the number of hours men work does not
increase but sleeping time is reduced in favour of leisure time
• consistent with prior research, men’s supply of work hours is much less
sensitive to changes in the wage rate than women’s. The labour supply
elasticity is measured as –0.021 for men and 0.191 for women
• sleeping time seems to be a normal good: in a sample of economies, a
higher GNP is associated with significantly longer sleep duration.

Risk and return20 20

Read Solberg (1992)
Chapter 7; Varian (2006)
Chapter 13
Further reading
If, after reading this section, you are interested in learning more about
finance (and if you have a lot of spare time!), I recommend:
Brealey, R.A. and Myers, S.C. Principles of corporate finance. (Singapore:
Mc-Graw Hill, 2005) eighth edition [ISBN 0073130826] this a very
accesible but thorough text.
In the chapter on decision analysis, I touched on the subject of investment
choice between risky assets. Suppose you are offered two alternative
investment opportunities each with a given set of uncertain outcomes and
a probability distribution over these outcomes. A straightforward approach
would be to use the expected utility model (i.e. calculate a weighted sum
of utilities of each outcome, with the weights equal to the probability
of the outcome materialising) and choose the asset which delivers the
highest expected utility. Although this approach is valid, major progress
in financial economics was made by using an assumption to simplify
this process. Harry Markowitz (1959) who, in 1990, was awarded the
Nobel Prize in Economics jointly with William Sharpe and Merton Miller,
proposed a model of choice among risky assets where utility depends on
expected value and standard deviation of income rather than on the entire
probability distribution.
When risk averse investors decide on their investment in a financial
asset they consider not only expected return (dividend or interest
receipts plus capital gain or loss) but also the risk involved. When given
a choice between investment A which gives a return of 10 per cent for
sure and investment B which gives five per cent or 15 per cent with
equal probability, A is chosen. In general, an investor is only willing to
take on higher risk when there is a compensating increase in expected

90
 Chapter 6: Topics in consumer theory

return (i.e. his indifference curves in the risk-return plane are upward
sloping: expected return is a ‘good’ whereas risk is a ‘bad’). The level of
risk associated with an investment is usually measured as the standard
deviation of the return. (This may be a good time to look for your statistics
notes!) Investments with larger standard deviations have to be rewarded
by greater expected returns. For example, average annual returns for
venture capital from 1945 to 1993 were 15.9 per cent and the standard
deviation in the returns was 25.5 per cent; stocks included in the S&P
500 averaged a return of 11.7 per cent over the same period and a
correspondingly lower standard deviation of 16.3 per cent. The exceptions
to the rule were gold and silver which had low average returns of about
five per cent and very high standard deviations of 25.8 per cent and 55.8
per cent respectively.21 21
See ‘Risk and return’

Given an investor’s preferences we could predict which of a list of assets,

with given risk and return, he would invest in. In reality, however, the
investor is not restricted to investing his entire budget in one asset. He
could invest in several assets or a portfolio. Investing in several assets
reduces risk when the returns on the assets are not perfectly positively
correlated. I will come back to this later. I want to turn to a simple
example illustrating the role of correlation between returns first. Suppose
I can invest in a business selling umbrellas and/or a business selling ice
cream. The profitability of these businesses depends on the weather.
Assume I have an investment budget of £10,000 and my net pay-off from
investing £1 in either business is according to the table below.

weather umbrellas ice cream probability

good –1 1 1/2
bad 10 –1 1/2

If I invest my £10,000 only in umbrellas, my expected pay-off is £45,000

and if I invest only in ice cream, my expected pay-off is £45,000.
However, suppose I invest £5,000 in both then, when the weather is good,
I gain £45,000 for sure and, when the weather is bad, I gain £45,000 for
sure. In other words, rather than running considerable risk by investing in
one business, I can eliminate all risk without reducing my expected return
by diversifying my investment.
Before you get too excited about this I have to warn you that in practice
it is impossible to find two stocks which have perfect negative correlation
and so you have to settle for some uncertainty. By choosing a portfolio
carefully this uncertainty can be reduced. Let us see how that works if
we have two assets A and B with expected returns rA and rB and standard
deviation of the returns A and B. Empirically it turns out that the normal
distribution works well to describe the volatility of returns but we do not
have to make any assumption about the distribution here. Assume rA < rB
and A < B otherwise we could do no better than invest in the asset with
the highest return and lowest risk. If of every £1 in my budget I invest a
share α in asset A and the remaining share (1 – α) in asset B, the expected
return and risk of my portfolio are given by:
rp = αrA + (1–α) rB (7)
and
Varp = α2 σA2 + (1–α)2 σB2 +2α (1–α) ρσA σB (8)
where ρ is the correlation between the returns of assets A and B. This
means that, by varying the share of my budget I invest in A, it is possible

91
28 Managerial economics

to attain the combinations of expected return and risk given by (7) and
(8). Consider the (easier) case of ρ = 1 first. For ρ = 1,
Varp = (α σA + (1–α) σB)2 or σB = α σA + (1–α) σB

which combined with (7) implies that (p, rp) lies on a straight line
between (A, rA) and (B, rB) as is shown in Figure 6.10.

r
B
ρ <1
r ρ =1
A

σ σ
A B σ

Figure 6.10: Portfolio consisting of assets A and B

For ρ < 1, p < A + (1 – )B and hence the combinations of risk and
expected return which are feasible will be to the left of the upward sloping
line segment in Figure 6.10. The lower the correlation, the lower the
risk of the portfolio. Ideally the two assets in the portfolio should have
negative correlation but, as mentioned before, in practice finding stocks
which are negatively correlated is difficult.
We could extend our exercise to more than two assets and determine the
possible risk – expected return combinations which are possible by varying
the shares of the budget allocated to each asset. The set of these feasible
combinations could look like the figure below with the x’s representing
expected return – risk combinations for individual assets in the portfolio.
Given this feasible set which portfolio would you choose? Of course it will
depend on how risk averse you are but whatever your utility function is,
we can slim down the choice set using a dominance argument. Given two
portfolios with the same , you would never choose the one with the lower
r because it is dominated by the portfolio which has the same risk and
a higher return. So, for each possible risk level , we need only consider
the portfolio which gives the highest expected return. Similarly, given
two portfolios with the same r, you would always prefer the one with the
lower risk. Hence, for portfolios with a given level of r, the portfolio with
the lowest s dominates the others. The portfolios which are not dominated
are called efficient and are on the curve between A and B in Figure 6.11.
There are computer programs available which calculate the set of efficient
portfolios given the risk and expected return of each available asset.

92
 Chapter 6: Topics in consumer theory

r
B
C
x
x
x
xx
x
xx
x x x
r x
f xx
xx
x
x x x

xx
x

Figure 6.11: Efficient portfolios

Now suppose we have a risk free asset (e.g. a Treasury bill) available in
addition to our collection of risky assets. The risk-free asset gives a return
rf and (by definition) risk f = 0. If we invest a share  in the risk-free
asset and the remaining in a portfolio (σp, rp) then we know from (7) and
(8) that our investment will have an expected return of r = αrf + (1 – α)rp
and a variance of Var = (1 – α)2σp2. This means that, by varying the share 
invested in the risk free asset, we can attain all points on the line between
(0, rf ) and (σp, rp ). We could do this for any portfolio in the feasible set but
of course the portfolios on the efficiency line are more desirable and, of
those, the portfolio at which the highest line from (0, rf ) is tangent to the
efficiency line is the optimal one to combine with investment in the risk-
free asset (point C in Figure 6.11).
In fact not only points between (0, rf ) and C are feasible but points to the
right of C as well. By borrowing money at interest rate rf and investing
it in the optimal portfolio we can extend our ‘budget line’ to the right
of C. Clearly, investors should choose a point on this budget line and
lend or borrow money at fixed interest according to whether they pick
to the left or the right of C. Exactly which combination of risk-free and
risky investment is chosen depends on the investor’s utility function, as
represented by the indifference curve in the figure above. It is not difficult
to derive the equation for the budget line through (0, rf ) and (σp,rp ): r = rf +
((rp – rf )/σp)σ. The slope of this line (rp – rf )/σp is the price of risk which
at the investor’s optimum equals his MRS.

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• derive demand functions and labour supply functions given a utility
function
• derive the Slutsky equation
• derive Slutsky and compensated demand functions
• calculate elasticities
• explain the state-contingent commodities model
• set up the intertemporal choice problem and analyse the effect of
changes in the interest rate
93
28 Managerial economics

• set up labour supply model

• derive the shape of the feasible set of risk-return combinations for
portfolios of two risky assets.

Sample exercises
1. Goods x and y are perfect substitutes and consumers have utility
functions of the form U(x, y) = x + y. Consumer i has income mi.
Derive and draw his demand for good x. Derive and draw the market
demand for good x.
2. True/False. If the income elasticity is positive, the compensated
(Hicks) demand curve is steeper than the ordinary demand.
3. Joanna likes lipstick (x1) and earrings (x2) which give her utility
U(x1, x2) = x1x2. Her pocket money is M per week and lipsticks and
earrings cost p1 = 4 and p2= 1 respectively. Suppose p1 falls to 1. By
how much would her pocket money have had to increase to make
her as well off as after this price change? By how much could her
pocket money be reduced while keeping her as well off as before the
price decrease? Which is the compensating variation? Which is the
equivalent variation? Calculate her increase in consumer surplus due
to the price decrease.
4. The price elasticity of demand for wixies is 1.5. The wixies producers
form a cartel and agree wixies quotas which will have the effect of
reducing output such that price rises by 10%. What is the effect on
total revenue? How does your answer change if the price elasticity of
demand is 0.75?
5. Using the state-contingent commodities model, show graphically
that, for full insurance, a risk averse individual is willing to pay an
insurance premium which is higher than his expected loss.
6. A consumer lives for two periods and receives income 100 in each
period. In the first period he can invest some of his income in a
stock which has a return r of 10% or 20%, each equally likely. His
utility function U(c1, c2) is a function of his consumption levels in
both periods: U(c1, c2) = c12 + 0.7 c22 for c1, c2 ≥ 0. Find the optimal
consumption plan.22 22
Hint: maximise
expected utility
7. A worker can choose the fraction of a day, y, that he is not working
(so he works a fraction 1 – y of a day). He derives utility from
consumption goods c (assumed to have price = 1) and leisure y
according to U(c, y) = c4y2. He has a fixed nonlabour income of M per
day and the wage rate is w per day. Write his budget constraint and
derive his labour supply curve. Is it backward bending at high wages?
8. Empirical studies show that (a) for male employment small changes
in the wage rate do not affect the number of hours worked and (b)
the number of hours women work increases when taxes fall because
of their increased labour market participation. Explain these findings
graphically.
9. Portfolio theory assumes that investors care about expected return
and risk. Draw indifference curves for a risk neutral investor.
10. The return on stock A will be –10% or +20%, each with probability
0.5. The return on stock B will be –5% or +15%, with probabilities
0.3 and 0.7 respectively. The correlation coefficient between the
returns of the two stocks is 0.6. Calculate the expected return and
the variance of the return for each stock. Calculate the covariance

94
 Chapter 6: Topics in consumer theory

between the returns of the stocks. Can you decrease your risk by
investing in a combination of these two stocks rather than in either of
the stocks?
11. Suppose the risk-free rate of return is 5% and you consider investing
in a risky asset with r = 10% and σ = 4%. Show all combinations of
risk and expected return which are feasible if you invest in a portfolio
consisting of the risk-free asset and the risky asset. If you are willing
to carry a risk of 3% but not more what is the best portfolio?
12. During the past six years the returns on Air Angleterre and French
Airways have moved as follows:

Year 1 2 3 4 5 6
AA 25 –16 18 5 13 2
FA 10 –3 15 3 20 8

Calculate all the parameters you need to derive the locus of feasible
expected return-risk combinations if you invest a share of your budget
in AA and the remainder in FA.

95
28 Managerial economics

Notes

96
 Chapter 7: Production, factor demands and costs

Chapter 7: Production, factor demands

and costs

Aims
The aim of this chapter is to consider:
• how the smooth isoquants continuous production model relates to the
linear production model
• the difference between conditional and unconditional input demands
• the problems involved in estimating production and cost functions
• the MRP = ME rule for determining unconditional input demands
• why imposition of a minimum wage may increase employment in a
monopsony
• why the industry demand for an input is not necessarily the horizontal
sum of firm demands
• why MR should equal MC in each producing plant of a multiplant firm.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• derive the cost function from a given production function i.e. derive
conditional input demands (input as a function of output) and write
the expenditure on these inputs as a function of output
• derive unconditional input demands for monopsony, monopoly
and price-taking firms
• solve the problem of which plants(s) to operate and at what level for
a two plant firm given the plants’ cost functions and the demand
function.

Essential reading
Varian, H.R. Intermediate Microeconomics. Chapters 18, 19 and 26.

Further reading
‘Doleful’, The Economist, 9 October 1993, p.17.
‘The trade unions scent a victory’, The Economist, 3 September 1994, pp.27–28.
‘Turning in to the future’, The Economist, 5 September 1992, pp.27–28.
Moroney, J. ‘Cobb-Douglas production functions and returns to scale in US
manufacturing industry’, Western Economic Journal 1967, pp.39–51.

Introduction
What I want to do in this chapter goes under the heading of ‘theory of the
firm’ in most microeconomics textbooks. However, over the last 15 years or
so, there has been a growing interest in what determines the boundaries
of the firm, how firms are organised internally, how their organisational
design can be explained in terms of efficiency and so on. All of these topics
are elements of a ‘theory of the firm’. The neoclassical view of the firm

97
28 Managerial economics

as a black box, hiring or buying inputs, transforming inputs into outputs

and selling these outputs so as to maximise profits has ceased to be the
only theory of the firm worth studying. This is not to say that traditional
neoclassical economics has nothing to contribute to our understanding of
how firms operate. It in fact complements the new organisational theories
in that it emphasises the technical relationships within the firm and its
market interactions. Furthermore, just as the theory of the consumer
is used to derive market demand, the neoclassical model of production
provides a building block in the determination of market supply.

Production functions and isoquants

The theory of production is very similar to the theory of the consumer.
The same type of reasoning is used, for example, to find conditional input
demands (amount of input used to produce a given level of output) as
to determine the consumer’s optimal consumption bundle. Isoquants are
contour lines of the production function whereas indifference curves are
contour lines of the utility function. Given this similarity and that you
have studied this material in your introductory economics course, I will
only give a brief review here. Although the models in this chapter are
presented in terms of the firm producing one output using one or two
inputs, most of the analysis can be generalised to multiproduct firms using
many inputs.
It is important to realise that a production function f(x1, x2) indicates
the maximum amount of output which can be produced for the given
quantities of inputs x1 and x2. Similarly, any input bundle (x1, x2) lies on
the isoquant corresponding to the highest level of output which can be
achieved using these levels of inputs. To derive input demands we will
need to know the slope of the isoquants. Since output is constant on the
isoquant, we have:
∂f (x1, x2) ∂f (x1, x2)
df (x1, x2) = dx1 + dx2 = 0
∂x1 ∂x2

so that the slope of the isoquant or the marginal rate of technical

substitution (MRTS) equals the ratio of marginal products:
∂f (x1, x2) ∂f (x1, x2)
dx2 / dx1 = – / ≡ – MP1/MP2
∂x1 ∂x2
The MRTS is the counterpart of the MRS in consumer theory. It shows the
rate at which one input can be substituted for another while maintaining
the same output level. For convex isoquants, the MRTS decreases in x1
which means that if a large quantity of x1 (and hence a relatively small
quantity of x2) is used, we can reduce the amount of x1 significantly while
increasing the level of x2 slightly and still produce the same amount. The
MRTS is sensitive to the scale of measurement which is why the elasticity
of substitution is sometimes used to measure the curvature of the isoquant.
It is defined as:

σ = % change in the input ration x2 / x1 (1)

% change in MRTS

In Figure 7.1, the change in the input ratio corresponds to the difference in
slope of rays OA and OB and the change in the MRTS corresponds to the
difference in slope of the lines a and b tangent to the isoquant. Clearly, if
for the input ratios determined by A and B, the change in MRTS is larger,

98
 Chapter 7: Production, factor demands and costs

then the isoquant is more convex and  is smaller. The extreme values for
s are 0 for perfect complements (L shaped isoquants) and ∞ for perfect
substitutes (linear isoquants).

Activity
Why? Convince yourself by drawing a graph.

x2 B

a
0 x1

Figure 7.1: Elasticity of substitution

To justify the assumption of smooth continuous isoquants it is useful
to think of production processes (or production activities) represented
by a ray from the origin. The angle of the ray indicates in which fixed
proportions the inputs have to be combined in that particular production
process. Suppose that using production process 1 (PP1) you can produce
100 units of output using 10 man-hours and 50 units of capital; using
production process 2 (PP2) you can produce 100 units of output using 100
manhours and 10 units of capital. For both production processes output
increases and decreases proportionally with increases and decreases in
inputs. If you want to produce 100 units you have a choice between using
exclusively PP1 or PP2 or a combination of the production processes. For
example, you could produce 50 units using PP1 (employing five manhours
and 25 units of capital) and 50 units using PP2 (employing 50 manhours
and five units of capital). If you combine the production processes, the
combination of the total amount of inputs used will be on the straight line
AB in Figure 7.2.

capital

PP1

50 A
PP2
B
10
man-hours
10 100

Figure 7.2: The linear model

Although the linear model of production discussed above is appealing, it
is not convenient to work with mathematically. The assumption of smooth
convex isoquants, which are convenient, can be seen as a reasonable
approximation to the linear model when there are many possible
production processes.

99
28 Managerial economics

Students often get confused about the difference between ‘diminishing

returns’ and ‘decreasing returns to scale’. Whereas the first expression
refers to decreasing marginal product of one input, the latter refers to
what happens to the level of output if all input levels increase by the same
proportion (e.g. for a production function f of two inputs x1 and x2,
f (x1, x2) < f(x1, x2) for  > 1 indicates decreasing returns to scale). Note
that it is possible for a production function to have increasing returns to
scale for some output levels and decreasing returns to scale for others. You
should be able to show that the Cobb-Douglas production function f (x1, x2)
= ax1b x2c has decreasing, constant or increasing returns to scale depending
on whether b + c < = > 1.
The notion of decreasing returns to scale is difficult to understand in a
realistic setting. If we can produce 100 units with a given set of inputs
why can’t we just duplicate the set of inputs to produce 200 units?
It must be that we are not really able to duplicate all inputs and we
are therefore keeping some inputs constant. Maybe it is impossible to
duplicate the entrepreneurial input or the R&D effort without loss of
quality. Increasing returns to scale can arise for example when an increase
in man hours allows for specialisation. Also inventories of spare parts and
back-up equipment may not have to increase with an increase in output.
Sometimes a basic engineering relationship is responsible for generating
increasing returns to scale. Think of the output of an oil pipeline as
the amount of oil which flows through per time unit. This output is
proportional to the square of the radius r of the pipeline (the cross section
has area πr2) whereas the input in terms of materials is proportional
to the radius (the circumference is 2πr). Ignoring other inputs such as
horsepower needed to create a flow, output is a quadratic function of
input. In real world industries, production characterised by constant
returns to scale or close to constant returns seems the norm.1 1
See Moroney (1967)

The production function reflects the current state of technology; if there is

technological progress, the production function changes. If you consider
a production function with one variable input, say labour, then the effect
of technological progress is to shift the production function up: with the
same amount of labour, a higher level of output can be achieved. For two
variable inputs, the effect of technological progress can be visualised on
the isoquant map: isoquants shift inwards towards the origin. For example,
in the steel industry, American and Japanese ‘mini-mills’ – which are a
quarter of the size of giant European firms (and involve a quarter of their
capital investment) – use cheap scrap rather than iron ore and coke.
To make a tonne of steel they use a quarter of the labour needed in the
European firms.
Empirical estimation of production functions is not easy. One popular
approach is to use statistical techniques such as regression analysis on
time series or cross section data (data on inputs and outputs for several
firms or plants in the industry) to estimate a specific functional form. The
Cobb-Douglas specification plays a role which is similar to the constant
elasticity demand function in consumer theory because of its mathematical
convenience: the parameters are estimated directly in linear regression.
When cross section data are used to estimate production functions the
implicit assumption is made that the same technological possibilities
are available to all firms in the industry and hence the same production
function applies to all of the firms. Similarly, when using time series data,
it is assumed that the state of technology is stable over the period under
study. Apart from measurement problems (it is difficult to determine
precisely the amount of all inputs used) there is the problem of whether

100
 Chapter 7: Production, factor demands and costs

the observations reflect efficient production, which is what the production

function should describe. An alternative to using existing data is of course
to create new data through experiments.

Firm demand for inputs

It is important to make a distinction between the following two questions:
• given that the firm wants to produce a given level of output, what is
the optimal choice of inputs?
• what are the firm’s profit maximising input demands and hence what is
the firm’s optimal level of output?
The answer to the first question gives the conditional input demands
(i.e. the amount of inputs the firm demands conditional on the level of
output it produces). The conditional input demands are used to determine
the cost function as we will see later on. The answer to the second
question gives the unconditional input demands (i.e. the amount of
inputs the firm hires or buys to produce the optimal level of output).
Clearly the optimal level of output is determined by the structure of the
market in which the firm operates. As we will see, there are two avenues
to determining the firm’s unconditional demand for inputs. We can find
the optimal output level first and then set demands for input equal to
the conditional demands for this optimal output level. Alternatively,
we calculate unconditional demands directly and the optimal output is
determined as the maximum output level the firm can produce given these
levels of inputs.
Let us start by deriving the conditional input demands. The easiest
scenario to analyse is the case of a firm which is a price taker on the two
input markets. For such a firm the locus of points with equal expenditure E
on inputs is a straight line, an isocost line (p1 x1 + p2 x2 = E), the analogue
of the budget constraint for the consumer. Given the input prices and
the output level it wishes to produce with corresponding isoquant I (see
Figure 7.3) the firm wants to minimise expenditure by finding the lowest
(furthest to the south-west) isocost line tangent to the isoquant I. The
graph looks the same as for the consumer problem.

x2

isocost line
x2*

0 x1
x1*

Figure 7.3: Optimal input mix

At the optimum (minimum cost) solution (if it is not a corner solution),
the slopes of isoquant and isocost line are equal, hence:
p1 / p2 = MRTS = MP1/MP2.

101
28 Managerial economics

This implies that, at the optimum, the elasticity of substitution, defined in

(1) equals:
 σ = % change in the input ration x2 / x1
% change in p1 / p
Hence, at the optimum, the input ratio changes significantly when the
input price ratio changes if the inputs are substitutes (σ large) whereas, for
complementary inputs (with σ small), the input price ratio has to change
dramatically to have an effect on the combination of inputs used.
For more than two inputs the optimality condition for cost minimisation
subject to an output constraint generalises to ‘the ratio of MP to price is
equal for all inputs which are actually used’. The intuition behind this
result is easy to grasp. If it were the case that the MP-price ratio of input
1 was higher than that of input 2 then, by diverting some money from the
budget for input 2 to that for input 1, you could increase output or you
could keep output constant and reduce total expenditure. It is important
to remember that the input levels which satisfy the tangency conditions
are the input demands for the given level of output. We can determine
what happens to these input levels when the output level q changes. The
relationships we find then are precisely the conditional input demands
x1(q), x2(q). Clearly these conditional input demands depend on the prices
of the inputs and the substitutability of the inputs.
Using this simple model of conditional input demands, we can make some
interesting predictions. When the interest rate (cost of capital) falls, firms
tend to invest in labour-saving equipment. If labour costs differ between
economies, economies with relatively cheaper labour will tend to use
labour intensive production processes, even when they have the same
technological knowledge.

Example 7.1

The production function is given by q = f (x1, x2) = a x11/2 x21/2. To

determine the optimal use of inputs, set the MP-price ratio’s equal:
(a/2)(x2 /x1)1/2 /p1 = (a/2((x1/x2)1/2 /p2.
This can be simplified to:
x1/x2 = p2/p1 or x1 = x2(p2 /p1).
Using the production function we can substitute x2 = (q/a)2 /x1:
x1 = (q/a)2 (p2 /p1)/x1.
The conditional input demands are:
x1 = (q/a) (p2/p1)1/2 and x2 = (q/a) (p1 /p2)1/2

We are now in a position to discuss the abuse of labour productivity

measures to construct arguments about technological change and to make
international comparisons. To view an increase in labour productivity as
technological progress is clearly wrong when the increase could be due
to a rise in wages which affects labour’s MP-price ratio so that less labour
will be used. If less labour is used then, because of the law of diminishing
return, its MP has to increase. When international comparisons regarding
productivity are made, what is often ignored are the differences in capital
intensity between production techniques used in different countries. If
Germany uses relatively capital intensive production methods (maybe
because of its high wage costs), then its labour productivity will be higher
than elsewhere.

102
 Chapter 7: Production, factor demands and costs

Suppose we want to determine how much of each input to use while at

the same time deciding how much output q to produce. Assume there is
only one variable input labour L with unit cost w. (The model generalises
to any number of inputs.) We will first consider the relatively simple case
where the firm is a price taker in the input and output markets. The firm’s
problem is to use the amount of input necessary to maximise profit:
max p f (L) – w L
where p is the price per unit of output. Let us assume that the second
order condition is satisfied. The solution to the maximisation problem is
obtained by setting L such that:
p f ' (L) = w, where f ' = df /dL.
The marginal return from labour should equal the marginal expenditure
on it. This implicitly gives the unconditional demand for labour:
L = f ' –1 (w/p), where f ' –1 is the inverse function of f '.
Of course we still have to check that positive profits are made at this
solution. Substitute p f ' (L) = w in the profit function to get:
p f (L) – p f ' (L) L
which is positive when:
f (L)/L > f '(L)
(i.e. when average product of labour (AP) exceeds marginal product of
labour (MP)). We therefore have to qualify our result by saying that the
unconditional demand for labour is:
L = f ' –1 (w/p) when AP > MP and zero otherwise.

Example 7.2

A competitive firm sells its output at price p and has production

function q = L1/2. To calculate the unconditional labour demand set
p × MP equal to w:
p (1/2 L–1/2) = w or L = (p/2w)2.
Labour demand is increasing in price of output and decreasing
in the wage rate. We have to check that no losses are incurred which
is the case since:
AP > MP (1/L1/2 > (1/2) L –1/2).

Now let us consider the more complicated case of a firm which is not a
price taker. In the output market this means that the market price of the
firm’s output depends on how much it decides to sell. In the input market
similarly it means that the price the firm pays per unit of input depends
on the quantity it buys. As an example consider a university which is the
only employer of cleaners in a small town. If the university wants to hire
10 cleaners it can offer a wage of £7 per hour. However, if 11 cleaners are
needed, the wage has to rise to £7.50 per hour (the supply of cleaners
is upward sloping). If the university cannot discriminate between the
workers, the marginal cost of the 11th worker is not just his salary of £7.50
but also the increase in the other workers’ salary: £7.50 + £0.50*
10 = £12.50. Intuitively we can deduce that, if a firm is not a price taker, it
will hire fewer people because it takes into account the effect of its demand
on the wage level whereas a price taking firm ignores this externality.
A monopsony is a firm which is the only buyer in a market with many
sellers who take the market price as given. A monopsony could be a

103
28 Managerial economics

monopoly or it could sell its output in a competitive market. In some

countries farmers have to sell their produce to a national marketing board
which (if the domestic market is isolated) operates as a monopoly or
sells the produce on the international market at the world price (in this
case price taking is likely if the country does not represent a large share
of the world market). Let us now consider the most general case of the
unconditional input demand model, the profit maximisation problem for
the monopsonist – monopolist:
max p(f (L)) f (L) – w(L) L.
The monopsonist takes into account the supply of labour function w(L) (i.e.
he knows that if he hires more workers he has to offer a higher wage). The
first order condition for the optimisation problem is:
p' f ' f(L) + p f ' (L) – w(L) – L w' (L) = 0
or
(p'f(L) + p) f ' (L) = w(L) + L w' (L).
The left-hand side is the marginal return from L or its marginal revenue
product (MRP): the product of marginal revenue and MP and the right-
hand side is the marginal expenditure (ME) on L. ME is larger than the
wage if we assume supply of labour to be upward sloping (w' > 0). The rule
MRP = ME captures the demand for inputs in the most general setting; it is
applicable to any combination of market power or price taking behaviour
in the input and output market. The model of course reduces to the simple
case we analysed above: for a price taker, marginal revenue equals price
and marginal expenditure equals wage. Figure 7.4 illustrates how the
monopsonist’s demand for labour is determined. Since ME equals MRP
and exceeds the wage, we are justified in saying that a monopsonist pays
workers below the value of their marginal product.

ME

w*
MRP

L
L*

Figure 7.4: The MRP rule – monopsony

Many textbooks make comparative statements of the type ‘since marginal
revenue is less than the price for a monopolist, a monopolist uses less input
than a competitive firm’. The implicit assumption here is that a competitive
firm operates with the same production function as the monopoly. Also it
seems necessary to assume that it is the only firm in the industry because,
otherwise, would it not be more interesting to compare the industry
demand with the monopoly demand? However, if the firm is the only firm in
the industry, assuming price taking behaviour is not very realistic.

104
 Chapter 7: Production, factor demands and costs

Example 7.3

The demand in an industry is given by p = 100 – 2Q, where Q is

industry output. The production function for each firm in the industry
is q = f (L) = 2 L and the wage rate is fixed at w = 4. If the industry
consists of one monopolistic firm, we determine its labour demand
from MRP = ME. Marginal revenue equals MR = 100 – 4 q = 100 – 8L
and MP = 2. The optimality condition thus reads (100 – 8L) 2 = 4 or
L* = 98/8 which results in output of q* = 98/4. If the industry consists
of a number of competitive firms, then each firm will set MRP = p MP
equal to w, or 2p = 4. At the optimum, p has to equal 2 so that industry
output equals Q = (100 – 2)/2 = 98/2 (from the demand function). This
requires L* = 98/4 units of labour. (Since each firm uses q/2 units of
labour, the industry demand for labour is Q/2.)

Case: monopsony and minimum wages

An important objection to minimum wage policy is that it tends
to create rather than solve problems for the people it is trying to
help because of its effect on unemployment. This argument is very
compelling: increase the wage rate and you will reduce the demand
for labour, creating unemployment; at the same time a higher wage
induces more people to enter the labour market, thereby further
increasing unemployment. Surprisingly, the unemployment effect of
introducing a minimum wage in a sector differs according to whether
the sector is a monopsony. Figure 7.5a represents a labour market with
firms acting as price takers. The equilibrium wage w* and employment
L* are determined by the intersection of supply and demand.
Imposition of a minimum wage wmin is equivalent to making the supply
curve horizontal at wmin up to where the horizontal line intersects
the original supply curve. No labour is available at wages below wmin
and, at wages above wmin, labour supply is unchanged. The result is
unemployment, corresponding to the distance AB. Employment has
decreased from L* to L'.

ME

w(L) MRP
supply
demand w(L)
A B
wmin
w* w*

L' L* L L* L' L
(a) Price takers (b) Monopsony

Figure 7.5: Minimum wages

Figure 7.5b represents the situation in monopsony. The effect of
the minimum wage on the labour supply curve is as for the price
taking sector. However for the monopsonist this has a dramatic
effect on ME. He can now hire as many people as are willing to work
for the minimum wage without driving wages up. His ME curve is
now horizontal at wmin, up to the intersection with the supply curve

105
28 Managerial economics

at L′ from where it coincides with the original ME curve since the

expenses on labour are E = wmin L for L < L′ and w(L)L otherwise. The
monopsonist ends up employing L′ workers so that employment has in
fact increased!
Empirical studies suggest that minimum wages do not automatically
reduce employment. Consider the recent rise in employment in New
Jersey’s fast-food industry. In 1992 New Jersey had raised its minimum
wage from $4.25 per hour to $5.05 per hour whereas neighbouring
Pennsylvania stuck to $4.25. Employment in New Jersey rose by 13 per
cent more than in it did in Pennsylvania. A recent forecast suggests that
in Britain, a minimum wage equal to half male median earnings (£4
per hour) would have no significant impact on jobs if it did not apply
to young workers under 21 years of age.2 This is not to say that all the 2
See ‘The trade unions
evidence points towards harmlessness of minimum wage legislation. In scent a victory’
France, where the minimum wage is more than 50 per cent of average
earnings, a quarter of people under 25 years old are unemployed.
Contrast this with the United States where the minimum wage is
approximately 30 per cent of average earnings and there is far less
youth unemployment.3 3
See ‘Doleful’

The model of unconditional input demand is interesting because it

provides a link between input and output markets. It allows us to predict
the effect of changes in the output market, such as moves towards more
or less concentration, on the input market. Price wars in the PC market
have the effect of reducing the MRP of microprocessors putting pressure
on microprocessor prices. Similarly, fare wars in the airline industry have
an effect on the aircraft market. As satellite and cable TV gain ground and
more TV channels compete for writers, directors and producers, it is likely
that wages in these professions will increase.4 4
See ‘Turning in to the
future’
The model of unconditional input demand summarises the firm’s problem.
Once we have determined the unconditional demand for inputs for a firm
we have solved the problem of how much to produce (i.e. the maximum
amount which can be produced given the inputs available) and what price
to charge (this can be read off the demand curve given the output level).

Industry demand for inputs

In the consumer theory chapter we concluded that the market demand
curve is simply the horizontal sum of individual demand curves. Since I
am devoting a separate section to the topic of industry demand for inputs
you may already suspect that the situation is not quite as simple here. Of
course, if the industry is a monopoly we have no aggregation problem:
the industry demand is the firm demand. So let us consider the case of
demand for an input, say labour, by a competitive industry. Each individual
firm in the industry takes the input price or wage w as given and
determines its requirement from MRP = p MP = w where p is the output
price. Now suppose the wage decreases from w0 to w1. Each firm, given
p, wants to hire more workers. However, p is not fixed in the sense that
when all firms in the industry start hiring more workers and producing
more output, the output price falls (if the demand curve does not shift).
When the output price falls, each individual firm’s demand for labour, and
hence the horizontal sum of the demands ∑MRP, shifts to the left, as is
indicated on Figure 7.6 for a drop in price from p0 to p1. We conclude that
the industry demand (in bold on the figure) is steeper or less elastic than
the sum of firm demands.

106
 Chapter 7: Production, factor demands and costs

There are alternative possible scenarios however. In the analysis above

we have ignored the effect of entry and exit in the industry. If the lower
wages lead to positive profits, new firms will enter the industry with two
counteracting consequences:
• these new firms help to increase total industry output and hence
depress output price but
• there are now more firm demands to sum horizontally.
Depending on which of these effects is strongest, the industry demand
could be more-or-less elastic than ∑MRP. This analysis has a mirror image
in the determination of market supply.5 5
See the section on
‘Perfect competition’ in
Chapter 9
w

w0

w1
∑ MRP (p0)

∑ MRP (p1)

Figure 7.6: Industry demand for inputs

Example 7.4

In Example 7.2 we determined the demand for labour by a competitive

firm with production function q = L1/2 as L = (p/2w)2. Suppose there are
100 identical firms in this industry and industry demand is given by
p = 50 – Q/50 where Q is total industry output (Q = 100q). Applying
p MP = w for each firm but now taking the effect of industry output on
price (via the demand function) into account, we find:
(50 – 2q)(1/2)L–1/2 = w
and after substituting the production function:
(50 – 2L1/2)/(2L1/2) = w or L = (25/(w + 1))2.
The industry demand for labour is thus 100(25 / (w + 1))2 rather than the
sum of firm demands 100(p/2w)2.

107
28 Managerial economics

Case: A worked out example of the effect of industry

concentration on input price

It is often argued that competition among buyers should be encouraged

if the objective is to get a good deal (a high price) for input producers.
Concentrated industries are said to be mean buyers of inputs. This
argument has been used to abolish state marketing boards for
agricultural products for example, and in the same sector, to promote
competition among food processing plants. The reasoning is intuitively
appealing. When firms act as price takers, they will demand more
inputs since they do not worry about the effect of their demand on
the input price. Also, a competitive industry is likely to produce more
output than a monopoly which again should increase demand for
inputs and hence their price. What I want to show you in this example
is that, although there may be very good arguments to promote
competition in an industry, it is not necessarily always the case that
more competition leads to higher input prices.
For simplicity we look at an industry with one input, labour. The supply
of labour is given by w = 1 + L. The production function is q = L1/2 for
each firm in the industry and market demand for the output is given by
Q = 2 – p, where Q is total industry output and p is output price.

Monopoly
Let us consider first what happens if the industry consists of one firm, a
monopoly – monopsony. Its total expenditure on labour equals
Lw = L (I + L) so that marginal expenditure equals ME = 1 + 2L.
Marginal revenue for the monopolist is MR = 2 – 2q = 2 – 2L1/2 and
MP = (1/2) L–1/2. Applying the rule MRP = ME, (and doing some
numerical work because this equation cannot be solved analytically),
gives an unconditional input demand of L* = 0.18. The corresponding
wage can be found from the labour supply function as w = 1.18.
Given its demand for labour, we know from the production function
how much output the monopolist produces: q = 0.42, and hence final
product price (from the demand function) is p = 1.58.

Perfect competition
Now let us see what happens if a number of (identical) firms in this
industry behave as price takers in input and output markets. Each firm
sets MRP = p MP equal to ME = w or p /(2L1/2) = w. To determine the
industry demand for labour we cannot just add these individual demand
curves as we have seen above. We need to take the output price effect
of increased input demand into account. The optimality condition for
the firm becomes (2 – nq)/(2L1/2) = w, where n is the number of firms
in the industry or, after substituting for q, (2 – nL1/2)/(2L1/2) = w. We can
solve this latter equation for L and find L = (w + n/2)–2 as the amount of
labour used by each firm. The total demand for labour in the industry
is n times this amount: L′ = n (w + n/2) –2 = 4n(2w + n)–2. The equilibrium
wage is such that demand and supply of labour are equal: 4n(2w + n)–2 =
w – 1. This equation cannot be solved analytically. The table below gives
the equilibrium wage (rounded to two decimal places) for some specific
values of n. It is not difficult to show that the limiting value of w, as the
number of firms n increases, equals 1.

108
 Chapter 7: Production, factor demands and costs

number of firms equilibrium wage w

5 1.34
10 1.26
20 1.16
100 1.04
1000 1

At first sight we have a counterintuitive result. The more competitive

the industry (the larger n) the lower the equilibrium wage. Also, when
the number of firms is large, the competitive industry pays lower wages
than a monopsonist! To see what is happening here look at the table
below which gives the values of the key variables for a monopsonist, a
competitive industry with n = 10 firms and a competitive industry with
n = 20 firms. L* is total labour demand in the industry and P is profit per
firm.

w L* q Q p Π
monopsony 1.18 0.18 0.42 0.42 1.58 0.45
n = 10 1.26 0.26 0.16 1.6 0.4 0.03
n = 20 1.16 0.16 0.09 1.79 0.21 0.01

If you compare the monopsony with the competitive industry with

n = 10 firms, you notice that total output Q is larger in the competitive
industry and the demand for labour is higher, hence the higher wage
rate. However, as the number of firms in the industry increases from
10 to 20, more output is produced with significantly less labour and,
as a consequence, the wage rate drops. The explanation for this
phenomenon is the decreasing returns to scale production function. An
industry consisting of many small plants can produce the output more
efficiently than one consisting of a few large plants. I admit that this
makes my example rather artificial since if there are decreasing returns,
the monopsonist could operate several small plants rather than one large
operation. However the analysis above was carried out for zero fixed or
sunk costs. If a sunk cost is incurred when a plant is put into operation,
there is a tradeoff between efficiency of production and setup costs. (It
is a good exercise for you to check how the assumption of positive sunk
cost would affect the analysis.) The purpose of my example was to show
you that you should analyse an industry and its related industries in
detail before you can make general statements.

From production function to cost function

Whereas the production function gives the maximum amount of output
which can be produced as a function of the level of inputs, the cost function
gives the least cost at which a given level of output can be produced. If
there is only one variable input, it is very easy to derive the cost function
from the production function. For a production function q = f (L), the input
requirement for production of q units is L = f –1 (q) and the corresponding
cost function C(q) = w f –1 (q). If there are several variable inputs, the
problem is a bit more involved. By definition of the cost function, we have
to determine, for each output level, the cheapest way to produce it. For
each output level q we have to find the best combination of inputs. But this
is exactly the problem we solved when we determined conditional input

109
28 Managerial economics

demands! The procedure to find the cost function corresponding to some

production function is thus to find the conditional input demands and then
determine the budget necessary to cover these requirements.

Example 7.1 cont’d

In Example 7.1 we showed that the conditional demands for the

production function:
q = f(x1, x2) = ax11/2x21/2 are x1 =(q/a)(p2/p1)1/2 and x2 = (q/a)(p1/p2)1/2.
The cost function is therefore:
C(q) = p1x1 + p2x2 = (q/a)(p2/p1)1/2p1 + (p1/p2)1/2p2) = (2q/a)(p1p2)1/2.
In the example above we had a constant returns to scale production
function and a linear cost function. In fact, this observation holds
more generally whenever the firm behaves as a price taker in its input
markets. Similarly, increasing (decreasing) returns to scale production
functions lead to cost functions which are less than (more than) linearly
increasing. For example, the cost function corresponding to the Cobb-
Douglas production function:
f (x1, x2) = ax1bx2c
is
C(q) = k p1b/(b+c)p2c/(b+c)q1/(b+c)
where k is a constant depending on a, b and c.

Division of output among plants

Suppose a firm operates or is considering operating several plants or
establishments from which it serves the same market. A natural planning
question which arises is: ‘How much output should each plant produce?’
Intuitively, we could say that, if several plants are in operation, the
marginal cost (MC) in each plant should be equal otherwise the firm could
produce the same amount of output at lower cost by transferring output
from a plant with high MC to a plant with lower MC. Also, the MC in each
plant should equal marginal revenue (MR). To show this mathematically,
assume I produce a total output q, q1 in plant 1, q2 in plant 2 (q1 + q2 = q)
and the demand curve is given by p(q). My profit maximisation problem is:
max П = p(q1 + q2)(q1 + q2) – C1(q1) – C2(q2).
q1, q2
The first order conditions for an interior solution ∂π/∂q1 = ∂π / ∂q2 = 0
result in:
MR(q1 + q2) = MC1 (q1) = MC2 (q2)
which confirms our intuition. We can represent the multi-plant problem,
and the optimality condition that MR equals MC at each plant, graphically
as is done in Figure 7.7. Suppose there are two plants with increasing MC
and MC1(0) < MC2(0). Clearly, if the firm’s total output is less than q' where
MC1(q' ) = MC2(0), then only Firm 1 should be used. As output increases
beyond q' the firm should start to use Plant 2, adding units of output to the
plant with minimal MC. In other words, the firm’s aggregate marginal cost
curve is the horizontal sum of the plants’ MC curves. The firm’s optimal
output level q is determined by the intersection of this aggregate MC
curve with the MR curve. We can trace back from this intersection to the
individual plants’ MC curves to find the optimal output level for each plant.

110
 Chapter 7: Production, factor demands and costs

MC1 MC2

∑ MC

MR

q' q*1 q1 q*2 q2 q' q* q

Figure 7.7: Division of output among plants

The optimality condition above is only valid for an interior solution
(i.e. a production plan in which the firm actually produces positive
output in both plants). In general we have to compare the profit obtained
with such a plan with profit obtained by operating one plant only. If, for
example, the fixed costs of Plant 1 are high, it may be better to produce
only in Plant 2 even if Plant 2 is less efficient in the sense that its MC is
higher. Example 7.5 shows how fixed costs determine where production
should take place.

Example 7.5

A firm which faces a demand curve p(q) = 50 – 10 q is considering

whether it should operate from two establishments and, if so, how much
to produce in each. The cost functions associated with the two plants
are: C1(q1) = F1 + 10q1 + 10q12 and C2(q2) = F2 + 5q22 respectively, where F1
is fixed cost in Plant 1, so that MC1(q1) = 10 + 20 q1 and MC2(q2) = 10q2.
Let us first check the interior optimum by setting MR(q1 + q2) = MC1(q1)
= MC2(q2). The marginal revenue corresponding to the demand curve
is MR(q1 + q2) = 50 – 20 (q1 + q2). Setting this equal to MC1(q1) = 10 + 20
q1 and solving for q1 gives q1 = 1 – q2/2(*). Setting MR(q1 + q2) equal to
MC2(q2) = 10q2 and substituting (*) for q1 results in q20 = 3/2 and, using
(*) again, q10 = 1/4. You should check this result by calculating the
marginal revenue and marginal cost functions at q10 and q20. You should
also check that if we had MC2(q2) = 60 + 10q2 instead of the original
MC function at Plant 2, there is no interior solution i.e. under these
conditions it does not pay to produce at Plant 2.
Even when we obtain an interior solution, we have not necessarily
identified a profit maximising production plan. The profit level
corresponding to (q10, q20) is calculated as:
П = (50 – 10(7/4))(7/4) – F1 – 10(1/4) – 10(1/4)2 – F2 – 5(3/2)2 = 42.5 – F1 – F2.
We compare this to the maximum profit attainable if we operate a
single plant. If only Plant 1 is in operation (q2 = 0), its optimal output
level is such that MR = MC1 or 50 – 20 q1 = 10 + 20 q1 such that q1 = 1.
The corresponding profit level is 1 = (40)(1) – F1 – 10 – 10 = 20 – F1.
Similarly, operating just Plant 2 (q1 = 0) leads to 50 – 20q2 = 10q2 or
q2 = 5/3 and corresponding profit level Π2 = (50 – 10 (5/3))(5/3) – F2
– 5(5/3)2 = 41.67 – F2. Depending on whether , 1, or 2 is largest,
we should produce at both plants, at Plant 1 only or at Plant 2 only.
Clearly, which is optimal depends on the fixed costs as is summarised
in the figure below. For example,  is largest if 42.5 – F1 – F2 > 20 – F1
and 42.5 – F1 – F2 > 41.67 – F2 which implies that operating both plants
(the interior solution) is profit maximising if F1 < 0.83 and F2 < 22.5.
You should check that 1 is largest when F2 > max(21.67 + F1, 22.5) and

111
28 Managerial economics

2 is largest when F1 > 0.83 and F2 < 21.67 + F1. The north-east corner
of Figure 7.8 corresponds to the scenario where fixed costs are too
large at both plants so that it is impossible for the firm to make positive
profits.

F2

just plant 1 shut down

41.67

22.5

21.67 just plant 2

both

0.83 20 F1

Figure 7.8: Fixed costs and mulitplant firms

The example above shows how firms can decide whether to operate one or
two manufacturing plants. Using this type of reasoning we can explain why
real-life firms shift production from one plant to another due to exogenous
cost shocks. Hoover, the domestic appliance maker, shut down its plant
in Dijon thereby ending 600 jobs in France in January 1993. It decided
to concentrate vacuum-cleaner production in Scotland, creating 400 new
jobs at its plant near Glasgow. By shifting its production, Hoover slashed its
costs by a quarter, partly due to economies of scale (all production in one
plant) and the rest from lower wages and non-wage costs such as health
insurance which are a lower percentage of overall wages in Britain than in
France. The French have accused Britain of ‘social dumping’.

Estimation of cost functions

Our analysis of how firms decide on optimal output levels and how they
allocate production between plants rests heavily on the firm’s knowledge
of its cost function. How do firms obtain this knowledge? One approach
is to use econometric analysis, using time series and/or cross section data
(for several firms or plants in the same industry). Using regression analysis
the relationship between cost and output level, input prices, plant size,
and so on can be calculated.
As in other applications, we have to decide which functional form to use.
Is the cost function linear, quadratic or cubic in output or does a Cobb-
Douglas function give a reasonable fit to the data? As we have seen before,
the Cobb-Douglas specification, which is very popular, takes the form:
C(q) = k pLα pKß pEγ qδ.
where pL, pK, and pE are labour, capital and energy (or other input) prices
and α, β, γ, δ are the parameters to be estimated.
In practice however, estimating cost functions is problematic. To
start with, there are serious problems with the availability of suitable
data. Expenditures on capital are difficult to measure. Most firms are
multiproduct firms and it is difficult to isolate costrelated data for a single
product. Accounting data reflect various ways of allocating overheads over
112
 Chapter 7: Production, factor demands and costs

products and may hide important factors such as economies of scope (i.e.
the scenario in which it is cheaper to produce several products together
rather than separately).
Using time series data leads to additional problems. To get reliable results,
we need a reasonable sample size of, say, 40 observations. Ideally we
should use monthly observations; it is unreasonable to assume that firms
are using the same production technology (and hence operate according
to the same cost function) over a period of 40 years! In agriculture for
example mechanisation has dramatically changed the labour intensity of
many production activities over timespans of 10 or 20 years.
Remember that in the definition of the cost function it is assumed that
firms are using the most efficient combination of inputs. If the real life
firms in the sample are not operating efficiently, the estimated cost
function will be biased. In fact, we cannot expect firms to be operating
‘efficiently’ at all times; this would require perfect foresight and
immediate adaptation of scale of inputs to changed market environments.
An alternative method which is used to estimate the cost function is the
engineering approach. Engineers are asked to estimate the quantity of
inputs such as plant size and machinery required for given levels of output.
To estimate the cost function, these requirements are simply multiplied by
the price of the inputs. Because of the technical nature of this exercise, the
costs of controlling and managing are often overlooked.

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• derive the cost function from a given production function i.e. derive
conditional input demands (input as a function of output) and write
the expenditure on these inputs as a function of output
• derive unconditional input demands for monopsony, monopoly
and price-taking firms
• solve the problem of which plants(s) to operate and at what level for
a two plant firm given the plants’ cost functions and the demand
function.

Sample exercises
1. Draw an isoquant map representing increasing returns to scale up to
some output level and then decreasing returns to scale.
2. You can cook meals from ‘scratch’ using fresh ingredients or you can
buy frozen dinners. Draw an isoquant assuming the only inputs in meal
production are time and money.
3. For the linear version of the production model where every production
process is characterised by a ray through the origin in the isoquant
plane, show (graphically) the effect of technological progress which
enables the same production levels as before with less labour input.
4. In many LDCs, production is constrained by the availability of capital.
Show graphically (on an isoquant map) how the cost function reflects
this constraint on the amount of capital available. How would you find
the cost function analytically?
5. A firm’s production function is given by q(K, L) = K1/3 L2/3 and it faces
fixed input prices r (for capital) and w (for labour).

113
28 Managerial economics

a. Derive the conditional input demands and the cost function.

b. If the demand function is given by q = (21 – 4p)/8, w = 1 and r =1/2,
what is the firm’s optimum output level? At this output level how
much capital and how much labour is used?
c. Suppose the wage increases from w = 1 to w = 8. If the firm
continues to produce the output level in (b) what is its new demand
for labour i.e. what is the substitution effect of the change in
input price?
d. Given the wage increase in (c), what is the new optimal output level
and its associated demand for labour? What is the output effect
of the change in input price?
6. Find the cost function corresponding to the following production
functions:
a. the inputs are perfect complements: q(x1, x2) = min (x1, x2)
b. the inputs are perfect substitutes: q(x1, x2) = x1 + x2.
7. True/false
a. The industry demand for an input is less elastic than the horizontal
sum of the marginal revenue product curves if the industry is
perfectly competitive
b. same question if the industry consists of local monopolists with
independent markets.

114
 Chapter 8: Topics in labour economics

Chapter 8: Topics in labour economics

Aims
The aim of this chapter is to consider:
• the role of unemployment both as a consequence and a necessary
condition in efficiency wage theory
• the dependence of the efficiency wage on the size of a potential
gain from cheating, the probability of being caught and unemployment
benefit
• why an individual firm’s demand for labour could increase when
it introduces efficiency wages
• why we observe long-term employment
• the role of promotions in the internal labour market.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• discuss the difficulties in attempting to apply standard neoclassical
models of labour supply and demand to internal labour markets
• analyse a simple efficiency wage model
• set up and solve the minimum cost implementation problem
• give explanations for increasing wage profiles
• discuss Frank’s theory of wage compression
• discuss managerial and executive compensation and the role of
performance related pay in this context

Further reading
‘A bit rich’, The Economist, 26 November 1994, p.89.
‘A racket in need of reform’, The Economist, 27 August 1994, pp.21–28.
‘Bosses on the run’, The Economist, 28 January 1995, p.26.
Frank, R.H. ‘Are workers paid their marginal products?’ American Economic
Review 74(4) 1984, pp.549–71.
Lazear, E.P. ‘Agency, earnings profiles, productivity, and hours restrictions’,
American Economic Review 71(4) 1981, pp.606–20.
‘Low risks, high rewards’, The Economist, 11 December 1993, pp.123–25.
Main, B.G., A. Bruce and T. Buck ‘Total hoard remuneration and company
performance’, (Discussion paper, Department of Economics, University of
Edinburgh, 1994).
Malcomson, J.M. ‘Work incentives, hierarchy, and internal labour markets’,
Journal of Political Economy 92(3) 1984, pp.486–507.
‘Nicely does it’, The Economist, 19 March 1994, p.94.
‘Ordinary deaths’, The Economist, 5 November 1994, p.74.
Ransom, M.R. ‘Seniority and monopsony in the academic labour market’,
American Economic Review 83(1) 1993, pp.221–33.
Shapiro, C. and J.E. Stiglitz, ‘Equilibrium unemployment as a worker discipline
device’, American Economic Review 74(3) 1984, pp.433–44.
‘The high price of freeing markets’, The Economist, 19 February 1994, pp.43–44.
‘The sour taste of gravy’, The Economist, 5 November 1994, p.50.
115
28 Managerial economics

Yellen, J. ‘Efficiency wage models of unemployment’, American Economic

Review, Papers and Proceedings 74(2) 1984, pp.200–08.

Introduction
The previous chapters have provided all the ingredients of the traditional
neoclassical model of the labour market. The supply of labour is derived
from individual utility maximising decisions regarding consumption of
leisure and other goods. The demand for labour by a firm is derived from
the marginal revenue product rule which says that firms hire an amount
of labour such that the marginal return (the marginal revenue product
or MRP) equals the marginal expenditure. The total demand for labour
(as for other inputs) is obtained by summing the firms’ demands taking
into account any effects of this aggregation on the marginal revenue of
output (see the section on industry demand for inputs). The equilibrium
employment and wage levels are determined by the intersection of supply
and demand.
Although the traditional neoclassical framework summarised above is
undoubtedly an elegant construction which provides a wealth of testable
predictions, as it stands, it is neither realistic nor a great help to managers.
For a start, the model cannot explain involuntary unemployment. If
there is unemployment (demand less than supply), wages simply adjust
downwards until an equilibrium is reached. This ignores, among other
things, the effect of minimum wage regulations, welfare payments and
union power. Furthermore, downward pressure on wages can only exist
if the pool of unemployed workers is available where they are needed; if
they are in a geographically different location they may face prohibitive
moving costs. Also, the theory predicts that wages should be continually
adjusted to reflect a worker’s productivity whereas in practice wages are
fairly stable and are almost never adjusted downwards. Similarly, firms are
assumed to adjust the size of their workforce continually in response to
changes in demand conditions whereas in practice labour turnover, at least
in Europe, is limited.
The MRP rule is hardly a practical guideline for managers making hiring
decisions and setting wage levels. At the time a worker is hired, there
is uncertainty about his productivity. This problem is likely to be worse
for skilled or professional workers for whom, even after they have been
hired, it is difficult to measure performance or productivity. Managers do
not treat workers like other inputs. In addition to making decisions about
how many workers to hire and how much to pay them, there are the
important problems of whom to hire, whether and how to train them and
how to motivate them. To explain certain features of the internal labour
market such as promotions, life-time employment with the same company,
increasing wages over a career path etc. different models or at least
variations on the conventional framework are needed.

Efficiency wages
Standard labour economics models assume that labour is of uniform
quality and that this quality can be observed by the employer. Indeed,
the employer is assumed to base the worker’s reward on his quality
or productivity. In reality of course employers do not have perfect
information about workers’ quality or (lack of) effort. If the employer
cannot detect ‘cheating’ such as entertaining friends instead of customers
on the business expense account, selling company secrets, taking bribes,
working less than contracted for, etc. then workers will cheat. Even when
116
 Chapter 8: Topics in labour economics

an employer can detect these activities, the possibilities for punishment are
limited. According to recent reports in China’s Legal Daily, a worker in an
American joint venture company was forced to lick clean a glass product
which the supervisor found dirty and a woman who stole two pairs of
shoes from a Taiwanese factory was made to sit in a dog cage.1 These types 1
See ‘Ordinary deaths’
of punishments are not ordinarily available to employers.
In many cases, the only means of disciplining a worker is to fire him or
her. However, if the terminated worker can immediately find a new job
at the same wage (there is no unemployment), even termination is not
an effective punishment and hence threat of termination will not induce
good behaviour. Only when workers fear that they will not be able to find
a job or at least not as good a job, are they deterred from cheating by the
threat of termination. The question then arises whether by paying workers
a wage premium in the form of an efficiency wage, the employer can
induce good quality work. Note that ‘good quality work’ can have many
interpretations ranging from higher productivity and better attitudes
towards customers to reduced absenteeism and staff turnover rate. If
workers are paid a wage higher than the market wage, they incur a real
penalty when they are fired. This question was examined by Shapiro
and Stiglitz (1984) in their seminal paper on efficiency wages.2 Shapiro 2
Yellen (1984) provides
and Stiglitz focused on the consequences of payment of efficiency wages a good overview of
efficiency wage models
on unemployment. If firms find it profitable to pay higher wages., the
argument goes, then the demand for labour decreases which creates
unemployment. Furthermore, unemployment is not only a byproduct of
efficiency wages, it is a necessary ingredient in the efficiency wage model.
If there is full employment, as mentioned above, the threat of termination
is not effective. Unemployment is a worker discipline device.

A simple efficiency wage model

Let us look at a simple version of the efficiency wage model. Suppose the
employee gains g from cheating if undetected. If the market wage is w0 and
there is no unemployment, firing is not a threat because a fired worker
can get another job at the same wage w0. It is important in this model
that there is no stigma attached to being fired. All workers are assumed
identical and equally likely to misbehave given the same circumstances.
This implies that it is not rational for an employer to discriminate
against workers fired from a previous job. Given that termination is not
a punishment, all workers cheat and firms do not spend any resources
stopping them. Workers get a payment w0 + g per period; w0 from the
employer and g as a bribe for example.
Now consider the problem of an employer who is willing to pay a higher
wage w to induce workers to be honest. How much does he have to pay?
The employer will monitor workers and catch cheaters with probability
p. Let’s just look at one period and assume that cheaters are caught (and
fired) at the beginning of the period. If a worker is fired, he receives
unemployment benefit b. We assume here that the higher wages lead to
unemployment. A risk neutral worker cheats if the expected benefit from
doing so exceeds his wage:
(w + g)(1 – p) + bp > w or w < b + g(1 – p)/p ≡ w1. (1)
The efficiency wage is w1; it is the lowest wage the employer can pay if
he wants the workers to be honest. The efficiency wage is increasing in
the gain the worker gets from cheating. If the temptations to cheat are
large so has to be the compensation for refraining from cheating. A rise
in unemployment benefit also drives up the efficiency wage because it
lowers the punishment value of getting fired. If workers are not likely
117
28 Managerial economics

to be caught (p low) the efficiency wage has to be relatively higher than

when the fine can detect cheating easily. This last relationship illustrates
the tradeoff a firm faces if it can determine the probability p, through
appropriate choice of monitoring intensity, as well as the wage w1.
Either the firm pays high wages and does little monitoring or it monitors
seriously and can get honesty for lower wages. This is illustrated in figure
8.1 below. All combinations to the north-east of the curve ensure that
workers are honest. Generally, monitoring activities designed to increase
the detection probability p are costly. This implies that a cost minimising
firm chooses a combination of p and w on the curve.

p(w)

w
b

Figure 8.1: Monitoring-wage tradeoff

Suppose the firm can fix p by allocating resources to monitoring and
incurring a monitoring cost M(p) per worker. If the firm minimises
total wage and monitoring expenditure per worker subject to inducing
honest behaviour, its problem, which is known as the minimum cost
implementation problem is:
min w1(p) + M(p) = b + g(1 – p)/p + M(p).
Assuming an interior solution and assuming the second order condition is
satisfied, we find:
p2 M' (p) = g. (2)
If M' (p) > 0 and not decreasing in p, (2) implies that, if the gain to the
worker is high, it is in the firm’s interest to devote a significant resource to
monitoring.
For the firm, the move from paying w0 and suffering dishonesty to the
efficiency wage scenario is profitable if:
w0 + C > w1 + M(p) (3)
where C is the cost to the firm caused by the worker’s cheating. Hence,
the model does not predict that efficiency wages are paid by all firms
in all circumstances. It is quite possible for an employer to be better
off permitting bad behaviour than punishing it if the necessary rise in
wage and monitoring cost is high. For example, restaurant owners often
allow staff to ‘steal’ meals or snacks. Many business employees ‘cheat’ by
inflating their business travel expenses.
We have determined the wage w1 the firm would have to pay to keep a
worker who joins the firm from cheating. We still have to check that the
firm can actually attract or retain workers at this wage. If the alternative
for the worker is to stay at wage level w0, it has to be the case that:
w1 > w0 + g. (4)
118
 Chapter 8: Topics in labour economics

Clearly, C has to be larger than g; the cost to the firm due to the worker
shirking has to be larger than the gain to the worker for both condition
(3) and condition (4) to be satisfied. If these conditions are satisfied, the
move to an efficiency wage situation constitutes a Pareto improvement:
both the firm and the workers are better off. Efficiency is increased hence
the terminology ‘efficiency wages’! The efficiency wage is higher than the
original market wage w0 and, if the MRP is unaffected, the demand for
labour decreases and unemployment is generated. In this simple model, a
rise in unemployment benefit is likely to increase the efficiency wage and
hence increase unemployment.
Real-life firms and some enlightened management consultants know that
being nice to workers may pay off. In the efficiency wage model, it is
shown that it may be possible to increase the quality of work, reduce staff
turnover or eliminate unproductive behaviour by paying higher wages.
Nordstrom, an American department store chain, pays its sales staff twice
the industry average in the hope of attracting and retaining good workers.
Clearly similar arguments could be made for non-pecuniary rewards for
workers. It may be more profitable (less costly) for a firm to increase
job satisfaction than to increase wages. Employees may appreciate non-
monetary aspects of their jobs such as company cars, training programmes,
pride in their work, more responsibility and autonomy. Chaparral Steel, a
Texan steel producer, allows workers to travel around the world to select
their own equipment. Every employee of Motorola spends at least a week
every year in training. Levi Strauss ‘empowers’ workers by allowing them
to redesign parts of the production process. In half of America’s large
companies, workers are organised in self-managing teams.3 When firms 3
See ‘Nicely does it’
create a worker-friendly environment, the effect on monitoring costs is
equivalent to that of an increase in wage. If firms are nice to workers,
their tradeoff between wage and detection probability (see Figure 8.1)
can be shifted to the left so that, for the same amount of monitoring, firms
can pay a lower efficiency wage or, for the same wage, they can do less
monitoring. Within organisations, the latter possibility creates conflict as
the reliance on middle managers to do the monitoring is diminished.

Case: Politicians, sleaze and efficiency wages

Politicians regularly exploit an efficiency wage argument to award
themselves salary increases. In theory at least, paying a cabinet
minister or any other government official a high wage should make him
think twice about accepting bribes and getting involved in corruption
and other scandals. In 1994 government ministers in Singapore saw
their salaries increase by about 25 per cent. The Singapore government
argues that high wages for officials have ensured that Singapore, as
one of very few Asian countries, does not have a corruption problem.
The starting pay for a cabinet minister in Singapore is $419,285 a
year. The annual salary of the prime minister is $780,000, a very nice
reward indeed compared to the $200,000 Bill Clinton has to get by on
or the measly $122,000 for John Major. We can offer at least a partial
explanation for these differences in pay in terms of our simple model of
tradeoff between monitoring and wage. In most western democracies, 4
See ‘A bit rich’
politicians are monitored by the media. The more intrusive the media,
the lower the efficiency wage! Singapore, for example, would not
tolerate a press as aggressive as the British press.4
In the United States scandals involving politicians are relatively
rare compared to Europe. This is even more surprising when one

119
28 Managerial economics

considers the enormous efforts of the American lobbying industry to

influence politics. The US offers its elected representatives a salary
of about $140,000, nearly three times the amount a British Member
of Parliament is paid. The secretarial and research allowance for an
American politician is more than eight times the British allowance
($577,000 versus $68,000).5 5
See ‘The sour taste of
gravy’
In Russia, public officials and soldiers are demoralised and underpaid.
Corruption, bribery and even sale of government property such as
weapons is the order of the day, fuelling crime. The police are paid
so badly that in 1992 more than 2000 crimes were blamed on police
officers.6
6
See ‘The high price of
freeing markets’
Efficiency wages and minimum wages
It is possible to discuss minimum wage legislation in the simple efficiency
wage framework outlined above. Suppose the initial situation is as
above with all firms paying low wages w0, all workers cheating and no
unemployment. The workers receive a pay-off of w0+ g and the cost per
worker to the firm is w0 + C. Assume this is an equilibrium (i.e. it is not
in any firm’s interest to deviate by paying a higher (efficiency) wage).
If a firm considers paying an efficiency wage w1, then to induce honest
behaviour w1 has to satisfy:
(w1 + g) (1 – p) + (w0 + g)p < w1 or w1 > w0 + g/p
where the first term in the first inequality represents the pay-off
corresponding to cheating and not being caught and the second term
represents the pay-off corresponding to cheating, being caught and getting
a job elsewhere at the low wage w0. Suppose the solution to the firm’s
minimum cost implementation problem is the wage w1 and detection
probability p(w1). Hence the firm would incur a total cost of w1 + M(p(w1))
per worker. It is not profitable for the fine to implement this solution when
this cost exceeds its original cost (i.e. when w1 + M(p(w1)) > w0 + C). Only
an employer who incurs a high cost of worker misbehaviour would find it
profitable to pay the high wage w1.
However, if all firms paid a higher wage and this leads to unemployment
then the worker’s fall back position is lowered to the unemployment
benefit. The resulting efficiency wage (see (1)) w1 is now lower:
w1 > g(1 – p)/p + b
Suppose the solution of the minimum cost implementation problem is w'
and p(w' ) and the cost per worker is lower than the original cost:
w' + M(p(w' )) < w0 + C. This means that, if all firms could coordinate
their employment policies and offer the efficiency wage w', they would
all be better off. They are in fact stuck in a low pay-low worker quality
equilibrium from which unilateral deviations are not profitable. The
government can facilitate firms’ coordination by setting a minimum wage
level at w'. If it becomes illegal to offer wages below w', all firms will pay
w' and no firm has difficulties attracting workers. Do any welfare gains
result from the introduction of the minimum wage legislation? It turns
out that, as long as the monitoring costs are not too high, there may
be welfare improvement. Total welfare, is initially g – C and, after the
minimum wage, ignoring unemployment, it is –M(p(w' )) which represents
an increase if M(p(w' )) – C < – g. The minimum wage legislation could
improve welfare if the monitoring costs are low relative to the cost of
shirking and if the worker’s gain from shirking is not too large. Of course,
unemployment increases and so the net welfare effect of introducing a
minimum wage is ambiguous.
120
 Chapter 8: Topics in labour economics

Are the workers better off when the government introduces minimum
efficiency wages? The answer is ambiguous. The workers who keep their
jobs are now receiving w' = g(1 – p)/p + b compared to w0 + g. Given b < w0,
workers are likely to be worse off unless p is low which it would be if firms
face high costs of monitoring. Of course the workers who lose their jobs are
worse off. The move to efficiency wages cannot be a Pareto improvement.

Firm demand for labour

Let us return now to efficiency wage theory and see if we can model the
effect of the introduction of an efficiency wage on the firm’s demand for
labour. We already know that, for the efficiency wage theory to work,
there should be some unemployment. However, this is not to say that
each firm should reduce its labour demand. Suppose workers choose a
low or high effort level eL or eH which gives them a disability of cL and
cH respectively and delivers an expected output of qL or qH respectively.
It is important that an individual worker’s output cannot costlessly be
observed by the firm. However, if the firm monitors the worker, it can gain
information about whether the worker is working hard (eH) or not (eL). In
particular, if the worker is shirking (eL), he has a chance p of being caught.
The firm decides on p and incurs a monitoring cost M(p) per worker. All L
workers are identical and, if they all work hard, the expected output is LqH
whereas, if they shirk, total expected output is LqL. Both the workers and
the firm are risk neutral.
If it is impossible for the firm to monitor workers, workers have no
incentive to work hard since their shirking is not detected. The firm thus
determines its demand for labour from:
max LqL P(LqL) – w L
L
where P(Q) is the firm’s demand function (i.e. the price the firm can
charge if it sells Q units of output) and w > cL is the wage rate. The result
of this optimisation problem is the familiar MRP = ME condition:
(5)
(P(Lq ) + ∂Q∂P Lq ) q = w.
L L L

Now suppose it becomes possible for firms to monitor. Consider the

problem of a firm wishing to pay an efficiency wage. Assume the market
wage is 100 i.e. a worker has a choice between working for the market
wage w′ and shirking or working for the efficiency wage w' at a high effort
level. The firm, when it detects shirking, fires the worker and pays him
zero. A shirking worker who is not caught gets the same wage w' as a non-
shirker. Workers are therefore attracted to the firm (anticipating that they
will have to work hard there) if:
w' = cH > w – cL. (6)
Once they join the firm, they work hard if the pay-off of working hard
exceeds the expected pay-off of shirking:
w' – cH > (1 – p)(w' – cL). (7)
It is assumed in (7) that firms detect worker shirking before workers have
suffered the disutility and that a worker who is caught shirking cannot
start working elsewhere at wage w immediately. The two conditions
(6) and (7) which are very similar to the participation and incentive
compatibility constraints in principal – agent problems, are represented
graphically in Figure 8.2. Note that (7) can be rewritten as:
w' > (cH – (1 – p) cL) /p = (cH – cL)/p + cL.
121
28 Managerial economics

w' A

join &
don’t cheat
join &
cheat
don’t join
B
w + cH – c L

c
(cH – cL)/p + cL
don’t join
H
0 p′ 1 p

Figure 8.2: Conditions on the efficiency wage

Now the firm maximises total profit subject to conditions (6) and (7):
max LqH P(LqH) – w' L – M( p) L (8)
w', p, L
s.t. w' > w + cH – cL and w' > (cH – cL) / p + cL.
The feasible set of combinations of w' and p is the ‘join & don’t cheat’ area in
the figure above.

Define p' as the minimum detection probability necessary to induce honest

behaviour if the firm pays the minimum wage w + cH – cL which attracts
workers.

We can find p' at the intersection of the curves corresponding to conditions

(6) and (7) as p' = (cH – cL )/(w + cH – 2cL ). Clearly it can never be optimal for
the firm to set p higher than p' because it can guarantee that the worker does
not shirk at any wage above w + cH – cL with a lower detection probability p.
The solution to the firm’s problem therefore has to be a point on the curve AB.
In addition, at the optimal solution of (8), the MRP = ME rule has to hold:
MRPH= MR(qHL) qH = w' + M(p) (9)
where MR(Q) is the marginal revenue function. This is of course very similar
to the optimality condition (5) before the introduction of an efficiency wage
which can be rewritten as:
MRPL = MR(qLL) qL = w. (10)

MR MRP

MRPH
w′ + M(p)

MRPL
xq
x H
qLL0 qHL0 L0 L
Q
(a) Marginal revenue (b) Marginal revenue product

Figure 8.3: Effect of efficiency wage on firm demand for labour

122
 Chapter 8: Topics in labour economics

To keep things simple, let us assume the MR function is linear and set
qL = 1 and qH > 1. Figure 8.3a represents the MR function. Figure 8.3b gives
the corresponding MRP functions before (MRPL) and after (MRPH) the
introduction of the efficiency wage. These curves are derived from the MR
curve as follows. Consider MRPL first. From (10) we know that:
MRPL (L) = MR(L) for qL = 1.
Hence for L = 0, MR(L) is given by the intercept in Figure 8.3a. For
L = L0, we can read MR(L0 ) on Figure 8.3a. These two points are sufficient
to draw MRPL in Figure 8.3b. Now consider MRPH. For L = 0, MRPH = MR(0)
qH which gives us the intercept for MRPH in Figure 8.3b. For L = L0, we see
that MR(qHL0) = x in Figure 8.3a which gives MRPH(L0) = qH x in Figure
8.3b.
Conditions (9) and (10) tell us that the firm demand for labour is
determined by the intersection of MRPL and w and the intersection of
MRPH and w' + M(p) respectively. If the initial wage w is relatively high as
on the figure and if the efficiency wage and monitoring costs are not too
high, the firm’s demand for labour could increase when it starts paying
efficiency wages!

Internal labour markets

Real-life companies often have employment and compensation policies
which seem to bear no resemblance to the neoclassical result that labour
should be hired and paid according to the MRP rule. Workers are not laid
off whenever there is a temporary shift in marginal revenue; wages are
often determined according to a fixed scale and the generally low variance
in wages does not seem consistent with workers being paid the value of
their marginal product in each period. Many employees are in an internal
labour market and move up the hierarchy inside an organisation. Mobility
in and out of the internal labour market is limited so that conditions in
the external market have only a limited effect on the internal market.
This is especially true for white collar jobs, professionals and managers or
primary sector workers. The internal labour market is insulated to some
extent from the external labour market except at a few entry points such
as the hiring of graduates where firms compete with other employers.
Workers may have outside options during their careers and firms may hire
from outside for higher level jobs. Only in these situations is a firm forced
to pay market wages. In this section we look at some of these features of
employment policies in organisations and discuss alternative explanations
economists have put forward.

Why do wages rise over a career path?

In most jobs employees are paid a relatively low wage when they start
work and the wage gradually increases with years in the job or seniority.
We could explain this phenomenon by referring to the higher value of
more senior workers because of their acquired (firm-specific) human
capital which means the knowledge and skills which determine
productivity. More senior workers have had more formal training as well
as on-the-job experience which is likely to make them more productive
relative to job entrants. Certainly, in some situations this explanation is
valid but overall it does not seem reasonable to maintain that the most
senior workers are the most productive. Wages generally rise faster than
productivity. It is interesting to note that university professors (in the
USA) are an exception to the rule of salaries increasing with seniority. In

123
28 Managerial economics

a recent study, Ransom (1993) found that for his samples of university
professors, after controlling for experience and some other variables, there
was actually a negative correlation between seniority and salary.
In some circumstances, human capital theory offers an explanation for
increases in wages. Firms investing heavily in training which increases
the workers’ value not only in their current job but elsewhere as well,
tend to pay new trainee employees a low wage. This wage is increased
after the training is completed. Examples of this are found in accountancy,
law and architecture. The reason for the jump in wage is that, while the
employee is developing marketable skills at the firm’s expense, he pays for
this training by working at a low wage. As soon as the training phase ends
and outside offers would be forthcoming, the firm is in a position to offer
a higher wage and retain the trained workers. What is harder to explain
is that older workers are paid more than younger and sometimes more
productive workers with the same training and qualifications.
Lazear (1981) uses an efficiency argument to explain such rising wage
profiles. Workers are paid less than their MRP early in their careers and
more than their MRP later in their careers so that over their lifetime
they are getting the value of their marginal product. Given this wage
profile it is obviously painful for a worker to lose his job after he has
worked with a firm for a few years: he loses his ‘investment’ in the form of
larger wages later on. Therefore workers with rising wages are more likely
to work hard and avoid being fired whereas workers whose wages do not
increase do not have the same incentive to put in high effort. By offering
increasing wages the employer can succeed in making both the worker
and himself better off because higher efforts by the worker lead to higher
productivity. The increased cost of shirking leads to a Pareto improvement.
Because workers are paid very high wages when they are older, mandatory
retirement is necessary otherwise workers would prefer to continue
working after they had been paid their lifetime value to the firm. Also,
there has to be some mechanism to prevent the firm from firing workers
(without cause) before it has paid them the higher wages to which they
are entitled. If firms care about their reputations, they will not renege on
the (implicit) agreement to pay the workers their lifetime productivity
value by firing them early because then they will not be able to attract
workers in the future.
The efficiency explanation above rests on the idea of using rising wages as
an incentive mechanism. In jobs where the employer can easily monitor
whether the worker is performing adequately or where alternative
incentive mechanisms such as piece-rates or commissions can be used,
we would not expect to see wages rise much with seniority. Conversely, in
large organisations where monitoring costs are high, rising wage profiles
are likely.
Efficiency wage theory can also explain why wages rise over a career
path up to a point where they exceed marginal revenue product. Initially
employees are usually offered low responsibility jobs which are easier
to monitor. In terms of the minimum cost implementation problem, this
implies that the firm can set a high detection probability p at low cost M(p)
which in turn leads to a relatively low wage. Later on, when the employee
has more responsibility, the monitoring cost increases so less monitoring
will be done and a higher wage results.
Firms also offer seniority-based pay for jobs where it is important to retain
people. Paying a low salary initially and compensating later screens for
loyalty in the sense that only workers who intend to stay with the firm

124
 Chapter 8: Topics in labour economics

over a longer period find this an attractive proposition. Firms also use
‘golden handcuffs’ such as deferred compensation and unvested pensions
as screening devices. For example, Bell South, an American regional
telephone company, encourages employees to put up to 25 per cent of
their pay in a special account. The company augments this contribution
and, when they retire, the employees get more than twice the normal
market interest whereas, if they quit, they just get the market interest.
When an employer hires a new worker neither she nor the worker knows
for sure how productive the worker will be. Workers are typically risk
averse and would prefer a contract which offers them a secure job at a
wage which corresponds to average productivity to a contract offering
them a wage corresponding to their actual productivity. However, if a firm
pays a wage equal to the value of the average productivity it will make
a loss: the low productivity workers will stay and the highly productive
workers will leave for better outside offers. To guard against this, the firm
has to start everyone at a low wage and increase wages of workers who
turn out to be productive. The difference between the low initial wage and
MRP for the productive workers is like an insurance premium they pay at
the time when they do not know their productivity.
Finally, in cross section data, wages can show positive correlation with
seniority even if individual workers do not experience rising wage
profiles. There are several explanations for this. Firstly, workers who are
productive in their jobs are highly paid and are likely to stay whereas
workers who are not productive and get low pay tend to leave. As a
consequence, we find that the highly paid workers have higher seniority.
Efficiency wage theory offers an alternative explanation: firms which pay
efficiency wages have low turnover which means that workers in these
higher paying firms have higher seniority again generating the positive
correlation.

Why is there long-term employment?

In contrast to the assumption of perfect mobility of labour underlying the
neoclassical model, firms do not lay off workers when there is a temporary
reduction in demand and they do not start hiring immediately when
demand for their products increases. Workers similarly do not usually
change jobs when slightly higher wage offers are made. Apart from the
obvious explanations of moving costs for workers and legal impediments
and costs of firing workers, are there other reasons for the firm-worker
relationship to be permanent?
One reason for long-term employment is the existence of firm-specific
human capital. Employees may acquire knowledge and skills which
increase their productivity as long as they stay in the same job but which
are not useful for other jobs. From an efficiency point of view then, long-
term commitment is necessary for the firm and/or the worker to have an
incentive to invest in firm-specific human capital.
Sometimes it is necessary to offer a long-term contract to ensure that
the employee’s incentives are aligned with the firm’s long-term
objectives. If the employee’s horizon is one year away, he may not care
about the firm’s profitability several years into the future. It may also be
difficult to judge an employee’s performance in the short-term if the effect
of his work is not felt until much later. This is likely to be the case for
managerial jobs.
Some versions of efficiency wage theory can also explain long-term
employment. If workers are not offered job security they are more likely to
shirk because the punishment of losing their job is not as severe.
125
28 Managerial economics

What is the role of promotions?

Generally promotions help to assign people to jobs where they are most
productive. However, promotions are also used as rewards to provide
incentives for good performance.
A problem arises when these two goals are in conflict. Good performers at
one level of the hierarchy may not be very productive when promoted to
a higher level. This is especially likely when the nature of the job changes
dramatically after promotion. To avoid the problem of good technical
employees (engineers and researchers) becoming mediocre supervisors
and managers, companies such as IBM and 3M use separate career ladders
for scientists and managers.
Promotion is a reward not only because it brings with it a higher salary
but it also gives the worker a chance to compete in the next round to
get promoted further. Since the 1980s, firms have eliminated levels of
management and, as a consequence, promotion is less of an incentive:
at any level there are more candidates and hence a lower probability of
getting promoted. Using promotion as a reward is not entirely harmless.
When workers are involved in team work, rewards based on relative
performance may lead to reduced cooperation. Another problem which
can arise when relative performance assessment is used is that of worker
collusion.
Malcomson (1984) proposes an interesting moral hazard explanation
for promotions. The starting point is that workers are paid more after
promotion. In many situations, especially when a substantial amount of
team work is involved, the employer cannot assess absolute individual
performance accurately. Relative performance is easier to judge. If the
employer would promise to pay according to individual performance,
workers may doubt his honesty. There is a moral hazard problem on the
part of the employer in that he could claim that no workers performed
up to the standard. The workers anticipate that the employer will
misbehave in this way and are not encouraged to work hard. If instead
the employer promises to reward relative performance by promoting the
top 10 performers, workers can easily check that the promise was kept. Of
course it is then in the employer’s interest to select the best performers to
encourage all workers to work hard.
Usually workers at higher levels of a firm’s hierarchy have been promoted
from within the firm. There are only a few ports of entry where outsiders
may be recruited. This is consistent with firms ignorant of individual
abilities paying low wages for young workers of mixed productivity and
screening them to learn which are best. This points to a reason for the firm
to delay promotion whenever it can. Promotion makes an employee more
attractive to outside firms because it acts as a signal of worker quality.
Higher wages have to be offered after promotion to prevent the worker
taking up attractive outside offers.

Wage compression
Another feature of internal labour markets is that within firms there seems
to be less variability in wages than in productivity. An explanation for this
fact could be that it is just too difficult to measure individual productivity
especially where team work is involved. A more interesting explanation is
given by Frank (1984) in terms of status. The starting point is the idea that
people care about their position in the income hierarchy of the firm and
they care more the more they interact with their co-workers. Employees
are willing to sacrifice some income if they can be the highest paid in the

126
 Chapter 8: Topics in labour economics

firm and conversely they need to be compensated for being the being the
one with the lowest pay. This is illustrated in Figure 8.4. The three lines
crossing the 45 degree line represent wage schedules for three firms. An
individual with marginal productivity x could choose to work in Firm
A where he is near the top of the salary scale and enjoy relatively high
status. He would then pay a price ab in terms of the difference between his
wage and his marginal productivity. If he does not care at all about relative
status the individual with marginal productivity x could earn a wage
premium bc by working for Firm C near the bottom of its salary scale.

wage 45 degree line

C
c

B
b

A
a

marginal product

Figure 8.4: Status and wage compression

Managerial and executive pay

It may be difficult to measure performance or productivity at the lower
echelons of an organisation; it is very hard to gauge performance of
managers and executives. This is no doubt why many firms have fixed
salary scales attaching salaries to jobs rather than people in these jobs.
It is unrealistic to assume that everyone in a particular job would have
the same productivity but, since productivity is so hard to measure, this
may be the best that can be done. Fixed salary scales show increased
compensation for jobs which carry higher responsibility and higher
diversity of tasks. When measures such as number of people reporting to a
manager are used to determine compensation, perverse incentives may be
given to encourage bureaucratic expansion.
When a manager can reasonably be expected to have a significant
influence on the profitability of the firm, performance-related pay in the
form of bonuses of a percentage share in profits can be used. However, this
may interfere with managers being risk neutral with respect to investment
decisions in the firm. When pay is related to performance, managers may
take decisions which maximise their own risk averse utility functions.
This could even be a problem when pay is not performance-related but
managers consider that their future employment possibilities depend 7
See ‘Low risks, high
on their current company’s results. In Japan, where job mobility is very rewards’
low, these considerations are less important and this, in addition to the
concept of management by consensus where no single manager carries
responsibility for failure of projects, should make Japanese managers less
risk averse. 8
See ‘Bosses on the run’
In 1990, total CEO (Chief Executive Officer) compensation for the 200
largest American companies averaged almost US$3 million per year.7 The
figures for Europe and Japan are much less although there has recently

127
28 Managerial economics

been public outrage in Britain over the ‘excessive’ salary increases top
executives of newly privatised industries award themselves. In January
1995, Cedric Brown, the CEO of British Gas was asked to justify to
Parliament his 75 per cent salary increase to £475,000.8 One of the
reasons for these huge remuneration packages seems to be the use of
incentive schemes for high-level executives. The fixed salary part of the
compensation package may be relatively modest but CEOs typically also
receive stock options which are rights to buy company shares at or above
the current share price. Clearly, there is no downside risk involved here.
If the share price is low during the specified time in which the options
can be exercised, the manager does not buy any shares. Nevertheless
share options can be a useful device to lengthen the CEO’s horizon and
give him an incentive to maximise the long-term value of the firm. The
remuneration package may also contain straightforward stock awards
which consist of shares given or sold at a discount. Sometimes there
are restrictions on the sale of these shares, for example, they may only
be sold after retirement. This is another example of the use of golden
handcuffs to align an employee’s interest to the long-term future of the
company.
Another reason for what seems like excessive compensation at the top of
large firms is that, once an employee reaches this highest level, promotion
can no longer be used as a reward. There is also a ‘political’ explanation
for high executive compensation. Nonexecutive directors who determine
executive pay clearly have an interest in exaggerating these pay awards
as they are often executives elsewhere and can only benefit if the ‘going
rate’ for top jobs rises. A similar story can be told to explain the inactivity
of institutional shareholders. Top executives of banks and pension funds
hardly have an interest in calling attention to high pay.9 The absolute 9
See ‘A racket in need of
size of CEO pay should not be a major concern of shareholders as long as reform’
there is a link with performance. The structure of pay and the provision of
incentives is what matters. It seems that too often executives are paid large
amounts of money when their companies are not doing so well. It is all the
more surprising therefore that large British institutional investors call for
ceilings on the amount of share options that can be issued to executives.
The Association of British Insurers (ABI) and the National Association of
Pension Funds (NAPF) recommend a limit of four times the fixed salary
component for stock options.10 10
Main et al. (1994)

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• discuss the difficulties in attempting to apply standard neoclassical
models of labour supply and demand to internal labour markets
• analyse a simple efficiency wage model
• set up and solve the minimum cost implementation problem
• give explanations for increasing wage profiles
• discuss Frank’s theory of wage compression
• discuss managerial and executive compensation and the role of
performance related pay in this context.

128
 Chapter 8: Topics in labour economics

Sample exercises
Review the minimum cost implementation problem for efficiency wages
and use this framework to argue whether efficiency wages will be higher
or lower than in the standard model if:
a. workers dislike monitoring
b. workers incur a large psychological cost when they are unemployed
c. workers are risk averse
d. monitoring is very costly
e. workers have high switching costs (moving expenses, loss of firm
specific human capital) when they lose their jobs.

129
28 Managerial economics

Notes

130
 Chapter 9: Market structure

Chapter 9: Market structure

Aims
The aim of this chapter is to consider:
• the importance of entry barriers as determinants of market structure
and firm behaviour
• the concepts concentration curve, CRm, HHI and Lerner index
• why in a perfectly competitive industry the supply curve is not
necessarily the horizontal sum of individual firm supply curves
• why we may not observe MR = MC in a monopoly
• the Dorfman-Steiner model.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• describe the important factors in the determination of market
structure giving (preferably your own) examples
• calculate the market structure measures
• derive perfectly competitive firm and industry supply
• find profit maximising price and output for a monopoly
• explain short run and long run equilibrium in a monopolistically
competitive industry.

Essential reading
Varian, H.R. Intermediate Microeconomics Chapters 22, 23 and 24.

References cited
‘Ahead for now’, The computer industry survey, The Economist, 17 September
1994, pp.12–16.
Dorfman, R. and P.O. Steiner, ‘Optimal advertising and optimal quality’,
American Economic Review 44 1954, pp.826–36.
Gourvish, T.R. and R.G. Wilson, The British Brewing Industry, 1830-1980
(Cambridge: Cambridge University Press, 1994) [ISBN 0521452325]
‘Insurers get that sinking feeling’, The Economist, 20 August 1994, pp.65–66.
Martin, S. Advanced industrial economics, (Oxford: Blackwell, 1993)
[ISBN 063117852X]
‘National lottery’, The Economist, 28 May 1994, pp.28–33.
‘Squeezing into Hong Kong’, The Economist, 4 December 1993, p.88.
Sutton, J. Sunk costs and market structure. (Cambridge, Mass: The MIT Press,
1992) [ISBN 0262193051].
‘The Post Office’, The Economist, 8 January 1994, p.29.
‘Unsure about insurance’, The Economist, 31 October 1992, p.101.

131
28 Managerial economics

Introduction
Industries differ in the way they are structured. An important defining
characteristic of market structure is the number of firms serving the
market. However, it is certainly not the case that we can predict how firms
will behave just on the basis of how many firms there are in the industry.
The presence of barriers to entry also plays a crucial role in maintaining
market structure. It is generally believed that, when entry is easy, firms are
disciplined in their pricing policies even when the industry is concentrated.
Empirically, high entry barriers coincide with high profit margins.
Economists classify industries according to the degree of market power
firms have (i.e. do they behave as price takers or price makers?), whether
firms behave strategically and whether the product is homogenous
or differentiated. For example, in a competitive industry all firms are
assumed to be price takers; in an oligopoly there is strategic interaction
between firms.

Determinants of market structure

The question of why real-life industries are structured the way they are is
very difficult to answer. There are a multitude of factors which contribute
to the likelihood of one or the other market structure evolving. The
following list is not exhaustive.

Economies of scale
The world market for disposable plastic syringes is dominated by three
firms: Becton Dickinson, Sherwood/Brunswick and the Japanese firm
Turumo, with worldwide market shares in 1992 of 31 per cent, 16 per
cent and 18 per cent respectively. There are significant economies of
scale in this industry: the production technology is such that, as output
increases, average costs fall dramatically. The source of economies of
scale may be large setup or fixed costs. In extreme cases the presence of
significant economies of scale leads to a ‘natural monopoly’ which refers
to the situation where average costs decrease beyond the output level
corresponding to market demand. In such cases it is obviously cheaper
for one firm to satisfy market demand. The natural monopoly argument
has been used as an argument against privatisation of the Royal Mail in
Britain. Local newspapers are also characterised by significant economies
of scale so that often there will be just one local newspaper in a town.
The size of the market relative to the degree of economies of scale
seems a relevant predictor of market structure. A useful concept in this
context is the minimal efficient scale (MES) of an establishment in
the industry. It is usually defined as the output level at which long-run
average cost (LRAC) is minimised. An alternative definition of MES is
‘the output level beyond which further increases in output would lead to
a reduction in LRAC of no more than 10 per cent’. In practical terms this
means it is the output level where the LRAC curve begins to flatten out.
If we know the MES for a given production technology and we know the
total consumption in the market, we can calculate how many firms can be
accommodated in the industry. For example, if the MES for commercial
aircraft manufacturing corresponds to 10 per cent of the US market, then
there is room for 10 fans to serve this market. If the MES is large relative
to the size of the market, the industry is likely to be concentrated.
The importance of economies of scale as a determinant of market structure
is made evident by the changes in structure after technological innovation.

132
 Chapter 9: Market structure

The beer industry, for example, used to be fragmented and consisted of

thousands of local breweries. Beer was unpasteurised and had to be kept
cold so it could not be transported far. Since the introduction of bottling,
however (a process which has significant economies of scale), the number of
beer producers has decreased dramatically, mainly through takeovers, and
now a handful of large companies dominate the industry.1 1
Gourvish and Wilson
(1994)

Strategy of incumbents
Market structure and the presence of entry barriers undoubtedly affect the
behaviour of established or incumbent firms in an industry. Conversely,
incumbents may engage in activities designed to maintain the market
structure or alter it in their favour. An example of the former is limit
pricing where a monopolist, rather than maximising short-run profits,
prices such that it is not possible for an entrant to make a profit at the
current price. An example of the latter is predatory pricing, where a firm
sets a low price (often below cost) in order to drive a rival out of business.
Firms have instruments other than prices to their disposal when trying to
drive competitors out of business. For example, after the deregulation of
buses in some British cities, some bus companies tried to organise their
timetables so that they would pick up passengers a few minutes before their
rivals. Note that limit pricing and predatory pricing are only possible if the
firm has a cost advantage. In some cases a monopolist does not have to set
a low price to deter entry. The threat to retaliate and flood the market when
a competitor enters may be sufficient. Such a threat can be made credible if
the monopolist has excess capacity so that increasing output (after entry) is
not very costly.
Incumbent firms manipulate dealer relationships to shut out
competition. An extreme case is that of exclusive dealership where a
manufacturer or wholesaler agrees to supply retailers only if they do not
sell any substitute products. Supermarkets typically carry a fairly limited
number of brands of consumer goods. Manufacturers negotiate deals which
involve the supermarkets allocating them a certain amount of shelf space.
Some PC makers sell their machines with a copy of MS-DOS or Windows
already installed. Some airlines have developed computer reservation
systems for use by travel agents. Of course flights by the airline which
developed the system will be more prominently displayed and easier to
book than others and, once travel agents have invested in learning how to
use such a system, they are ‘locked in’.
Incumbent firms often try to erect entry barriers by heavily advertising and
promoting their products. This is effective in industries where customer
loyalty is important or where buyers prefer established brands so that
it becomes very difficult for an entrant to convince consumers to try his
product. Large advertising budgets also increase an entrant’s capital
requirements.2 Companies can increase customer loyalty by offering loyalty 2
See next section
rebates. Airlines’ frequent flyer programmes are a good example of this.

Capital requirements
Large capital requirements can be a source of a cost disadvantage for a
potential entrant. Entrants are usually perceived as riskier prospects by
banks and other money-lenders. This is not without reason as in many
industries entrants are much more likely than established firms to go
out of business. They therefore pay a higher risk premium. Large sunk
costs (investments such as R&D which cannot be recuperated upon exit
from the market) make the entry decision particularly risky. They are

133
28 Managerial economics

sources of barriers to entry in for example car manufacturing, defence

industries, oil refining, deep sea drilling, chemicals, electronics, aerospace,
pharmaceuticals, etc. Empirically it is found that, the larger an industry
(in sales), the larger the number of firms, but this is not true for markets
with high advertising and R&D costs.3 Large fixed costs even when they 3
Sutton (1992)
are not entirely sunk make entry more difficult. Capital requirements are
obviously larger if the industry is vertically integrated and it is impossible
to enter unless you enter all layers of the industry. The computer industry
used to be vertically integrated with computer manufacturers making
most of the parts as well as the software in-house. They used to do most
of their own marketing, distribution, sales and service as well. There were
few independent suppliers of parts. Currently, PCs are assembled from
readily available parts used to supply the consumer electronics industry.
Most PC makers were never vertically integrated. Barriers to entry in the
computer industry are much lower than they used to be and there is more
competition in every layer of the industry.

Patents and other legal controls

Governments clearly influence the structure of some industries by setting
up legal barriers to entry. Taxi drivers in many cities have to obtain a
license before they are allowed to establish themselves. In New York
the powerful taxi industry lobby has been able to keep the number of
licenses (‘medallions’) unchanged at about 12,000 for almost 50 years.
Governments decide which airlines are allowed to fly a certain route. By
taking protectionist measures, governments can make or keep domestic
industries concentrated. South Africa uses very high import tariffs and
most South African industries contain a few major players. Patent law,
which grants monopoly rights for a specified period of time, is used to
encourage innovation.

Control of resource
An obvious instance of market structure determined by entry barriers is
when incumbents control a special non-replicable resource such as mineral
deposits. Some well-known examples are De Beers in diamonds, OPEC
and French Champagne. Take-off and landing slots at airports are also
very important resources without which entry into the airline industry
is impossible. In the same industry, the practice at most airports of long-
term leasing of gates represents an entry barrier. There are however limits
to the market power which derives from owning a special resource. The
Organisation of Petroleum Exporting Countries (OPEC) found that, when
they raised the price of oil too much, other countries started searching
for oil. The oil crisis of the early 1970s was largely responsible for the
exploration of oil fields in the North Sea.

Cost advantage and first mover advantage

In many industries the firms which entered first have a significant
advantage in terms of entrepreneurial skills and/or technological know-
how. Late entrants tend to have a cost disadvantage especially when
learning curve effects are important (i.e. average production cost
decreases with cumulative output). Sunk costs are also a source of cost
advantage as they are only incurred on entry.

134
 Chapter 9: Market structure

Measures of market structure

Given the complexity of real industries and the many factors which
determine how firms behave in these industries we cannot expect that a
simple measure of market structure paints the entire picture. Nevertheless,
a summary measure of how many firms there are and their market share
can convey useful information. The concentration curve represents
a complete statistical summary of the number of firms and their market
shares. On the X-axis the firms are listed starting with the largest in terms
of market share. Market share can mean percentage of sales or assets or
employment in the industry. On the Y-axis cumulative market share is
measured so that by definition the concentration curve is concave.
Concentration ratios (CRm) give the total market share accounted for
by the largest m firms in the industry. UK official sources use CR3 and
CR5 whereas, in the US, CR4, CR8 and CR12 are used. If there is only one
firm in the industry the concentration ratio is 1. When there are n firms
of equal size the concentration ratio is m/n which is close to 0 if n is large.
Empirically it is found that concentration ratios are similar across industries
from one country to another (e.g. consumer durables industries – electrical
appliances, washing machines, refrigerators – tend to be concentrated
CR5 = 95%; for salt, CR4 = 99.5% in UK, 93% in Germany, 98% in France).
Most manufacturing industries have CR4 < 50%.4 Roughly speaking, 4
Sutton (1992)
an industry is classified as a monopoly if CR1 > 90% and as perfectly
competitive when CR4 < 40%. Monopolistic competition and oligopoly
correspond to CR4 > 40%.
The Herfindahl-Hirschman-Index (HHI), like the concentration curve,
uses information about all the firms in the industry. It is defined as the sum
of the squares of all market shares:
n
Σ
HHI = i = 1si 2
If there is only one firm in the industry HHI = 1 and if there are many firms
with equal market share, HHI is close to 0. The HHI has the interesting
property that its inverse corresponds to the number of equally sized firms
which would give this value of HHI. For example, if the HHI is 0.25, the
industry is as concentrated (according to HHI) as an industry consisting of
four firms each with market share 25% since 4(0.25)2 = 0.25. The advantage
of the HHI is that it uses information on all firms in the industry. This is
important when analysing the evolution of industries through merger
activity for example. Whereas the CR does not change after a merger unless
the very largest firms in the industry are involved, HHI always increases by
2sisj where si and sj are the market shares of the merged firms. (Make sure
you know why!) The HHI is used in the US Department of Justice Merger
Guidelines to indicate when a merger is likely to be opposed. For example,
a guideline might be formulated as ‘a merger which increases HHI by x
when HHI is currently over y will be opposed.’
As I mentioned before, these measures only give a very crude idea of what
an industry is like and they have to be interpreted with care. For a start,
CRs are usually measured at national level. This leads to two important
distortions. Firstly, it ignores imports and this may lead to overestimation of
concentration in the domestic industry. Secondly, the relevant market may
be regional due to high transportation costs or other factors.
Although there may be many cement and sugar producers nationally, within
a region in which these goods are normally distributed there may be only
two or three firms. Similarly, there are many local newspapers in Britain
but in each town only one or two are usually distributed. National CRs

135
28 Managerial economics

are therefore meaningless in this market. The definition of the product

or the market is also very important. In antitrust cases the defendants
often claim that the market should be defined more widely to include
substitute products. If we consider the computer industry, for example,
we find different concentration levels depending on whether we just
consider laptop manufacturers or we include mainframes, desktops
and workstations. In the pharmaceutical industry concentration may be
relatively low if we consider the market for ‘drugs’ whereas, if we define
the market as ‘drugs for gout’ it may be highly concentrated. Given the
importance of barriers to entry for firms’ behaviour in an industry, the
simple measures of concentration have to be supplemented by further
analysis. In particular the evolution of concentration measures may give an
indication of whether firms in an industry compete fiercely. If the shares
are relatively stable, this may indicate the absence of competition.
There are some measures which, unlike CRs, are not officially published
but can be constructed from industry data. The Lerner monopoly
index measures the wedge between price and marginal cost:
L = (p – MC)/p.
It is meant to give an indication of market power and entry barriers. When
an industry is concentrated and has low entry barriers, the concentration
ratio is high but the Lerner index would be expected to show evidence
of limit pricing. Alternative measures of market power are cross-
elasticities. Firms producing a good which has many close substitutes
(large positive cross elasticity) have less market power than those
producing goods with few distant substitutes.

Perfect competition
In a perfectly competitive market there are many buyers and sellers who
act independently (non-strategically) and have perfect knowledge of the
market and in particular of the price charged by all sellers. The product is
homogenous so that firms do not engage in non-price competition. There
is no point in advertising for example if consumers are perfectly informed
and all products are identical. This also means that there is no price
dispersion (i.e. there is a unique market price at which everyone buys and
sells). No firm has control over this market price. It is often also assumed
that all firms have access to the same technology and this assumption,
combined with the fact that firms face the same input prices, implies
that the firms will have identical LRAC curves. Inputs are assumed to be
perfectly mobile and responsive to monetary incentives and there are no
barriers to entry in the long-run. Entry then occurs until economic profit
(of the marginal firm if they are not all identical) is zero. Zero economic
profit of course allows for a normal rate of return on equity.
As a result of these assumptions a firm behaves as if the demand for its
product is horizontal. Clearly the firm is in a sense making a mistake
here: when it decides to increase its supply the price of output is affected
because the industry demand curve is downward sloping. The same
comment applies to consumers. If any consumer increases his consumption
of the good he has a (very small) effect on price since industry supply is
upward sloping. The idea that competitive firms and consumers ignore
their effect on output price should be seen as a convenient simplification.
The assumptions of the perfect competition model may seem unrealistic
but there are several real world industries which conform relatively closely
to the competitive framework. Markets for some agricultural products
operate in a competitive fashion as do financial markets for savings and
136
 Chapter 9: Market structure

stocks. The packaged tour business is also very competitive with operators
regularly going out of business. There are virtually no barriers to entry in
this business. Chartering aircraft or setting up a travel agency does not
require major capital investment.

From cost function to supply function

Competitive firms are assumed to maximise profits which, given their price
taking behaviour, means:
max pq – C(q) (1)
q
where q is the firm’s output level, p is the market price and C is the firm’s
cost function.
The first order condition for (1) is:
p = MC(q) (2)
which means that the firm chooses its output level so that its marginal cost
equals the market price. The firm’s supply curve therefore coincides with
its MC curve. We need to qualify this to ensure that the firm is not making
negative profits. If the best output level according to (2) leads to losses,
the firm should shut down (set its supply q = 0). The shutdown conditions
are as follows:
1. in the short-term, if revenue exceeds variable costs, the firm should
continue to produce since revenue contributes to fixed cost
2. in the long-term, if revenue does not cover all costs, the firm should
shut down.
These conditions imply that the firm’s supply function is only a section of
the MC curve, the section above average variable cost for short-run supply
and the section above average cost for long-run supply. Short-run supply
coincides with the short-run marginal cost curve (some inputs are held
fixed) and long-run supply with the long-run marginal cost curve.

Case: Competition in the insurance industry

In many countries the insurance industry is fiercely competitive. The
‘insurance cycle’ is a well-documented phenomenon. During such a
cycle the industry goes through a phase of overcapacity and price wars.
This leads to losses which cause the marginal (usually smaller) firms
to exit. Historically when the industry has been free of price wars for a
short period of time a natural catastrophe such as Hurricane Andrew
has struck eliminating the need for a price war to rid the industry of
the weaker players. In any case a new equilibrium settles in with fewer
firms and higher premiums and profits start to appear, encouraging
new entry and buildup of excess capacity. A typical cycle lasts three to
five years.5 5
See ‘Unsure about
insurance’; ‘Insurers get
From firm supply to industry supply that sinking feeling’

As a first approximation for industry supply we could sum the individual

firm supply curves horizontally (i.e. for each price we determine (using
the MC curve) how much each firm is willing to supply and we add these
output levels to find industry supply at that price).

137
28 Managerial economics

Example 9.1

Suppose the annual market demand for wild pasta is given by

D(p) = 100 – 5p and there are 10 firms in the industry, each with
production function q = (LK)1/2 where L is water and K is wheat. In the
short-run the amount of wheat is fixed at 1 unit. The input prices are
r = 2 for wheat and w = 1 for water. Since q = L1/2 in the short-run
(K = 1) each firm’s requirement of water is L = q2 so that the cost
function is given by C(q) = q2 + 2 (since w = 1, K = 1 and r = 2). Each
firm’s supply function is thus p = MC = 2q or q = p/2. We do not have to
worry about shutdown here since MC always exceeds average variable
cost. The industry supply curve or the sum of the individual supply
curves is then Q = 10(p/2) = 5p and we find the equilibrium in this
industry by equating demand and supply: 100 – 5p = 5p which results
in a market price of p = 10 and quantity Q = 50. At the market price of
10 each firm wants to produce five units and makes a profit of (10)(5) –
C(5) = 50 – 25 – 2 = 23.

As an exercise you should check how many firms could enter this
industry and still make a profit.6 6
Hint: Let the total
number of firms be
n and calculate the
The distinction between short and long-term is also important for industry
equilibrium market
supply. Long-run supply is more elastic than short-run supply because all output and price as a
inputs are variable and there is entry and exit. Sometimes the notion of function of n. Then find
intermediate run is used meaning the period of time in which existing the maximum n for
firms can adjust most or all of their inputs whereas in the long-run market which profit is positive.
entry and exit can take place. If all firms are identical the entry and exit
process ensures that each firm produces the output level at which its LRAC
are minimised (i.e. at its efficient scale) and price equals LRAC. Why is
this so? Clearly price cannot go below minimum LRAC or nothing would
be produced. If it is above LRAC then firms can, by choosing their output
level carefully, make positive profits which would induce entry. Note that,
if firms are not identical, long-run supply is no longer constant but upward
sloping. At the equilibrium in such an industry only the ‘marginal’ firms
make zero profits; the other firms make positive profits.

p
∑ MC ∑ MC
1 0
p
1

p
0

Figure 9.1: Input price effect on industry supply

Recall the derivation of industry demand for inputs where we concluded
that the output price effect of a change in input price meant that we
could not simply sum individual firm demands for the input to obtain
industry demand. A very similar story can be told here. As the output price
increases the demand for inputs will shift out which will cause input prices
to rise (see Figure 9.1). At the changed input prices each firm’s marginal
cost or supply curve will shift inwards which makes industry supply less
138
 Chapter 9: Market structure

elastic than the horizontal sum of firm supply curves. This input price
effect of a change in output price is likely to be significant when the
industry is a large user of any of its inputs.
It is very difficult to assess the size or the significance of the input price
effect. In some cases it may be counteracted by new entry into the
industry as higher product prices lead to (temporary) larger profits. When
industry output increases there may be some positive externality effects
as well such as technological improvements through learning by doing for
example or improved support services. These factors would tend to make
industry supply more elastic than the horizontal sum of MC curves.

Monopoly
A monopoly is an industry consisting of one firm. Apart from state owned
or heavily regulated industries such as electricity there are not many
markets which conform to this extreme market structure. In practice
however, we assume that a firm will behave as a monopolist when it is
has a dominant market share of, say, 60–70 per cent and there is no major
rival. Microsoft and Intel for example have market shares of about 80 per
cent in operating software and microprocessor chips for PCs respectively.7 7
See ‘Ahead for now’
Monopolies are relatively common in emerging economies and former
communist countries because of protectionist measures and the lack of
antitrust legislation. For example, in Mexico, Vitra has a market share of
90 per cent in flat glass. Philip Morris took over Tabak (a Czech tobacco
company) which gives it control over 80 per cent of the market in the
Czech Republic. A few other examples of monopolies are:

Monopolies based on patents

When firms invent a new product or process they are entitled by
patent law to have temporary exclusive rights to market or license
this innovation. Companies such as Polaroid and Xerox were in the
fortunate position of building up a significant first mover advantage due
to patenting. In the pharmaceutical industry each new drug is patent
protected. When such a patent expires, competition from generic drugs
cuts price dramatically. In addition to the patent laws, governments can
create monopolies by granting a government franchise or ‘sole rights’
to operate a certain business. In 1994 the British government awarded
Camelot, a consortium of several companies, the contract to run the
national lottery as a monopoly for a seven year period. Other firms are
prohibited from running a lottery with the exception of clubs, charities and
local councils and then only if the turnover is small (less than £5 million
a year). This setup mirrors what goes on in other European countries.8 In 8
See ‘National lottery’
Sweden alcohol stores are state-owned and have a monopoly on the sale
of strong beer, wines and spirits. This should change now that Sweden has
entered the EU as an alcohol monopoly is against EU competition law.

Utilities
Whereas in Britain, New Zealand and the US you can choose your
telephone company, in many countries the telecom industry is a monopoly.
Hongkong Telecom will face competition from three new competitors from
1995 but is allowed to keep its monopoly on international calls (which 9
See ‘Squeezing into
generate most of its profits) until 2006.9 In Britain the Post Office keeps its Hong Kong’
monopoly on letter deliveries below £1.9.10 10
See ‘The Post Office’
The standard monopoly model is very simple. Industry demand equals
firm demand and, although we could make a distinction between short-
and long-term with respect to the monopolist’s cost function, there is no
139
28 Managerial economics

entry or exit. The profit maximisation problem can be stated in terms of

price or quantity. Both formulations give the same result of course. Using
the quantity formulation leads to the familiar ‘marginal revenue equals
marginal cost’ result. If we think of the monopolist’s problem as one of
choosing an optimal price, we have:
max pq(p) – C(q(p))
p
The first order condition is:
Pq' (p) + q(p) – MC(q(p)) q' (p) = 0.
With some algebraic manipulation this can be rewritten in terms of the
price elasticity as:
(p – MC)/p = 1/η or p = (η/(η – 1)) MC.
The markup η/(η – 1) clearly depends on how price elastic demand is.
For example, for the constant elasticity demand curve q = ap–2, the markup
equals 2 so that the optimal price is double the marginal cost: p = 2 MC.
This result also shows that, if the firm is profit-maximising, the Lerner
index is the inverse of price elasticity.
The markup formula bears a strong resemblance to the popular practice
of cost-plus pricing. Many pricing decisions are made using the rule
of thumb that you should take an estimate of average variable cost, add
a charge for overhead and a profit margin. If the profit margin reflects
market conditions (price elasticity) and if economic costs (i.e. opportunity
costs rather than accounting costs) are used, there is not too much
wrong with this. The main problem is the overhead charge. For profit
maximisation only marginal cost is relevant, not average cost.
The simple monopoly model is not very interesting and there are several
reasons, such as regulation, why we do not observe the behaviour
it predicts. The really interesting questions arise when we consider
how monopoly arises as a market structure and how a firm can stay a
monopoly. Unless the barriers to entry are very high or the monopolist
has a significant cost advantage, potential entrants are attracted by juicy
monopoly profits. If the monopolist is not prepared or forced to share
the market, how can he discourage entry? If the monopolist engages in
limit pricing, we would not observe MR = MC. If he has a cost advantage,
he could threaten to flood the market when entry occurs. If this threat is
credible the monopolist may use the MR = MC rule. It is likely, however,
that the threat is not credible (subgame perfect) as it may be more
profitable to accommodate the entrant.11 If the incumbent could commit 11
Recall the discussion
to a ‘fighting’ strategy by announcing or advertising that he is willing to of the chainstore
paradox in Chapter 2,
match or undercut any price offered elsewhere, entry can be deterred.
‘Game theory’
Similarly, if the monopolist can obtain some large long-term contracts from
major buyers he is more likely to be in a position to discourage entry. We
will return to these issues later when we have looked at oligopoly. At this
point these questions are hard to tackle because we have to model what
happens after entry to decide whether it is worthwhile to try to deter entry
for example.
Another reason why we may not observe the MR = MC rule in real-life
monopolies is that some of these monopolies operate in contestable
markets. Such markets have very low entry and exit barriers and the
reason they are monopolies is usually due to economies of scale. Average
cost decreases over the relevant output range so that only one firm can
survive but many firms may compete to be the single supplier. Contestable
markets are markets in which sunk costs are low. A frequently cited
140
 Chapter 9: Market structure

example is that of an airline route between two small cities. Entry is easy
for anyone who has planes available (on other less lucrative routes) and
exit is easy because planes can be sold or leased or switched to other routes.
In these circumstances, a monopolist cannot exercise his market power
because, if he did, he would be replaced by a willing and able entrant. In
fact monopolists may even behave as if they were in a perfectly competitive
industry if the market is contestable.

Monopolistic competition
Monopolistic competition is the market structure which emerges if we relax
the assumption of homogenous goods in the perfect competition model.
There are many sellers producing a differentiated product. For example, in
the beer industry, each brewery markets a unique product (although it has
many close substitutes). The main consequence of relaxing the assumption
of homogeneity of products is that some forms of non-price competition
such as advertising and providing different types of service can be profitable.
Monopolistic competition is associated with industries of firms with large
advertising budgets. Many ‘fast moving consumer goods’ (FMCG) sectors
fall into this category. Firms make enormous efforts to convince their
consumers that the washing powder they market, for example, is not
identical to the one marketed by their rivals. You could say that advertising
is necessary in monopolistic competition to ensure that buyers and sellers
have perfect information about the market which is what we continue to
assume. There are, apart from advertising, many methods firms use to create
differentiation. The following list contains a few examples:
• packaging: two chemically identical dishwasher detergents are
perceived as different by consumers if one is packaged in a plastic bottle
and the other in a cardboard box
• product design: shampoo and conditioner are sold separately and
combined
• type of service: in the PC industry it is becoming increasingly difficult
to produce a differentiated product. The same components are used in
the assembly of most PCs. Almost all PCs use Intel’s microprocessors. Dell
however has been very successful in this industry because it invented a
new way to sell PCs. Dell computers are sold by mail order and there is a
telephone hot-line which offers technical advice and after-sale service
• location: supermarkets and restaurants, even when they offer the same
range of products, are differentiated by their location. It is unrealistic to
assume that, if a supermarket charges a penny more for butter than the
supermarket at the other end of the high street, it will lose all its business,
or conversely that, when it undercuts its rival by a small amount, it
captures the whole market.

Short-run and long-run

A crucial assumption in the monopolistic competition model is that, as in
perfect competition, fines do not act strategically when formulating price and
output decisions. They do not take potential rivals’ responses into account
and they do not collude. This distinguishes monopolistic competition from
differentiated oligopoly.
Whereas perfectly competitive firms are forced to sell at the market price,
in the monopolistic competition model, a firm does not lose all of its sales
if it prices slightly above its competitors. This implies that each firm faces a
downward sloping demand curve which is highly but not completely elastic.

141
28 Managerial economics

Each firm has some monopoly power. Indeed one of the reasons to engage
in non-price competition such as improving service is that a firm which
produces a differentiated product does not have to cut its price whenever a
rival cuts his price.
The short-run monopolistic competition model is in fact exactly the same
as the monopoly model with firms setting output levels such that MR =
MC. As in perfect competition there are no or low barriers to entry so that
in the long-run entry and exit ensure that profits are zero. The long run
equilibrium conditions for a monopolistically competitive industry are as
follows:
• at the optimal output level q* for each firm, price should equal average
cost (AC) so that profit is zero
• at its optimal output level each firm maximises profit and hence MR =
MC at q*.
From the long-run equilibrium conditions we know that, for the optimal
output level q*, MR intersects MC. The price corresponding to q* can be
read off the demand curve and has to equal AC as is illustrated in Figure
9.2. This leaves the possibility of demand intersecting AC rather than
being tangential to it at q*. However, this would mean that there are
output levels for which the firm could make a positive profit contradicting
the fact that q* is profit maximising. The figure shows that, at any output
level other than q*, the firm makes a loss. Because demand has to be
tangent to AC at the long-run equilibrium, firms operate at output levels
to the left of the efficient AC minimising level. If they could increase
their output, their AC would decrease. The cost of this ‘excess capacity’
is often referred to as the price of variety. Consumers are willing to pay a
premium, so the argument goes, to have a choice between differentiated
products. When, during your weekly food shopping, you have to find your
way through the enormous variety of breakfast cereals for sale on the
supermarket shelves you may be forgiven for being sceptical about this. In
the US in 1990 alone, 48 new cold or cough remedies were introduced. It
is hard to believe that consumers want this much variety.

p
MC

AC

p*

D
MR

q* q

Figure 9.2: Monopolistic Competition – Long Run Equilibrium

Many industries satisfy the assumptions of the monopolistic competition
model although inevitably there are entry barriers because of large
advertising budgets for example. In such cases the analysis is still valid but
we cannot insist on zero profits at the long-run equilibrium.

142
 Chapter 9: Market structure

Because of product differentiation it is generally not easy to define a

monopolistically competitive industry. For example, does the market for
soap include liquid soap or can we restrict it to soap bars? Whereas a
perfectly competitive industry consists of firms producing perfect substitutes,
in monopolistic competition products are close but not perfect substitutes.
What ‘close’ means in practice is difficult to say and always entails some
arbitrary decisions. In empirical work a judgement has to be made when
specifying demand functions as to which cross price effects to estimate and
which to ignore.

Advertising and the Dorfman-Steiner model

Advertising plays a major role in monopolistically competitive industries. This
gives us an opportunity to look at a model of advertising expenditures here.
Even if two shampoos are chemically identical their demand curves are not
perfectly elastic as in the perfect competition model if one is advertised by
Naomi Campbell and the other by ...let’s just say someone like me. In general,
advertising a product has the effect that consumers consider fewer products
as good substitutes for the product. Advertising shifts the demand curve and
changes price elasticity. Analytically this implies that the decisions of how
much to advertise and which price to set cannot be taken in isolation.
Deriving the optimal advertising budget and price is not hard if we know how
demand responds to changes in price and advertising. Let q(p, a) be
the demand function. Then the firm’s profit function is:
π = pq(p, a) – C(q(p, a)) – a.
The first order conditions are:
∂π
=p
∂q(p,a)
+ q(p,a) –
∂C ∂q(p,a)
=0 (3)
∂p ∂p ∂q ∂p
and
∂π ∂q(p,a) ∂C ∂q(p,a) (4)
=p – –1=0
∂a ∂a ∂q ∂a

After some algebraic manipulation, we can rewrite these conditions in terms

of the price and advertising elasticity of demand:
η = p / (p – MC) (5)
and
∂q(p,a) a a
α≡ (6)
∂a q = q(p – MC ) .
Equation (5) is the familiar markup formula for monopoly pricing. From
equation (6) we deduce that advertising increases with the responsiveness
of demand to advertising, the output level and the profit margin. At the
optimum, the ratio of advertising expenditures to sales revenue equals the
ratio of advertising and price elasticities (divide (6) by (5)):
a / (pq) = α/η (7)
This result was first shown by Dorfman and Steiner (1954) and has
generated a huge literature in the field of marketing models, In empirical
studies it is found that the advertising-sales ratio is positively related to
price-cost margins12 which confirms the Dotfinan-Steiner result (substitute 12
Martin (1993) p.159
(5) into (7)) since price elasticity is negatively correlated with the price cost
margin. Of course this simple model does not capture all of the interesting
aspects of the advertising decision. For example, it ignores dynamic effects
of advertising. A company’s current advertising budget is unlikely to just
affect current sales. By advertising now a stock of goodwill is built up which
leads to higher sales in the future as well.

143
28 Managerial economics

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• describe the important factors in the determination of market
structure giving (preferably your own) examples
• calculate the market structure measures
• derive perfectly competitive firm and industry supply
• find profit maximising price and output for a monopoly
• explain short run and long run equilibrium in a monopolistically
competitive industry.

Sample exercises
1. The market demand for a product is as indicated in the table below.
The industry is perfectly competitive and the second table gives each
firm’s long-run total cost. Each firm can produce only integer numbers
of units of output. How many firms will be in the industry in the long-run?
price (£) quantity demanded
30 200
20 300
10 400
5 600
3 800
output total cost (£)
1 10
2 12
3 15
4 30

2. The demand for a textbook is given by p = 20 – 0.0002q. The publisher’s

marginal cost function is MC = 6 + 0.00168q. The author of the textbook
gets a royalty of 20% of total revenue. Suppose the publisher decides
on the price. What is his preferred price-quantity pair? Suppose the
author could decide on the price. What is her preferred price-quantity
pair? In the first scenario (the publisher setting the price), how much
would the publisher offer the author as a lump sum payment in return
for the author giving up the royalty. Should the author accept?
3. Mickey Mouse Publishing Ltd. is a price-taking firth in the market for
microeconomics textbooks. It produces books with total cost function
C(q) = 3g2 + 6q + 3. What is its short-run supply function? What is its
long-run supply function?
4. The Seow Food Corporation has bought exclusive rights to sell
chocolate bars at Wembley arena. The fee it paid for the concession
was £1000 per event. The cost (excluding this fee) of obtaining and
marketing each candy bar is 10 pence. The demand schedule for candy
bars in the arena is as indicated in the table below. Prices must be in
multiples of five pence. What price should Seow charge for a candy bar
and what is the maximum amount that it should pay for a concession
for a single event?
144
 Chapter 9: Market structure

Price per candy bar (pence) Thousands of candy bars sold per game
20 10
25 9
30 8
35 7
40 6
45 5
50 4
5. Demand for a monopolist’s goods is given by q = S/p0 if p ≤ p0 and q = 0
if p > p0. S is a constant. What is the monopoly price?

145
28 Managerial economics

Notes

146
 Chapter 10: Monopolistic pricing practices

Chapter 10: Monopolistic pricing

practices

Aims
The aim of this chapter is to consider:
• the differences between first, second and third degree price
discrimination
• why under first degree price discrimination all consumer surplus can be
extracted
• why under two part pricing, at least for identical consumers, unit price
should be set equal to marginal cost
• the relationship between price in a market and its price elasticity when
third degree price discrimination is practised
• intertemporal price discrimination and the Coase conjecture
• the peak-load pricing model
• the difference between pure and mixed commodity bundling
• the concepts of loss leaders, cannibalisation, joint products and
economies of scope
• the joint products model and why a monopolist might waste rather
than sell some of his output
• the three transfer pricing scenarios
• the relevance of taxation systems on the practice of transfer pricing.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• give (your own) examples of price discrimination and commodity
bundling
• derive the optimal take-it-or-leave-it-offer in first degree price
discrimination
• derive the optimal two-part pricing scheme
• derive optimal prices and quantities in third degree price discrimination
(analytically and graphically) and determine the optimal price and
quantity when the monopolist faces two demand curves and is not
allowed to price discriminate
• solve a simple two period skimming model with myopic consumers
• solve a simple peak-load pricing problem
• show graphically how consumers with different reservation prices
distribute themselves over the various options in commodity
bundling
• for a multiproduct firm derive prices and quantities for two
products when
a. each product is handled by a separate division and divisions act
noncooperatively and

147
28 Managerial economics

b. when the decisions are taken to maximise firm profit

• solve a joint products problem
• derive the optimal transfer price (analytically, graphically and
numerically).

Essential reading
Tirole, J. The Theory of Industrial Organisation. Sections 1.1.2, 1.5.2, Chapter 3.
Varian, H.R. Intermediate Microeconomics. Chapter 25.

Further reading
Adams, W. and J. Yellen ‘Commodity bundling and the burden of monopoly’,
Quarterly Journal of Economics 90 1976, pp.475–98.
‘Ad nauseam’, The Economist, 5 February 1994, p.71.
‘A hell of an operating system’, The Economist, 18 January 1995, pp.17–18.
‘Better than price fixing?’, The Economist, 3 October 1992, p.100.
‘By the seat of their pants’, Airlines Survey, The Economist, 12 June 1993,
pp.26–28.
‘Europe’s car market. Carved up’, The Economist, 31 October 1992, p.94.
‘Death of the brand manager’, The Economist, 9 April 1994, pp.79–80.
Fisher, K. ‘Return to sender’, The Economist, 4 February 1995, p.8.
‘Hard sell’, The Economist, 4 March 1995, pp.89–92.
Hirshleifer, J. ‘On the economics of transfer pricing’, Journal of Business 29 1956,
pp.172–84.
‘IOU all over again?’ The Economist, 2 July 1994, pp.40–41.
‘Labouring in obscurity’, The Economist, 17 September 1994, p.94.
Lan, L. and A Kanafani ‘Economics of park-and-shop discounts: a case of bundled
pricing strategy’, Journal of Transport Economics and Policy 3/27, pp.291–303.
‘Learning to fly all over again’, Airlines survey, The Economist, 12 June 1993,
p.13.
‘Life after Lenin’, The Economist, 29 January 1994, p.74.
‘Managing the future’, The Economist, 19 December 1992, pp.70–75.
‘Movie mystery’, The Economist, 19 November 1994, p.36.
‘Not dead yet’, The Economist, 28 January 1995, pp.76–80.
Oi, W. ‘A Disneyland dilemma: two-part tariffs for a Mickey Mouse monopoly’,
Quarterly Journal of Economics 85 1971, pp.77–90.
Schmalensee, R. ‘Gaussian demand and commodity bundling’, Journal of Business
57(1,2) 1984, S211–S230.
‘Tax deficient’, The Economist, 22 May 1993, p.83.
Stigler, G.J. ‘United States v. Loew’s Inc: A note on block booking’, Supreme
Court Review (1963) pp.152–57. Reprinted Stigler, G.J. The organisation
industry. (Homewood, I11: Irwin, 1968).
‘Taxing questions’, The Economist, 22 May 1993 p.83.
‘The high-tech war’, The Economist, 26 December – 8 January 1993, pp.83–84.
‘The richest islands in the world, maybe’, The Economist, 6 November 1993, p.80.
‘Unhappy returns’, The Economist, 2 April 1994, p.88.
‘Unhappy returns for Barclays’, The Economist, 25 June 1994, p.97.
van Ackere, A. and Reyniers, D.J. ‘A rationale for trade-ins’, Journal of Economics
and Business 45 1993, pp.1–16.
van Ackere, A. and Reyniers, D.J. ‘Trade-ins and introductory offers in a
monopoly’, RAND Journal of Economics (1995) 26 pp.58–74.
Webster, T. ‘Rooms with a view to saving’, Evening Standard, 30 November 1993,
p.38.
‘What went wrong at IBM’, The Economist, 16 January 1993, pp.23–25.

148
 Chapter 10: Monopolistic pricing practices

Introduction
In our discussion of monopoly pricing (Chapter 9) we found that the
monopolist sets price such that marginal revenue equals marginal cost.
We implicitly assumed that the monopolist had no option but to set a
single price, the same for all buyers, independent of the quantity they buy,
whether they buy any other products from the monopolist, etc. In reality,
sellers often have the opportunity to market their products at different
prices (price discrimination) or sell them as part of a package (commodity
bundling). So far, we have also assumed that the firm markets one product
or that we could analyse production and marketing of individual goods in
isolation. However, most firms produce a range of products and they find
it profitable to keep the pricing of the entire product line in mind when
making pricing decisions for each product.
Many firms are structured as multidivisional organisations with one
division buying goods from another division within the firm. The transfer
price at which such goods are sold between divisions is of crucial
importance for the profitability of the firm. If selling divisions are rewarded
on the basis of their divisional profit, they will tend to set the transfer price
too high which is detrimental to overall profitability. Clearly, transfer prices
have to be determined centrally in such situations. In this chapter we turn
our attention to these types of pricing problems.

Price discrimination
Of all pricing practices, monopolistic or otherwise, price discrimination
in its many forms is certainly the most visible. Price discrimination occurs
whenever the same good or service (i.e. a good or a service produced at
identical cost) is sold at different prices. More generally, we can say that,
when price differences between consumers are larger than is warranted
by cost differences, there is price discrimination. For example, British Gas
offers a reduction of your gas bill if you pay by direct debit. To the extent
that collecting revenue by direct debit is less costly than by alternative
methods, the discount is not an indication of price discrimination. It is
more difficult to argue that price differences between first and second class
on trains, or business and economy on planes, are justifiable on the basis of
cost differences. There is something else going on here.
For a firm in a perfectly competitive industry it is not possible to indulge
in price discrimination. If such a firm sets a price above the market price,
it loses all of its customers. Some degree of market power is necessary
to enable a firm to price discriminate. It is easiest to analyse price
discrimination by a monopolist although in reality it occurs mostly in
oligopolies. (Models of price discrimination by oligopolists are the subject
of recent and current research in industrial organisation.) It is easy to
understand the attraction of charging different prices for the same good.
Clearly, profits cannot decrease as a consequence of price discrimination
since one option is to set the same price in all markets. In general, profits
increase if price discrimination is possible.
Why is price discrimination more common in the sale of services than
manufactured goods? For services, resale is generally not possible. I
cannot get a haircut or a massage and sell it to you, for example. Utilities
such as telephone and electricity companies often successfully use price
discrimination. Customers cannot resell their services.

149
28 Managerial economics

When resale is possible, price discrimination may not succeed because

of arbitrage. The term arbitrage is used when it is possible for low-price
buyers to sell in a high price market. If a wholesaler offers quantity
discounts to retailers, he has to ensure that it is impossible for a retailer
to buy a very large quantity taking advantage of the discount and then
resell the product to the wholesaler’s other customers. Similarly, when
a Japanese car manufacturer wants to charge different prices in Europe
and in the US, the potential price differential is limited by transportation
costs. If the price difference is too large (e.g. when Japanese cars in the
US are much cheaper than in Europe), they will be shipped from the US
to Europe (in the absence of quotas, etc.).
Economic theory classifies price discrimination practices according to
how much information the monopolist has, whether consumers can
choose from a price menu (e.g. when quantity discounts are offered)
and whether markets are isolated (no arbitrage possibility) so that
consumers in each market or market segment are quoted a segment
specific price.

First degree price discrimination

Suppose a consumer with income m derives utility from widgets x and
income remaining y after paying for the widgets. A seller of widgets who
knows the utility function can offer the consumer a take-it-or-leave-it
deal. She can determine the maximum amount the consumer is willing
to pay by comparing the utility of consuming widgets and paying a
certain amount P, to the utility without widgets. You can think of the
seller choosing a point on the consumer’s indifference curve through
(0,m) as is illustrated in figure 10.1. If the seller asks for a payment P*
in return for x* widgets, the consumer will just about accept this offer.
If the seller asks for more than P* or offers less than x* in return, the
consumer rejects the offer.

m- P*

x
x*

Figure 10.1: First degree price discrimination

150
 Chapter 10: Monopolistic pricing practices

Example 10.1

A consumer with income m derives utility from a good x produced by

a monopolist and from his remaining income y according to utility
function U(x, y) = x1/2 y + y. Hence his utility is U(x, m – P) = x1/2 (m – P)
+ (m – P) when he buys x units and pays the monopolist P and utility
U(0, m) = m if he goes without product x. The monopolist who knows
the consumer’s utility function is able to make a take-it-or-leave-it offer,
that is, he says ‘I will give you x units if you give me £P; if you refuse,
you will get nothing’. The consumer accepts such an offer if U(x, m – P)
> U(0, m):
2
1/2
(
x (m – P) + (m – P) ≥ m or x ≥ m – P
P
) .
(1)
Assume the monopolist has constant marginal costs c. To determine his
offer P optimally he maximises his profit P – cx subject to (1). Clearly he
is not going to give the consumer more than is necessary to make him
accept the offer. This implies that he sets x = (P/(m – P))2. His problem
then becomes one of finding the optimal payment P:
2
(2)
Max ( P
= P – cx = P – c m – P ).
The first order condition corresponding to (2) is:
P
(
1 – 2c m – P )( m
)
(m – P) 2 = 0 .
which can be rewritten as:
1 = Pm .
2c (m – P) 3 (3)
You should check that the second order condition is satisfied for
P < m. The optimality condition (3) does not have an analytical solution
but given any values of m and c we can solve for P. If marginal cost
is c = 4/3 and income m = 3, then the monopolist should set
P = 1 and x = (P/(m – P))2 = 1/4, i.e. he should make an offer of 1/4 unit at
a payment of 1. We can check in (1) that the consumer will just accept
this offer: (1/2)(3 – 1) + (3 – 1) = 3. The monopolist, who is in the most
enviable position of knowing the consumer’s utility function, gets profit
π = 1 – (4/3)(1/4) = 2/3. You should check that, when the monopolist is
restricted to setting a fixed unit price, his optimal profit would be less
than 2/3.1 1
Hint: determine the
consumer’s demand
function and set
There is an alternative way to analyse the problem of first degree price
MR = MC
discrimination. It starts out with a consumer’s demand function rather
than the utility function and uses the idea that the area under the demand
curve (consumer surplus) represents ‘willingness to pay’ or a monetary
equivalent of the utility a consumer derives from consuming the good.
However, this interpretation is only valid for special classes of utility
functions such as quasi-linear utility. A utility function of type U(x, y)
= v(x) + y, where x is the quantity of widgets produced by the monopolist
and y is remaining income, is quasi-linear. For such a utility function, the
indifference curves in the x – y plane are vertically parallel and hence the
income elasticity of widgets is zero. When we can use consumer surplus
as a measure of how much consumption of a good is worth to a consumer,
the monopolist’s problem can be illustrated as in Figure 10.2. For each
choice of x, the monopolist knows how much the consumer is willing
to pay e.g. for x = x′, the seller can demand the consumer surplus ab x′
0. His costs are cd x′ 0 so that his profits are abdc. Profits are maximised
when the monopolist offers to sell x* units and demands payment of the
151
28 Managerial economics

entire consumer surplus. In other words, the monopolist is selling each

unit at the highest possible price the consumer is willing to pay. Although
we have concentrated on the idea of the monopolist offering take-it-or-
leave-it deals to individual consumers, we can think of D as the market
demand curve. The same conclusion holds in that the monopolist will sell
the quantity which maximises consumer surplus minus costs and demand
payment of the entire consumer surplus which is the sum of individuals’
consumer surplus.
Under first degree price discrimination, assuming constant marginal cost,
the seller deals with each consumer and determines individual take-it-or-
leave-it-offers. The offer (P, x) varies from person to person. Of course this
is all quite unrealistic. In practice you might come reasonably close to the
situation described above when there are few buyers and the seller has
individual contact with each of them. For example, a car dealer may have
some idea of how much a potential customer is willing to pay. In any case,
the perfect price discrimination model provides a useful benchmark.

p
a
D
b

c
d

x
0 x′ x*

Figure 10.2: Extraction of consumer surplus

Second degree price discrimination

Second degree price discrimination is also referred to as nonlinear
pricing or block pricing. Under this scheme, the seller offers
consumers a menu consisting of prices corresponding to different
quantities. Usually price is lower if a higher quantity is bought (quantity or
bulk discount). Foods and drinks are often sold at lower per unit prices in
‘family’ packs. This is price discrimination unless the price difference with
single item sales can be justified by differences in distribution or packaging
costs. Utilities (gas, electricity, water, telephone) offer price menus on
which price decreases with the number of units bought. In theory a seller
could offer a price menu with prices increasing when larger quantities are
bought. In practice consumers would buy small quantities several times
rather than a big quantity unless of course the seller can keep track of all
this by making sure that sales are not anonymous. Indeed, for bank loans
where price (interest) increases with quantity, buyers (borrowers) are not
anonymous.
In contrast to first degree price discrimination, under second degree price
discrimination every individual who buys the same quantity pays the same
price. In a sense consumers self-select to pay different average prices
through their choice of consumption quantity.

152
 Chapter 10: Monopolistic pricing practices

A frequently used special case of nonlinear pricing is two-part pricing.

Consumers pay a lump sum ‘entry fee’ which gives them the right to
purchase the good plus a regular price per unit. Of course, this type of
pricing is only possible if customers cannot resell (to customers who did
not pay the fixed fee). One of the first papers2 on two-part pricing quotes 2
Oi (1971)
Disneyland amusement park which used to charge per ride in addition to
the entrance fee.
The following can all be interpreted as two-part pricing:
• telephone, electricity and gas companies charge a fixed rental fee plus
a price per unit
• nightclubs use entry fees in addition to a fixed price per drink
• for products which need complementary inputs or replacement
parts such as cameras and shavers, the price of the product is like an
entrance fee and in addition the consumer pays a fixed price per roll of
film or per blade
• taxis often charge a fixed fee plus a price per distance travelled.
In some cases, there is a cost based justification for two-part pricing,
for example, for telephones there is a connection cost and printing jobs
involve setup cost so that larger batch sizes are less costly per unit.
As an illustration of two-part pricing in the same framework as first degree
price discrimination, consider the problem of the monopolist who knows
an individual’s utility function. He offers the individual a combination of
a fixed fee F and a unit price p. Figure 10.3 shows that this problem is
equivalent to the one we encountered in first degree price discrimination.
The seller can in fact choose any point on the indifference curve through
(0,m) by an appropriate choice of F and p. The entry fee F determines the
intercept of the consumer’s budget line whereas p determines its slope.

m-F
a

x
x* (m-F)/p

Figure 10.3: Two-part pricing

If we analyse the problem in terms of the demand function (which we
are allowed to do only if the income effect is negligible), then for any
price p the seller charges, the maximum fee is the consumer surplus at
price p. Again the best the monopolist can do leads to the same result as
in the take-it-or-leave-it scenario, where the entire consumer surplus was
captured. Here, price is set equal to marginal cost and F is the consumer
surplus corresponding to p = c. This result is not dependent on the
assumption of constant marginal costs. The seller’s optimisation problem
is:
max p(q)q – C(q) + CS(q) = max p(q)q – C(q) + ∫q0 p(t)dt – p(q)q

153
28 Managerial economics

where CS(q) is the consumer surplus of consuming q units. The solution to

this maximisation problem is to set p equal to marginal cost.
The two-part pricing problem becomes more realistic and more interesting
but also much more complicated when the seller does not know individual
utility functions or demand curves. Suppose the seller knows there are
two types of consumers out there with demand functions q1(p) and q2(p)
respectively but he cannot distinguish between them. He also knows the
fraction of consumers belonging to each of these groups. The seller has
to offer the same two-part pricing offer (F, p) to all consumers. How is he
going to determine the best (F, p) offer? A first consideration is whether
the seller should aim to sell to both groups. It may be that one of the
groups has very low willingness to pay for the product and is not worth
trying to sell to. If we assume that both groups are served, then we could
determine the optimal F for any price p. It will be the minimum of the
consumer surplus at p for a Type 1 and a Type 2 consumer. (Remember
that we want both groups to buy.) This implies that the seller does not
succeed in extracting all consumer surplus. One group will enjoy positive
consumer surplus whereas the other’s will be zero. Average profit per
consumer consists of this fee F plus the profit margin times the average
demand (at price p), which is dependent on the relative size of the groups.
Although all consumers will be offered the same deal, the two groups
end up paying different average prices (F + pq)/q = F/q + p because their
quantity choice differs.

Third degree price discrimination

Third degree price discrimination is ‘group’ price discrimination. Different
people pay different prices according to which group they belong.
Typically they cannot change their membership status. For example, my
hairdresser gives a 50 per cent discount to elderly ladies (No, I do not
qualify!). It is important that it is impossible or unprofitable for low-price
buyers to resell to potentially high-price buyers. This is why third degree
price discrimination is most frequently used when selling services. Usually
in third degree price discrimination schemes you pay the same price
for each unit you buy (i.e. there are no quantity discounts). (In theory
you could combine second and third degree price discrimination.) The
following list gives a few examples of third degree price discrimination (I
am sure you can think of others):
• senior citizen discounts for movies and buses; student discounts for the
theatre; children’s discounts for the zoo
• academic journals which can be three times more expensive for
libraries than for individuals
• seasonal pricing at resorts
• some restaurants which offer the same menu for lunch and dinner but
dinner prices could be double the lunch prices
• afternoon and evening performances at the theatre may differ in price
• trade-in discounts are offered for many appliances
• fare classes on trains (first and second class) and planes (first, business
and economy class)
• companies with large travel budgets are given corporate discounts by
airlines and hotel chains.

154
 Chapter 10: Monopolistic pricing practices

The analysis of third degree price discrimination is similar to that of

division of output between plants. Indeed, you could analyse the (realistic)
scenario of a company manufacturing a product in several plants and
selling in several markets in this same framework. For simplicity we will
focus on price discrimination here and assume that everything is produced
in one plant with cost function C(q). The seller can charge different prices,
p1 and p2, in his two markets with demand functions p1(q1) and p2(q2).
The seller’s profit maximisation problem is:
max П = p1(q1 )q1 + p2(q2 )q2 – C(q1 + q2 )
q1, q2
with first order conditions:
MR1(q1 ) = MC (q1 + q2 )
MR2(q2 ) = MC (q1 + q2 ). (4)
The intuition behind this result is straightforward: you should always
allocate a unit of output to the market where it will have the highest
revenue. If you supply both markets, their marginal revenue should be
equal. Furthermore you should only sell another unit if its MR exceeds
its MC. You could decide not to sell at all in one market. This means that
the MR in the market which is not supplied is always below the MR in
the market which is supplied and is below MC at the optimal output
level. It is very important to realise that the marginal cost should be
measured at the total output q1 + q2. When MC is constant it obviously does
not matter where it is measured but in the general case of increasing MC it
does. Figure 10.4, which illustrates the solution we have just discussed, is
similar to the ‘division of output’ figure. The intersection of the horizontal
sum of the marginal revenue curves with the marginal cost curve
determines optimal total output. We can then trace back to the individual
MR curves to determine how much is sold in each market. The optimal
prices are read off the demand curves. If marginal costs are sufficiently
high (MC), only the ‘high willingness to pay market’ (Market 1) will be
supplied.

p2
p1

p1' MC'
D1 Σ MR

p1*
MC

MR1 p2* D2
D
MR2
q1*+ q2* = q0
q1* q 0 q2* q2 q
q1' q0 q1 2
1 q 1' + q2'

Figure 10.4: Third degree discrimination

Figure 10.4 also illustrates the monopolist’s pricing problem when he is
prevented from price discriminating between the two markets. In this case,
his demand curve is the horizontal summation of the market demands

155
28 Managerial economics

D = D1 + D2 and optimal output q0= q01+ q02 is determined where MC

intersects the MR corresponding to D. Since D is kinked, the MR (the bold
lines) is not continuous. In the figure, because the demand functions are
linear and it continues to be optimal for the monopolist to serve both
markets, the optimal output is unchanged. Of course the prices in the
two markets are different: the low price is increased and the high price
is reduced. In general, when price discrimination is not allowed, the
new common price should be between the two previous prices so that
consumers in the previously high price market benefit at the expense of
consumers in the previously low price market. For example, Distillers had
a dual pricing system in the seventies. Whisky exported to the continent
was more expensive than that sold in the domestic market. When this
was ruled to be illegal by the EU Commission, prices in the UK rose
considerably for brands such as Red Label.
When price discrimination is not allowed, a monopolist is more likely to
not to serve the low willingness to pay market.

Can you draw a MC curve on Figure 10.4 for which the monopolist
only sells in both markets when he is allowed to charge different
prices?

Generally output is lower when price discrimination is not allowed. If

the monopolist is constrained to charge the same price in all markets, he
may not want to expand production since it would lower the price of all
output but if he can price discriminate he can expand sales in one market
without hurting revenue in another. This is why the welfare effects of price
discrimination are ambiguous. Generally, some consumers gain and others
lose; the seller of course always gains.

Example 10.2

The domestic and foreign annual demands for a textbook entitled

Pricing for monopolists are pd = 100 – 15qd and pf = 60 – 2.5qf respectively
where p is the price in dollars and q is the number of copies sold (in
thousands). Any copy of the textbook can only be sold in the country
to which the publisher has consigned it (i.e. re-export is illegal). The
publisher’s annual production costs for this textbook are:
C(q) = 10.8 + 20q + 0.1q2.
The MR curves corresponding to the two demand functions are:
MRd = 100 – 30qd
and
MRf = 60 – 5qf.
The optimality condition (4) implies that these two expressions are set
equal to:
MC(qd +qf ) = 20 + 0.2 (qd + qf ).
Solving for qd and qf results in:
qd = 2.6 and qf = 7.6
which suggests (use the demand functions) prices:
pd = 61 and pf = 41.
At output qd + qf = 10.2, MC equals 22 which is also the marginal
revenue in both markets for these sales levels. Profit equals:
П = (2.6)(61) + (7.6)(4l) – 225.2 = 245 or $245,000 per year.

156
 Chapter 10: Monopolistic pricing practices

Now suppose the publisher is accused of dumping (i.e. selling at a

lower price abroad than domestically) and is forced to sell the textbook
at the same price domestically and abroad. Assuming the publisher still
finds it profitable to sell in both markets, he faces a total demand:
q = qd + qf = (100 – p)/15 + (60 – p)/2.5.
The publisher’s profit maximisation problem can be expressed with price
as the decision variable:
max П=pq( p) – C(q( p)).
p
The optimality condition is:
(5)
( p – ∂C∂q ) ∂q∂p = – q.
If we substitute the total demand q in (5), and solve for p we find
p = 43.83. At this price 3,740 copies (qd = 3.74) are sold on the domestic
market and 6,470 (qf = 6.47) are sold on the foreign market. The no
dumping restriction hurts the publisher’s profit which is now
П = (43.83)(10.2) – (10.8 + 20(10.2) + 0.1(10.2)2) = 221.9
so that profit has fallen to $221,900 annually.

Case: The European car market

Although price discrimination is illegal within the European Union, it is
a fact that, on 13 January 1993, the Opel Corsa pre-tax price was 9,276
ecus ($12,000) in Britain but only 6,137 ecus in Belgium; a Mercedes
200 cost 30,351 ecus in Ireland but 21,545 ecus in Germany; and a
Peugeot 205 cost 9,339 ecus in Germany but 7,205 ecus in Portugal.
These large price differentials are sustained by the bureaucratic
difficulties involved in importing a car in an EU country. Documents
have to be filled out, special registration procedures followed and rules
on technical standards protect domestic manufacturers. The Greek
government does not allow its citizens to buy foreign currency for the
purpose of importing a car. Dealers may not honour manufacturer
guarantees when cars bought overseas break down. Unlike the situation
in the US, dealers in Europe only sell one manufacturer’s cars. EU
competition rules forbid exclusive dealerships but, due to pressure
from the car manufacturers lobby, a 1985 regulation exempted the car
industry from these rules for 10 years. This means that car makers have
influence over dealers to keep prices up and to discourage cross-border
sales (e.g. Renault ordered its dealers in Belgium to charge much more
for right-hand-drive models destined for British roads).3 3
See ‘Europe’s car
market. Carved up’
Rewriting the optimality conditions (4) in terms of elasticity of demand
allows us to explain some features of third degree price discrimination in
practice. At the optimum, we find:
η1
p1 =( )
η1 – 1 MC(q1+ q2) and
η2
p2 =( )
η2 – 1 MC(q1+ q2).
η1 η2 when η < η , higher price will be set in the market with
Since > 1 2
η1 – 1 η2 – 1

157
28 Managerial economics

the least elastic demand. Every day, you can see applications of this
principle. People are charged more when they want to have things done in
a hurry (their demands are likely to be inelastic): for example, one-hour
versus 24-hour photo processing. This type of price discrimination based
on urgency of service also applies to same day versus same week dry-
cleaning etc.
Japanese car manufacturers charge more for their cars in Japan than
abroad because the overseas car market is more competitive (and hence
its price elasticity higher) than the protected Japanese market. Dumping
is illegal for goods imported in the US. To offset the price differential,
extra duties are imposed on importers found dumping. The situation in
the European Union is similar. However, in an interesting case of double
standards, the EU persistently dumps surplus agricultural commodities,
arising from farm support programs, in poor countries.
Because the income effect of a price change is likely to be larger for poor
customers, their demands are generally more elastic. In many instances we
do observe that poorer customers are charged lower prices. Now you know
that this may have nothing to do with charity as it can be explained by
pure profit maximising behaviour on the part of the seller. It is rational for
doctors and dentists to charge poor people a lower fee. Often geographical
proxies are used in the sense that everyone in a generally poorer area
benefits from lower prices and conversely everyone in a relatively richer
area is asked to pay more. For example, Warner Brothers’ cinema in
Leicester Square in London charges £7 for a ticket about double of what
is charged (£3.80) in its cinema in Bury near Manchester.4 Apparently, 4
See ‘Movie mystery’
Russian newspapers have caught on to this trick. In January 1994,
Izvestia charged 20m roubles ($13,300) for a full page advertisement
if the customer was a Russian company and $30,000 if it was a foreign
company.5 In China, outdoor advertising rates for foreign products are up
5
See ‘Life after Lenin’

to five times those for domestic products.6 Keith Fisher (1995), the director 6
See ‘Not dead yet’
of Royal Mail International, points out in a letter to The Economist that the
Royal Mail charges a developing country one penny to deliver an incoming
letter whereas an advanced country pays 16 pence. Measures have to be
taken to avoid abuse by ‘remailers’ who essentially arbitrage by shipping
bulk mailings to developing countries.
Marketing managers use various devices to discriminate between segments
of the market. Although grocery coupons have other uses such as
gathering information on demand by experimenting with the price, it
seems that sellers may be trying to discriminate between people who
collect and redeem coupons and those who do not bother. The assumption
is that the redeemers have more elastic demand. Coupons are used for
many FMCG such as breakfast cereals. The importance of this type of
promotion is illustrated by the fact that grocery-coupon redemption
amounted to $6 billion in the US in 1992.7 Similar reasoning applies 7
See ‘Death of the
to price matching which refers to the practice of offering to match brand manager’
any lower price the consumer can find. Usually an advertisement of the
lower price has to be presented. Again there are other explanations for
this marketing practice but, since very few customers collect on price
guarantees, price matching is an easy way to discriminate between
customers with low search costs (for lower prices) and those with high
search costs. It can be argued that the latter have more inelastic demands.

Skimming or intertemporal price discrimination

A monopolist wants to market a new durable good such as a CD player,
the type of product that, within a horizon comparable to its lifecycle, a

158
 Chapter 10: Monopolistic pricing practices

consumer buys at most once. A frequently observed pattern for these types
of goods is that they are initially sold at extremely high prices. After a
few years these innovative goods become more affordable. This pricing
pattern of decreasing prices over time is referred to as intertemporal price
discrimination. It is a form of third degree price discrimination in the sense
that consumers who buy later pay a lower price than those who buy early
on. At the same time you could argue that rational consumers can predict
that prices will decrease and can therefore ‘self-select’ when they will buy,
which makes intertemporal price discrimination more like second degree
price discrimination.
Price decreases over time are the norm in publishing. Publishers are in fact
monopolists since each book is sold by only one publisher. Although there
are some cost differences between hardbacks and paperbacks, it is difficult
to justify the price differences on a cost basis. Furthermore, paperbacks
are always sold when the demand for the hardback version has virtually
dried up. Between the hardback and the paperback stage, books are sold
through bookclubs as well. The Softback Preview bookclub is an English
company which specialises in high-quality paperback versions of original
hardback editions at around half the publisher’s hardback price. An
example of market skimming, where price cuts are not disguised by quality
differences is provided by Intel. Until Intel lost its monopoly, (when AMD
(Advanced Micro Devices) entered the market), its pricing strategy was
to keep new microprocessor prices very high and ease them down slowly
over time.
Sometimes intertemporal and third degree price discrimination are used
simultaneously. Tickets for the Kasparov-Short chess championship in
London, sponsored by the Times, were initially offered at up to £150.
When they did not prove popular and very few seats were sold, prices
were reduced down to £20 (intertemporal price discrimination)
and readers of the Sun (a London daily newspaper) were offered tickets
for £10 (third degree price discrimination). The frequently used
marketing practice of offering a ‘trade-in discount’ on goods such as
vacuum cleaners and other domestic appliances can be explained as an
attempt to price discriminate both intertemporally and between customers
who have bought before and those who have not.8 8
van Ackere and
Reyniers (1993, 1995)
If we assume that consumers are myopic, which means that they do not
think ahead and hence do not anticipate price changes, price skimming
is easy to model. The population of buyers and their willingness to pay
can be represented by a demand curve p(q), as in Figure 10.5. If the
monopolist can costlessly change the price he ends up with the entire
consumer surplus (above marginal cost c). He ‘skims’ the market by
selling at a high price to the most eager (or rich) consumers with very
high willingness to pay first and then gradually moves down the demand
curve selling to consumers with lower and lower reservation prices. In
reality of course durable goods prices do not change every hour. The
monopolist may have to keep the price constant for a given period of time,
for example, because the price has appeared in a printed advertisement or
a catalogue. In that case we would expect the price to jump from period
to period. As in Figure 10.5, during the first year, the price may be p1 and
q1 units are sold. At the beginning of the second year, since all consumers
with reservation price above p, have left the market, the monopolist faces
a new residual demand curve representing consumers still in the market
(line AB in Figure 10.5). He then picks a new (lower) price p2 and sells
q2 – q1 units at this price and so on. Clearly, a large fraction of the
consumer surplus can be captured in this way.

159
28 Managerial economics

A
p1 p(q)

p2

p3

c q
q1 q2 q3 B

Figure 10.5: Market skimming

Example 10.3

A monopolist seller of microprocessors has a horizon of two years. Any

microprocessor not sold after two years is considered obsolete and is
scrapped. The demand for this durable product is given by p(q) = 1 – q.
Consumers are myopic and buy a microprocessor at most once. The
production costs are negligible. Assuming the seller can set different
prices in year 1 and in year 2, what is his optimal pricing strategy?
As we know from Figure 10.5, the price set in the first period
determines second period demand. The easiest way to deal with this
analytically is to work out what the seller should do in period 2 when
he has set a price p1 in year 1. When we have worked out this problem,
we can solve the problem of setting p, optimally. In year 2, the
(residual) demand is p2(q) = p1 – q or q = p1 – p2 so that period 2 price is
determined by
max П2 = p2(p1 – p2 ).
The optimal second period price is half the first period price:
p2 = p1 /2 and П2 = p12/4.
In year 1, the monopolist who foresees what will happen in year 2, sets
the price to maximise total discounted profits:
max П1 + dП2 = p1(1 – p1 ) + dp12/4,
where  is the discount factor. The optimal first period price is thus
p1 = 1/(2 – /2). If the seller is patient and does not discount the second
year ( = 1), the price path is p1 = 2/3 and p2 =1/3. A more impatient
( < 1) seller sets higher prices.

So far we have discussed the price skimming problem assuming consumers

do not think ahead. If consumers are rational and forward-looking, the
problem becomes more complicated. A consumer, anticipating a lower
price in the future, may decide to postpone his purchase unless he is very
impatient. This has a moderating effect on the monopolist’s skimming
plans. However, as long as prices cannot change instantaneously, there
will be some intertemporal price discrimination with eager consumers
buying earlier and paying more than others. When prices can change
instantaneously, consumers anticipate a very rapid decline in price which
means that they will refuse to buy until the price is very low (equal to

160
 Chapter 10: Monopolistic pricing practices

marginal cost). This is the essence of the Coase conjecture: a durable

goods monopolist who can change price instantly loses all his monopoly
power when faced with rational consumers.
When consumers are rational, a seller who can commit to keeping the
price up may be better off than a seller who cannot make such a (credible)
commitment. It is not easy to make such a commitment and convince
consumers you will not renege. Suppose the monopolist charges the one
period monopoly price in the first period. Given that not everyone buys
at this price, there is a very tempting residual demand to be satisfied
in period 2. Generally, consumers will not believe that the seller will
be able to resist the temptation to lower the price later on. If, however,
a most-favoured customer clause is used so that the monopolist has to
compensate earlier buyers if he ever lowers the price, the commitment to
keep prices up may become credible. (Models of these phenomena involve
consumers forming rational expectations about future price movements
and are outside the scope of this course.9) 9
If you are interested
you could read more in
Tirole (1990) pp.79–87
Case: Yield management
Airlines are champions of intertemporal and third degree price
discrimination. It is relatively rare to find more than a handful
of passengers on a plane who have actually paid the same fare.
The number of tickets differentiated by a myriad of restrictions is
overwhelming. The Saturday night stopover requirement is an attempt
by airlines to steer business travellers away from cheap fares offered
to tourists. The price difference is usually so large that it makes sense
to buy two restricted tickets with a Saturday night stopover, one
starting when you want to leave and the other returning when you
want to return and just use the outgoing part of the first one and the
return part of the second one. In trying to squeeze as much revenue
as possible out of certain classes of passengers, airlines sometimes
create what seem absurd pricing patterns. It is not uncommon for a
return ticket to be cheaper than a one-way ticket, for example. Holiday
packages including a flight and self-catering accommodation are
sometimes cheaper than the flight only.
Due to computer reservation systems (CRS) it has become feasible to
change fares instantaneously in response to rivals’ price changes or
demand predictions. Travel agents use a CRS to check which flights are
available, at which fare and to make bookings. A CRS may be owned by
one airline (e.g. the Sabre network belongs to American Airlines) but
most are operated by groups of airlines. Amadeus – which is operated
jointly by Lufthansa, Air France and Iberia – also has links with
American and Asian airlines.
In addition to the CRS which is publicly accessible, airlines run their
own sophisticated programs aimed at maximising revenue from each
flight. A revenue management system (RMS) has as its objective to
sell each seat and to obtain the maximum fare possible. A RMS checks
several times a day on how forthcoming flights are filling up and
updates its predictions accordingly. If bookings are coming in quickly,
the number of low fares on offer is reduced; if it looks like the plane
may not be full, more cheap seats are offered. Some full fare seats
are always held back until close to take-off for late booking business
passengers or passengers who connect to other (long-haul and more

161
28 Managerial economics

profitable) flights. Airlines now even have flexibility with respect to the
proportion of seats allocated to first, business and economy classes. The
new Boeing 777 airliner uses quick changing bulkheads so that these
proportions can be changed rapidly.10 10
See ‘Learning to fly
all over again’; ‘By the
Although yield management is mainly associated with the airline seat of their pants’; ‘The
industry, hotels face very similar problems and use similar revenue high-tech war’
maximising techniques. Discounts for guests booking in advance,
differential pricing between weekdays and weekend nights, loyalty
bonuses and quantity discounts (e.g. 50 per cent off for a two-night
stay) are just a few of the marketing ploys characterising this business.
One of the nicest examples of third degree price discrimination must
be the deal offered by The Royal Orchid Sheraton Hotel in Bangkok. To
celebrate its 10th birthday in 1993, it offered a percentage discount on
the room rate rising according to the guest’s age.11 11
Webster (1993)

Peak-load pricing
For many goods and services, demand fluctuates with the seasons or time
of day so that the seller faces several demand curves. The demand at these
different times is satisfied by a common facility but typically different
prices are charged for peak and off-peak use. The demand for electricity
peaks in the morning and the evening and there is seasonal variation
because of heating and air conditioning. Telephone companies have lower
rates for long-distance calls in the evenings and at weekends. Health clubs
sometimes offer off-peak membership at a lower fee. Hotels are cheaper
off-season when they have spare capacity.
Charging different prices at different times of day is not really price
discrimination if costs vary with output because of overtime or use of
less efficient production methods when demand is high. Often marginal
cost is constant until a critical capacity is reached. Electricity generating
companies use technologically efficient methods for lower levels of
output but when these are exhausted they have to use older technologies
which could be ten times as expensive per unit of output (Kilowatt hour).
The same is true for airlines which charge high fares in times of peak
demand when older and less fuel-efficient planes have to be brought into
operation. It is not always easy to establish whether a price difference
can be justified by a cost difference. In many instances, demand in the
peak period is less elastic and therefore, even if costs were the same, a
monopolist would charge a higher price.
In its most general form, peak-load pricing is not easy to model. We
should really start out with a model of consumers deciding whether they
want to consume in the peak or off-peak period depending on the relative
prices. However, we will skip this step here and take the demand functions
which would result from such a process as given. Let us look at a simple
model of peak-load pricing in which the peak (Dp( p)) and off-peak (D0( p))
demands have the same elasticity so that the standard third degree price
discrimination argument would give identical prices. An example of such
a situation is when, for every possible price, peak demand is a multiple of
off-peak demand: Dp( p)= mD0( p), where m is a constant. Assume further
that the cost of providing the service, ignoring capacity costs, is C(q), the
same for both periods. The same capacity K is used in peak and off-peak
periods. The opportunity or investment cost per unit of capacity is constant
at c per unit. The profit maximisation problem is:
max П = pp(qp ) qp + p0 (q0)q0 – C(qp ) – C(q0) – cK (6)

162
 Chapter 10: Monopolistic pricing practices

subject to qp , q0 ≤ K, where qp and q0 are the units sold in peak and off-peak
periods respectively. There are two cases we need to consider. The first case
is when capacity is fully used in the peak period only (q0 < qp = K). Optimal
prices are derived by rewriting (6) as:
max П = pp(qp ) qp + p0 (q0 )q0 – C(qp ) – C(q0 ) – cqp
for which the first order conditions are:
MRp(qp) = MC(qp ) + c (7)
and
MR0(q0) = MC(q0).
The other case we need to consider is when capacity is fully used in both
periods (q0 = qp = q = K). This leads to rewriting (6) as:
max П = pp(q) q + p0(q) q – C(q) – C(q) – cq
which leads to the following result:
MRp (q) + MR0 (q) = 2MC(q) + c. (8)
In either case, the optimality conditions (7) and (8) determine quantity
sold in each period, which by substitution in the demand functions gives
the optimal prices. To determine whether it is best to set prices so that
capacity is only binding in the peak period or whether the seller should
attempt to equalise demand in both periods, profit has to be calculated for
both cases.

Commodity bundling
Sellers sometimes offer special deals when goods or services are bought
in packages. For example, it is often cheaper to buy a holiday package
which includes flight, accommodation and meals than to buy these items
separately. Some computer manufacturers (e.g. Gateway 2000 and Zeos)
sell bundled software as part of a package with a computer. PCs are often
sold in a package with peripherals. Many durable goods sellers offer
financing with their products. Luxury cars may be sold with leather seats,
sunroof, wooden dashboard, CD player and driver and passenger airbags
included at no extra cost. Again this marketing practice can be explained
in terms of price discrimination but, as it is rather special, it deserves a
special section.
Commodity bundling is sometimes used when the seller has to offer all
buyers the same deal because explicit price discrimination is impossible or
illegal. As for most forms of price discrimination, there may be cost-based
justifications for offering discounts when a bundle of products is bought.
For example, when a stereo system consisting of a specific tape deck, tuner
and CD player is sold as a bundle, the manufacturer and the retailer may
benefit from this standardisation and resulting economies of scale and
lower handling costs. Similarly, a restaurant which offers set meals reduces
waste of perishable items.
Commodity bundling, also called joint purchase discounts, was first
analysed by Adams and Yellen (1976) although the idea that bundling
can be used to extract consumer surplus first appeared in a paper by
Stigler (1963) who discusses the case of ‘block booking’ by the American
film industry. Monopoly producers used to (before it was ruled illegal)
bundle films when offering them to theaters. Theaters could not buy the
film they were interested in if they were only interested in one film; they
had to buy the bundle or nothing. This is an example of pure rather than
mixed bundling. Pure bundling refers to the situation in which the

163
28 Managerial economics

goods are only available as part of a package or bundle; they cannot be

bought separately. A newspaper may sell advertising space in the morning
paper only if an advertisement is also placed in the evening edition. Mail
order film developers include free films when prints are returned. Most
tape decks and cassette-radios have built-in speakers. Many professional
societies and charities sell journals or magazines to their members. Often
membership ‘includes’ subscription to these publications and members are
not allowed to opt out and just pay a membership fee. Mixed bundling
on the other hand allows for goods to be sold separately as well as in a
bundle. Theatre groups and concert organisers usually sell separate tickets
for individual performances as well as season tickets. Airlines sell one-way
as well as round-trip tickets. A restaurant may offer dishes which feature
in the day menu on the a la carte menu as well. An interesting marketing
practice which is equivalent to mixed bundling consists of offering
customers discount coupons for another product in the seller’s range.

To illustrate how bundling can be profitable, let us consider the pricing

problem of a cafeteria manager who sells apple pie and cappuccino.
He figures that 50 per cent of his customers are mainly thirsty and the
other 50 per cent are mainly hungry. A thirsty customer is willing to pay
up to £2 for a cappuccino and up to £2 for a slice of apple pie. A hungry
customer is willing to pay up to £1 for a cappuccino and up to £3 for
the apple pie. For simplicity we will ignore costs. (You should, after
reading this, be able to repeat the analysis taking costs into account.)
If the manager aims to sell apple pie to both groups of customers, he
cannot charge more than £2 per slice. This is of course better than just
aiming to sell to hungry customers since that would result in expected
revenue of £1.50 per customer. Similarly, if the manager wants everyone
to buy cappuccino, he should charge £1, which gives him the same
revenue as selling to thirsty customers only at a price of £2. Using this
pricing strategy, the manager gets an (expected) revenue of £3 per
customer. Now let’s see how bundling can make him better off. For each
customer, the sum of reservation prices for cappuccino and apple pie
is £4. This means that, if the two items are offered in a bundle at £4,
revenue per customer increases from £3 to £4.

The groups of customers in this example have negatively correlated

reservation prices for the two items. This feature makes the bundling very
profitable since the alternatives of either selling to the high willingness to
pay group only, for each item, or charging low prices so everyone buys, are
not attractive.
It is however not necessary for reservation prices to be negatively
correlated for bundling to be profitable. Schmalensee (1984) points out
that pure bundling reduces buyer diversity in the sense that the standard
deviation of reservation prices for the bundle is. always smaller than
the sum of the standard deviations of reservation prices for the goods
which compose the bundle. Let σ1 and σ2 be the standard deviations of
reservation prices r1 and r2 for potential consumers of goods 1 and 2. The
correlation between the reservation prices is ρ. If consumers’ reservation
price for the bundle is the sum of the reservation prices for the product
(i.e. rb = r1 + r2) then
Var(rb ) = σ12 + σ22 + 2Cov(r1, r2 ) = σ12 + 2ρ σ1 σ2 < (σ1 + σ2 )2.
The lower the correlation between reservation prices, the greater the
reduction in heterogeneity achieved by bundling. Reduced diversity allows
for more efficient extraction of consumer surplus and hence increased
profits.
164
 Chapter 10: Monopolistic pricing practices

Figure 10.6 illustrates pure and mixed bundling for consumers with
independent demands for two goods. Each consumer is represented by a
point in the square OABC, marking the pair of reservation prices for both
goods. Consumers buy at most one unit of each product and a consumer’s
reservation price for the bundle equals the sum of his reservation prices for
the two goods. Consumers cannot resell goods. If goods are sold separately
(unbundled sales) then all consumers with reservation prices above the
set prices p1 and p2 buy Product 1 and Product 2 respectively (see Figure
10.6a).

C
r B
2
Buy 2 Buy
only both

p
2
Buy Buy
neither 1 only
r
0 1
p A
1
Figure 10.6a: Unbundled sales
If the seller only offers the goods as a bundle, then only consumers whose
reservation price for the bundle exceeds the bundle price (r1 + r2 ≥ pb ) will
buy (see Figure 10.6b).

r
2
Buy
bundle
p
b
Don’t
buy
r
p 1
b

Figure 10.6b: Pure bundling

In the case of mixed bundling, the seller offers the individual components
of the bundle as well as the bundle. Obviously the individual products
are priced such that their prices add up to more than the bundle price,
otherwise nobody would ever buy the bundle.
Now consumers have four options: (a) buy nothing, (b) buy Product I
only, (c) buy Product 2 only and (d) buy the bundle. They will make their
choice according to which option gives them the highest consumer surplus,
that is, according to max (0, r1 – p1, r2 – p2, r1 + r2 – pb ). On Figure 10.6c, the
consumers who buy nothing are in area OABCD and the consumers who
prefer option (b) are in area DCEF since there:
165
28 Managerial economics

r1 > p1, r2 < r1 + p2 – p1 and r2 < pb – p1

Similarly, consumers in area AHGB prefer option (c) since there:
r2 > p2 , r2 > r1 + p2 – p1 and r1 < pb – p2.
This leaves consumers in area GBCE to buy the bundle.

p - p
b 2
r
2 G r +p - p
H 1 2 1
p
b
B
p A
2

C E p -p1
D E b
0 p r
p 1
1 b

Figure 10.6c: Mixed bundling

Commodity bundling is a current research topic. Different assumptions
regarding consumers’ reservation prices can be made. It may not always
be reasonable to assume, as we have done here, that a consumer’s
reservation price for the bundle equals the sum of the reservation prices
for individual items. Also, the effect of different market structures has not
been conclusively studied. If the seller is an oligopolist for example he may
bundle products to gain an advantage over his competitors. When a good
for which the seller is a monopolist is bundled with goods which are sold
competitively the monopolist may be trying to extend his monopoly power.
In a current lawsuit Microsoft is accused of doing just that. By bundling
software packages with the operating system it makes PC buyers who get
this software installed ‘for free’ reluctant to buy a package from one of
Microsoft’s rivals. Several firms which used to sell programs to do tasks
now performed by Windows, Microsoft’s operating system, have gone out
of business.12 12
See ‘Hard sell’; ‘A hell
of an operating system’
Sometimes different sellers offer components of the bundle. In an
interesting application of commodity bundling, Lan and Kanafani (1993)
discuss park-and-shop discounts. Parking facilities in city centres and
commodities people shop for in these city centres, are owned by different
firms. Nevertheless the authors quote examples of successful bundling
strategies through which shoppers get a discount on car parking. The case
of pure bundling in this context is also quite common. It occurs whenever
a shop or supermarket parking lot is only accessible to customers (who
may have to show a receipt). Schemes which give discounts to shoppers
who use public transport are an alternative form of commodity bundling
which may be preferable if congestion and pollution in the city centre is a
problem. In this spirit, London Underground’s one day Travelcards have
been used to offer discounts on museum entrance fees and films.
The decision whether to market a product on its own or in combination
with another product can be of crucial importance. Commercial Brake
is the name of a new gadget which, when used with a video recorder,
automatically skips over the commercial breaks. This invention is for sale
for $199 and one million units are expected to be sold during its first

166
 Chapter 10: Monopolistic pricing practices

year. Total yearly VCR sales however are about 50 million. Arthur D. Little
Enterprises (ADLE), the consultancy marketing the new technology, not
surprisingly, tries to convince VCR manufacturers to sell their products
with Commercial Brake fitted.13 13
See ‘Ad nauseum’

The term tied sales is often used to describe a marketing practice which
is related to commodity bundling. Whereas in commodity bundling the
proportions of the items in the bundle are fixed (e.g. one bar of soap and
one towel), under tied sales, the consumer can decide on the quantities
in the bundle. Under tied sales, a manufacturer refuses to supply a
product for which he has a monopoly unless the consumer agrees to
also buy some (complementary) product from the manufacturer. When
this complementary product is available in a competitive market, tying
sales can be seen as an attempt to extend monopoly power. Usually there
is another reason to tie sales, namely to discriminate between types of
customers. As part of their standard leasing arrangement for photocopiers,
Xerox used to insist that customers bought Xerox paper. This allowed
it to price discriminate between high intensity and low intensity users
by charging a high paper price. If higher use of paper indicates a higher
willingness to pay for the photocopier lease then customers reveal their
type through the amount of paper they use. Tied sales is a mechanism to
extract more consumer surplus from the keener users. Of course Xerox
could have negotiated different rentals with low and high intensity users
but this leaves the problem of identifying which class any user belongs to.
Manufacturers who supply replacement complementary goods to the
durable good they sell may be tempted to design their products so
that only products in their range are compatible. This constitutes a
technological tie. For example, a printer manufacturer could design
his products so that only cartridges from the same make can be used.
Camera manufacturers design lenses and camera bodies so that they have
limited compatibility - usually only within the brand name - and often
there is incompatibility between successive generations of products. In
the US there have been several lawsuits concerning technological ties and
designed incompatibility.

Multiproduct firms
In mainstream microeconomic models the firm is usually modelled as a
single product manufacturer. In our discussion of production and pricing
we have focused on the case of a firm producing a single output. This is
done for convenience of course as we know that practically all firms are
multiproduct firms. In fact some of the most interesting pricing questions
arise when firms have to take into account that their pricing policy for
one of their products has significant implications on the demand and
profitability of another. Generally, when pricing and marketing decisions
for individual products are not coordinated the firm does not perform
as well as it could. Particularly, a company structure’ in which separate
divisions are given responsibility for the profitability of an individual
product can be undesirable.

167
28 Managerial economics

Example 10.4

Noindent, a company which markets game computers and videogames,

has set up two divisions. The hardware division has responsibility for
pricing the game machines whereas the software division is in charge
of pricing the game cartridges. The demand functions for hardware and
software respectively are as follows:
ph = 1000 – 30qh + 2.5qs
ps = 2.5qh + 500 – 20 qs
where ph and ps are the prices and qh and qs the quantities (in thousands)
demanded per week at these prices. Clearly, the two products are
complements and increases in the demand for one have a positive effect
on the demand for the other. Game computers and video game cartridges
are produced at constant marginal cost of £100 and £50 respectively.
The hardware division of Noindent maximises its profits, taking the
software division’s pricing as given. In game theory terms, we will be
looking for a Nash equilibrium. Profit from selling game computers is:
Пh = (900 – 30qh + 2.5qs ) qh
which is maximised (set the derivative with respect to qh zero) for:
qh = 15 + qs /24. (9)
Substituting this expression for qh in the demand function results in:
ph = 550 + 1.25qs (10)
We can do a similar exercise for the software division with profits:
Пs = (2.5 qh + 450 – 20qs )qs
maximised at:
qs = 11.25 + qh /16. (11)
Substitution in the demand function gives:
ps = 275 + 1.25 qh (12)
Solving (9) and (11) for qh and qs gives qh = 15.509 and qs = 12.219 so that
the prices set by the divisions are determined by (10) and (12) as ph =
565.27 and ps = 294.39. The maximum profit when divisions act separately
can be calculated for these results as Пsep= 10,202.
Suppose Noindent restructures so that hardware and software pricing
decisions are made within one division or profit centre. Such a division
has incentives to take demand interactions into account by maximising
overall profit:
П = (900 – 30qh + 2.5qs )qh + (2.5qh + 450 – 20qs )qs
Setting derivatives with respect to qh and qs zero and solving for qh
and qs gives qh = 16.105 and qs = 13.263 so that the profit maximising
prices are determined by (10) and (12) as ph = 550 and ps = 275. The
maximum profit when the software and hardware pricing decisions are
made together is Пlog = 10,232. Prices are set lower when the demand
complementarity, between products is taken into account. When divisions
only care about their own profitability, they tend to set prices too high
because they ignore the externality effect of their pricing decision on the
other division’s profitability.

Firms selling complementary goods often price one product at or below

marginal cost in order to stimulate demand (at high prices) for the other
products. The cheap products are loss leaders aimed at attracting

168
 Chapter 10: Monopolistic pricing practices

customers for more profitable goods or services. For example, engine

manufacturers may sell engines rather cheaply and charge a lot for
service and spare parts. Amstrad wordprocessors were priced low to fuel
the market for peripherals and software. Airlines sometimes take a loss
on certain routes because some passengers on these routes connect to
profitable flights. In extreme cases, companies offer gifts of free products
to build market share for other products.
An important issue for many multiproduct firms is that of
cannibalisation. When firms add new products or new varieties of
existing products to their range, they may hurt their own sales. Whenever
Intel introduces a new generation of microprocessors, it competes against
its own previous generation. Firms are sometimes multiproduct firms
not by choice but because of technological reasons. This is the case when
they create valuable byproducts in the course of manufacturing some
other product. There are many examples of such joint products in
the agriculture and chemical industries. When raising cattle for beef, the
farmer also produces leather. Shell creates propylene basic ingredient in
polypropylene, as a byproduct when it makes ethylene.14 In contrast to 14
See ‘Better than
the other situations described in this section, joint products are related price-fixing?’
in production rather than consumption. The joint product scenario forms
an extreme case of economies of scope which refers to a production
technology whereby it is cheaper to produce two products together than
separately. For example, it is relatively cheap for a company laying cables
for cable television to add a telephone connection.
Let us focus on a monopolist manufacturing and marketing joint products.
Each ‘unit’ consists of a unit of Product A and a unit of Product B. If you
find this too abstract, think of the ‘unit’ as an egg: Product A is eggyolks
(which could be used by a mayonnaise manufacturer) and Product B is
eggwhites (sold to a manufacturer of low cholesterol egg substitutes). If
we know the cost function of the ‘composite good’ (eggs) and the demand
functions of Product A and Product B, how much should be produced?
Note that the quantity produced of A always equals that of B. Intuitively
we can deduce that production should be increased until the marginal cost
exceeds the sum of the marginal revenues for A and B. To derive this rule
mathematically, write the profit function as:
П = pA(qA )qA + pB(qB )qB – C(q ), q = max (qA> qB ) (13)
where qA and qB are the quantities of A and B sold and C(q) is the cost
of producing q units of A and of B. If we assume that everything which
is produced will be sold (see below for reasons why this is not always
optimal), then we can set q = qA = qB. Optimising (13) then gives the
following result:
MRA(q) + MRB(q) = MC(q) (14)
When the seller is a monopolist an interesting problem can arise. A
monopolist’s marginal revenue decreases in output and may tend to
zero for large enough output levels. It is thus possible that, while MR for
Product A is still positive and above marginal cost, MR for Product B is
negative. Of course a firm never sells at negative MR since it can avoid this
by selling up to zero MR and wasting the excess production. We will see
exactly how this works in an example but, before tackling the example,
let’s think about why this type of waste does not occur for joint products in
a perfectly competitive market. In such markets, MR equals price which is
positive (as long as the good is a ‘good’ and not a ‘bad’!) and a competitive
firm does not need to worry about selling so much that it drives marginal
revenue down. Of course there may be waste in a competitive market if

169
28 Managerial economics

the price net of transportation costs is not high enough. Gas which
is released when drilling for oil is often burnt because its transportation
cost from remote oil fields is too high. In the Falkland Islands where
sheep are farmed for wool, the meat goes to waste presumably because
of the high transportation costs. This state of affairs is not helped by
British regulations requiring that all meat for the British forces is imported
including ‘mutton granules’.15 15
See ‘The richest
islands in the world,
Example 10.5 maybe’

A pineapple grower in the Bahamas produces cans of sliced pineapples

and cans of pineapple juice. The demand functions for cans of
pineapples and cans of juice are respectively:
qp = 80 – 5 pp
and
qj = 50 – 5 pj.
At a cost of C(q) = 25 + 8q + 0.05q2, the production process produces
q cans of pineapple slices and q cans of pineapple juice. Applying rule
(14) to this problem gives:
(80 – 2q)/5 + (50 – 2q)75 = 8 + 0.1q
so that q = 20 pineapple cans and juice cans should be produced. Figure
10.7 shows how to find the optimal quantity graphically by adding the
marginal revenue curves vertically and identifying the intersection of
this sum with marginal cost. Note that we assume that the seller does
not sell at negative marginal revenue and hence for q > 25 the sum
of marginal revenues coincides with the marginal revenue curve for
pineapple cans. This intersection is at 20 units and the prices can be
read off the demand curves for q = 20 as pp = 12 and pj = 6. At 20 units
of output, marginal revenue is positive for both products.

26
∑ MR

16 MC

10
8 MC ′

D D
j p
25 40 q
50 80
Figure 10.7: Joint products
Now suppose the cost function is C(q) = 25 + q2/35 and there is free
disposal (i.e. it is costless to discard excess pineapple juice). Note in
Figure 10.8 that marginal cost (MC' ) now intersects the sum of the
marginal revenue curves where marginal revenue of juice would be
negative if all of the juice produced is sold. The profit maximising
producer should set MRp(qp) = MC'(qp) or (80 – 2q)/5 = 2q/35 which
implies production of qA = 35 cans of pineapple. Rather than selling 35
cans of pineapple juice, the seller should restrict sale of juice to the
level where marginal revenue is zero (i.e. qj = 25 cans).

170
 Chapter 10: Monopolistic pricing practices

Transfer pricing
A transfer price is the price which is charged when one division of
a vertically integrated firm (say the production division) sells an
intermediate good or service to another division (say the marketing
division). When such internal sales take place, maybe in addition to sales
of the intermediate good on the external market, the determination of
the transfer price is important for the firm’s profitability. At first sight
this statement may puzzle you: why should it matter what price GM’s
parts division charges its assembly plants? By definition, the payment
for parts is a transfer; it is like putting money from the left pocket into
the right pocket. The reason transfer prices are important is that they are
used to provide proper incentives for individual divisions to take actions
which maximise total firm profit. Many large-scale enterprises have a
decentralised structure consisting of semi-autonomous profit centres.
Such stand-alone divisions are often established to avoid excessive
communication and coordination costs in large firms. Since in this
scenario divisions are rewarded on the basis of their performance, the
selling and buying divisions have opposing interests in the determination
of the transfer price. The selling division has an interest in setting the
transfer price high, whereas the buying division prefers a low transfer
price. If transfer prices are not set properly or if the transfer price is left to
negotiation between divisions, total firm profit suffers.
Following Hirshleifer’s (1956) seminal paper on transfer pricing we will
develop the analysis for three scenarios distinguished by whether there
is an outside or external market for the intermediate good and if there
is, whether this market is competitive or not. To simplify the analysis
we assume that units are defined so that one unit of the final good
requires one unit of intermediate good. We also assume that there are
no technological synergies between the production and the marketing
division (i.e. the cost function for either division is not affected by the
output of the other).

Scenario 1: no external market for the intermediate good

The simplest scenario is that of a vertically integrated firm which does not
sell or buy the intermediate good on the external market. There may be
several reasons for this arrangement. An important consideration is that of
transaction costs the firm incurs when it deals with the external market.
Hirshleifer gives the example of a steel mill consisting of two divisions
exchanging molten iron. Buying and selling of molten iron on the external
market is prohibitively expensive. It could be an important strategic
decision not to have a market for the intermediate good. American
Airlines, for example, has blacklisted some of its information systems such
as flight-scheduling and yield-management programmes for sale to its
domestic competitors whereas other services such as data punching into
computers are sold on the external market.16 IBM’s bad fortune in the early 16
See ‘Managing the
nineties has been blamed on its decision to buy the essential components future’.
for its personal computers from the outside market. Instead of using its
own microprocessor chips and software, it bought chips from Intel and
operating system software from Microsoft. This allowed standardisation in 17
See ‘What went
the PC market and made IBM machines very easy to clone.17 wrong at IBM’
In what follows we use subscripts f, i and e to denote variables related
to the final good, the intermediate good and the external market for the
intermediate good. Since, in this first scenario, the firm does not buy or
sell the intermediate good externally, the final good quantity qf equals the

171
28 Managerial economics

quantity of the intermediate good produced. Given the demand function

pf (qf ) this quantity qf is the only decision variable. The integrated firm’s
problem is simply to maximise revenue on the final good market minus the
costs incurred by the two divisions:
max pf (qf )qf – Ci (qf ) – Cf(qf ).
For profit maximisation the firm should produce a quantity qf such that
the marginal revenue equals total marginal cost:
MRf(qf ) = MCi(qf ) + MCf(qf ).
This result is illustrated in Figure 10.8 where ∑MC is the vertical
summation of the marginal cost curves. The intersection of ∑MC and MR,
the marginal revenue for the final good, determines optimal output. Of
course this analysis also applies when the firm sells the final good in a
perfectly competitive market. The only difference there is that marginal
revenue is horizontal at the market price.

p, MC

∑ MC

p
f MCi

MCf
p
t

D
MR

q
f q

Figure 10.8: No external market

Given this determination of the optimal output level, the firm has to find
a way of implementing a suitable transfer pricing scheme. If the marginal
cost curves are known, the transfer price should be set at pt in Figure 10.8,
such that pt = MCi(qf ). With this transfer price rule in place, the firm can
then instruct both divisions to maximise profit subject to this restriction
and this will lead to the desired result: the manufacturing division
produces the output qf for which marginal cost equals ‘marginal revenue’
pt and the marketing division wants to use exactly what is supplied by the
manufacturing division since at output qf its marginal cost, pt + MCf(qf ),
equals marginal revenue.
However, we have really only solved the problem for an ideal world where
the cost functions are known and where managers do not try to negotiate
‘a better deal’ for their division. In practice, when the headquarters does
not know divisions’ marginal cost functions, the divisions may not find it in
their interest to volunteer this information or to give accurate information
about costs. Especially when conditions are such that one of the divisions
is making very low profits, managers will try to bargain to increase or
decrease the transfer price. For example, when the manufacturing division
has constant marginal costs, its profits are zero under the optimal transfer
price rule. If costs are known it may be better in terms of providing
incentives to make such a division a cost centre with rewards tied to cost
reductions rather than reward it on the basis of profitability.

172
 Chapter 10: Monopolistic pricing practices

Scenario 2: perfectly competitive external market for the

intermediate good
In this scenario the output of the production division need not be equal
to the output of the final product. If there is excess demand the firm buys
on the external market and if there is excess supply it sells on the external
market. In addition to the assumptions of scenario 1, we assume demand
independence between the production and marketing division (i.e. sale
of an extra unit by either division does not affect demand for the other).
This is a strong assumption since generally when the firm sells more
of the intermediate good we would expect that less of the final good is
demanded and vice versa. This is obviously the case when the firm sells
the intermediate product to its competitors in the final goods market.
Therefore demand independence is only realistic if the intermediate good
is sold to an industry which produces an imperfect substitute or better,
a product unrelated to the final good. Hirschleifer quotes the example
of a copper concern selling copper to its wire manufacturing division
and on the external market to firms which use copper as an input in the
production of products other than wire, such as pots and pans.
It is useful to consider the situation in which the firm sells to the external
market first. The firm now has to make two decisions: how much of the
intermediate good and of the final good to manufacture. It maximises
revenue from the final good and from the intermediate good minus costs:
max pf (qf ) + peqe – Ci(qf + qe) – Cf (qf ).
Since the intermediate good market is perfectly competitive, the firm takes
the price on that market pe as given. For profit maximisation we thus have:
MRf (qf ) = MCi (qf + qe) + MCf (qf )
and
pe = MCi (qf + qe).

p, MC

MCf + p
p e
f MCi

p =p
t e

MCf

D
MR
q q
f i q

Figure 10.9: Competitive external market

Figure 10.9 illustrates these results graphically. To determine the amount
of final good to produce, the firm sets marginal revenue equal to the
sum of marginal costs at the optimum output levels for final and
intermediate goods. The optimal output level for the intermediate good qi
= qf + qe is determined where its marginal cost MCi equals the market price
pe The quantity of the final good produced is then found at the intersection
173
28 Managerial economics

of MCfc + pe and MR. The external market price represents the opportunity
cost of using a unit of the intermediate good internally. Of course the same
analysis applies when the final good is sold in a competitive market so that
its MR curve is horizontal at the market price.
If the firm is a net buyer of the intermediate good, its optimisation
problem becomes:
max pf (qf )qf – pe qe – Ci (qf – qe ) – Cf (qf )
which results in
MRf (qf) = MCi(qf – qe) + MCf (qf )
and
pe = MCi (qf – qe ).
This situation could be illustrated graphically as in Figure 10.9 but with
MCi and pe intersecting to the left of the optimal output level for the final
good.
When there is an external market, profit maximisation generally calls
for the vertically integrated firm to use it. The optimal quantities of
intermediate and final good produced are not equal. As a consequence,
when the divisions are prevented from using the external market,
profitability suffers. For example, a cement producer cannot increase
profitability when he buys a ready-mix-concrete firm with the sole
intention of using it as a captive customer.
The proper transfer price when there is a competitive external market
is the external market price. This will induce both divisions to produce
the optimal quantities if they are instructed to maximise profit. This
transfer price rule is also intuitively appealing to managers of the separate
divisions. The manufacturing division would not want to sell below the
market price and forego a more profitable outside sale and the marketing
division does not want to pay more than the market price. Headquarters,
when setting the transfer price in this scenario, does not need any cost
information. Because of its simplicity the rule of equating transfer price to
external market price is used almost without exception in practice.
Note that the marketing division does not benefit from being vertically
integrated with the manufacturing division. It could buy the intermediate
good under the same conditions on the external market. There is no
rationale for the vertically integrated structure under this scenario. We
obtain this rather surprising result because of the assumption of demand
independence. In the more realistic setting, when an increase in the
quantity of the intermediate good sold on the outside market has a
depressing effect on the demand for the final good, we would expect that
the firm sets the transfer price below the external market price. On the
other hand, when demand dependence and technological dependence
are negligible, there is a case to be made for ‘outsourcing’ or spinning off
a manufacturing division which operates in a competitive market. This
process has been taking place in the computer industry which used to
have a multi-layered vertical structure. Most PC makers in fact restrict
themselves to little more than contracting with electronics suppliers,
assembling and marketing their machines. Several of these electronics
manufacturers arose from restructuring in computer companies which
spun off some of their production capacity.18 18
See ‘Labouring in
obscurity’
The issues of make-or-buy decisions and demerger are however very
complex and we cannot hope that a simple transfer price model like
the one outlined in this section will give us all the answers. Arguments

174
 Chapter 10: Monopolistic pricing practices

in favour of outsourcing have to be balanced with the firm’s product

differentiation strategy for example. Consumer perceptions and hence
valuations of a product may be determined to a larger or lesser extent by
what is outsourced. Consumers may not care whether a part in a domestic
appliance engine is made in-house. They may, however, due to Intel’s
advertising campaigns, care about which microprocessor is built into their PC.

Scenario 3: imperfectly competitive external market for the inter-

mediate good
The last scenario we consider is the one in which the firm has monopoly
power in both of its markets. It sells the intermediate good internally at
a transfer price and externally according to a downward sloping demand
curve pe (qe ). Hence its profit maximisation problem is:
max pf (qf )qf + pe (qe )qe – Ci(qf +qe) – Cf (q),
which results in the following optimality conditions:
MRf (qf ) = MCi (qf + qe ) + MCf (qf ) (18)
and
MRe (qe ) = MCi (qf + qe). (19)
These conditions are similar to the conditions for optimal third degree
price discrimination. The manufacturing division sets its marginal cost
MCi(qf + qe ) equal to marginal revenue in the external market MRe (qe )
and to net marginal revenue in the final good market MRf (qf ) – MCf (qf ).
Figure 10.10 shows graphically what is going on. The left panel represents
conditions in the external market for the intermediate good. The middle
panel shows the derivation of the net marginal revenue curve NMR as
the vertical distance between MRf and MCf . In the panel on the right the
manufacturing division sets its output level where marginal cost intersects
∑MR, the horizontal sum of the NMR and MRe curves. Again, the
analysis for a firm which sells its final product in a competitive market is
very similar.

p
e
∑ MR

De
p* Df
e
p* MCi
f
MCf

MRe NMR MRf

q* q* q q
e q f f q* i
e i

Figure 10.10: Imperfectly competitive external market

The optimal transfer price rule consists of selling internally at a price equal
to marginal revenue on the outside market. Because price on the outside
market exceeds marginal revenue, the internal division gets a discount
(i.e. the intermediate good is sold for less internally). The intuitive reason 19
see Chapter 12,
for this is the double marginalisation problem19 which would occur if the section on successive
monopoly
manufacturing division used its monopoly power internally.
175
28 Managerial economics

Example 10.6

Siemips, a major producer of microprocessor chips for telecoms

equipment, sells these chips on the outside market as well as internally.
It has monopoly power in the chips market as is reflected in the
downward sloping external demand pe =1200 – 0.8 qe. The demand for
Siemips’ final good is pf = 1200 – 0.625qf . Each unit of the final good
requires one microprocessor chip. Since Siemips has a policy of only
supplying chips to manufacturers it is not in competition with, there
is demand independence between the chips and equipment divisions.
The equipment division has marginal cost MCf (qf ) = 20 + 0.025qf and
the chips division has marginal cost MCi(qi) = 200 + 0.375qi = 200 + 0.375
(qf+ qe).
To determine the optimal output level on the final good market, we use
(18) and set marginal revenue equal to the sum of marginal costs:
1200 – 1.25qf = 200 + 0.375(qf + qe ) + 20 + 0.025qf . (20)
Siemips acts as a monopolist on the external market and sets marginal
revenue equal to marginal cost as in (19):
1200 – 1.6qe = 200 + 0.375(qf + qe ). (21)
Rewriting (20) and (21) gives
qf = 593.93 – 0.227 qe and (22)
qf = 2666.67 – 5.267 qe. (23)
Equating these two and solving for qe results in qe = 411.27 and hence
from the demand function pe = 871. Substituting into (22) or (23) gives
us the optimal output level for equipment qf = 500.6. The transfer price
is set equal to marginal revenue in the external market 1200 – 1.6 qe =
541.95 which means that the internal division gets a discount of about
330 on a chip.

In all three scenarios we have considered, the transfer price should be set
equal to marginal cost. In scenario 1, marginal cost corresponding to the
optimal output level of the final good is relevant; in scenario 2, the market
price is used but this is equal to marginal cost when the manufacturing
division maximises profits; and in scenario 3, marginal cost corresponding
to total production (for internal and external use) is the valid price. In
practice, because it is convenient to set transfer price equal to market
price, mistakes are made when market conditions are as in scenario 3.
More complicated scenarios can be modelled for firms in different market
structures and where the assumptions of demand independence and
technological independence do not hold. This is however beyond the scope
of this subject guide.

Case: Transfer pricing and taxation

Multinational enterprises manipulate their transfer pricing systems in
order to redistribute profits between countries so as to minimise their
overall tax liability. If country A has lower profit tax than country B
then the transfer price will be set low in B so that profits are realised
in A. This is illegal but it is very difficult to monitor especially if there
is no external market. If there is an external market, the price on the
external market can in some cases be used as an estimate of the proper
transfer price. In the early nineties the US International Revenue
Service investigated American subsidiaries of foreign firms, especially

176
 Chapter 10: Monopolistic pricing practices

Japanese ones, on the suspicion that they had underpaid US corporate

income taxes by as much as $12 billion. Of the nearly 37,000 foreign-
owned companies filing returns in 1986, more than half reported no
taxable income.
Most countries tax multinationals based on the profits that the firm
earned within their borders as if it had made those profits as a stand-
alone business at arm’s length from the parent company (the arm’s
length system). Firms are expected to use proper transfer prices
which, in many cases, means the prices they would pay if they had to
buy the internally transferred goods and services from outside. This
is of course problematic when there is no outside market. For many
intangibles there is no comparable external market. The advantage of
the arm’s length system is that profits are normally taxed only once.
California has operated a unitary tax where the tax base is simply a
slice of the worldwide profits of the multinational determined by the
proportion of the firm’s worldwide workforce, property and sales sited
within the taxing country (or state in this case). The state of California
which is strapped for cash can guarantee in this way that foreign
companies with Californian subsidiaries pay tax irrespective of whether
the subsidiaries (claim to) make a profit. The advantages of a unitary
tax system are that the rules for taxing are clear, simple and not easily
manipulated but double taxation of profits is likely. Barclays Bank
was involved in a 17-year legal battle with California claiming that
its profits were subject to double taxation and challenging the state’s
right to use the unitary tax system. Since Barclays first took the state
to court, California has given up unitary taxation. From 1988 firms
which are willing to pay a fee and go through the paperwork required
to get an exemption can do so and in October 1993 the arm’s length
system was adopted. Barclays did not drop the case and demanded a
refund of taxes paid under the unitary system. On 20 June 1994, the
‘tax war’ ended with the Supreme Court ruling in favour of California.
If California had lost, it would have had to refund $1.5 billion in taxes
to Barclays and other multinational firms.20 20
See ‘Taxing questions’;
‘Tax deficient’; ‘Unhappy
returns for Barclays’;
A reminder of your learning outcomes ‘Unhappy returns’; ‘IOU
all over again?’
Having completed this chapter, and the Essential reading and activities,
you should be able to:
• give (your own) examples of price discrimination and commodity
bundling
• derive the optimal take-it-or-leave-it-offer in first degree price
discrimination
• derive the optimal two-part pricing scheme
• derive optimal prices and quantities in third degree price discrimination
(analytically and graphically) and determine the optimal price and
quantity when the monopolist faces two demand curves and is not
allowed to price discriminate
• solve a simple two period skimming model with myopic consumers
• solve a simple peak-load pricing problem
• show graphically how consumers with different reservation prices
distribute themselves over the various options in commodity
bundling

177
28 Managerial economics

• for a multiproduct firm derive prices and quantities for two

products when
a. each product is handled by a separate division and divisions act
noncooperatively and
b. when the decisions are taken to maximise firm profit
• solve a joint products problem
• derive the optimal transfer price (analytically, graphically and
numerically).

Sample exercises
1. This is an extract from ‘Serving your needs. A guide to telephone
services in your home 1993–94.’ published by British Telecom:
‘We work out the cost of each call individually, and measure the call in whole
units. Each unit buys a period of time. (The basic unit currently costs 4.935 p
including VAT.) The length of time given for each unit depends on when and
where you are phoning from, and where your call is to. (The unit rate may
vary depending on how many calls you make... We have introduced three
`Customer Options’ which give discounts on the basic unit rate depending on
the number of calls you make:
Standard Personal: A 5% discount is automatically given on all direct
dialled calls you make over £58.75 in one quarter. This discount increases to
8% on calls made over £293.75 in a quarter. (A quarter is an average of 91
days.)
Option 15: For a £4 quarterly charge you can apply for our new high-value
scheme which offers a 10% discount on all direct-dialled calls at the basic unit
rate. You will benefit from this if your call charges are consistently more than
£40 per quarter including VAT.
Supportline: If you make very few calls (less than 125 units per quarter) you
can get: half-price standard rental and the first 30 units free, On the next 120
units per quarter, you pay a rate which is higher than the standard unit rate.
After you have used 150 units the price falls to the standard unit rate.’
a. Draw budgetlines corresponding to each of the three ‘Customer
Options’. The x-axis is the number of units per quarter and the
y-axis is remaining income. Remember the rental.
b. If my current phone bill is less than £40 per quarter, it does not
make sense for me to go for Option 15. True or false?
c. Why does BT offer these options?
2. In 1979 the Belgian government forced BMW to charge a low price for
cars sold in Belgium (price ceiling). BMW attempted to ban its dealers
from exporting the cheaper cars, offered for sale in Belgium, to other
countries. The European Commission condemned these bans. Show
the possible effect of the Commission’s decision by graphing BMW’s
price-discriminating strategy (selling at price ceiling in Belgium and at
higher price elsewhere) and graphing the non-discriminating scenario.
Show that BMW might decide to quit the Belgian market altogether or
alternatively, charge the Belgium controlled price throughout the EEC.
3. Off-peak demand for a service is given by q1 = 100 – 2p1 and peak
demand is q2 = 300 – p2. Marginal cost per unit is c = 10 and capacity
cost g = 90 per unit. Find the optimal peak and off-peak price.
4. Suppose Gillette knows it has two types of customers: about 50% are
of Type 1 and 50% are of Type 2. A Type 1 customer is willing to pay
up to £1 for a razor and up to £1.50 for a package of 10 razorblades. A

178
 Chapter 10: Monopolistic pricing practices

Type 2 customer is willing to pay up to £0.80 for a razor and up to

£1.70 for a package of 10 razorblades. The marginal cost of razors is c1
per unit and a package of 10 razorblades costs c2 to produce.
a. Given that Gillette has to charge the same prices to all customers,
what is the best pricing policy? How does it depend on c1 and c2?
b. What is the best pricing policy if Gillette sells its products as a
bundle consisting of a razor and 10 blades, and the individual
products are not on sale?
c. If it uses mixed bundling how should it price the bundle and the
individual products?
5. Due to increased awareness of the dangers associated with a high-
fat diet, the demand for skimmed milk has increased. Assuming milk
products are sold through a monopoly distributor (a milk marketing
board), what is the effect of this shift in demand on the price of cream?
6. The Croucher Corporation which is in the woolly hat industry is
composed of a marketing division and a production division. The
marginal cost of producing a woolly hat is £10 per unit and the
marginal cost of marketing it is £4 per unit. The seasonal demand
for woolly hats is p = 100 – 0.01q. There is no external market for the
woolly hats Croucher produces (they look rather odd and are basically
worthless unless properly marketed which requires special skills only
Croucher’s own marketing division can provide).
a. How many woolly hats are sold each season?
b. What is the price of a woolly hat?
c. How much should the production division charge the marketing
division for each hat?
7. Harry enjoys conversations about economics with Sally and wants her
to be his tutor. Sally does not enjoy tutoring but she enjoys getting
paid. She knows that Harry has utility function U(x, y) = x2 + y, where
x is number of hours of tutoring per week from her and y is income
remaining after paying for tutoring. She also knows that he has an
income m per week. Sally’s disutility associated with tutoring (or, to
be precise, the monetary equivalent of her disutility) is given by C(x) =
x3/3.
a. Because of her monopoly power, Sally can make Harry a take-it-or-
leave-it offer (P, x), which means she offers x hours per week for a
fee of P. Determine the optimal offer.
b. Harry’s friends get to hear about Sally. Now Larry, Barry and Terry
are also interested in economics tutoring. If each has the same
utility function as Harry, what offer or offers should Sally now
make? [Hint: she does not have to tutor all of them.]
8. Hummer & Co has two divisions. Division 1 produces cotton which
is used by division 2 in the manufacture of bandages. The market for
cotton is competitive and the price is pe = $350 per unit. One unit
of bandages can be produced from one unit of cotton. The marginal
costs for the two divisions are MC1= 200 + 0.375q and MC2 = 100 + 0.5q.
Demand for the company’s bandages is given by: p = 1000 – 0.625q.
a. Determine the optimal quantity and price for bandages and the
transfer price for cotton. How much cotton is bought or sold on the
external market?

179
28 Managerial economics

b. Suppose now that Hummer & Co has market power in the external
market for cotton. It cannot buy cotton from the external market,
but it can sell cotton according to an external demand function
pe = 358 – 0.09qe. Determine the optimal quantity and price for
bandages, and how much cotton (if any) is sold on the external
market and, if any is sold, the transfer price.

180
 Chapter 11: Oligopoly

Chapter 11: Oligopoly

Aims
The aim of this chapter is to consider:
• the differences in assumptions with respect to decision variables and
strategic (a)symmetry between the various oligopoly models discussed
in this chapter
• why there is a temptation to cheat in a cartel
• the name of the Nash equilibrium in the Bertrand model
• the concept of a profit possibility frontier
• why the division of profit can be problematic in a cartel
• how MFC clauses can facilitate collusion
• the nature of regime-switching models, the role of uncertainty in
these models and their predictions about price wars.

Learning outcomes
By the end of this chapter, and having completed the Essential reading and
activities, you should be able to:
• solve Stackelberg, Cournot, Betrand and collusion models
• solve a dominant firm model analytically and graphically
• demonstrate why in a repeated Bertrand model cheating is more
likely when there are many firms are impatient.

Essential reading
Tirole, J. The Theory of Industrial Organisation. Chapters 5 and 6.
Varian, H.R. Intermediate Microeconomics. Chapter 27.

Further reading
‘A bonus for Saddam’, The Economist, 11 March 1995, pp.105–06.
Abreu, D., D. Pearce and E. Stacchetti ‘Optimal cartel equilibria with imperfect
monitoring’, Journal of Economic Theory 39(1) 1986, pp.191–225.
Allen, B. and J.-F. Thisse ‘Price equilibria in pure strategies for homogenous
oligopoly’, Journal of Economics and Management Strategy 1(1) 1992,
pp.63–82.
Anderson, S.P. and M. Engers ‘Stackelberg versus Cournot oligopoly
equilibrium’, International Journal of Industrial Organisation 10 1992,
pp.127–35.
Bertrand, J. Book review of ‘Theorie Mathematique de la Richesse Sociale’ and
of ‘Recherches sur les Principes Mathematiques de la Thoerie des Richesses’,
Journal des Savants 68 1883, pp.499–508.
Boyer, M. and M. Moreaux ‘Being a leader or a follower’, International Journal
of Industrial Organisation 5 1987, pp.175–92.
Brander, J.A. and A. Zhang ‘Dynamic oligopoly behaviour in the airline
industry’, International Journal of Industrial Organisation 11 1993,
pp.407–35.

181
28 Managerial economics

Brannman, L.E. and J.D. Klein ‘The effectiveness and stability of highway
bid-rigging’ in Audretsch, D.B. and Siegfried, J.J. (eds) Empirical Studies in
Industrial Organisation: Essays in Honour of Leonard W. Weiss. (Dordrecht:
Kluwer, 1992) [ISBN 0792318064] pp.61–75.
‘Business and Finance’, The Economist, 3 December 1994, p.7.
‘Clean streets’, The Economist, 12 March 1994, pp.53–56.
Cooper, T.E. ‘Most-favoured-customer pricing and tacit collusion’, RAND
Journal of Economics 17(3) 1986, pp.377–88.
Cournot, A.A. Recherches sur les Principes Mathematiques de la Theorie
des Richesses. (1838) English edition, Bacon, N. (ed) Researches into the
Mathematical Principles of the Theory of Wealth. (New York: Macmillan,
1897).
‘Disputes are forever’, 1986, 17 September 1994, pp.93–94.
Edgeworth, F. ‘La Teoria Pura dle Monopolio’, Giornale degli Economisti 40
1897, pp.13–31.
Encaouna, D. and A. Jacquemin ‘Degree of monopoly, indices of concentration
and threat of entry’, International Economics Review 21(1) 1980, pp.87–105.
Geroski, P.A., L. Philips and A. Ulph Oligopoly, competition and welfare. (Oxford:
Blackwell, 1985) [ISBN 063114479X].
Gravelle, H.S.E. and R. Rees Microeconomics. (Harlow: FT Prentice Hall, 2004)
third edition [ISBN 0582404878] Chapter 16.
Green, E.J. and R.H. Porter ‘Non-cooperative collusion under imperfect price
information’, Econometrica 52(1) 1984, pp.87–100.
Hall, R.L. and C.J. Hitich ‘Price theory and business behaviour’. Oxford
Economic Papers 2 1939, pp.12–45.
Kreps, D.M. and J.A. Scheinkman ‘Quantity precommitment and Bertrand
competition yield Cournot outcomes’, Bell Journal of Economics 14(2) 1983,
pp.326–37.
Levinstein, M. ‘Price wars and the stability of collusion: a study of the pre-
World War I bromine industry’, NBER Working Paper Series on Historical
Factors in Long-Run Growth (August 1993) p.50.
Neilson, W.S. and H. Winter ‘Bilateral most-favoured-customer pricing and
collusion’ RAND Journal of Economics 24(1) 1993, pp.147–55.
‘Oil rolls over’, The Economist, 2 April 1994, p.75.
Ono, Y. ‘The equilibrium of duopoly in a market of homogenous goods’,
Economica (1978) 45, pp.287–95.
‘Scored’, The Economist, 20 November 1993, p.5.
Spulber, D.F. ‘Bertrand competition when rivals’ costs are unknown’, The
Journal of Industrial Economics XLIII(1) 1995, pp.1–11.
‘Smelt a rat’, The Economist, 23 July 1994, pp.72–73.
‘Steel woes’, The Economist, 19 February 1994, p.18.
Stigler, G. ‘A theory of oligopoly’, Journal of Political Economy 72(1) 1964,
pp.44–61.
‘Still smokin’, The Economist, 11 March 1995, pp.93–94.
Sweezy, P.M. ‘Demand under conditions of oligopoly’, Journal of Political
Economy 47(4) 1939, pp.568–73.
‘Then there were seven’, The Economist, 5 February 1994, pp.19–24.
von Stackelberg, H. Marktform und Gleichgewicht. (Berlin: Springer, 1994).

Introduction
In terms of number of firms and consumer price, oligopoly is an
intermediate market structure between monopoly and perfect competition.
Firms may price at marginal cost as in perfect competition, they may
set a monopoly price or a price between these two extremes. An
oligopolistic industry consists of a few firms who recognise their strategic
interdependence. This strategic interaction, rather than the number of
182
 Chapter 11: Oligopoly

firms, defines oligopoly. Whereas monopoly, monopolistic competition and

perfect competition could be analysed from an optimisation perspective,
oligopoly requires a game theory framework. Actions taken by each firm
(choice of price or output level) have implications for the pay-offs of all
firms in the market and therefore individual firms cannot ignore the effect
of their actions on their rivals’ behaviour. In contrast with monopoly and
perfect competition, there is no unified theory of how oligopolies behave.
Furthermore, the predictions of the various models diverge enormously
and depend on exactly which assumptions are made regarding decision
variables and strategic symmetry (firms making decisions simultaneously)
or asymmetry (leader-follower models). The diversity of models and their
predictions correspond to the diversity of real-life behaviour patterns
of oligopolistic industries. To decide which model is appropriate, the
particular industry under study has to be analysed carefully. There is an
abundance of industries which could be described as oligopolies. The US
domestic car market contains three major players (the Big Three), namely:
General Motors, Ford and Chrysler. In the airline industry, most routes are
served by a few airlines.
The oldest and best-known models are models of pure oligopoly (i.e. firms
producing a homogeneous good such as zinc or salt). In this chapter we
restrict ourselves to homogenous good industries. Oligopoly models can
be classified on the basis of their assumptions regarding the toughness
of price competition in the industry; on the basis of whether sequential
(strategic symmetry) or simultaneous moves (strategic asymmetry) are
assumed and according to which variable(s) are used as the decision
variable(s). The models we discuss in this chapter are listed in Table
11.1. In addition to these static (one-period) models we analyse repeated
versions of some of these standard models.1 1
See the section on
‘Dynamic interaction’ in
tough price sequential/ decision this chapter.
competition simultaneous variable
Stackelberg (section 2.1) no seq. quantity
Dominant firm (section 2.2) no/yes seq. price
Cournot (section 3.1) intermediate sim. quantity
Bertrand (section 3.2) yes sim. price
Joint profit max. (section 4) no sim. quantity

Table 11.1 Oligopoly models

Given that in pure oligopoly the product is homogenous, there is an
industry demand curve, relating price and total market demand. If firms
choose output levels then the price they obtain is determined from this
market demand function. When quantity is the decision variable, firms
are not making any choices regarding price. Price is determined by the
demand curve. This seems unrealistic in that it is hard to imagine how
real-life firms can operate without posting prices or publishing price
lists. But remember that in the study of monopoly the same problem is
encountered and the monopolist’s optimal quantity decision generally has
to be ‘translated’, using the demand curve, into a pricing policy to become
operational. There is no reason why the same method of analysis cannot
also be applied to oligopoly. The oligopolists could determine their and
their rivals’ ‘optimal’ or equilibrium production quantities and infer the
price from the demand function. You may wonder how price could be
a decision variable for firms selling a homogenous product. Surely they
will all have to charge the same price? Without running too far ahead
183
28 Managerial economics

let’s think of an industry consisting of two firms, A and B, and assume A

charges a lower price than B (pA < pB ). If A has sufficient capacity, it can
serve the entire market (i.e. the quantity demanded at pA according to
the demand function). If A has limited capacity however, A sells up to its
capacity and B faces the residual demand. When A is capacity-constrained
it is not a priori impossible for B to charge a different (higher) price. In all
of these scenarios we can model firms as taking decisions with respect to
the quantity or the price variable, the other one being determined through
market demand. Firms choose price in the Bertrand model and quantity in
the other one-period models we will discuss.
In practice, there are several factors determining whether price or quantity
is used as the decision variable in a particular industry. For strictly
homogenous goods it is generally accepted that there are no price choices
to be made by individual firms. The price is determined by the market
(demand function) and quantity is therefore the strategic variable. For
heterogenous goods, price is more likely to be the decision variable but
firms could also compete on advertising, capacity investments, service,
etc. If the firms in a particular industry customarily print catalogues which
list prices, then price is likely to be the strategic variable. Production is
adjusted to meet short-term variations in demand when the production
process allows for quick changes in the rate of production. Alternatively,
inventories are used to deal with variable demand. If the firm has high
inventory costs and/or has to plan production ahead, it is likely to consider
quantity as its strategic variable and deal with any situations of demand
shocks by adjusting its pricing policy. An easy way to adjust pricing in the
short-term is through offering discounts from a book price which is set
as the maximum the firm would consider charging. In the car industry,
quantity is the strategic variable since dealer discounts are adjusted more
easily than production schedules. Also when firms are operating close
to capacity and changing capacity is costly, a model with quantity as the
decision variable is likely to be a good approximation. In this case capacity
choice is really the strategic decision. In more sophisticated dynamic
models, firms can choose price and quantity and if the chosen price output
combination does not lie on the demand curve, inventory is used to absorb
the difference between supply and demand.
For convenience we will mainly concentrate on the two firm or duopoly
case. Duopoly was the prevalent market structure in the European airline
industry before deregulation. On the route between Country A and B
an airline from Country A and an airline from Country B (usually the
state subsidised airlines) shared the market. Countries had bilateral
agreements which determined how many passengers each airline could
carry and at what fare. Some well-known duopolies in the US market are
Pepsi and Coke (soft drinks) and Procter & Gamble and Kimberley-Clark
(disposable diapers). In the UK market for white salt, there are effectively
two producers: British Salt (a subsidiary of Stavely Industries) and ICI
Weston Point (part of the Mond division of ICI). In the models we consider
here it is also assumed that every firm knows the demand function and
the cost parameters of all firms in the industry. Each firm has complete
information.
You may be familiar with the ‘kinked demand curve model’ studied by
Sweezy (1939) and Hall and Hitch (1939). In this model, oligopolists
assume that their rivals will follow any downward change in price but
not an upward change in price. This implies that each firm’s demand
curve is elastic above the status quo price and inelastic below it (a
price increase leads to significant reduction in quantity whereas a price

184
 Chapter 11: Oligopoly

decrease is followed by a relatively small increase in sales): hence, the

kink in the demand curve. Because of the kink, the marginal revenue
curve is discontinuous and, as a consequence, small changes in marginal
cost do not call for changes in the optimal price. Because of its relative
unpopularity in modern industrial organisation, I will not discuss the
kinked demand curve model in detail. The kinked demand curve theory
is incomplete in the sense that it does not explain how the initial price
(corresponding to the kink) is established. At first sight the model is
appealing in that it seems to give an explanation for price rigidities in
oligopoly. However, empirically it has been found that prices in oligopoly
are not more rigid than monopoly prices.

Strategic asymmetry
In this section we study models in which firms for whatever reason do not
make their decisions simultaneously. Simultaneity does not necessarily
have a chronological meaning. In game theory ‘simultaneous decisions’
refers to the situations in which players make decisions ignorant of the
decisions taken by the other players. In sequential models there is a
‘leader’ who makes a first move, after which the ‘followers’ make their
decisions. This type of analysis is obviously only appropriate when there is
an industry leader and it is clear which firm assumes the leadership role.

Stackelberg
In the Stackelberg model, the ‘leader’ (Firm 1) makes a quantity decision,
which can be observed by the ‘follower’ (Firm 2) who in turn decides on
quantity. Price depends on total output and the exact relationship between
output and price is given by the demand curve p(q1 + q2). Let’s consider
the follower’s problem first. The follower has to determine an optimal
output level q2 given that the leader has chosen q1. Hence, his optimisation
problem is:
max p(q1 + q2) q2 – c2(q2 ).
q2
The result (i.e. the optimal choice of q2) will generally depend on q1: q2 =
q2 (q1). The leader can anticipate the follower’s choice of quantity and its
dependence on his own quantity decision and therefore his problem is:
max p(q1+q2(q1 ))q1 – c1(q1 ).
q1

Example 11.1

Demand in an industry consisting of two firms is given by p = 24 – q1 –

q2. Firm 1 is the leader and has constant marginal costs of 8 per unit;
Firm 2 is the follower with constant marginal costs of 4 per unit. Given
q1, Firm 2 determines its optimal quantity by maximising:
(24 – q1 – q2) q2 – 4 q2.
From the first order condition we find q2=10 – q1/2. Firm 1 can
anticipate this dependence of q2 on q1 and maximises (24 – q1 – 10
+ q1/2)q1 – 8q1 which leads to a quantity choice of q1 = 6. So the
Stackelberg model predicts q1 = 6, q2 = 7 and p = 11 for this industry. The
profits are 18 and 49 for Firm 1 and Firm 2 respectively.

185
28 Managerial economics

Because of the strategic asymmetry assumed in the Stackelberg model

(i.e. the leader can influence the follower’s quantity choice but not vice
versa), the leader generally has an advantage. This first mover advantage
is an essential ingredient of the Stackelberg model and makes the model
applicable only to industries where such an advantage clearly exists.
If we interpret quantity choice as capacity choice, it may be easier to
find situations in which the Stackelberg model is empirically plausible.
It would mean that the firm which invests first acts as the leader. With
this interpretation, one could argue that IBM had a leader’s role in the
mainframe computer market.

Dominant firm
The dominant firm model applies to industries in which there is one large
firm (or a cartel) acting as a price leader and several small firms. The
small firms in the industry (the fringe) act as competitive firms in that
they take the price set by the dominant firm as given and they supply all
they want at this price. In other words, they maximise their profits by
determining their supply quantity as the quantity for which marginal cost
equals the given price. In contrast to the perfectly competitive model there
is no condition of zero profits. The dominant firm supplies the residual
demand. As in the Stackelberg model, the dominant firm as the leader has
the advantage of influencing the fringe firms’ or followers’ quantities, in
this case by imposing the price they have to sell at.
Figure 11.1 illustrates how the price decision is made. For any price, the
dominant firm can determine how much will be sold by the fringe firms
by horizontally adding their MC curves. This is indicated by the SMC (sum
of marginal cost) curve. The horizontal difference between the industry
demand curve D and the SMC is the residual demand (RD) faced by the
dominant firm. Profit maximisation by the price leader requires setting
marginal revenue (MR) corresponding to the residual demand equal to the
fern’s marginal cost (MC). This leads to an output of q* by the dominant
firm. The optimal price can be read from the residual demand curve as p*.
Fringe firms supply q′ at price p*.

SMC

p*
MC
RD D

MR

MC
q

q* q′

Figure 11.1: Dominant firm model

Note that when the dominant firm has a significant cost advantage as
indicated by MC′ in Figure 11.1, it finds it optimal to set the price below
186
 Chapter 11: Oligopoly

the fringe firms’ marginal cost SMC. The fringe firms therefore do not
produce and the dominant firm is the monopoly producer. In practice, a
dominant firm may decide not to undercut the fringe firms even when
that would deliver the highest profit. The reason is that refusing to
accommodate the fringe firms or taking over some of them may provoke
legal action.
If there are no barriers to entry and the small firms are making positive
profits, entry will occur, shifting the SMC curve to the right and lowering
the dominant firm’s market share. For example, US Steel’s market share
decreased from 75 per cent in 1903 to less than 25 per cent in the 1960s.
American Can which used to supply 90 per cent of the tin market in 1901,
saw its market share decrease to 40 per cent by 1960. Such market entry
may alter the way in which the industry operates. The dominant firm may
not be willing to tolerate a very large fringe sector.
Theoretically, for the dominant firm model to apply, the dominant firm
should be powerful so that it can credibly threaten to punish fringe firms
if they do not accept its price leadership. This punishment would take
the form of driving the fringe firms out of business through a price war.
Therefore the dominant firm should have low costs, substantial market
share and large production capacity to force the other firms in the industry
to set the same price. In the UK white salt market, however, there is
evidence of British Salt following ICI in its pricing whereas British Salt is
the low-cost producer. Examples of industries for which the dominant firm
model may be plausible include (dominant firms in brackets):
• the UK chemicals industry (ICI)
• the European soft drinks industry (Coca-Cola)
• the US aluminium industry (ALCOA)
• the US airline industry (American Airlines)
• the European steel industry (British Steel)
• photocopiers (XEROX)
• cars (GM)
• cameras and film (KODAK).
The reasons why particular firms get to be leaders may be related to size,
management style or they may be historical. For example, Coca-Cola was
exempt from sugar rationing during the war in return for supplying cheap
Coke to American troops in Europe. This gave Coca-Cola a significant
advantage over Pepsi Co at least in the European market. Federal Express,
probably because of first mover advantages, is seen as the price leader in
overnight delivery.
The leadership role can switch between large players in an industry. In the
US candy bar market Hershey was dominant in the 1960s. This dominance
was later challenged by Mars in the early 1970s. The term barometric
price leadership refers to firms deliberately alternating the dominant
firm role to avoid attracting the authorities’ attention. The firm which
takes the leader role is the first to announce a price change when a change
in cost or demand conditions warrants it.
The dominant firm model can be modified to allow for several large firms
dominating the market.2 These market leaders can be modeled as joint 2
See, for example,
profit maximisers (a cartel) in which case our analysis above is basically Encaoua and Jacquemin
unaltered or they may be assumed to behave in Cournot fashion with (1980)

respect to residual demand.3 Industries with a few(rather than one) major 3

See the section on
players are empirically important. In the US tobacco industry in 1995 for ‘Cournot’ in this chapter

187
28 Managerial economics

example, three giant companies (BAT Industries, Philip Morris and RJR
Nabisco) jointly have a 90 per cent market share.4 4
See ‘Still smokin

Example 11.2

Assume industry demand is given by p = a – bQ, where Q is total

industry output. There are n identical small firms in the industry with
cost function C(qs ) = csqs2. The dominant firm has constant marginal
costs cd and produces an output qd so that Q = nqs+ qd. Suppose the
dominant firm sets a price p. The small firms set marginal cost equal
to p so that p = 2csqs or qs = p/(2cs ). Hence, residual demand for the
dominant firm is given by:
qd = Q – nqs = (a – p )/b – np/(2cs ).
This can be rewritten as:
p = (a – bqd )/(1 + bn/(2cs )).
The dominant firm sets marginal revenue corresponding to this demand
equal to its marginal cost:
cd = (a – 2bqd ) /(1 + bn/(2cs )).
which results in:
qd = (a – (1 + bn/(2cs)) cd )/(2b).
The optimal price set by the dominant firm can be found by substituting
qd in the residual demand curve, which gives:
p = cd / 2 + csa/(2cs + bn).
For n = 0 this reduces to p = (cd + a)/2, the monopoly price. The optimal
price is decreasing in the number of fringe firms.

Symmetric models
The theoretical difficulty of justifying asymmetric models lies in the fact
that the asymmetry is not explained within the model. There have been
some recent attempts5 to endogenise the strategic (a)symmetry in oligopoly 5
See, for example,
models. Unless we know the particular industry we are studying well, and Ono (1978); Boyer
and Moreaux (1987);
can appeal to, say, historical reasons for the asymmetry, it is perhaps more
Anderson and Engers
sensible to assume that no firm has a strategic advantage. This assumption (1992)
is the starting point for simultaneous moves oligopoly models.

Cournot
In the Cournot (1883) model firms make simultaneous quantity decisions.
In game theory terminology, the duopolists are engaged in a non-cooperative
game and we are interested in the Nash equilibrium of this game (i.e. a pair
of quantity choices such that it is in neither firm’s interest to alter its choice
unilaterally). These strategies are consistent in the sense that no firm has
ex post regret when it observes its competitor’s quantity decision. The good
produced by the Cournot duopolists is homogenous and therefore the price
is determined by the total quantity supplied: p(q1 + q2). To find the Nash
equilibrium, we find best response functions (i.e. we find the best quantity
choice of Firm 1(2) as a function of Firm 2 (1 )’s quantity choice). The
intersection of these best response functions gives the Nash equilibrium.
Mathematically this means that for Firm 1 we solve:
max p(q1 + q2 ) q1 – c1(q1 )
q1
and find q1 = f1 (q2) and for Firm 2 we find q2 = f2(q1) similarly. The Nash
equilibrium is then found by solving q1 = f1( f2(q1)).
188
 Chapter 11: Oligopoly

Example 11.3

Assume the same demand function and marginal costs as in example 1.

The firms’ profits can be written as π1 = (24 – q1 – q2 – 8)q1 and π2 =
(24 – q1 – q2 – 4)q2. The first order conditions lead to the response
functions q1 = f1(q2) = (16 – q2)/2 and q2 = f2(q1) = (20 – q1)/2. These
response functions are drawn in Figure 11.2. The unique Nash
equilibrium is at the intersection where q1 = 4 and q2 = 8. Hence, the
Cournot model predicts a price of 12. Compared to the Stackelberg
model, the Cournot model predicts a lower output for Firm 1 (the
Stackelberg leader) and more for Firm 2 (the follower). Firm 1 has
lower profits and Firm 2 has higher profits than in the Stackelberg
model. These conclusions hold generally, not just for this example.
In the Stackelberg scenario the leader has the opportunity to select a
high output to induce the follower to cut back his production. In the
Cournot model this is not possible because firms take output decisions
simultaneously.

q
2

16

10

8
f
2

q
f 1
1
4 8 20

Figure 11.2: Cournot response functions

Example 11.4

Consider an industry consisting of n identical firms with constant

marginal cost c and demand function p(Q) where Q is total industry
output. Firm i’s profit function is given by:
πi = p(Q)qi – cqi
Its optimal quantity decision given the other firms’ quantity choices is
determined by the first order condition:
∂p
p(Q)+qi –c=0
∂Q

Since all firms are identical we look for a symmetric Nash equilibrium
in which all output decisions are identical i.e. qi = Q/n so that the
condition above can be rewritten as:

()
Q ∂p
p+ n
∂Q
= c,

189
28 Managerial economics

which, using the definition of demand elasticity, reduces to:

p
p – nη = c

or
p–c 1
.
p = nη

We conclude that, at the Cournot solution, each firm’s monopoly power,

measured as its markup-price ratio, varies inversely with the demand
elasticity and the number of firms in the industry. If the number of firms
n is very large, the Cournot solution approximates the equilibrium of a
perfectly competitive industry.

Bertrand
Bertrand (1883) wrote a book review of Cournot’s work and disagreed
with Cournot about his assumptions on the behaviour of oligopolists.
Bertrand thought that oligopolists would collude rather than compete and
to demonstrate how unrealistic Cournot’s description was, he developed
a model in which the decision variable is price. He showed how Cournot’s
way of thinking leads to the implausible result that in a duopoly the
perfect competition price is charged.
We can think of the Bertrand model as a non-cooperative game in which
each firm decides on price, taking the other prices as given. We then
look for a Nash equilibrium as in the Cournot model. The assumptions
of the Bertrand model are that the product is homogenous, consumers
have complete information and there are no search or transport costs
so that they buy from the firm charging the lowest price. If firms charge
the same price they are assumed to share the market and have identical
market shares. Assume firms have identical and constant marginal costs
and unlimited capacity. Under these assumptions we find the very striking
result that there is a unique Nash equilibrium in which both firms set
price equal to marginal cost. Why is this the Nash equilibrium? Clearly,
price could not be below marginal cost since then firms would be making
losses. Price cannot be above marginal cost since then one of the firms
could reduce its price by a very small amount, capture the whole market
and make a positive profit. If firms differ in (constant) marginal costs,
then the low cost firm prices just below the second lowest marginal cost at
equilibrium unless that price is above the monopoly price.

Example 11.5

Assume the same data as in Example 11.1. We look for a Nash

equilibrium in prices. For any price at or above Firm 1’s MC, Firm 2,
which has the lower MC, can undercut Firm 1 and capture the whole
market. At the Nash equilibrium Firm 1 charges p1 = 8 and Firm 2
charges p2 = 8 –  so that Firm 2 is the sole supplier. Note that Firm 2
is not charging the monopoly price however (check that a monopolist
would charge p = 14) since that would induce Firm 1 to enter the
market. The Bertrand model thus predicts a price of 8 – e and a quantity
of 16 + . Compare this to the predictions of the Cournot model in
which the high cost producer, because of the lower intensity of the price
competition, obtains a 1/3 market share.

190
 Chapter 11: Oligopoly

The assumption of consumers buying from the firm charging the lowest
price, even if the price difference is very small, is quite restrictive and
justifiable only when the goods are very close substitutes. In fact,
oligopolists strategically differentiate their products to reduce tough price
competition. When products are differentiated, a firm can charge a higher
price than its rival without losing all its business. Even when the good
is homogenous, it is not realistic to assume that all demand goes to the
lower price firm. Allen and Thisse (1992) have developed a pure oligopoly
model in which some customers do not care about small price differences.
This implies that firms do not lose all of their sales if they are undercut
slightly by a rival. The equilibrium of their pricing game allows for some
market power by the firms. Spulber (1995) shows that the assumption of
firms knowing rivals’ costs is a crucial ingredient in the Bertrand model. If
firms know only their own costs and have probabilistic information about
rivals’ costs, they set price above marginal cost at the equilibrium of the
pricing game.
It is difficult to think of situations in which the Bertrand model is a
reasonable approximation of reality. The scenario which fits the original
Bertrand model best is maybe that of firms submitting sealed bids for a
procurement contract where the contract is awarded to the firm quoting
the lowest price. However, even in these circumstances, firms find ways
to avoid tough price competition. In the 1950s, General Electric and three
other firms rotated sealed-bid business for circuit breakers. They held
secret meetings to decide market shares in advance. More recently, two
of America’s biggest bakeries, Continental Baking and Campbell Taggert,
have been investigated for alleged bidrigging in sales of bread to schools
and hospitals. In Japan in 1993, the Fair Trade Commission conducted
raids on some big electronics companies, including Sony and Toshiba, who
allegedly colluded on bids for electronic billboards in sports stadiums.6 6
See ‘Scored’

Whereas Bertrand had assumed that firms have sufficient capacity to

serve the entire market, Edgeworth (1897) modified Bertrand’s analysis
by restricting firms’ output levels to exogenously determined capacity
levels. As an example, think of airlines offering scheduled flights between
London Heathrow and New York JFK. In the short-run at least, an airline
cannot change the number of seats available on the route (unless the flight
frequency is increased which is problematic because of a shortage of take-
off and landing slots). If price is set equal to marginal cost, there may be
excess demand. When this is the case, both firms pricing at marginal cost
and making zero profit is not an equilibrium. One firm could increase price
slightly and make positive profits. Of course, customers would prefer to
buy from its rival (pricing at marginal cost) but they are rationed there.
The analysis of the Edgeworth model is quite complicated and I will not
discuss it here other than to state the conclusion: with capacity constraints
duopolists competing on price set price above marginal cost unless they
have excess capacity to the extent that one firm could serve the entire
market demand for price equal to marginal cost.
Since the predictions of Bertrand and Cournot are very different, it is
important to decide whether price or quantity is the strategic variable. In
a much quoted paper, Kreps and Scheinkman (1983) allow duopolists to
use both variables. In a first stage, ‘capacity’ is chosen which limits each
firm’s output in the second stage. In this second stage, firms set prices. The
model is in fact an Edgeworth model with capacities endogenised. The
firm which sets the lowest price can produce and sell up to its capacity
whereas the high price firm is left with any residual demand. Kreps and
Scheinkman show that, at the unique subgame perfect equilibrium of this

191
28 Managerial economics

game, the capacities chosen in the first stage are the Cournot equilibrium
quantities and in the second stage firms choose (identical) prices such that
demand equals joint capacity.

Collusion
Chamberlin was one of the first economists (after Adam Smith) to suggest
that oligopolistic sellers would form a cartel and cooperate to set joint
profit maximising price and output levels. Cartels were common before
the antitrust laws were passed. Levinstein (1993) vividly describes how
the US and Europe bromine producers colluded in the 30 years before
World War I. Currently, in most of the western world (Switzerland is a
notable exception), the norm is that collusion is illegal although some
cartels are sanctioned by governments. For example, the International Air
Transport Association, consisting of American and European airlines flying
transatlantic routes, used to set uniform fares for transatlantic flights. The
US allows agricultural committees to determine prices and production
quotas for some products. Sometimes collusion between firms in the
domestic market is explicitly forbidden but governments allow firms to
participate in international cartels. The European steel industry operated
as a price fixing cartel until 1988 and was fined in 1994 by the European
Commission.7 In the same year, the Commission fined 42 cement firms and 7
See ‘Steel woes’
trade groups for running a cartel.8 8
See ‘Business and
Finance’
Because price fixing is illegal, there are not many open cartels but firms
may nevertheless find ways to cooperate at the expense of customers.
Powerful firms use the media to announce price changes which are followed
by the other firms in the industry. All firms in the industry use the leader’s
price as a focus point and the risk of ‘misunderstandings’ is minimal. This
is a form of implicit or tacit collusion. Trade associations or professional
associations are often vehicles for collusion. This certainly seems to be the
case for the Associations of Trade Waste Removers and their central council
which hold a tight grip on the rubbish-collection business for commercial
customers in New York City. The rate charged by New York carters is fixed
at three or four times the rate in Los Angeles or Chicago. The business is
notorious for its connections with the Mafia.9 9
See ‘Clean streets’

Technically, the problem of a cartel is at first sight identical to that of

a monopoly operating several plants. The optimal output is found by
horizontally summing marginal cost curves and determining the intersection
of this cartel marginal cost curve with the marginal revenue curve (or the
horizontal sum of the marginal revenue curves if there are several markets).
The output is allocated to individual plants such that marginal cost is equal
in all plants and equal to marginal revenue. For joint profit maximisation, it
is necessary to allocate high output quotas to low cost firms and low quotas
to high cost firms which may even shut down if they are too inefficient.
In an oligopoly, finding the optimal output division between cartel members
is only a first step. An important but often ignored problem is how to
divide the cartel profit. If firms can bargain over sidepayments, it is not as
difficult to agree on optimal quotas as when no sidepayments are allowed.
By definition of the collusive solution (i.e. that it is the solution which
maximises joint profits) it is always possible to make all cartel members
better off than they would be in a non-cooperative setting such as Cournot.
This is not to say that the division of profits is trivial when side payments
are possible but it seems that the bargaining should not be too difficult.
Because of the problems associated with divisions of profits, as well as
uncertainties about market demand, output allocations are often clearly

192
 Chapter 11: Oligopoly

suboptimal in real life cartels. When there are no sidepayments, each

cartel member wants a large share of the output. Production allocations
may be based on past sales, capacity or a geographic segmentation of the
market (‘market sharing cartels’). Having the most skilled negotiator may
be more important than relative production efficiency.
Usually cartel members do not make contractual agreements about
sidepayments or redistribution of profits since this would be evidence of
(illegal) collusion. In this case (if firms are not identical), some firms may
be better off at the Cournot solution than at the joint profit maximising
solution. This is illustrated for a duopoly in Figure 11.3 where a profit
possibility frontier is drawn. By definition, all combinations of profits
under this curve can be attained by a suitable choice of q1 and q2. The
curve itself is derived by solving:
max π1 + (1 – ) π2
q1, q2
for all values of the parameter  between 0 and 1. The point which
maximises joint profits corresponds to the solution to the optimisation
problem for  = 1/2. NE in Figure 11.3 indicates the Cournot-Nash
equilibrium pay-off pair. By moving from NE to the point (max) where
joint profits are maximised, firm 2’s profit decreases.

profit 2

NE o max

profit 1

Figure 11.3: Profit possibility frontier

Mathematically, a cartel maximises joint profit:
max π = p(Q)(q1+ q2 ) – c1(q1 ) – c2(q2)
q1, q2
where Q = q1 + q2. The first order conditions are:
∂π = ∂p
× (q1+q2 ) + p(q1+q2 ) – MC1(q1 ) = 0
∂q1 ∂Q
and
∂π = ∂p
× (q1+q2 ) + p(q1+q2 ) – MC2(q2 ) = 0
∂q2 ∂Q

which gives the result mentioned above: marginal revenue equals each
firm’s marginal cost at its optimal output level.

193
28 Managerial economics

We can use this analysis to show that, for cartel members, since they are
playing a prisoners’ dilemma game, there is always a temptation to cheat
by overproducing. Given that firms, even when they have formally agreed
to collude, have no recourse to a court of law (because collusion is illegal),
this temptation is very real. Consider firm l’s profit given that firm 2
produces the joint profit maximising quantity q2*:
π1 (q1, q2*) = p(q1 + q2*) q1 – c1 (q1 ).
If firm 1 assumes that firm 2 is going to stick to q2* it finds its optimal
output level q1* by differentiating this function with respect to q1 which
gives:
∂p ∂c1
p(q1+ q2* ) + q1 – (2)
∂Q ∂q1
Comparing this to the first order condition for q1 above in (1) reveals that
(2) is equal to:
∂p
– p(q2)
∂Q
which is positive. This indicates that Firm 1 can increase its profits by
increasing output if Firm 2 keeps output constant at the ‘collusive’ amount.
The intuition for this is that marginal revenue for the cheater is close to
the cartel price whereas his marginal cost equals marginal revenue and is
thus lower than the cartel price.
Even when cartels can monitor and enforce prices effectively, cartel
members may engage in non-price competition by offering free extras such
as after-sale service or free delivery and installation. Airlines offer better
meals, free in-flight entertainment etc., in order to lure customers away
from other cartel members. This type of non-price competition can be
interpreted as cheating and it is not as easily detected as price cutting.

Case: OPEC
The standard example of a cartel is OPEC which restricts oil output
by its members so as not to spoil the market. OPEC’s history has
repeatedly illustrated the difficulties of maintaining a cartel and
avoiding cheating by cartel members. Within OPEC there is tension
between the rich countries with large oil reserves and a small
population (e.g. Kuwait and Saudi-Arabia, which alone accounts for a
third of OPEC’s output) and poor countries with small oil reserves and
a large population (e.g. Libya and Indonesia). The countries with large
reserves want a low price since they are concerned about the long-
term effect of high oil prices: customers may start switching to other
energy sources and there may be market entry in the form of non-
OPEC exploration and production such as in the North Sea. Countries
with low reserves do not worry about the long-term effects and, since
short-term demand elasticity for oil is low, they argue for a high price.
The poor OPEC countries have consistently overproduced their quotas 10
See ‘Oil rolls over’; ‘A
and even the United Arab Emirates decided to increase its production bonus for Saddam’
to 1.5 million barrels a day from its allocation of one million barrels in
1988. The lifting of sanctions against Iraq, imposed by the UN after the
invasion of Kuwait, may put more pressure on the cartel.10

194
 Chapter 11: Oligopoly

Case: Diamonds
When industries succeed in filtering sales through a common
distributor, collusion is not an unlikely mode of behaviour. The
Central Selling Organisation (CSO), a subsidiary of De Beers, aims to
control the world wholesale market in rough diamonds. In 1993, De
Beers itself accounted for 50 per cent of CSO sales and Russian sales
amounted to 26 per cent. Even this seemingly cosy cartel arrangement
has had problems. Russia agreed in 1990 to sell 95 per cent of its
output through the CSO for a period of five years but it is unlikely
that this agreement was honoured. In particular, Russia has exploited
loopholes such as classifying diamonds as ‘technical’ and hence not
covered by the agreement. De Beers has had to buy this leaked output
to keep control of the flow of diamonds on to the market and hence
their price.11 11
See ‘Disputes are
forever’
Given the incentive to overproduce, cartels use several mechanisms to
detect and prevent cheating e.g. forcing firms to publicise prices. Firms
may adopt ‘facilitating practices’ which eliminate the incentive to cheat.
For example, cartel members may agree to offer a ‘meet or release clause’.
Such a clause applies when a customer finds a lower price offered by
another supplier, in which case the low price is met by the original seller
or the customer is released from the obligation to buy. Thus way cartel
members are informed if any undercutting takes place. Another ‘trick’
used by colluders is the ‘most favoured customer clause’ (MFC). Under
MFC, firms promise their customers that if they ever lower the price, they
will offer a rebate, equal to the difference between the price customers
pay now and the new price. In this scenario, if firms can agree to collude
for a small number of periods while offering the MFC, then their pay-offs
are such that deviating from the collusive outcome by lowering price is
no longer profitable because of the penalties which would have to be
paid to previous customers. Both General Electric and Westinghouse,
manufacturers of turbine generators, used MFC, effective for six months
after a sale, in the 1960s and 1970s until they agreed to end the practice
to avoid antitrust prosecution.12 12
See Cooper (1986)
and Neilson and Winter
(1993)
Dynamic interaction
So far we have discussed models which describe oligopolistic behaviour
when the firms play a one-shot game. More realistic models assume a
dynamic setting and in particular the repeated prisoner’s dilemma
provides a natural framework. As we have seen before, collusion can be
an equilibrium if the time horizon is infinite or if there is uncertainty
about the length of the horizon. Such a collusive equilibrium is sustained
by players’ threats to retaliate (i.e. to revert to the non-cooperative Nash
outcome as soon as one player deviates). To illustrate this idea consider a
simple infinitely repeated Bertrand game. All n firms are identical and the
price which maximises joint profit is pm, the monopoly price. If all firms
charge pm each makes a profit equal to a share 1/n of the monopoly profit
Пm. If one of the firms undercuts however, it captures the whole market.
Although repetition of the one-period Bertrand solution with all firms
pricing at marginal cost in each period is an equilibrium in this game,
there may be a Nash equilibrium in which firms collude. Suppose firms
use trigger strategies of the following type. collude (i.e. set price pm)
until at least one firm deviates; when one or more firms deviate in a given

195
28 Managerial economics

period, set price equal to marginal cost from the next period onwards. For
this combination of trigger strategies to form an equilibrium it should be
the case that it does not pay for any firm to do something else assuming all
the other firms are sticking to their trigger strategies. If a firm decides to
deviate, it gains:
Пm – Пm/n = Пm (n – 1)/n
immediately as it becomes a monopoly. Its loss, compared to the collusive
outcome, due to retaliation from the next period onwards is:
(Пm/n)(δ + δ2 + ...) = (Пm/n) (δ/(1 – δ))
where δ is a discount factor. The firm cheats if the gain exceeds the loss or
when:
δ < 1 – 1/n.
This proves that, at least in this simple pricing model, cheating is more
likely when the number of firms in a cartel is large. When n = 50, a
discount factor above 0.98 is needed to sustain collusion whereas for an
industry with n = 2 firms, a discount factor above 0.5 is sufficient. The
dependence on the discount factor is clear: when firms are patient
(high δ), they evaluate their future losses more highly and this deters them
from cheating.
We have analysed repetition of the Bertrand pricing game here but a
similar story can be told about repeating the static Cournot game. The
trigger strategies there involve playing Cournot equilibrium output forever
as soon as cheating occurs. In contrast to the Bertrand situation where you
don’t sell at all when someone cheats, firms in a Cournot industry may not
be able to observe cheating immediately but maybe two or three periods
after it has taken place. Cheaters could get away with earning profit
above their share of the collusive profit for several periods. Detection lags
therefore make cheating more tempting.
Stigler (1964) has pointed out that, when demand fluctuates and there
is a lot of uncertainty, there is larger scope for misunderstanding moves
which merely reflect changed demand conditions as attempts to cheat.
This observation has formed the basis for recent game theoretic work on
dynamic oligopoly.13 Recall that the repeated Bertrand and Cournot models 13
See for example Green
predict that, when collusion is sustained, prices and outputs are stable. and Porter (1984); Abreu
et al. (1986)
Firms choose the joint profit maximising price or output and there are no
price wars. In the recent models which allow for uncertainty, price wars
do occur. To illustrate the main ideas here, let’s assume a homogenous
oligopoly with identical firms. The time horizon is infinite or there is
a chance, in each period, that the game ends. Firms decide on output
levels and they observe only their own output level and the market price
(which, as always, depends on total industry output). Each firm uses these
observables to deduce whether a rival has defected from the (tacitly)
collusive arrangement (i.e. when the market price is low, overproduction,
compared to the joint profit maximising output, has taken place). So
far, we are describing a repeated Cournot game. The ingredient which
has to be added to generate equilibrium price wars is uncertainty. The
relationship between industry output and price is stochastic so that a low
price can be caused by a rival flooding the market or an exogenous fall in
demand.
Green and Porter (1984) show that, when there is uncertainty, there is an
equilibrium in which firms periodically revert from the collusive solution
to static Cournot output levels for a few periods, after which they return
to the collusive output levels. At this equilibrium, firms are using trigger

196
 Chapter 11: Oligopoly

strategies which tell them to start behaving non-cooperatively when price

falls below a given level. The non-cooperative punishment phases are
called price wars. What is interesting in this and other regime-switching
models is that no firm ever defects at this type of equilibrium yet price
wars do and have to occur to sustain the equilibrium. In other words, if
there is no trigger in terms of a low price level setting off a price war, it
would not be in any individual player’s interest to continue to collude.
Paradoxically, price wars can be seen as evidence of collusion. If firms
act non-cooperatively (play static Cournot in each period), there is no
reason for them to revise prices periodically and consequently price and
output is stable. However, when they try to collude, punishment periods
are necessary. In the Green-Porter model, the ‘collusive’ output is not the
output a monopolist would set. When determining output, firms have
to take into account that when a low output level is set, the temptation
to cheat is higher and in order to sustain the collusive equilibrium the
punishment periods would have to be longer.
In terms of empirical predictions from this type of model, we should
expect to observe price wars in recessionary periods when there is surplus
capacity. There is some casual evidence for price wars triggered by an
exogenous fall in demand. Green and Porter offer the American rail freight
industry in the 1880s as an example of an industry which behaved in
a manner consistent with their model. The European car market went
through price wars when it declined by 16 per cent in volume in 1993. It
has overcapacity of at least three million cars per year.14 In an attempt to 14
See ‘Then there were
test Green-Porter type models, Brander and Zhang (1993) have analysed seven’
data for the 1984-1988 period on economy and discount fares on routes
to or from Chicago for which American Airlines and United Airlines
are duopolists. They find some evidence in favour of regime switching
behaviour with quantity as the decision variable and reversion to Cournot.
Punishments appear to have taken place during the first three quarters
of 1985 and the first quarter of 1987. The hypotheses that American
and United were playing one shot Bertrand or Cournot or joint profit
maximising, were rejected.

Case: Aluminium
Aluminium producers certainly seem to be aware of the dangers of
excess capacity. They agreed to cut back capacity in 1994 and prices
and profits have increased since this agreement. However, Russia -
which is a large producer and exporter - is believed to be cheating
by cutting down its smelting capacity by much less than the agreed
500,000 tonnes. Furthermore, it is quite possible that, without the
agreement, Russia would have had to close more capacity than it has
done now. Russian costs have increased and inefficient Russian smelters
are kept in business only because prices have climbed. Also, as firms
become more profitable, the temptation to cheat and bring capacity
back into production may be hard to resist even for ‘honest’ western
producers.15 15
See ‘Smelt a rat’

There are models which reach exactly the opposite conclusion to Green- 16
See, for example,
Porter, namely: that cartels are less stable during expansionary periods
Brannman and Klein
because firms then have less capacity or inventories to punish the (1992)
cheaters. Also the gain from cheating is larger when there is a boom in
demand. Studies of construction industry auctions concur with this latter
prediction.16

197
28 Managerial economics

Conclusion and extensions

In this chapter we have discussed Stackelberg, dominant firm, Cournot,
Bertrand and collusion models as the basic models of oligopoly. These
models vary with respect to the decision variable firms are assumed to
manipulate and the toughness of price competition between sellers. The
Bertrand model assumes an extreme degree of price competition; the joint
profit maximisation model assumes no price competition; the Cournot
model is somewhere in between. The Bertrand model predicts a low price
and high output whereas the joint profit maximisation or collusion model
predicts a high price and low output. Whereas Bertrand, Cournot and
joint profit maximisation models assume simultaneous decision-making,
in the Stackelberg model and the dominant firm model, it is assumed
that one firm (the leader) makes a decision which can be observed by the
follower(s) who in turn make their decisions taking the leader’s decision
into account. From a theoretical point of view, the simple leader-follower
models are not complete because they do not explain how firms adopt
leader or follower roles. More recent versions of the Stackelberg model
endogenise the role choice and are therefore more sound.
The predictions of the standard models are not always very convincing and
this is the reason why industrial organisation economists have proposed
alternative and often more complex models. In developing these models,
they have dropped or revised some of the assumptions in the older models,
for example the one-shot assumption which leads one to ignore the fact
that real-life oligopolistic firms interact with each other over time. We
already know from the discussion of the repeated prisoners’ dilemma,
that repetition can make a difference. In the infinitely repeated Cournot
game, firms could play trigger strategies (or other strategies which allow
for punishment if one player cheats) as equilibrium strategies so that they
both set low output rates (and hence a high price is sustained). This could
ensure that a point on the profit possibility frontier is reached but the
Nash equilibrium of the one-shot game, as well as a large number of other
combinations of strategies, is also an equilibrium in the repeated version
and so one cannot claim that collusion must necessarily be achieved. When
it is achieved, prices and output levels are stable in the repeated versions
of the Cournot and the Bertrand games. This is not the case when demand
is uncertain and firms cannot be sure whether cheating has occurred. In
regime switching models firms alternate between collusive periods and
punishment periods. Price wars start when (stochastic) demand falls below
a trigger level.
We have only considered pure or homogenous good oligopolies but real
life products are often differentiated. Fortunately, differentiated oligopolies
can be analysed in essentially the same way as pure oligopolies.17 The 17
see, for example,
main difference is that, when products are differentiated, each products Gravelle and Rees
(2004) Chapter 16
has its own demand curve whereas in pure oligopoly it is assumed that
price depends on total industry output. As a result, prices charged by
firms in a differentiated oligopoly are not identical. Demand for a good
produced by one oligopolist depends on the prices of the other products
as well as its own price. It can be argued that collusion is less likely in
differentiated oligopoly because of the larger informational requirements
and the difficulty of monitoring. When a manufacturer of a substitute
product lowers his price it is not known whether this is cheating or
whether it reflects changed cost or demand conditions. Given that in a
differentiated oligopoly technologies are less likely to be identical, and
even at the collusive outcome prices differ, monitoring is an awesome task.

198
 Chapter 11: Oligopoly

In the models we have discussed, firms have complete information. In

practice it is unlikely that firms have exact information about each other’s
costs. When firms have private information they can strategically reveal
this information.18 In experiments when information about competitors 18
See, for example,
is limited and no communication is allowed, prices tend to the Bertrand Tirole (1988) Chapter 6
solution and there is some evidence in favour of the Cournot model
when subjects are asked to decide on quantity. The collusive outcome
becomes likely when there are few experienced players who have perfect 19
See Geroski, Phlips
information about each other’s costs and actions.19 and Ulph (1985)

A reminder of your learning outcomes

Having completed this chapter, and the Essential reading and activities,
you should be able to:
• solve Stackelberg, Cournot, Betrand and collusion models
• solve a dominant firm model analytically and graphically
• demonstrate why in a repeated Bertrand model cheating is more
likely when there are many firms are impatient.

Sample exercises
1. Let the inverse demand curve for a homogenous product be
p = 70 – Q. Suppose there are two firms in the industry, each with
constant marginal costs of 10. Assuming they behave as Cournot
duopolists, what will be the price and output? What price and output
levels does the Stackelberg model predict? What price and output
levels does the Bertrand model predict? What is the collusive price
and output level? Now assume the firms interact each period and
the time horizon is infinite. For which values of the discount factor is
there a collusive equilibrium sustained by trigger strategies if firms
play Cournot forever after cheating has occurred? What if they play
Bertrand forever after cheating?
2. An industry consists of two firms, each of which produce output at a
constant unit cost of 10 per unit. The demand function for the industry
is Q = 1,000,000/p. Find the Cournot and Stackelberg equilibrium price
and quantity.
3. PowerGen and National Power are the only two electricity generating
companies in the UK. The demand for electricity is p = 580 – 3q. The
total cost function of PowerGen is TCP = 410qP and the total cost
function of National Power is TCN = 460qN.
a. If these two firms collude to maximise their combined profits, how
much will each firm produce?
b. How will they divide profits?
4. An industry consists of one dominant firm and five small (fringe) firms
which behave competitively, taking the price set by the dominant firm
as given. Industry demand is Q = 10 – p. The fringe firms have identical
cost functions C(q) = q2 and the dominant firm has constant marginal
cost of 1 per unit. Derive the dominant firm’s residual demand and
corresponding marginal revenue. Illustrate all your derivations on a
graph. What price will the dominant firm set? At this price, how much
does it supply and how much is supplied by the fringe firms?

199
28 Managerial economics

5. Demand is given by p = 1 – Q and production costs are zero. Find

Cournot, Bertrand and Stackelberg price, output and profit levels.
Suppose firm 2 (which is the follower in the Stackelberg game) has a
fixed cost F. How does this change your answers?

200
 Appendix 1: Sample examination paper

Appendix 1: Sample examination paper

Important note: This Sample examination paper reflects the

examination and assessment arrangements for this course in the academic
year 2010–2011. 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

Candidates should answer SIX of the following TEN questions: FOUR
from Section A (12.5 marks each) and TWO from Section B (25 marks
each). Candidates are strongly advised to divide their time
accordingly.
A handheld non-programmable calculator may be used when answering
questions on this paper. The make and type of machine must be stated
clearly on the front cover of the answer book.

Section A
Answer all four questions from this section (12.5 marks each)
1. Each of two firms (A and B) must decide, independently and without
knowing the other’s decision, whether to use its proprietary technology
to manufacture a product or to use its competitor’s technology. Each
firm would prefer that they both use the same technology rather than
different technologies but naturally each prefers that its technology be
the industry standard. The pay-offs are given in the matrix below.

B
use A’s use B’s
technology technology
A
use A’s
2,1 0,0
technology
use B’s
0,0 1,2
technology
a. Find all (including mixed strategy) equilibria.
b. Draw the pay-off region and mark the equilibrium pay-off pairs
corresponding to your answer to (a).
c. Which (if any) of the equilibria are Pareto efficient?
2. ‘It is foolish for a seller to use second price sealed bid auctions because
the winner who bids £10,000 may end up paying only £l which is much
less than his bid.’ Comment.
3. Low quality workers have a value (constant marginal revenue product)
of £10,000 per year whereas the value of high quality workers is
£15,000 per year. Firms cannot distinguish between the two types
of workers but they know that there are as many low quality as high
quality workers.
a. How much are the workers paid assuming the labour market is
competitive?

201
28 Managerial economics

b. Assume firms offer workers a course on economics for management

during office hours. Although the course is offered free to the
workers, they experience a disutility of taking the course, mainly
because of the time it will take them to do the assignments. A high
quality worker will not have to spend too much time and for him
the disutility of taking the course is £5,000. For the low quality
workers the disutility is £20,000. Assume the workers expect a job
tenure of three years irrespective of whether they take the course.
Establish that there is a separating equilibrium and determine the
competitive wage for workers who take the course and for workers
who do not take the course, at this equilibrium.
c. i. Are the high quality workers better or worse off? By how much?
ii. Are the low quality workers better or worse off? By how much?
4. LSE Ltd. produces plastic statues of famous economists. The products
are made by hand in the Virgin Islands. LSE Ltd. has a plant in each of
the three main inhabited islands (St. Croix, St. Thomas and St. John)
with the following marginal cost (MC) schedules (in £):

Output per day MC St. Croix MC St. Thomas MC St. John

1 28 27 20
2 32 29 26
3 36 31 32
4 40 33 36
The statues are shipped to London and from there they are marketed in
several American and European university towns. Transportation costs
are negligible because all employees at LSE Ltd. have a contractual
obligation to take as many statues as required along on any business
trip. Although demand in Chicago is particularly inelastic, LSE Ltd.
cannot price discriminate. The demand function is p = 96 – 4q where p
is the price and q is the number of statues sold per day.1 1
Hint: Construct on
overall MC schedule for
a. How should LSE Ltd. price the statues? How many will it sell? LSE Ltd
b. Where should it manufacture the statues?

5. Abel & Co. is the price leader in the market for MM-disorder drugs.
The industry demand is given by Q = 90 – 8p. There are four ‘follower’
or ‘fringe’ firms who accept the price set by the dominant firm Abel &
Co. Each of these followers has long run cost function C(qf) = 10qf +
qf2. The leader has a constant returns to scale technology with long run
cost function C(q1) = 10q1.
a. Derive the followers’ supply curve.
b. Derive the net demand facing Abel & Co. and determine the optimal
price and output for the dominant firm.
c. How much is supplied by the followers?

202
 Appendix 1: Sample examination paper

Section B
Answer two questions from this section (25 marks each).
6. Lisa Listens supplies services for which the demand function is p(q) =
1– Q/2. The technology exhibits constant returns to scale. Labour is
the only variable input and the firm employs L workers. An individual
worker’s productivity cannot be costlessly observed but it is known
that the expected output of a shirking worker is 1 unit per hour (Q = L)
whereas that of a hard working employee is q (q > 1) units per hour
(Q = qL). The workers and the firm are risk neutral.
a. Draw the ‘marginal revenue product of labour’ curve for the case of
shirking employees and for the case of hard working employees. Do
these curves intersect? Give an interpretation of your findings.
b. At wage w = 1, how much labour is hired if Lisa Listens knows that
its workers will shirk?
c. Suppose Lisa Listens decides to induce its workers to work hard
by paying an efficiency wage and monitoring the workers. As a
consequence the cost per worker equals we+M(p) where we is
the efficiency wage and M(p) is the monitoring cost per worker
incurred to ensure a probability p of catching a shirking worker. A
worker suffers a disutility c from working. If she works hard, this
disutility is c = 0.75 whereas if she shirks it is c = 0. Lisa Listens
pays its workers we unless they are caught shirking in which case
they get nothing. A worker who is caught cheating cannot get a job
elsewhere. Workers have a choice between a job with Lisa Listens or
an outside low effort job (giving disutility c = 0) paying w = 1.
i. Given a detection probability p, what is the minimum wage
we Lisa Listens can pay and still attract workers if the workers
anticipate that they will have to work hard
ii. How large should we be to induce the workers to work hard?
d. Show that the solution to Lisa Listens’ minimum cost
implementation problem is p =1/7 and we = 21/4 for monitoring
cost M(p) =147 p/4. How much labour is hired in this scenario for
q=12?
7. Many charities offer differential membership fees. Depending on
whether the member wishes to receive the charity’s magazine or
whether only membership is desired, a different fee is charged. It is
usually not possible for non-members to purchase a subscription to the
magazine only. Assume that among potential members, willingness to
pay for membership ve is uniformly distributed on [0,E] and willingness
to pay for the magazine va is uniformly (and independently of va)
distributed on [0,A].
a. OUT, a charity for overworked university teachers, charges a
‘package’ price p for membership and a subscription to its magazine
Termtime Blues. Indicate on the [0, E] × [0, A] rectangle which
potential members are interested in this package. Determine
the revenue maximising price p and OUT’s optimal revenue per
potential member. Assume that 2A/3 < E < 3A/2 so that the optimal
price does not exceed A or E.
b. Now suppose OUT changes its membership policy and allows
potential members to buy membership only at a fee f or to ‘opt in’
and buy the magazine as well for a total cost of f + g. Indicate on
the [0, E] × [0, A] rectangle the potential members who opt in, those

203
28 Managerial economics

who are interested in membership only and those who will not
become members.
c. Interpret your findings.
8. An individual with wealth W and strictly concave von Neumann-
Morgenstern utility function U can face either of two states of nature.
In state 1 which occurs with probability 1-p he suffers no loss and
in state 2, which occurs with probability p, he suffers a loss V. The
individual has no control over probability p (there is no moral hazard).
a. An insurance firm offers the individual a policy (R, D), where R is
the premium and D is the payment the individual receives from the
insurance company in state 2. Write down the individual’s expected
utility and the (risk neutral) firm’s expected profit from this policy.
b. Show that if the insurance company maximises the individual’s
utility subject to its expected profit being nonnegative, full
insurance results.
c. Assume now there are two classes of individuals, low and high
risk, with probabilities of loss V of p1 and ph respectively. Although
individuals know their own risk, an insurer cannot distinguish
between high and low risk individuals. Explain why this may result
in only the high risk individuals being insured.
d. Given the scenario in (c) an insurance company decides to offer
a ‘menu’ of two policies: a low premium policy with a deductible
(R1,D1) where D1< V, designed for the low risk group and a high
premium full insurance policy (Rh,V) for the high risk group.
i. Assuming a perfectly competitive insurance industry, write down
the constraints on contract design.
ii. In what way does a low risk individual suffer from the existence
of high risk individuals?
9. Steve derives utility from Ben-and-Jerry’s ice cream (x1) and CDs (x2)
according to utility function U(x1, x2) – √x1+x2.
a. Given p1 and p2, the prices for tubs of Ben-and-Jerry’s ice cream
and CDs respectively, and Steve’s income m, solve his utility
maximisation problem.
b. The price of Ben-and-Jerry’s ice cream has recently gone up from
£2 per tub to £3 per tub. Calculate the equivalent variation and
the compensating variation of this price increase assuming the
price of CDs is £10 and m = 100.
c. Calculate the change in consumer surplus due to the price
change in (b).
d. Due to a MMC investigation into pricing practices in the music
business, the price of CDs decreases from £10 to £7. Calculate
the compensating variation of the combined price changes i.e the
price increase of ice cream and the price decrease of CDs
10. a. Discuss, using graphs where necessary, how the demand for an
input by a competitive industry differs from the horizontal sum
of its firms’ individual input demands.
b. Under World Bank pressure, the Zambian state marketing
monopoly for maize was recently dismantled. Its role was
subsequently taken over by a handful of private companies
which, according to Oxfam reports, pay farmers less than half
the old state price. The World Bank argues that farmer prices will

204
 Appendix 1: Sample examination paper

be driven up when competition among middlemen increases. Do

you agree with the World Bank’s prediction?
11. Publishers prefer that their books are sold at not less than their
recommended retail prices. This seems a puzzle because when retailers
set lower prices we would expect the manufacturer to sell more and
make higher profits. Why do some manufacturers want to control the
retail price of their products?
12. a. Give a moral hazard explanation for the phenomenon of
promotion in the internal labour market.
b. Discuss Frank’s theory of wage compression and status.
c. Why do Chief Executive Officers (CEOs) earn so much and is this
a good thing?

END OF PAPER

205
28 Managerial economics

Notes

206
 Appendix 2: References for older versions of textbooks

Appendix 2 References for older versions

of textbooks

The following earlier versions of books, which might be easier to borrow

from libraries than the most recent versions, also contain the relevant
material for this course:

Main texts
Varian, H.R. Intermediate Microeconomics. (New York: W. W. Norton & Co.,
2003) sixth edition [ISBN 9780393978308].
Varian, H.R. Intermediate Microeconomics. (New York: W. W. Norton & Co.,
1999) fifth edition [ISBN 9780393973709].

Other books
Gravelle, H.S.E and R. Rees Microeconomics. (London: Longman, 1992) second
edition [ISBN 0582023866].

Chapter and page references for older books

Chapter 1
Page 9: Main texts:
• Varian, sixth edition, Chapter 12
• Varian, fifth edition, Chapter 12

Chapter 2
Page 21: Main texts:
• Varian, sixth edition, Chapters 28 and 29
• Varian, fifth edition, Chapter 28

Chapter 4
Page 43: Main texts:
• Varian, sixth edition, Chapter 36
• Varian, fifth edition, Chapter 36

Chapter 5
Page 57: Main texts:
• Varian, sixth edition, Chapter 17
• Varian, fifth edition, Chapter 17

Chapter 6
Page 73: Main texts:
• Varian, sixth edition, Chapters 2, 3, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15
• Varian, fifth edition, Chapters 2, 3, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15
Page 73: References cited:
• Gravelle and Rees, second edition, Chapters 4, 20
Page 73: Margin note 1:
• Varian, sixth edition, Chapters 2, 3, 4, 5, 6;

207
28 Managerial economics

• Varian, fifth edition, Chapters 2, 3, 4, 5, 6

Page 74: Margin note 2:
• Varian, sixth edition, 29-31
• Varian, fifth edition, 29-31
Page 74: Margin note 3:
• Varian, sixth edition, Chapter 8
• Varian, fifth edition, Chapter 8
Page 77: Margin note 4:
• Varian, sixth edition, Chapter 14
• Varian, fifth edition, Chapter 14
Page 77: Margin note 5:
• Gravelle and Rees, second edition, 116-23
Page 78: Margin note 6:
• Varian, sixth edition, Chapter 15
• Varian, fifth edition, Chapter 15
Page 80: Margin note 10:
• Varian, sixth edition, Chapter 12
• Varian, fifth edition, Chapter 12
Page 82: Margin note 11:
• Gravelle and Rees, second edition, 586-94
Page 82: Margin note 12:
• Varian, sixth edition, Chapter 10
• Varian, fifth edition, Chapter 10
Page 85: Margin note 14:
• Varian, sixth edition, Chapter 9
• Varian, fifth edition, Chapter 9
Page 88: Margin note 20:
• Varian, sixth edition, Chapter 13
• Varian, fifth edition, Chapter 13

Chapter 7
Page 95: Main texts:
• Varian, sixth edition, Chapters 18, 19, 26
• Varian, fifth edition, Chapters 18, 19, 26 Chapter 9
Page 127: Main texts:
• Varian, sixth edition, Chapters 22, 23, 24
• Varian, sixth edition, Chapters 22, 23, 24 Chapter 10
Page 141: Main texts:
• Varian, sixth edition, Chapter 25
• Varian, fifth edition, Chapter 25

208
 Appendix 2: References for older versions of textbooks

Chapter 11
Page 173: Main texts:
• Varian, sixth edition, Chapter 27
• Varian, fifth edition, Chapter 27
Page 189: Margin note 17:
• Gravelle and Rees, second edition, Chapter 12

209
28 Managerial economics

Notes

210
 Appendix 3: Maths checkpoints

Appendix 3: Maths checkpoints

This appendix was written by Till Fellrath and first appeared as an

appendix to the subject guide in 2000.
This course does not require any advanced maths skills. You are, however,
expected to have a good working knowledge of some main concepts,
all of which were introduced in the first year Mathematics courses. It is
absolutely vital that you are completely familiar with these before you
start this course and if you encounter any mathematical problems you
should address them immediately. In particular, you are strongly advised
to check critically your understanding in the areas listed below.

1. Functions – a few general remarks

Functions indicate a relationship between two or more variables. They
are one of the central tools in applied economics and are used throughout
the subject guide. For instance, a demand function models the relation
between price and quantity within a particular market. It indicates the
demand for a particular good at any given price.
If q(p) = 100 – 4p, we can see that the market demand q at a price p = 10
will be q = 100 – (4 × 10) = 60. In this function, quantity is the explained
variable, with the price being the explaining variable. This function
answers the question: What is the market demand at a price p?

Inverse:
Any function can be expressed in its inverse form by changing the
explained and the explaining (endogenous) variable. Mathematically this
means solving the equation for the other variable. Formally the inverse of
a function f is indicated as f –1.
For the above demand function we get
4p = 100 – q
p(q) = 25 – (¼)q
We can see that for a market demand q = 60, the resulting market price
p = 25 . (1/4 × 60) = 10. Of course the relation between the two variables is
still the same, but the logic is now reversed. This function now answers the
question: What is the price p that could be charged (i.e., by a company)
for a specific market demand q?
Very often it is necessary to find the inverse of the demand function
in order to calculate a company’s revenues. Similarly to calculate the
marginal revenue product function (refer to section 5 of this appendix)
the inverse of the Production function needs to be derived (subject guide
reference: Chapter 7, Production, p.100, firm demand for inputs).

Variables and constants:

It is important to distinguish between variables and constants. For a cost
function C(q) = 100 + aq, the figure 100 indicates a level of fixed cost. It
is a constant and occurs regardless of the production level q. The letter ‘a’
represents a constant exogenous factor. Although we do not know how
large ‘a’ is, we cannot influence or choose its level. The letter ‘q’ denotes
the explaining variable. The company can choose its production level
q. Functions are sometimes written down as q(p), stating the explaining

211
28 Managerial economics

variable(s) in brackets. Often, this is omitted. For any function, students

should check that they understand what the explaining and the explained
variable is, distinguishing them clearly from any constants.

General suggestions for studying and revising:

The following suggestions might be useful for students who are uncertain
about the idea of a function. They might facilitate a general understanding
and avoid confusions when attempting to solve sample questions.
• Translate the function in plain English. For example a demand function
q(p) = 100 – 4p could be read as follows: The market demand ‘q’ for a
particular good equals a constant factor 100 minus 4 times its market
price ‘p’.
• Distinguish between the explained variable, the explaining variable(s)
and constant factors. It might be a good idea to clearly write this down,
or to at least spot the explaining variables where this is not stated
explicitly.
• Check if you understand how this function works by substituting some
simple values for the explaining variable(s).
• Where a function includes some unknown constants, i.e. if q(p) =
100 – ap, simplify this function for study purposes by substituting the
constant for a particular value, i.e. a = 3, so that the function reads as
q(p) = 100 – 3p.
• Illustrate a function graphically if you are uncertain about its form.

2. Differentiation
Students are expected to be able to differentiate functions of several
variables. Typically this is done to find the maximum value a function can
reach and to identify the corresponding value for its explaining variable(s).
For example, a company’s profit function can be maximised to find the
profit optimising output quantity (see section 9). Below is a summary
of the main rules and the derivatives of some specific functions. Partial
derivatives are found by treating the other variables as a constant.

Main rules:
Constant factor y = a.f (x)y'=a. f '(x)
Example: y = 5x 5
a = 5 and f(x) = x5
y' = 5x 5x4 = 25x 4

Summation y = f(x) + g(x) y' = f '(x) + g'(x)

Example: y = x +x
7 5
f(x) = x' and g(x) = x5
y' = 7x 6 + 5x4

Products y = f(x) . g(x)

y' = f '(x) . g(x) + f(x) . g'(x)
Example y = x2(x3 + 7x - 1) f(x) = x2 and g(x) = (x3 + 7x – 1)
y' = 2x(x3 + 7x – 1) + x2(3x2 + 7)
= 5x4 + 21x2 – 2x
Fractions y = f(x)/g(x)
y' = [f '(x).g(x) – f(x).g' (x)] / [g(x)]2
212
 Appendix 3: Maths checkpoints

Example: f(x) = x3 and g(x) = x2 -7

y' = [3x2 (x2 – 7) – x3 (2x)]/[(x2 – 7)2]
= (x4 – 21x2)/(x2 – 7)2

Chain rule y = f(g(x)) y' = f ' (g(x)).g'(x)'

Example: y = (x – 7x )
2 3 21

f(u) = u21 and g(x) = x2 – 7x3

y' = 21(x2 – 7x3)20 (2x – 21x2)

Derivatives of some main functions

y = c (constant) y' = 0
y=x n
y' = nxn-1
y = x½ [= √x] y' = ½x(-½) [= ½(1/√x) = 1/(2√x)]
y = ex y' = ex
y = ax y' = ln a.ax [e.g. y = 2x y' = ln 2.2x]
y = ln x y = 1/x
y = alog x y' = 1/(ln a) . 1/x

Formal notations:
Differentiations can be expressed in several ways, all meaning exactly the
same. For example, for a cost function C(q):
C'(q) Typically used for functions with one variable
∂C(q)/∂q Often used for partial differentiations of functions with two
variables
MC(q) MC = Marginal Cost, implies the differentiation of a total
cost function C(q)

Some applications in the subject guide:

Chapter 2, Game Theory, page 29 (mixed strategy equilibrium)
Chapter 5, Auctions, page 62 (optimal bid V)
Chapter 6, Consumer Theory, page 75 (Slutsky equation)
Chapter 7, Production, page 101 (Monopsony)
Chapter 9, Market Structure, page 138 (Dorman-Steiner model)
Chapter 10, Monopolistic Pricing Practices (virtually in every model)
Chapter 11, Oligopoly (all models)

The concept of differentiating a function is used very frequently

throughout the subject guide and the cited applications are just a few
examples. Students are expected to be at ease with this fundamental
concept of calculus (check the old exam papers for the level of
difficulty that is expected). Anyone who does not feel comfortable with
differentiation should revise the relevant material of their first year course
in Mathematics or check out any calculus book.

213
28 Managerial economics

3. Logarithmic functions / properties of In and exp.

log uv = log u + log v Example: ln(2x) = ln 2 + ln x
log u/v = log u - log v Example: ln(2/x)= ln(2) – ln(x)
log u = n. log u
n
Example: log 2x = x. log 2
log ul/n = (1/n).log u Example: log x1/2 = (1/2). log x
(remember that √x = x1/2)
Although these rules are not central to the understanding of the unit,
it might be necessary to apply these rules in order to simplify given
functions, for instance in order to differentiate a logarithmic function, or
to solve a system of equations.

Some applications in the subject guide

Chapter 1, Decision Analysis, page 19, exercise 6 (risk attitude)
Chapter 4, Asymmetric Information, page 51 (utility curves)
Chapter 6, Consumer Theory, page 84 (utility curves)

4. Integration
To calculate the area under a function, which can for instance represent
consumer surplus, students need to understand integration as the
‘opposite’ of differentiation. Integration is used to calculate the area below
a function, for instance, to find the consumer surplus. In general, integrals
between the borders a and b can be calculated as follows
b

∫
a
f(x) dx = F(b) - F(a)

for any integral function.

Example:
Demand is given by a function D: q (p) = 100 – 2p. The consumer surplus
equals the area between the price and the demand curve. For price 40,
consumer surplus is represented by the area below the demand curve and
above p = 40. At the price of 40, the market demands a quantity q = 20.
50

∫ (100 - 2p) dp = [100p - p ]

40
2 50
40

= (100 × 50 – 502) – (100 × 400 – 402)

= 2500 – 2400
= 100
What is the gain in consumer surplus if the price p drops from 40 to 30?
The new consumer surplus at p = 30, where quantity demanded equals
q = 40, can be calculated as
50

∫ (100 - 2p) dp = [100p - p ]

30
2 50
30

= (100 × 50 – 502) – (100 × 30 – 302)

= 2500 – 2100
= 400

214
 Appendix 3: Maths checkpoints

Thus, the gain in consumer surplus equals the difference between 400 –
100 = 300. Alternatively it can directly be calculated as
40

∫ (100 - 2p) dp = [100p - p ]

30
2 40
30

= (100 × 40 – 402) – (100 × 30 – 302)

= 2400 – 2100
= 300
For linear functions, it might be easier to use simple geometry to calculate
the integral. For example, the consumer surplus at price p = 30 could be
calculated as the triangle under the demand function and above the price.
At the ‘y’-axis we get the length of (50–30). At the ‘x’-axis the demanded
quantities at the two prices equal q = 0 for p = 50 and q = 40 for p = 30, thus
we get the length (40 – 0). The area of this triangle can now be calculated
as (50 – 30) × (40 – 0) × ½ = 400.

Example:
If a firm’s marginal cost function is given as MC (q) = 10, it is clear that
marginal cost (or the incremental production cost for the next unit)
is a constant 10 for any given production unit. Thus, total cost can be
calculated as the product of 10 times the level of production, or 10q, plus
any fixed cost F which occurs regardless of the level of production. The
total cost function equals C (q) = 10q + F, which is the integrated marginal
cost function or ∫MC. And marginal cost is the differentiated total cost
function. (See for instance Chapter 11, Oligopoly, p.177).
Similarly, if MC (q) = 10 + 4q + 6q2, then ∫MC =∫(10 + 4q + 6q2) dq = C(q) =
10q + 2q2 + 2q3 + F

Some applications in the subject guide:

Chapter 5, Auctions and Bidding, page 68 (variance of winning bids)
Chapter 6, Consumer Theory, page 78 (consumer surplus)

5. Systems of equations / manipulating equations

To find a maximum of a function with two variables, partially differentiate
with respect to each variable and set the results equal to zero. These two
optimality conditions have to be satisfied at the same time. Thus, they
represent a system of two equations with two variables. The values for
each variable can be found for example by substituting one equation
into the other. Students are expected to be comfortable with algebraic
manipulation of equations.

Some applications in the subject guide:

Chapter 1, Decision Analysis, page 19 exercise 6 (utility curves)
Chapter 7, Production, page 107 (a vision of output among plants)
Chapter 10, Monopolistic Pricing Practices, page 147 (price discrimination)
Chapter 11, Oligopoly, page 183 (collusion)

215
28 Managerial economics

6. Uniform distribution
The concept of uniform distribution is used in Chapter 5, Auctions and
Bidding. It describes the way in which a sample of data is distributed
within a certain interval.

Example:
What are the expected valuations of bidders in an auction, if we know the
number of bidders as well as the interval from which their valuations are
drawn?
If the valuation of n bidders is uniformly distributed on [0,1] we can make
the following predictions:
n = 1, expected valuation: ½ (middle of the interval)
n = 2, expected minimum and maximum valuations: 1/3 and 2/3
n = 3, expected minimum, middle and maximum valuations: ¼, ½ and ¾
Thus, when n > I the expectation of the valuation of the highest bidder can
be calculated as (n)/(n + 1), of the second highest as (n – 1)/(n + l), and so
on.
If the distribution is not normalised, for instance for an interval [0, 5]
the expected valuations of 4 bidders would be at equal distances on the
interval [0....v1....v2.... v3….v4….5], or 1, 2, 3, and 4 respectively. The highest
valuation can be calculated as
(n)/(n + 1) × 5 = (4)/(4 + 1) × 5 = 4
For uniformly distributed valuations between [10, 15], the expected
valuations of 4 bidders are expected to be at equal distances on the
interval
[10.... v1….v2.... v3.... v4.... 15] or 11, 12, 13, 14 respectively.
Thus, the expected highest valuation can be calculated as
10 + (n)/(n + l) × 5 = 14
In general for n bidders with uniformly distributed preferences on [L,U]
the expected highest valuation equals
L + (n)/(n + 1) . (U – L)

Some applications in the subject guide:

Chapter 5, Auctions and Bidding, page 64, (auction revenue)
Chapter 5, Auctions and Bidding, page 71, sample questions 1, 5, and 6.

7. Probabilities
Probabilities are often denoted by the variable ‘p’ which represents a
number between 0 and 1. If a decision maker faces two (or more) possible
scenarios, which he cannot influence, assumptions could be made about
the likelihood of each outcome. The probabilities of all outcomes must add
up to 1 (representing 100%). Suppose one in ten bikes gets stolen each
year. Outcome 1, bike gets stolen, occurs with probability p = 0.1 (equal to
10%) and outcome 2, bike does not get stolen, occurs with probability
(1 – p) or (1 – 0.1) = 0.9 (equal to 90%).

216
 Appendix 3: Maths checkpoints

Some applications in the subject guide:

Chapter 1, Decision Analysis, page 11, example 1.1 and 1.2 (decision tree)
Chapter 2, Game Theory, page 29, example 2.7 (mixed strategy equilibrium)
Chapter 4, Asymmetric Information, page 47, example 4.4 (demand for insurance)
Chapter 6, Consumer Theory, page 80 (state-contingent commodities model)
Chapter 8, Labour Economics, page 115 (efficiency wage model)

To understand Chapter 5, Auctions and Bidding, students need to be able

to perform basic probability calculations, which go back to the concept of
the uniform distribution.

Example:
One bidder wants to buy a rare painting which he values at $400,000.
He does not know what the value is to the owner, but he knows that it is
uniformly distributed between $100,000 and $500,000. The bidder can
only make one offer, which the owner can either accept or reject. What is
the optimal offer that the bidder should make?
The owner’s valuation is said to be uniformly distributed, meaning that
any point in the interval [U, L] = [100,000; 500,000] is equally likely to be
the valuation. The painting will only be sold if the submitted bid exceeds
the owner’s valuation, otherwise the owner would consider the bid to be
too low. If the bidder offers $500,000 he would definitely win the auction,
but his ‘gain’ would be a negative $100,000 as the bid would be above
his own valuation. Thus, the bidder faces a basic trade-off between the
potential gain and the likelihood of the painting being sold. He can lower
his bid, but at the same time the probability of his bid being larger than
the owner’s valuation decreases and he becomes less likely to win the
auction. The bidder’s expected gain can be calculated as the gain from
buying the painting multiplied with the probability of winning.
(v – b) . P(win)
(v – b). [(b – L) /(U – L)]
(400,000 – b) . [(b –100,000) / (500,000 – 100,000)]
To understand the transformation of the owner’s distribution to the
probability values, consider a few examples:
Probability of
Bid Gain Expected gain
winning
500,000 -100,000 1 -100,000
400,000 0 0.75 0
300,000 100,000 0.5 50,000
200,000 200,000 0.25 50,000
100,000 300,000 0 0
By interpreting the above equation as a function of the variable b, we can
calculate the optimal bid by differentiating the function with respect to b.
Expected gain (b) = (400,000 – b) . [(b – 100,000) / 400,000]
∂Expected gain / ∂b = –1[(b – 100,000)/400,000] + (400,000 – b).(1/400,000)
(product rule!)
Set equal to 0 and solve for b to find optimal bid
– 1/400,000. b +1/4 + 1 – (1/400,000 . b) = 0
5/4 = (1/200,000) . b

217
28 Managerial economics

b = 250,000
For the optimal bid of 250,000 the expected gain equals 56,250.

8. The discount factor ‘δ’

The discount factor δ represents the value of money over time. As used
in this subject guide δ equals any number between 0 and 1. Given the
alternative between either accepting $100 now, or waiting for the same
amount for another year, an individual would most likely take the money
now. But the same individual might accept $125 in the following year,
rather than $100 now. If this individual would be exactly indifferent
between $100 now, or $125 one year later, we can calculate his personal
discount factor  as 100/125 = 0.8. In other words, money loses its value
by 20% each year. To be compensated for the wait, this consumer would
want to receive $125 which equals in ‘today’s money’ $125 x 0.8 = $100.
Thus, a large discount factor close to 1 represents a rather small discount
and money retains a high proportion of its value in subsequent periods.
Similarly, a small discount factor close to 0 indicates a very large discount
in that money loses its value very rapidly.
It is possible to calculate the present value of future payoffs. Suppose that
the individual receives $100 every year for a very large number of years.
How much would this be worth in today’s money? If one does not account
for the value of money over time, the answer would simply be $100
multiplied by the number of years. But if one includes a discount factor in
the analysis, e.g.  = 0.8, the result changes as follows:

Discounted Value Sum of discounted

Year Face Value $
=0.8 values
1 100 100 100
2 100 100 = 80 180
3 100 1002 = 64 244
4 100 100 = 51.2
3
295.2
5 100 1004 = 40.96 336.16
6 100 100 = 32.768
5
368.928
Converges to 100 x 1/
∞ 100 100∞ = 0
(1 – ) = 500
In general: 1 + d + d2 + d3 + ... = 1/(1 – d) (to aggregate all periods including
period one)
d + d2 + d3 + ... = d (1 + d + d2 + d3 +...) = d/(1 – d) (to aggregate from period
two onwards)
The discount factor can be interpreted in several ways, such as the rate of
inflation, the interest rate for outside investment, or simply the patience of
an individual.

Some applications in the subject guide:

Chapter 3, Bargaining, page 39 and exercise on page 42 (alternating offers bargaining)
Chapter 10, Monopolistic Pricing Practices, page 152 (skimming)
Chapter 11, Oligopoly, page 187 and exercise I (dynamic interaction)

218
 Appendix 3: Maths checkpoints

9. The company’s profit function

Chapters 7–11 in the subject guide consider a company’s decision
problems from various angles. The basic assumption is that a company
wants to maximise profits. On the one hand the company receives
revenues for the products it sells in the market. On the other hand it faces
costs which typically increase if the level of production is increased. The
company’s profits can therefore be expressed by a profit function as the
difference between revenues (typically as a function of the production
level q) and costs (typically also as a function of production level q). Thus
profits depend on the production level:
П(q) = Revenues(q) – Costs(q)
If revenues exceed costs, the company makes a profit. Should cost be
larger than the revenues, the company makes a loss.
In order to find the optimal production level, we need to find the
function’s maximum value. Thus, we can differentiate this function with
respect to q.
∂П/∂q = ∂Revenues/∂q – ∂Costs/∂q
set this derivative equal to zero
∂Revenues/∂q – ∂Costs/∂q = 0
or Marginal Revenue (q) = Marginal Cost (q)
or MR= MC
and solve for the profit optimising value of the variable q. Any problem
can thus either be solved by differentiating a (total) profit function or by
directly applying one of the commonly known optimality conditions, such
as Marginal Revenue = Marginal Cost. It is important, however, never to
confuse the two levels of analysis, i.e. total or marginal (differentiated)
values.

Example:
If demand function is given as D: q = 100 – 4p
Marginal cost = 12q
(i) starting with the firm’s profit function П(q) = Revenue(q) – costs(q)
Revenues: Total Revenues TR (q) = price x number of units sold, or p.q, or (25-
1/4q) .q or 25q – ¼q2
Costs: Marginal Cost MC = 12q (given), Total Cost C(q) = ∫MC = 6p2 + F
П(q) = Revenues (q) – Costs (q)
П(q) = 25q – ¼q2 (6q2+F)
П(q) = 25q – 6.25q2 – F
∂П/∂q = 0
25 – 12.5q = 0
q=2
Thus, the firm should produce 2 units, which gives it a maximum profit of
П(q) = 25q – 6.25q2 – F
П(2) = (25 × 2) – 6.25(2)2 – F
П = 50 – 25 – F
П = 25 – F

219
28 Managerial economics

The market price can be found by substituting q back into the (inverse)
demand function
p = 25 – (¼)q
p = 25 – (¼ × 2)
p = 24.5
(ii) applying the optimality condition ‘MR = MC’
MR: Derive Marginal Revenue from the Demand function
- Find the inverse demand curve as p = 25 – (¼)q
- Total Revenue (TR) = price x number of units sold, or p.q, or (25 – (¼)q) . q or
25q – (¼)q2
- Marginal Revenue (MR) = ∂TR/∂q (25-q (¼)q2) = 25 – (½)q
MC: Given as 12q
MR = MC
25 – (½)q = 12q
q=2

General suggestions for studying and revising

• Check the models in Chapters 7-11 and study the way in which they
modify this basic example.
• Check whether you are able to understand both the profit function and
the optimality condition for each model.
• Be absolutely certain what variable(s) you are solving for in each
model, for example input ‘L’ or outputs ‘q1’ and ‘q2’.
• Check whether an inverse function needs to be found, typically for
marginal revenue or marginal revenue product curves.
• Check whether you understand all of the curves listed below and how
they relate to one another.

Overview and definitions of some important functions

П Profits The difference of TR and TC.
TC Total Cost Includes all Cost Factors.

MC Marginal Cost Incremental cost of one additional unit

differentiation of TC.
AC Average Cost Average total cost to produce one unit, TC
divided by output, MC intersects AC at its
minimum.

FC Fixed Cost Independent of level of production,

constant factor of TC.
VC Variable Cost Increases with output, variable fraction of
TC.
AVC Average Variable Average variable cost per unit, VC divided
Cost by output.
TE Total Expenditure Cost of one particular input factor, e.g.
labour.

220
 Appendix 3: Maths checkpoints

ME Marginal Incremental expenditure if one additional

Expenditure input unit (e.g. one worker) is demanded,
differentiation of TE, increasing for
monopsonist, constant (e.g. wage) for
competitive input market.

TR Total Revenue Number or units sold multiplied by the

price.

MR Marginal Revenue Incremental revenue if one additional unit

is sold, differentiation of TR with respect
to quantity, decreasing for monopolist,
constant (price) for competitive firm.
MP Marginal Differentiation of production function with
Productivity respect to input, typically decreasing, e.g.
productivity of labour.

MRP Marginal Revenue MR x MP, could be constant, typically

Product decreasing, used to solve input problems
such as monopsony model.

10. A few hints on solving questions

Very often students find it difficult to understand sample questions and to
relate them to the existing theory. The following example gives a few ideas
on how a question can be attempted in a structured manner. Secondly, it
shows that very often there is more than one way to arrive at the correct
solution. As long as the reasoning and the results are correct, students are
free to choose any alternative that suits them best.

Example
Tom, Dick and Harry manage a company which markets balloons. The balloon market is
perfectly competitive and the price of balloons is $5 per unit. The amount of capital is
fixed and the investment in capital can be considered a sunk cost. The cost of material
inputs is negligible but labour is expensive. Tom has asked Dick to gather some data on
the productivity of labour. Dick presents him with the figures in the middle of the table
below. Meanwhile Tom has asked Harry to get some information about labour costs.
Harry told him that these costs would depend on how many workers were employed, as
indicated in the table below on the right. Tom now has to decide how many workers to
hire. What should he do and what will the wage rate be?
Number of Total number of
Wage rate
workers balloons per hour
0 0 10
1 10 11
2 15 12
3 19 15
4 21 17
5 22 20
6 20 25

Step 1
Read the question carefully!

Step 2
Write down the most important information.
221
28 Managerial economics

• Company wants to maximise profits.

• Output market for balloons: Perfect Competition, p=5 per unit.
• Input market for balloons: number of workers drives the wage rate up, thus the
company is a monopsonist (it is the only buyer of labour).
• Cost other than labour can be neglected.

Step 3
Relate it to the studied material.
• Company is a Monopsonist facing perfect competition in its output market.
• Chapter 7, Production, pages 98-103.

Step 4
Solve the question
• Plan ahead and be certain about what you are looking for. Here we need to find the
optimal number of workers, variable ‘L’ (explicitly asked in the question)!
• Remind yourself of the analytical steps required to solve the question.

Solution Alternative 1:
Solve in Total Values, finding maximum profit
L Total Revenue (p.q) Total Expenditure (L.w) Profits II
p q Rev. L w Exp. (Rev – Exp.)
0 5 0 0 0 10 0 0
1 5 10 50 1 11 11 39
2 5 15 75 2 12 24 51
3 5 19 95 3 15 45 50
4 5 21 105 4 17 68 37
5 5 22 110 5 20 100 10
6 5 20 100 6 25 150 –50
Answer: Hiring two workers and paying them a wage of 12 gives the maximum profit
of 51.

222
 Appendix 3: Maths checkpoints

Solution Alternative 2:
Solve in Marginal Values, applying MRP = ME rule (company should continue to expand
the workforce as long as the additional worker’s revenue contribution is larger than the
incremental expenditure).

Marginal Revenue Product Marginal Expenditure

L
(Marginal Revenue x Marginal Productivity) (additional expenditure)

Total Marginal Total Additional Profit

MR MRP ME
Productivity Productivity Expenditure (MRP-ME)

0 5 0 0 0 0 0 0

1 5 10 10–0 = 10 5x10=50 11 11–0=11 50–11= 39

2 5 15 15–10 = 5 5x5=25 24 24–11=13 25–13= 12

3 5 19 19–15 = 4 5x4=20 45 45–24=21 20–21= –1

4 5 21 21–19 = 2 5x2=10 68 68–45=23 10–23= –13

5 5 22 22–21 = 1 5–32= –27

6 5 20 20–22 = -2 5x(–2)= –10 –10–50= –60

Answer: Workers are hired as long as MRP is larger than ME. Thus, two workers should
be hired at wage 12. (If the third worker would be hired, MRP would be very close to ME.
But since ME exceeds MRP by exactly 1, total profits would be driven down by 1 if the
company hires worker number 3.)

Solution Alternative 3:
It is also possible to derive the actual functions from the data in the table and to proceed
analytically. Clearly this would be more complicated than the two ways suggested
above. If a table lists a limited number of possibilities, it is usually quicker to check each
alternative to find the optimal solution.

223
 Notes

224
 Notes

Notes

225
 Notes

226