Very Extensive Lectures and The Book Economics of Markets and Organizations

lOMoARcPSD|1733731
Economics of Markets and Organizations
Lecture notes
20132014
Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

lOMoARcPSD|1733731
Table of Contents
Economics of Markets and Organizations................................................................................................................1
Lecture notes...................................................................................................................................................................1
20132014 ..................................................................................................................................................................1
Part I: Introduction............................................................................................................................................................7
Chapter 1: Organizations and efficiency...................................................................................................................8
1.1 Efficiency ...................................................................................................................................................................8
1.2 General equilibrium ........................................................................................................................................... 10
1.3 Perfect competition............................................................................................................................................ 11
1.4 Market failures ..................................................................................................................................................... 13
1.5 Set-up and research methods ........................................................................................................................ 14
1.6 Case study: Apple Inc. ....................................................................................................................................... 15
Bibliographical Notes ............................................................................................................................................... 18
References ..................................................................................................................................................................... 19
Exercises ........................................................................................................................................................................ 19
Chapter 2: Organizational Economics .................................................................................................................... 22
2.1 Hiring decisions ................................................................................................................................................... 22
2.2 The Principal-Agent problem ........................................................................................................................ 23
2.3 Critical assumptions in the Principal-Agent model .............................................................................. 24
2.4 Case study: Moral hazard in the mortgage market and the financial crisis of 2008 ............... 25
References ..................................................................................................................................................................... 28
Exercises ........................................................................................................................................................................ 29
Chapter 3: Industrial Organization .......................................................................................................................... 31
3.1 Market power ....................................................................................................................................................... 31
3.2 Market concentration........................................................................................................................................ 33
3.3 Monopoly: The inverse elasticity rule ........................................................................................................ 35
3.4 Government intervention ................................................................................................................................ 37
3.5 Case study: Cartels and market power ...................................................................................................... 39
References ..................................................................................................................................................................... 41
Exercises ........................................................................................................................................................................ 41
Part II: Strategic interaction ....................................................................................................................................... 44
Chapter 4: Game theory................................................................................................................................................ 45
4.1 The prisoners dilemma ................................................................................................................................... 45

lOMoARcPSD|1733731
4.2 Nash equilibrium ................................................................................................................................................ 46

4.3 The Hotelling game ............................................................................................................................................ 48
4.4 Case study: Game theory in football Do professionals play Nash? ............................................. 49
References ..................................................................................................................................................................... 50
Exercises ........................................................................................................................................................................ 51
Chapter 5: Team incentives ........................................................................................................................................ 54
5.1 The model .............................................................................................................................................................. 54
5.2 Paying for team performance ........................................................................................................................ 55
5.3 Comparative performance evaluation ....................................................................................................... 56
5.4 Case study: The downside of relative performance pay ..................................................................... 59
References ..................................................................................................................................................................... 61
Exercises ........................................................................................................................................................................ 61
Chapter 6: Oligopoly ...................................................................................................................................................... 65
6.1 Bertrand competition........................................................................................................................................ 65
6.2 Cournot competition ......................................................................................................................................... 67
6.3 Other ways to resolve the Bertrand paradox .......................................................................................... 72
6.4 Case study: Why do airplanes arrive late?................................................................................................ 74
References ..................................................................................................................................................................... 75
Exercises ........................................................................................................................................................................ 76
Part III: Dynamic interaction ...................................................................................................................................... 78
Chapter 7: Dynamic games.......................................................................................................................................... 79
7.1 Subgame perfect Nash equilibrium ............................................................................................................. 79
7.2 Pricing in the vertical chain ............................................................................................................................ 81
7.3 Case study: How chess players fail to backward induct ..................................................................... 83
References ..................................................................................................................................................................... 85
Exercises ........................................................................................................................................................................ 86
Chapter 8: Optimal incentive contracts ................................................................................................................. 88
8.1 The optimal incentive contract ..................................................................................................................... 88
8.2 Risk sharing ........................................................................................................................................................... 90
8.3 Case study: Pay enough or dont pay at all ............................................................................................... 92

lOMoARcPSD|1733731
References ..................................................................................................................................................................... 94
Exercises ........................................................................................................................................................................ 95
Chapter 9: Market entry and product positioning ............................................................................................ 98
9.1 Market entry ......................................................................................................................................................... 98
9.2 Product positioning ........................................................................................................................................ 101
9.3 Case study: General Motors road to success ....................................................................................... 103
Bibliographical Notes ............................................................................................................................................ 105
References .................................................................................................................................................................. 105
Exercises ..................................................................................................................................................................... 105
Part IV: Repeated interaction.................................................................................................................................. 108
Chapter 10: Repeated interaction and the value of revenge ...................................................................... 109
10.1 A simple example .......................................................................................................................................... 109
10.2 Infinitely repeated games and the trigger strategy ........................................................................ 111
10.3 Analyzing the stability condition ............................................................................................................ 113
10.4 Case study: Within-game trigger strategies in spectrum auctions ........................................... 114
References .................................................................................................................................................................. 116
Exercises ..................................................................................................................................................................... 117
Chapter 11: Relational contracts ........................................................................................................................... 119
11.1 What are relational contracts? ................................................................................................................ 119
11.2 Bulls model ..................................................................................................................................................... 120
11.3 Case study: Large Stakes and Big Mistakes Why big bonuses dont always work.......... 124
References .................................................................................................................................................................. 126
Exercises ..................................................................................................................................................................... 126
Chapter 12: Collusion in markets .......................................................................................................................... 129
12.1 Collusion in markets: Why, how, and when ....................................................................................... 129
12.2 Factors facilitating collusion .................................................................................................................... 131
12.3 Collusion and competition policy ........................................................................................................... 134
12.4 Case study: Collusion beyond theory .................................................................................................... 136
References .................................................................................................................................................................. 138
Exercises ..................................................................................................................................................................... 138
Part V: Information asymmetry ............................................................................................................................. 141
Chapter 13: Games with asymmetric information ......................................................................................... 142

lOMoARcPSD|1733731
13.1 The adverse selection problem ............................................................................................................... 142

13.2 Screening .......................................................................................................................................................... 143
13.3 Signaling............................................................................................................................................................ 145
13.4 Case study: How to distinguish a good seller from a bad one on eBay ................................... 146
References .................................................................................................................................................................. 148
Exercises ..................................................................................................................................................................... 148
Chapter 14: Hiring and firing .................................................................................................................................. 150
14.1 Credentials ....................................................................................................................................................... 150
14.2 Assessments .................................................................................................................................................... 150
14.3 Probation .......................................................................................................................................................... 151
14.4 Incentive contracts ....................................................................................................................................... 152
14.5 Case study: Performance pay as a screening device....................................................................... 153
References .................................................................................................................................................................. 154
Exercises ..................................................................................................................................................................... 155
Chapter 15: Price discrimination........................................................................................................................... 158
15.1 First-degree price discrimination .......................................................................................................... 158
15.2 Third-degree price discrimination ........................................................................................................ 160
15.3 Second-degree price discrimination ..................................................................................................... 161
15.4 Price discrimination and competition policy .................................................................................... 164
15.5 Case study: Damaged goods Lets make things worse ............................................................... 165
References .................................................................................................................................................................. 167
Exercises ..................................................................................................................................................................... 168
Part VI: Commitment .................................................................................................................................................. 171
Chapter 16: The value of commitment ................................................................................................................ 172
16.1 The value of commitment .......................................................................................................................... 172
16.2 First-mover (dis)advantages .................................................................................................................... 173
16.3 The hold-up problem ................................................................................................................................... 174
16.4 Case study: Should the European Central Bank serve as a lender of last resort in order to
save the euro? ........................................................................................................................................................... 176
References .................................................................................................................................................................. 178
Exercises ..................................................................................................................................................................... 178

lOMoARcPSD|1733731
Chapter 17: Make or buy ........................................................................................................................................... 181

17.1 Advantages and disadvantages of vertical integration.................................................................. 181
17.2 Vertical restraints ......................................................................................................................................... 183
17.3 Vertical restraints and competition policy ......................................................................................... 185
17.4 Case study: Vertical restraints in the European car market........................................................ 186
References .................................................................................................................................................................. 188
Exercises ..................................................................................................................................................................... 189
Chapter 18: Predation ................................................................................................................................................ 192
18.1 Examples of predation ................................................................................................................................ 192
18.2 Building overcapacity to deter entry .................................................................................................... 193
18.3 Predation and competition policy .......................................................................................................... 196
18.4 Case study: Predatory bundling in the market for web browsers ............................................ 197
References .................................................................................................................................................................. 199
Exercises ..................................................................................................................................................................... 199

lOMoARcPSD|1733731
Part I: Introduction
Welcome to this course on the economics of markets and organizations! The course combines
insights from two fields in Economics: Organizational Economics and Industrial Organization.
Organizational Economics focuses on the organizational design of firms, which includes a firms
motivation of its personnel, its hiring and firing policy, and make or buy decisions. Industrial
Organization considers the interaction of firms in markets, including their decisions with respect
to price, quantity, quality, product positioning, and advertising. The two fields are each others
complements in the sense that Organizational Economics examines transactions that occur
within firms (e.g., employees devoting time and effort in the firms production process, and
company divisions exchanging goods and services) while Industrial Organization studies firms
external transactions (e.g., transactions with suppliers, retailers, and final consumers).
In this part, I will introduce you to the course. In chapter 1, I will discuss several key concepts
and results, in particular those related to the neoclassical general equilibrium model. In chapter
2, I will give you a first impression of Organizational Economics. Chapter 3 includes an
introduction to Industrial Organization.

lOMoARcPSD|1733731
Chapter 1: Organizations and efficiency

In this course, we will study economic decisions in markets and organizations. A market is a
place (in the broadest sense of the word) where buyers and sellers can trade goods and services,
usually in exchange for money. Markets sometimes follow strict trading rules, in the case of
auctions and stock exchanges, for example, but they are usually organized less formally in that
the trading price and quantity are not decided on the basis of well-defined rules. Organizations
are entities in which people interact to reach economic goals. Organizations come in many
shapes and sizes: universities, student associations, soccer clubs, joint ventures, government
agencies, labor unions, households, and so on. In this course, we will mainly focus on for-profit
business companies.
In this chapter, I will introduce you to several concepts and key results that play a central role in
this course. I start in section 1.1 by defining the efficiency of markets and organizations. In
section 1.2, we study the neoclassical general equilibrium model. In this model, the fundamental
theorem of welfare economics holds true: An efficient outcome emerges in a competitive
equilibrium. I illustrate some of the lessons from the general equilibrium model in section 1.3 by
discussing perfect competition in isolated markets. I show that perfectly competitive markets
produce an efficient market outcome in equilibrium. In Section 1.4, we will see that inefficient
market outcomes may emerge when the assumptions underlying the general equilibrium model
are not satisfied. The resulting market failures seem highly relevant for markets and
organizations in practices and motivate our analysis in the remainder of these lecture notes. In
Section 1.5, I say a few words on both the structure of the lecture notes and the research
methods that we will apply. Section 1.6 contains a case study on Apple Inc. that highlights
several of the issues that we will study in this course.
1.1 Efficiency
In this course, we will judge the functioning of markets and organizations by their efficiency. In
markets, people trade particular goods and services to enhance their own well-being. The goal of
an organization is to satisfy the wants and the needs of its individual members. Students become
members of student associations to acquire valuable skills for their future career, to participate
in sports, or to meet other students. University professors may want to teach interesting courses,
engage in path-breaking research, and earn a good salary. Firms form a research joint venture to
share knowledge, to develop a new product, and to benefit from a joint patent.
Transactions of goods and services between individuals are the fundamental units of analysis in
this course. In student associations, students offer services to fellow students, such as organizing
debating tournaments, sports competitions, and dinner parties. University professors allocate
courses, research facilities, and divide management tasks among themselves. Participants in a
research joint venture share research responsibilities, management tasks, and the proceeds
related to newly developed products. In the economy as a whole, workers offer their time and
skills to firms, and firms distribute goods and services to consumers.
In this course, we will examine how successful markets and organizations are in establishing
efficient transactions. An allocation of goods and services is efficient if no reallocation of goods
and services exists that makes somebody better off without making someone else worse off. An
alternative way of evaluating markets and organizations is by measuring the total value they
create. In fact, the concepts of value and efficiency are closely linked as the value maximization
principle shows:

lOMoARcPSD|1733731
The value maximization principle: An allocation of goods and services is efficient (only)
if it maximizes the total value among the affected agents.
Let us consider an example from the car market to illustrate the value maximization principle.
Suppose Alfa Benz sells a particular car brand for a price of 20 (thousand euros). Four potential
buyers are interested in the product: Adle, Bono, Cher, and Dido. Adles value for the car
equals = 30, while Bono, Cher, and Didos are = 26, = 22, and = 18 respectively.
Figure 1.1 plots the data in an inverse demand function.
Figure 1.1: Inverse demand for Alfa Benz
How much value does this market generate? In economics, the usual term for value is welfare,
which is the sum of consumer surplus and producer surplus. Let us start with consumer
surplus, which is the net value gained by consumers. Adles surplus equals 10, i.e., the
difference between her value ( = 30) and the price ( = 20). Similarly, Bono and Chers
surpluses are = 6 and = 2 respectively. What about Dido? Her surplus equals zero,
because she will not buy a car as its price ( = 20) exceeds her maximum willingness-to-pay for
it ( = 18). Consumer surplus is the sum of the surpluses of the four potential buyers:
= ( ) + ( ) + ( ) = 10 + 6 + 2 = 18.
Figure 1.2 plots consumer surplus for this example. Note that consumer surplus is equal to the
surface enclosed by the price-axis, the inverse demand curve, and the line representing the
products price. Producer surplus generated in a market equals the sum of the profits of the
firms that are active in the market. In our example, Alfa Benz is the only firm in the market.
Producer surplus equals Alfa Benz profits, which is the difference between the firms revenue
and its costs of producing the three cars it sells. Therefore, if the costs of producing each car are
equal to = 16, producer surplus equals
= = ( ) = 3 (20 16) = 12.
Welfare is the sum of consumer surplus and producer surplus, i.e.,
= + = 18 + 12 = 30.

lOMoARcPSD|1733731
Figure 1.2: Welfare generated by Alfa Benz
Is the car market efficient, i.e., is welfare maximized? The answer is no. As you may recall from a
Fundamentals of Microeconomics course, welfare is maximized when price equals marginal
costs:
= .
The intuition is that a transaction between a seller and a buyer enhances welfare (only) if the
buyers additional utility from the transaction is higher than the production costs for the seller. A
buyer buys an additional unit (only) if the marginal utility from this unit exceeds the price while
the seller sells it (only) if the marginal costs of producing it are lower than the price. If the price
is above marginal costs, some value-enhancing transactions do not take place because the buyer
will not buy the units when his marginal utility lies below the price but above marginal costs.
Too few units will be traded. Similarly, a price below marginal costs will lead to excessive trade
because some units will be sold for which the buyers marginal utility is lower than the sellers
marginal costs. In other words, welfare is maximized at the marginal cost price.
In the car example, the price ( = 20) exceeds marginal costs ( = = 16). Note that the
allocation resulting from a price of 20 is not efficient because after selling cars to Adle, Bono,
and Cher at this price, Alfa Benz can make both itself and Dido better off by selling Dido a car at a
price of 17. Indeed, welfare is not maximized if the price equals 20. At a price equal to marginal
costs ( = 16), even Dido will buy a car. Note that producer surplus equals zero so that welfare
equals consumer surplus. It is readily verified that welfare is equal to
= = ( ) + ( ) + ( ) + ( ) = 14 + 10 + 6 + 2 = 32,
which is indeed higher than welfare at a price of 20.
1.2 General equilibrium

A serious challenge to reaching an efficient allocation of goods and services is that it depends on
information that may be scattered throughout the economy, including individual preferences,
technological possibilities, and resource availability. How could a market, an organization or,
10

lOMoARcPSD|1733731
more broadly, the economy as a whole channel this information to achieve an efficient allocation?
Two extreme possibilities are: (1) individuals communicate their information to a central
planner who makes all relevant decisions and (2) individuals make independent decisions on
the basis of prices of goods and services. In practice, most organizations and economies use a
mix of these two extremes.
The neoclassical general equilibrium model, developed by Nobel Prize laureates Kenneth Arrow
and Grard Debreu, formalizes the idea that a system of prices can achieve an efficient allocation.
The model analyzes an economy with many producers and consumers that may trade a great
number of goods and services between them. Arrow and Debreu assume that each producer
maximizes its own profits while each consumer maximizes his utility at the prevailing prices of
all goods and services in the economy.
The key result from the neoclassical general equilibrium model is the fundamental theorem of
welfare economics.
The fundamental theorem of welfare economics: An efficient allocation of goods

emerges at a competitive equilibrium.
This result is remarkable for two reasons. First, producers and consumers only need to know the
prices of goods and services to reach an efficient allocation: No central coordination of decision
is required. Prices play the role of Adam Smiths invisible hand by leading individuals to take
decisions necessary for a coordinated and efficient recourse allocation. Second, producers and
consumers behavior is in line with the interests of the entire economy despite all individuals
only pursuing their narrow self-interest. In the words of Adam Smith:
It is not from the benevolence of the butcher, the brewer, or the baker, that we can
expect our dinner, but from their regard to their own interest.
In a general equilibrium world, the role of a government is limited to the protection of property
rights: No central planner is needed to gather and disseminate information to coordinate
decisions on the economy and individual decision makers need not be forced to make decisions
that are not in their own self-interest.
1.3 Perfect competition

In this section, I discuss the model of perfect competition to illustrate how prices can lead
decision makers to efficient choices. The model focuses on a single market and relies on the
following assumptions:
1. There are many small buyers and sellers in the market: None of them can influence
the market price.
2. A homogeneous product is traded on the market: There is no product differentiation.
3. No entry barriers: Firms can freely enter and exit the market.
4. Perfect information: All buyers and sellers have perfect knowledge of the prices of all
sellers and every firm has access to the same production technology.
The market has a long-run equilibrium where the price equals both average costs and marginal
costs:
= = .
11

lOMoARcPSD|1733731
Figure 1.3 indicates how the market reaches the long-run equilibrium from an initial situation
where price exceeds average costs. The upper panel shows a potential market outcome in the
short run. On the left, we see the cost structure of a typical firm in the market (recall that all
firms have access to the same production technology so that each faces the same cost structure).
The profit-maximizing quantity ensures that a firms marginal revenue equals marginal costs.
Because a firm cannot influence the market price, its marginal revenue is equal to the market
clearing price, i.e., the price where supply equals demand. As a consequence, each firm produces
a quantity such that price equals marginal costs. Note that an individual firm makes a profit
because the market price exceeds the firms average costs. The profits will attract more firms
into the industry up to the point where none of them makes a profit. As the lower panel of figure
1.3 indicates, supply will increase until quantity reaches the point where the average costs and
marginal costs cross.
Figure 1.3: Towards an equilibrium in a perfectly competitive market
Observe that the equilibrium outcome in the perfect equilibrium model is efficient. As we
discussed before, at the marginal cost price, the market is allocatively efficient. Moreover, at the
quantity level where the marginal costs equal average costs, average costs are minimized so that
12

lOMoARcPSD|1733731
firms produce at the minimum costs per unit of production. As a consequence, the market is
productively efficient as well.
The model of perfect competition shows the invisible-hand role of prices in several ways. First of
all, the price informs consumers about the quantity they should purchase to maximize their
utilities. Second, the price informs the firms about the quantities each should produce to
maximize profits. Third, the price indicates whether firms should stay in or enter the market
(when the maximum profit at the current price is positive) or exit (when the maximum profit at
the current price is negative). Finally, and most importantly, the price guides consumers and
firms into an efficient market outcome.
1.4 Market failures

In the ideal world of the general equilibrium model, markets produce an efficient outcome. In
particular, markets are efficient in three important ways. First of all, the market outcome is
allocatively efficient: given the cost structure of producers, producers sell goods and services for
which the consumers value exceeds production costs. Second, goods and services are produced
at the lowest possible cost: The market is productively efficient. Third, markets are dynamically
efficient in that they establish an efficient balance between production and consumption over
time. In a dynamically efficient economy, firms engage in both process innovation (they develop
new production processes) and product innovation (they develop new products) up to the point
where the marginal social benefits of those innovations are equal to their marginal costs.
If markets work so well, two questions beg for an answer: Why do we see governments
intervene in markets? And why do so many transactions take place within firms (and not in
markets)? The answer to both questions is that markets may be plagued by market failures if the
assumptions underlying the general equilibrium model do not hold. In fact, most of the analysis
in the remainder of this course is motivated by market failures. We distinguish between four
potential sources of market failure: market power, information asymmetry, externalities, and
transaction costs.
In the general equilibrium model, it is assumed that producers and consumers are price takers.
However, if the number of producers (or consumers) in a particular market is low, they may be
able to influence the market price, i.e., they have market power. In chapter 3, we will see how a
monopoly firm charges prices above marginal costs resulting in an inefficient market outcome.
Similarly, in the case of oligopoly (chapter 6), product differentiation (chapter 9), and collusion
between firms (chapter 12), firms may be able to maintain prices above marginal costs. In
chapter 18, we discuss ways in which firms can establish market power by using aggressive
business strategies that induce the exit or deter the entry of rival firms. As we will discuss in
more detail in chapter 3, the government may intervene in markets to curb market power using
competition policy or economic regulation.
Information asymmetry is another source of market failure. Information asymmetry emerges

when one party engaged in a transaction has more or better information than another. A
decision maker may be tempted to act opportunistically if the other party cannot observe his
actions. In chapter 2, we will discuss the moral hazard problem that is imminent if a principal
cannot observe how much effort an agent expends. Potential solutions to the moral hazard
problem include incentives contracts (chapters 2, 5, and 8) and relational contracts (chapter 11).
We speak of adverse selection if one party to a transaction is better informed about the quality
13

lOMoARcPSD|1733731
or other characteristics of the traded product than another. In chapter 13, we will examine the
adverse selection problem that may emerge in such settings, i.e., only low-quality goods are
traded on the market. Two potential solutions to the adverse selection problem are screening
and signaling. In chapter 14, we will study screening and signaling devices employers and
employees may employ to mitigate adverse selection problems in the labor market. Chapter 15
contains a discussion of price discrimination as a screening device in markets.
Externalities may also cause transactions to result in inefficient outcomes. In the case of
negative externalities (such as air pollution) too much of the good is traded because the parties
involved in the transaction do not take into account the negative consequences of their
transaction to third parties. In chapter 9, we will observe that too many firms may enter a
market because they do not take into account that an incumbent firms profits decrease because
entrants steal business from the incumbents. Similarly, in the case of positive externalities, too
few transactions may take place because parties outside the trade benefit from it. A good
example of a setting where positive externalities emerge is team production. All team members
benefit from the effort provided by an individual member. As we will see in chapter 5, team
members may undersupply effort because they can free-ride on the efforts of others.
Finally, transactions may be plagued by transaction costs. There may be three sources of
transaction costs: coordination costs, information asymmetry, and imperfect commitment.
Coordination costs refer to costs parties incur to complete the transaction. These
include the costs trading partners incur to learn about each others existence, to
determine the price and the other terms of the transaction, and to come together to
complete the transaction. For example, on markets, sellers may have to advertise their
products to make potential buyers aware of their existence while buyers may face search
costs to find a satisfactory product. Within firms, central management incurs costs to
collect the relevant information from inside the organization to make strategic decisions
and to communicate the decisions to the relevant players in the organization.
As we saw earlier, transactions may be plagued by information asymmetry. In the case
of moral hazard, a party may have to invest in a way to monitor another party to prevent
him from acting opportunistically. Similarly, parties may have to implement costly
screening and signaling mechanisms to mitigate adverse selection problems.
Trading partners suffer from imperfect commitment if they cannot bind themselves to
fulfill promises they would like to make before the transaction takes place. In part VI, we
will examine several settings where the lack of commitment could be costly to trading
partners.
Nobel prize laureate Ronald Coase argues that transaction costs are the very reason why firms
exist. His point is that for particular transactions, the transaction costs are lower within firms
than in the market so that it makes perfect sense to organize them within firms. In chapter 17,
we will look in more detail at reasons why firms may prefer to make inputs themselves over
buying them in the market.
1.5 Set-up and research methods

The lecture notes consist of six parts. All parts are divided into three chapters. We will start each
part by devoting a chapter to theoretical techniques. In the two subsequent chapters, we will
examine applications to Organizational Economics and Industrial Organization respectively.
14

lOMoARcPSD|1733731
Each chapter contains a case study. Case studies play a double role. The first role is to show that
our analysis is not just of theoretical interest but that the issues discussed in the chapter are
relevant for decision making in practice. Secondly, in most case studies I take a further step by
presenting the results of empirical tests of the theoretical results established in the chapter. As
you will notice, I have attempted to provide evidence for and against the theory, to indicate both
its strengths and its limits. The current chapters case study is mainly illustrative: It involves the
success story of Apple Inc. The case study highlights several of the issues in Organizational
Economics and Industrial Organization that we will study in this course.
1.6 Case study: Apple Inc.

At the time of writing, Apple Inc. is the second-largest publicly traded company in the world, by
market capitalization. Its journey in reaching this point is one of the great business success
stories and it illustrates many key concepts from this course. It gives examples of market entry
by new firms, but also shows how successful innovation can enable a firm to create a new
market. It shows how, with heterogeneous goods, competition and strategic behavior can take
on many different forms. Patent war chests can create barriers to entry. Finally, Apple is also a
clear example of how management styles and employee relations can have far-reaching effects
on a companys identity and on its fortunes.
As we know it today, Apple is a well-diversified company, but its history begins in the computer
market with a single product: the Apple I. Until the 1970s, the idea of a small computer for
personal use seemed unfeasible. Computers were large, costly, and complicated systems mostly
used by companies, universities, and government agencies. The market was dominated by IBM
(International Business Machines), with a market share of 81.2% during the 1960s. Several
prototypes of small computers were launched during this period, but they were still too
expensive to appeal to the consumer masses.
Apple Computer Company was the first company to foresee the potential of a market for
personal computers, and the opportunity that it presented. Officially founded on April 1, 1976,
by Steve Jobs, Steve Wozniak and Ronald Wayne, Apple Computer had a clear mission that
reflected the founders belief one person one computer to change the world by bringing
computers to everyone. Indeed, a rather ambitious object for three young men (Jobs had not
even graduated) who headquartered their company in their parents garage.
However, the Apple I did not attain the desired results. It was only with the production of their
second PC, the Apple II in 1977 that the company really took off. In six years following its
foundation, Apples earnings rose exponentially: from $793,000 to $76,714,000. This fast growth
prompted the company to offer its shares on the stock exchange. It was a great success, and the
IPO (initial public offering, i.e., stock market launch) oversubscribed. At the beginning of the
1980s, Apple counted thousands of employees and established itself as one of the major players
in the early stages of the personal computer market.
However, its competitors did not sit idly by. In 1981, IBM broke into the personal computer
market with its first PC using Microsofts software (called IBM PC). In three years, IBM was able
to conquer 50% of the market share. The challenge between Apple and IBM involved their
business models and their different technological standards (their operating systems, for
example). On the one hand, Jobs company focused on innovation by following a strict licensing
and patent-regulated policy. Apple Computer was reluctant to give away any information
15

lOMoARcPSD|1733731
regarding its hardware and software, not allowing third-party program writers to develop
different features for Apple technology. On the other hand, IBM followed an open-architecture
model, giving the opportunity to other manufacturers to produce and sell peripheral
components and compatible software without purchasing a license. IBM even diffused its PC
circuit schematics and other engineering and programing information.
As a consequence, Apple products were qualitatively better than their competitors, but their
price was widely above the industry average. The IBM open-architecture model created a sub-
market devoted to the production of IBM PCs components. At the same time, though, it enabled
other companies (such as Dell, Compaq and HP) to enter the market by cloning the IBM PC. The
increase in competition implied a decrease in IBMs returns, but it established IBM PC
technology as the industry standard.
In 1985 Steve Jobs left the company, and the years that followed were probably the darkest in
Apples history. The market for personal computers was saturated by IBM PC clones, and the
software market was mainly dominated by Microsofts standards. During this period, Apple
faced a 70% decline of its market share, and the companys structure was reorganized. In the
early 1990s Apple gave up its restrictive patents policy, and started to license both hardware
and software to third-party companies, somehow imitating IBMs business model. These
companies, though, were diffident to invest time and money for a technology that had only a
small portion of the market.
It was only with the return of Steve Jobs, in 1997, as iCEO (interim CEO) that the company rose
again. Jobs drastically changed the structure of Apple by announcing an alliance with Microsoft
Apples main competitor in the software market and by acquiring several companies to create
a portfolio of professional and consumer-oriented software products. The hardware side of the
company evolved following the strategy that was to determine Apples future success: design
and innovation. The iMac and iBook were the firstborn of the new Apple era, and within a year
the company returned to profitability.
The arrival of the 21st Century coincided with a new phase in Apples history. Apple effectively
ceded the business segment to Microsoft and focused on making desirable consumer products.
Following a product differentiation strategy, Jobs company aimed at new markets, where the
company achieved astonishing results. As consumer technology came to play an ever-growing
part in our everyday lives, Apple reacted dynamically to capitalize on emerging trends,
particularly with regard to growing demand for highly portable devices and the potential for
integration with online platforms, such as iTunes.
The first Apple retail store was opened in 2001; the stores would become a key channel for
Apple to maintain its iconic status as a brand. In the same year, Apple entered the nascent mp3
player market with the iPod, selling more than 100 million units in six years. Three years later,
Jobs launched the iTunes Store, offering online music downloads. The company then entered the
mobile phone market in 2007, with its groundbreaking iPhone. Eventually, Apple entered, or
rather, created the market for tablets in 2010, introducing the iPad.
As technology markets evolved, so did management styles within the worlds leading
corporations. In recent years, new standards of management have been established by highly
innovative and forefront companies such as Google, in which the employees gratification and
creativity play a central role. Under Steve Jobs management, Apple continued to represent an
16

lOMoARcPSD|1733731
exception to this stream. Nothing came out of the company if it did not meet Jobs strict
requirements. While other high-tech companies motivated their employees with carrots, Jobs
used the stick. A former Apple engineer reported: More than anywhere else Ive worked before
or since, theres a lot of concern about being fired. Nevertheless, Jobs autocracy was always
balanced by his famous charisma. He was able to create a strong feeling of membership, and the
employees remained devoted to the company.
The evolution of consumer technology and Apples entry into new markets also changed the
competitive landscape. Whereas Apple Computer Company used to do battle with IBM and
Microsoft, Apples key 21st Century rivals are Google and Samsung. On the surface there are
similarities with the old rivals: Samsung manufactures popular alternatives to the iPhone and
iPad, while Google offers a flexible, open-source operating system Android. However, the
nature of competition has perhaps become more complex. While core microeconomic theory
tends to focus on competition though price, quality, or product differentiation, it is clear that
Apple and Samsung (among others) are also increasingly engaged in fierce competition in
another dimension: that of intellectual property.
Economic theory tells us that patents play an important but delicate role, especially in markets
where innovation is crucial. On the one hand, patents give a crucial incentive for firms to invest
in R&D, by allowing them to appropriate ideas and reap the rewards of any substantial
innovations as opposed to having ideas immediately copied by rivals. On the other hand, if
patents are excessively broad in the way they are defined, or if they last for too long, they can
give the innovating firm significant market power and thereby disadvantage consumers,
unreasonably preventing competition from other firms. Thus, a well-functioning patent system
is needed to strike the right balance and promote efficient market outcomes.
At present, patent offices are overrun with applications. Americas Patent and Trademark Office
has a backlog of over half a million patents to review, while the Chief Economist at the European
Patent Office has admitted that the increase in the number and complexity of applications
being examined cant be neutral in terms of quality. In many markets, widespread uncertainty
now prevails over what can be patented and how far the breadth of patents extends, particularly
in the US, where in recent decades it has become possible to patent software and business
methods (Amazons one-click payment method is a recent and controversial example).
Apple has been involved in numerous legal battles with Samsung (as well as other companies).
In August 2012s landmark ruling, a Californian court deemed Samsung to have copied design
features and functions of the iPhone (such as sweep-to-zoom) and ordered the company to pay
Apple damages of $1 billion. The ruling remains the subject of a lengthy appeals process at the
time of writing, but according to many commentators it gives encouragement for firms to build
up stocks of patents. Indeed, there is evidence of Apple doing just that: as of June 2012, Apple
had already filed nearly as many patent suits as it did in 2010 and 2011 combined. A patent war
chest could be used strategically against competitors, even as a form of predatory behavior.
However, other analysts predict that legal cases between firms such as Apple and Samsung will
tend to be ultimately resolved by confidential out-of-court settlements, with little money
actually trading hands, and it will continue to be the court of capitalism that determines the
fortunes of each company. After all, Samsung produces parts that account for one quarter of an
iPhones component cost, so Apple may have little incentive to eliminate its rival!
17

lOMoARcPSD|1733731
Although Apple continued to see success after Steve Jobs death on October 5 2011, by 2013
many investors had begun to feel that the company had peaked. Apples share price fell by
around 40% between September 2012 and April 2013, amid lower-than-expected profit figures,
concerns about increasing competition and doubts about Tim Cooks ability to fill the gap left by
Steve Jobs. However, the companys remarkable history of ups and downs should serve as a
reminder to not write them off too quickly. Whether or not Apple has peaked could depend on
the success of further product diversification, as it is widely expected to go after the smart TV
market, while it may have to loosen its principles and launch lower-quality, lower-price
products in emerging markets in order to tap in to new channels of growth. Will Apple continue
to thrive in the face of increasing competition, without compromising the philosophy that gives
the brand its essence?
Bibliographical Notes
[Every individual] neither intends to promote the public interest, nor knows how much he is
promoting it () by directing [his] industry in such a manner as its produce may be of the greatest
value, he intends only his own gain, and he is in this, as in many other cases, led by an invisible hand
to promote an end which was not part of his intention (). By pursuing his own interest he
frequently promotes that of the society more effectually than when he really intends to promote it.
The passage comes from Adam Smiths famous Wealth of Nations, published in 1776, and
illustrates the importance of free competitive markets and organizations in the development of
modern society and the achievement of economic efficiency. Smiths 1776 work is where many
central issues in economics, such as economic organizations and efficiency, find their first
significant treatment. These ideas remained topical in the writings after Smith, with important
contributions to organizational efficiency coming from the work of Karl Marx in the 19th century
(Marx, 1867/1976). Efficiency is a topic of predominant focus also in Frank H. Knights Risk,
Uncertainty and Profit (1921), where economic activity in organizations is addressed and the
first departures from the competitive equilibrium are established.
At this point it is important to mention the rise of the so-called marginal revolution within
economics, the idea that economic decisions are made (or should be made) on the margin and
not by looking at the big picture. That idea, which underlies our development of the competitive
equilibrium, can be attributed to the work of Carl Menger (1871), William Stanley Jevons (1866)
and Marie-Esprit-Lon Walras (his original work was published in 1874 in French; the
translation available in the references dates back to 1954) in the late 19th century and is
popularized in Alfred Marshals Principles of Economics (1920), where actual supply and
demand curves and tables are first illustrated. These contributions, together with the work of
Ronald Coase (1937 and 1960) on the organization of economic activity between agents and
firms and the role of transaction cost economics, have largely denoted the importance of value
maximization and efficiency and have made economists aware of their power as helpful
explanatory principles.
This leads us on to the formal development of the neoclassical general equilibrium model, which
formalizes the idea that a system of prices can achieve an efficient allocation. In his influential
monograph The Theory of Value, published in 1959, Grard Debreu gives an initial
mathematical explanation of the prices of commodities resulting from the interaction of the
agents of a private ownership economy through markets and the role of prices in an optimal state
of an economy. Along with Kenneth Arrow they developed the neoclassical general equilibrium
18

lOMoARcPSD|1733731
model (Arrow and Debreu, 1954) for which they both won a Nobel Prize in economics. More
recent contributions to the competitive equilibrium stem from the work of Vernon Smith (1987)
who embarked on some of the first experimental explorations of market incentives that came to
support the theoretical predictions of efficiency in competitive outcomes.
The fundamental theorem of welfare economics was mathematically demonstrated principally

by Kenneth Arrow and Grard Debreu (1954). For a more accessible treatment of the theorems
and the key ideas underlying the competitive equilibrium you can refer to the work of Werner
Hildenbrand and Alan Kirman (1975).
The Apple case study is based on several articles from the popular press including Macworld
(2006), Business Insider (2012), The Economist (2012, 2013) and The Guardian (2012).
References
Arrow, K.J. and G. Debreu (1954). Existence of an equilibrium for a competitive economy,
Econometrica, 22, 265-290.
Business Insider (2012). How Apple really lost its lead in the '80s, December 9.
Coase, R. (1937). The nature of the firm, Economica, 4, 386-405.
Coase, R. (1960). The problem of social cost, Journal of Law and Economics, 3, 1-44.
The Economist (2012). Apple vs. Samsung: iPhone, uCopy, iSue, September 1.
The Economist (2013). Has Apple peaked?, January 24.
The Guardian (2012). The secret of Apple's success: simplicity, June 15.
Hildenbrand, W. and A.P. Kirman (1975). Introduction to Equilibrium Analysis, Amsterdam:
North-Holland.
Jevons, W.S. (1866). Brief account of a general mathematical theory of political economy, Journal
of the Royal Statistical Society, Section F, 282-287.
Knight, F.H. (1921). Risk, Uncertainty and Profit, London: London School of Economics.
Macworld (2006). Thirty pivotal moments in Apple's history, March 30.
Marshall, A. (1920). Principles of Economics, Revised edition (eds.), London: Macmillan.
Marx, K. (1867/1976). Capital, Harmondworth, UK: Penguin Books.
Menger, C. (1871). Principles of Economics, New York: New York University Press.
Smith, A. (1776/1976). An Inquiry into the Nature and Causes of the Wealth of Nations, Oxford:
The Clarendon Press.
Smith, V.L. (1987). Experimental methods in economics, in J. Eatwell, M. Milgate and P. Newman
(eds.), The New Palgrave: A Dictionary of Economics, volume II, London: Macmillan, 241-249.
Walras, L. (1874/1954). Elements of Pure Economics, London: Allen and Unwin.
Exercises
Exercise 1.1 Welfare

The city of Cabernet is very famous for its production of wine. The inhabitants of the city have an
aggregate demand for wine that can be described as follows:
() = = 150 2,
where is the demanded quantity in bottles of wine and is the price in euros per bottle. The
producers aggregate supply is:
19

lOMoARcPSD|1733731
() = = 13 15,
where is the supplied quantity in bottles of wine. The city produces only one quality of wine.
a) Find the equilibrium price and quantity.
b) Compute the consumer and producer surplus, as well as the total welfare in the city of
Cabernet.
The municipality of Cabernet decides to put a tax of 7.50 on every bottle of wine sold.
c) Compute the new equilibrium price and quantity and calculate the allocative inefficiency.
d) Assume that the municipality equally redistributes tax revenues to the inhabitants. Are they
better off compared to the situation before the tax was introduced?
Exercise 1.2 Utility and consumer surplus

Jeremy is an economics student who loves hamburgers. He could eat any number of them for
dinner, but he gets a really bad stomachache after eating a certain amount. In fact, his utility
function for hamburgers is given by:
2
() = 10
2
where is the number of hamburgers he eats for dinner (with 0).
a) How many hamburgers can Jeremy eat before he gets a stomachache (that is, before his utility
becomes negative)?
b) Calculate the optimal number of hamburgers that Jeremy can eat as a function of , the price
per hamburger. (Hint: You have to maximize Jeremys net utility. That is, his utility minus the
amount he spends on hamburgers.)
c) Derive Jeremys inverse demand for hamburgers.
d) Compute Jeremys consumer surplus as a function of .
e) Show that Jeremys net utility as a function of coincides with his consumer surplus.
Exercise 1.3 Perfect competition

In the city of Gelato the market for ice cream is perfectly competitive. Aggregate demand for ice
cream is:
() = 1200 25,
where is the price for one cone of ice cream. All ice cream producers in the city have the same
total cost function:
( ) = 10 ,
where represents the number of ice cream cones firm produces. Assume that the market is
in equilibrium.
a) Derive the firms marginal and average cost.
b) Compute price and quantity in equilibrium.
Assume that there are 50 firms present in the market.
20

lOMoARcPSD|1733731
c) Calculate the number of ice cream cones that each firm produces. Will they produce the same
quantity? Why (not)?
d) In general, is it possible that a firm in a perfectly competitive market produces at a price
greater than its average cost?
Exercise 1.4 Perfect competition and entry decisions

ChoppinAxe is a small Swedish firm that produces wood planks and operates in a perfectly
competitive market. Every firm in the market has the following total cost function:
( ) = 2304 20 + 42 ,
where represents the tons of wood planks each firm produces. ChoppinAxe and all other
firms sell their products to only one buyer, IPEA, an international furniture company. IPEAs
total demand for wood planks can be described as follows:
() = 3448 4,
where is the price for a ton of wood planks. You may assume that the market is in equilibrium.
a) Find ChoppinAxes supply curve.
b) How many tons of wood planks will ChoppinAxe sell to IPEA? At which price?
c) Repeat sub questions (a) and (b) for KindCutters, another wood planks producer present in
the market.
FallingTrees is a start-up firm that is considering entering the wood planks market. Along with
ChoppinAxe and KindCutters there are another 113 wood plank producers in the market.
d) Should FallingTrees enter the market? Why (not)?
21

lOMoARcPSD|1733731
Chapter 2: Organizational Economics

In this chapter, I introduce you to Organizational Economics. In section 2.1, we discuss hiring
decisions: Which workers will a firm hire? In section 2.2, we study how managers can motivate
workers in a baseline Principal-Agent model. In section 2.3, we extend our analysis of this
Principal-Agent model by discussing the consequences of relaxing some of its crucial
assumptions. Section 2.4 contains a case study on the financial crisis of 2008.
2.1 Hiring decisions

Most firms are continuously looking for workers to fill their vacancies. When several people
apply, the key question is: whom to hire? The obvious answer to this question is: the best
candidate. However, it might not be at all obvious which of the candidates is the best. Consider
the example included in table 2.1. Web design company Macrosoft is growing rapidly and
desperately needs new employees. Two people apply: Bert and Ernie. Bert has a Masters in
Computer Science. Ernie has not enjoyed any formal education related to web design, but he has
some experience in building websites. Bert is the more productive of the two: He is able to sell
100,000 worth of web designs annually, while Ernies yearly production potential equals
80,000. Ernie is cheaper for the firm, though. As he has no degree, his outside options are less
attractive than Berts. Ernie is willing to take the job for an annual salary of 40,000, while Bert
is only satisfied with 55,000 a year. Whom would you hire?
Table 2.1: Berts and Ernies yearly production and salary
Education Yearly production Yearly salary

Bert MSc Computer Science 100,000 55,000
Ernie No 80,000 40,000
At first sight, Bert seems to be the more attractive than Ernie for Macrosoft. As table 2.2
indicates, Macrosofts profit equals 45,000 for Bert and 40,000 for Ernie. So, if Macrosoft has
only one workplace available, it should hire Bert. However, Ernie may be more attractive if
Macrosoft has many vacant workplaces. The reason is that Ernie generates a higher profit per
unit of output than Bert: If Macrosoft hires Ernie, its profit per unit of output equals 0.50 which
is greater than 0.45, its profit per unit of output if it hires Bert. To illustrate this point further,
suppose Macrosoft has been awarded a contract to design 1,000,000 worth of websites. To
fulfill this task, it can hire 10 educated workers like Bert or 12.5 uneducated workers like Ernie.
Uneducated workers are clearly more attractive for Macrosoft than educated workers.
Table 2.2: Macrosofts hiring decision when it needs to produce 1,000,000 worth of output
Profit per Profit per unit Number of Employee

employee of output employees needed costs
Bert 45,000 0.45 10 550,000
Ernie 40,000 0.5 12.5 500,000
In our analysis, we made several simplifying assumptions. First of all, we assumed that
Macrosoft is able to motivate the workers one way or the other to produce the desired output. In
section 2.2, we will see how firms can do so. Second, we assumed that Macrosoft can observe
applicants productivity. In practice, this assumption is often not fulfilled. In chapter 14, we will
22

lOMoARcPSD|1733731
have a deeper look into firms hiring and firing decisions when it is incompletely informed about
relevant characteristics of potential employees.
2.2 The Principal-Agent problem

In this section, I will introduce you to the Principal-Agent problem. This problem refers to
motivating a person or organization (the agent) to act in the interest of another (the principal).
Principal-Agent relationships are often plagued by moral hazard problems, i.e., behavior by the
agent that is inefficient arising from conflicts of interest between the parties and the principal
not being able to observe the agents actions.
To illustrate the idea, consider agent Antonio who directs a movie on behalf of principal
Penlope. Antonio has to decide how much effort to put into directing the movie. We assume
that each unit of Antonios effort yields one additional movie theater ticket sold, i.e., Antonios
output equals
= .
Suppose that Penlope obtains a price per unit of output. Antonios cost () of exerting effort
equals
() = 2
where > 0. There are no costs other than the costs of effort.
What are the properties of an efficient contract between Penlope and Antonio? Penlopes
profits are given by
= ,
where denotes the wage Penlope pays Antonio. Antonios utility is equal to
= 2.
So, the value () generated within Penlopes and Antonios Principal-Agent relationship as a
function of Antonios effort equals
() = + = 2 = 2 .
Observe that Antonios wage is irrelevant for the total value generated: It is a welfare-neutral
transfer from Penlope to Antonio. As we will see in chapter 8, Antonio may have attractive
outside options available so that his wage must be sufficiently high for him to do business with
Penlope in the first place. We ignore this for the moment.
Value is maximized if
() = = 0,
or, equivalently, if

= .

23

lOMoARcPSD|1733731
Suppose that Penlope cannot enforce a contract with Antonio that specifies that he must exert
the above level of effort. Such a contract may not be enforceable because Penlope cannot
observe Antonios effort. Even if Penlope can observe Antonios effort, the contract may be
meaningless because his effort cannot be verified by a court.
Penlope can still encourage Antonio to exert effort indirectly by offering him an incentive
contract that specifies that Antonios wage is increasing in his output, i.e., the number of movie
tickets sold. In practice, incentives contracts come in many shapes and sizes. We will mainly
focus on linear contracts, i.e., contracts that specify a base wage plus a bonus for each unit of
output. An example of a non-linear contract is one that pays the agent a bonus if his output hits
some pre-set target. Another example is managers obtaining a set of stock options in their
company. Both contracts may give the agent perverse incentives. Under the first contract, the
agent will stop working as soon as he has reached the target that guarantees him the bonus. In
the case of stock options, managers have a strong incentive to undertake overly risky projects if
the underlying stock price is below the strike price to maximize the probability that the stock
price does exceed the strike price. Linear contracts are immune to such perverse incentives. In
fact, as we will see in a moment, they can induce the agent to exert an efficient level of effort.
Antonios utility () from the above linear contract as a function of his effort equals
() = + 2 = + 2 .
Antonio chooses the effort level that maximizes his utility. The first-order condition of the
corresponding maximization problem yields
() = = 0
or, equivalently,

= .

When we compare this effort level with the value-maximizing effort, we find that the optimal
bonus is given by
= .
This result is striking for at least three reasons. First, the principal can use a simple, linear
contract to implement an efficient outcome. Second, the optimal bonus does not depend on
Antonios effort costs. This makes the efficient contract robust to particularities of the
environment. Third, the bonus in the efficient contract is equal to the price per unit of output. In
other words, Penlope should allow Antonio to keep all the fruits of his efforts. Penlope can still
obtain some of the gains from trade by making sure that Antonios fixed wage is negative, i.e.,
Antonio pays her for the opportunity to direct her movie. We conclude that it is optimal that
Penlope makes Antonio the residual claimant of her movie project. More generally, it can be
shown that it is optimal for the principal to sell the firm to the agent.
2.3 Critical assumptions in the Principal-Agent model

We made several simplifying assumptions in the above Principal-Agent model. Relaxing those
assumptions may change the conclusion that selling the firm to the agent is efficient. First of all,
24

lOMoARcPSD|1733731
the agent may be budget restrained so that he is unable to pay the principal an upfront payment.
Second, accepting the contract may be unattractive for a risk averse agent when output
depends not only on his effort but also on events outside his control. For example, the revenue
from Penlopes and Antonios movie may be affected by press coverage, the weather, competing
movies released in the same period, and so on, all of which are outside Antonios control. In
chapter 8, we will discuss the consequences of the agents risk aversion in such environment.
A third critical assumption is that output is perfectly measurable. In practice, principals often
have to rely on imperfect performance measures. One risk for the principal if she gives the
agent an incentive contract that is based on an imperfect performance measure is that she gets
what she pays for. The agent will focus his attention on effort that improves his score on the
imperfect performance measure, which is not necessarily in the principals interest. This parts
case study includes the example of incentive contracts in the US mortgage market. Banks paid
mortgage brokers higher fees the greater the mortgage volume the broker originated. The
consequence was that the mortgage brokers had little incentive to verify whether the borrower
could repay the loan. According to most analysts, this was one of the major causes of the 2008
financial crisis.
A fourth crucial assumption is that the agent performs only one task. In the case of multiple
tasks, the principal should incentivize the agent on all tasks as otherwise there are some he will
neglect. The equal compensation principle makes this statement more precise:
The equal compensation principle: Suppose that the agent can expend effort on two
tasks and that the principal cannot monitor how he divides his efforts over the two tasks.
The agent will only expend effort on both tasks if their marginal rate of return is the
same. Otherwise, he will exert zero effort on the task with the lower marginal rate of
return.
The equal compensation principle has sharp consequences for the design of incentive contracts
in practice. The principal should be careful when giving incentives on one task if it is difficult for
her to obtain an objective performance measure on another. For example, it may not be a good
idea to give teachers a bonus depending on the performance of their pupils on a test. The reason
is that teachers may be tempted to teach to the test ignoring pupil skills that are not part of the
test.
A final assumption is that the agents output is objectively measurable, i.e., it can be verified by a
court. If the agents output is not objectively measurable, but the principal can assess it
subjectively, she may rely on subjective performance evaluation. While the principal cannot
write a contract with the agent that is enforceable in court, she can rely on a relational contract
with the agent: The principal promises the agent to pay a bonus if he performs well on the
subjective performance measure. In chapter 11, we will see that such a relational contract can
work if the principal and the agent have a long-run relationship in which they interact
repeatedly.
2.4 Case study: Moral hazard in the mortgage market and the financial crisis
of 2008
According to most analysts, the global financial crisis in 2008 was rooted in ill-designed
incentive contracts in the US mortgage market. Initially, only the most credit-worthy borrowers
25

lOMoARcPSD|1733731
were able to obtain residential mortgages. However, the emergence of financial innovations in
the mortgage market around the turn of the century led to an enhanced quantitative evaluation
of the credit risk of borrowers. Households with poor credit history and a high likelihood of
default on their mortgage repayment, namely, the subprime borrowers could now borrow at
attractive rates. As a result, bank lending increased sharply leading to a flourishing mortgage
market and a dramatic escalation in housing prices. Between 1997 and 2006, the price of the
typical American house rose by 124 per cent. The growth of the subprime mortgage market
inevitably resulted in the formation of a housing price bubble, which is considered to be one of
the triggering factors of the recent financial crisis.
Figure 2.1: Development of sales price of new homes in the US over time
The underlying problem in the development of subprime lending was the incentive
incompatibility of the mortgage contracts, which resulted in excess and disregarded risk-taking
by banks and other financial institutions. Traditionally, banks originate commercial loans and
hold them until their maturity. When the borrower successfully pays off the loan, the bank
receives the principal payment back and derives its profit from the loans interest payments. In
the case that the borrower defaults on their loan payment, the bank either repossesses and sells
the loans collateral (in the case of a mortgage, the house) or suffers a loss depending on whether
the loan is collaterized or not. Therefore, according to the originate-to-hold business model
banks have high incentives to avoid information asymmetry problems, select carefully their
customers and monitor their behavior.
Before the emergence of the crisis, however, there was a switch from an originate-to-hold to an
originate-to-distribute business model in the writing of residential mortgages. According to the
latter, the mortgage was originated by a separate party, typically a mortgage broker, and then
distributed by the bank to an investor as an underlying asset in a security. Unfortunately, the
originate-to-distribute model is subject to a moral hazard problem: the mortgage originator has
little incentive to make sure that the borrower is a good credit risk for the bank. Once the
mortgage is written and passed on to the financial investor, the mortgage broker has no
incentive to ensure that the borrower will repay the loan. This moral hazard problem was
intensified by the incentive contract between the bank and the mortgage broker: The more
volume the broker originates, the higher the fee she earns.
26

lOMoARcPSD|1733731
Not surprisingly, given these incentives, the mortgage brokers did not make a strong effort to
evaluate whether the borrower could pay off the loan. Instead, they were incentivized to
encourage households to take on mortgages they could not afford, or even to commit fraud by
falsifying information on borrowers mortgage applications in order to qualify them for their
mortgages. Lax regulations, which did not require the disclosure of information that could help
to assess the borrowers risk, ensured that borrowers who were most likely to produce an
undesirable outcome could still be selected for mortgages. In addition, as a result of the
originate-to-distribute model, brokers had no incentive to restrain borrowers from engaging in
undesirable activities after undertaking their mortgage, and therefore, moral hazard problems
soared.
Figure 2.2: Mortgage brokers encouraged households to take on mortgages they could not afford
Banks, in turn, had also few incentives to monitor the risk associated with the brokers incentive
contracts, since that was passed on to other financial investors. The mortgages were bundled
together with other securities and then sold off as underlying assets in a debt security through a
process know as securitization. Historically low interest rates and the failure of the regulatory
framework to keep pace with financial innovation encouraged borrowing, which led to excessive
risk taking by banks as more and more subprime borrowers were served. In addition, the
function of the FED and most central banks around the world as the lender of last resort created
another moral hazard problem for banks; knowing that they would be bailed out in the case
losses from their mortgages occurred, further contributed to an increasing debt burden, or over-
leveraging.
The burst of the housing bubble in 2007, forcing banks to write down several hundred billion
dollars in bad loans caused by mortgage failures, led to the most severe financial crisis since the
Great Depression. The stock market capitalization of the major banks declined by more than
double. While the overall mortgage losses were large on an absolute scale, they were still less
than 10 per cent of the $8 trillion of US stock market wealth lost between October 2007, when
the stock market reached an all-time high, and October 2008. The recent financial crisis led to a
slowing of worldwide economic growth and massive government bailouts of financial
institutions and threatens to have large repercussions on the real economy until today.
The field of organizational economics is a recent one and developed with the application of
standard microeconomic theory to organizations and the labor market.
27

lOMoARcPSD|1733731
Organizational economics is closely linked to the analysis of human capital, which can be
attributed to University of Chicago professor Gary Becker. See, for instance, his 1988 work
Human Capital: A Theoretical and Empirical Analysis. Although departing from the focus of this
chapter, Beckers research has been recognized by many authors as a cornerstone in the fields of
labor economics and employment relations, which will be discussed in later chapters. For a
formal theory of employment relationships you can refer to Herbert Simons influential paper
published in 1951 in Econometrica.
The main intuition behind hiring decisions discussed in this chapter stems from the work of
Lazear (1995). See also Lazear and Gibbs (1995) textbook. Considered by many as the founder
of personnel economics, Edward Lazears publications have contributed a lot to matters such as
the recruitment of risky workers, internal labor market incentives, and performance evaluation.
George Baker, Michael Gibbs and Bengt Holmstrm (1999) have developed these ideas further,
which will appear as recurring themes in later chapters. For a general overview of
organizational economics you can also refer to the textbook of Besanko et al. (2000) or the
Handbook of Organizational Economics by Gibbons and Roberts (2013).
A classic textbook is Milgrom and Roberts (1992). Among other things, Milgrom and Roberts
discuss several principles of contract design including the equal compensation principle, the
informativeness principle, and the incentive-intensity principle. We will discuss the latter two in
chapter 8.
References
Baker, G., M. Gibbs and B. Holmstrm (1994a). The internal economics of the firm: evidence from
personnel data, Quarterly Journal of Economics, 109, 881-919.
Baker, G., M. Gibbs and B. Holmstrm (1994b). The wage policy of a firm, Quarterly Journal of
Economics, 109, 921-955.
Becker, G.S. (1988). Human Capital: A Theoretical and Empirical Analysis with Special Reference
to Education, New York: Columbia University Press.
Bentham, J. (1824/1987). An introduction to the principles of morals and legislation, in J.S. Mill
and J. Bentham, Utilitarianism and Other Essays, Harmandsworth: Penguin.
Besanko, D., D. Dranove, M. Shanley, and S. Schaefer (2000). Economics of Strategy, New York:
Wiley.
Gibbons, R. and J. Roberts (2013). The Handbook of Organizational Economics, Princeton:
Princeton University Press.
Lazear, E. (1998). Hiring risky workers, in I. Ohashi and T. Tachibanaki (eds.), Internal Labour
Markets, Incentives and Employment, New York: St. Martins Press.
Lazear, E. and M. Gibbs (2009). Personnel Economics in Practice (Second Edition), Hoboken, NJ:
John Wiley and Sons.
Milgrom, P. and J. Roberts (1992). Economics, Organization and Management, New York: Prentice
Hall.
Mill, J.S. (1887/1987). Utilitarianism, in J.S. Mill and J. Bentham, Utilitarianism and Other Essays,
Harmandsworth: Penguin.
Simon, H.A. (1951). A formal theory of the employment relationship, Econometrica, 19, 293-305.
28

lOMoARcPSD|1733731
Exercises
Exercise 2.1 Shirking workers and the efficiency wage

Jamie owns a posh restaurant in the center of Amsterdam that has gained reputation for having
a friendly, personal, and efficient service. The key to its success is that Jamie offers his personnel
a higher wage (efficiency wage) than the market average as an incentive to avoid shirking.
Gordon, following Jamies success, has decided to introduce a similar incentive scheme at his
restaurant. He noticed that his employees have recently started to shirk and as a result the
restaurants profit has decreased. All employees are assumed to be risk-neutral.
a) Show that a risk-neutral worker will not shirk if and only if ( ) , where
> 0: The probability that a shirking worker is caught and fired,
> 0: The efficiency wage,
> 0: The average market wage, and
U > 0: The utility the worker obtains from shirking
b) Suppose that cooks get utility = 1,000 from shirking, the average market wage is =
20,000 a year and the probability that a cook is caught shirking is 20%. How much should
Gordon pay if he wants to hire cooks that will not shirk?
c) Assume Gordons utility in terms of the combination of wages and time spent monitoring
are given by the function:
(, ) = 4/5 1/5
His labor costs can be described by:
(, ) = + 100,000
Gordon has decided to spend no more than 35,000 a year in labor costs. Find the profit
maximizing pair of monitoring level and efficiency wage.
Exercise 2.2 Linear and non-linear incentive contracts

A firms short-run revenue is given by = 10 2, where is the level of effort of a typical
worker (all workers are assumed to be identical). A worker chooses his level of effort to
maximize his wage net of effort (the per-unit cost of effort is assumed to be 1).
a) Determine the level of effort and the level of profit for each of the following wage
arrangements:
i. = 2 for 1; otherwise = 0

ii. = 2
iii. = 12.5
b) Explain why these incentive contracts generate different outcomes. Which contract provides
the greatest incentives?
Exercise 2.3 The optimal bonus

A medical insurance company offers its salespeople the following compensation scheme: every
worker receives a fixed salary and, in addition to that, a commission depending on the volume
29

lOMoARcPSD|1733731
of sales they generate. At the time, the salespeople are paid 10% of every sale they make that is
above 3,500 per day. In order to motivate its salespeople to work harder, the company has
decided to change the commission structure and offer 25% for sales above 5,000. In all cases,
the workers have to generate a minimum of 2,500 or else they are fired.
a) Illustrate both compensation schemes in a graph.
b) Will the new commission structure motive salespeople to work harder, and why?
c) Which scheme would the salespeople prefer?
d) The company is trying to develop the optimal compensation scheme for its workers and is
considering the following two options:
Offer 20% of all sales above some level .
Offer 25% of all sales above some level .
The company can freely choose the levels and respectively and the firms variable costs
equal 80% of the sales price. Assume the fixed salary paid by the insurance company is
enough to induce the workers not to shirk. Which commission scheme is more profitable for the
firm? Why?
Exercise 2.4 Choosing Employees

A profit-maximizing firm faces the following options for hiring employees:
Revenues Wage
Worker without diploma 80,000 20,000
Worker with diploma 100,000 35,000
a) Suppose the firm has limited space so that it can only hire one worker. Which type of worker
should the firm hire?
b) Suppose the firm can rent additional space so that it can employ two workers at a yearly rate
of 25,000. Should the firm rent the additional space? If yes, which type of worker should the
firm hire?
c) Suppose now that the firm has unlimited space and that it estimates its market demand for
this year to be 800,000. Which type of worker should it hire?
30

lOMoARcPSD|1733731
Chapter 3: Industrial Organization

Industrial Organization studies the interaction of firms on markets. In this chapter, we introduce
you to two key concepts of Industrial Organization: market power (section 3.1) and market
concentration (section 3.2). We will argue that market power may lead to market inefficiencies.
In section 3.3, we examine market power in markets with the highest possible market
concentration: monopolies. Section 3.4 highlights how the government may intervene in
markets to fight the negative welfare consequences of market power. Section 3.5 includes a case
study on how firms establish market power by forming cartels.
3.1 Market power

In chapter 1, we concluded that a market is efficient (only) if the market price equals marginal
costs. The more the price deviates from marginal costs the lower the markets welfare is. Market
power is the ability of a firm to sell its products at a price above marginal costs. In other words,
market power indicates how much the market price deviates from marginal costs and, as a
consequence, how inefficient the market is relative to its welfare optimum. The most commonly
used measure of a firms market power is the Lerner index (), which is defined as follows:

= .

So, the Lerner index equals the difference between price and marginal costs, normalized as a
fraction of the price. This normalization is useful because it allows us to compare market power
for products with totally different prices. Note that in the neoclassical general equilibrium model
and in perfectly competitive markets, = , which implies that = 0. This chapters case
study shows that firms can establish non-negligible market power by forming cartels: In a meta-
study on more than 800 instances of cartel formation, the estimated average Lerner index turns
out to be equal to 30%. A more general lesson from this case study is that it is not just a
theoretical possibility for firms to sustain substantial market power.
To measure market power in a market (rather than for a single firm), we use a weighted average
of the Lerner indices of the firms active in that market. The weights for calculating the average
are the corresponding firms market shares. Suppose firms are active in the market. Individual
firm s market share is given by

= ,

where and denote firm s revenue ( = 1, , ) and total market revenue respectively. If
firm produces at marginal costs and sells its product for a price , its Lerner index equals

= .

The markets Lerner index is given by
= 1 1 + 2 2 + + = =1 .
Companies can obtain and retain market power in many ways. In the case of patents or
copyrights, the law protects firms against entry by competing firms so that firms can establish a
monopoly position on their products, which could guarantee their market power. As a
31

lOMoARcPSD|1733731
consequence, firms have a strong incentive to obtain patents and copyrights and challenge
competitors when they breach them. Companies can also obtain and retain market power
without the protection of the law, sometimes even violating the law: Economic ways of getting
and maintaining market power include advertising, collusion, mergers and acquisitions, and
predatory behavior. I will discuss each of these business strategies in much more detail in the
remainder of the lecture notes.
What are the welfare consequences of market power? At first sight, the price does not seem to
matter much: If consumers buy a good at a higher price, the loss in consumer surplus is exactly
compensated by a gain in producer surplus, so the net effect on welfare is zero. However, some
consumers will decide not to buy, so that a high price reduces quantity sold and prevents some
value-enhancing transactions to take place. In short, market power induces allocative
inefficiency. Allocative inefficiency may be substantial. Consider a market with a linear inverse
demand function () given by
() = .
and constant marginal costs:
= .
Figure 3.1: The deadweight loss of above marginal cost pricing.
Figure 3.1 depicts the deadweight loss (i.e., welfare loss) of above marginal cost pricing. Note
that market demand is equal to

() = .

Now, it is easy to verify that the deadweight loss equals
32

lOMoARcPSD|1733731
1 1
= ( )(() ()) = ( )2 .
2 2
So, the deadweight loss of above marginal cost pricing is a quadratic function of the difference
between price and marginal costs.
The efficiency loss may be even greater because of productive inefficiency, dynamic inefficiency,
and influence costs. Firms with market power may have an easy life in the sense that they feel
less competitive pressure then in a competitive market. More in particular, the firms may
experience little pressure to reduce costs (leading to productive inefficiency) or to innovate
(leading to dynamic inefficiency). Influence costs may add to the efficiency loss: Firms with
market power could lobby the government to protect their market power, for example, by
restricting imports by foreign competitors. The resources spend on lobbying are wasted in the
sense that they do not create consumer value.
Still, market power could be beneficial for welfare. In particular, the very prospect of getting
market power may induce firms to innovate. A firm will only take the trouble of developing new
products or production technologies if it can reap the benefits in the form of high post-
innovation profits. This is precisely the reason why government agencies award patents and
copyrights to give innovative firms market power for some period of time. While the resulting
market power leads to allocative inefficiencies once the product is developed, welfare may be
even lower without it because firms will develop fewer new product and production
technologies.
3.2 Market concentration

Market concentration is another key concept in Industrial Organization. Roughly, a market is
concentrated if only a handful of firms have a serious market share. We will see throughout the
course that there is an intuitive link between market concentration and market power: Only in
concentrated market do firms manage to establish market power. In contrast, competition is
likely to be fierce in markets where many firms are active, resulting in little market power. As a
consequence, it is not coincidence that market concentration plays an important role in public
policy. As we will discuss in more detail in section 3.4, the government mainly intervenes in
concentrated markets. Competition authorities tend to be less lenient toward mergers in more
concentrated markets. Economic regulation is often only relevant for highly concentrated
markets.
The Herfindahl-Hirschman index () is the most commonly used measure of market

concentration. A markets Herfindahl-Hirschman index is the sum of the squares of the market
shares of the firms active in the market:
= (1 )2 + (2 )2 + + ( )2 = =1( )2 .
The Herfindahl-Hirschman index is always a number between 0 and 1. = 1 corresponds to a

monopoly, while if is close to zero then the market consists of many very small firms.
Sometimes, the index is multiplied by 10,000, which is the index outcome when market shares
are entered as percentages rather than fractions. Table 3.1 includes an example based on the
Dutch market for mobile telecommunications.
Note that in the case of equally sized firms,
33

lOMoARcPSD|1733731
1
= .

This indicates that is a natural measure for market concentration: The more equally sized
firms are active in the market, the lower the index is. Of course, in most markets, market shares
differ between firms. Still, we could use the expression for equally sized firms to get a
meaningful idea about how concentrated the market is. A market having market concentration
is as concentrated as a market with
1
=

equally sized firms. In the example from table 3.1,
1 1
= = = 3.01.
0.3326
So, the Dutch market for mobile telecommunications is roughly as concentrated as a market with
only three firms each serving one-third of the market. It seems fair to say that the market is quite
concentrated.
Table 3.1: Market concentration in the Dutch market for mobile telecommunications
Company ( )
KPN 0.48 0.2304
Vodafone 0.26 0.0676
T-Mobile 0.15 0.0225
Orange 0.11 0.0121
Herfindahl-Hirschman index 0.3326
Competition authorities typically estimate the effect of the merger on the Herfindahl-Hirschman
index to get an idea about the impact of the merger on the market power. Although market
concentration is not the same as market power, the two concepts are closely related. Specifically,
under quite general assumptions, it can be shown that market concentration and market power
are linked according to the following formula:

= ,

where
()
= .
()
represents the price elasticity of demand. The relation between market power, market
concentration, and price elasticity indicates where we can find firms with considerable market
power: in concentrated markets (high ) where demand is relatively price inelastic (low ).
There is a practical reason for competition authorities to calculate the Herfindahl-Hirschman
index instead of the Lerner index. To calculate the Herfindahl-Hirschman index it is sufficient to
know the market shares, which is often much easier to obtain than information about marginal
costs, which is needed to calculate the Lerner index,
34

lOMoARcPSD|1733731
A practical concern for the determination of the Herfindahl-Hirschman index could be that the
competition authority does not know the market shares of all firms in the market. In that case,
the competition authority can only estimate the index. A lower bound for is obtained by
calculating the squares of the known market shares. This number results in a lower bound
because adding the squares of the unknown market shares can only increase the index. In fact,
the resulting estimate is the actual lower bound of the value the Herfindahl-Hirschman can take
because theoretically the remaining market share could be catered by a zillion very small firms.
An upper bound of emerges by assuming that as few as possible firms are responsible for the
remaining market share. For example, if in table 3.1 the market share of Orange was missing, we
would be able to find the upper bound of by assuming that the remaining 11% market share
was served by a single firm. If we also knew that the remaining firms have at most 5.5% market
share each, would be maximized in the case that two firms served 5.5% of the market.
3.3 Monopoly: The inverse elasticity rule

Monopolies are markets in which only one supplier is active. In most countries, electricity
transmission, passenger railway transportation, and water supply are monopolies. Other
examples include markets that are protected by patents and copyright, such as markets for
medical drugs or movies. A monopoly is at the extreme end of the distribution of possible
market concentrations. As the sole supplier in the market holds a 100% market share, the
Herfindahl-Hirschman index equals = 1.
Figure 3.2: The cyclists inverse demand for a soft drink.
In general, a firm can find its profit-maximizing quantity by equating its marginal revenue to its
marginal costs. To illustrate this rule, consider a simple example based on the market for soft
drinks depicted in figure 3.2. Imagine a thirsty cyclist entering an isolated French mountain
village. The cyclist is looking for something to drink. He soon finds out that the village has no
bars and only one store. Luckily, the store sells soft drinks, which is exactly what the cyclist is
looking for. The figure shows the cyclists inverse demand curve for bottles of soft drink. For the
first bottle of soft drink, he is willing to pay 5, for a second 4, and so on, down to 0 for a sixth
35

lOMoARcPSD|1733731
bottle. The marginal cost for each bottle equals 1. What price should the shop owner charge the
cyclist to maximize profits?
Table 3.2 is helpful in deriving the answer. The shops marginal revenue can be derived by
calculating the additional revenue for each bottle sold. For the first bottle, the marginal revenue
equals 5. For the second, it is the difference between 8 (the shops revenue if it sells two bottles
at a price of 4) and 5 (revenue if the shop sells one bottle), and so on. Note that it is profitable for
the shop to sell up to three bottles, because the marginal revenue of each bottle covers its
marginal costs. Clearly, selling a fourth bottle is not profitable as the corresponding marginal
revenue is negative. Indeed, the optimum lies at the quantity where marginal revenue equals
marginal costs
Table 3.2: Marginal revenue in the market for soft drink
Quantity Price Revenue Marginal revenue Marginal costs

1 5 5 5 1
2 4 8 3 1
3 3 9 1 1
4 2 8 1 1
5 1 5 3 1
In general, a shortcut to a monopoly firms profit-maximizing price is the inverse elasticity rule:
1
= .

The inverse elasticity rule follows directly by equating the firms marginal revenue to its
marginal costs. The firms revenue equals
= ()
so that its marginal revenue is given by
= () + ().
Equating this to marginal costs, we have
= () + () = ,
which is equivalent to
() = ()
Dividing by () implies
() ()
= .
() ()
This equation can be rewritten as
1
= .

36

lOMoARcPSD|1733731
In words: a profit-maximizing monopolist makes sure that its Lerner index equals one over the
price elasticity of demand. The inverse elasticity rule implies that in general, a monopoly firm
has market power because > 0 implying that > 0. Finally, note that the inverse elasticity rule
is a special case of the formula
because for a monopoly, = 1.
3.4 Government intervention

One of the central questions of Industrial Organization is whether there is a role for government
intervention in the case of market power. In this section, we will highlight two ways in which
governments can deal with market power: competition policy and economic regulation.
The main goal of competition policy (or antitrust policy) is to promote competition in markets
in general. Competition policy is embodied in competition law. In most developed countries,
competition law consists of three pillars: anti-cartel law, abuse of a dominant position, and
merger control. Anti-cartel law forbids cartel agreements between firms, i.e., agreements that
restrict competition in markets, including agreements about price, production, and market
division. For example, Paragraph 1 of Article 101 of the Treaty on the Functioning of the
European Union reads:
The following shall be prohibited [...]: all agreements between undertakings [...] which:
(a) directly or indirectly fix purchase or selling prices or any other trading conditions;
(b) limit or control production, markets, technical development, or investment;
(c) share markets or sources of supply;
(d) apply dissimilar conditions to equivalent transactions with other trading parties,
thereby placing them at a competitive disadvantage;
(e) make the conclusion of contracts subject to acceptance by the other parties of
supplementary obligations which, by their nature or according to commercial usage, have
no connection with the subject of such contracts.
When firms do make cartel agreements, they risk heavy punishment including fines for the firms
and, in some countries, imprisonment for their managers.
Most competition authorities make exceptions to the general rule that cartels are forbidden.
Group exemptions in the European Union are an example: The European Commission explicitly
allows for particular agreements in specific markets, such as in the car industry. Sometimes,
firms can get dispensation for agreements between them. Dispensation is usually granted to
promote technological progress, for instance, in research joint ventures.
Abuse of a dominant position is the second pillar of competition law. The law stipulates that
dominant firms (firms with a substantial market share, usually more than 40%) are not allowed
to abuse their dominant position. Examples of abusive practices include price discrimination,
product bundling, and predatory pricing. Keep in mind that establishing a dominant position is
37

lOMoARcPSD|1733731
not illegal in itself. In the European Union, abusing a dominant position is illegal under Article
102 of the Treaty on the Functioning of the European Union, which states:
Any abuse by one or more undertakings of a dominant position within the internal market
or in a substantial part of it shall be prohibited as incompatible with the internal market in
so far as it may affect trade between Member States.
Such abuse may, in particular, consist in:
(a) directly or indirectly imposing unfair purchase or selling prices or other unfair trading
conditions;
(b) limiting production, markets or technical development to the prejudice of consumers;
(c) applying dissimilar conditions to equivalent transactions with other trading parties,
(d) making the conclusion of contracts subject to acceptance by the other parties of
Merger control is the third pillar of competition law. A competition authority may block
concentrations (as legal jargon has it) such as mergers, acquisitions, and far-reaching joint
ventures when it expects adverse effects on competition. According to the horizontal merger
guidelines of the European Commission,
the Commission must assess [...] whether or not a concentration would significantly impede
effective competition, in particular as a result of the creation or strengthening of a
dominant position, in the common market or a substantial part of it.
In its evaluation, the European Commission takes both market shares and (changes in) market
concentration into account to obtain a useful first indications of the market structure and of the
competitive importance of both the merging parties and their competitors. The analysis in this
chapter indicates that this makes sense because market power is closely linked to market
concentration.
While competition policy is relevant for markets in general, economic regulation only applies
to specific markets. Regulated markets are usually monopolies or near monopolies including
markets for electricity transmission, passenger railway transportation, and water. Those
markets are supervised by a regulator that has legal tools to intervene in the market in order to
mitigate welfare losses, for example, by imposing maximum prices to prevent the monopolies
from charging excessive prices. Our analysis above suggests that such intervention is justified
because the markets are not only highly concentrated but are often also characterized by price
inelastic demand. In an unregulated market, the resulting market power and the corresponding
welfare loss can be substantial.
The two most commonly used types of economic regulation are rate-of-return regulation and
price-cap regulation. In the case of rate-of-return regulation, the regulator sets prices such that
the firm is exactly compensated for the costs it incurs, including a fair rate-of-return. Rate-of-
return regulation has the advantage that allocative inefficiency is reduced as much as possible
38

lOMoARcPSD|1733731
because price moves down from the monopoly price toward the marginal cost price. A potential
disadvantage is that firms have hardly any incentives to reduce costs because cost savings
reduce the price a firm is allowed to charge. As a consequence, the market could become highly
productively inefficient (which conveniently coincides with the stereotype of sluggish utilities in
the 1980s, which was the period in which the rate-of-return regulation was a common way of
regulating utilities in Europe and the US).
Under price-cap regulation, the regulator simply sets a maximum price, usually for a long period
of time, taking inflation and technological progress into account. The firm can charge any price
up to this maximum (and keep the profits). Price-cap regulation reduces allocative inefficiency
as much as possible. Moreover, in contrast to rate-of-return regulation, it gives the firm
incentives to cut costs because the firm can keep every euro it saves by reducing costs. There are
disadvantages as well. First of all, it may not be straightforward for the regulator to find the
optimal price cap. Ideally, the price cap should be equal to the firms marginal costs, but often
the regulator lacks information about the firms costs. Both too high and too low a price cap
result in market inefficiencies, in particular when the firm goes bankrupt and is forced to leave
the market if the price is too low. Second, price-cap regulation requires strong commitment from
the regulator. The regulator may be tempted to tighten the price cap when the firm decreases
costs because this would bring the price closer to marginal costs. This is called the ratchet
effect because the regulator ratchets up the standards in response to the firms efforts to cut
the costs. However, when firms anticipate that the regulator may do so, they will think twice
before they invest heavily in cost reductions. Third, the regulated firm may be tempted to cut
costs by reducing quality, which may harm consumers.
3.5 Case study: Cartels and market power

How much market power do cartels manage to sustain? To answer this question, John M. Connor
and Yuliya Bolotova of Purdue University employed a meta-study of quantitative estimates of
cartel market power. Their sample contains 395 price-fixing cartels that were active in different
markets in several countries during the last 250 years. Connor and Bolotova measured market
power by comparing observed cartel price changes to some competitive benchmark, a method
resulting in a market power measure similar to the Lerner index. To obtain a competitive
benchmark, the researchers collected prices for comparable markets that they believed to be
free from cartelization. For example, the competitive yardstick can include prices in a nearby
city or a neighboring region with similar demand or cost conditions; the cartel price can then be
compared to the yardstick during the collusive period.
In accordance with their hypothesis, Connor and Bolotova found cartel prices to be significantly
higher than their competitive benchmarks. The mean value derived using various statistical
methods was estimated to be about 29% above the competitive yardstick price. In addition, the
organizational characteristics of cartels and their market environment appear to have a
substantial impact on the variability of the market power estimates. For instance, cartels tend to
have less market power in countries with strongly enforced competition laws: The degree of
market power attained by cartels operating in North America and the EU was found to be lower
than the market power of the same or similar cartels operating in Asia or Latin America.
Moreover, market power tends to be greater for international cartels: Their market power is
about 14 percentage points higher than that of domestic cartels. Finally, durable cartels tend to
be very effective in driving up the price: For each five additional years of cartel operation,
market power rose by about 4 percentage points.
39

lOMoARcPSD|1733731
The results of this article contain a number of suggestions for anti-cartel policies. First of all, it
pays to strongly enforce competition laws. The main priority of antitrust policies should be to
uncover and weaken durable, global price-fixing collusive violations. In addition, programs
aimed at the development of antitrust expertise in underdeveloped regions will also serve the
interest of consumers in advanced countries. Weak implementation of competition law in
underdeveloped regions not only spurs high rates of local cartel formation but also increases the
number of global cartels that affect consumers of high-income regions. Furthermore, according
to Connor and Bolotova, an effective deterioration of international and global cartels requires
more active intervention. More resources and greater effort should be deployed toward harsher
penalties for international cartels compared to domestic ones. In addition, durable cartels ought
to be fined at higher rates than short-lived ones, as the longitude of existence tends to increase
market power.
It seems that competition authorities are moving in the right direction: Bolotova and Connor
observe that over time, competition policy has become significantly more effective in the sense
that cartel overcharges have decreased: The market power by cartels between 1991 and 2004
was 13 percentage points lower than in the period 1770-1919.
In this chapter, we began our study of industrial organization by reviewing the basic tools of
analyzing a market. Industrial organization is a field within economics that rapidly developed in
the 1980s as a result of advances in microeconomic theory, including transaction cost theory,
information theory, and game theory.
Although French mathematicians Antoine Augustin Cournot and Joseph Louis Franois Bertrand
developed the first models of imperfect competition as early as the 19th century (see chapter 6),
the initial development of the field can be attributed to the Harvard School economists in the
1930s. They were among the first who started thinking out of the sphere of perfect competition
and shaped the structure-conduct-performance approach in industrial economics. Notably, the
study of the enterprise and imperfect competition find their first extensive treatment in
Chamberlins (1933) Theory of Monopolistic Competition. Another study of the structure of
large-scale enterprises can be found in Masons influential paper, published in 1939 in the
American Economic Review. The field of industrial organization was definitely established with
the publication of Joe Bains Industrial Organization: An analysis of competing markets in
1958.
For more recent publications that provide a good introduction to the field you can refer to the
textbooks of Cabral (2000) and Pepall et al. (2008). More advanced textbooks on Industrial
Organization include Tirole (1988), Belleflamme and Peitz (2010) and Pepall et al. (2011).
Viscusi et al. (2005) covers both competition policy and economic regulation. Mottas (2004)
textbook discusses competition policy at an advanced level. You can also find a citation to the
treaty of the European Union in the reference list below. We will frequently refer to the articles
of the treaty in later chapters, especially as far as European competition policy is concerned.
The Lerner index is named after the economist Abba Lerner, who studied extensively the
measurement of market power in his 1934 paper. In addition, the HerfindahlHirschman Index
(HHI) is named after the economists Orris Herfindahl and Albert Hirschman. You can read more
40

lOMoARcPSD|1733731
about the paternity of the index in a note Albert Hirschman published in the American Economic
Review in 1964.
The case study of this chapter is based on the findings of Connor and Bolotova (2006) that
appeared in the International Journal of Industrial Organization.
References
Bain, J.S., J. Lipczynski and J.O.S. Wilson (1958). Industrial Organization: An Analysis of Competing
Markets, New York: Pearson Education.
Belleflamme, P. and M. Peitz (2010). Industrial Organization: Markets and Strategies, Cambridge:
Cambridge University Press.
Cabral, L.M. (2000). Introduction to Industrial Organization, Cambridge, MA: MIT Press.
Chamberlin, E. (1933). The Theory of Monopolistic Competition, Cambridge, MA: Harvard
University Press.
Connor, J. and Y. Bolotova (2006). Cartel overcharges: survey and meta-analysis, International
Journal of Industrial Organization, 24(6), 1109-1137.
European Commission (2012). Consolidated versions of the treaty on European Union and the
treaty on the functioning of the European Union, Official Journal of the European Union, 55(C
326), 3-390.
Hirschman, A.O. (1964). The paternity of an index, American Economic Review, 54(9), 761-762.
Lerner, A.P. (1934). The concept of monopoly and the measurement of monopoly power, The
Review of Economic Studies, 1(6), 157-175.
Mason, E.S. (1939). Price and production policies of large-scale enterprise, American Economic
Review, 29(3), 61-74.
Motta, M. (2004). Competition Policy: Theory and Practice, Cambridge, MA: Cambridge University
Press.
Pepall, L., D. Richards and G. Norman (2008). Industrial Organization: Contemporary Theory and
Empirical Applications, Malden, MA: Blackwell Publishing.
Pepall, L., D. Richards and G. Norman (2011). Contemporary Industrial Organization: A
Quantitative Approach, Hoboken, NJ: John Wiley and Sons.
Viscusi, W.K., J.M. Vernon and J.E. Harrington (2005). Economics of Regulation and Antitrust,
Cambridge, MA: MIT Press.
Exercises
Exercise 3.1 Monopoly

You own a pharmaceutical company that is specialized in the production of medicine for
smokers. You recently patented an innovative drug called Clealung, which drastically decreases
the smokers likelihood of lung cancer. You are the only producer of the medicine and there are
no similar products in the market. Your costs for producing Clealung can be described by:
() = 30,
where is the amount of Clealung boxes you produce. The market demand for your medicine is:
() = 2000 5,
where is the price per box of Clealung.
41

lOMoARcPSD|1733731
a) Write down your firms profits in two ways: (i) as a function of the price per box and (ii) in
terms of the quantity.
b) Maximize both profit functions, and show that in both cases = at the optimum. In
addition, compute the profit.
c) Calculate the producer and consumer surplus.
Being an altruist, you realize that Clealung should be used for the good of the society, instead of
your firms profits. In fact, your altruism prompts you to maximize social welfare at the
condition of not incurring a loss.
d) What price would you impose in order to maximize welfare? How many boxes of Clealung
will you sell at that price?
e) To what extent is the society better off at the new price? Compute the new consumer surplus,
producer surplus and estimate the total gain for the society.
Exercise 3.2 Monopoly and Market Power

The fez is the typical Arabic hat in the shape of a short red cylinder. Historians believe it was
developed in the city of Fes, in Morocco, during the 17th century. It has been used as a head cover
of the Arabic world for centuries and was also adopted by the Ottoman Empire as part of its
military uniform. The fez is still the most common hat in North Africa and the Middle East; the
hat was banned at the beginning of 20th century in Turkey. Assume that the city of Marrakech in
Morocco has a daily demand for fezzes given by:
() = 180 2,
where is the price for a traditional hat in Moroccan dirham. The cost for producing fezzes is
given by:
() = 4,
where is the total amount of fezzes produced.
a) Derive the equilibrium price and quantity assuming that the market for fezzes is a monopoly.
b) Derive the equilibrium price and quantity assuming that the market of fezzes is perfectly
competitive.
c) Compute the elasticity of demand in both equilibria.
d) The demand elasticity in the case of a monopoly is really close to 1. In which cases does a
monopolist produce with an elasticity of demand lower than 1? Explain.
1
e) Show that = in the monopoly case, where is the Lerner Index and is the demand
elasticity both in absolute values.
Exercise 3.3 Market Concentration and Competition Policy

Consider the following table. It presents the market shares of seven clothing stores (A to G) in
five different cities.
42

lOMoARcPSD|1733731
City A B C D E F G Rest
Amsterdam 0.20 0.15 0.06 0.15 0.11 0.17 0.16 0.00
Rotterdam 0.10 0.40 0.10 0.10 0.00 0.08 0.15 0.07
Den Haag 0.70 0.06 0.02 0.03 0.03 0.00 0.00 0.16
Utrecht 0.30 0.05 0.25 0.25 0.00 0.15 0.00 0.00
Eindhoven 0.25 0.25 0.25 0.25 0.00 0.00 0.00 0.00
a) Calculate the Herfindahl index () for each city. If an exact calculation is not feasible, give an
upper and lower bound.
1
b) Calculate for Utrecht. Give an interpretation of this value.
c) Suppose that in Den Haag clothing stores A to E are the largest ones. What is the maximum ?
d) Suppose that firms C & A intend to merge and in turn notify the European Commission. For
which city is it less likely for the Commission to block the merger according to its guidelines?
You may assume that the market shares do not change after the merger.
Exercise 3.4 Market Concentration

In the country of Sleep-well, the inhabitants main activity is sleeping. Despite the loss of
productivity that this entails, the country has a profuse and renowned production of blankets,
mattresses, bed covers, and the like. Particularly, the market for pillows is Sleep-wells flagship;
the market is perfectly competitive and has firms present, all of which have the same cost
function. The aggregate production is 8,000 super-comfort pillows per day.
a) Compute the Herfindahl-Hirschman Index () of Sleep-wells market for pillows as a function
of .
In order to increase their market power, firms decide to merge (with > 1). Assume that the
merger does not imply any changes in production, i.e., the mergers total production will be the
sum of the production of the firms.
b) Compute again the as a function of and . (Hint: Start by computing the firms post-
merger market share. Keep in mind that the market has + 1 firms now, the merged firm
and firms that did not participate in the merger).
c) Show that, according to the assumptions made so far, a merger always increases .
d) What would happen to the index if = ? What does this imply?
Assume now that = . Although the pillow markets demand is difficult to estimate, Sleep-
wells economists are sure that it has a constant price elasticity equal to 1.5. The production of
super-comfort pillows has a constant marginal cost of 8 Dreamies (Sleep-wells current
currency).
e) Given that the newly merged firm is maximizing its profit, what will be the price of a super-
comfort pillow?
43

lOMoARcPSD|1733731
Part II: Strategic interaction

In part I, we studied situations in which decision makers in organizations acted in isolation:
They did not have to take into account that their actions could affect the actions taken by others,
which might have led them to reconsider their decisions. We discussed workers in a firm who
maximize effort conditional on the incentives the contract with the firm might give them.
Because a workers payoff did not depend in any way on the decisions by other workers, he or
she did not have to take decisions by others into account when deciding on how hard to work.
Similarly, a firm in a perfectly competitive market has no impact on the prices competitors pay
for their inputs or outputs so that its decisions do not affect the decisions by other firms. At the
other extreme (in terms of market concentration), a monopoly firm does not have any
competitors by definition, so that it simply does not interact with other firms in the same market.
But what if decision makers do interact? What if workers payments do depend on how hard
their colleagues work? What if firms are active in a moderately concentrated market and
compete against only a handful of similarly sized firms? Those are precisely the settings we will
discuss in this part and, in fact, in all remaining parts. Specifically, we will use game theory to
study strategic interaction between several decision makers. Game theory has been successfully
applied to explain a wide range of human (and even animal) behavior, not only in economics and
business, but also in psychology, sociology, political science, computer science, and biology. In
chapter 4, I will introduce you to game theory and its concepts. In chapter 5, we will apply game
theory to workers decisions in firms. Chapter 6 contains applications to strategic interaction on
markets.
44

lOMoARcPSD|1733731
Chapter 4: Game theory

Game theory is a powerful tool to study strategic interaction among two or more decision
makers. In this chapter, I will introduce you to game theory. I will start in section 4.1 by
presenting what is probably the most famous game from game theory: the prisoners dilemma.
In section 4.2, I will discuss the Nash equilibrium, which makes a prediction about the actions
the decision makers will take. This section also includes an application to the worlds most
important game (in a more popular interpretation of the word game): football. In section 4.3,
we will apply game theory to a market game called the Hotelling game. The case study in
section 4.4 contains some evidence that peoples behavior is in line with the Nash equilibrium
predictions.
4.1 The prisoners dilemma

Imagine that the police catch two robbery suspects and put both in separate prison cells. So far,
there is not sufficient evidence against either prisoner for the robbery, but there is a strong case
against both of them for other, minor offenses. If the police presented this evidence in court,
both prisoners would be imprisoned for one year. The robbery is considered a much more
serious offense, however, for which the suspect risks a two-year imprisonment. The police are
aware that they will have sufficient evidence against one prisoner if the other prisoner testifies
against him. A police officer proposes the following: Let us ask each prisoner to testify against
the other prisoner. To reward a prisoner for confessing, we will not prosecute him for the minor
offenses he committed. Table 4.1 includes the utilities both prisoners will obtain depending on
who denies or confesses that he has evidence against the other prisoner.
Table 4.1: The prisoners dilemma
Prisoner 2
Deny Confess
Deny 4,4 2,5
Prisoner 1
Confess 5,2 3,3
Let me explain table 4.1 in more detail. The numbers in the table refer to the prisoners utilities
given their own action (Deny or Confess) and the other prisoners action. The first number
represents the utility of prisoner 1, the second refers to the utility of prisoner 2. If a prisoner
does not go to prison, his utility equals 5. For each year he spends in prison, his utility is reduced
by one. So, if both prisoners confess, both will go to prison for two years for the robbery and
both will be relieved of the minor crimes, resulting in a utility of 5 2 = 3 for both. If one
prisoner confesses and the other denies then the former will obtain utility 5 (because he will not
go to prison) and the latter will get utility 1 (because he will go to prison for both the robbery
and the minor offences). Finally, if both deny, their utility will be 4 because they will get a one-
year prison sentence for the minor offences and they will not be sentenced for the robbery by
lack of evidence. The prisoners dilemma is a typical example of a simultaneous move game.
A simultaneous move game has the following four elements:
A set of players
For all players, the actions they can take
For all players, the information they possess when they decide which action to play
45

lOMoARcPSD|1733731
For all players, the utility they obtain given their own actions and the actions of other
players
In the prisoners dilemma, the set of players consists of the two prisoners. They can choose
between two actions: Confess and Deny. Both prisoners have complete information about the
game, that is, both are aware of how much utility either of them obtains conditional on the
actions chosen by both. Finally, their utilities, conditional on the actions, are given in table 4.1.
What will the prisoners do? Both might be attracted by the action Deny, because this will give
either of them a payoff of 4 if the other does the same, which is quite a bit better than the payoff
of 2 if they were both to confess. However, if one prisoner denies, the other has an incentive to
confess because his utility will increase from 4 to 5. In fact, the other prisoner is always better
off by choosing Confess instead of Deny regardless of the first prisoners action. It seems that
both prisoners are likely to confess resulting in a sad outcome for both of them (but a good one
for the police and the rest of society). In the next section, we will explore in greater generality
how to predict outcomes of the prisoners dilemma and other simultaneous games.
4.2 Nash equilibrium

In 1994, John F. Nash shared the Nobel Memorial Prize in Economic Sciences with fellow game
theorists Reinhard Selten and John Harsanyi. John Nash became well known to a wide audience
because he is the main character in the Hollywood movie A Beautiful Mind. The movie actually
focuses mainly on Nashs suffering from paranoid schizophrenia but it also highlights one of
Nashs most important contributions to economics (albeit not very convincingly): his solution
concept for simultaneous games, the Nash equilibrium, as we call it in his honor.
Let me give a definition of the Nash equilibrium. Players strategies constitute a Nash
equilibrium if none of the players has a reason to choose another strategy given the other
players strategies. The prisoners dilemma has a Nash equilibrium in which both prisoners pick
the action Confess. Indeed, a prisoner has no incentive to deviate to Deny given that the other
prisoner testifies because by doing so his utility will decrease from 2 to 1. You can check that
(Confess, Confess) is the only Nash equilibrium. Any other combination of two actions is not an
equilibrium because the player opting for Deny is strictly better off by deviating to Confess.
Table 4.2: Best responses in the prisoners dilemma
Prisoner 2
Deny Confess
Deny 4,4 2,5
Prisoner 1
Confess 5,2 3,3
The definition of the Nash equilibrium can be rephrased in terms of best responses, resulting in
a definition that might be easier to apply. A players best response against the actions by other
players is the players action that maximizes his payoff. In other words, given the choices by the
other players, a players best response is an action that results in at least as high a payoff as any
other action he can choose from. Let me illustrate this in the above prisoners dilemma. What is
Prisoner 1s best response against Prisoner 2 choosing Deny? It is Confess because if
Prisoners 2 chooses Deny, Prisoner 1 will obtain greater utility by choosing Confess than by
46

lOMoARcPSD|1733731
choosing Deny. In table 4.2, I highlight the best responses of both prisoners against the two
possible actions by their opponents by underlining the corresponding payoffs.
The Nash equilibrium of a game can now be expressed as follows in terms of best responses:
Players actions constitute a Nash equilibrium if each player plays a best response against the
actions of the other players. So, we have identified a Nash equilibrium when observing that all
players play a best response against the others. From table 4.2, it becomes clear that (Confess,
Confess) is a Nash equilibrium because both prisoners play a best response against the action
chosen by the other prisoner. Moreover, it is also immediately clear that (Confess, Confess) is the
only Nash equilibrium of the prisoners dilemma because it is the only set of actions in which
both players play a best response.
In many cases it is quite straightforward to discover all Nash equilibria of a game by

systematically identifying the best responses of all players against all actions of the opponents
(for instance, by underlining the corresponding payoffs). The game in table 4.3 (which is
inspired by a scene in the movie A Beautiful Mind) illustrates this point. Suppose two male
students, Antonio and Brad, sit in a bar vividly discussing deep philosophical questions about
economics and business. Then, three beautiful ladies enter the bar, one with blonde hair and the
other two with brown hair. The men are immediately distracted from their discussion and start
talking vigorously about the women. They agree that the blonde is the most attractive of the
three, although the brunettes certainly look good, too. Then they decide to go for it and approach
the ladies. Whom to approach first? If both go for the blonde, they will block each other and
neither will get her. The brunettes, in turn, will give them the cold shoulder because no one likes
to be second choice. The sad outcome is that both will go home alone. However, if both approach
the brunettes, they will not be turned down. Finally, both will get the woman they choose if one
of them goes for the blonde and the other for one of the other girls. Table 4.3 presents the
resulting payoffs for both students.
Table 4.3: The bar scene
Brad
Blonde Brunette
Blonde 0,0 3,2
Antonio
Brunette 2,3 2,2
From table 4.3, it becomes clear that the resulting game has two Nash equilibria. In either
equilibrium, one student chooses the blonde girl while the other goes for one of the brunettes.
You can find both equilibria by underlining the payoffs from the best response actions as shown
in the table.
You might wonder whether all games like the ones above have at least one Nash equilibrium.
The following case study on penalty shoot-outs in football shows they do not, at least if we do
not allow players to randomize over the actions they take. Indeed, so far we have restricted our
attention to non-random strategies (or pure strategies in game theory lingo): Antonio could
choose either the blonde or the brunette with certainty, but we did not allow him to choose the
blonde with probability 1/3 and the brunette with probability 2/3, for example. If we allow for
random strategies (or mixed strategies), generally all simultaneous move games with complete
information have at least one Nash equilibrium, a result established by John F. Nash himself.
47

lOMoARcPSD|1733731
4.3 The Hotelling game

We now move to the Hotelling game. In contrast to the games above, in the Hotelling game,
players can choose from more than two actions. In fact, they have infinitely many alternatives at
their disposal. Although such continuous games can become more complicated to analyze, the
Nash equilibria can still be found by systematically looking for combinations of actions that are
best responses for each player against the actions of all others.
In 1929, Harold Hotelling proposed the following game to study strategic interaction of firms in
the case of product differentiation, i.e., in settings where consumers differ in terms of
preferences over products. For instance, some consumers prefer a red car while others prefer a
blue car. The Hotelling game models the firms location choice, either literally or in terms of
product positioning. Suppose that two ice cream vendors, Clint and Dennis, have to decide where
to locate their ice cream stands on a beach. The beach is one kilometer long, with 1,000
consumers uniformly distributed over the entire beach, each willing to buy at most one ice
cream. The marginal cost of the ice cream equals for both vendors. The ice cream vendors both
sell the exact same kind of ice cream (there is no product differentiation apart from the shops
location) at the same price > . Assume that all consumers are willing to walk a kilometer to
buy an ice cream at price . As a consequence, all consumers will buy their ice cream at the
nearest stand. For instance, if both ice cream vendors locate at opposite ends of the beach, all
consumers to the left of the middle will go to the stand on the left and all remaining consumers
will walk to the stand on the right. In that case, both vendors will attract 500 customers.
So, where will the vendors locate? Given the above facts, their only target is to attract as many
customers as possible. Will both firms locate at the extreme ends of the beach in the Nash
equilibrium? No, they will not. If Clint locates at the extreme left end of the beach, Dennis will get
750 customers if he locates in the middle, which is clearly better than the 500 customers he gets
when locating at the extreme right end of the beach. In fact, the best response of Dennis is the
location just to the right of Clint. Doing so, he will be able to attract almost all 1,000 consumers.
More generally, the best response of Dennis is to locate right next to Clint making sure that he is
just somewhat closer to the middle of the beach. The only exception is if Clint locates exactly in
the middle. In that case, Denniss best response is to locate in the middle as well. Because the
same holds true for Clint, the only point at which both vendors play a best response is the middle
of the beach. In other words, the only Nash equilibrium is that Clint and Dennis both locate in the
middle.
The Hotelling game has many real world applications. For instance, it predicts that similar stores
tend to cluster, which is indeed the case in many shopping streets. Relatedly, in elections in
which only two (serious) parties compete for votes, both parties are likely to pick viewpoints
close to the middle of the political spectrum. The reason is that voters prefer the political
program that is closest to their preferences just as the beachgoers prefer to buy ice cream at the
stand that is located nearest to them. Indeed, in two-party elections, it is unlikely that either
party will present extreme viewpoints.
Of course, the Hotelling game relies on the strong assumption that prices are fixed. In most
markets, this assumption is not realistic. Indeed, if your competitor locates in the same spot as
you, you have an incentive to lower your price to attract additional consumers. In part III, we
will see that the predictions of the Hotelling game might be completely reversed if firms
determine the price after choosing to be located as far away from each other as possible.
48

lOMoARcPSD|1733731
4.4 Case study: Game theory in football Do professionals play Nash?

Ignacio Palacios-Huerta, professor at the London School of Economics, answers this question
using data from one of the worlds most popular games: football. In particular, he studies the
decisions made by professional football players in penalty kicks. By doing so, he provides an
empirical test of the Nash equilibrium concept and argues that professional football players
indeed maximize their expected payoff and end up being in equilibrium.
In football, a penalty kick is given against a team that commits one of the 10 punishable
violations inside its own penalty area while the ball is in play. These violations are described in
detail by the Federation Internationale de Football Association (FIFA) and set the rules of the
game. To test the implications of the Nash equilibrium concept, Palacios-Huerta uses a unique
data set of 1,417 penalty kicks from professional football games, including detailed information
on relevant aspects of the play, players actions, and final outcomes. He collected the data during
the period from September 1995 to June 2000 from a number of professional football games in
Spain, Italy, England and other countries. The data mostly comes from weekly TV programs in
these countries that review professional games played during the week, including all penalty
kicks that take place in these games.
Penalty kicks in football is a particular two-person, simultaneous move game. It is a game

characterized by unpredictability and one that requires mutual outguessing. Each penalty kick
involves two players: a kicker and a goalkeeper. In the typical kick the ball takes about 0.3
seconds to travel the distance between the penalty mark and the goal line. Hence, both the
kicker and the goalkeeper must move simultaneously in this game. The penalty kick has only
two possible outcomes, namely score or no score. Given the two possible outcomes, the kicker
wishes to maximize the expected probability of scoring, while the goalkeeper wishes to
minimize it. Players have a limited number of available strategies and their actions are directly
observable. The game cannot be repeated since there are no second penalties if a goal is not
scored. Furthermore, the initial location of both the ball and the goalkeeper is always the same:
the ball is placed on the penalty mark and the goalkeeper positions himself on the goal line,
exactly in the middle between the goalposts. Finally, the outcome is decided immediately after
players choose their strategies.
Table 4.4: Penalty kicks
Goalkeeper
Left Right
Left 0,1 1,0
Kicker
Right 1,0 0,1
These clearly defined rules and the general setting of the game prove to be very attractive in
providing an adequate empirical verification of the notion of Nash equilibrium. Consider, for
example, the following simple game, in which we assume that both players have only two
possible actions. The kicker can only choose to shoot to the left or to the right and the
goalkeeper must choose to dive to the left or to the right (from the kickers viewpoint). We also
assume that if both players choose the same side, the goalkeeper will always stop the ball, while
if they choose opposite sides, the shot will always result in a goal.
49

lOMoARcPSD|1733731
Note that there is no Nash equilibrium in pure strategies, i.e., where both players choose one
side with certainty. The reason is that the best response of each player is to choose the opposite
side of what the other player chooses. Still, the game does have a Nash equilibrium. Suppose that
both players choose either side with probability . Clearly, if the goalkeeper does so, it does not
matter for the kicker which corner to choose, so choosing each side with an equal probability is a
best response for the kicker. Using similar reasoning for the goalkeepers choice, we can
conclude that both players choosing either side with probability is indeed a Nash equilibrium.
As a result, equilibrium would require each player to randomize his action. This conclusion does
not only hold true in the particular game above, but also in more realistic settings where the
kicker may not score even if he chooses a different corner to the goalkeeper, where the kicker
may score even if he chooses the same corner as the goalkeeper, where the kicker has a
preferred corner, and so forth. In general, one can yield two testable predictions about the
behavior of kickers and goalkeepers. Firstly, the probability that a kicker scores a goal (or a
goalkeeper prevents it) is independent of the corner chosen. Second, a player's choice does not
depend on his choices in previous penalty kicks. Palacios-Huerta finds support for both
hypotheses. Indeed, his findings suggest that the fundamental notion of Nash equilibrium can be
supported with real data.
Modern game theory originates from the work of John von Neumann on the proof of the
existence of mixed-strategy equilibria in two-person games in 1928. More than a decade after
his initial paper, von Neumanns ideas were extended in the Theory of Games and Economic
Behaviour (1944), coauthored by Oskar Morgenstern, which is considered as the first book on
game theoretic analysis. Von Neumann and Morgenstern (1944) have also made significant
contributions in theoretical microeconomics, especially with the development of the theory of
expected utility (also in the Theory of Games and Economic Behaviour), which encouraged
economists to examine situations involving uncertainty and risk in economic decision making.
The field really boomed in the following decades, as mathematicians contributed significantly to
its development and academics realized the potential game theoretic analysis has in examining
problems in a variety of fields including most social sciences, biology and philosophy.
In addition, we should make a note on the concept of the Nash-equilibrium, which was
developed by the mathematician John Forbes Nash. Nashs work in game theory (1950 and 1951)
identified him as a significant figure in economics with his contributions setting a new paradigm
in game theory. You should refer to his papers "Equilibrium points in n-person games" (1950)
and "Non-Cooperative Games" (1951) for a rigorous treatment of game theoretic analysis of
non-cooperative games with more than two persons. You can read Sylvia Nasars A Beautiful
Mind (1998) and the 2002 Hollywood film for John Nash life story.
Applications of game theory abound. Introductory textbooks include Osborne (2004) and
Harrington (2009). Hotellings location model and some first insights on product differentiation
and strategic interaction can be found on his 1929 paper Stability in Competition, published in
the Economic Journal. The case study of this chapter stems from the 2003 paper by Ignacio
Palacios-Huerta in the Review of Economic Studies.
References
Harrington, J. (2009). Games, Strategies and Decision Making, New York: Worth Publishers.
50

lOMoARcPSD|1733731
Hotelling, H. (1929). Stability in competition, Economic Journal, 39(3), 41-57.

Nasar, S. (1998). A Beautiful Mind: The Life of Mathematical Genius and Nobel Laureate John Nash,
New York: Simon Schuster.
Nash, J. (1950). Equilibrium points in n-person games, Proceedings of the National Academy of
Sciences, 36(1), 48-49.
Nash, J. (1951). Non-cooperative games, The Annals of Mathematics, 54(2), 286-295.
Neumann, J.v. (1928). Zur theorie der gesellschaftsspiele, Mathematische Annalen, 100(1), 295-
320.
Neumann, J.v. and O. Morgenstern (1944). Theory of Games and Economic Behavior, Princeton, NJ:
Princeton University Press.
Osborne, M.J. (2004). An Introduction to Game Theory, New York: Oxford University Press.
Palacios-Huerta, I. (2003). Professionals play minimax, Review of Economic Studies, 70(2), 395-
415.
Exercises
Exercise 4.1 Nash Equilibrium

Two competing firms are each planning to introduce a new product. Firm 1 will decide whether
to produce product A, product B or product C, while firm 2 can choose between products D, E,
and F. They will make their choices at the same time. The resulting profits are shown below.
Firm 1
A B C
D 10, 20 0, 35 30, 80
Firm 2 E 50, -20 40, 80 30, -10
F 0, 20 35, 30 -10, 50
What are the Nash equilibria of this game?
Exercise 4.2 The battle of the sexes

Ingrid and Jeff would like to spend Saturday night together but have different tastes in
entertainment. Jeff would like to go to the opera but Ingrid would prefer to watch soccer. As the
following payoff matrix shows, Jeff would most prefer to go to the opera with Ingrid, but prefers
watching soccer with Ingrid to going to the opera alone, and similarly for Ingrid.
Jeff
Soccer Opera
Soccer 2, 1 0, 0
Ingrid
Opera 0, 0 1, 2
a) Determine the Nash equilibria of this game, assuming that Ingrid and Jeff do not randomize
over actions.
Strategies in which a player makes a specific choice or takes a specific action will not always
result in a Nash equilibrium. In some cases, players can make a random choice among two or
51

lOMoARcPSD|1733731
more possible actions, based on a set of chosen probabilities. A strategy like this is called a
mixed strategy.
1
b) Suppose Jeff chooses to watch soccer with probability = . What will Ingrids expected
3
payoff be? Assume that Ingrid randomizes.
c) Should we expect Jeff and Ingrid to use these mixed strategies?
Exercise 4.3 Rock-Paper-Scissors

A mother is torn between choosing her son Leonardo and her daughter Meryl to have the last
bar of chocolate in her cupboard. As both her kids want the chocolate and she wants to be fair,
she decides to resolve the issue by having her children compete in the playground game of rock-
paper-scissors. According to the rules of the game, a rock (fist) breaks scissors (two fingers
striking out), scissors cut paper (flat hand), and paper smothers rock. If both players end up
choosing the same there is a tie and the game is repeated.
a) Show the payoff matrix for this rock-paper-scissors game (Hint: You may assume that the
payoff is -1 if you lose, 0 if you tie and 1 if you win).
b) What are Meryls best responses to the possible actions Leonardo might choose? What pure
strategy would you recommend when playing this game? What mixed strategy?
After playing the game, Meryl won the bar of chocolate: scissors beat paper. She told her mum:
Everybody knows you should start with scissors. Rock is way too obvious and scissors beats
paper. Scissors is the safest choice.
c) Is always choosing scissors part of the Nash equilibrium of this game?
Exercise 4.4 The beach location game

Suppose that Nicolas and Orson plan to sell soft drinks on a beach this summer. The beach is 400
meters long and sunbathers are spread evenly across its length. Nicolas and Orson sell the same
soft drinks at the same prices, so customers will walk to the closest vendor. Carrying the soft
drink a distance back to ones beach umbrella incurs transportation costs because walking to
and from the vendor takes time and effort, and the soft drink heats up more (and tastes worse)
the further a customer has to walk. Furthermore, the coast guard only allows the vendors to
locate in one of the following places on the beach, as indicated by the figure below.
l=0 l = 100 l = 200 l = 300 l = 400
a) Where will Orson locate if Nicolas puts a stand at the extreme end of the beach? Does Orsons
choice result in a Nash equilibrium?
b) What is the Nash equilibrium?
Now assume that prices are flexible. Nicolas and Orson will first pick a location on the beach and
then determine their prices after observing the location choice of the other vendor. You may also
assume that the marginal cost of selling an additional drink is equal to zero.
52

lOMoARcPSD|1733731
c) Does a Nash equilibrium exist in which both vendors locate in the middle?
d) Assume that transportation costs amount to = 0.002 per meter and that the utility that a
sunbather derives from a soft drink is equal to 2. Where should Nicolas and Orson locate if they
want to maximize welfare?
53

lOMoARcPSD|1733731
Chapter 5: Team incentives

In this chapter, we will study how members of teams can be motivated. Typically, the output of
an organization depends on the effort exerted by several people. A large range of workers take
part in assembling a single car, including mechanics, painters, and interior technicians. At your
university, you follow lectures by many professors, all contributing to you obtaining your degree.
In football, 11 players in a team join forces against the other team. We will consider several ways
to motivate individual team members, including rewarding each team member on the basis of
the performance of the entire team (section 5.2) and his comparative performance relative to
other team members (section 5.3). We begin in section 5.1 by presenting a model that will allow
us to evaluate the strengths and the weaknesses of those incentive schemes. In section 5.4s case
study we will compare the performance of teams of fruit pickers in two compensation schemes:
piece rates and a comparative performance scheme.
5.1 The model

Consider a team consisting of workers ( = 2, 3, ). If workers 1, 2, . . . , exert effort
1 , 2 , . . . , respectively, total team output is given by
= 1 + 2 + . . . + .
A team members cost () of exerting effort equals
() = 2
where > 0. There are no costs other than the costs of effort. Suppose that the output is sold in
the market at price per unit.
How much effort should each team member expend to maximize the teams profits? The teams
profits are given by
= (1 ) (2 ) . . . ( ) = (1 + 2 + . . . + ) [(1 )2 +
(2 )2 + . . . + ( )2 ].
To find the profit maximizing effort levels, you can write down the first-order condition for each
individual team members effort:
1 : 1 = 0
2 : 2 = 0
. ..
: = 0.
Therefore, for each team member optimal effort equals
= .
The outcome is intuitive. Optimal effort increases the greater the benefits from effort (the price
) and the lower the costs of effort (as measured by ).
54

lOMoARcPSD|1733731
A linear contract can implement the efficient outcome in a very similar way as we saw in chapter
2. Suppose that each team member obtains base wage plus a bonus for each unit of effort. An
individual will choose effort to maximize its utility:
= + 2 .
The first-order condition yields
= 0
or, equivalently,
= /.
So, if = , it immediately follows that every team member exerts effort equal to the profit
maximizing effort . Like we saw in chapter 2, it is optimal to let each decision maker become
the residual claimant of the fruits of his effort.
5.2 Paying for team performance

Often, it is impossible or prohibitively costly to measure each individuals contribution to the
teams output. How much does an individual professor add to the likelihood of you obtaining a
university degree? When you make a group assignment for a course, how much does each group
member contribute to the end result? What is the contribution of a single football player to his
clubs championship? As a consequence, the above optimal contract rewarding individual effort
may not be implementable. An alternative is to reward each team member on the basis of team
output. Indeed, it is not uncommon for firms to link their employees pay to the performance of
worker teams, divisions, or even the entire firm. Examples include profit sharing (employees are
given a bonus which depends on the profits generated by their team, division, or the firm),
employees owning stocks or stock options in their company, and group bonuses (employees get
a bonus if their team reaches a pre-specified target).
When team members are rewarded on the basis of team output, they become involved in a game
because each members effort has an impact on other team members payoffs. Let us consider
such a game using the above model with just two team members, Emma (team member 1) and
Francis Ford (team member 2). Emma and Francis Ford produce a movie together. Let be the
number of people who will watch the movie in a movie theater, each paying , so that the
movies box office revenue equals . Suppose Emma and Francis Ford agree to share the
revenue equally.
How much effort will the two exert? We start by looking at Emma. Her utility is given by
1 = (1 ) = (1 + 2 ) (1 )2 .
Because Francis Fords effort adds a constant to her utility function, Emma will ignore it in
finding her optimal effort. In a similar way as before, we find that Emma will choose effort
= /2.
By symmetry, the same holds true for Francis Ford. Note that Emmas and Francis Fords effort
equals half the first-best effort .
55

lOMoARcPSD|1733731
Things become even worse if the team consists of more than two players. In the -player case,
team member 1s utility equals
1 = / (1 ) = (1 + 2 + . . . + )/ (1 )2 .
When maximizing her utility, team member 1 can ignore the efforts by the other team members
because they just add a constant to her utility. It is now straightforward to show that the
equilibrium effort of team member 1 and of all other team members is equal to
= /.
Because effort is decreasing in , we can conclude that the larger the team the more the
incentives to exert effort are diluted.
Clearly, team incentives are not quite as strong as individual incentives. The intuition is that
team members can free ride on each others effort. Another way of saying this is that a team
members effort creates a positive externality on the other team members so that effort in teams
is typically undersupplied. How to solve the apparent free rider problem? One straightforward
way is to increase the bonus per unit of team output substantially. Of course, it can become very
costly to cover those bonuses. In the movie example with two team members, both members
should obtain a bonus equal to the box office revenue. I doubt whether a film distribution
company would be willing to offer such a contract. Repeated interaction may offer another
solution, as we will discuss in more detail in part IV. A team can enforce the first-best outcome if
team members can find a credible way to punish a shirking team member the next time the team
interacts. Emma may have a good reason to work hard if she knows that otherwise, Francis Ford
will not cooperate with her in the next movie.
5.3 Comparative performance evaluation

Even if an individuals contribution to the teams output is perfectly measurable, individual
incentive contracts can have at least two practical drawbacks. First, the optimal contract dictates
that each team member becomes the residual claimant of his contribution to the teams output.
It can become very costly for a firm if it pays its employees in such a way. In chapter 8, we will
see that the firm can mitigate this problem by paying employees a relatively low (or even
negative) base salary. A second difficulty emerges if individual output is partly determined by
unobservable random factors outside the control of a team member. A risk-averse team member
may only be willing to accept a contract that pays a high reward per unit of individual output if
the base wage is high enough to compensate for the risks.
Comparative performance evaluation may solve both problems simultaneously. We speak of

comparative performance evaluation if a workers payment is based on his performance relative
to the performance of other workers. Why might it work? Suppose that each workers output is
affected by the same random shock (for instance, the weather, a demand shock, or the sentiment
on the stock market). If workers are paid based on relative performance rather than on absolute
performance, the contract is less risky because the random shock has less impact on the relative
output than the absolute one. We will discuss this point in more detail in the current chapters
case study and in chapter 8.
A tournament is an example of comparative performance evaluation. Participants in a

tournament compete for one or several prizes that will be allocated to the best performer(s).
56

lOMoARcPSD|1733731
Tournaments are commonly used in practice. The UEFA Champions League, the Tour de France,
and Wimbledon are examples of tournaments in football, cycling, and tennis respectively.
Tournaments occur not only in sports, but also in firms where people get promoted to a higher
rank in the corporate hierarchy when they perform well on the job.
I will show now that a tournament can incentivize team members to expend the first-best level
of effort. Suppose that the team members in our model compete for a single prize with value .
The prize is awarded to the team member who produces the highest individual output. An
individuals output is determined both by his effort and a random factor that is outside his
control. We assume that the probability that individual produces the highest output given his
own and the others efforts equals

= .
1 + 2 + . . . +
Note that the players are engaged in a game in which their effort impacts both their own and
the other team members probability of winning. The resulting utility for team member 1 is
equal to the probability of winning the prize times the prizes value minus the costs of his effort:
1
1 = 1 (1 )2 = (1 )2 .
1 + 2 + . . . +
To get the best response of team member 1, we maximize his utility keeping the effort choices of
the others constant. The first-order condition for the maximization problem is given by
2 + . . . +
1 = 0.
(1 + 2 + + )2
Because the game is symmetric from the viewpoint of all team members, we can assume that all
will choose the same effort in the Nash equilibrium:
1 = 2 = . .. = = .
Substituting this in the first-order condition for team member 1, we obtain
( 1)
= 0
2 2
which implies that
1 ( 1)
= .

To obtain the efficient effort = , it must be the case that
1 ( 1)
= =

57

lOMoARcPSD|1733731
2 2
= .
( 1)
Observe that the bigger the team, the greater the value of the prize must be. According to this
model, team members work harder the bigger the prize they can win in the tournament.
The main conclusion from the analysis so far is that both absolute performance compensation
(based on piece rates for instance) and comparative performance compensation (based on
tournaments) can achieve an efficient outcome. Of course, the two mechanisms are quite
different in terms of informational requirements so that in practice, one may be preferable over
the other. For example, in settings where output is readily observed, paying by piece rate is
preferable over a tournament because this saves the need to make direct comparisons between
workers. As a consequence, it might not be surprising that salesmen are compensated using
piece rates. On the other hand, for some jobs, it is easier to obtain information about the ranking
of individual performance than about individual output. For such jobs, tournaments may
perform better than piece rates. For example, managers in large firms typically engage in
tournaments where the best performing manager is promoted to a higher rank in the corporate
hierarchy obtaining a substantial salary raise. The tournament structure of such compensation
scheme might explain why the salaries of CEOs are often substantially higher than what their
contribution to the firms productivity seems to justify.
Another way to implement incentives on the basis of comparative performance is to reward

team members on the basis of how much better (or how much worse) they perform compared to
the rest. Consider, for simplicity, a team consisting of two workers, Gwyneth and Harvey, who
are hired by a farmer to harvest oranges. On top of a base wage , both Gwyneth and Harvey
obtain a bonus per orange they harvest more than the other. Similarly, if they harvest fewer
oranges than the other, they pay a penalty per orange. Suppose the assumptions of our model
hold true so that Gwyneth utility is given by
= + ( ) ( ) = + ( ) ( )2
where and represent the number of oranges harvested by Gwyneth and Harvey
respectively. Note that Gwyneth can ignore Harveys effort when deriving the maximizing her
utility. You can verify that her equilibrium effort equals
= /.
The same holds true for Harvey. Gwyneth and Harvey obtain the efficient effort if = .
Note that this comparative performance incentive scheme has the additional attractive property
for the principal that it gives the workers incentives to work hard while the principals net
payment equals zero by construction: Harveys penalty covers Gwyneth bonus and the other
way around.
While comparative performance evaluation can have several advantages over contracts based
on absolute performance, they may also suffer from several practical disadvantages. First of all,
it provides team members with little incentive to help each other. Helping another is doubly
costly in that it reduces the own output and increases the others output at the same time. Even
worse, team members may be inclined to sabotage other team members to improve their own
relative performance. Second, team members have an incentive to conspire against the principal
58

lOMoARcPSD|1733731
not to work hard because such conspiracy need not have an impact on their relative
performance. In part IV, we will study how team members can sustain such collusion.
5.4 Case study: The downside of relative performance pay

Several studies use laboratory experiments to examine how productivity changes under
different incentive schemes. While the laboratory approach has its advantages, for instance in
terms of allowing precise control of the environment and accurate observation of participants
behavior, the external validity can often be called into question (the extent to which
conclusions can be generalized to apply to real-world situations). The alternative approach is
that of field studies, which use real-world data. This can directly address the external validity
problem, but obtaining adequate real-world data is often an issue. Since the data does not come
from a controlled environment, it is subject to bias from all kinds of external influences,
measurement errors and so on.
However, one example of real-world data being used to good effect comes from a study by
Oriana Bandiera, Iwan Barankay, and Imran Rasul. The authors use data from a farm where
workers have the task of picking fruit. The amount of fruit picked by each worker is accurately
measured on a daily basis. Workers were initially paid using a relative incentive scheme and
subsequently using a piece rate scheme. Since the data available covers a sufficient period of
time before and after the change, it is possible to compare productivity before and after.
As mentioned in section 5.3, a relative incentive scheme has the advantage of being unaffected
by common productivity shocks that affect all workers, meaning that the workers benefit from
being insured against the uncertainty that these shocks bring. Indeed, this was given by the
farms management as the reason why relative incentives were chosen initially. Common
productivity shocks, due to factors such as changing weather conditions, are quite prevalent in
this line of work. Later, however, the management decided to try moving to a piece-rate
incentive scheme, since productivity had been lower than expected.
For both schemes, daily compensation can be written as , where is the total kilograms of
fruit picked by worker i on that particular day. The difference is that, under the relative scheme,
is determined by the average productivity of all workers on the day. Specifically, = /,
where is average hourly productivity (in kilograms fruit picked) of all workers on the day and
is a constant set by management at the start of the season, slightly above minimum wage. On
the other hand, under piece rates, is set by management at the start of the each day and then
remains fixed. Thus, a workers productivity influences the wage rate of other workers under
relative incentives, whereas under piece rates it does not.
If, for simplicity, we assume that effort is equal to productivity, so Ki = ei and = , then under
piece rates a worker will maximize utility = ( ) (wage earned minus cost of effort).
Under the relative scheme, however, average effort influences the wage rate, and a workers
own chosen level of effort ei affects . Therefore, one worker increasing his effort with all else
equal imposes a negative externality on all other workers, because it has the effect of increasing
average effort and therefore reducing the wage rate. This suggests the possibility that workers
may care about the well-being of others and take this into account when choosing ei. Formally,

they would maximize =

( ) + (

( )) (wage earned, minus cost of
effort, plus weighted sum of other workers utility). Here, indicates to what extent the payoffs
of other workers influences an individual workers behavior. If = 0 then the payoffs of other
59

lOMoARcPSD|1733731
workers have no influence at all, whereas in the extreme case, = 1 and the payoffs of others
have as much weight as the individuals own payoff. (You might argue that < 0 is also a
possibility, if a worker enjoys the suffering of others!)
The study uses the data to test the hypothesis that, with relative incentives, workers do indeed
behave as if they care about others, in other words that they behave in accordance with a utility
function that has > 0. If this was the case, then the productivity observed under relative
incentives would be lower than under piece rates, because high individual effort penalizes
others under the relative scheme, but has no effect on others under piece rates. The authors take
steps to ensure that the periods before and after the incentive change are comparable, for
example in terms of the work environment, the amount of daily hours worked, the quality of the
fruit picked and so on. The comparison reveals that productivity was 50-70% higher under piece
rates, which is clear evidence of apparently altruistic behavior. The estimation of indicates a
value between 0 and 1, meaning that the negative externality was partly but not fully
internalized by the workers.
Having confirmed that workers respond to relative incentives with apparently altruistic
behavior, the authors consider two explanations. One is that, as alluded to earlier, workers are
altruistic and genuinely care about their colleagues wellbeing. Alternatively, the observed
behavior may be due to collusion: workers realize that, with relative incentives, it is to their
mutual benefit to conspire so that they all choose a relatively low level of effort. For example, if
all workers reduce effort by 10%, this lowers the negative utility accruing from the cost of effort
for all workers, without affecting the wages paid. However, this collective strategy also creates a
temptation for an individual worker to deviate from the agreement: if a worker maintains the
same effort level while all others reduce their effort by 10%, that worker benefits (his wage rises
while cost of effort is unchanged). (The line of reasoning here parallels the theory of collusion in
repeated oligopoly games, which will be discussed in chapter 12.)
To further investigate behavior, the authors use data that shows which workers describe each
other as friends. They exploit the fact that the group of fruit-pickers changed every day, to see
how individual behavior varies depending on the number of friends that are in the group. With
piece rates, the number of friends in the group has no effect, but with relative incentives, a
worker exerts less effort when more of their friends are present. At first sight, this supports the
idea that workers are genuinely altruistic towards their friends and take into account their
payoffs. However, the authors also have data on two different types of fruit-picking tasks: one
task has workers working alongside one another and therefore able to observe each others
behavior, while the other involves working in dense shrubs where it is impossible to see others.
The data shows that, when behavior is unobservable, productivity is the same regardless of the
incentive scheme. It is only when others behavior is visible that productivity under relative
incentives is much lower than under piece rates. This evidence strongly supports the collusion
explanation, because maintaining a collusive strategy requires being able to see the behavior of
others, to ensure that they do not deviate.
The study gives an important insight into the considerations that a firm should make, when
choosing an incentive scheme. Relative incentives have the benefit of reducing worker
uncertainty, but we have seen that they can be exploited by collusive behavior, to the workers
benefit but to the firms loss. This possibility requires that workers can observe one another, but
the study highlights other conditions that can facilitate this outcome. Friendship and social ties
60

lOMoARcPSD|1733731
between workers make collusion easier perhaps because bonds of trust are formed, but also
because interaction between workers is frequent and there are many opportunities to form
collusive agreements and to punish deviants who break the agreement. The study also finds
that, under relative incentives, effort is lower in smaller groups: collusion is always easier to
sustain when fewer parties are involved and more difficult among a large group of agents. Those
findings are very much in line with the theory of collusion that we will discuss in chapter 12.
The model developed in this chapter is based on Lazear and Rosens (1981) work on
tournaments. In their paper, Lazear and Rosen suggested the idea of tournament theory by
looking at performance pay in teams. In turn, the new field they developed within personnel
economics compacts situations in which wage differentials are based on relative differences
between individuals and not upon their marginal productivity. Rosen further extended this work
in 1986 with a paper he published in the American Economic Review. You can also refer to Bull,
Schotter and Weigelt (1987) for an experimental study of tournament theory. The case study of
this chapter is based on the results from Oriana Bandiera, Iwan Barankay, and Imran Rasuls
(2005) field experiment.
References
Bandiera, O., I. Barankay and I. Rasul (2005). Social preferences and the response to incentives:
evidence from personnel data, The Quarterly Journal of Economics, 120(3), 917-962.
Bull, C., A. Schotter and K. Weigelt (1987). Tournaments and piece rates: an experimental study,
Journal of Political Economy, 95(1), 1-33.
Lazear, E and S. Rosen (1981). Rank-order tournaments as optimum labor contracts, Journal of
Political Economy, 89(5), 841-864.
Rosen, S. (1986). Prizes and incentives in elimination tournaments, American Economic Review,
76(4), 701-715.
Exercises
Exercise 5.1 Individual Incentives

A few months ago, Russell and Sharon opened two different fruit stands to sell oranges at the
Albert Cuyp market in Amsterdam. At first sight they should compete against each other to sell
as many oranges as possible. However, Woody, a wholesale fruit seller who is trying to become
the monopolist of Amsterdams market of oranges, hired both of them. Woody wants to find the
right incentive scheme to optimize Russells and Sharons effort, taking into account that both
sellers incur an implicit cost in working at the open-air market they would rather put effort in
filming a movie. Russells and Sharons preferences can be summarized by the following utility
functions:
( ) = 10002 ,
( ) = 15002 ,
where , are the effort levels of Russell and Sharon, respectively. and are the bonuses
paid by Woody per effort level. For each unit of own effort, both Russell and Sharon sell 10
61

lOMoARcPSD|1733731
oranges, regardless of the others effort. Assume that Woody can observe the quantity sold by
both sellers at the end of a working day.
a) Find the optimal level of effort for Russell and Sharon.
b) How much should Russells bonus be with respect to Sharons, in order to let them exert the
same level of effort? What do you conclude?
c) Assume that both fruit stands are allowed to set different prices. Find the bonus per effort
level that Woody would pay to Russell and Sharon as a function of prices in an efficient contract.
d) Suppose Woody wants to set the same price in both fruit stands. How will he set the bonuses?
Does it make sense?
Exercise 5.2 Individual Incentives

Woody, the wholesale fruit seller we encountered in the first exercise, continues with his plan to
conquer Amsterdams market of oranges. His next target is Dappermarkt, where he sent two
promising orange sellers recently hired: Tom and Uma. Woody wants to exploit the two sellers
charm to capture all the demand for oranges of the neighboring Turkish district. Tom is able to
sell 25 oranges for each unit of effort he puts in his job, while Uma sells 20. The strategy is
similar to the one used in the Albert Cuyp market: two fruit stands, same price. However, Tom
and Uma have higher costs than the sellers present in Albert Cuyp. Their utility functions are
described as follows:
( ) = 20002
( ) = 20002 ,
where , are Toms and Umas level of efforts, respectively, and , are the bonuses per
effort level that Woody grants to the two sellers. Once again, Woody observes the quantity sold
at the end of the working day.
a) Compute Toms and Umas optimal effort levels and their bonuses assuming equal efforts. Is
there any difference with respect to the previous exercise?
b) Assume that the two fruit stands act as one and set the same price. Calculate the profit-
maximizing effort levels for both Tom and Uma.
c) What level of bonuses will Woody set as part of an optimal linear contract? Is there any
difference between Toms and Umas bonus per unit of effort? Does it make sense?
Exercise 5.3 Team Incentives

Things are not going that well for Woody. So far, his plan to dominate Amsterdams oranges
market has resulted in a failure. He blames his sellers in both the Albert Cuyp market and the
Dappermarkt, claiming that they do not put enough effort in their job. To change their behavior,
Woody establishes a different type of incentive scheme. Maybe, he thinks, if they work as a
team, and their earnings depend on each others effort, their total effort will be doubled! Woody
will not differentiate the bonus per effort level among the sellers anymore. Instead, he will now
pay the same bonus per unit of quantity sold to each seller in the open-air markets, and their
total earnings will depend on the total quantity sold divided by the number of sellers.
Specifically, if (with = , for the Albert Cuyp market and the Dappermarkt) is the
62

lOMoARcPSD|1733731

bonus for each open-air market, the total earnings of a seller will be , where is the total

quantity and is the number of sellers (for both = , ) on the Albert Cuyp market and
the Dappermarkt. Recall that Russell and Sharon are the sellers in the Albert Cuyp market, each
with a marginal productivity of 10 oranges per unit of effort level, whereas Tom and Uma work
in the Dappermarkt, selling 25 and 20 oranges per unit of effort level respectively. Each seller
incurs the same costs as in the previous exercises.
a) Write down the utility functions and the optimal level of effort for the sellers in the Albert
Cuyp market. Compare the results with the one in Exercise 5.1. Is there an improvement in the
level of effort, assuming equal bonuses between the two incentive schemes?
b) Do the same as in part (a) for the Dappermarkt sellers, and compare with the results found in
Exercise 5.2.
Woody is not satisfied by the performance of his sellers. He is convinced that the team incentive
scheme works better with more team members. Thus, he will now pay the same piece-rate
wage to each seller, without any difference between the two open-air markets, and their
earnings will depend on the sum of the quantity sold in both markets divided by the number of
all the sellers.
c) Compute Sharons and Umas optimal effort level with this incentive scheme.
d) Does the team incentive scheme perform better if more members are in the team? Why?
Exercise 5.4 Comparative performance evaluation

Having abandoned the team incentive schemes, Woody thinks he may have found the right way
to motivate his sellers: more competition. He will try two different comparative performance
schemes, one in each open-air market in which he operates. Unfortunately, Tom, one of the
sellers in the Albert Cuyp market, left the orange-seller career to seek his fortune as a Hollywood
actor. He was replaced by Val, who will join Uma in the open-air market. Both Val and Uma are
able to sell 20 oranges per unit of effort, and they face the same costs (20002 , with =
, for Val and Uma). In The Albert Cuyp market, Woody sets a prize with value for the
seller who first reaches a certain amount of oranges sold. The probability of winning the prize
depends on the ability and the effort of both sellers, and can be summarized as follows:

P {} = , = , ,
+
where is the quantity sold by one seller and denotes the quantity sold by the other seller.
a) Write down the utility of Val and Uma as a function of their effort level and the prize.
b) Compute the prize as a function of the optimal effort level. What do you conclude? (You can
use the symmetric condition.)
At the Dappermarkt, Woody tries a different approach, based on relative performance. He will
pay a marginal salary to a seller only for the quantity that exceeds the quantity sold by the other
seller. At the same time, if one of the fruit stand sells a lower amount of oranges than the other,
he/she will have to pay the difference. Thus, the income of Russell and Sharon can be written as:
= ( ) , = , ,
63

lOMoARcPSD|1733731
where is the bonus per unit in excess, and clearly if < the seller will have to pay the
difference to Woody.
c) Write down Russells and Sharons utility functions, and find the optimal level of effort. Are
they different from the ones that Woody obtained with the individual incentive scheme
(Exercise 5.1), assuming equal bonuses?
d) Assume that both the individual incentive scheme and this comparative performance scheme
are available, which one do you think Woody will choose?
e) What are the risks of this comparative performance scheme?
64

lOMoARcPSD|1733731
Chapter 6: Oligopoly
The topic of this chapter is competition between firms in an oligopoly. We call a market an
oligopoly if a handful of firms are active in this market. Oligopoly can be seen as a market
structure that lies between monopoly (a market with only one firm) and perfect competition (in
which case many small firms compete). Because an oligopolistic market only consists of a few
firms, each firm has to take into account that its strategic choices have an impact on the
decisions by other firms and the other way around. This strategic interconnection may be
relevant for all strategic decisions including product, price, place, and promotion (the Marketing
mix or the 4Ps of Marketing).
The set-up of the remainder of this chapter is as follows. In section 6.1, I will introduce you to a
stylized model of price competition, the Bertrand model. We will observe that this model
produces the surprising result that the equilibrium price equals marginal costs even in the case
of as few as two firms, a result that has become known as the Bertrand paradox. In sections 6.2
and 6.3, we will consider several models of oligopolistic competition that resolve the Bertrand
paradox in the sense that the equilibrium prices exceed marginal costs. Those models include
constraints on production capacity, competition on quantity (rather than on price), cost
asymmetries, and consumer search and switching costs. Section 6.4 includes a case study on
airline markets, where firms not only compete in price but also in terms of quality.
6.1 Bertrand competition

Suppose that in the direct neighborhood of your university campus, you can get coffee at two
coffee shops: Coffee Bucks and Star Company. Suppose that those two coffee shops are the only
ones in the market. Other places are either too far away from the campus or they sell hot drinks
that do not deserve to be called coffee. The total demand () for coffee on campus (in cups a
day) equals
() = 960 240
where represents the price for coffee. Coffee Bucks and Star Company are located right next to
each other and both sell the very same quality of coffee. In other words, consumers consider the
coffee on campus as a homogeneous good. As a consequence, all will buy the coffee at the less
expensive place. For both firms, the marginal costs per cup of coffee are constant and equal to
= 1.
All fixed costs are sunk and can be ignored.
What price will the two firms charge for a cup of coffee? Before we can answer this question, we
should be somewhat more specific about the rules of the game Coffee Bucks and Star Company
are involved in. At the start of the day, both firms will independently choose their price for a cup
of coffee. If the price of one firm is lower than the price of the other firm, all market demand will
be served by the less expensive firm at his own price. If the two firms choose the same price,
both will serve half the demand at that price. Each firms profit is its price-cost margin times the
number of cups it sells. This game is called the Bertrand game after the French mathematician
Joseph Louis Franois Bertrand who published this model of price competition as early as 1883.
The Bertrand game has a unique Nash equilibrium in which both firms set their price equal to
marginal costs:
65

lOMoARcPSD|1733731
= = 1.
To verify that marginal cost pricing is the only equilibrium, let us consider the firms best
responses against the prices chosen by the other firm. If Coffee Bucks charges a price above the
monopoly price (which is equal to 2.50), Star Companys best response is the monopoly price. At
this price, Star Company will attract all demand (because its coffee is less expensive than Coffee
Bucks) at the same price a profit-maximizing monopolist would choose. If Coffee Bucks price is
at or below the monopoly price and above marginal costs, Star Companys best response is a
price just below Coffee Bucks (let us say, one cent lower). It does not make sense for Star
Company to choose a higher price than Coffee Bucks because it will not sell any coffee. If Star
Company picks the same price as Coffee Bucks, Star Company will cater to half the consumers.
However, it could almost double its profits by decreasing the price slightly (let us say by 1 cent)
because it would then serve the entire market. A further decrease in the price is not profitable
because the increase in demand is not sufficient to compensate for the lower price-cost margin.
If Coffee Bucks price is equal to marginal costs, Star Company is indifferent between the
marginal cost price and any other price because its profits will be zero. Finally, if Coffee Bucks
chooses a price below marginal costs, Star Company should take care not to serve any
consumers because it will only lose money on them. Any price above Coffee Bucks is a best
response.
Figure 6.1: Bertrand competition
Figure 6.1 depicts the best responses of both companies. Because the two firms are completely
symmetric, you will find Coffee Bucks best responses in exactly the same way as we found Star
Companys. Note that the figure only presents the marginal costs price as a best response against
prices at or below marginal costs while from the discussion above we know there are many
more. A firms reaction curve connects the firms best responses against all potential prices by its
opponent. Recall that prices constitutes a Nash equilibrium if both firms play a best response
against the other firms price. Therefore, the Nash equilibrium is the point where the two
reaction curves cross. From the figure, it becomes clear that marginal costs pricing is a Nash
66

lOMoARcPSD|1733731
equilibrium. You can check that any other pair of prices is not a Nash equilibrium, so that
marginal costs pricing is the unique Nash equilibrium.
The outcome of the Bertrand game may come as a surprise. While the market is quite
concentrated in the sense that only two firms serve the entire market, the outcome is the same
as in a perfectly competitive market: The price in the Nash equilibrium equals marginal costs.
Because the result is surprising, at least at first sight, it is often dubbed the Bertrand paradox.
The Bertrand paradox emerges not just in the above stylized example of competition between
Star Company and Coffee Bucks, but also in a large range of settings in which the following
assumptions hold true:
Two or more firms are active in the market

Each firm can serve the entire demand at any price
The firms strategic choice variable is the price
The firms interact only once
All firms have the same marginal costs
All consumers buy at the firm(s) with the lowest price
Clearly, if the first assumption is not satisfied, that is, in the case of a monopoly, the Bertrand
paradox is resolved because the monopoly firm will charge a price above marginal costs. As we
will discuss in the remainder of this chapter, we can resolve the Bertrand paradox by relaxing
any of the other assumptions.
6.2 Cournot competition

The second assumption underlying the Bertrand game is that each firm can serve the entire
demand at any price. This assumption may be unrealistic in many practical situations, at least in
the short run. In practice, most companies face a capacity constraint in that they can serve
demand only up to some point. A firm selling bottled water can only extract up to a maximum
amount of spring water from its source. Bank employees can only process a limited number of
applications for a mortgage within a reasonable time. In the above example, Coffee Bucks and
Star Company are restricted by the capacity of their coffee machines and the speed of their
personnel in terms of how many cups they can brew a day.
When we introduce capacity constraints into the Bertrand model, the Bertrand paradox may no
longer hold true. To see why, let us return to the coffee example. Recall that demand () for
coffee equals
() = 960 240.
Suppose that Coffee Bucks can produce at most 240 cups of coffee a day while Star Company has
a limited capacity of 180 cups a day. At the marginal cost price of 1, Coffee Bucks will sell at most
240 cups of coffee, even if demand at this price is 720. Star Company can now act as a
monopolist on the residual demand, that is, the demand not being served by Coffee Bucks:
() = 960 240 240 = 720 240.
Star Company can now sell 180 cups of coffee at a price of 2.25, which clearly results in higher
profits than if it charged a price equal to marginal costs. In other words, marginal cost pricing is
not a best response so it is not a Nash equilibrium.
67

lOMoARcPSD|1733731
In this case, the Nash equilibrium price is equal to the market clearing price when both firms
produce as much as their capacity constraint allows them to. So, if Coffee Bucks and Star
Company sell 240 and 180 cups respectively, the market clearing price follows by equating
demand to the two firms total production capacity of 420:
() = 960 240 = 420.
The resulting market clearing price equals = 2.25. Both firms choosing this price constitutes
a Nash equilibrium if neither has a reason to deviate to another price. Suppose Star Company
chooses a price = 2.25. The more Coffee Bucks decreases its price from = 2.25, the lower
its price-cost margin will be while the demand it serves stays constant at 240 cups, since it
cannot sell more than that because of its capacity constraint. So, a price below = 2.25 is not a
best response for Coffee Bucks. A higher price can be profitable, but only if the costs saved on
producing fewer units exceed the revenue lost, which is the case if marginal costs exceed
marginal revenue. Given Star Companys supply of 180, Coffee Bucks residual demand equals
() = 960 240 180 = 780 240.
The corresponding inverse demand is given by
780
() = .
240
Therefore, Coffee Bucks revenue equals
780 2
= () =
240
so that marginal revenue is equal to
780
= .
240 120
If Coffee Bucks produces at capacity, = 240, resulting in
780 240
= = 1.25 > 1 = .
240 120
So, it is not beneficial for Coffee Bucks to increase the price and produce less than = 240
because at that point its marginal revenue exceeds its marginal costs. You can check that a
similar reasoning applies for Star Company.
More generally, if firms face serious capacity constraints, they tend to choose prices above
marginal costs. This finding shows that the Bertrand paradox is not a universal law. Note,
however, that we assumed that the firms production capacities are fixed. While this may be the
case in the short run, in the longer run, firms can expand their production capacity. The water
company can acquire stronger pumps or additional water sources. The bank can hire more
employees to increase the number of mortgage applications it can handle. Coffee Bucks and Star
Company can replace the current coffee machines by quicker ones or place extra personnel
behind the counter. (The two firms may have a reason to do so because at their current
capacities the marginal revenue exceeds the marginal costs, as we saw above.) Indeed, you may
68

lOMoARcPSD|1733731
argue that in the long run, the firms key strategic decision is not the price but the production
capacity, which could be seen as a proxy for the quantity they will sell in the market.
As early as 1838, the French philosopher and mathematician Antoine Augustin Cournot modeled
strategic interaction between oligopolistic firms where the decision variable is not price but
quantity. Cournot assumed that firms in a market make independent decisions as to how much
quantity to produce and sell to the market. All firms sell their output in the market at the market
clearing price, which depends only on the total quantity the firms offer for sale. Nowadays, we
use the term Cournot competition for this type of strategic interaction on markets.
Let us reconsider the example with the two coffee bars. Suppose that both offer coffee to the
market at constant marginal costs
= 1.75.
So, both firms add 0.75 to their marginal costs of 1 per cup of coffee, which could be the cost of
increasing production capacity because of additional personnel working hours or more
productive coffee machines. Again, we assume fixed costs to be sunk. Both firms decide
independently how many cups of coffee to sell. If Coffee Bucks produces cups and Star
Company produces cups, total quantity offered to the market will be equal to
= + .
The resulting market clearing price follows from
= () = 960 240,
which implies
= (960 )/240.
To establish the Nash equilibrium of this game, we need to find the point where both firms play a
best response against the quantity chosen by the other firm. To derive the best response of
Coffee Bucks against the quantity choice by Star Company, we need to maximize Coffee Bucks
profits as a function of its own quantity assuming the quantity of Star Company to be fixed.
Coffee Bucks revenue equals
= = (960 ) /240 = (960 ) /240.
By equating marginal revenue to marginal costs, we get:
= (960 2 )/240 = = 1.75.
Because the two firms are entirely symmetric, it makes sense to assume that both will produce
the same quantity in equilibrium:
= = .
Substituting this in the above = equation, we obtain
(960 2 )/240 = 1.75.
69

lOMoARcPSD|1733731
Therefore, both firms will produce
= 180
cups in the Nash equilibrium.
We can also find the Cournot equilibrium by crossing the two firms reaction curves as shown in
figure 6.2. Like in the Bertrand case, a firms reaction curve represents the firms best responses
against all potential choices by its competitor. We can find Coffee Bucks reaction curve by
writing its optimal quantity as a function of Star Companys. We can do this by rewriting Coffee
Bucks = equation
= (960 2 )/240 = = 1.75
in the following way:
= (540 )/2.
Similarly, Star Companys best response curve is described by the equation:
= (540 )/2.
The point where the two curves cross is the Nash equilibrium. Not surprisingly, this is the point
where both firms produce = 180 cups.
Figure 6.2: The Cournot equilibrium
The resulting Nash equilibrium price equals
= (960 2 )/240 = 2.5.
Observe that the equilibrium price is greater than the marginal cost price = = 1.75. As a
consequence, in Cournot competition, firms generate higher producer surplus, lower consumer
70

lOMoARcPSD|1733731
surplus and lower welfare than an otherwise comparable market with perfect competition. In
contrast, the price under Cournot competition is lower than the monopoly price (which equals
= 2.875). So, in Cournot competition, firms generate lower producer surplus, higher
consumer surplus and higher welfare than an otherwise comparable monopoly market.
What happens if more than two firms compete la Cournot? Would the price converge to the
marginal costs if an increasing number of firms are active in the market? Let me show that this is
indeed the case. Suppose there are firms in the market characterized by inverse demand
() = .
All firms produce at constant marginal costs = , with 0 < . All fixed costs are sunk.
Suppose that firm 1 produces output 1 and that the other firms produce quantity 1 together
resulting in a total output = 1 + 1 . In equilibrium, firm 1 plays a best response against
the quantity chosen by the other firms. Firms 1s revenue equals
1 = ()1 = [ ]1 = [ 1 1 ]1.
The first-order condition of firm 1s profit maximization problem follows by equating its
marginal revenue to its marginal costs:
1 = 21 1 = = .
By symmetry, we may assume that all firms produce the same quantity in equilibrium:
1 = 2 = = =
which implies that
1 = ( 1) .
Plugging this back into the first-order condition, we can find the equilibrium quantity:

= .
( + 1)
The corresponding equilibrium price equals
+
= = = =+ .
+ 1 + 1
Indeed, if the number of firms grows large ( ) the price converges to marginal costs =
.
In a more general model of Cournot competition, one that allows for firms having different cost
structures and for the demand curve being non-linear, it can be shown that in the Nash
equilibrium,

= ,

where denotes the markets Lerner index, represents the Herfindahl-Hirschman index, and
stands for the price elasticity of demand. One important conclusion from this formula is that in
71

lOMoARcPSD|1733731
the Cournot model, the Lerner index is always greater than zero except for the extreme case of
perfect competition where = 0 or . In other words, in a very large range of settings,
we can resolve the Bertrand paradox by assuming that firms compete in quantities or
production capacities rather than on prices.
6.3 Other ways to resolve the Bertrand paradox

We may relax each of the remaining assumptions in the above Bertrand model to resolve the
Bertrand paradox. Let us start with the assumption that firms interact only once. Of course, this
assumption is unrealistic for many markets. Typically, firms can change prices from one day to
the next, not only responding to changes in their marginal costs, but also to pricing decisions by
their competitors in the past. In chapter 12, we will argue that if firms interact repeatedly, they
can sustain prices above marginal costs in equilibrium. The intuition is that firms have no
incentive to deviate to a lower price in some period because if they do so, other firms will react
in the next period by reducing the price to an even lower level.
Another critical assumption in the Bertrand model is that all firms produce under the same
marginal costs. Suppose that in the coffee bars example, Coffee Bucks incurs marginal costs
= 1
for each cup of coffee while Star Company is less efficient, producing cups of coffee at marginal
costs
= 1.23.
Then the Bertrand game no longer has an equilibrium in which both firms choose a price equal
to their marginal costs. If Star Company does so, Coffee Bucks is better off by choosing price
= 1.22 than the marginal cost price = 1. Therefore, marginal cost pricing is not a best
response for Coffee Bucks if Star Company picks a price equal to its own marginal costs. You can
check that there is a Nash equilibrium in which = 1.22 and = 1.23 if the smallest
currency unit is one cent. In this equilibrium, Coffee Bucks makes a handsome profit, so that the
Bertrand paradox does not arise.
A final assumption underlying the Bertrand model is that all consumers buy from the firm(s)
with the lowest price. There are at least three situations in which this assumption is violated.
First of all, if firms sell differentiated goods instead of a homogeneous good, a consumer may
prefer not to buy the least expensive good. For instance, consumers may be willing to pay more
for Coffee Bucks coffee than for Star Company coffee if Coffee Bucks coffee tastes better than
that of Star Company. Moreover, even if the quality of coffee is the same at both companies and
Coffee Bucks is more expensive than Star Company, consumers may prefer to go to Coffee Bucks
simply because Coffee Bucks is around the corner. In chapter 9, we will study such situations in
more detail to conclude that product differentiation typically results in equilibrium prices above
marginal costs.
Another reason why not all consumers go to the least expensive shop is that they simply are not
aware of the existence of this shop. In most cases in practice, consumers incur at least some
search costs to find out about the existence of firms supplying the goods they need. If it is costly
to search, some consumers may decide to buy at the first firm they encounter because they
expect that the additional costs of searching for another firm outweigh the expected price
72

lOMoARcPSD|1733731
decrease they might obtain by finding another firm. Typically in such an environment, different
firms offer the same good for a different price. Why can such price dispersion emerge in a Nash
equilibrium? The reason is that all firms make sufficiently high profits for it to be attractive to
deviate. The high-price firms obtain a high price-cost margin (although they mainly serve
consumers that do not search because those are likely to find a low-price firm). The low-price
firms will serve high demand because they serve both consumers that search and consumers
that do not (albeit at a low price-cost margin). Deviating to another price is not profitable for
either type of firm.
Switching costs is a third reason why consumers need not buy from the least expensive firm. A
consumer incurs switching costs if it is costly for him to change supplier. Switching costs come in
many shapes and sizes. In some cases, switching costs are monetary, for example, in the case
where your new supplier charges you administration costs for initiating a contract or in the case
where you have to pay your old supplier a fine for breaking up an existing contract. Switching
costs may also refer to the hassle related to switching. For example, consider the first-mover
advantage that Microsoft enjoys with Windows OS. Although almost everybody relishes the
design and performance of an Apple computer, not everyone is willing to invest the time
necessary to learn, or risk a potential disruption of having to adjust to a new operating system.
Switching costs may even be purely psychological in the sense that you may feel attached to
some firm and you experience costs when switching to another even if, objectively, the two
firms sell the very same product.
If switching costs are sufficiently high, the Bertrand paradox does not hold true. Even worse:
firms may even be able to sustain the monopoly price in equilibrium. To fix the idea, suppose
two companies, Vodamobile and T-Fone, are active in a market for mobile telecommunications.
Each consumer attaches value to mobile services. Marginal costs of either firm are constant at
= < per consumer. At the moment, both firms cater to half the consumers. When
consumers switch provider, they incur switching costs > 0. In other words, consumers will
only switch if the price difference between the two firms is at least . Now, suppose that
Vodamobile and T-Fone simultaneously choose a subscription price. In contrast to the original
Bertrand game, marginal cost pricing is not a Nash equilibrium. The reason is that if Vodamobile
picks price = , it is not in T-Fones best interest to choose price = . For instance, T-
Fone can increase its profits by charging a price equal to = + . In that case, it will still
make a profit margin of on all its current consumers because for them it is (just) not attractive
enough to switch to Vodamobile.
Both firms choosing the monopoly price is a Nash equilibrium if switching costs are sufficiently
high. Note that the monopoly price equals = : A monopoly firm can maximize profits by
extracting the entire willingness-to-pay of each consumer. = = is a Nash equilibrium if
it is not attractive for either firm to deviate. The most attractive deviation for Vodamobile is to a
price just below = because then the price difference with T-Fone is sufficiently high for
T-Fones current customers to switch to Vodamobile. Therefore, Vodamobile will double the
number of customers. The resulting increase in demand is not sufficient to compensate for the
lower price-cost margin if
2( ) < ( )
73

lOMoARcPSD|1733731
> ( )/2.
What we can learn from this exercise is that firms can obtain market power if consumers face
switching costs. Indeed, firms may have reasons to deliberately increase consumers switching
costs. For example, a mobile telecommunications company may not allow its customers to keep
their current mobile number if they change provider, it may write long-term service contracts
with its customers that are costly for customers to break, or it may offer discounts to loyal
customers creating opportunity costs of switching.
6.4 Case study: Why do airplanes arrive late?

As you have witnessed, economic theory analyzes various types of markets and competitions.
The three basic models monopoly, oligopoly, and perfect competition differ from each other
by the number of firms operating in the market and by their behavior, which consequently
determines market outcomes. So far, we have mainly focused on two dimensions of market
outcome, price and quantity, and we have studied how those are affected by market
concentration. In reality, though, consumers care about more than price and quantity alone. For
example, a monopolist can increase its profits not only by setting a price above the competitive
level, but also by reducing the quality and hence the associated costs of its products.
Figure 6.3. Frequency of minutes early/late in the dataset
Michael Mazzeo, Management & Strategy professor at Northwestern University, tries to draw the
connection between market concentration and quality. In his empirical study, he analyzes the US
airline industry in January, April, and July of 2000. His dataset contains over 800,000 flights. As a
measure of quality of the service, he uses the on-time performance of all the US airline
companies that is, the delay of the flights arrival time. Figure 6.3 shows the frequency of
observations in 15-minute intervals. Mazzeos idea is simple: if we assume that the on-time
performance (among other characteristics) enters into the consumers utility function, then the
expected utility of a flight would be conditional to the previous performances of a company.
Based on this, the consumer will choose which airline company she wants to fly with. Said
another way, if consumers care about being on time and they know that one companys flights
are always late, they will look for another company that offers the same route and the always-
74

lOMoARcPSD|1733731
late company will see its profits diminish. However, if the airline is the only carrier serving a
particular route, the future loss in revenues of delayed flights will be less severe because the
consumers cannot choose any other company. Therefore, a profit-maximizing airline invests
more in delay prevention the more competitive is the market.
To control for factors other than market concentration that may have an impact on airlines on-
time performance, Mazzeo included several variables in his regression. For example, he knows
whether on a certain day there was rain, snow or fog, or if there was a thunderstorm above the
destination airport. Additionally, he controls for the age of the airplane, whether it was built by
Airbus or Boeing, the number of flights scheduled each day, etc. To measure market
concentration, the author uses the Herfindahl-Hirschman index and a variable that describes
whether there is only one carrier which serves a certain non-stop route or not.
Mazzeo finds market concentration to have a statistically significant positive effect on the flights
delay. That is, a more concentrated market decreases the quality of the flight in terms of on-time
performance. On average, a flight on a monopoly route arrives 1.35 minutes later than a similar
flight on a more competitive route, ceteris paribus. Despite the statistical significance, the
magnitude of these effects is relatively small. It is hard to believe that a delay of less than two
minutes can make much difference for the consumers. In addition, the on-time performance is
probably not the only relevant characteristic of a flights quality. Consumers may also care about
the comfort of the chairs, noise on board, the quality of the food, and so forth.
We have mentioned before that the very first contributions to the study of imperfectly
competitive markets and the development of industrial organization as a branch of economics
came as early as the 19th century. French mathematicians Antoine Augustin Cournot (1838) and
Joseph Louis Franois Bertrand (1883) developed the first models of imperfect competition that
we studied in this chapter. Many academics consider them as the precursors of game theory and
industrial organization due to their pioneering study approach of duopoly markets and
oligopolies. Although their original work was published in French, you can refer to the more
recent reprints in English included in the reference list below.
The following authors studied the solutions to the Bertrand paradox that we discussed in this
chapter. Kreps and Scheinkman (1983) show that the Bertrand paradox disappears if capacity
constraints are introduced. Farrell and Shapiro (1988) examine the effects of switching costs in
markets characterized by dynamic competition. More insights in the modeling of consumer
switching costs can be found in the work of Klemperer (1987) and Beggs and Klemperer (1992).
Janssen and Moraga-Gonzlez (2004) examine oligopolies with search costs.
The case study of this chapter is based on the findings of Michael J. Mazzeo, which were
published in 2003 in the Review of Industrial Organization.
References
Beggs, A. and P. Klemperer (1992). Multi-period competition with switching costs, Econometrica,
60(3), 651-666.
Bertrand, J. (1883/1988). Review, Journal des Savants, 68, 499-508, in A.F. Daughety, Cournot
Oligopoly Characterization and Applications, Cambridge: Cambridge University Press.
75

lOMoARcPSD|1733731
Cournot, A.A. (1838/1927). Researches into the Mathematical Principles of the Theory of Wealth,
translated by N.T. Bacon, New York: Macmillan.
Farrell, J. and C. Shapiro (1988). Dynamic competition with switching costs, The Rand Journal of
Economics, 19(1), 123-137.
Janssen, M.C.W. and J.L. Moraga-Gonzlez (2004). Strategic pricing, consumer search and the
number of firms, The Review of Economic Studies, 71(4), 1089-1118.
Klemperer, P. (1987). Markets with consumer switching costs, The Quarterly Journal of
Economics, 102(2), 375-394.
Kreps, D.M. and J.A. Scheinkman (1983). Quantity precommitment and Bertrand competition
yield Cournot outcomes, Bell Journal of Economics, 14(2), 326-337.
Mazzeo, M.J. (2003). Competition and service quality in the U.S. airline industry, Review of
Industrial Organization, 22(4), 275-296.
Exercises
Exercise 6.1 Cournot competition

Air Canada and KLM compete for customers on flights between Amsterdam and Toronto. The
total number of passengers () flown by these two firms is the sum of passengers who fly KLM,
, and those who fly Air Canada, . Assume that no other companies can enter the route
because they cannot obtain landing rights at both airports. Market demand is given by:
() = = 339
where is the cost of a one-way flight (in Euros), and is measured in thousands of passengers
flying one-way per quarter. Each airline has a constant marginal and average cost of 147 per
passenger per flight.
a) Determine the Cournot equilibrium.
b) Assume both airlines receive a subsidy of 54 per passenger. Derive the new Cournot
equilibrium. What is the effect of the subsidy? How much of the subsidy is passed on to the
passengers?
c) Now return back to the situation in (a). How would the Cournot equilibrium change if KLMs
marginal cost falls to 100 and Air Canadas increases to 200?
d) Check whether the following condition holds in the equilibrium you derived under c):

=

KLM and Air Canada consider introducing a Frequent Flyer Program for their passengers. A
marketing study indicated that their total market has a constant elasticity of demand and that if
both firms introduce FFP programs, that elasticity would change from 2 to 1.75. The FFP will
increase each firms marginal cost per passenger from 147 to 157 per trip because of
additional service and extra meals for members. Each customer can only join one FFP.
e) Assume both airlines are engaged in Cournot competition. What will be the change in the
price of a one-way flight? Do the airlines benefit from the introduction of an FFP?
76

lOMoARcPSD|1733731
Exercise 6.2 Cournot with n firms

Consider a market for a homogeneous product with demand given by the following function:
1
= 37.5
4
All firms have a constant marginal cost equal to 40.
a) Derive the Cournot equilibrium as a function of the number of firms, .
b) Using your results from (a), determine output and price under a Cournot equilibrium with
= 2 firms.
c) What is the efficiency loss associated with this equilibrium? Compute the efficiency loss as a
percentage of the efficiency loss under monopoly.
Exercise 6.3 Bertrand competition

The computer graphics chip industry is one with a small number of competitors that earn
normal economic profit. Two chip manufacturers, NVIDIA and ATI both face the prospect of low
profits, largely on account of each others existence. Consider the homogeneous-good Bertrand
model in which each firm has a positive fixed and sunk cost and a zero marginal cost.
a) What are the Bertrand equilibrium prices? What are the Bertrand equilibrium profits?
b) What happens to the homogeneous-good Bertrand equilibrium price if the number of firms
increases? Why?
Now, consider a solution to that Bertrand paradox. Assume that NVIDIA and ATI are still
Bertrand competitors that set prices simultaneously. The demand for chips is given by:
= 100
where is the price in Euros. Both firms face a capacity constraint of 1 = 2 = 20 (measured
in millions) units per period, because of restrictions in labour and machinery usage.
c) Is firms charging a price equal to marginal cost a Nash equilibrium of this two-stage game?
d) Show that a Nash equilibrium exists where both firms charge = 1 = 2 = 60.
Exercise 6.4 Switching costs

In the market for Internet connections in Amsterdam, KPN and UPC compete la Bertrand.
500,000 clients are in the market, each willing to pay at most 120 per year to be connected to
the Internet. Both firms have constant marginal costs of 40 per connection per year. At the
moment, clients are equally distributed over the firms. If consumers switch, they incur switching
costs.
a) Suppose that switching costs are zero. What is the Nash equilibrium?
b) What is the Nash equilibrium if switching costs are equal to 50?
c) If switching costs are equal to 30, does the game have a Nash equilibrium in which both
firms charge the same price?
77

lOMoARcPSD|1733731
Part III: Dynamic interaction

In the previous part, we discussed situations in which decision makers (prisoners, team
members, firms, and so forth) make decisions simultaneously. In many applications, however, it
is not reasonable to assume that all decision makers make the relevant choices at the same time.
A retailer decides what prices to charge for its products after observing the wholesalers prices.
An employees decision to work hard or not may depend on the contract his employer has
offered him. A firms decision to compete fiercely or not may depend on how many other firms
have decided to enter in the same market. In this part, we will consider settings where decision
makers move sequentially. We start in chapter 7 by introducing an equilibrium concept for
dynamic games and applying it to several specific games. In chapter 8, we continue our
discussion of the Principal-Agent problem by modeling it as a three-stage game. In chapter 9, we
will study market entry and product positioning.
78

lOMoARcPSD|1733731
Chapter 7: Dynamic games

In this chapter, I introduce you to dynamic games. In section 7.1, I will start with a simple
Principal-Agent model to give an example of a dynamic game. Moreover, I will show you how to
solve dynamic games using the notion of subgame perfect Nash equilibrium. In section 7.2, we
will consider pricing in the vertical chain of production as another example of a dynamic game.
In section 7.3 I will present the results of an experiment that was conducted to test to what
extent peoples behavior is in line with the predictions of the subgame perfect Nash equilibrium.
7.1 Subgame perfect Nash equilibrium

Suppose that a principal (P) and an agent (A) interact in the following way. The agent first
decides whether to work hard or to shirk. The principal observes the agents decision and
chooses whether or not to give the agent a bonus if he works hard. If the agent shirks, both the
principal and the agent receive a zero payoff. If the agent works hard, the principals payoff
equals 3, but the agent will obtain a negative payoff equal to -1. By giving a bonus of 2, the
principal can ensure that each obtains a payoff of 1. This situation is a typical example of a
dynamic game, i.e., a game in which players move sequentially. Figure 7.1 summarizes the
interaction between the principal and the agent.
Figure 7.1: A simple Principal-Agent game
The German economist Reinhard Selten proposed the subgame perfect Nash equilibrium as a
solution concept for dynamic games. For this contribution to game theory, Selten shared the
1994 Nobel Memorial Prize in Economic Sciences with John Nash and John Harsanyi. Technically,
players strategies form a subgame perfect Nash equilibrium if the players play a Nash
equilibrium in each subgame of the entire game. I will not go into the details of this definition.
The main thing to remember is that you can find the subgame perfect Nash equilibrium by using
a straightforward and intuitive method called backward induction. The idea is to start by
solving the last stage of the game, then the penultimate stage, and so forth, up to the first stage
of the game. The solution for each stage is the Nash equilibrium in that stage, where players take
into account the actions that will be chosen in later stages of the game.
Let me return to the Principal-Agent game above to illustrate the backward induction method.
We start by solving the final stage of the game, which is the stage where the principal moves. At
this stage, the principal can choose between awarding the bonus or not. Clearly, he is better off
by not paying the bonus because this will pay him 3 rather than 1. So, the principal will choose
No bonus. Now, we move up to the first stage of the game. In this stage, the agent has to decide
between working and shirking. When making his decision, he has to take into account the
principals response in the second stage. Clearly, it is in the agents best interest to shirk because
79

lOMoARcPSD|1733731
if he works, the principal will respond by not awarding a bonus so that the agent is worse off. So,
in the subgame perfect Nash equilibrium, the agent chooses Shirk and the principal No bonus.
(Even though the principal does not actually make any decision if the agent chooses Shirk, No
bonus is nevertheless part of the Nash equilibrium, since it is this potential decision that
motivates the agent to choose Shirk.)
Figure 7.2 presents an example of a dynamic game in a market context. Suppose that a potential
entrant (E) decides whether or not to enter into a market. Currently, only one firm, the
incumbent (I), is active in this market. If the entrant decides to stay out, the incumbent enjoys a
nice monopoly profit of 80. The incumbents profit will decrease substantially in the case of
entry. Therefore, the incumbent would like to prevent entry at all costs. In fact, it has a strategy
at its proposal that might scare the entrant away from the market: If the incumbent initiates a
price war as soon as the entrant enters, the entrant will make a loss of 20. The question is
whether the incumbent can credibly announce that it will start a price war in the event of entry.
It may also employ a soft strategy, in which the entrant enjoys a profit of 20. The soft strategy is
also attractive for the incumbent: instead of making a loss of 20 in the case of a price war, it will
still make a nice duopoly price equal to 20. Indeed, when solving the resulting game using
backward induction, we observe that the incumbent will choose a soft strategy. Knowing this,
the entrant will decide to enter the market. So, in the subgame perfect Nash equilibrium, the
incumbent will accommodate the entrant despite the fact that the incumbent has a strategy at
its disposal (price war) that may deter entry. In this game, the threat of a price war is not
credible.
Figure 7.2: The accommodation game
Note that the outcome of the two games is a bit unfortunate from the viewpoint of the player
that moves second. In the Principal-Agent game, the principal (and the agent) would be better
off if the agent worked and the principal paid the bonus. Indeed, the principal has an incentive to
commit to paying the bonus, for instance by writing this down in the agents contract. If the
principal can do so, he can increase his payoff from 0 to 1 ensuring him a commitment value of 1.
Similarly, in the accommodation game, the incumbent would increase its profits if it could
credibly commit to starting a price war in the case of entry. The value of such commitment
equals 60 because the incumbent can increase its profits from 20 (its profits in the subgame
perfect Nash equilibrium without commitment) to 80 (its profits in the subgame perfect Nash
equilibrium with commitment). We will come back to the value of commitment in part VI.
80

lOMoARcPSD|1733731
7.2 Pricing in the vertical chain

What is worse than a monopoly? A chain of monopolies. The production of most goods in
modern economies takes place in what is called a vertical chain: a series of firms, starting with
the firms exploiting raw materials to firms selling the final products to end consumers. For
example, in the cheese market, farmers sell milk to cheese factories that sell cheese to
supermarkets that, in turn, cater to end consumers. The quote suggests that the welfare
consequences of having several monopolies in the chain can be disastrous. The intuition is that
each firm in the chain adds a profit margin to the original marginal costs. In contrast, if one firm
controls the entire vertical chain, it only adds a single profit margin to its marginal costs, leading
to a lower price and higher welfare. This phenomenon is known as the double marginalization
problem. The theory of dynamic games provides a useful tool to systematically study
interaction in the vertical chain so that we can verify why the double marginalization problem
arises (and why it is a problem).
Let us consider the hypothetical vertical chain of smartphones shown in figure 7.3. Suppose that
Strawberry is the sole supplier of smartphones. Strawberry does not sell directly to end
consumers, but leaves this to retailer Red Apple Store, a monopolist in the market for
smartphones. Visualizing the chain of production as a river running down from supplier to
consumers, we call Strawberry the upstream firm and Red Apple Store the downstream firm.
Figure 7.3: The vertical chain in a hypothetical market for smartphones
Suppose that Strawberrys marginal costs of producing a smartphone are equal to
= 50.
Strawberry supplies its smartphones to the Red Apple Store at price and the Red Apple Store
resells to consumers for a price equal to . Consumer demand for smartphones is given by
() = 750 .
The corresponding inverse demand equals
() = 750 .
We can model the interaction between Strawberry and Red Apple Store as a dynamic game:
81

lOMoARcPSD|1733731
1. Strawberry chooses the wholesale price . Red Apple Store observes .

2. Red Apple Store chooses the retail price and serves demand () = 750 .
The subgame perfect Nash equilibrium can be found using backward induction. We start by
solving the second stage of the game. We can find Red Apple Stores optimal price by equating its
marginal revenue to its marginal costs. The retailers revenue equals
= () = (750 )
so that its marginal revenue is equal to
= 750 2.
Note that Red Apple Stores marginal costs are equal to Strawberrys wholesale price . The
retailers optimal quantity follows by equating its marginal revenue to its marginal costs:
= 750 2 = =
which implies
750
= .
2
Note that this is also the quantity Strawberry sells to Red Apple Store for the simple reason that
Red Apple Store has to buy each smartphone it sells from Strawberry. Consequently, Strawberry
faces inverse demand
() = 750 2.
Its revenue equals
= () = 750 2 2 .
Strawberrys optimal quantity follows from
= 750 4 = = 50.
So, we find that in the subgame perfect Nash equilibrium,
700
= = 175.
4
The corresponding prices are
() = 750 2 = 400
and
() = 750 = 575.
Observe that the equilibrium price is way above the marginal cost price of 50.
Indeed, we can speak of a double marginalization problem because the price is substantially
higher than the price Strawberry and Red Apple Company would charge if the two firms were to
82

lOMoARcPSD|1733731
merge. The combined firm would act as a monopolist with marginal costs equal to 50. The
corresponding monopoly price is = 400. Because of double marginalization, welfare is lower
in the case of two monopolies in the vertical chain than in the case of one integrated monopoly,
exactly as the quote suggests (see figure 7.4). The reason is that both consumer surplus and
producer surplus are lower in the case of two monopolies. Consumer surplus is lower for the
simple reason that the price is higher. Producer surplus is lower because the price is different
from the monopoly price in the case of integration, which is the price that would maximize the
combined profits of Strawberry and Red Apple Company. A merger between the two firms is an
obvious remedy to the double marginalization problem. In chapter 17, we will examine various
alternative solutions.
Figure 7.4: Welfare consequences of two monopolies in the vertical chain
7.3 Case study: How chess players fail to backward induct

We have seen that backward induction is a very important concept in game theory. In simple 2-
stage games such as the ones seen in section 7.1, it seems to be a compelling solution to help
predict the outcome of the game. However, we may wonder to what extent the predictions of
backward induction truly conform to the decision-making that we see in the real world. Do
people really use the method of backward induction in practice? Even if they attempt to do so,
are they able to apply the method effectively in games that are more complex than the ones we
have seen?
Many experiments have been carried out to try and answer questions like these, with mixed
results. Overall, the evidence does suggest that economic agents do not engage in backward
induction as frequently as we might hope. Particularly in games that can potentially go on for
many stages, such as the well-known Centipede games, observed outcomes consistently depart
from the predicted subgame perfect Nash equilibrium found using backward induction. One
exception to this is provided by Palacios-Huerta and Volijs study, which used chess players as
participants and found that the majority of them played the game according to backward
induction. They claim that, although we perhaps should not expect everyone to always behave as
predicted by theory, backward induction reasoning is second nature to expert chess players.
83

lOMoARcPSD|1733731
Supporting this, they also find that chess Grandmasters are particularly successful in following
optimal strategies.
However, even this result is called into question by subsequent experiments carried out by
Levitt, List and Sadoff. They also use chess players as participants, but their results are sharply
at odds with the aforementioned study. As well as the Centipede game, these authors use a game
which is arguably a more direct test of backward inductive reasoning: the race to 100 game.
The game involves two players who take turns in choosing a number between 1 and 9. Player 1
starts by choosing a number. Then Player 2 chooses a number, which is added to Player 1s
number to create a sum. Then Player 1 chooses again; his number is added to the previous sum
to make a new sum, and so on. The winner of the game is the player that eventually chooses a
number to make the sum equal to 100. In this game, Player 2 (i.e. the second mover) has the
advantage and is guaranteed to win, if he follows backward induction. Reasoning from the last
stages of the game, Player 2 wants to be the one to make a sum equal to exactly 90. If he does
this, then regardless of which number 1-9 is chosen by Player 1, Player 2 can always respond by
choosing the number that takes the sum to 100. Following the same reasoning, Player 2 wants to
be the one making a sum of 80, 70, and so on. Therefore, the equilibrium strategy has Player 2, in
his first move, choosing the value that takes the sum to 10 and then in subsequent moves taking
the sum to all the key numbers: 20, 30, 40, , all the way to 100. Of course, if Player 2 does not
realize this in the beginning, Player 1 has the opportunity to respond by himself targeting the
key numbers 10, 20, , 100.
In fact, only 39% of players solved the game in their first move. Notably, when Player 2s fail to
use their advantage, Player 1s do not necessarily see the opportunity to capitalize. Ultimately, in
one quarter of the games played, neither player has solved the game even by the time the sum
has reached 90.
The authors also test a slightly different variant of the game, where players can choose numbers
1-10. This difference means that backward induction indicates the sum of 89 as being key to
ensure victory. The other key numbers become 78, 67, , 12, 1. Thus, while the logic involved
in the optimal strategy is exactly the same, the pattern is perhaps harder to see. Interestingly,
this minute variation appears to make the game much more difficult. Although the same
proportion of games is solved by the time the latter stages are reached, very few players are able
to solve this variant of the game in the early stages. Even Grandmasters fare only slightly better
than other chess players in this respect.
The results suggest that even chess players, who we assume are accustomed to thinking several
steps ahead, may not behave in accordance with the logic of backward induction. If that is the
case, we should certainly exercise caution when using game theory and backward induction to
predict real-world outcomes, particularly in complex scenarios. Even if some agents do behave
according to backward induction, the fact that not all of them do (or that some of them make
errors) would mean that the best response of a rational agent is not necessarily to follow
backward induction himself.
At least two qualifications should be made to this warning, however. The first relates to a
general point about experimental economics and comparability between experiments: the
evidence on use of backward induction remains quite mixed and although these authors tried to
replicate the conditions of a previous experiment (where, in fact, chess players did behave
84

lOMoARcPSD|1733731
according to backward induction), it is always difficult to know how seriously the participants
took the games that they were asked to play. Secondly, it is important to note that this
experiment involved only one-shot games, so that there was no opportunity for players to learn
the optimal strategy by playing the games several time. In many real-world contexts it is quite
plausible to assume that economic agents learn from experience. Then, the outcomes may more
often resemble the predictions of game theory.
As mentioned in the main text, the notion of subgame perfection stems from the work of
Reinhard Selten (1965, 1975, 1988). Selten shared the 1994 Nobel Prize in economics with John
Nash and John Harsanyi for their pioneering analysis of equilibria in the theory of non-
cooperative games. In his 1965 paper Ein Oligopolmodell mit Nachfragetrgheit (An Oligopoly
Model with Demand Inertia), he introduced the concept of subgame perfection to solve a
dynamic oligopoly model. He argued that some of the Nash equilibria are economically
unreasonable and should therefore be eliminated in a similar manner to what backward
induction would dictate. He further developed the notion of subgame perfection in his later work
published in 1975. You should also refer to his influential book Models of Strategic Rationality
published in 1988 for an extensive treatment of subgame perfection and equilibrium stability in
non-cooperative dynamic games.
The double marginalization problem can be traced back to Cournot (1838), even though
Edgeworth (1889) and Pareto (1896) probably were the first to formally analyze the problem.
For more recent contributions, extensions and applications alternative settings you should also
refer to Mueller (1969), Jeuland and Shugan (1983), Salinger (1988) and Riordan (1998).
This chapters case study is based on articles by Levitt et al. (2011) and Palacios-Huerta and Volij
(2009) published in the American Economic Review.
References
Cournot, A.A. (1838/1927). Researches into the Mathematical Principles of the Theory of Wealth,
translated by N.T. Bacon, New York: Macmillan.
Edgeworth, F.Y. (1889). On the application of mathematics to political economy, the address of
the president of section F (economic science and statistics) of the British association at the
fifty-ninth meeting held at Newcastle-upon-Tyne in September 1889, Journal of the Royal
Statistical Society, 52(4), 538-576.
Jeuland, A. and A. Shugan (1983). Managing channel profits, Marketing Science, 2, 239-272.
Levitt, S.D., J.A. List and S.E. Sadoff (2011). Checkmate: exploring backward induction among
chess players, American Economic Review, 101(2), 975-990.
Mueller, D.C. (1969). A theory of conglomerate mergers, Quarterly Journal of Economics, 83(4),
643-659.
Palacios-Huerta, I. and O. Volij (2009). Field centipedes, American Economic Review, 99, 1619-
1635.
Pareto, V. (1896). Cours d economies politique, Lausanne, Switzerland: F. Rouge.
Riordan, M.H. (1998). Anticompetitive vertical integration by a dominant firm, American
Economic Review, 88(5), 1232-1248.
Salinger, M.A. (1988). Vertical mergers and market foreclosure, Quarterly Journal of Economics,
103(2), 345-356.
85

lOMoARcPSD|1733731
Selten, R. (1965). Spieltheoretische Behandlung eines Oligopolmodells mit Nachfragetrgheit

(An oligopoly model with demand inertia), Zeitschrift fr die Gesamte Staatswissenschaft, 121,
301-324 and 667-689.
Selten, R. (1975). Reexamination of the perfectness concept for equilibrium points in extensive
games, International Journal of Game Theory, 4, 25-55.
Selten, R. (1988). Models of Strategic Rationality, Dordrecht: Kluwer.
Exercises
Exercise 7.1 Subgame perfect Nash equilibrium

Woody, the wholesale fruit seller you encountered in chapter 5, is still trying to become the
monopolist in the oranges market of Amsterdam. Along with oranges, he now wants to introduce
another product in his fruit stands, orange juice. However, he fears the reaction of his main
competitor, Steven, who may begin a price war. The two orange sellers currently compete
normally and earn 1,000 per week. Woody believes that in the case of normal competition,
his revenues would rise by 40% with the introduction of orange juice, and his competitors
revenues would fall by 30%. Steven observes Woodys choice and decides whether to behave
normally or aggressively. In the latter case, both sellers revenues would decrease by 80%.
Woody faces 200 of fixed costs if he introduces the orange juices in the market.
a) Calculate the payoffs of both sellers and describe the game in an extensive form.
b) Find the subgame perfect Nash equilibrium.
Woodys expectations of the increase of his revenues after the introduction of the orange juices
may not be completely accurate. To describe this uncertainty, assume that his payoff also
depends on the probability that Woody obtains a positive revenue. With probability , his
expectations are fully accurate. With probability 1 , he will not have any increase in his
revenues. Assume that the uncertainty only afflicts Woodys increase in revenues, not the
expected decrease in Stevens revenues, which do not depend on .
c) Describe how the subgame perfect Nash equilibrium changes as a function of .
Exercise 7.2 Subgame perfect Nash Equilibrium

The airplane industry is dominated by two competitors, Aireing and Bobus. The two companies
compete la Cournot in the production of airplanes (for simplicity, assume there is only one
type of airplane). Both have the same cost function ( ) = 20 , where is the number of
airplanes produced by each company, with = , for Aireing and Bobus, respectively. The
market demand can be described as follows:
() = 900 3,
where represents the price of one airplane in millions of euros. Recently, Jimmy the engineer
developed a new production process with which the costs of producing an airplane are
drastically decreased to ( ) = 10 . Jimmy can sell his invention to only one of the airplane
companies for 3,000 million, and he decides to first offer it to Aireing. If the company refuses to
buy his invention, Jimmy will offer it to Bobus. Moreover, if Aireing buys the invention, Bobus
may hire a spy for 2,000 million who will steal the blueprints of Jimmys new process, and the
86

lOMoARcPSD|1733731
company will be able to produce airplanes at the same cost as Aireing (10 ). For simplicity,
assume that Aireing cannot hire a spy. Who, if any, is going to buy Jimmys invention? Describe
how the companies will strategically behave. (Hint: there are four possible scenarios, and two of
them are mirror images. When needed, approximate to the first decimal).
Exercise 7.3 The double marginalization problem

Assume that Canon is a monopolist in providing ultra-high capacity memory cards for cameras.
Assume also that its marginal cost of supplying a memory card for one more camera is zero.
Denote by the price charged by Canon for its memory cards.
Nikon and Kodak assemble high-end cameras. Every camera is equipped with a memory card
developed by Canon. Suppose that the cost to them of all the parts necessary to build a camera
(excluding memory card) adds up to 800 per camera, and that assembly costs are another
200 per camera. Finally, cameras are a homogeneous good and their annual demand is given by:
= 50 0.01
where is quantity (in millions) and is price as usual. Suppose that because of intense
competition between Kodak and Nikon, the downstream market is perfectly competitive.
a) For any given price of a memory card, what will be the price and sales of cameras?
b) What price maximizes Canons profits? Calculate the profits in both markets and the price
of a camera.
c) How much money would a vertically integrated firm controlling both the supply of memory
cards and the assembly of cameras make? What price would such a firm charge for its final
product?
d) Could Canon make more money by integrating downstream into photographic camera
assembly?
Suppose now that a single firm, Sony, has a monopoly over the assembly of cameras.
e) For a given price for a memory card, what price would Sony set for its cameras and how
many of them would be sold in the market?
f) What price should Canon set for its memory cards? Calculate Canon and Sonys profit and
the final price of a high-end camera.
g) Could Canon and Sony make more money by merging? Would such a merger benefit or harm
camera users, and by how much?
87

lOMoARcPSD|1733731
Chapter 8: Optimal incentive contracts

In chapter 2, we discussed the Principal-Agent model and constructed value-maximizing
incentive contracts. In our analysis, we ignored an important aspect in the relationship between
the principal and agent: The agent may have outside options at his disposal so that he may
choose not to accept the principals contract. In this chapter, we will study the effect of the
agents outside option on the incentive contract the principal will offer the agent. In section 8.1,
we will return to the Principal-Agent game between Penlope and Antonio to derive the
incentive contract that is optimal from the viewpoint of the principal. In section 8.2, we will
discuss the consequences of the agents risk attitude for the optimal contract. In the case study in
section 8.3, we will examine the potential negative effect of small monetary incentives.
8.1 The optimal incentive contract

Reconsider the Principal-Agent game we introduced in chapter 2, in which agent Antonio directs
a movie on behalf of principal Penlope. Antonios cost () of exerting effort equals
() = 2
where > 0. There are no costs other than the costs of effort. Each unit of Antonios effort
yields one additional movie theater ticket sold, i.e., Antonios output equals
= .
Suppose that Penlope obtains a price per unit of output. Penlope offers Antonio a linear
contract which specifies that Antonio will receive a base wage plus a bonus for each movie
theater ticket sold. Of course, Antonio is free to reject Penlopes offer. Suppose that his most
attractive outside option gives him utility . For Antonio to accept Penlopes contract, he
should at least obtain utility from it.
Let me summarize the interaction between Penlope and Antonio in terms of a dynamic game
with the following three stages:
1. Penlope offers Antonio a contract that pays Antonio + (= + ).

2. Antonio accepts or rejects the contract. If Antonio rejects the contract, the game ends.
Antonio obtains utility and Penlope obtains zero profits.
3. Antonio chooses effort .
We can solve this game using backward induction, starting at stage 3. Antonio will choose effort
to maximize his utility:
= + 2 .
The first-order condition yields
= 0
or, equivalently,

= .

Antonios resulting utility equals
88

lOMoARcPSD|1733731
2
= + 2 = + .
2
In stage 2, Antonio will accept the contract (only) if
2
=+ .
2
In stage 1, Penlope has to decide on both Antonios base wage and his bonus per unit of
effort. Given the bonus, she would like to keep the base wage as low as possible. Optimally, she
fixes at such a level that Antonio will just accept the contract, which is the case if he obtains
the same utility from the contract as from his best outside option. Therefore,
2
=+ =
2
or, equivalently,
2
= .
2
What remains is the optimal bonus. To determine this, it is useful to write down Penlopes
profits:
= .
Penlope obtains the revenue from Antonios output () and pays Antonios salary consisting
of his fixed wage () and a bonus (). When we substitute = , Antonios effort = / and
the optimal wage = 2 /2 in Penlopes profits, we obtain
= ( )
( ) 2
= +
2
(2 )
= .
2
We can now find the optimal bonus by maximizing with respect to . The first-order
condition of this maximization problem is
2 2
= = 0,
2
which implies
= .
Observe that the optimal bonus equals the unit price of Antonios output, which coincides with
Penlopes marginal benefits from Antonios effort. The result implies that Penlope allows
Antonio to keep all the fruits of his efforts. This finding may not be as surprising as it seems at
first sight. Recall that we derived in chapter 2 that the bonus in the value-maximizing contract is
89

lOMoARcPSD|1733731
also equal to the price per unit of output. In the optimal contract, Penlope ensures that the
contract maximizes total value. In order to retain a maximum share of this value, Penlope sets
the fixed wage she pays Antonio as low as possible. Moreover, in the event that Penlope is in
the luxury position of attracting several applicants for the job to direct her movie, she can select
the most productive one by offering the contract to the agent that is willing to accept the lowest
fixed wage.
8.2 Risk sharing

Note that the optimal fixed wage Penlope pays Antonio may be negative. For instance, in the
extreme case that = 0, Antonios fixed wage equals
2 2 2
= = = .
2 2 2
So, Antonio should pay Penlope to work for her! Such contracts actually exist in practice. A
good example is a franchise contract, for example between the hamburger chain McDonalds and
the restaurant manager. The restaurant manager pays McDonalds a fixed amount a year. In
exchange, he is allowed to keep the profits he obtains from the McDonalds products he sells.
Still, contracts where the agent pays the principal a lump sum are rare in practice, in particular
between employers and employees. To get some intuition why, think about a football player in a
team that participates in the Champions League. Decisive goals in this tournament may be
worth millions of euros for his club. According to the logic above, if the football player scores
such an important goal, he should obtain all the benefits from it. At the start of the tournament,
the expected benefits for the football player may be worth several million euros. Would the
football player accept a contract that specifies that he pays such an amount upfront to his club?
Probably not. First of all, the football player may not have the cash available to pay this amount.
Banks are likely to be reluctant to provide a loan to the player because of the risks involved.
What if the player gets injured so severely that he misses important matches? What if his team is
eliminated in an early round of the tournament? What if the player misses some good goal-
scoring opportunities because of bad luck or incorrect referee decisions? In all cases, the player
would get a lower bonus than expected and might not be able to repay his loan.
Second, even if the player could pay upfront, he may be concerned about the risks of such a
contract. Indeed, if the agent is risk averse, the above contract is no longer optimal for the reason
that the agent may simply reject it and opt for a safe outside option. In the remainder of this
chapter, we will look at optimal contracts in the case of a risk averse agent. In order to do so, we
include some uncertainty in the model: We assume that the agents output depends not only on
his effort but also on events outside his control. In particular,
= + ,
where is a random number with expected value zero ({} = 0). If the principal does not
observe the agents effort but only his output , she can only base the agents payment on .
Another interpretation of this model is that the principal makes a measurement error when
assessing the agents effort.
This model may apply to many real-world settings. In agriculture, a farmers output () depends
on how much time and effort he puts in his land () and the weather conditions (which has an
90

lOMoARcPSD|1733731
impact on ). A companys stock market value depends on both the CEOs effort and the
sentiment on the stock market. Your result for an exam may depend on how much time and
effort you invest in the course and how well you sleep the night before the exam (although this
might not be entirely outside your control because you may sleep better if your feel confident
that you studied hard for the exam).
In our model, the agents payoff from the above linear contract equals
+ = + + .
Note that for a risk neutral agent, the expected utility from this contract equals
= { + + } = + + {} = + .
So, a risk neutral agents expected utility is not affected by potential events outside his control or
a measurement error. Therefore, a risk neutral agent is willing to accept a contract where he
becomes the residual claimant of the fruits of his effort despite the risks that accompany such a
contract. However, a risk averse agent may obtain disutility from such randomness in his
payment. The principal should compensate the agent by paying him a risk premium on top of his
base wage for the disutility he suffers. How to calculate this risk premium is beyond the scope of
this course. Instead, we will discuss two general principles that are relevant for the construction
of the optimal contract.
The principles are based on the trade-off between incentives and risk sharing. In the extreme
case of no incentives ( = 0), the contract is completely risk free so that the principal does not
have to pay the agent a risk premium. The higher the power of incentives, i.e., the greater , the
greater the random component in the agents payoff and so the higher the risk premium the
principal has to pay to compensate the agent for the additional risk. It can be shown that in
general, it is optimal for the principal to provide a risk averse agent with some incentives, i.e.,
> 0 but less than full incentives, i.e., < . The incentive intensity principle specifies how the
optimal depends on the type of task and the agents risk preferences.
The incentive intensity principle: The optimal intensity of incentives () is higher (1)
the more responsive the agent is to incentives (i.e., the lower ), (2) the greater the
principals incremental profits from additional effort (i.e., the higher ), (3) the more
precisely performance can be measured (i.e., the lower the variance of ), and (4) the
greater the agents risk tolerance.
Sometimes, the principal can measure some of the events beyond the agents control. If she can
do so, it is in her best interest to correct for such events as much as possible. By doing so, she
reduces the variance of the agents payoffs. As a result, according to the incentive intensity
principle, the principal can give the agent higher incentives which will induce the agent to exert
more effort. The informativeness principle summarizes this observation.
The informativeness principle: Total value is increased by correcting for events

beyond the agents control in such a way that the error with which the principal
measures the agents performance is reduced.
The informativeness principle is not without application. An example is comparative

performance evaluation, which we introduced in chapter 5. To fix the idea, suppose that the
91

lOMoARcPSD|1733731
principal employs several agents and that the agents performance depends on a common
random factor (e.g., the weather). If this random factor affects each agents performance in the
same way, the principal can cancel it out in the agents payments by looking at their relative
performance instead of their absolute performance. For example, the regulation of regional
monopolies is based on comparative performance evaluation. More in particular, the regulator
can impose on each regional monopoly a maximum price that depends on the costs of the other
regional monopolies. Such a regulatory scheme is called yardstick competition. Yardstick
competition is applied in the electricity market in many countries. Usually, the electricity
networks are monopolies: Households are connected to exactly one network. However, in
different regions different companies may own the network. Because each firm has a monopoly
in the regions it supplies, regulation is still needed to prevent the firms from charging
skyrocketing (and welfare-reducing) prices. Yardstick competition is one way of doing so.
Yardstick competition has the advantage that it takes account of common shocks (such as an
unanticipated increase or decrease in fuel prices) that affect all firms. As a consequence, the
regulator can give the firms greater incentives to operate cost efficiently, which will ultimately
lead to lower electricity prices.
One cautionary note on the applicability of the informativeness principle: The principal may be
tempted to condition an agents performance standard on the basis of the agents past
performance. Indeed, the principal could argue that the agents past performance is an indicator
of how easy or hard the circumstances are under which the agent operates. When those
circumstances are difficult to assess before the agent signs the contract with the principal, they
may add to the perceived risk of the contract. The principal can partly take away this risk by
letting the agents performance standards depend on his performance in early periods. This
phenomenon is known as the ratchet effect because the standards are ratcheted up in
response to good performance. In chapter 3, we saw an example of the ratchet effect in
situations where a regulator does not commit to the price cap it set once the regulated firm
decreases its costs. So, why is the ratchet effect undesirable? The reason is that the agent will
have all the incentives to underperform in early periods as otherwise he will be punished later
by higher performance standards.
8.3 Case study: Pay enough or dont pay at all

According to standard economic theory, performance on an activity improves the greater the
financial incentives provided for that activity. Performance is positively related to effort and
since effort is unpleasant and money is good, a monetary compensation will tend to increase
performance. However, in certain situations, factors other than money and effort, such as social
norms or intrinsic motivation, may enter into the decision of the agent and counterbalance or
even reverse the effect on performance. Let us do a quick thought experiment to see where such
a negative effect might come from. Image your professor promised you 1 as an award for
getting a very high grade in your final exam. Would you prepare harder for your exam? Most
probably not, since you might think that such a small bonus does not compensate for all the
extra effort of preparing for the exam. It might even be the case that this monetary incentive
would have the opposite effect on your performance, since you might feel that the professor is
offending you by offering such a small bonus! Think of the following case, too. Suppose you are
on a first date with someone. At the start of the evening, he or she promises to pay you a bit of
money at the end of the night if you give him or her a great evening. Would you really make an
92

lOMoARcPSD|1733731
additional effort to entertain such a person? I guess not. You would probably say goodbye right
after your date made the proposal.
Uri Gneezy (UCSD) and Aldo Rustichini (University of Cambridge) felt that the negative effect of
small monetary incentives might be more common, so they designed several smart experiments
to test it. Their first experiment took place at daycare centers in Israel. Before the experiment,
parents regularly collected their child late, forcing teachers to stay at the daycare after closing
time. To incentivize parents to pick up their kid on time, the daycare centers introduced a small
fine for latecomers. Guess what happened: the number of late coming parents increased! Clearly,
the monetary incentive failed to act as an effective mechanism. A potential explanation is that
before the imposition of the fine, parents were bound by a social contract, where social norms
about being late inclined them to show up on time and avoid the guilt that comes with being late.
However, when the fine was imposed, parents became bound by a market contract; they
suddenly started paying for their tardiness with money instead of guilt. Once this happened, the
parents could decide for themselves if they wanted to be late or not.
This example raises a lot of additional questions. In which cases, if any, does a monetary
incentive have a positive effect on performance? Do small monetary incentives crowd out
intrinsic motivation? And what effect does the interplay between market and social norms have
on choosing an optimal bonus? Gneezy and Rustichini designed two smart experiments to
answer these questions and test the potential negative effect of small bonuses.
In their first experiment a group of 160 students at the University of Haifa in Israel were asked
to take an IQ quiz. At the beginning of the experiment, each student was promised a fixed
payment of 60 NIS (New Israeli Shekel) for participation (1 NIS corresponds to around 0.265 in
the period when the experiment was conducted). The students were divided into four different
groups, with each group consisting of 40 students and corresponding to a different treatment.
The first group was simply asked to answer as many IQ questions as they could. In addition to
that, the second group was promised a payment of 10 cents (of an NIS) for each question they
answered correctly. Subsequently, the third group was offered a higher monetary incentive of 1
NIS, and the fourth group the highest one corresponding to an amount of 3 NIS per question.
As the authors expected, the introduction of high monetary incentives had a positive influence
on performance. However, the results indicate that you should pay enough if you wish to
increase performance; weak monetary incentives appear to be ineffective. The average number
of questions answered correctly declined from slightly more than 28 in the first group to 23
questions in the second group, where an additional 10 cents of a NIS were offered. In contrast,
this number increased to more than 34 in the third group, and was stable at 34 in the fourth
group, where higher monetary incentives were offered. Their results seem to indicate that the
effect of the use of financial compensation affects individuals in the same way, despite diverse
individual characteristics such as greater talent or higher willingness to put in effort.
Gneezy and Rustichinis second experiment involved high school students, again in Israel, who
did volunteer work. Once a year, students go from house to house collecting monetary donations
that households make to the society for causes like cancer research, assistance to disabled
children and so on. In this experiment, 180 such students were divided into three separate
groups. The first group was simply given a short speech recalling the importance of the activity
the students were going to perform. Each student in the second group, in addition to the speech,
93

lOMoARcPSD|1733731
was promised a bonus pay of 1 per cent of the total amount he or she raised. Finally, each
student in the third group was promised a bonus pay of 10 per cent of the amount he or she
raised. In both cases it was made clear that the bonus payment was financed by the
experimenters and was not part of the amount collected. The average amount collected from
students in the first group (with no payment) was 238.67 NIS. That average fell to 153.67 NIS in
the second group and was 219.33 NIS in the third group, which is higher than in the second
group but still lower than in the first group.
In both experiments, the effect of small rewards turned out to be detrimental for performance.
One possible explanation of this results is that a monetary reward crowds out the intrinsic
motivation for doing the task which could result in lower incentives if the monetary reward is
small. For example, students may lose some of their intrinsic motivation to raise money for
charity when they get a financial compensation for their efforts. Another explanation could be a
change in perception of the contract between the principal (the experimenters in this case) and
the agents (the subjects). Without monetary incentives, the students may have felt that they
should perform well to compensate the experimenters for receiving a show-up fee to participate
in the experiment. In the case of monetary incentives, the subjects might consider the show-up
fee as a sunk benefit and only condition their performance on the level of the incentives.
Whatever the interpretation, the message from this article is clear: If you desire high
performance, pay enough or dont pay at all!
You should primarily refer to the references and notes of chapters 2 and 5, as the theory of this
chapter is an extension of the ideas presented there.
Modern incentive theory developed in the 1970s by various authors who proposed a number of
optimal incentive contracts in a variety of contexts such as insurance (Michael Spence and
Richard Zeckhauser 1971), taxation schemes (James Mirrlees 1971, 1976, 2006), and managerial
compensation (Robert Wilson 1968 and Stephen Ross 1973). The optimal linear incentive
contracts and main intuition of incentive pay developed in this chapter stem largely from the
work of Bengt Holmstrm and Paul Milgrom (1987, 1991). Holmstrm (1979, 1982), together
with Steven Shavell (1979), also originated the informativeness principle, as a solution to
information asymmetry problems in incentive pay. Extensions to that idea and a more
generalized treatment can be found in later work by Grossman and Hart (1983). Green and
Stokey (1983) show that relative performance schemes like tournaments dominate piece rates
in the present of a sufficient diffuse common random shock affecting the performance of all
agents.
The case study of this chapter is based on the work of Uri Gneezy and Aldo Rustichini (2000ab).
References
Gneezy, U. and A. Rustichini (2000a). A fine is a price, Journal of Legal Studies, 21(1), 1-17.
Gneezy, U. and A. Rustichini (2000b). Pay enough or don't pay at all, Quarterly Journal of
Economics, 115, 791-810.
Green, J.R. and N.L. Stokey (1983). A comparison of tournaments and contracts, Journal of
Grossman, S.J. and O.D. Hart (1983). An analysis of the principal-agent problem, Econometrica,
51(1), 7-45.
94

lOMoARcPSD|1733731
Holmstrm, B. (1979). Moral hazard and observability, Bell Journal of Economics, 10(1), 74-91.
Holmstrm, B. (1982). Moral hazard in teams, Bell Journal of Economics, 13(2), 324-340.
Holmstrm, B. and P. Milgrom (1987). Aggregation and linearity in the provision of
intertemporal incentives, Econometrica, 55(2), 303-328.
Holmstrm, B. and P. Milgrom (1991). Multitask principal-agent analyses: incentive contracts,
asset ownership and job design, Journal of Law, Economics and Organization, 7, 24-52.
Mirrlees, J.A. (1971). An exploration in the theory of optimum income taxation, Review of
Economic Studies, 38(2), 175-208.
Mirrlees, J.A. (1976). The optimal structure of incentives and authority within an organization,
Bell Journal of Economics, 7(1), 105-131.
Mirrlees, J.A. (2006). Welfare, Incentives and Taxation, Oxford: Oxford University Press.
Ross, S.A. (1973). The economic theory of agency: the principal's problem, American Economic
Review, 63(2), 134-139.
Shavell, S. (1979). Risk sharing and incentives in the principal and agent relationship, Bell
Journal of Economics, 10(1), 55-73.
Spence, M. and R. Zeckhauser (1971). Insurance, information, and individual action, American
Wilson, R. (1968). The theory of syndicates, Econometrica, 36(1), 119-132.
Exercises
Exercise 8.1 Optimal Incentive Contracts

Sharon is a brilliant cook, currently working in one of the most famous restaurant of Amsterdam,
where she receives a monthly utility of . Recently a renowned hotel offered her the job of
master chef of its restaurant. The contract provides a fixed monthly income of , plus a bonus
() for each costumer served (). With each unit of effort (), Sharon is able to prepare meals
1
for 3 customers a month. However, she also incurs a monthly cost of () = 6 2 . Assume that
Sharon first decides whether to accept or reject the hotels offer (in the latter case, she continues
to work at her current restaurant), and if she accepts, she decides on her effort level.
a) Find Sharons optimal level of effort.
b) Compute the fixed wage that the hotel will offer to convince Sharon to change jobs.
c) What is the optimal bonus for the hotel?
d) Assume that the hotels price for a meal is 50, and that Sharons utility obtained from her
current contract equals 12,000. What fixed wage will the hotel offer Sharon?
Exercise 8.2 Optimal Incentive contracts

Two producers of smartphone applications, Zaz and Pep, offer programmer Tom two different
job contracts. On the one hand, Zaz will pay Tom 40 a day, plus 1 for each application Tom
invents. On the other hand, Peps contract provides 10 as a fixed daily wage, and an additional
10 for each application developed. Tom can create one application per unit of effort he puts in
his job, and he has a cost of effort equals to: () = 3 2 . However, most of Toms friends work
for Zaz, and he would be intrinsically motivated to work with them. In fact, on top of the usual
95

lOMoARcPSD|1733731
utility, he obtains two additional units of utility for each unit of effort he exerts at Zaz. Therefore,
his utility function if he works for Zaz is: = + 2 + 3 2 , where is the fixed wage
and is the bonus he receives per application developed. Tom could also decide not to work at
all, in which case he would gain a utility of = 10.
a) In which company will Tom exert more effort? Which company will he choose to work for?
b) Assume that the two companies pay Tom only the bonus ( = = 0), to what extent will
Toms effort level change? Where will he decide to work?
Exercise 8.3 Optimal Incentive Contracts

Arctic Ice is a new company in the market of ice cubes. The company had a great idea: It wants to
transform ice cubes into a luxury good. How? By taking the ice directly from the North Pole!
Unfortunately, not many people are willing to go to the Arctic icy lands to shovel ice. For this job,
the company offers a fixed wage , plus a bonus for each kilogram of Arctic ice shovelled. One
of the candidates is Russell, who can shovel 3kg of ice for each unit of effort and incur costs equal
to ( ) = 43. However, Russell may decide not to accept Arctic Ices offer, in which case he
would receive utility.
a) Find Russells optimal effort level, and the contract (optimal fixed wage and optimal bonus)
that the Arctic Ice offers him.
b) Compute Arctic Ices profits. What do you conclude?
Exercise 8.4 Risk Sharing

Angelina is an orange harvester hired by the Woody&Co with a linear contract, which provides a
fixed wage , and a bonus for each kilogram of oranges harvested. Of course, the amount of
oranges that Angelina harvests depends not only on her effort level, but also on the weather
2
conditions, which are substantially random. In particular = + , where and
3
represent the quantity of oranges harvested by Angelina and her effort level. The variable is a
1 1
random variable with mean 3 during summer periods and + 3 during winter, which reflects
how, for a given effort, the orange harvest will be more abundant during winter than during
1
summer. Angelina faces a cost of ( ) = 2, and she could quit her job gaining utility.
6
Woody&Co is aware of the seasonal expected change of the harvest.
a) Assume that Angelina is risk-neutral, and that Woody&Co maximizes its expected profits.
How does Angelinas optimal effort change passing from winter to summer periods? And the
optimal fixed wage and bonus? What do you conclude?
b) Without doing any calculus, how do you think Angelinas optimal effort would change in case
she was a risk-averse person? What if she was a risk lover?
c) In general, what kind of risk compensation do you expect if Angelina is either risk-averse or a
risk lover?
d) How could Woody&Co decrease Angelinas risk?
96

lOMoARcPSD|1733731
97

lOMoARcPSD|1733731
Chapter 9: Market entry and product positioning

In many market settings, firms make strategic decisions sequentially. In this chapter, we will
consider two such settings. Section 9.1 contains a model of market entry. This model is a
dynamic game where companies first decide whether or not to enter into a market, after which
the entrants compete against each other. In Section 9.2, we will reconsider the Hotelling model.
In contrast to what we assumed in chapter 4, firms can decide on their product location before
fixing prices. Section 9.3 includes a case study on product positioning in the car industry in the
US in the early twentieth century.
9.1 Market entry

In chapter 6, we found that in Cournot markets, the more firms are active in the market the
lower is the equilibrium price. This result might suggest that in terms of welfare, it is always
desirable that as many firms enter the market as possible. However, does this logic still holds
true if firms have to incur costs before entering the market? Entry costs can be quite substantial.
Entry typically involves building production facilities, opening distribution channels, advertising
the firms products, and so forth. It might be quite efficient if only a limited number of firms
enters the market to avoid a welfare reducing duplication of entry costs. Of course, if only a few
firms enter the market, they might not compete very fiercely. In other words, there is a trade-off
between competition and mitigating entry costs. The resulting outcome is that generally, it is
optimal from a welfare point of view that only a handful of firms enters the market, sometimes
as few as one. Then, the question is whether too few or too many firms will enter the market if
market entry is unregulated.
To study this question, we study a model of entry where firms face fixed entry costs. Consider a
market with demand
() = (1 ).
We refer to as the market size. If supply equals , the resulting market price is given by

() = 1 .

Suppose that firms interact in the following two-stage game:
1. Firms simultaneously decide whether or not to enter into the market.

2. The firms that enter compete la Cournot.
If a firm enters, it incurs fixed costs > 0. For simplicity, we assume that the marginal costs of
production are zero for all firms.
The subgame perfect Nash equilibrium follows using backward induction. Suppose that in stage
1, 1 firms enter. We can find the equilibrium quantity in the second stage using the formula
that we derived in chapter 6:

= = .
( + 1) ( + 1)
The equilibrium price equals
98

lOMoARcPSD|1733731
1
=+ = .
+ 1 + 1
The corresponding profits for an entrant are equal to

= = .
( + 1)2
In the first stage, firms will enter only if they make positive profits. Therefore, the number of
firms that enter follows from the break-even condition:

= =0
( + 1)2
which implies

= 1.
If is not an integer, we round it downwards to obtain the number of firms that enter (e.g., if
= 4.567, four firms will enter the market).
From the formula for , we learn that the larger the size of the market () and the smaller the
fixed costs () the more firms will enter the market. The other way around, could be so small
relative to that only one firm enters the market. This is not only a theoretical possibility: In
many markets, there is room for only one firm. Examples of such natural monopolies are
railways, water supply, and cable television. The fixed entry costs for those markets mainly
relate to the network that has to be built to be able to serve clients.
Another lesson from the above expression for is that the equilibrium number of entrants
increases less than proportionally with respect to market size . For instance, if market size
doubles, the number of entrants increases but it does not double, as table 9.1 shows. The
intuition behind this result is the following. If market size doubles, more firms will enter the
market. Because there are more firms in the market, competition will become fiercer, which has
a negative effect on each firms price-cost margin. Therefore, total industry profits will less than
double so that the number of firms making sufficient profits to cover their fixed costs will not
double either.
Table 9.1: The equilibrium number of firms increases less than proportionally with market size.

1,000 100 2
2,000 100 3
4,000 100 5
8,000 100 7
16,000 100 11
32,000 100 16
We conclude this section by considering the question of whether too few or too many firms
enter a market from a welfare point of view. To do so, let us return to our market entry model.
99

lOMoARcPSD|1733731
You can calculate total welfare in this market as follows. If the market price equals , consumer
surplus is equal to
1
= (1 )2 .
2
Recall that we assumed marginal costs to be zero and that firms incur the fixed costs of entry
so that producer surplus equals
= ( ) {fixed costs} = (1 ) .
Because the equilibrium price equals
= 1/( + 1)
welfare can be written as
1
= + = (1 )2 + (1 )
2
1 1
= ( + 2 ) + ( 2 )
2 2
1
= (1 2 )
2
1 1
= (1 ) .
2 ( + 1)2
When we maximize welfare with respect to , we obtain the first-order condition for the
efficient number of entrants :

= /( + 1)3 = 0

which implies
3
= 1.

Because a cubic root is smaller than a square root for numbers greater than 1, we can conclude
that too many firms enter in equilibrium from a welfare point of view. Table 9.2 shows that the
gap between the equilibrium number of entrants and the efficient number of entrants can be
substantial.
So, why do we find excessive entry? The reason is that an entrant imposes a negative externality
on the incumbents: If it enters, it steals business from the incumbents in the sense that their
equilibrium output is reduced. Because the entrant will only consider its own fixed costs of entry
when making the entry decision and not the additional welfare loss from reduced supply by the
other firms, it may also enter when it is not desirable from a welfare point of view.
In many settings, market entry may not be as excessive as the above analysis suggests. For
example, the disadvantages of additional entry may be mitigated in the case of product
100

lOMoARcPSD|1733731
differentiation. The reason is that consumers not only benefit from a lower price but also from
greater product variation. Another countervailing factor could be cost efficiencies. If an entrant
is much more efficient than the current firm, it is likely that its entry enhances welfare.
Table 9.2: Equilibrium entry versus efficient entry

800 100 1 1
2,700 100 4 2
6,400 100 7 3
12,500 100 10 4
21,600 100 13 5
34,300 100 17 6
9.2 Product positioning

Now, we return to the Hotelling model that we already discussed briefly in chapter 4. To refresh
your memory, let me repeat the example of the two ice cream vendors, Clint and Dennis. Clint
and Dennis have to decide where to locate their ice cream stands on a one-kilometer-long beach.
Both produce the very same ice cream and they do so at constant marginal costs = . We
assume that 1,000 consumers are uniformly distributed over the entire beach. In this chapter,
we relax the arguably unrealistic assumption that prices are fixed at the point that Clint and
Dennis decide where to locate. More in particular, suppose that the two ice cream vendors
interact in the following dynamic game consisting of three stages:
1. Clint and Dennis simultaneously decide where to locate on the beach.

2. Clint and Dennis simultaneously fix the price of their ice cream.
3. Consumers decide whether to buy ice cream at Clints or Dennis stand (if they buy at
all).
In each stage, all players know the decisions made in the previous stage(s).
We can solve this game using backward induction. Let us start with the third stage. Assume that
in stage 1, Clint and Dennis locate at the opposite ends of the beach, that is, Clint locates at the
left-hand corner of the beach and Dennis at the right-hand corner, one kilometer away from Clint.
Suppose that in stage 2, Clint sells his ice cream at price , while Dennis charges price . For
simplicity, assume that all consumers attach value to a scoop of ice cream where is large
(meaning that all consumers buy ice cream in equilibrium). Consumers face travel costs equal to
when they buy ice cream at the stand that is distance away from their location (in meters)
where > 0. In other words, if a consumer is located meters from the left-hand corner, her
total cost of buying an ice cream at Clints stand equals + and her total cost of buying an ice
cream at Dennis stand equals + (1000 ).
So, which consumers will go to Clint and which to Dennis? A straightforward starting point of
answering this question is to find the location of the marginal consumer, that is, the consumer
who is indifferent between the two. Recall that Clint and Dennis sell the exact same ice cream. So,
consumers will base their decision where to buy ice cream only on the costs of acquiring it. A
consumer located in is indifferent if the two firms do not differ in terms of the costs of ice
cream:
101

lOMoARcPSD|1733731
+ = + (1000 ).
Or, equivalently,

= 500 + 2
.
Now, we are ready to establish which consumers go to which ice cream vendor: All consumers
left of will go to Clint because they face lower costs to go there and higher costs to go to Dennis
than the marginal consumer. Indeed, demand for Clints ice cream is equal to :

( , ) = = 500 + .
2
Similarly, all consumers to the right of will go to Dennis. So, demand for Dennis ice cream
equals

( , ) = 1000 = 500 .
2
Check that if both firms charge the same price, both will serve 500 clients, which makes perfect
sense. Figure 9.1 summarizes the results so far.
Figure 9.1: Identifying demand in the Hotelling model
What prices will Clint and Dennis charge in stage 2? In equilibrium, both choose a price that
maximizes profits, given the price chosen by the other ice cream vendor. So, Clint maximizes

= ( , )( ) = (500 + ) ( ).
2
The first-order condition of this maximization problem is given by

= (500 + ) = 0.
2 2
102

lOMoARcPSD|1733731
Because Clint and Dennis are perfectly symmetric (they face the same marginal costs and both
are located at the extreme ends of the beach), we may assume that both will charge the same
price in equilibrium: = = . The above first-order condition can be rewritten as follows:

500 = 0,
2
which implies
= + 1000.
So, in equilibrium, the ice cream vendors charge a higher price the higher the travel costs of
their customers. In the extreme case of zero travel costs, both will charge a price equal to
marginal costs, which coincides with the Nash equilibrium price under Bertrand competition.
This result is not very surprising as in the case of zero travel costs, the ice cream at either shop
are perfectly homogeneous goods from the viewpoint of the customers. Moreover, in the case of
positive travel costs, this model offers a solution to the Bertrand paradox because equilibrium
prices exceed marginal costs so that both firms make positive profits in equilibrium.
Let us now examine the first stage of the Hotelling game. Where will Clint and Dennis locate? In
chapter 4, we concluded that both would locate in the middle if prices were fixed. Could that
happen in equilibrium when Clint and Dennis can adjust their prices after choosing location?
The answer is no: If both locate in the middle, from the viewpoint of consumers, both sell the
very same product. As a consequence, the situation becomes a Bertrand setting. Both will charge
a price equal to marginal costs so that neither of them will make a positive profit. Therefore, it is
better for Clint to move to another location if Dennis locates in the middle, for instance to one of
the outer edges. In that case, consumers observe ice cream at either ice cream stand as
differentiated products so that both Clint and Dennis will charge a price above marginal costs.
Indeed, Clint now makes a profit, in contrast to the case where he locates in the middle.
Therefore, locating in the middle is not a best response for Clint. We can conclude that in
equilibrium, both firms will locate in different spots. In other words, in equilibrium, there is at
least some product differentiation. Unfortunately, it is not easy to derive an equilibrium. To
solve stage 2, we should calculate the equilibrium price for all possible locations the two firms
might choose, which will often lead to asymmetric situations that are cumbersome to solve.
Under some assumptions, it can be shown that in equilibrium, both vendors will locate their
stand at the extreme ends of the beach. By doing so, they differentiate their products as much as
possible, allowing them to reap a handsome profit margin. In this chapters case study, we will
see how car producer General Motors successfully employed a strategy of product
differentiation in the early 20th century.
9.3 Case study: General Motors road to success

The US automobiles industry began around 1895, 10 years after the invention of the first
automobile with petrol engine built by Karl Benz in Mannheim, Germany. As Figure 9.2 depicts,
the number of companies in the US industry rose sharply in the early 1900s, peaking at 271 in
1909, then drastically decreasing during the following decade. The high exit rate favored the
expansion of the top two firms in the market, Ford Motor Company (Ford) and General Motors
(GM), which, by 1920, accounted for an aggregate share of output of over 60%. However, by
looking at the share of the single firms the picture is much different.
103

lOMoARcPSD|1733731
In the early 1900s, Ford revolutionized the automobile industry by introducing the moving
assembly line for the production of cars. To exploit the mass production that the invention
granted, the company sold only one type of car, the Model T, without giving its costumers any
options at all. These were the dictates of the companys founder, Henry Ford, who allegedly said
that his customers could have any color they liked, as long as it was black. In contrast, General
Motors was essentially a collection of car companies, including Cadillac, Buick, Oakland, Olds and
Chevrolet. The negative demand shock that hit the market at the beginning of 1920 triggered a
25% price cut of Fords Model T a reduction that GM could not match. While Ford was
expanding its production by benefitting from its economies of scale, GM was not able to produce
a car at the same price/quality ratio as its main competitor, and GMs various brands ended up
competing with each other. By 1921, Fords Model T held 55% of the US automobile market,
whereas General Motors only 11%.
Figure 9.2: Number of firms, entry, and exit, in the US automobiles market from 1885 to 1966
Source: Klepper (2002).
Facing a constant decline in market share, the new CEO of GM, Alfred Sloan, changed the strategy
of the company into a new and more aggressive one, in order to curb Fords widening dominant
position. Sloans idea was to differentiate GMs product spectrum to compete against Ford by
exploiting the substantial profit margins on higher quality cars to finance production of more
competitive lower quality models. Now, GM multi-brands turned out to be an important asset for
the company. The Cadillac division was designated to produce costly luxury cars, while
Chevrolet would compete against Ford by producing cars with an even lower price than Model T.
The other divisions of the company had each one segment of the market characterized by a
different quality, to appeal to consumers with diverse willingness to pay. The new strategy was
surely a dangerous bet. It needed new production facilities, new car designs, and a new company
structure. Without an appropriate coordination and sharing components among the divisions,
the costs would have risen exponentially, making it impossible to reach the efficient scale.
However, the bet paid out in both the short and the long run. General Motors was not forced out
of the market. Instead it acquired more and more power, reaching 45% of market share by the
end of 1940, while Ford saw its share shrinking to a mere 16%. Additionally, GMs product
104

lOMoARcPSD|1733731
differentiation strategy allowed the company to continue its product line expansion. In
subsequent years, General Motors entered several other markets, ranging from trucks to kitchen
appliances, and it is now one of the leading companies in the automobile market.
For a formal treatment of the equilibrium entry in markets with large fixed costs you can refer to
Bain (1951, 1956). His 1956 book is considered a classic handbook of industrial organization.
Entry and efficiency is also examined by Mankiw and Whinston (1986), who find that private
return exceeds social return, which results in excess entry.
Sutton (1991) investigates further the effect of market structure on entry and distinguishes
between endogenous and exogenous sunk costs in determining the efficient equilibrium. You
should also refer to the work of Bresnahan and Reiss (1991), Berry (1992), Mazzeo (2002ab)
and Seim (2006) for empirical applications and extensions.
The Hotelling model was developed by mathematician statistician Harold Hotelling. He

introduced the linear city model in his early work Stability in Competition in 1929.
The case study is based on Milgrom and Roberts (1992) textbook and Kleppers (2002) article
in the RAND Journal of Economics.
References
Bain, J.S. (1951). Relation of profit rate to industry concentration: American manufacturing
1936-1940, Quarterly Journal of Economics, 65(3), 293-324.
Bain, J.S. (1956). Barriers to New Competition: Their Character and Consequences in
Manufacturing Industries, Cambridge, MA: Harvard University Press.
Berry, S.T. (1992). Estimation of a model of entry in the airline industry, Econometrica, 60(4),
889-917.
Bresnahan, T.F. and P.C. Reiss (1991). Entry and competition in concentrated markets, Journal of
Hotelling, H. (1929). Stability in competition, Economic Journal, 39(153), 41-57.
Klepper, S. (2002). Firm survival and the evolution of oligopoly, RAND Journal of Economics,
33(1), 37-61.
Mankiw, N.G. and M.D. Whinston (1986). Free entry and social inefficiency, RAND Journal of
Economics, 17(1), 48-58.
Mazzeo, M.J. (2002a). Competitive outcomes in product-differentiated oligopoly, Review of
Economics and Statistics, 84(4), 716-728.
Mazzeo, M.J. (2002b). Product choice and oligopoly market structure, RAND Journal of Economics,
33(2), 221-242.
Hall.
Seim, K. (2006). An empirical model of firm entry with endogenous product-type choices, RAND
Sutton, J. (1991). Sunk Costs and Market Structure: Price Competition, Advertising and the
Evolution of Concentration, Cambridge, MA: MIT Press.
Exercises
105

lOMoARcPSD|1733731
Exercise 9.1 Bertrand competition revisited

The computer graphics chip industry is one with a small number of competitors that earn
normal economic profit. Two chip manufacturers, NVIDIA and ATI both face the prospect of low
profits, largely on account of each others existence. Consider the homogeneous-good Bertrand
model in which each firm has a positive fixed sunk cost and zero marginal cost.
a) What are the Bertrand equilibrium prices? What are the Bertrand equilibrium profits?
b) What happens to the homogeneous-good Bertrand equilibrium price if the number of firms
increases? Why?
c) Assume now that the fixed cost the firms need to incur at the initial stage is not sunk. How
many firms will enter in equilibrium?
Consider now a solution to that Bertrand paradox. Assume that NVIDIA and ATI produce
differentiated products and are still Bertrand competitors. The demand for NVIDIAs chip is:
= 1 2 + 3
Similarly, the demand for ATIs chip is given by:
= 1 2 + 3
The constants 1 , 2 , and 3 are (positive) coefficients of the demand function. Suppose each
manufacturers marginal cost is constant and equal to .
d) Without deriving the Nash equilibrium, argue that there exist values of 1 , 2 , and 3 for
which the duopoly will result in an equilibrium that sets a price that is higher than marginal cost.
e) Derive the Nash equilibrium. For which values of 1 , 2 , and 3 does the duopoly result in an
equilibrium price that is higher than marginal cost?
f) What happens to the equilibrium prices and quantities if 3 = 0?
Exercise 9.2 Bertrand duopoly with asymmetric costs

Suppose that identical duopoly firms have constant marginal costs. Firm 1 faces the demand
function:
1 = 50 21 + 2
and has a constant marginal cost of 30 per unit. Similarly, the demand Firm 2 faces is:
2 = 50 22 + 1
Firm 2s marginal cost is 10 per unit.
Find the firms reaction curves and solve for the Bertrand equilibrium. What can you conclude?
Relate your answer to the discussion about the Bertrand paradox.
Exercise 9.3 Market Entry

In developing his model of imperfect competition, the French mathematician Antoine Augustine
Cournot used the example of a mineral spring, from which the owner can sell the water. Each
spring owner, which can be viewed as an identical firm, has a large supply of healthy water and
106

lOMoARcPSD|1733731
faces the problem of how much to provide the market. A firms cost of pumping, purifying, and
bottling liters of water is given by the cost function:
() = 10
Market demand () for mineral water is:
= 150
a) Determine the equilibrium individual firm output, total market output, and market price
under Cournot competition. What is the total profit for all spring owners?
Finding the natural spring and acquiring the necessary bottling facilities entails an initial fixed
cost to be incurred by each firm in the entry stage. Suppose there is one period of Cournot
competition after entry.
b) Determine the long-run equilibrium number of firms (). Which condition is required for
there to be any entry?
Consider now the social planners problem. Suppose the social planner can choose the number of
firms by restricting entry through licenses or promoting entry through subsidizing the entry
cost , but cannot regulate prices or other competitive conduct of the firms once in the market.
c) Find the number of firms () that the social planner would choose if she wishes to maximize
total welfare. Which condition is required for there to be any entry?
d) Suppose = 1150. Is the long-run equilibrium efficient?
e) If the social planner could set both the number of firms and the price in this case, what
choices would she make?
Exercise 9.4 Hotellings beach

Two ice cream stands are located along a small beach with a length of 100 meters. Both stands
sell identical ice cream cones. Assume that all production costs are sunk. The ice cream vendors
locate 1 and 2 meters from the left-hand corner of the beach, respectively. 100 sunbathers are
located uniformly along the beach, one in every meter. Carrying ice cream distance back to
ones beach umbrella costs 2 because ice cream melts more the higher the temperature and
the further one must walk. Assume that firms engage in Bertrand competition and choose their
prices simultaneously.
a) Assume that 1 = 40 and 2 = 70. Find the demand of each ice cream stand.
b) Let = 0.001. What price will each firm charge in the Nash equilibrium? How many ice
cream cones are demanded at those prices? Is this equilibrium efficient?
c) Determine the equilibrium profits. How would these profits change if transportation costs
doubled?
d) What happens to prices and profits if transportation costs go down to zero? If the ice cream
stands locate in the same spot? What if they locate at the opposite ends of the beach?
107

lOMoARcPSD|1733731
Part IV: Repeated interaction

Up till now, we have assumed that decision makers (workers, firms, consumers) play a game
only once. However, in practice, it is often unrealistic to assume that decision makers meet each
other only once. Children interact with their parents day after day. Workers and their bosses are
faced with similar choices year after year. Firms choose their price anew every year, month,
week, day, hour, or even minute.
In this part, we will consider the possibility that decision makers play the same game several
times in a row. In other words, they interact in a repeated game. When decision makers interact
repeatedly, they share a common history in later stages of the game. This allows them to
condition their behavior based on what other decision makers did in the past. An important
consequence of this is that decision makers may manage to establish attractive outcomes that
are not feasible in the case where they interact only once. More in particular, a player may have
a reason to cooperate because she believes that the other decision makers will be nice in the
future if she cooperates while they will punish her if she does not. In chapter 10, we will
develop a framework for the analysis of repeated games, deriving conditions under which
decision makers can sustain cooperation. In chapter 11, we will apply this framework to
repeated interaction in the Principal-Agent model. In chapter 12, we will analyze how firms can
collude on markets when they interact repeatedly.
108

lOMoARcPSD|1733731
Chapter 10: Repeated interaction and the value of revenge

When games are repeated, decision makers may obtain opportunities to establish better
outcomes for themselves than in the case where they interact only once. The reason is that they
can punish non-cooperative behavior in future interactions. We start in section 10.1 by showing
this result in a game that is repeated only once. In section 10.2, we extend the analysis to
infinitely repeated games and we will establish a condition under which cooperation is stable. In
section 10.3, we will study this condition in more detail. Finally, section 10.4 includes a case
study related to an auction of licenses for mobile telecommunications where the bidders
managed to coordinate on low prices using punishment strategies in the auction itself.
10.1 A simple example

Let us see how this might work in the following game. Suppose Adle and Bono are cheese
sellers at a market. Each can charge a high, medium, or low price for cheese. The resulting
payoffs are given in table 10.1. Observe that a low price yields relatively low profits. The reason
is if only one of the two sellers chooses a low price, all consumers will buy at the seller with the
low price. Both will obtain zero profits because the seller with the higher price does not sell any
cheese, while the other sells at a price equal to marginal costs. If both do so, both will still make
some profit because they will only serve half the market and they can sell the remaining cheese
at another market for a higher price.
Table 10.1: The market game
Bono
High Medium Low
High 4,4 2,5 0,0
Adle Medium 5,2 3,3 0,0
Low 0,0 0,0 1,1
Note that the resulting game extends the prisoners dilemma game we discussed in chapter 4 by
adding a third action to both players action set (Low). At first sight, this additional action
seems irrelevant: (Medium, Medium) is a Nash equilibrium in which both Adle and Bono obtain
payoff 3, so why would they even think of choosing a low price which gives them 1 at best? Well,
strictly speaking, (Low, Low) is a Nash equilibrium: For both players, Low is a best response to
Low. More importantly, when the players play this game several times in a row, they can use the
low price as a threat when the other player does not cooperate.
Suppose that Adle and Bono interact in two consecutive periods, period 1 and period 2. In each
period, they play the extended prisoners dilemma. Observe that both are better off if they both
charge a high price than if they both charge a medium price. In fact, they can establish a high
price in equilibrium in period 1 despite the fact that (High, High) does not constitute a Nash
equilibrium in the case where the players interact only once. More in particular, the strategies in
table 10.2 form a subgame perfect Nash equilibrium.
Why do those strategies constitute a subgame perfect Nash equilibrium? The logic of backward
induction requires us to look at the last stage of the game first, which is period 2 in this game. In
the second period, the players will choose either (Medium, Medium) or (Low, Low), depending
on what happened in period 1. Both outcomes are indeed a Nash equilibrium. In the first stage, if
both Adle and Bono stick to the above strategy, Adle obtains 4 in the first stage and 3 in the
109

lOMoARcPSD|1733731
second, so 7 in total. Adle may deviate in the first period by choosing a medium price instead of
a high price. This will yield her 5 in period 1 but only 1 in the next because according to Bonos
strategy, he will choose a low price in period 2 in response to Adle choosing Medium in the first.
As a consequence, Adles total payoff will be 6 rather than 7. Indeed, she has a good reason to
stick to the above strategy. The same holds true for Bono, so that the strategies constitute a
subgame perfect Nash equilibrium.
Table 10.2: A subgame perfect Nash equilibrium of the repeated market game
Action in period Action in period 2 if both Action in period 2 if at least one

1 choose High in period 1 player did not choose High in period
1
Adle High Medium Low
Bono High Medium Low
From this exercise, we learn that if games are repeated, decision makers can reach attractive
outcomes that they cannot reach if they play the game only once. The intuition is that they can
punish other decision makers misbehavior by choosing unattractive Nash equilibrium actions in
future periods. In other words, in the case of repeated interaction, decision makers can establish
cooperation in equilibrium if they have punishment strategies at their disposal which they can
play in case other decision makers do not cooperate. Observe that in the game above, adding the
action Low to the prisoners dilemma game gives the players a stick they can use to achieve a
better outcome for themselves. Without it, the only subgame perfect Nash equilibrium would be
for both players to choose Medium in both periods. Including the threat option Low increases
the equilibrium payoff for both Adle and Bono from 6 (twice the payoff when both choose
Medium) to 7 (a payoff of 4 in period 1 plus a payoff of 3 in period 2). Therefore, the value of
revenge equals 7 6 = 1.
Let me conclude this section with two cautionary remarks. First, we assumed that decision
makers do not discount earnings over periods: For both Adle and Bono we assumed that a euro
earned in period 2 has the same value as a euro earned in period 1. This assumption need not be
realistic. Adle and Bono may be impatient in the sense that they attach more value to their
earnings today than to their earnings tomorrow, next week, or next year. Suppose that Adle
values a euro she earns in period 2 as being worth euro in period 1 (where 0 < < 1). If this
is the case, she has an incentive to choose a medium price in period 1 instead of a high price if
5 + > 4 + 3
or, equivalently,
1
< .
2
In other words, if Adle is impatient (i.e., if is small), the above strategies do not form an
equilibrium. Adle does not attach sufficient weight to Bonos threat of choosing a low price in
the future if she does not cooperate today. A similar result arises if it is not sure whether there is
a second period in the first place, for instance, because the market is closed unexpectedly due to
110

lOMoARcPSD|1733731
bad weather. Indeed, Adle and Bono will not cooperate if the probability that they compete in
the next period is less than one half even if they are patient. You can see this by interpreting as
the probability that the game continues.
A second cautionary remark relates to the credibility of the players punishment strategy. You
could argue that choosing a low price is a credible threat for both Adle and Bono because it is
part of a Nash equilibrium in the one-shot game. However, this Nash equilibrium itself is not
very attractive for either player: They are better off if both of them choose Medium. In fact, if
Adle and Bono talk to each other in period 2, they will soon agree that starting a price war is in
nobodys interest, even when one of them deviated in period 1 by charging a medium price
instead of a high price. They are likely to convince each other to choose Medium in period 2.
However, if Adle and Bono realize in period 1 that they will renegotiate in period 2, neither of
them has a reason to stick to a high price in period 1 because their misbehavior will not be
punished. Indeed, cooperation can break down if a punishment strategy is not credible, for
example, because players have a strong incentive to renegotiate. In part VI, we will discuss the
credibility of future actions in more detail.
10.2 Infinitely repeated games and the trigger strategy

In the previous section, we saw that decision makers can cooperate in early periods if they can
use threat strategies in later periods. Let us return to the prisoners dilemma from chapter 4
(see table 10.3) as a stepping stone to a more general framework of analyzing cooperative
behavior in the case of repeated interaction. Suppose the prisoners play this game twice in a row.
We have already seen that (Deny, Deny) is not an equilibrium outcome in the first period. The
reason is that Confess cannot be used as a threat, because this strategy is optimal for either
prisoner in the last period of the game regardless of whether the other prisoner cooperated or
not. So, a prisoner cannot use it as a stick to punish misbehavior by the other prisoner.
Table 10.3: The prisoners dilemma
Prisoner 2
Deny Confess
Deny 4,4 2,5
Prisoner 1
Confess 5,2 3,3
However, if the prisoners play this game an infinite number of times, they can still reach the
favorable outcome (Deny, Deny) in equilibrium. How might this be possible? The answer is that
the prisoners can use a trigger strategy. This strategy tells a player to cooperate as long as the
other player cooperates in all previous periods and punish the other player as much as possible
in all future periods as soon as she does not cooperate in one period. Table 10.4 presents the
trigger strategy according to which a player cooperates up to the point that one of the other
players sets off a trigger by not cooperating. As soon as another player does not cooperate, the
player punishes the deviating player(s) as much as possible in all future periods. A trigger
strategy constitutes a subgame perfect Nash equilibrium if none of the players has an incentive
to deviate from it. Each player makes the following trade-off in each period: He can either stick
to the cooperative action or he can deviate and obtain a higher payoff now but a lower payoff in
all future periods because of the other player(s) punishing.
111

lOMoARcPSD|1733731
Suppose that the very same game is repeated infinitely many times and that a player discounts
each period with discount factor < 1 relative to the previous period. There are at least two
reasons why players may discount earnings in future periods. First of all, a euro earned today is
worth more than a euro earned next year because you can put a euro in a bank account and
obtain some interest on it. If the yearly interest rate equals , an euro earned today is worth 1 +
next year. Or, the other way around, a euro earned tomorrow is equivalent to
1
=
1+
euro today. Alternatively, the interaction between players only continues in the next period with
probability < 1 so that a player values a euro in next periods interaction (ignoring
discounting) as
= .
If a player discounts next periods payoffs, the prospect of earning one euro in the next period,
given that the next period takes place with probability , is worth

=
1+
in terms of todays money.
Table 10.4: The trigger strategy
Period Action
1 Choose the cooperative action
Choose the cooperative action when the other players cooperated in all previous
periods
= 2,3,
Otherwise, choose the one-shot Nash equilibrium action that harms the deviating
player the most
It can be shown the trigger strategy is a subgame perfect Nash equilibrium (only) if for all
players it holds true that
where denotes the players payoff if all players cooperate, represents her payoff if she
deviates from the cooperative outcome while all other players stick to it, and stands for her
payoff in her least preferred one-shot Nash equilibrium. We call this condition on the trigger
strategy stability condition.
Where does the trigger strategy stability condition come from? By definition, strategies
constitute a (subgame perfect) Nash equilibrium if none of the players has an incentive to
deviate from it. If a player sticks to the trigger strategy and all other players do so as well, she
112

lOMoARcPSD|1733731
obtains the cooperative payoff in all periods. The corresponding net present value of todays and
all future periods payoff is equal to

= + + 2 + 3 + = .
1
Now, if the player deviates in some period, while all other players stick to their strategy, she
obtains in the current period and her punishment payoff in all future periods. As a
consequence, the net present value from deviating equals

= + + 2 + 3 + = + .
1
The trigger strategies form a subgame perfect Nash equilibrium if for all players

+
1 1
(1 ) +
( )
which is precisely the trigger strategy stability condition we intended to derive.
Let me illustrate the trigger strategy stability condition using the prisoners dilemma from table
10.3. The cooperative outcome is that neither prisoner testifies so that = 4. Confess is the
most attractive deviation action (in fact, it is the only deviation action), resulting in deviation
payoff = 5. Finally, the payoff in the only Nash equilibrium equals = 3. We plug those
numbers into the trigger strategy stability condition to find:
5 4 1
= = .
5 3 2
10.3 Analyzing the stability condition

The trigger strategy stability condition shows that for all players the discount factor should
exceed the minimum level

= .

We will refer to as the critical discount factor. As figure 10.1 shows, cooperation is only
feasible if the discount factor exceeds the critical discount factor.
Figure 10.1 Feasibility of cooperation
113

lOMoARcPSD|1733731
Observe that the critical discount factor is decreasing in , a players payoff in the case of
cooperation. In other words, the higher a players cooperative payoff, the lower and, in turn,
the greater the range of discount factors for which cooperation is feasible. This makes a lot of
sense: The more a player earns if she cooperates, the stronger her incentives are to cooperate.
Likewise, cooperation is more likely the lower a players punishment payoff is. Again, this
makes sense: Cooperation becomes more attractive for a player the harsher others can punish
her if she does not cooperate. To consider the effect of the deviation payoff , let us rewrite the
critical discount factor as follows:

= = 1 .

From this expression, it follows that is increasing in : The higher the deviation payoff, the
more interesting it becomes to deviate from the cooperative outcome and the smaller the range
of s for which cooperation is feasible.
10.4 Case study: Within-game trigger strategies in spectrum auctions

In this chapter you have studied how players can exploit trigger strategies in repeated games to
coordinate their actions in order to reach an equilibrium that is more attractive for them than
the one-shot Nash equilibrium. In some cases, players may even use a trigger strategy to
coordinate on an attractive outcome in a one-shot game. Peter Cramton and Jesse Schwartz,
auction specialists from the University of Maryland, provide an example of an auction in which
bidders managed to do so.
More in particular, they studied how telecommunication firms coordinated their bids in a
spectrum auction held by the United States Federal Communications Commission (FCC) in
1996-1997. In this auction, the FCC sold rights to use radio spectrum for mobile
telecommunication services. The FCC bundled the rights in licenses. The FCC auctioned three
licenses for each of 493 areas of the US. A license allowed the winning bidder to use a specific
frequency band to provide wireless communication services to customers in a particular area.
The FCC opted for a simultaneous ascending auction to sell the licenses. Essentially, this auction
runs over an unspecified number of rounds in which the bidders are able to bid on as many of
the licenses as wanted. At the end of each round, the auctioneer reports the amount of each bid
on each license, along with the bidder who placed the bid. If a license receives a higher bid in a
later round, the corresponding bidder becomes the new standing high bidder. The auction
continues until a round passes with no new bids, that is, until no bidder raises the bid on any
license. Then, licenses are awarded to the highest bidders, who pay their final bids.
Clearly, bidders have a strong incentive to keep the license prices low. They may coordinate on
low prices by splitting up the licenses among them and abstaining from bidding on each others
licenses. However, explicit agreements about dividing the market in such a way are illegal
according to competition law. So, how can bidders divide the market without an explicit
agreement? Some companies found a smart way to coordinate their bids legally.
Table 10.5 shows all the bids for the license in Canton and Harrisburg after round 56. At first,
there was a lot of competition on the license in Canton, where NextWave and NorthCoast
repeatedly overbid each other. On the contrary, Harrisburgs license seemed solidly in
NextWaves hands, until round 161 when NorthCoast submitted a higher bid just after
114

lOMoARcPSD|1733731
NextWaves last bid on Canton. NorthCoasts intention was clear: it wanted to signal to
NextWave to stop topping its bids on Canton, otherwise it would continue bidding on
Harrisburgs license. The message was received and understood; in the subsequent two rounds
NextWave bid and obtained Harrisburgs license whereas NorthCoast obtained Cantons.
Table 10.5: Bids in dollars on Canton and Harrisburg
Canton, OH, 65 F Harrisburg, PA. 181 F

Round NextWave NorthCoast OPCSE NextWave NorthCoast
56 358,000 1,217,000
57 409,011
78 460,000
82 511,011
125 562,000
136 618,011
158 680,000
159 748,011
160 861,000
161 1,339,011
162 1,473,000
163 947,011
Table 10.6: Bids in dollars on Marshalltown, Rochester, and Waterloo
Marchalltown, IA, 283 E Rochester, MN, 378 D Waterloo, IA, 452 E

Round McLeod USWest McLeod USWest AT&T McLeod USWest
24 56,000 287,000
46 568,000
52 689,000
55 723,000
58 795,000
59 875,000 313,378
60 345,000
62 963,000
64 62,378 1,059,000
65 69,000
68 371,000
A more subtle signaling scheme was used by USWest. Table 10.6 shows the bids on the licenses
for Marshalltown, Rochester, and Waterloo. From round 24 onward, McLeod was the highest
bidder in both Marshalltown and Waterloo, whereas for Rochesters license the company faced
intense competition by USWest. In round 59, USWest bid simultaneously on Rochester and
Waterloo, ousting McLeod from both. USWest tried to use its bid on Waterloo to send a signal to
McLeod. Do you see what USWest wanted to signal and how? Well, USWests bids on Waterloo
115

lOMoARcPSD|1733731
and Marshalltown ended with 378, the number which represented the code of Rochesters area.
USWest wanted to signal McLeod to stop bidding on Rochester. The first attempt failed though:
McLeod outbid USWest again on Rochester in round 62. In a second attempt to get the message
through, USWest bid on Marshalltown letting the bid end with 378. Now, McLeod got the
message. It reacted by overbidding USWest in Marshalltown but not in Rochester.
Let me stress that this type of communication is not illegal despite the harm it may represent for
the auctioneer (and for taxpayers, when the auctioneer is the government). Rather, they are the
result of poor auction rules. Nowadays, the FCC applies better rules. For example, bidders can no
longer submit bids in small denominations so that USWests signaling strategy is no longer
feasible. Nevertheless, auction designers should be aware that companies are always ready to
find new and more creative ways to coordinate on low prices in auctions.
In his 1971 paper in the Review of Economic Studies, James W. Friedman proves the folk theorem
which states that infinitely repeated games have subgame perfect Nash equilibria in which all
players are better off than in the one-shot Nash equilibrium provided that the discount factor is
sufficiently large. The trigger strategy that we studied in this chapter is commonly referred to as
grim trigger. Friedman used this trigger strategy to prove the folk theorem. Rubinstein (1976)
and Fudenberg and Maskin (1986) analyze more general versions of the folk theorem.
Political scientist Robert Axelrod (1984) organized a famous experiment to study behavior in
repeated games. He invited researchers to submit strategies for a repeated prisoners dilemma
game. He let the strategies compete against each other and the strategy that obtained the highest
overall payoff won. The tit-for-tat strategy, submitted by Anatol Rapoport, was the winner. The
tit-for-tat strategy is similar to the grim trigger strategy. The difference is that the tit-for-tat
strategy is forgiving in the sense that a player returns to cooperate if the opponent chooses a
cooperative action after having deviated in an earlier stage. A survey of experimental evidence
on the prisoners dilemma game can be found in Cooper, DeJong, Forsythe, and Ross (1996).
The case study of this chapter is based on Cramton and Schwartz paper published in 2002 in the
B. E. Journal of Economic Analysis.
References
Axelrod, R. (1984). The Evolution of Cooperation, New York: Basic Books.
Cooper, R., D. DeJong, R. Forsythe, and T. Ross (1996). Cooperation without reputation:
experimental evidence from prisoner's dilemma games, Games and Economic Behavior, 12(2),
187-218.
Cramton, P. and J.A. Schwartz (2002). Collusive bidding in the FCC spectrum auctions, B.E.
Journal of Economic Analysis and Policy, 1, Article 11.
Friedman, J.W. (1971). A non-cooperative equilibrium for supergames, Review of Economic
Studies, 38(1), 1-12.
Fudenberg, D. and E. Maskin (1986). The folk theorem in repeated games with discounting or
with incomplete information, Econometrica, 54, 533-554.
Rubinstein, A. (1979). Equilibrium in supergames with the overtaking criterion, Journal of
Economic Theory, 21(1), 1-9.
116

lOMoARcPSD|1733731
Exercises
Exercise 10.1 Two-period game without discounting

Consider the following two persons 3 x 3 game.
Bono
X Y Z
A 5, 5 8, 4 0, 0
Adle B 4, 8 7, 7 1, 9
C 0, 0 9, 1 0, 0
a) Find all Nash equilibria of this game.

b) Consider the twoperiod repeated game in which the above stage game will be played twice.
Suppose the payoffs are simply the sum of the payoffs in each stage game. Then, is there a
subgame perfect Nash equilibrium that can achieve (B, Y) in the first period? If so, describe the
equilibrium. If not, explain why.
Exercise 10.2 Finitely repeated game with discounting

Gotye and Lana are soft drink sellers on a beach during a three-day windsurfing event. Each of
them can charge a high, medium, or low price for their soft drinks. They repeat the game shown
in the following table three times, discounting their payoffs for the second and third day of the
event to the present using the same discount factor . In answering the first three questions you
may assume that the players discount factor is significantly high (close to one).
Gotye
High Medium Low
High 10, 10 -1, -12 -1, 15
Lana Medium -12, -1 8, 8 -1, -1
Low 15, -1 8, -1 0, 0
a) Is there a subgame perfect Nash equilibrium in which both players play High in all three
periods? Why (not)?
b) Describe a subgame perfect equilibrium in which both players play High in the first two
periods.
c) How would the equilibrium described in (b) change:
In a twice-repeated game?
In an times repeated game ( > 3)?
d) What is the critical discount rate for which the equilibrium is stable in the 3-period game?
Exercise 10.3 Infinitely repeated game

Justin has a company that makes iced tea. His only customer is Jennifer. Justin has to decide
whether to make his tea good or bad. Good tea is more expensive to make. Jennifer has to decide
whether to buy one or two bottles. All the bottles in a given production run are of the same
quality. Jennifer cannot tell the quality of the tea when she decides how much to buy, but she
117

lOMoARcPSD|1733731
does discover the quality later once she drinks it.

Jennifer's payoff is 3 if she buys two bottles of tea and it is good, 2 if she buys one bottle and it is
good, 1 if she buys one bottle and it is bad and 0 if she buys two bottles and it is bad. Justin's
payoff is 3 if he makes bad tea and sells two bottles, 2 if he makes good tea and sells two bottles,
1 if he makes bad tea and sells one bottle and 0 if he makes good tea and sells one bottle.
a) Write down the payoff matrix for this game and find the Nash equilibrium.
Now suppose that Justin and Jennifer have an on-going business relationship. That is, in each
period, Justin has to choose the tea quality for that period; Jennifer has to choose the quantity
she wants to purchase in that period. Let 1 be Justins discount factor, and let 2 be Jennifers
discount factor.
b) Determine the subgame perfect equilibrium of the two-stage repeated game. For simplicity,
assume that 1 = 2 = 1.
Now consider the case where the game is infinitely repeated, with 0 < 1 , 2 < 1.
c) Prove that there exists a subgame perfect equilibrium of this game in which Justin makes good
tea in each period and Jennifer buys two bottles in each period. Specify the minimum discount
factors such that this equilibrium is stable.
Exercise 10.4 Trigger strategies and sabotage

A manager (M) and a worker (W) repeat the following game infinitely many times:
Worker
Work Shirk
Bonus 10, 10 0, 16
Manager
No Bonus 15x, 0 8, 9
In this game, the worker chooses between working and shirking. The manager can award a
bonus or not. The worker sabotages the manager if he does not obtain a bonus when he worked.
The sabotage decreases the managers original payoff by 0. The manager and the worker
discount future payoffs using the same discount rate.
a) Suppose that = 0. Assume that W picks Work only if M has given him a bonus in all
previous periods. In other words, as soon as M chooses No Bonus in a period, W will play
Shirk forever. For which values of the discount factor will M award a bonus?
b) Suppose that = 0 and assume that both M and W consider playing a trigger strategy. Does
the worker have a greater incentive to deviate from (Work, Bonus) than the manager? (Hint:
answer this question using the critical discount rate.)
c) Can M benefit from offering W the possibility to sabotage him? Answer this question by
analyzing how the critical discount factor changes when increases.
Now, suppose that if W chooses Shirk, he leaves the firm. As a consequence of this, he and the
manager obtain a utility of zero in all future periods and not the payoffs shown in the table
above.
d) For which values of is (Work, Bonus) an equilibrium strategy in all periods if both
players play a trigger strategy? Answer this question for all positive values of .
118

lOMoARcPSD|1733731
Chapter 11: Relational contracts

In this chapter, we apply the analysis of repeated games to the Principal-Agent model. In section
11.1, I will introduce you to the concept of a relational contract and show how a principal and an
agent may enforce such contracts if they interact repeatedly despite the contract not being
enforceable in court. In section 11.2, we will discuss Bulls more general model of relational
contracts. Bull shows that only moderate bonuses may be feasible in equilibrium because it is
too tempting for the principal not to pay the agent a big bonus in the case of good performance.
Section 11.3 describes the results of two experiments that identify another reason why big
bonuses may not work in practice: The agent may feel too much pressure to perform well.
11.1 What are relational contracts?

In chapter 7, we discussed the simple Principal-Agent game presented in figure 11.1. We
concluded that in the subgame perfect Nash equilibrium of this game, the principal (P) pays no
bonus which induces the agent (A) to shirk. The outcome of this game is that both the principal
and the agent obtain zero utility, while if the agent worked and the principal gave a bonus, both
would be better off because each would get a payoff of 1. Implicitly, we assumed that the
principal and the agent cannot write a binding contract which states that the principal pays the
bonus if the agent works. More precisely, we assumed that the principal and the agent cannot
sign a formal or explicit contract, i.e., a contract that can be enforced in court. Contracts may not
be enforceable by the court because the court cannot verify whether the agent worked or
whether the principal benefited from the agents output.
We will call agreements between two parties that cannot be enforced by the court implicit
contracts. An implicit contract can be an agreement of any kind: written, verbal or even tacit.
We still speak of a contract because the agreement is self-enforcing in that both parties have a
reason to stick to it, even without any external enforcement. A relational contract is a special
case of an implicit contract. We speak about relational contracts if an implicit contract involves
agreements that are enforceable because parties interact repeatedly, i.e., they have a long-term
relationship with each other. As we saw in the previous chapter, when parties play a trigger
strategy in the case of repeated interaction they may reach outcomes that would not be feasible
if they met only once. Relational contracts are relevant for interactions within organizations.
Workers usually interact with the same colleagues day after day. Everyone can understand that
if they do not cooperate today they can expect their colleagues not to cooperate tomorrow.
In the simple Principal-Agent game in figure 11.1, how could the principal and the agent enforce
an implicit contract that states that the agent works and that the principal pays a bonus if the
119

lOMoARcPSD|1733731
agent works? A potential answer is that the agent plays the trigger strategy described in table
11.1. If the principal is sufficiently patient, her best response is to always award the bonus if the
agent works hard. (Why do we not have to assume the agent to be patient?) Suppose that the
principal discounts next periods payoffs with discount factor . If the principal pays the bonus,
she obtains 1 now and in all periods in the future because the agent will work in all future
periods. If the principal does not pay the bonus, he obtains 3 in the current period, but zero in all
coming periods because according to the agents trigger strategy, the agent will shirk forever.
The principal prefers paying the bonus (only) if
1 2
1 + 1 + 1 2 + 1 3 + = 3 .
1 3
Table 11.1: The agents trigger strategy
Period Action
1 Choose Work
Choose Work when the principal awarded the bonus in all previous periods
= 2,3, Otherwise, choose Shirk
We conclude that the principal and the agent cooperate in equilibrium if the principal is
sufficiently patient. This conclusion has several interesting implications for the interaction
between managers and workers. In the above game, the principal could be seen as the
management of a firm and the agent as one of the firms workers. There may be several reasons
why management is patient (so that the worker will work hard) or impatient (so that the worker
will shirk). For example, if the firm is active in a growing market, it is likely to continue its
activities in the far future, so that it is likely that the firms management follows a long-run
policy corresponding with a high . The other way around, if a firm is close to bankruptcy,
management will focus on todays profits rather than the profits the firm may generate in the
future, if it still exists. Another important factor is the age of the worker. Workers close to
pension age have only a short future with the firm so that management will not attach much
weight to those workers potential future earnings. In the terminology of our model, the in the
relationship between management and older workers is probably low so that it is unlikely that
beneficial cooperation will arise. This may be one reason why it is relatively difficult for older
workers to find a new job even if they have much more experience than younger workers.
11.2 Bulls model

While the simple Principal-Agent model provides valuable insights, it is also unrealistic in
several dimensions. First of all, it assumes that the principal can observe whether the agent
works or shirks. This assumption is often unrealistic: The government cannot observe how
much effort the dean of a university faculty expends in motiving his professors. The dean, in turn,
cannot verify how well a professor prepares her lectures. For the professor, it is difficult to tell
how hard students study for the exams of the courses she teaches. Of course, in many cases, the
agents output is observable for the principal. The government observes the number of
graduates from the faculty, the dean obtains course evaluations, and the professor can evaluate a
students performance on the exam.
120

lOMoARcPSD|1733731
A second assumption is that the principal always benefits from the agents effort. Although in
most cases, the agents effort and his output are closely related, the relation need not be perfect.
The university may perform poorly in terms of the time it takes the average student to graduate
because it tends to attract lazy students. The professor may obtain poor course evaluations not
because she spends little time preparing her lectures but because the students found the exam
more difficult than she anticipated. A student may fail the exam despite having studied very hard
because she felt ill on the day of the exam. In other words, in many cases, the agents output is an
imperfect measure of his effort.
New York university professor Clive Bull relaxes both assumptions in his model of relational
contracts. Let me illustrate Bulls model using a modified version of the Principal-Agent model
we discussed in chapter 2, in which (risk-neutral) agent Antonio directs a movie on behalf of
(risk-neutral) principal Penlope. For simplicity, we assume that Antonios movie project can
have only one of two outcomes: It is either a blockbuster or a flop. In the case that the movie
flops, Penlope breaks even, i.e., she obtains zero profits. If it is a blockbuster, her net value for
the movie equals (0 < < 1). Antonio can exert effort at cost
() = 2
where > 0. The more effort Antonio expends the more likely it is that the movie becomes a
blockbuster. Let us assume for simplicity that the probability of a blockbuster equals
= .
We assume that 0 1 so that indeed represents a probability.
Suppose that Penlope and Antonio interact in the following four-stage game:
1. Penlope offers Antonio a contract that pays Antonio if his movie is a flop and + if
it is a blockbuster, where represents Antonios bonus.
2. Antonio accepts or rejects the contract. If Antonio rejects the contract, he does not direct
the movie and the game ends. Antonio obtains utility > 0 and Penlope obtains
zero profits.
3. Antonio chooses effort .
4. If the movie is a blockbuster, Penlope decides whether or not to pay Antonio the bonus
specified in Antonios contract.
If Penlope and Antonio play this game only once, we can find the subgame perfect Nash
equilibrium using backward induction. In stage 4, Penlope does not pay the bonus (why would
she?). As a consequence, the best thing Antonio can do in stage 3 is choose as low effort as
possible because effort is costly and has no impact on the possibility of obtaining a bonus. In
stage 2, Antonio accepts the contract (only) if . Finally, Penlope offers Antonio a
contract he will refuse (e.g., by proposing a fixed wage < ) because her net profits if
Antonio accepts the contract are negative. So, the outcome is that the movie will not be produced,
despite the possibility that this would be beneficial for both Penlope and Antonio.
Bull shows that the outcome may be much better for both parties if they repeatedly engage in
similar interactions, i.e., if they produce several movies consecutively. To see how this works,
suppose that in the case of repeated interaction, Antonio accepts the contract and Penlope pays
the bonus. Antonios expected utility equals
121

lOMoARcPSD|1733731
= + () = + 2 .
You can verify that Antonios optimal effort is given by
= /
so that his expected utility equals
= + 2 = + 2 /2.
In stage 2, Antonio will accept Penlopes contract (only) if
2
=+ .
2
Now, we turn to stage 1. If Antonio accepts the contract, Penlopes expected profits are equal to
() = = .
Penlope will fix Antonios base wage at such a level that Antonio will just accept the contract,
that is,
2
= .
2
When we substitute Antonios effort = /, = , and the optimal wage = 2 /2,

we obtain that Penlopes expected profits conditional on the bonus are equal to
{()} = ( )
( ) 2
= +
2
(2 )
= .
2
Observe that Penlopes expected profit () is an inverse U-shaped function of . Figure 11.2
depicts {()}.
You can verify that the optimal bonus equals
= .
Note that Antonio obtains a bonus equal to the value Penlope assigns to the movie becoming a
blockbuster. In other words, in the optimal contract, Antonio becomes the residual claimant of
the fruits of his effort, consistent with what we concluded in earlier chapters on optimal
incentive contracts. Similarly, Antonios bonus equals = , Penlope and Antonio generate the
value-maximizing outcome.
122

lOMoARcPSD|1733731
Figure 11.2: Graphical representation of Bulls model
Under what conditions can Penlope and Antonio sustain value-maximizing interaction in a
subgame perfect Nash equilibrium if they interact repeatedly? To answer this question, let us
assume that Antonio plays the trigger strategy described in table 11.2.
Table 11.2: Antonios trigger strategy
Period Action
1 Accept the contract and choose the optimal effort given bonus
= 2,3, Accept the contract and choose the optimal effort given bonus when Penlope
awarded the bonus in all previous cases in which the movie was a blockbuster
Otherwise, reject the contract
Penlope has to be sufficiently patient for Antonios trigger strategy to be part of a subgame
perfect Nash equilibrium. In stage 4, Penlope will weigh the benefits from paying the bonus (i.e.,
future payoffs from Antonio remaining productive) against the costs (the bonus itself). So, she
will pay the bonus (only) if

{()} + {()} 2 + {()} 3 + = {()} ,
1
which is equivalent to the following trigger strategy stability condition:
1
{()} ( 1) =

where denotes Penlopes interest rate. (We assume here that the interaction will continue
1
forever and that Penlope discounts future profits using a discount factor = 1+.) Figure 11.2
123

lOMoARcPSD|1733731
depicts both {()} and as a function of . From the trigger strategy stability condition, we
learn that Penlope pays the bonus (only) if {()} exceeds .
Figure 11.2 depicts three possibilities for Penlopes interest rate: low ( ), medium ( ), and
high ( ). If Penlope is patient, her corresponding interest rate is low. Consistent with what we
have seen so far, cooperation is feasible because Penlope is sufficiently patient: There is a
substantial range of for which the trigger strategy stability condition is satisfied, i.e., for which
{()} . Indeed, the optimal bonus = lies in this range so that Penlope and Antonio
can reach a value-maximizing outcome in equilibrium. In contrast, if Penlope is impatient, she
has a high interest rate, and figure 11.2 shows that the trigger strategy stability condition
{()} is never satisfied. For intermediates levels of impatience, corresponding to a
medium interest rate, figure 11.2 indicates that the trigger strategy stability condition is
satisfied for some s, but not for = . In this case, the best feasible bonus satisfying the trigger
strategy stability condition is . So, cooperation between Penlope and Antonio is feasible, but
not at the highest possible level. The reason is that the optimal bonus = is so high that it
becomes attractive for Penlope not to pay it if the movie turns out to be a blockbuster. Antonio
will understand this so he will not accept the contract: it is too good to be true.
11.3 Case study: Large Stakes and Big Mistakes Why big bonuses dont
always work
Performance-based payment is nowadays conventional in a wide variety of jobs. It is commonly
seen as a pay scheme that enhances effort and productivity relative to a non-bonus payment.
The intuitive logic behind it is that the more you motivate people to achieve something, the
harder they will work to reach their goal. After all, this is part of the rationale behind offering
CEOs and top-executives sky-high bonuses: pay people a very high bonus and they will be
motivated to perform at very high levels!
Well, perhaps the story is not quite that simple. Imagine a professor offering you a 10,000
bonus if you pass his course. While you might probably study a bit harder than you otherwise
would, you may also feel much more nervous during the exam. On balance, you might perform
quite badly as a result of the extremely generous bonus. In line with this, many observers claim
that sky-high bonuses in the financial sector were an important cause of the financial crisis of
2008. In order to test which story is closest to reality, Dan Ariely (Duke University), Uri Gneezy
(UCSD), George Loewenstein (Carnegie Mellon University) and Nina Mazar (University of
Toronto) set up a smart series of experiments. In these experiments, they varied the amount of
bonuses the participants could receive when they performed well and measured the influence
on performance. In analogy to the question posed above, they were particularly interested in
whether very high monetary incentives would increase or decrease performance.
To keep their research budget low, their first experiment was conducted in rural India where the
average persons monthly spending was about 500 rupees (approximately 7) at the time.
Eighty-seven residents were recruited to participate in the experiment, which took place in 2002.
The participants had to play a number of games that required creativity, concentration, memory,
and problem-solving skills. The researchers believed these qualities to be highly relevant for
CEOs and other executives. Participants were randomly assigned to one of three treatments in
which they faced incentives (in all games) that were either relatively small, moderate or very
large. In each treatment, participants played six different games in a random order and were
promised payments for each game if they reached certain performance levels.
124

lOMoARcPSD|1733731
The results of the first experiment point to two main conclusions. First, there was no significant
difference in the performance between the low-payment and moderate-payment groups. Thus,
despite the large relative difference in magnitude of reward across the treatments (i.e., 10 times
higher for the moderate-payment condition relative to the low-payment condition),
performance did not seem to increase. One interpretation of this result is that the incentives in
the low-payment condition were not that low after all and therefore created a level of
performance that was already at a peak. Second, and more importantly, the performance of
participants was always lowest in the high-payment group compared to the low-payment and
moderate-payment groups.
The researchers were surprised that both results applied to all of the games they used.
Therefore, in their follow-up experiment, they included a task that only required physical effort
(key-pressing) and another that mostly required cognitive skills (adding). This time, the
participants were undergraduate students at MIT and each student received both high and low
levels of monetary incentives. In the key-pressing task the students were asked to alternate
between pressing the v and n keys on the keyboard. In the adding task the students were given
a set of 20 matrices (similar to the one in table 11.3), one at a time, with 12 numbers in each
matrix and were asked to find the two numbers in that matrix that would add to 10. Each
student participated three times in each of the two tasks: once for practice, once for low pay, and
once for high pay.
Table 11.3: Sample matrix
9.38 6.74 8.17

5.15 6.61 3.06
9.71 0.91 4.88
3.58 4.87 6.42
In this second experiment, the researchers observed the following: Like in the India experiment,
performance in the adding task went down in the case of a very high bonus. In contrast, the
performance for the key-pressing task increased with the level of the bonus. The results indicate
that tasks that require reasoning or other cognitive skills seem to have a level of incentive
beyond which further increases can have negative effects on performance, while tasks involving
only physical effort are likely to benefit from increased incentives.
These findings strongly suggest that higher incentives do not always lead to higher performance!
Still, determining the optimal incentive scheme or level of reward is not an easy job. Factors such
as the characteristics of the task, the characteristics of the individual and his or her experience
with the task can prove to be quite significant in determining the final effect on performance.
Does this imply that we should stop rewarding people for their work and contribution? Well,
most certainly not; the problem lies with the types of compensations people receive and how
stressful such large stakes might be.
Jonathan Levis 2002 paper in the American Economic Review is an influential paper in the
literature on relational contracts. Levi studies self-enforced incentive contracts and extends our
analysis to a variety of settings with hidden information and asymmetry problems (e.g. moral
125

lOMoARcPSD|1733731
hazard). Other work concerning relational contracts include MacLeod and Malcomson (1998),
Che (2001), Levin (2002, 2003) and Rayo (2007).
For a comprehensive review of Bulls repeated-game model of a bonus based on a subjective

assessment of a workers total contribution to firm value, refer to his 1987 paper published in
the Quarterly Journal of Economics. The version presented in the chapter is a simplified form of
Bulls formal analysis.
The case study of this chapter is based on Ariely et al.s article in the Review of Economic Studies.
References
Ariely, D., U. Gneezy, G. Loewenstein, and N. Mazar (2009). Large stakes and big mistakes, Review
of Economic Studies, 76, 451-469.
Bolton, P. and D. Mathias (2004). Contract Theory, Cambridge, MA: MIT Press.
Bull, C. (1987). The existence of self-enforcing relational contracts, Quarterly Journal of
Economics, 102, 147-159.
Che, Y. (2001). Optimal incentives for teams, American Economic Review, 91(3), 525-541.
Holmstrm, B. (1982). Moral hazard in teams, Bell Journal of Economics, 13(2), 324-340.
Levin, J. (2002). Multilateral contracting and the employment relationship, Quarterly Journal of
Economics, 117(3), 1075-1103.
Levin, J. (2003). Relational incentive contracts, American Economic Review, 93(3), 835-857.
MacLeod, W.B. (1989). Implicit contracts, incentive compatibility and involuntary
unemployment, Econometrica, 57(2), 447-480.
Milgrom, P. (1988). Employment contracts, influence activities and efficient organization design,
Journal of Political Economy, 96(1), 42-60.
Milgrom, P. and R. John (1988). An economic approach to influence activities in organizations,
American Journal of Sociology, 94, 154.
Rayo, L. (2007). Relational incentives and moral hazard in teams, Review of Economic Studies,
74(3), 937-963.
Exercises
Exercise 11.1 Principal-Agent

Consider a firm that sells its products to the market for a given price . Production depends on
how much effort the firms workers exert. For simplicity, assume that the firm employs only one
worker. Production technology is given by the production function:
=
Where is the firms output, and is the workers effort. The workers utility function is given
by:
1 2
=
2
Where is the workers wage. The workers participation constraint is . Suppose that the
firm can observe the workers effort. The firm offers a contract consisting of a fixed wage and a
minimum required effort level.
126

lOMoARcPSD|1733731
a) Derive the profit-maximizing contract, taking into account the workers participation
constraint.
b) Calculate the resulting level of profits when the firm offers the profit-maximizing contract
calculated in (a) and interpret your result.
Suppose now that the firm cannot observe effort, but only output. The firm pays the worker +
.
c) Derive the profit-maximizing level of and also taking into account the workers
participation constraint. Calculate the resulting level of profits and compare it with your result
in (b).
Assume that, due to public regulation, the workers fixed pay must be at least equal to with:
1 2
0 < < +
2
d) What type of contract will the profit-maximizing firm offer?
Exercise 11.2 Principal-Agent

Consider the same setting as in Exercise 11.1, but now assume that the worker likes to exert
effort to some extent. This is reflected in his utility function, which is in this case given by:
1 2
= +
2
The term represents the joy this worker obtains from work. All other variables remain the
same. Suppose the firm offers a fixed salary to its worker but no bonus ( > 0, = 0).
a) Determine the level of fixed salary with which the firm attracts workers at the lowest possible
cost. How hard do workers work? What is the resulting level of profits?
Suppose now that the firm offers the following wage scheme:
= +
b) Derive the profit-maximizing level of and also taking into account the workers
participation constraint. Compare your results with those obtained in exercise 11.1(c).
Exercise 11.3 Relational contracts (Part 1)

Consider the relational contract problem that arises between a risk-neutral firm and a risk-
neutral agent. At the beginning of each period , with {1, 2, , } the firm offers a contract
(, ) to the agent, where is the agents salary and the bonus if output is high. The agent
receives the salary and privately chooses action at a cost given by:
2
() = + 9,000
2
Output of the firm is either 0 or = 25,000, with probability:

{} = {|} =
100
After observing the output, the firm chooses whether or not to pay the bonus . The firm pays
127

lOMoARcPSD|1733731
the bonus if and only if output is equal to . As long as the firm obeys this rule in the previous
periods, the relationship is continued. If the firm deviates once from paying the bonus the agent
terminates the relationship at the beginning of the next period. In the case that the relationship
is terminated both parts receive a utility of zero in all future periods. Output is observed by the
firm and the agent but cannot be enforced by court. Both the firm and the agent discount future
payoffs using the same interest rate .
a) Suppose = 8,000 and = 5,000. If the agent believes that the firm will stick to the bonus
agreement, what action will he choose?
Continue assuming the same contact agreement offered by the firm.
b) For which rates of the interest rate will the firm pay the bonus?
Exercise 11.4 Relational contracts (Part 2)

Consider the same setting as in exercise 11.3. However, instead of using the given contract
scheme the firm is interested in deriving the optimal contract ( , ) from its perspective for
all levels of the interest rate . In that respect, the firm knows that the events determining the
optimal relational contract should occur according to the following sequence:
First of all, if the agent accepts the contract she will choose an action in order to
maximize her utility.
Secondly, the firm realizes that the contract ( , ) should be sufficiently attractive for
the agent to take it. In fact, the firm will decrease the agents salary exactly to the level
that the agent will just accept it.
Finally, the bonus offered by the firm should maximize the firms expected profits per
period given that the firm will choose a salary .
a) Using the above-mentioned sequence of events derive the agents optimal action given a
contract (, ). What is the minimum required level of salary that the firm should offer to the
agent if the bonus equals ?
b) Determine the first-best optimal bonus. For which levels of the interest rate will the firm
offer the first-best optimal bonus in equilibrium?
c) What bonus will the firm offer to the agent for any other level of the interest rate than the
one calculated in (b)?
128

lOMoARcPSD|1733731
Chapter 12: Collusion in markets

So far, we have seen that players who interact repeatedly can reach attractive outcomes they
would not be able to reach if they interacted only once. In this chapter, we will apply this idea to
competition between firms on markets. We start in section 12.1 by discussing collusion
between firms, i.e., explicit or implicit agreements between firms, for instance, about the prices
they will charge. In particular, we will argue why firms would like to collude, how they can do so,
and under what circumstances collusion is stable. In section 12.2, we will further zoom in on
circumstances under which collusion is stable. In particular, we will identify market
characteristics and business practices that facilitate collusion. Section 12.3 contains a discussion
on how competition authorities can fight collusion. Section 12.4 contains a case study on the
worldwide lysine cartel.
12.1 Collusion in markets: Why, how, and when

Let us return to the coffee market on your campus we introduced in chapter 6. We assumed that
Coffee Bucks and Star Company are the only firms on this market, where total demand for coffee
equals
() = 960 240.
We observed that if the firms compete la Bertrand, both will charge a price equal to marginal
costs resulting in zero profits. Clearly, the firms have a strong incentive to collude so that they
maintain a higher price. We saw that in the case of Cournot competition, firms do establish an
equilibrium price above marginal costs. For instance, if the marginal cost of a cup of coffee
equals = 1.75, the equilibrium price equals
= 2.50
which is quite a bit above marginal costs. Still, firms have an incentive to collude because the
monopoly price is even higher: = 2.875.
How can Coffee Bucks and Star Company establish the monopoly price? They have many
possibilities. The simplest is a price agreement: Coffee Bucks and Star Company agree to charge
the monopoly price. Another possibility is an agreement on quantity: The two firms agree to
restrict their combined production to the monopoly quantity, for example, by each producing
half the monopoly quantity. Alternatively, the companies may divide the market if the market
consists of several separate submarkets. For instance, they may agree that Coffee Bucks only
sells coffee and Star Company only tea, or that Coffee Bucks only caters to students and Star
Company only to professors. The firms can charge the monopoly price in each submarket
because they are the sole supplier in each submarket. Another possibility is that one of the two
firms pays the other to leave the market so that the remaining firm becomes a monopolist. Such
an agreement is in the interest of both firms if the side-payment equals half the monopoly profits.
When firms establish any of the above agreements, we say that they collude. We distinguish
between two types of collusion: explicit collusion and tacit collusion. Companies collude
explicitly if independent firms take joint decisions, i.e., they formed a cartel. Firms can also
collude tacitly, which is the case when they make independent decisions that have the same
consequences as in the situation in which the decisions had been taken jointly. For example,
Coffee Bucks and Star Company may both realize that it is in their best interest to charge the
129

lOMoARcPSD|1733731
monopoly price. If both charge the monopoly price without having communicated about it, we
speak about tacit collusion.
Regardless of how companies establish collusion, they face several challenges. The first one is to
establish an agreement both can live with. In the coffee market example, this is relatively easy
under the assumptions we made. The firms are symmetric in that they face the same marginal
costs and consumers view their coffee as homogeneous goods. Both will be happy with an
agreement that gives them both half the monopoly profits. Life becomes more difficult if
asymmetries enter the picture. For example, if Coffee Bucks has lower marginal costs than Star
Company, the monopoly price does not exist: Coffee Bucks would charge a lower price than Star
Company if it were the only firm in the market. So, when the two companies try to fix a price,
they face a difficult decision as to what price to coordinate on. Next to establishing the
agreement, the companies face the challenge of not being caught by the competition authority
because such agreements are illegal when firms communicate about them, as we saw in chapter
3.
Second, once they have reached an agreement, colluding firms should find a way to enforce it.
They cannot write a binding contract about their agreement as such contracts typically are not
enforceable in court for the simple reason that they are illegal according to competition law.
Because the firms cannot write a binding contract, each may have an incentive to deviate from
an agreement. For instance, suppose Coffee Bucks and Star Company agree to produce half the
monopoly quantity. As figure 12.1 indicates, Star Company has a reason to deviate from this
agreement by producing more, and so does Coffee Bucks. Of course, as we saw before,
cooperative outcomes can be enforced when the parties have credible ways of punishing
deviation. So, deviations should be observable in the first place, which implies that firms should
be able to monitor each others actions. Depending on the type of agreement, monitoring can be
quite easy. For example, an employer of Coffee Bucks could buy coffee at Star Company every
day to check whether Star Company sticks to the agreed upon price.
Figure 12.1 Incentives to deviate
In addition, the firms should have a credible punishment device ready in case they observe the
other firm deviating. As we saw before, firms can play a trigger strategy in the case of repeated
130

lOMoARcPSD|1733731
interaction, according to which the punishment device is playing the Nash equilibrium strategy
that gives the other firm the lowest payoff. For example, Coffee Bucks and Star Company could
establish the monopoly price in equilibrium if both employ the trigger strategy presented in
Table 12.1. As we observed in chapter 10, this trigger strategy is a subgame perfect Nash
equilibrium if both firms are sufficiently patient.
Table 12.1 Trigger strategy in the case of a price agreement
Period Action
1 Choose the monopoly price
Choose the monopoly price when the other firm chose the monopoly price in all
previous periods
= 2,3,
Otherwise, choose the marginal cost price
A third challenge for the cartel is to keep other firms out of the market. When Coffee Company
and Star Bucks establish the monopoly price one way or another, a third firm may find it
attractive to enter the market and grab a share of the monopoly profits, leaving the market as
soon as Coffee Company and Star Bucks decrease the price (a hit-and-run strategy). It is not
unlikely that entry may induce the firms to stop colluding. The threat of entry may even prevent
the firms from colluding in the first place because the gains from collusion are diminished.
12.2 Factors facilitating collusion

In this section, we consider in more detail in which markets collusion is most likely. Moreover,
we study business strategies that firms may employ to facilitate collusion among them. Recall,
that in a repeated game setting, collusion (enforced using a trigger strategy) is feasible in
equilibrium only if

= .

We concluded in chapter 10 that collusion is more likely, the greater the collusive payoffs ,
the lower the Nash-payoffs , and the lower the deviation payoffs . We can use this insight
to establish what market characteristics and what business strategies facilitate collusion.
We start with probably the most obvious market characteristic facilitating collusion: market
concentration. The more concentrated the market the easier it is for firms to collude. There are
at least two reasons why this is the case. First, the more concentrated the market, the fewer the
number of active firms in the market so that it becomes easier to coordinate behavior. For
instance, in the case of three firms it is easier to establish a collusive price which all firms can
live with than in the case of 10 firms. Second, the fewer active firms there are in the market, the
less reason they have to deviate from the cartel agreement. The reason is that in the case of just
a few firms in the market, each already has a substantial share in the collusive profits. Therefore,
there is relatively little to gain from deviation. In other words, the collusive payoffs are large
compared to the deviation profits . Of course, the Nash payoffs are greater in a more
131

lOMoARcPSD|1733731
concentrated market, but generally, this is not sufficient to compensate for the relatively large
collusive profits.
High entry barriers is another market factor facilitating collusion. As said, a challenge for
colluding firms is preventing others from entering the market. Entry barriers are very helpful in
that respect. An entry barrier is a cost of producing that must be borne by firms seeking to enter
a market but is not borne by firms already in the market. Entry barriers come in many shapes
and sizes, including the cost of building production facilities, advertising costs needed to attract
customers, and licenses firms have to buy from the government. The lower the entry barriers are
the more difficult it is to generate above normal profits because of the entry those will provoke.
In other words, entry decreases collusive profits resulting in relatively low .
Transparency and frequent interaction are two other market characteristics facilitating
collusion. If the market is transparent, firms can monitor each others actions easily, and detect
and punish deviations quickly. So, the more transparent the market the sooner the next period
starts. Similarly, the more frequently firms interact, the nearer the next period is. For example,
oil companies may change the price of petrol on a day-by-day basis so that the next period may
already start tomorrow. In contrast, firms selling holiday packages have to make crucial capacity
decisions several months before the start of each travel season, including the number of hotel
rooms or airplane seats they reserve. The next period may start next year or even later. The
earlier the next period starts, the greater the discount factor is and therefore is more likely
to lie in the range of discount factors for which collusion is feasible.
Stable demand facilitates collusion as well. Let us distinguish between two possibilities:
Fluctuations in market demand are observable or not. We start by considering the case where
demand fluctuations are observable. For example, the consumption of ice cream varies quite
predictably with the outside temperature. Suppose, for simplicity, that demand is either high or
low. It is most attractive for a firm to deviate from a collusive agreement if demand is high: by
deviating, the firm grabs a relatively big share of the pie ( is greater in high-demand
periods than in low-demand periods), while the level of punishment is not affected (assuming
that low- and high-demand periods occur randomly independently of demand in the past). The
more demand fluctuates, the more attractive it becomes to deviate in the case of high-demand
periods so the less likely collusion is. In some situations, firms may be able to collude only in
low-demand periods. In that case, a surprising pattern emerges: high prices in busts (low-
demand periods) and price wars during booms (high-demand periods).
Now suppose that demand fluctuations are not observable. If market demand is stable over time,
a company knows when another firm deviates from a collusive agreement because it will
experience a decrease in its own demand. As a consequence, monitoring is relatively easy. If
market demand fluctuates over time, monitoring becomes more difficult and so does collusion. A
firm can only speculate whether a decrease in its demand is caused by another firm deviating or
market demand being low. Should firms punish competitors in the case when its demand it low?
Yes, they should. If they do not, it becomes attractive for competitors to deviate because they
will never be punished. However, note that in this setting the trigger strategy does not work to
keep collusion stable. According to the trigger strategy, punishment will last forever. But if that
is the case, firms will start punishing at some point for the simple reason that demand will be
low some time. The best way to collude is to start a phase of temporary punishment after which
the firms return to the collusive outcome.
132

lOMoARcPSD|1733731
In general, firm symmetry facilitates collusion. Of course, firms may be symmetric (or
asymmetric) in many ways. For example, they may have a similar cost structure. Symmetry in
terms of cost structure facilitates (tacit) collusion. There are at least three reasons why this is
the case. First, the smaller cost differences are between firms, the easier it is to coordinate
behavior because all will have roughly the same collusive price in mind. Second, in the case of
strong cost differences, the most efficient firm has a great incentive to deviate because it is likely
that the collusive price lies substantially above its marginal cost (otherwise the other, less
efficient firms would not be willing to collude). In terms of our model, is relatively large for
the most efficient firm. Third, in the case of cost asymmetries, when the most efficient firm
deviates, it will be difficult for the other firms to punish it seriously, which gives the most
efficient firm an additional reason to deviate (its is relatively high).
Production capacity is another dimension on which firms may differ. Symmetry in terms of
production capacity facilitates (tacit) collusion for the following reasons. The greater the
differences in production capacity the more attractive deviation becomes for the firm with the
largest production capacity, i.e., the higher its . The reason is that in the collusive agreement,
this firm has substantial capacity available to expand production and benefit from deviating. In
addition, firms with low capacity cannot substantially punish deviation because they simply
cannot expand their production much, resulting in a relatively high for a high-capacity firm.
A third relevant firm characteristic is their number of product lines. For example, the Coca-Cola
Company has at least three product lines of products, including Coca-Cola, Fanta, and Sprite. The
same holds true for PepsiCo Inc. selling Pepsi Cola, Gatorade, and 7UP. Symmetry in the number
of product lines facilitates (tacit) collusion if the products are substitutes. The greater the
differences in the number of product lines the more interesting it becomes for a firm with a few
product lines to deviate (i.e., the higher its ). The reason is that this firm can steal business
from the many product lines of its competitors to its own product line. Additionally, firms with
many product lines will find it unattractive to inflict serious punishment because they will have
to cannibalize on their other product lines (resulting in a relatively high for the small firm).
In order not to oversell the idea that symmetries facilitate collusion, let me stress that the exact
opposite might occur as well. Probably the most straightforward case in which asymmetries
facilitate collusion is when a market has a clear market leader (e.g., the biggest firm). This
market leader could act as a price leader, which may help firms coordinate on a collusive price.
Such coordination may be more difficult if none of the firms stand out against the others.
Another example where asymmetries may facilitate collusion is product differentiation. In the
case of strongly differentiated products, firms have less incentive to deviate from a cartel
agreement than in the case of homogeneous products. In the latter case, firms sell close
substitutes so that they may steal some business from the other firm by deviating from a
collusive agreement (high ). In the former case, firms are already close to being a monopolist
so they cannot steal much business from the other firm. Of course, the more differentiated
products are the higher Nash equilibrium profits so that deviations can be punished less
severely (higher ), which has a disciplining effect on collusion. In total, the effect of product
differentiation on collusion is ambiguous.
So far, we have looked at the effect of market characteristics on collusion. In addition, certain
business strategies may facilitate collusion. We call such strategies facilitating practices.
Clearly, firms have strong incentives to engage in facilitating practices because these could
133

lOMoARcPSD|1733731
soften competition between them. The most obvious example is probably a merger. If two firms
merge, the market becomes more concentrated. Greater market concentrate may facilitate
collusion, as we saw before.
Another example of a facilitating practice is the lowest price guarantee. If a firm gives a
customer a lowest price guarantee, it promises the customer a refund if she finds the same
product for a lower price at another firm. Typically, the refund equals the difference between the
original price and the price offered by the other firm. In the case of a lowest price guarantee, it
will soon become clear whether another firm deviates from a collusive agreement by offering a
lower price, because in that case customers are eager to return to collect their refund. In other
words, the market becomes more transparent which facilitates collusion as we saw above.
A most-favored-customer-clause represents another facilitating practice. If a customer

obtains a most-favored-customer-clause, she pays a lower price if the price decreases in the near
future. For example, a bank may assign such a clause to a person who obtains a mortgage for
their house. The interest rate on the mortgage typically fluctuates from one month to the next.
Many homeowners prefer to get a mortgage with an interest rate that is fixed for a long period of
time (let us say, five to 30 years) so that they do not face the risk inherent in interest rate
fluctuations. Typically, homeowners arrange their mortgage and the corresponding interest rate
months before the actual transaction. A most-favored-customer-clause specifies that the buyer
will pay the lowest interest rate the bank offers up to the moment of the actual transaction for
similar mortgages. This arrangement sounds very attractive for the banks customers because
they benefit from decreases in the interest rate while not being harmed by interest rate
increases. However, a most-favored-customer-clause may also serve as a facilitating practice.
The reason is that it becomes relatively unattractive for a bank to deviate from a collusive
agreement by decreasing its interest rate because it has to decrease the interest rate on all its
current customers as well.
12.3 Collusion and competition policy

In this section, we will discuss the welfare consequences of collusion and a potential role for
competition authorities. As we saw in chapter 1, welfare is lower the more the price deviates
from marginal costs. So, if collusion increases the price consumers pay, it will usually have a
negative impact on welfare. Notable exceptions are agreements related to research and
development (R&D) and between firms in different layers of the vertical chain. In chapter 1, we
concluded that for dynamic efficiency, R&D is crucial. Firms may need to cooperate when
developing new products and improving production processes because otherwise they may not
have access to sufficient capital and knowledge. As we observed in chapter 7, in a vertical chain,
prices can be excessively high because of double marginalization. Agreements between firms in
the vertical chain can mitigate the double marginalization problem benefiting both the firms and
the consumers. Later, in chapter 17, we will discuss several examples of such welfare enhancing
agreements.
Competition authorities have several instruments available to fight collusion. In fact, each of the
three pillars of competition policy we introduced in chapter 3 may be effective. The most
obvious is anti-cartel law: In most countries of the world, agreements about price, production,
market division, and so forth are forbidden. For example, Paragraph 1 of Article 101 of the
Treaty on the Functioning of the European Union reads:
134

lOMoARcPSD|1733731
Note that the article refers to agreements. In case law, this is usually interpreted as explicit
agreements. Typically, in a cartel case, the competition authority looks for written and recorded
evidence that parties communicated with each other about prices, quantities, market division,
and so forth. In other words, tacit collusion is not illegal in most cases because the firms only
make implicit agreements. The distinction case law makes between explicit and tacit collusion
might surprise you because both types of collusion can have equally detrimental effects on
welfare. Of course, it is hard for outsiders to judge whether firms coordinate or compete
normally in the absence of evidence of an agreement.
Paragraph 3 of Article 101 indicates that the European Commission makes an exception for
agreements related to R&D activities, in line with the potential benefits of R&D cooperation
between firms:
The provisions of paragraph 1 may, however, be declared inapplicable in the case of any
agreement [...] which contributes to improving the production or distribution of goods or to
promoting technical or economic progress [...]
On the basis of the same provision, the European Commission tends to be lenient with respect to
agreements between firms in different layers of the vertical chain. This makes sense because
such agreements are welfare improving in general, for instance by mitigating the double
marginalization problem.
Business practices facilitating (tacit) collusion may be considered an abuse of a dominant

position. For example, competition authorities found that particular most-favored customer
clauses violated competition law because firms could abuse their dominant position by applying
this business strategy.
Finally, competition authorities may block mergers that are likely to give rise to a market in
which firms can collude much easier than before the merger. In fact, the European Commission
makes a distinction between two effects a merger may have on competition in the markets on
which the merging firms are active:
There are two main ways in which horizontal mergers may significantly impede effective
competition [...]:
135

lOMoARcPSD|1733731
(a) by eliminating important competitive constraints on one or more firms, which

consequently would have increased market power, without resorting to coordinated
behaviour (non-coordinated effects);
(b) by changing the nature of competition in such a way that firms that previously were not
coordinating their behaviour, are now significantly more likely to coordinate and raise
prices or otherwise harm effective competition. A merger may also make coordination
easier, more stable or more effective for firms which were coordinating prior to the merger
(coordinated effects).
Note that non-coordinated effects refer to situations in which firms do not collude after the
merger. A merger may still result in undesirable effects, because the merger decreases the
number of firms in the market so that the remaining firms may find it in their interest to
compete less fiercely. Coordinated effects relate to market conditions that facilitate collusion.
12.4 Case study: Collusion beyond theory

The factors facilitating collusion that we have seen in this chapter find support in both theory
and common sense. However, other stylized facts and practices emerge from the analysis of
recent cases of collusion. In an article in the American Economic Review in 2011, Joseph
Harrington, professor at Johns Hopkins University, and Andrzej Skrzypacz, professor at Stanford
University, review some of these cartels, highlighting the recurring aspects that complement
(and sometimes contrast) the classic theory on collusion.
One of the most famous collusive agreements in recent years is the global lysine cartel formed by
the five major lysine producers in 1990, which lasted approximately five years. (The movie The
Informant is based on this cartel agreement.) Lysine is an amino acid usually added to livestock
feed to develop body tissue in pigs and poultry. As Table 12.2 shows, the companies agreed upon
allocation of output based on market share and geographical area of production. From the
competition authoritys investigation it emerges that the cartel members also coordinated on
price. For example, in July 1992 the members agreed to raise the price twice before the end of
the year.
Table 12.2: Lysine Output Allocation in 1992 (tons)
Company Global Europe

Ajinomoto 73,500 34,000
Archer Daniels Midland 48,000 5,000
Kyowa 37,000 8,000
Sewon 20,500 13,500
Cheil 6,000 5,000
To monitor possible deviation, each company had to report their sales volumes on a monthly
basis. Additionally, guaranteed buy-ins were used: A company that reported sales above its
quota had to buy output from companies reporting sales below their quota. At first sight, self-
reporting volumes of sales does not appear to be an efficient mechanism to monitor the cartel
members behavior because it leaves strong incentives to underreport. Indeed, the investigation
showed how underreported volumes happened occasionally. However, the collusive agreement
was largely successful overall.
136

lOMoARcPSD|1733731
Also at the beginning of the 1990s, the five largest producers of citric acid (a component used
primarily in the food and beverage industry, but also for cleaning products, pharmaceuticals,
and cosmetics) formed a cartel. Each company had a sales quota based on the average market
share of the previous three years, as shown in table 12.3. Similarly to the lysine cartel, each citric
acid producer self-reported its volume of sales on a monthly basis to an executive of Hoffman
LaRoche, who gathered all the information and distributed them back to the companies. The
reported sales were checked by independent Swiss auditors that provided external validity. Also
in the citric acid cartel a buy-back system applied: The companies that produced above quota
were obliged to purchase output from the companies below quota. The investigation underlined
how the actual production by each member adhered very closely to the quotas imposed by the
cartel.
Table 12.3: Citric Acid Market Allocation
Company Market share (%)

Haarman and Reimer 32.0
Archer Daniels Midland 26.3
Jungbunzlauer 23.0
Hoffman LaRoche 13.7
Cerestar Bioproducts 5.0
A common denominator of the markets of lysine and citric acid is their intransparency: The price
is usually set bilaterally between a seller and a buyer so that it is not observable for other parties,
including other cartel members. In this chapter, we identified lack of transparency as a factor
hindering collusion. The cartel cases show that collusion is still feasible in such intransparent
markets, albeit the firms had to rely on a rather complicated cartel mechanism:
The firms made agreements on both output allocation and prices.

The firms in the cartel reported their actual sales monthly to the other cartel members.
A monitoring mechanism compared the self-reported sales with the quotas assigned by
the cartel.
The collusive agreement used side-transfers whereby firms that reported sales above
their quota effectively paid those who reported sales below their quota.
Harrington and Skrzypacz prove theoretically that such cartel mechanism is stable in the sense
that collusive strategies satisfy a trigger strategy stability condition if firms are sufficiently
patient. The mechanics resemble the one for settings with non-observable demand fluctuations
that we discussed before in the sense that low reports can be followed by a phase of temporary
punishment. Harrington and Skrzypacz theory is rather advanced (way beyond the scope of this
course) so that it is quite impressive how the firms managed to develop this cartel mechanism
and sustain it for many years.
In the 1980s, it became common to study factors facilitating collusion using repeated game
theory. Famous articles include Rotemberg and Saloners (1986) work on price wars during
booms and Green and Porters (1984) study of markets with non-observable demand
fluctuations. More recently, Compte et al. (2002) study the effect of asymmetries in production
capacity on the stability of collusion. See Ivaldi et al. (2003) for an excellent overview of factors
137

lOMoARcPSD|1733731
facilitating collusion in markets. This chapters case study is based on Joseph Harrington and
Andrzej Skrzypacz 2011 article published in the American Economic Review.
References
Compte, O., F. Jenny and P. Rey (2002). Capacity constraints, mergers and collusion, European
Economic Review, 46, 1-29.
Green, E. and R. Porter (1984). Non-cooperative collusion under imperfect price information,
Econometrica, 52, 87-100.
Harrington, J.E. and A. Skrzypacz (2011). Private monitoring and communication in cartels:
explaining recent collusive practices, American Economic Review, 101, 2425-2449.
Ivaldi M., B. Jullien, P. Rey, P. Seabright, and J. Tirole (2003). The economics of tacit collusion,
IDEI Working Paper, 186, Final Report for DG Competition, European Commission.
Rotemberg, J. and G. Saloner (1986). A supergame-theoretic model of business cycles and price
wars during booms, American Economic Review, 76, 390-407.
Exercises
Exercise 12.1 Collusion under price competition

Suppose you own a firm that produces mineral water, and you are currently competing with
only one other firm. The consumers perceive mineral water as a homogeneous good. The market
is characterized by price competition. Aggregate demand can be described as:
() = 240 2,
where represents the price per bottle of mineral water. You and your competitor have the
same cost function ( ) = 20 (with = , ). Your own demand (and the one of
your competitor) is such that:
( ) <
( )
( ) = = , = ,
2
{0 > .
a) What price would you and your competitor set to maximize your joint profits?
b) Which is the one-shot Nash Equilibrium?
c) Assume that you and your competitor interact an infinite number of times. Therefore, you
think it may be profitable to make a collusive agreement with the other firm. Which trigger
strategy would grant maximum profits to both? Find the critical discount factor () that makes
collusion sustainable.
d) Assume now that in each period, demand declines by 10% of the original demand. Is your
collusion still feasible? How is the critical discount factor going to change?
Exercise 12.2 Collusion under price competition

In the country of Sleep-well two firms, HighJumps and StrongSprings, produce mattresses. The
two firms produce only one type of mattress and they believe they will compete an infinite
number of periods. The consumers perceive the good as homogeneous. Aggregate demand is:
138

lOMoARcPSD|1733731
= 180 ,
where is the amount of mattresses produced and is the price for one mattress. Despite the
similarity of the product, HighJumps R&D granted it a cost advantage with respect to the
competitor. In fact, ( ) = 20 and ( ) = 40 , where the subscripts and stand for
HighJumps and StrongSprings, respectively. Assume that if both firms charge the cartel price, the
two firms split the market evenly.
a) Call the cartel price, which trigger strategy will permit the firms to sustain collusion if both
are sufficiently patient?
b) Assume the firms agree to set the cartel price equal to HighJumps monopoly price. Find the
critical discount factor ().
c) How does the critical discount factor change if the parties agree to set StrongSprings
monopoly price as the cartel price?
d) Assume now that in case of collusion, HighJumps will serve 60% of the market. Without
making any calculations, do you think the collusion will become more sustainable?
Exercise 12.3 Collusion under competition on quantity

In the city of Seville there are producers of sangria, the famous Spanish wine punch. The citys
aggregate inverse demand is:
= 48 ,
where is the price for a bottle of sangria and is the total quantity produced. The producers
compete la Cournot and they all have the same cost function ( ) = 12 .
a) Find equilibrium quantity, price and profits as functions of . (You can use the symmetry
condition.)
Assume that instead of firms, Sevilles market of sangria is a duopoly ( = 2). The firms form a
cartel whereby they set the price equal to the monopoly price, dividing the monopoly profits
evenly.
b) Find the critical discount factor of this cartel.
c) Now assume that there are three firms competing in the market, and compute again the
critical discount factor.
d) What do you conclude comparing your answer in b) and c)?
Exercise 12.4 Factors facilitating collusion

Imagine there are firms in the market that compete on price. They agree to set the monopoly
price and divide the demand evenly. All the firms have the same marginal cost, and they believe
they will interact for an infinite number of periods. If any firm deviates from the collusive
agreement, the others will play their Nash equilibrium price forever after.
a) Find the critical discount factor for the firms and show that it is increasing in .
Assume that the demand is increasing at a constant rate (with 0 < < 1), so that:
139

lOMoARcPSD|1733731
() = (1 + ) ().
b) Does the increasing demand hinder or facilitate collusion? (Hint: the profits increase at the
same rate of the demand).
Assume now that the demand is constant, but the interactions among the firms happen every
periods. That is, firms compete in periods 1, + 1, 2 + 1, 3 + 1, etc.
c) How does the critical discount factor changes as the periods between the interactions
increase (i.e., as increases)?
Imagine now that the demand is constant, and the interactions happen in every period. However,
the market has entry barriers to some extent. More precisely, in each period there is a
probability that a new firm enters the market and sets a price equal to marginal costs. Here
represents the entry barriers effectiveness (i.e., if = 0 the entry barriers are so high that no
firms can enter the market).
d) Compute the critical discount factor of a firm already present in the market and show that the
higher the entry barriers the more likely the cartel is sustainable.
140

lOMoARcPSD|1733731
Part V: Information asymmetry

In the analysis made up until now, we assumed that decision makers possess all the information
relevant for the decisions they make: Consumers observe the quality of products before they buy
them, managers know the quality of the applicants for their vacancies, firms are completely
informed about consumer demand, and so on. In practice, the assumption of complete
information is often unrealistic. This week, we will study the consequences of information
asymmetry, i.e., settings where one decision maker has better information than the other. We
start in chapter 13 by studying games with asymmetric information. We will see how markets
with asymmetric information may be highly inefficient because high-quality goods will be kept
from the market. We will also propose two potential solutions to this problem: screening and
signaling. In chapter 14, we will apply the insights obtained in chapter 13 to hiring decisions by
firms. Chapter 15 contains a discussion of price discrimination as a screening device in markets.
141

lOMoARcPSD|1733731
Chapter 13: Games with asymmetric information

In this chapter, we will introduce you to settings with asymmetric information. We will do so by
discussing the work of three important contributors to the analysis of games with asymmetric
information, George Akerlof, Michael Spence, and Joseph E. Stiglitz, the winners of the 2001
Nobel Memorial Prize in Economic Sciences. In section 13.1, we will analyze the market for
second-hand cars and we will identify what Akerlof called the adverse selection problem, i.e., a
situation where only low-quality goods are traded on the market if sellers of second-hand cars
are better informed about the quality of the car than potential buyers. In section 13.2, we will
discuss screening, a potential solution to the adverse selection problem pushed by Stiglitz. In
section 13.3, we will present another potential solution, put forward by Spence: signaling. The
case study in section 13.4 reveals how a buyer may distinguish a good seller from a bad one on
eBay.
13.1 The adverse selection problem

Consider the following interaction in the market for second-hand cars. A seller, Sting, offers
potential buyer Bono his 10-year-old car for sale. The car can have two quality levels: high
quality (the car is in excellent shape) or low quality (the car has some important defect). Bonos
willingness-to-pay for a high-quality car equals 10,000 while he is willing to pay at most
6,000 for a low-quality car. Stings value for a high-quality car is equal to 9,000 and his value
for a low-quality car is 5,000. Observe that the market outcome is efficient (only) if Sting sells
his car to Bono. The reason is that Bono always attaches greater value to the car than Sting,
regardless of its quality. Indeed, if both Bono and Sting know the cars quality, they will trade
both types of car. In the price range between 9,000 and 10,000, Sting is willing to sell and
Bono is willing to buy a high-quality car. Similarly, both parties are willing to trade the low-
quality car if the price is between 5,000 and 6,000.
What happens in the case of information asymmetry? Suppose that Sting perfectly knows the
cars quality (because he has used it for many years) but Bono does not. More in particular, if the
car is of low quality, it has a defect that Bono cannot detect either from its appearance or in a
test drive. Imagine that Bono believes that the car is of high quality with probability 60% and of
low quality with probability 40%. Consistent with Akerlofs theory, Sting and Bono will not trade
high-quality cars. To see why, assume to the contrary that both types of cars are traded. The
price at which Sting can sell his car is at most 8,400 as Bono is willing to pay up to
60%*10,000 + 40%*6,000 = 8,400 for it. However, for this price, Sting prefers to keep his
car in the case of high quality because his value for it (9,000) is higher than the price (8,400).
We have established a contradiction to the assumption that both cars will be traded on the
market. Indeed, Sting will only sell his car to Bono in the case of low quality.
Because high-quality cars remain unsold, some value-enhancing transactions do not take place:
The market of second-hand cars is inefficient. Akerlof labeled this result the lemons problem
because low-quality second-hand cars are sometimes called lemons. An alternative term for the
lemons problem is the adverse selection problem as the uninformed party tends to select the
wrong kind of trading partner: Bono will only induce Sting to offer his used car if the car is of
low quality. Note that adverse selection is not the same as moral hazard (which we discussed in
chapter 2). In the case of adverse selection, a decision maker lacks information on the
characteristics of the informed party while in the case of moral hazard a decision maker lacks
142

lOMoARcPSD|1733731
information on the performance of the other party. An example of moral hazard is Sting
deliberately refraining from a routine check before he offers his car on the second-hand market.
The adverse selection problem has many more applications in practice than just the second-
hand car market. Health insurance companies may only sell insurances to bad risks if they
cannot find a way to distinguish between healthy and unhealthy people. When banks cannot
observe the creditworthiness of potential borrowers, they face the risk of only attracting people
who are not creditworthy. Employers who cannot verify the extent to which job applicants are
suited for the job may only attract unqualified workers.
There may be a role for government intervention to mitigate potential market inefficiencies
caused by adverse selection. Lemon laws are one example. According to a lemon law, the buyer
of a low-quality good has the right to be compensated financially by the seller if the buyer can
show that the seller advertised that the good was of high quality. Similarly, in some markets, the
government imposes a duty to disclose all relevant information about the product, in particular
if the seller possesses information that the product is of inferior quality. A good example is the
real estate market, where sellers are obliged to inform potential sellers about hidden defects
such as a leaking roof, rotten wood, or a malfunctioning heating system. The government may
also impose quality regulation. For instance, the government decrees that a good can only be
sold as a high-quality good if the seller can show the buyer some sort of certification issued by
an independent expert. Note that those solutions for the adverse selection problem have in
common that the government intervenes in the market. In practice, trading partners have found
many ways to solve or attenuate the adverse selection problem without government
intervention. In the remainder of this chapter, we will discuss some of those mechanisms.
13.2 Screening
Most market solutions for the adverse selection problem can be divided into two rough
categories: screening devices and signaling devices. The difference between the two is that it is
the uninformed party that screens while it is the informed party that signals. In this section, we
look at screening devices. The idea behind screening is that the uninformed party imposes some
mechanism that separates good trading partners from bad ones. Stiglitz won the Nobel Prize for
studying such mechanisms.
Own risk in the insurance market is a typical example of a screening mechanism. To see how this
might work, consider the following setting. Suppose that there are two types of customers in the
market: low-risk customers and high-risk customers. An insurance company prefers to sell
insurances only to low-risk customers because their expected damages are lower than for high-
risk customers. However, assume that for the insurance company it is impossible (or illegal) to
discriminate between a low-risk and a high-risk customer. An adverse selection problem may
emerge because the insurance company will only be able to cover the expected costs for the
high-risk customer if it charges a very high premium. The premium may even be so high that a
low-risk customer will decide not to insure herself at all. In other words, the insurance company
ends up with the exact opposite outcome that it desired: it only insures the people it prefers not
to insure.
Fortunately, the insurance company can do better by offering two types of insurances, one with
a high premium and a low own risk and another with a low premium and a high own risk. In
other words, the insurance company offers a menu of contracts to potential customers. By doing
143

lOMoARcPSD|1733731
so, the company intends to separate good risks (low-risk customer) from bad risks (high-risk
customer). If the menu of contracts is designed well, it ensures that people self-select: Low-risk
customers take the insurance with the high own risk and high-risk customers the one with the
low own risk. The idea is that an insurance with a high own risk is only attractive for low-risk
customers because it is unlikely that they incur high damages. In other words, a restriction on
the menu of contracts is that it must be incentive compatible: Both low-risk and high-risk
customers should prefer the insurance designed for them over the other insurance.
Incentive compatibility implies that the menu the insurance company will offer is probably
inefficient. To see why, let us assume that customers are risk-averse while the insurance
company is risk-neutral. Both assumptions seem a good approximation of reality. Only risk-
averse customers have a reason to insure themselves against risks. Moreover, the insurance
company can spread the risks over its entire portfolio of customers so that it can virtually ignore
the risks associated with a single customer. The efficient outcome is full insurance, i.e., the
insurance company takes the entire risk away from the customer by covering all his damages. In
other words, the efficient insurance does not include an own risk. It can be shown that the
insurance company sells the same insurance to high-risk individuals as in the case of complete
information, i.e., if the company could observe the consumer type. As a consequence, the
insurance for low-risk customers is inefficient because it does contain an own risk. If it did not,
the menu would not be incentive compatible because then the high-risk customer would also
buy this insurance because it has a lower premium. Indeed, low-risk customers suffer from the
presence of high-risk customers. Or, as Stiglitz and his co-author Michael Rothschild put it: If
only the high-risk individuals would admit to their having high accident probabilities, all
individuals would be made better-off without anyone being worse-off.
In some settings, it is difficult if not impossible for the uninformed party to screen their
informed trading partner. Still, the uninformed party may find smart ways to limit the number of
undesirable transactions. Credit rationing in financial markets is a good example. Banks usually
cannot perfectly distinguish low-risk borrowers (investors that can repay their loan with high
probability) from high-risk borrowers (investors that are unlikely to be able to repay their loan).
Stiglitz and his co-author Andrew Weiss argue that it may be profitable for banks to limit the
amount lent to borrowers despite the fact that those borrowers are willing to pay the banks
interest rate. By doing so, banks are able to partly screen out high-risk loans while retaining
some low-risk loans.
The intuition is the following. At the market clearing interest rate, demand for the banks credit
equals its supply. But at that interest rate, the bank will mainly attract high-risk borrowers as
the interest rate is too high for low-risk investors. Only high-risk investors are willing to take a
loan with a high interest rate because they are not liable for the downward risks as they are
protected by bankruptcy laws: a clear instance of adverse selection. A lower interest rate will be
attractive for low-risk borrowers as well so that it will decrease the riskiness of the banks
portfolio, and, in turn, increase its expected profits. Note that at the lower interest rate, credit
demand exceeds supply so that the bank will deny some of the credit demanded by borrowers,
i.e., the bank rations its credit.
Credit rationing may solve the adverse selection problem for two reasons. First of all, the total
amount of money lent to high-risk borrowers is limited so that the relative fraction of risky
assets is low. Secondly, the corresponding interest rate is below the market clearing interest rate
144

lOMoARcPSD|1733731
and therefore the probability that high-risk borrowers cannot repay their debts is lower than at
the market clearing interest rate. (Note that credit rationing may also attenuate a moral hazard
problem in the sense that at a lower interest rate, it becomes less attractive for borrowers to
invest in risky projects.)
13.3 Signaling
In the previous section, we saw how the uninformed party may be able to solve or to mitigate
the adverse selection problem by applying screening devices. In this section, we will focus on
signaling by the informed party. As Michael Spence stressed in his research, the informed party
may have the possibility to send a credible signal to the uninformed party. For example, workers
can send a signal to prospective employers about their talents by showing a college degree
which would be very hard to obtain for less talented workers. What I find interesting about this
example is that the education itself need not have any value for the worker or the employer
beyond being a credible signaling device. (Of course, I do not wish to make the claim that you are
not learning anything valuable from the current course for your future career.)
Advertising by firms may serve as a signal in the very same spirit as the college example. To see
how this might work, consider the market for restaurant meals in Saint Cournot, a small French
town on the Mediterranean coast. In Saint Cournot there is only one restaurant, Chez Bertrand.
Chez Bertrand intends to cater to tourists who spend their holiday in town. A priori, the tourists
cannot observe the quality of Chez Bertrands food. Tourists think it equally likely that Chez
Bertrand offers high-quality or low-quality food. Suppose that Chez Bertrands marginal costs
equal per meal. Tourists value good meals at > and bad meals at < . We assume
+
that 2
< , that is, without further information, tourists expect the restaurants quality to be
so low that the restaurant cannot attract them without making a loss.
It so happens that Chez Bertrand knows the quality of its food to be good. In fact, it would like to
charge a price close to for tourist meals so that it can make better use of its monopoly position
in Saint Cournot. However, this strategy can only be successful if the restaurant can convince the
tourists that its food is indeed worth paying for otherwise they will decide not to come at all.
Advertising that the restaurant serves high-quality food does not work because the restaurant
could make the same claim if it served low-quality food. So, Chez Bertrand should find a way to
send its potential customers an informative signal about the quality of the restaurant.
Now, travel guide The Rough Planet offers the restaurant the opportunity to advertise at a cost.
Suppose that advertisement costs are equal to per potential customer. If Chez Bertrand is a
good restaurant, consumers will return once. As a consequence, the restaurant is willing to
advertise (only) if < 2( ). In the case that Chez Bertrand offers low-quality food, after
one meal consumers will learn of the restaurants actual quality inducing them not to come back.
Therefore, the restaurant will not advertise if the price-cost margin for a single visit is not
sufficient to cover the advertising costs, i.e., if < . Indeed, if for the advertising costs it
holds true that < < 2( ), Chez Bertrand will only advertise if it serves high-
quality food. Tourists can now read the fact that Chez Bertrand advertises as a credible signal
that its food is good. Interestingly, the advertisement itself need not contain any useful
information: Saint Cournot is a small town, so tourists will find out about the existence of Chez
Bertrand, ad or no ad. The only information that an advertisement reveals to tourists is that Chez
Bertrand must be a fine restaurant because otherwise it would not have spent money on
advertising. In other words, the ad persuades consumers to go to the restaurant.
145

lOMoARcPSD|1733731
Persuasive advertising is particularly relevant for what economists call experience goods.
An experience good is a good of which the consumer only observes the quality after
consumption. Besides meals at a restaurant there are many examples of experience goods
including wine, movies, pop concerts, second-hand cars, and lectures. In contrast, search goods
are goods for which the quality is observable before consumption, such as bicycles, clothes, and
furniture. For search goods, it does not make sense for the seller to burn money. Indeed,
advertising for search goods is purely informative.
How often do we see persuasive advertising in practice compared to informative advertising?

The answer is: surprisingly often! Consider commercials on television. How many commercials
do you find informative in the sense that you learn about the existence or the characteristics of a
product? For example, what do you learn from a washing powder producer advertising that its
product cleans your clothes better than its competitors while preserving the colors? Indeed,
many commercials are persuasive ads for experience goods. This is no coincidence. According to
one empirical study, the advertising/sales ratio (advertising expenditures as a fraction of total
sales) for experience goods is three times higher than for search goods.
13.4 Case study: How to distinguish a good seller from a bad one on eBay
If you have ever bought something on eBay, one of the worlds largest online auction websites,
you know that it is not always easy to recognize a good seller. That is, using Akerlofs wording, it
is not easy to distinguish between lemons and high-quality products. The adverse selection
problem you have studied in this chapter is further exacerbated in online transactions by the
fact that the buyer is rarely able to check the product in person. Yet, the market of online
auctions has experienced a strong and fast growth over the past 15 years, despite what the
economic theory would suggest.
Gregory Lewis, professor of economics at Harvard University, has studied how the adverse
selection problem in online transactions is mitigated by disclosing some sellers private
information through pictures and descriptive text of the product added on the webpage of the
online auction. If the seller cannot lie about the products quality, the pictures and text can be
seen as a credible signal about the quality of the sellers good. In addition, if the disclosure is
costly, one would expect only sellers of high-quality products to provide rich additional
information.
Lewis focuses his attention on eBay Motors, the largest used car marketplace in the US. On eBay
motors, 36,000 cars are sold each month, which corresponds to almost one car sold every
minute. Most of the vehicles are sold to out-of-state buyers who, hence, cannot examine the cars
and must rely on the description provided by the sellers. However, since most of the buyers opt
to pick up the vehicle in person and material misrepresentations by the seller constitute fraud,
Lewis argues that enforceable contracts are present in the market and the sellers have little
incentive to lie. In addition, optional descriptive elements such as pictures cost $0.15 each, but
the opportunity cost is much higher since uploading photos, generating graphics, and writing
text are all time-consuming activities.
In his analysis, Lewis takes into account another important factor in online auctions: the sellers
reputation. For each seller, he observes all the feedback left by previous buyers, which is
commonly used as a reliability parameter by future buyers. His dataset counts over 80,000
observations of 18 vehicle types. Lewis focuses on the information that the sellers voluntarily
146

lOMoARcPSD|1733731
disclose to the buyers. He uses two simple measures: the number of pictures on an auctions
webpage, and a text analysis that recognizes when keywords such as rust or scratch are used
in a negative sense. For example, he is able to distinguish between full of rust and rust-free.
Lewis results are quite surprising: on average, one more picture added in the auctions webpage
increases the final price by 1.54%, roughly $171 more for an average car of the dataset.
Therefore bidders do rely heavily on pictures to decide their purchases. Lewis also shows how
sellers of high-quality cars provide a lot of pictures and detailed descriptive texts. In contrast,
sellers of low-quality cars provide fewer pictures and minimally descriptive webpages. In
addition, Lewis analysis demonstrates that disclosure costs affect the level of disclosure and
hence have a causal effect on the price levels.
In conclusion, this study shows that certain kinds of information asymmetries can be resolved
by the agents in the market, so that adverse selection is mitigated and the market equilibrium is
more efficient than Akerlofs theory might suggest.
As said, we have studied the contributions of three great economists: George A. Akerlof, A.
Michael Spence and Joseph E. Stiglitz. The Swedish Riksbank Prize in Economic Sciences in
Memory of Alfred Nobel was jointly awarded to them in 2001 for their analyses of markets with
asymmetric information.
We started with George A. Akerlofs study of markets where sellers of products have more
information than buyers about product quality. The results are illustrated by a simple example
in the first section of this chapter. The primary reference is his Market for 'Lemons': Quality
Uncertainty and the Market Mechanism, published in 1970 in the Quarterly Journal of
Economics. It is ironic that Akerlof was working on that paper since his first year as an assistant
professor at Berkeley and his work was turned down for publication three times. The American
Economic Review, the Review of Economic Studies and the Journal of Political Economy all rejected
his work on the grounds that they do not publish papers on subjects of such triviality. Akerlof
himself cites Samuelson (1938) and Schultz (1963) as works that facilitated the development of
a formal theory of asymmetric information in the market for lemons.
Spence published his Job Market Signaling model in 1973. Unlike the example given in the
chapter, he focused on education as a signaling device and he showed how agents could improve
the market outcome (equilibrium wage) by taking costly action to signal information to poorly
informed recipients. As mentioned in the chapter he showed that education does not need to
have intrinsic value; costly investment in education as such can signal high ability. You should
also refer to his updated work on job signaling published in 2002 in the American Economic
Review. Nelson (1974) provides a seminal study on the signaling properties of advertising.
Finally, Joseph Stiglitzs work on screening has provided valuable insight in many observed
market phenomena, including unemployment and credit rationing. For the scope of the material
covered in this chapter it is sufficient to refer to Greenwald, Stiglitz and Weiss (1984) and
Rothschild and Stiglitz (1970, 1971, 1973).
The case study of this chapter is based on Lewis findings from his 2011 article in the American
Economic Review.
147

lOMoARcPSD|1733731
References
Akerlof, G.A. (1970). The market for 'lemons': quality uncertainty and the market mechanism,
Quarterly Journal of Economics, 84(3), 488-500.
Lewis, G. (2011). Asymmetric information, adverse selection and online disclosure: the case of
eBay motors, American Economic Review, 101(4), 1535-1546.
Nelson, P. (1974). Advertising as information, Journal of Political Economy, 82(4), 729-754.
Rothschild, M. and J.E. Stiglitz (1970). Increasing risk I: a definition, Journal of Economic Theory,
9, 2(3), 225-243.
Rothschild, M. and J.E. Stiglitz (1971). Increasing risk II: its economic consequences, Journal of
Economic Theory, 3, 5(1), 66-84.
Rothschild, M. and J.E. Stiglitz (1973). Some further results on the measurement of inequality,
Journal of Economic Theory, 6(2), 188-204.
Samuelson, P.A. (1938). Foundations of Economic Analysis, Cambridge, MA: Harvard University
Press.
Schultz, T.W. (1963). The Economic Value of Education, New York: Columbia University Press.
Spence, M. (1973). Job market signaling, Quarterly Journal of Economics, 87(3), 355-374.
Spence, M. (2002). Signaling in retrospect and the informational structure of markets, American
Exercises
Exercise 13.1 Adverse selection with a uniform distribution of types

KLM is considering taking over a small regional airline that operates flights to the Alps (Air Alp).
Air Alp knows its own value. However, KLM does not know Air Alps value; it believes that this
value, when Air Alp is controlled by its own management, is between 0 and 100 million, and
assigns equal probability to each value in this range (that is, the value of Air Alp is a draw from
the uniform distribution [0,100]). Air Alp will be worth 50% more under KLMs management
than it is under its own management. Suppose that KLM bids 0 to take over Air Alp, and Air
Alps value equals under its own management. Then if Air Alp accepts KLM's offer, KLM's
3
payoff is 2 and Air Alp's payoff is ; if Air Alp rejects KLM's offer, KLM's payoff is 0 and Air
Alp's payoff is . Assume that the managers of both firms maximize the expected profit.
a) How much will KLM offer in equilibrium and what is the lowest possible offer that Air Alp will
accept? Explain why the logic behind the (Nash) equilibrium is called adverse selection.
b) Explain intuitively what will happen to the Nash equilibrium in (a) if the managers of the two
firms are both:
i. Risk-averse
ii. Risk-seeking
Exercise 13.2 Inverse adverse selection: The market for peaches

In this exercise, we examine a marketplace where buyers are better informed than sellers. We
will see that this creates an inverse adverse selection problem: The market tends to disappear
from the bottom rather than from the top. In contrast to the traditional model, it is the high-
148

lOMoARcPSD|1733731
value goods (peaches) that are traded on the market, rather than the low-value goods (lemons).
Penlope would like to sell an old gem that she found in a jewelry box in her attic. However, she
is uncertain of its value. For her, the gem could be worth 1,000 (if its a diamond), 400 (if its a
sapphire) or 100 (if its an amethyst). All values are equally likely. Potential buyers are expert
collectors who can immediately identify the value of the gem. In answering the questions below
you may assume that Penlope knows the collectors valuations.
a) Suppose collectors value the three types of gems at 1,200, 800 and 300 respectively.
What is the lowest price at which trade will emerge? Is the market efficient?
b) Now answer question a) under the assumption that the collectors valuations change to
1,200, 600 and 300 respectively.
c) Propose and discuss solutions to the inverse adverse selection problem.
Exercise 13.3 Labor market signaling

Suppose that two types of workers exist, high- and low-ability workers. Workers know their
own ability, but the firm cannot observe a workers type. Suppose getting a degree is easier for a
high-ability worker than for a low-ability worker. The monetary equivalent of the utility loss
from obtaining a degree is 500 for a high-ability worker and 1,100 for a low-ability worker.
Assume that the lifetime market wage for workers with a degree is 3,000 and for workers
without a degree 1,500.
a) Will having a degree reveal a workers ability to the firm?
b) What happens if the lifetime market wage for workers with a degree decreases to 2,500?
c) Determine the threshold lifetime wage for workers with a degree below which no type of
worker would find it profitable to obtain the degree.
149

lOMoARcPSD|1733731
Chapter 14: Hiring and firing

When a firm has a vacancy, the firms management typically receives several applications. The
firm often finds itself in a situation of information asymmetry. Who is the best candidate?
Typically, applicants are better informed about their suitability for the job than the firm. An
adverse selection problem is imminent unless applicants can signal their ability to the firm or
the firm finds a way to screen the applicants. In this chapter, we will look at several signaling
and screening devices applicants and firms may employ to solve or attenuate the adverse
selection problem, including credentials (section 14.1), assessments (section 14.2), probation
(section 14.3), and incentive contracts (section 14.4). In this the case study in section 14.5, we
examine how incentive contracts served as a screening device in a car glass company.
14.1 Credentials
Credentials (such as degrees and experience in past jobs) can serve as signaling devices. Two
conditions must be met for credentials to work as a signaling device. First, the credential must
be related to the job. A college degree in economics will not help you much if you apply for a job
at a law firm. Second, it must not be too easy or too difficult to obtain the credential. If
credentials are easily obtained, everyone will get one, so that the credential does not signal
anything. In the other extreme of credentials that are extremely difficult to obtain, none of the
candidates will be able to show one. College degrees usually serve as excellent credentials (at
least if the area is relevant).
14.2 Assessments
When unsure about a job candidates suitability for the job, the firm can ask the applicant to
undergo a job assessment. In a job assessment, the applicant is asked to perform several job-
related tasks, usually over the course of a single day. Job assessments are more profitable the
better the information the firm obtains about the qualities of the applicant, the cheaper the
assessment, and the higher the stakes, i.e., the more necessary it is for the firm to get the right
man or woman for the job. The following example shows how this might work.
Suppose that the applicant is highly productive with probability and less productive with
probability 1 . A highly productive worker contributes to the firms profits, while a less
productive worker contributes < . Without assessment, the firm expects its profits to
increase by + (1 ) > 0, where is the workers wage. In other words, without
further information, it is attractive for the firm to hire the candidate. Now, let us assume that
< < so that only the highly productive worker is truly profitable. Suppose that the
assessment costs are equal to and that the assessment allows the firm to learn the productivity
of the worker with certainty, so that the firm only hires if the worker is of the highly-productive
type . Then it pays for the firm to organize an assessment (only) if
( ) > + (1 )
< (1 )( ).
So, the firm will engage in the assessment (only) if (1) is low, i.e., the assessment is not too
costly, (2) is low, i.e., it is unlikely that the firm attracts a highly productive worker, and (3)
is high, i.e., the firm makes a substantial loss if it hires a less-productive worker. Note that
150

lOMoARcPSD|1733731
we derived the inequality under the assumption that the assessment reveals the workers
productivity perfectly. The less accurate the assessment is, the more extreme the above three
variables must be for the assessment to be valuable.
14.3 Probation
Probation is another commonly used screening mechanism. In many jobs, workers are first
hired on a fixed-term contract before they receive tenure. The time served under the fixed-term
contract could be viewed as the probation. (Only) if the worker performs well, will the firm offer
him tenure. Probation is quite like an assessment. The main difference is that a probation period
often takes much longer than an assessment. For instance, in the academic world, the probation
period of professors may take as long as six year before they are tenured! The advantage of
probation over assessments is that the employer may get much better information about the job
candidate because he performs job-related tasks for a much longer time period. An obvious
disadvantage of probation is that it may be very costly for the firm because it has to pay the
candidates salary during probation even if it does not hire him afterwards. To contain the costs,
the firm may pay the worker a much lower salary than after tenure.
An employer may very well be able to screen out workers, even if the probation itself does not
work as a perfect screening device. Unskilled workers are so unlikely to get tenure after
probation that they have a strong reason not to take probation, in particular if the salary in the
probation period is low. Skilled workers, in turn, are likely to be tenured after probation so they
will take the firms contract, in particular if the salary is high in the event of tenure. Indeed,
probation can effectively separate skilled workers from unskilled ones even before the start of
the probation itself. Two conditions must be met: First, the salary under probation must be
below the market salary for unskilled workers. Second, the salary after probation must exceed
the market salary for skilled workers.
In the analysis so far we assumed that workers know their type. This is not always realistic. How
many young football players who get a contract from a professional club at the age of 16 do not
believe they are very talented and can become at least as good as Lionel Messi? A probation year
will help the club weed out most of the bad types. In the case of asymmetric information on both
sides, the question emerges whether firms should hire risky or risk-free workers.
The following example shows that firms may prefer hiring risky workers. Suppose football club
Real United can choose between two talents for a position in their youth training program. Let
us call them George and Michael. Michael is a sure bet: With certainty, he is a good player. His
yearly value for the club is equal to 2 million. George is more risky. With probability 50%, he is
extremely good (type Best) and with probability 50%, he is a mediocre player (type Clooney).
The club attaches a value of 4 million a year to type Best and zero to type Clooney. Assume that
the club will hire one player on a five-year contract that pays the player a 1 million annual
salary. Note that Real United expects both George and Michael to yield an annual net value of 1
million. The contract contains a clause that specifies that at the end of the first year, the club can
fire the player without further costs. Assume that after the one-year probation, the club knows
perfectly whether George is of the Best type or of the Clooney type.
So, why does Real United prefer to hire George, the risky player, over Michael, the risk-free
player? Let us have a look at the clubs expected profit from the two players. Michael generates a
guaranteed profit of 1 million per year, i.e., 5 million over the five-year period. The profits on
151

lOMoARcPSD|1733731
George depend on his type. Real United will fire George after the one-year probation if he turns
out to be type Clooney. In that case, its loss amounts to 1 million, Georges one-year salary. Real
will keep George if the probation year reveals he is the Best type. In that case, its profits are
equal to 15 million. Because George is of either type with equal probability, Reals expected
1
profits are 2 (1 + 15) million = 7 million. Indeed, it is more attractive for Real United to
hire George than to hire Michael.
Some cautionary remarks are appropriate. It is not always optimal to attract the most risky
worker. First of all, we assumed that the firm could keep the productive workers salary fixed at
a level well below its productivity. However, competitors may be willing to offer a productive
worker a much higher salary once it is clear that he is one. So, the firm may have to raise a
productive workers salary to retain him, which decreases the firms expected profit on a risky
worker. Second, the assumption that a firm can fire the worker without invoking costs is often
not realistic in practice. In most developed countries, firms must pay workers some form of
compensation when they let them go. Third, the attractiveness of hiring a risky worker depends
on a large range of factors. Hiring a risky worker becomes less interesting for a firm the lower
the workers upward potential, the higher his downward risk, the more risk averse the firm is,
the longer the probation period, and the shorter the period of employment after probation.
14.4 Incentive contracts

A firm may also use incentive contracts as a screening device. To show how this might work, let
us revisit the Principal-Agent game we introduced in chapter 2, in which agent Antonio directs a
movie on behalf of principal Penlope. Recall that each unit of Antonios effort yields one
additional movie theater ticket sold, i.e., Antonios output equals
= .
Antonios cost () of exerting effort is given by
() = 2
where > 0. There are no costs other than the costs of effort. Penlope offers Antonio a linear
contract which specifies that Antonio will receive a base wage plus a bonus for each movie
theater ticket sold. In chapter 2, we showed that Antonios optimal effort is given by

= .

Now, suppose that Antonio accepts the contract (only) if the utility he obtains from the contract
exceeds the utility he obtained from his most attractive outside option.
To show how an incentive contract may serve as a screening device, assume that there are two
types of Antonio: a productive one having cost parameter = 1/2 and an unproductive one
having cost parameter = 2. Assume that for both types of Antonio, the utility for the outside
option equals = 4 and the Penlope pays a fixed wage equal to = 2 and a bonus equal to
= 2. Antonios utility from Penlopes contract is therefore
2 2
= + () = + 2 = + =2+ .
2
152

lOMoARcPSD|1733731
The productive Antonio will accept the contract because
2
=2+ = 6 > = 4.

In contrast, the unproductive Antonio prefers the outside option over working for Penlope as
2
=2+ = 3 < = 4.

Indeed, the contract serves as a screening device: It weeds out unproductive Antonio types. This
chapters case study provides some empirical evidence that is in line with this result.
14.5 Case study: Performance pay as a screening device

In this chapter, we have argued that incentive contracts could serve as a screening device in that
they allow a firm to attract a better pool of applicants for the job. In this case study, we examine
an instance which shows that this conclusion is not only of theoretical interest. Stanford
University professor Edward P. Lazear used a dataset from Safelite Glass Corporation, a large
auto glass company in the US, to find that performance pay not only increases existing workers
productivity but also helps the firm separate high-quality from low-quality workers.
Lazear bases his analysis on data collected during 1994 and 1995, when the company gradually
changed the compensation scheme for its workers. After a change in management, Safelite
decided to switch from hourly wages to a piece-rate pay. Up till January 1994, workers were
paid an hourly wage rate for installing glasses on cars. The hourly wage offered was not largely
varied amongst glass installers and did not depend in any direct way on the number of windows
that were installed in an hour. During 1994 and 1995, glass installers were shifted from an
hourly wage scheme to performance pay and in particular to a piece-rate arrangement. Instead
of being paid for the number of hours that they worked, glass installers were paid for the
number of glass units that they installed.
The primary motivation behind introducing a piece-rate scheme is to increase worker effort.
Although it cannot be certain that effort will increase, it should not be expected to decrease
when the firm switches from hourly wages to piece rates. To fix the idea, suppose there are two
worker types: low-quality workers and high-quality workers. Usually, an hourly wage payment
scheme is coupled with a minimum level of output that is acceptable by the firm that is retained
either by contractual agreements or by peer pressure. Thus, given that all of the firms workers
satisfy that requirement (and are not fired) average effort will increase as long as at least one of
the worker types supplies more output. The potential of the low-quality worker may not change
as a result of the switch in payment scheme, but the potential of the high-quality worker rises
since a piece-rate allows her to work harder and receive more from the job. The low-quality
worker is indifferent between the two schemes. At the same time, Safelite may be able to attract
better applicants by switching to a piece-rate payment scheme. The reason is that high-quality
workers prefer working for Safelite because the new payment scheme is more attractive than
the fixed salary they obtain from other firms in the industry.
Lazear finds that Safelites employees became much more productive after the switch from
hourly wages to piece rates: Productivity for the entire company increased by no less than 44
per cent! This productivity gain can be split into two components. First, average output per
153

lOMoARcPSD|1733731
worker increased as a result of the higher incentives. The incentive effect, as Lazear labels it,
was responsible for about half of the productivity gain. The other half of the gain confirms the
ability of incentive pay to help in screening the candidates. After the switch, Safelite attracted
relatively more high-quality workers because of its appealing compensation scheme. Lazear also
observes that Safelites profits increased because of the new payment scheme. Workers pay
increased by about 10%. Of course, this does not necessarily entail that they benefitted as well
because they had to work harder. However, the fact that the firm selected so many high quality
workers suggests that they did.
Although this is a single example where performance pay was shown to work, the case for piece
rates seems especially strong. However, it is important to keep in mind the perverse effects of
monetary incentives on performance that we encountered in the case studies of Chapters 5 and
8. Perhaps the most important lesson from Lazears study is that incentives can sometimes work
very well. In the case of Safelite, factors that contributed to the successful implementation were
that output was easily measured, quality problems could readily be detected, and errors could
be attributed to individuals. As the other case studies suggest, other tasks may not be as suitably
appropriate for performance pay, such as tasks that require creativity. Therefore, the fact that
the productivity gains are so large in this case does not imply that firms should always switch to
piece-rate pay.
The screening and signaling devices discussed in this chapter are extensions and applications
suggested by different authors to the problems analyzed in the models of Akerlof (1970) and
Spence (1973, 2002). Therefore, it would be useful to review chapter 13 again and refer to the
original papers for a comprehensive treatment.
The structure of this chapter follows Edward Lazears (1992) examination of hiring and firing
decisions. Other useful resources are Becker (1988), Milgrom and Roberts (1992), and Baker,
Gibbs and Holmstrm (1994).
This chapters case study is based on a paper by Edward Lazear, which was published in 2000 in
the American Economic Review.
References
Akerlof, G.A. (1970). The market for 'lemons': quality uncertainty and the market mechanism,
Quarterly Journal of Economics, 84(3), 488-500.
Baker, G., M. Gibbs and B. Holmstrm (1994a). The internal economics of the firm: evidence from
personnel data, Quarterly Journal of Economics, 109, 881-919.
Baker, G., M. Gibbs and B. Holmstrm (1994b). The wage policy of a firm, Quarterly Journal of
Economics, 109, 921-955.
Becker, G.S. (1988). Human Capital: A Theoretical and Empirical Analysis with Special Reference
to Education, New York: Columbia University Press.
Lazear, E.P. (1992). Performance Measurement, Evaluation, and Incentives, William Bruns (eds.),
Boston: Harvard Business School Press.
Lazear, E.P. (2000). Performance pay and productivity, American Economic Review, 90, 1346-
1361.
Hall.
154

lOMoARcPSD|1733731
Spence, M. (1973). Job market signaling, Quarterly Journal of Economics, 87(3), 355-374.
Spence, M. (2002). Signaling in retrospect and the informational structure of markets, American
Exercises
Exercise 14.1 Probation

The pizza restaurant Napoletana currently has a weekly profit of 1,000, but it is looking for a
new chef to attract more and more people to the restaurant. There are two candidates for the job:
Chris and Tony. The latter can prepare a super tasty pizza Margherita that will surely increase
the restaurants weekly profit by 1,200. Chris, on the other hand, is an eccentric cook who loves
to mix flavors from different traditional cuisines; unfortunately for him, the customers may not
like his creations at all. Chris can increase the restaurants weekly profit by 2,000 with a
probability of 2/3, or decrease the restaurants profits to zero with probability of 1/3. After one
week of work, it will be clear whether the customers like Chris pizzas or not. If hired, Tony and
Chris would receive the same wage. Assume that the restaurant stays in the market for a long
time (let us say 10 years). The restaurants weekly discount factor is negligible and the
restaurant is risk-neutral.
Suppose the restaurant has a one-week probation period, during which time it can hire only one
cook. Moreover, assume that if its clients do not happen to like Chris pizza, the firms reputation
will be so damaged that it has to leave the market.
a) Which cook will the restaurant hire? Would your answer change if the restaurant were risk-
averse?
Assume that after the first week the restaurant can replace the cook with the other one without
incurring any costs.
b) Which cook is the restaurant going to hire in the first week? Would your answer change if the
restaurant were risk-averse?
Now, assume that the restaurant is in the market for only two periods. There is a cost to
replace the cook after the first period.
c) Find the cost for which the restaurant is indifferent between hiring Chris and Tony.
Exercise 14.2 Probation

The University of Amsterdam (UvA) is looking for new talented PhD students. There are two
types of students, Good and Bad. The university offers a three-year PhD program during which
both types of students earn 30,000 per year, after which the university would hire the student
for an additional three years as a assistant professor, with a salary of 80,000 per year if the
student turns out to be Good after her PhD, or 70,000 per year if the student turns out to be
Bad. Instead of writing a PhD thesis, both types of students could become bartenders for six
years at the bar Kriterion, earning 50,000 a year. All the students have an annual discount
factor of = 0.9. Students have a limited time horizon in the sense that they do not care about
potential earnings after six years.
155

lOMoARcPSD|1733731
a) Which student types will write a PhD thesis? Does the low salary for PhD students serve as a
screening device?
The University of Amsterdam (UvA) is also looking for a new professor. A good professor could
increase the universitys fame, attracting additional funds. A bad professor might damage the
universitys reputation. Diederik has applied for the position at the UvA. Diederiks type is
unknown to the university, but with his credentials the university estimated the following table:
Type Value (Millions) Probability
Einstein 10 0.01
Nash 6 0.04
Stiglitz 2 0.15
Sonnemans 0.2 0.20
Stapel -2 0.60
The UvA can offer a two-year contract with a year of probation, after which the professors type
will be known. The contract specifies that the university can unconditionally fire the professor
after the first year of probation. Each year including the probation period he will be paid
300,000. The first year returns are zero for all types. UvA is risk neutral and the values contain
all the relevant costs and benefits for the UvA after the probation year.
b) Will the university offer the contract to Diederik?
Exercise 14.3 Screening

In ancient times, before the invention of printing, books and writings were reproduced by hand,
typically by monks. In the Middle Ages, university students also transcribed books to earn some
money.
Assume you are a nobleman in the 15th century who wants to get his family tree transcribed by
some university students. In the city there are two types of students: the Fast ones, who can
transcribe nine pages per hour, and the Lazy ones, who can transcribe only four pages per hour.
Instead of being employed as reproducer, the Fast types could earn 27 silver coins per hour by
working in one of the citys workshops, and the Lazy ones could harvest the crop, earning 14
silver coins per hour.
a) If you could readily observe the types of students, which ones would you hire in order to
minimize your expenditures?
b) Assume that you do not observe the students types. Which piece-rate salary would you pay
to make sure that only Fast types would accept the job?
Suppose the reproducers of the city organize a guild, similar to what we would call a union
nowadays. The reproducers guild imposes a minimum piece-rate wage of four silver coins.
c) Could you still attract only the students that you prefer? How?
A fire has destroyed half of the crop, pushing down the salary for harvesting to 10 silver coins
per hour.
d) Would this change your answer in part a) and b)?
156

lOMoARcPSD|1733731
Exercise 14.4 Incentive contracts

Lehman Sisters is a new investment bank looking for new brokers to hire. The brokers job will
be to contact people by phone and convince them to buy shares of Scottish Petroleum, a
company on the edge of bankruptcy. There are 100 broker types in the market, =
1.01, 1.02, ,2. Assume that each type is equal and that Lehmann Sisters cannot observe a
brokers type. Each broker can convince four persons a day by putting one unit of effort (), with
a cost of effort equal to (, ) = 8 2 . Lehman Sisters offers a linear contract with a fixed wage
() and a bonus () for each customer. All the brokers can be hired by a competitor investment
bank for 22 a day.
a) Find the brokers optimal effort and discuss its relation with .
b) Lehman Sisters wants to hire only the 50% most efficient brokers. Given a fixed salary of =
10, how can Lehman Sisters accomplish its purpose?
c) What if Lehman Sisters wanted to hire the 20% most efficient brokers?
d) Given the answers in part b) and c), what do you conclude?
e) Given the bonus as in part b), if the competitor investment bank increased its daily salary to
64, how many brokers would leave Lehman Sisters?
157

lOMoARcPSD|1733731
Chapter 15: Price discrimination

In this chapter, we study how firms can increase profits by employing price discrimination, i.e.,
charging different prices for the same good. Figure 15.1 indicates why a firm may have an
incentive to price discriminate. It shows that a monopoly firm leaves quite some money on the
table when charging a uniform price. First of all, consumers who buy the firms product at the
monopoly price gain some consumer surplus. Moreover, consumers might be willing to buy
more of the firms product if the price were lower. As we saw in chapter 3, the monopoly price is
high in the sense that it exceeds marginal costs so that several value-enhancing transactions do
not take place. In this chapter, we will show that price discrimination allows the firm to reap
some (and in some cases all) of this untouched surplus. In particular, we will discuss pricing
schemes that allow the monopolist to screen consumers with the high willingness-to-pay from
those with a low willingness-to-pay.
Figure 15.1: Surplus a monopolist leaves on the table in the case of a uniform price
The set-up of this chapter is the following. In section 15.1, we will discuss the benchmark case in
which a firm is completely informed about each and every customers willingness-to-pay for its
products. In section 15.2, we will consider settings where the firm can only observe the
aggregate demand of consumer groups rather than individual consumers. In section 15.3, we
will analyze how a firm can price discriminate in settings where it is asymmetrically informed
about consumers willingness-to-pay. In the case study in section 15.4, we will study how a seller
can implement a successful price discrimination strategy by offering damaged goods.
15.1 First-degree price discrimination

We start by discussing an extreme setting where a monopolist is completely informed about all
consumers willingness-to-pay for its products. In such a setting, the firm may be able to charge
each individual consumer a different price. In the economics literature, such a pricing strategy is
known as first-degree price discrimination. Other commonly used (and more intuitive)
terms are personalized prices (because each person potentially pays a different price) and
perfect price discrimination (because first-degree price discrimination allows the firm to
extract all potential surplus from the market, as we will see below). While first-degree price
158

lOMoARcPSD|1733731
discrimination is rare in practice, analyzing it serves as a useful benchmark for the practically
more realistic settings with asymmetric information that we will study in the following sections.
Let us discuss two examples to illustrate how a firm can implement first-degree price-
discrimination. We start with a setting where each consumer demands at most one unit of a
good, the single-unit demand case. Consider software company Cherry. Suppose that the firm
can observe each individual s value for its software package. Clearly, the firm prefers not to
sell to customers whose value is below its marginal costs because in that case the firm makes a
loss on that individual. So, if Cherrys marginal costs are constant and equal to , it can maximize
profits by charging each consumer a price equal to the consumers value as long as this value
exceeds the firms marginal costs. In a simple formula, Cherrys profit-maximizing personalized
price equals
for all individuals for whom > . (To ensure that individuals strictly prefer buying the
software over not buying it, Cherry can give a one cent discount.)
The second example concerns multi-unit demand. Mobile telecommunications company WorTel
caters to a set of individuals, each having a potentially different downward sloping demand
curve for (minutes of) mobile telecommunication services. WorTels marginal costs are constant
and equal to . If WorTel can observe each individuals demand curve, it can maximize its profits
in the following way. For each minute called, it charges each consumer a price equal to his value
for that minute as long as this value exceeds marginal costs. The firm can implement this scheme
using what is called a two-part tariff: WorTel charges each consumer a fixed fee (or
subscription fee) for the possibility to use its services, independent of consumption, and a
price per unit of consumption (e.g., minutes called). The surplus maximizing scheme
guarantees that consumers buy an additional unit as long as their value for that unit exceeds
marginal costs. The firm implements this scheme by putting the unit price equal to marginal
costs, i.e.,
= .
The profit-maximizing fixed fee ensures that the firm captures the consumer surplus it generates.
Therefore, for each individual , the optimal fixed fee equals his consumer surplus at the
marginal cost unit price:
= ( ) = ().
(Again, to ensure that individuals strictly prefer buying over not buying, WorTel may give a one
cent discount.)
Observe that in the two examples, welfare is maximized: Both firms serve each unit of demand
that generates more consumer value than the cost of producing it. Of course, welfare is divided
quite unequally between producers and consumers. The firms are able to extract all surplus
from the market leaving nothing (or a few cents) for consumers.
Why is first-degree price discrimination rare in practice? The reason is that at least four
conditions must be satisfied for a firm to be able to implement it. First of all, that a firm can
distinguish consumers perfectly. Second, that a firm can observe each individuals value or
159

lOMoARcPSD|1733731
demand curve. Those two conditions hardly ever apply in practice. In the following sections, we
will study what the effect is of relaxing them. A third crucial condition is that the firm has a
monopoly position. In the case of competition, profitable price discrimination is more difficult if
not impossible. Indeed, another firm can attract all consumers by offering a small discount
relative to the optimal prices so that the proposed pricing schemes do not maximize profits. A
fourth condition is the absence of resale opportunities, or arbitrage in economics parlance.
Price discrimination may fail miserably if the firm cannot prevent arbitrage. In the mobile
telecommunications example, one individual could pay the fixed fee to WorTel and buy a great
number of units at the marginal cost price, reselling them to other consumers at a discount
relative to WorTels price. Of course, in practice, firms can often make arbitrage difficult or
impossible. In the case of mobile telecommunications, selling calling minutes to other consumers
is feasible in principle: You can lend your SIM card to someone else and charge him or her by the
minute. However, this is clearly quite cumbersome and I do not expect many people to engage in
this type of arbitrage. Indeed, mobile telecommunication subscriptions typically involve two-
part (or multi-part) tariffs. In other markets, firms take actions to discourage or block arbitrage.
For example, airline tickets carry a persons name so that the airline makes arbitrage from one
passenger to another impossible.
15.2 Third-degree price discrimination

In this section, we relax the assumption that a firm can observe each individuals demand for its
product. Instead, we assume that the firm only knows the aggregate demand of several separate
consumer groups. We maintain the other assumptions: The firm can charge different prices to
different individuals, it has a monopoly position, and arbitrage is impossible. Because the firm
can observe the aggregate demand among different consumer groups, it makes sense to charge
members of different groups different prices. Indeed, if the firm can identify which consumer
belongs to which consumer group, it can treat each consumer group as a different market. This is
called third-degree price discrimination or group pricing (because the firm charges
different prices to different consumer groups). Third-degree price discrimination is common in
practice: children get discounts on transport and in museums, students pay less for books and
meals, firms charge different prices in different countries, and so on.
If all consumers demand at most one unit of the firms product, the firm optimally charges each
consumer the monopoly price related to its consumer groups aggregate demand. Consistent
with our analysis in chapter 3, you can find consumer group s monopoly price by applying the
inverse elasticity rule:
1
= .

The welfare consequences of third-degree price discrimination are ambiguous. It can be shown
that third-degree price discrimination decreases welfare compared to a uniform price if the total
quantity sold is the same in either case. We do not prove this here, but the inverse elasticity rule
provides some intuition. The price-cost margin (as measured by the Lerner index ) is relatively
high in markets with a low price elasticity , i.e., roughly, where consumer demand is high. If a
uniform price guarantees that the firm sells the same total quantity, the discriminatory price in
high-demand markets must be higher than the uniform price while the discriminatory price in
low-demand markets must be lower. As a consequence, price discrimination substitutes some
consumers in the high-demand market for consumers in the low-demand market. Therefore, the
160

lOMoARcPSD|1733731
market generates less value in the case of price discrimination because the dropped consumers
are willing to pay more for the firms products than the new consumers. In contrast, third-
degree price discrimination may increase welfare if total quantity increases relative to a uniform
price. For example, welfare may increase if the firm caters market segments under price
discrimination that would be left untouched in the case of a uniform price.
15.3 Second-degree price discrimination

In this section, we relax another crucial assumption we made in our analysis of first-degree price
discrimination: The firm is able to charge different prices to different consumers. In practice,
this is often difficult because the firm cannot identify which consumer group a consumer belongs
to or because it is illegal according to anti-discrimination laws. Still, the firm may be able to price
discriminate in a subtle way by offering consumers a menu of pricing schemes from which they
can choose. The idea is that different consumer types will choose a different pricing scheme so
that, implicitly, the firm is able to price discriminate. This screening mechanism is called
second-degree price discrimination or menu pricing (because consumers can choose from
a menu of pricing schemes).
Second-degree price discrimination comes in many shapes and sizes. Two-part tariffs, multi-
part tariffs, and quantity discounts are all examples of second-degree price discrimination
where the unit price a consumer pays depends on the quantity he buys. The firm screens
consumers by letting them choose the quantity they purchase. Low-demand consumers buy very
little while high-demand consumers buy a high quantity, usually at a lower price per unit than
low-demand consumers.
To give an idea about the profit-maximizing two-part tariff, let us assume that the market
consists of two segments. Within a segment, all consumers have the same downward sloping
demand curve which differs between the two market segments. Suppose that the two demand
curves are not too far apart but that the demand curve in one market is always strictly above the
one in the second market. It can be shown that the optimal two-part tariff features a unit price
between marginal costs and the monopoly price. Moreover, it has a fixed fee equal to the
consumer surplus of the low-demand consumers. This scheme differs in two ways from the
profit-maximizing two-part tariff under first-degree price discrimination: The unit price is
higher and the fixed fee is lower, in particular for high-demand consumers. The firm distorts the
scheme for two reasons. First, it lowers the fixed fee to ensure participation by low-demand
consumers. Second, it increases the unit price to increase the profit margin it obtains on the
high-demand consumers. Clearly, the firm leaves some money on the table but it is still better off
than if it charges a uniform unit price. Note that price discrimination increases welfare
compared to a uniform price because the optimal unit price is lower than the monopoly price.
The effect on consumer surplus is ambiguous. Price discrimination makes the consumers in the
low-demand market clearly worse off because the firm uses the fixed fee to absorb their entire
surplus. High-demand consumers may benefit in particular cases.
Quality discrimination is another example of second-degree price discrimination. Practical

examples of this type of price discrimination include economy versus business class in airplanes,
first versus second class train tickets, and paperbacks versus hard-cover books. Table 15.1
presents an example of a setting where quality discrimination can be profitable for a firm. An
airline offers a daily service between London and Kuala Lumpur using a plane that can carry 300
persons. It caters to two kinds of consumers: tourists and business people. Assume, for
161

lOMoARcPSD|1733731
simplicity, that the marginal costs per passenger are zero, independent of class (an assumption
that is probably not too far from reality as most costs are fixed, including depreciation costs for
the plane, personnel costs, and fees for landing slots). Table 15.1 indicates the willingness-to-pay
for the two classes for business people (100 in total) and tourists (200 in total).
Table 15.1: Willingness-to-pay for airline tickets
Willingness-to-pay
Number Economy class Business class
Business people 100 1,100 2,000
Tourists 200 800 1,000
For the moment, consider the case that the airline can distinguish between the two types of
consumers. What is its optimal strategy in terms of price/quantity pairs? Because both business
people and tourists are willing to pay the highest amount for business class, the firm will only
offer business class and it will do so at a price equal to the willingness-to-pay (minus one cent)
for each consumer type, i.e., 2,000 for business people and 1,000 for tourists.
In reality, the airline does not know whether someone who buys a ticket is a business person or
a tourist. What is the optimal menu of price/quality pairs? Table 15.2 contains an overview of all
natural candidates for a profit-maximizing strategy. (Why can we exclude other price/quality
pairs as being profit maximizing?) It shows that it is optimal for the firm to offer economy class
tickets for a price of 800 and business class tickets for a price of 1,700. Note that information
asymmetry forces the airline to move away from the first-best optimum in two ways. First, it
distorts quality downward for tourists. By doing so, the firm creates the opportunity to screen
its passengers: Tourists and business people fly different classes. It offers tourists lower quality
not because it wants to hurt them but to be able to offer business class for a high price to
business people. Second, it decreases the price for business class from 2,000 to 1,700. It has to
do so to prevent business people from choosing economy class, which would give them a net
surplus of 300 at the price of 800. Note that price discrimination reduces welfare compared to
the case where the airline can offer only one price/quality pair. In the latter case, it will
optimally offer business class for a price of 1,000. While the possibility to price discriminate
increases the firms profits by 30,000, it reduces consumer surplus from 100,000 to 30,000.
Table 15.2: The profitability of several price/quality pairs
Economy class Business class Profits

800 1,700 330,000
Not available 1,000 300,000
In contrast, welfare may increase when price-discriminating firms serve market segments they
would leave untouched if they only offered one price/quality pair. To see why, let us reconsider
our airline example. Now, suppose that 75 tourists (rather than 200) are interested in a flight
between London and Kuala Lumpur (see table 15.3). You can verify that the airline optimally
sells business class tickets for a price of 2,000 if it is restricted to selling only one type of ticket.
As table 15.4 shows, the profit-maximizing menu of price/quality pairs is the same as the one
above. Note that this menu increases both producer surplus and consumer surplus compared to
162

lOMoARcPSD|1733731
the case where the firm only sells business class. Welfare is higher indeed. In the spirit of this
example, this weeks case study shows that in some markets, firms sell deliberately damaged
goods in order to be able to cater to new market segments.
Table 15.3: Willingness-to-pay for airline tickets
Consumer type Number Willingness-to-pay

Economy class Business class
Business person 100 1,100 2,000
Tourist 75 800 1,000
Table 15.4: The profitability of several price/quality pairs
Economy class Business class Profits

800 1,700 230,000
A firm may also price discriminate by bundling products: The firm offers a package of two or
more products for a single, usually discounted, price. We distinguish pure bundling and mixed
bundling. In the case of pure bundling, the firm only sells the bundle, not the separate products.
We speak of mixed bundling if the firm offers both the bundle and the separate products. To
show why the firm may have an incentive to bundle products, let us consider fixed
telecommunications firm iPear. iPear sells two products: (fixed) telephony and internet services.
Table 15.5 shows the willingness-to-pay for telephony and internet services for three types of
consumers: Frequent callers, computer users, and general users.
Table 15.5: Willingness-to-pay for telephony and Internet services
Consumer type Number Willingness-to-pay

Telephony Internet
Frequent callers 2,500,000 100 0
Computer users 2,500,000 0 100
General users 1,500,000 60 60
Table 15.6: Pricing schemes
Pricing scheme Telephony Internet Bundle Profits

Telephony &
Internet
No bundling 60 60 120 480,000,000
No bundling 100 100 200 500,000,000
Pure bundling Not available Not available 100 650,000,000
Pure bundling Not available Not available 120 180,000,000
Mixed bundling 100 100 120 680,000,000
Table 15.6 contains several pricing schemes that iPear may offer. Those pricing schemes are
borderline cases in the sense that the firm reaps the total willingness-to-pay of at least one
163

lOMoARcPSD|1733731
consumer group. Consequently, those pricing schemes are natural candidates for the profit-
maximizing one. It turns out that in this example, the firm benefits from bundling its products.
By only selling a bundle of telephony and internet services at a price of 100, iPear increases its
profits compared to the case where it only sells services separately. So, pure bundling is
profitable. The firm can even do better by offering a mixed bundle. In fact, by doing so, the firm is
able to reap the entire potential consumer surplus. Note that the possibility to bundle increases
welfare: If iPear does not bundle its services, it blocks value-enhancing transactions by not
catering to general users. Of course, the welfare distribution in the case of mixed bundling is
quite skewed in the sense that the firm gets the entire surplus.
15.4 Price discrimination and competition policy

The analysis in this chapter indicates that the effect of price discrimination on welfare is
ambiguous. Clearly, firms can make sure they do not end up worse off: They always have the
option not to price discriminate. We saw that consumers may or may not benefit from price
discrimination. On the one hand, consumers may benefit when a price discriminating firm sells
to market segments that it would not serve if it could not price discriminate. On the other hand,
price discrimination may allow companies to distill a substantial fraction of consumer surplus.
Moreover, firms may have an incentive to distort their quality downward when price
discriminating, which could harm consumers as well.
According to this analysis, the question of what is sensible competition policy with respect to
price discrimination arises. In the European Union, price discrimination falls under Article 102
of the Treaty on the Functioning of the European Union, i.e., it may be interpreted as an abuse of
a dominant position. To refresh your memory, Article 102 reads:
Any abuse by one or more undertakings of a dominant position within the internal market
or in a substantial part of it shall be prohibited as incompatible with the internal market in
so far as it may affect trade between Member States.
Such abuse may, in particular, consist in:
(a) directly or indirectly imposing unfair purchase or selling prices or other unfair trading
conditions;
(b) limiting production, markets or technical development to the prejudice of consumers;
(c) applying dissimilar conditions to equivalent transactions with other trading parties,
(d) making the conclusion of contracts subject to acceptance by the other parties of
Three of the four cases may refer to price discrimination. The most obvious one is Article 102(c).
Price discrimination could be interpreted as an application of dissimilar conditions to
equivalent transactions. Article 102(a) relates to price discrimination as well in the case when it
results in excessive prices in particular market segments. Finally, (pure) bundling falls under
Article 102(d) because a firm obliges other parties to buy supplementary products even if they
are only interested in a single product in the bundle.
164

lOMoARcPSD|1733731
The landmark price discrimination case in European competition law is the United Brands case
that took place in the 1970s. United Brands sold bananas to wholesalers from various European
Union member states at significantly different prices. These price differentials were not the
result of cost differences or quality differences so it appeared to be a clear-cut case of price
discrimination. Moreover, United Brands took measures to prevent arbitrage by incorporating a
clause in its sales conditions which prohibited trade of bananas between countries. The
European Commission decided that United Brands abused its dominant position by its policy or
price discrimination based on Article 102(c).
Both lawyers and economists were seen to criticize the Commissions decision. Legally, the
conditions of Article 102(c) did not seem to apply, particularly the condition that other trading
partners be placed at a competitive disadvantage. Still, United Brands practice might call for
intervention on the basis of Article 102(a) if banana prices were excessively high in some
member states. Economists, in turn, argued that the welfare consequences of price
discrimination across member states are ambiguous in general. It was not clear that United
Brands pricing policy would harm welfare.
15.5 Case study: Damaged goods Lets make things worse

Computer company IBM announced in May 1990 the introduction of LaserPrinter E, a low-cost
version of its popular LaserPrinter. As it turned out, the two printers were substantially identical
in all their parts, with just one exception: IBM added more chips and processes in the
LaserPrinter E that served as idlers. Essentially these chips had no function other than to slow
down the printer. As a result, the low-cost LaserPrinter E printed an average of five pages per
minute, as opposed to the ten pages per minute of the original LaserPrinter. The LaserPrinter E
is a perfect example of what economists call a damaged good, a good which is intentionally
worsened by the producer in order to enhance price discrimination.
Besides the LaserPrinter E, it is easy to find many more examples of damaged goods. In
particular, they are routinely used by global companies in the software and hardware markets.
When Sony first released the PlayStation 3 console, both the 20GB and 60GB versions had the
components needed to play HD Blu-ray discs, but Sony deliberately removed from the 20GB
version an output connector needed to display HD images, so that the Blu-ray functionality could
not be utilized. Software companies such as Microsoft routinely offer reduced-price, light
versions of software packages with certain features disabled, even though the investment to
develop these features has already been made and the additional marginal cost of including
them is typically nonexistent or negligible.
As damaged goods appear to abound, two questions beg for an answer: Is a strategy of selling
damaged goods profitable? And, what are the welfare consequences of damaged goods?
Raymond Deneckere of the University of Wisconsin-Madison, and Preston McAfee of the
University of Texas at Austin answer both questions in a theoretical study published in 1996. To
start with the second question, Deneckere and McAfee showed that selling damaged goods can
be welfare improving. More in particular, all consumers can benefit from this type of price
discrimination. Consumers that would only buy the damaged good are clearly better off. Those
who have a strong preference for the high-quality good may benefit as well because the firm may
have to offer the original good at a discount to prevent them from buying the damaged good.
With respect to profitability, the possibility to sell damaged goods cannot decrease a firms
optimal profit for the simple reason that the firm can always decide not to sell them.
165

lOMoARcPSD|1733731
With regard to profitability, you might guess that when so many companies purposely damage
their products and sell them on the market, it has to be a profitable strategy. However, McAfee
(2007) shows in a more recent empirical study that certain conditions must hold in order for the
strategy to be profitable. A simplifying assumption can be made so that there are only two types
of consumer: high-value consumers, who will consider buying the full-price, original good, and
low-value consumers who will not consider the original good but will consider a discounted,
damaged good. The profitability of introducing a damaged good requires a large enough
difference in the preferences of two consumer types, so that a damaged good can be appealing
enough to attract low-value consumers, while at the same time high-value consumers will not be
tempted to switch and buy the damaged good instead of the original good.
An example from McAfees article helps to illustrate this point. It also shows that damaging
goods does not necessarily require specialized engineering as seen in the previous cases:
sometimes a simple trick serves the purpose. In recent years, electronics company Sharp was
selling two models of a DVD player, the DVE611 and the DV740U. The main difference between
the two DVD players was the ability of the DVE611 to play European (PAL) DVDs as well as
North American (NTSC) DVDs. Sharps DVD player DV740U was only able to play North
American DVDs, until someones curiosity piqued them to lift the cover of its remote.
Surprisingly, a button was hidden under the cover, a button that allowed the DVD player to also
read PAL DVDs. Sharp had hidden the button to price discriminate the two DVD players.
It is quite plausible that, to many US consumers, the ability to play European DVDs may have
seemed an unnecessary addition that they would happily forego in order to pay a lower price.
However, we can also speculate that there is another, quite distinct consumer group of wealthier
(i.e. high-value) consumers who frequently travel to Europe and who would therefore
appreciate the additional feature. Intuitively, the damaged goods strategy may well have been
profitable in this case. However, in other instances which do not fit this pattern, damaged goods
might backfire and bring a loss to the company. McAfee cites one example which is difficult to fit
into his theoretical framework: even the most basic versions of Microsoft Office include
programs such as Excel and Powerpoint, which are traditionally associated with business use
rather than home use. The suggestion here could even be that Microsoft is not damaging the
good enough for home (low-value) consumers, while business users could still be tempted to
buy the more basic packages.
Finally, there is one element which is absent from the above discussion: the consumer response
to the fact that a firm uses a damaged goods strategy. This factor is the subject of an article by
Gershoff, Kivetz and Keinan (2012), who examine the possibility that the use of damaged goods
strategies (which they refer to as versioning) can affect consumer perceptions of unfairness.
They make the point that the internet and social media make it easier than ever before to find
out about firms production processes and to disseminate such information. Consumers may feel
a sense of unfairness upon discovering that a product they have purchased was deliberately
worsened by the producer. Any sense of unfairness is likely to be stronger when the producer
has actually incurred additional costs in order to damage a good, for instance by adding extra
computer chips to an existing printer in order to sell a similar, but slower, version.
While it is difficult to quantify such effects, the study uses consumer surveys to show that there
is indeed a negative response to the practice of selling damaged goods and that this can force
consumers to consider switching away from the firm and towards its competitors instead, thus
166

lOMoARcPSD|1733731
highlighting another way in which the strategy can backfire. There are also real-world examples
to show that the negative consumer response can be strong and widespread. In 2005, a group of
Californian cellphone users filed a class action lawsuit against Verizon Wireless, charging the
firm with deceit for deliberately disabling some Bluetooth features on a handset. In this case,
Verizon eventually paid out over $10 million to settle, although such an outcome represents the
exception rather than the norm. In economic terms, we may describe consumers anger or sense
of unfairness as negative utility, which detracts from their consumer surplus (if they find out
about the production process after having already bought a damaged good). Thus, the various
mechanisms at play in the trading of damaged goods mean that the welfare consequences
remain ambiguous for both consumers and producers.
The formal analysis of price discrimination dates back to at least Pigous classic Economics of
Welfare, published in 1920. An all-encompassing work, including more discriminatory practices
of linear and non-linear type, is Louis Phlips Economics of Price Discrimination, first published
in 1983. These two books give a detailed overview of all topics covered in this section.
There is also extensive literature on practical applications of price discrimination. You should
refer to Stiglitz (1971) for the implications of information asymmetry on non-linear pricing
practices and Schmalensee (1981) for output considerations in third-degree price
discrimination.
Walter Ois Disneyland Dilemma (1971) is an interesting look into pricing by Disneyland from
the perspective of non-linear pricing.
The case study of this chapter is based on Deneckere and McAfees (1996) article on damaged
goods and follow-up studies by McAfee (2007) and Gershoff et al. (2012).
References
Deneckere, R. and R.P. McAfee (1996). Damaged goods, Journal of Economics and Management
Strategy, 5, 149-174.
Gershoff, A.D., R. Kivetz, and A. Keinan (2012). Consumer response to versioning: how brands
production methods affect perceptions of unfairness, Journal of Consumer Research, 39, 382-
398.
McAfee, R.P. (2007). Pricing damaged goods, economics: the open access, Open Assessment E-
Journal, Kiel Institute, 1.
Oi, W.Y. (1971). A Disneyland dilemma: two-part tariffs for a Mickey Mouse monopoly, Quarterly
Phlips, L. (1983). The Economics of Price Discrimination, Cambridge: Cambridge University Press.
Pigou, A.C. (1920). The Economics of Welfare, New York: Macmillan; St. Martin's Press.
Schmalensee, R. (1981). Output and welfare implications of monopolistic third-degree price
discrimination, American Economic Review, 71(1), 242-247.
Stiglitz, J.E. (1977). Monopoly, non-linear pricing and imperfect information: the insurance
market, Review of Economic Studies, 44(3), 407-430.
167

lOMoARcPSD|1733731
Exercises
Exercise 15.1 Comparing pricing schemes

Suppose super market chain Aha! is a monopolist on the islands of Texel and Terschelling in the
north of the Netherlands. The supermarket can sell any level of output it wishes at a constant
marginal and average cost of 5 per unit. Since the two islands are isolated and transportation is
costly, Aha! exploits the differences in demand by charging a different price in each location.
Demand in Texel is given by:
1 = 55 1 .
While the demand curve in Terschelling is given by:
2 = 70 22 .
a) Suppose Aha! can price discriminate. What prices will Aha! charge in each market? Calculate
Aha!s total profit in this situation.
Now suppose that a low-fare ferry is introduced that connects the two islands more frequently.
As a result, it costs consumers 5 per unit to transport goods between the two markets.
b) Calculate again the prevailing price(s) and the monopolists new profit level in this situation.
c) How would your answer under a) change if Aha! is forced to follow a single price policy?
d) Compare the total quantity offered in the market and the resulting welfare of the two pricing
schemes (under a) and c)).
Finally, assume that each consumer on Texel and Terschelling follows the above demand curve
corresponding to the island they live on. Suppose that Aha! could adopt a linear two-part tariff
where both the unit price and the lump-sum fee might vary.
e) What pricing policy should Aha! follow in order to maximize profits?
Exercise 15.2 Non-linear pricing models

In the mountain village of Utopia, people love swimming. However, there is only one swimming
pool close to the village to which we will refer to as the local monopolist. There are two types of
consumers in the village, in equal number. Kids have demand:
1 = 1 .
While adults have demand:
2 = 2 2.
The monopolist faces a constant average and marginal cost of zero. First assume that the local
council precludes the swimming pool from using non-linear pricing schemes.
a) Calculate the optimal price the monopolist can charge. What is the resulting profit and
welfare?
Now suppose the seller is allowed to use a single two-part tariff.
b) What is the resulting pricing scheme and how much does the monopolist earn in profits?
c) In which case is welfare greater in Utopia?
168

lOMoARcPSD|1733731
Exercise 15.3 Welfare consequences of different pricing schemes

Fixed line and mobile telecommunications company KPN has established a dominant position in
the Dutch market and as a result the Dutch competition authority (NMa) wants to examine ways
to regulate the firm. In the mobile telecommunications market, two types of consumers can be
identified; Chatterboxes with individual demand given by
= 18
and regular callers with an individual demand given by
= 5 .
Where is the price charged by KPN per minute per call. Suppose that there is an equal number
of the two types, 100 of each type, and that KPN can perfectly identify them. KPNs cost function
is given by:
= 800 + .
Where is the number of minutes called. Initially, the competition authority precludes KPN from
charging a subscription fee. More in particular, it only allows KPN to charge individual
consumers one unit price per minute called.
(a) What is the profit-maximizing unit price if the NMa forces KPN to charge the same unit
price to both groups?
(b) What is the profit-maximizing unit price if the NMa allows price discrimination? Does
welfare increase or decrease and why?
Now suppose that KPN is allowed to charge a subscription fee.
(c) Find a pricing scheme that will result in higher welfare than the one calculated under both a)
and b). Who obtains the surplus in this case?
Exercise 15.4 Bundling

The Amsterdam Music Theater wants to determine the optimal price to charge for its
performances. There are currently two kinds of performances. One consists of dance
performances organized by the National Ballet and the other of operas organized by the Dutch
Opera. Both are produced at constant zero marginal cost. The Amsterdam Music Theater faces
four different types of consumers with their willingness to pay as listed in the table below. Each
consumer type is equally well represented in the population.
Consumer Percentage in population Ballet () Dance ()
A 25% 10 90
B 25% 50 70
C 25% 70 50
D 25% 90 10
a) Consider two alternative pricing strategies: selling the goods separately and pure bundling.
For both strategies, determine the optimal prices to be charged and the resulting profits. Which
strategy should the Amsterdam Music Theater follow?
b) Suppose that the Amsterdam Music Theater considers mixed bundling as well. Is this strategy
more profitable than selling the gods separately or in a bundle?
169

lOMoARcPSD|1733731
Now, suppose there are only two different consumer types, each in equal number: Music lovers
who enjoy both kinds of performances and opera lovers who are mainly interested in the
performances organized by the Dutch Opera and who would only visit the ballet performances
occasionally. The table below presents the consumers willingness to pay in euros for a season
ticket for each performance, where .
Music Lovers Opera Lovers

Ballet performances
Opera performances 100 500
Consider three alternative pricing strategies: selling the tickets separately, pure bundling and
mixed bundling. The Amsterdam Music theatre wants to determine which pricing strategy is the
best. You may assume that the marginal cost of selling a ticket is zero.
c) Argue that mixed bundling is always at least as profitable as pure bundling in this case.
d) Is pure bundling always more profitable than selling the tickets separately in case both
consumer types have the same willingness to pay? (Note: you will have to construct a new table
to answer this question)
e) Give values for and for which selling the tickets separately is more profitable than pure
bundling.
170

lOMoARcPSD|1733731
Part VI: Commitment

In the previous chapters, we observed that decision makers often fail to reach favorable
outcomes for themselves if they interact only once. In part IV, we concluded that they can do
better in the case of repeated interaction. In this part, we will consider another possibility for
decision makers to reach outcomes that are not feasible in equilibrium in the case of one-shot
interaction: commitment to a particular strategy before the start of the game. We start in
chapter 16 by using a few simple examples to show how commitment can have strategic value
for a decision maker. In chapters 17 and 18, we will consider several ways in which decision
makers can create beneficial pre-play commitment in practice. In chapter 17, we will show how
splitting up a company in an upstream firm and a downstream firm may act as an efficiency-
enhancing commitment device. We will also see that upstream and downstream companies can
sign contracts between themselves that commit them to taking actions that are mutually
beneficial. In chapter 18, we will examine when a company has an incentive to build excessive
production capacity to commit itself to an aggressive strategy in the case of future entry so that
potential entrants will think twice before entering into the market.
171

lOMoARcPSD|1733731
Chapter 16: The value of commitment

In this chapter, we study how decision makers can benefit from committing to a particular
strategy. We will start in section 16.1 by discussing three simple games to examine why
commitment can be valuable for players. In section 16.2, we will study how early markets
entrants can establish credible commitment. In Section 16.3, we will study how trading partners
can benefit from committing not to renegotiate the terms of a deal. Section 16.4 contains a case
study on commitment by the European Central Bank to save the euro by acting as a lender of last
resort.
16.1 The value of commitment

Let us reconsider the simple Principal-Agent game that we introduced in chapter 7 and which is
replicated in figure 16.1. The agent (A) moves first and chooses between working hard and
shirking. The principal, after observing the agents choice, decides whether or not to give the
agent a bonus if he works hard. If the agent shirks, both the principal and the agent receive zero
pay-off. If the agent works hard and the principal pays the bonus, both obtain a pay-off equal to 1.
If the agent works hard and the principal does not pay the bonus, the principal receives 3 while
the agent gets -1. In chapter 7, we observed that in the subgame perfect Nash equilibrium, the
principal does not pay the bonus and the agent shirks (realizing that he will not obtain a bonus if
he works). Note that the outcome of this Principal-Agent game is a bit unfortunate from the
viewpoint of the principal. She would be better off if the agent worked and the principal paid the
bonus. Indeed, the principal has an incentive to commit to paying the bonus. If the principal can
do so, she can increase her pay-off from 0 to 1 ensuring her a commitment value of 1.
Figure 16.2 presents another example of a dynamic game that we discussed in chapter 7. A
potential entrant (E) decides whether or not to enter into a market in which only one firm, the
incumbent (I), is active at the moment. If the entrant stays out, the incumbent enjoys a nice
monopoly profit of 80 while the entrant obtains zero. Both the entrant and the incumbent obtain
a profit of 20 if the incumbent accommodates entry by choosing a soft strategy. If the
incumbent initiates a price war as soon as the entrant enters, both the incumbent and the
entrant will make a loss of 20. In chapter 7, we concluded that in the subgame perfect Nash
equilibrium of this game, the incumbent employs a soft strategy and the entrant enters. The
incumbent would increase its profits if it could credibly commit to starting a price war in the
case of entry. For the incumbent, the value of such commitment equals 60 because it can
172

lOMoARcPSD|1733731
increase its profits from 20 (its profits in the subgame perfect Nash equilibrium without
commitment) to 80 (its profits in the subgame perfect Nash equilibrium with commitment).
Figure 16.2: The accommodation game
The above examples might suggest that commitment to a strategy is only beneficial in dynamic
games. This is not true, as the simultaneous-move game in table 16.1 shows (for a similar game,
see exercise 7.3). Agent Antonio can choose between working and shirking. Principal Penlope
decides whether or not to monitor Antonio. If she monitors and Antonio works, both obtain pay-
off 40, while if Antonio shirks, Penlope fires Antonio and both get pay-off zero. If Penlope does
not monitor and Antonio works hard, she gets 50 and Antonio 30. Finally, in the case that
Penlope does not monitor and Antonio shirks, Penlope receives 30 and Antonios pay-off
equals 60. You can verify that the game has a unique Nash equilibrium in which Penlope does
not monitor and Antonio shirks. But what if Penlope can commit to monitoring Antonio? Well,
if she does, Antonios best response is to work resulting in a pay-off of 40 for Penlope, which is
more than 30, the pay-off she obtains without her commitment to monitor. Clearly, commitment
to a strategy can make sense in a simultaneous-move game as well.
Table 16.1: A simultaneous Principal-Agent game
Antonio
Work Shirk
Monitor 40,40 0,0
Penlope
Not monitor 50,30 30,60
16.2 First-mover (dis)advantages

The above examples indicate that players may enjoy a substantial advantage if they can credibly
commit to some strategy. You may wonder how players can do so. Of course, the answer
depends very much on the context. For instance, in a Principal-Agent setting, the principal may
establish commitment by signing a binding contract with the agent in which she ensures the
agent that he will obtain a bonus if he works hard. (Note that the agent is happy to accept the
contract in the example above because his pay-off also increases compared to the situation in
which the principal does not commit to paying the bonus.)
In markets, early entrants may establish first-mover advantages that allows them to commit to
aggressive behavior once a newcomer enters. How does early entry create such advantages? Let
173

lOMoARcPSD|1733731
me identifying three possibilities: technological leadership, preemption of assets, and buyer

switching costs.
Technological leadership, i.e., a sustainable comparative advantage in technology, can benefit

the first mover in two ways. First, unit production costs usually fall with accumulative output.
This generates a sustainable cost advantage for the early entrant if learning can be kept
proprietary so that the firm can maintain a leadership in market share. As a firm moves down its
learning curve, it can create substantial barriers to entry and, as a result, only few firms may be
able to compete profitably. Second, a technological advantage can be highly beneficial when
advances in product or process technology are a function of R&D expenditures. In this case,
firms with an early head start in research can exploit their lead to deter rivals by engaging in
patent races in order to maintain their prominent position in the market.
The first-mover firm may also be able to benefit from preemption of assets, i.e., preventing
rivals from acquiring scarce assets. Here, the first mover gains an advantage by controlling
assets that already exist, rather than those created by the firm through research or the
development of new technology. Such assets may be physical resources or other process inputs.
Although one can distinguish between different cases of asset purchases, it is always critical for
the first mover to have superior information. In such situations, a firm can gain a first-mover
advantage by procuring natural resource deposits, prime retailing or manufacturing locations,
by selecting the most attractive niche of the market while limiting the amount of available space
for subsequent entrants, or by heavily investing in plant and equipment in advance.
First-mover advantages may also arise from buyer switching costs. In chapter 6, we discussed
the effects of switching costs on firms pricing behavior in a more symmetric setting where firms
have an equally large customer base. In the case that one firm enters the market before another,
the setting becomes asymmetric as the second-mover has no customer base yet. With switching
costs, late entrants must invest additional resources to attract customers away from the first-
mover firm which gives the former an advantage.
Despite all the effects that benefit first movers, there are a number of disadvantages (or
conversely, relative advantages of being a late mover) that can mitigate or even reverse the
positive net gain that the incumbent might derive. First, since imitation costs are usually lower
than innovation costs in most industries, late movers may be able to free-ride on an initiating
firms investments. In addition, late movers can gain an edge through resolution of market or
technological uncertainty. Technological or customer needs are highly dynamic, in a sense that it
could prove to be more profitable to observe an incumbent and the development of a market
and only enter when the latter is stabilized and thus more opportunities have been created.
Finally, first movers may also suffer from inertia, such that they might appear to be locked-into a
specific set of fixed assets, may be reluctant to cannibalize existing product lines, or may become
organizationally inflexible.
16.3 The hold-up problem

Commitment can also be important in bilateral trade. Trading partners may be tempted to
renegotiate the terms of a deal after the other party has made relationship-specific investments.
We speak of hold-up if a party engages in an action to exploit his trading partners dependence.
To give a concrete example of hold-up, reconsider the market for second-hand cars that we
discussed in section 13.1. Recall that seller Sting offers potential buyer Bono his 10-year-old car
174

lOMoARcPSD|1733731
for sale. Suppose that Sting can exert effort to tailor the car to Bonos specific wishes. Assume
that effort costs are equal to effort (() = ) and that Stings investment increases the cars
value for Bono by = 8. If Bono pays Sting for his investment, Bonos and Stings utility are
given by
= and = ()
respectively.
An efficient outcome emerges if the total gains from trade are maximized, i.e., if Stings effort
maximizes
+ = () = 8 .
The first-order condition is
4
1=0

which implies that Sting should exert effort
= 16.
If Sting expends this level of effort, the total gains from trade created by Stings investment is
given by
+ = () = 8 = 816 16 = 16.
Now, suppose Bono and Sting are in an equal bargaining position. Before Sting makes the
investments, Bono promises to split the gains from trade 50-50, i.e., both get an additional utility
of 8. You can verify that this corresponds to a price of = 24.
However, after Sting has sunk his investment, Bono is in a good position to renegotiate the terms
of the deal, in particular if there are no other potential buyers in the market who are interested
in Stings car. Renegotiation is likely to result in a lower price, for instance equal to half of Bonos

value, i.e., = 2 = 16. If Sting anticipates this, he has a good reason to exert less effort. In that
case, Sting chooses to maximize

= () = 4 .
2
You can verify that the resulting optimal effort level equals
= 4,
which is clearly below the efficient level of effort. Because Sting anticipates that he will be held
up by Bono, we speak of a hold-up problem.
In general, a hold-up problem arises if two trading partners refrain from establishing an
efficient outcome because one of the parties fears being forced to accept disadvantageous terms
from the other party after having sunk a relationship-specific investment. Relationship-specific
investments come in many shapes and sizes. For instance, some investments are site specific. A
175

lOMoARcPSD|1733731
supplier has to locate close to the firm it supplies to save transportations costs or processing
costs. A classic example is steel manufacturing where iron production and steel production take
place close to each other to avoid having to reheat the iron for the production of steel. Another
example of a relationship-specific investment is an investment in human capital where workers
acquire skills that are more useful within their current company than elsewhere. Parties may
feel reluctant to make relationship-specific investments because this makes them vulnerable to
opportunistic behavior by their trading partners leading to inefficiencies.
How can trading partners avoid the hold-up problem, i.e., how can they commit not to
renegotiate after the other party has sunk relationship-specific investments? A straightforward
answer is writing binding contracts. In the example above, Bono could offer Sting a contract
specifying that Bono pays Sting = + where is a fixed amount of money independent of
the realized value. Such contract has similar properties as the optimal incentive contract that we
derived in the principal-agent model in chapter 8. The idea is that Sting becomes the residual
claimant of the fruits of his efforts. In the case of repeated interaction, a relational contract could
enforce commitment as we saw in chapter 11. Finally, vertical integration, where a company
merges with a supplier, sometimes serve as a commitment device to attenuate the hold-up
problem. We will discuss this in more detail in the next chapter.
16.4 Case study: Should the European Central Bank serve as a lender of last
resort in order to save the euro?
In May 2012, Europes leaders appeared to be running out of options to tackle the euro crisis.
With Greece, Ireland, and Portugal already part of a bailout fund, and Spain and Italy facing
financial market turmoil, pressure on the common currency was substantial and remedies
limited. But Europe had (and still has) one major weapon in its arsenal: the use of unlimited
liquidity that only the European Central Bank (ECB) can provide by means of its power to print
money. The ECB could credibly stand ready to buy the debt of countries like Portugal, Italy,
Greece, and Spain facing solvency problems. However, that would amount to the monetary
financing of governments, which is forbidden under the Banks own rules.
The ECB already discovered that there is more to central banking than price stability (its
primary goal) when it had to vastly increase liquidity to save the banking system during the
2008 financial crisis. The ECB did not hesitate to serve as a lender of last resort to the banking
system in order to ensure stability, despite fears of moral hazard and inflation. However, when
the sovereign debt crisis erupted in 2010 the ECB hesitated: The central bank followed a stop-
and-go policy that dented its credibility without offering a solution. The ECB provided liquidity
in the government bond markets one moment and withdrew it the next, through sterilized
market operations. This behavior raises the question: Is there a role for the ECB as a lender of
last resort in the government bond market? The interaction between the ECB, national
governments, and the bondholders has a structure that very much resembles a dynamic game.
First, the national governments choose their fiscal budget and debt exposure. Second, the
bondholders decide whether or not to renew their sovereign bonds. Third, in the case of a
liquidity crisis, the ECB decides whether or not to bail out governments, for example, by buying
government bonds.
With such a framework in mind, Paul De Grauwe, professor of international economics at the
University of Leuven and a former member of the Belgian parliament, argues that the ECB can
prevent a credit crisis by committing to bailing out governments. According to De Grauwe, the
176

lOMoARcPSD|1733731
single most important argument for allowing the ECB to bail out countries facing bond market
turmoil is to prevent countries from being pushed into a self-fulfilling debt crisis. When solvency
problems arise in one country (e.g., Greece), bondholders may start selling off the bonds of other
indebted countries out of fear of contagion. This loss of confidence can trigger a liquidity crisis in
the other markets (e.g., Spain and Italy) since there is no buyer of last resort. As the loss of
confidence increases it pushes up the premium governments must pay in order to issue new
bonds, which might lead a country to insolvency. The central bank can resolve this failure by
providing last resort lending. The point is that when people know that they will in any event get
their money back they do not panic and withdraw funds out of the financial system. Besides
which, the best thing about last resort lending is that it is rarely practiced; the very existence of a
lender of last resort prevents the loss of confidence in the first place.
Why then is the ECB not allowed to engage in any sort of monetary financing of Eurozone
governments? And what problem would a lack of commitment to this strategy spur? A popular
argument against an active role of the ECB as a lender of last resort in the sovereign bond
market is that this would lead directly to inflation. By buying government bonds the ECB
increases the money supply, thereby leading to a risk of inflation. This would be at variance with
the ECBs primal objective, which is price stability. This is known as a time inconsistency
problem that arises when it seems tempting to deviate from a long-run objective in order to reap
some short-term benefits. This, in turn, is the main reason why the ECB would want to commit to
its main goal. Avoiding doing so would result in stark moral hazard problems. By providing
lender of last resort assurance, the ECB gives an incentive to governments to issue an excessive
amount of debt. Only if the ECB could credibly commit to its no-bail-out clause would it give an
incentive to governments to tighten their fiscal budgets.
As De Grauwe argues, a potential solution to this problem would be to separate liquidity

provision from moral hazard concerns. Liquidity provision should be performed by the central
bank and the governance of moral hazard by another independent institution. The ECB would
thus commit to the responsibility of acting as a lender of last resort in the government bond
markets, thus ensuring stability. At the same time, another independent authority would take
over the responsibility of regulating and supervising the creation of debt by the national
governments. In order for the ECB to be successful in stabilizing the sovereign bond markets of
the Eurozone, it would have to make it clear that it is fully committed to providing liquidity as a
lender of last resort. By building up confidence, such a commitment would guarantee that the
ECB need not intervene in the government bond markets most of the time. Ideally, much like the
commitment of being a lender of last resort to the banking system, the central bank would rarely
have to act as one.
Nobel laureate Thomas Schelling gives much attention to the value of commitment in his classic
book Strategy of Conflict (1960). Schellings work inspired movie director Stanley Kubrick to film
his 1964 black comedy movie Dr. Strangelove or: How I Learned to Stop Worrying and Love the
Bomb. In this movie, the Soviet Union has built a doomsday machine that will destroy all
civilization if the Soviet Union is attacked. The doomsday machine is supposed to serve as a
commitment device to prevent the US from attacking. However, when the American general
Ripper decides to launch an unauthorized nuclear attack, the Soviet Unions strategy goes
tragically wrong. It turned out that Soviet Union failed to inform the US of the existence of the
177

lOMoARcPSD|1733731
doomsday machine. When asked why, the Soviet ambassador Alexei de Sadeski says they
planned to do so a week later at the Communist Party conference, in honor of the Premier who
loves surprises.
For seminal contributions to the theory of dynamic games you can refer to the bibliography of
chapter 7. The duopoly model presented in this chapter was initiated by German economist
Heinrich Freiherr von Stackelberg in his 1934 work Marktform und Gleichgewicht. You can find
the original paper as well as a citation of a good English translation in the references below. For
an analysis and a review of first-mover advantages and disadvantages you can refer to Sutton
(1991) and Lieberman and Montgomery (1998) respectively.
The case study of this chapter is based on publications and articles of KU Leuven professor Paul
De Grauwe (De Grauwe 2010, 2011 and De Grauwe and Ji, 2013).
References
De Grauwe, P. (2011). Managing a fragile Eurozone, CESifo Forum, 12(2), 40-45.
De Grauwe, P. (2010). The financial crisis and the future of the Eurozone, Bruges European
Economic Policy Briefings, Number 21.
De Grauwe, P. and J. Ji (2013). Self-fulfilling crises in the Eurozone: an empirical test, Journal of
International Money and Finance, 34, 15-36.
Lieberman, M. and D. Montgomery (1988). First-mover (dis)advantages: retrospective and link
with resource-based views, Strategic Management Journal, 9, 41-58.
Schelling, T. (1960). The Strategy of Conflict, Cambridge, MA: Harvard University Press.
Stackelberg, H.F.v. (1934/1952). Marktform und Gleichgewicht, in A.T. Peacock, Theory of the
Market Economy, London: William Hodge.
Sutton, J. (1991). Sunk Cost and Market Structure, Cambridge, MA: MIT Press.
Exercises
Exercise 16.1 The value of commitment

Consider the following game between an employer (Katharine) and an employee (Kevin).
Katharine wants Kevin to work hard rather than loaf around and that is why she considers
spending more time supervising him. She wants to choose between a low and a high level of
supervision. All else the same, Kevin prefers to loaf. Both players must simultaneously
determine their actions.
Kevin
Work Loaf
Low 4, 3 2, 4
Katharine
High 3, 2 1, 1
a) What are Katharines best responses in this game? Determine the Nash equilibrium of the
game. What are the rationality assumptions implicit in this equilibrium?
b) Suppose now Katharine has the option of committing to an action ahead of Kevins decision.
What action will she choose?
178

lOMoARcPSD|1733731
c) Comparing the answers to (a) and (b), what can you say about the value of commitment for
Katharines choice?
d) When pre-commitment has a strategic value, the player that makes that commitment ends
up regretting his actions, in the sense that, given the rivals choices, (s)he could achieve a higher
payoff by choosing a different action. Is that statement true in general? How does it apply in this
example?
Exercise 16.2 Location

The town of Stracciatella consists of one street, 1 kilometer long. The one thousand inhabitants
of the town () are uniformly distributed along the street. Currently, there is only one ice cream
producer, Cioccolato, which sells one type of ice cream at price = 5. Its marginal cost for
producing one ice cream is = 3. Each inhabitant is willing to buy one ice cream per period,
and their utility function is:
() = 10 20 ,
where represents the distance in kilometers between the consumers position and the closest
ice cream shop. Cioccolato can choose how many shops to open and their location along the
street. However, the municipality imposes that each ice cream shop must be at least 250 meters
distant from another. Opening a new shop costs = 600.
a) How many shops will Cioccolato open? Where?
Pistacchio is another ice cream producer willing to enter Stracciatellas market. It has the same
price as Cioccolato, but a lower marginal cost: = 2. The consumers perceive the ice creams
of the two sellers as perfect substitutes.
b) Given Cioccolatos choice of localization as in part a), how many shops will Pistacchio open?
Where?
c) What is the cheapest way for Cioccolato to deter entry?
d) Will Pistacchio enter the market? Why (not)?
e) In this framework, is there a first- or a second-mover advantage?
Exercise 16.3 ECB and (lack of) commitment

Consider a three-stage game between the European Central Bank (ECB) and private agents. In
this setting, private agents form their expectations of inflation in the first stage, the ECB
chooses the actual level of inflation in the second stage and the unemployment rate is
determined in the final stage based on the following rule:
= 0.06 0.5( )
For simplicity, assume that the ECB can choose the rate of inflation. In reality, the ECB can only
control the level of inflation imperfectly through its control of the money supply. The welfare
cost of inflation and unemployment is described by the following welfare loss function:
(, ) = + 5 2
179

lOMoARcPSD|1733731
Since the ECB dislikes high inflation and unemployment, it will choose a level of inflation so that
its loss function is minimized. Throughout the exercise, private agents are assumed to prefer to
predict inflation as well as possible.
a) Suppose the ECB announces a target for inflation and private agents find the announcement
credible. What is the level of inflation and unemployment in the Eurozone if the ECB targets =
2%?
b) Does the ECB have an incentive to deviate from its announced target? Determine the optimal
level of inflation according to the welfare loss function.
Assume now that private agents acknowledge the ECBs lack of commitment to its announced
rule and form their expectations accordingly.
c) What is the resulting level of inflation and unemployment in the Eurozone? Compare the two
outcomes.
d) What changes if the ECB commits to the initial target of 2% before private agents form their
expectations?
Exercise 16.4 The hold-up problem
Adle is also interested in purchasing a second-hand car from Sting. Suppose that Sting can exert
effort to tailor the car to Adles specific needs and that Stings investment increases the cars
value for Adle by = 12. You may assume that effort costs for Sting are given by () = 4 +
2 and that Adle will pay Sting a price for his investment.
a) Write down the utility functions of Adle and Sting.
b) Calculate the socially efficient level of effort.
Suppose the two parties are in an equal bargaining position and that Adle promises to Sting
that they will split the gains of trade exactly in half.
c) Is that a credible promise? Will Sting exert the optimal level of effort in this case? Explain your
answer and provide the relevant calculations.
180

lOMoARcPSD|1733731
Chapter 17: Make or buy

In this chapter, we will study how firms in a vertical chain can establish efficiency-enhancing
commitment by vertical integration or vertical separation. We will do so in the model of
production in the vertical chain that we initiated in chapter 7. One application is the firms make
or buy decision. This refers to the companys choice whether to produce an input itself or to buy
it from a firm higher up in the vertical chain. In other words, make refers to vertical integration
and buy to vertical separation. Similarly, a firm may choose between selling its products in its
own retail stores or leaving that to independent retailers.
The set-up of this chapter is as follows. In section 17.1, we will examine the pros and cons of
vertical integration and vertical separation. One of the conclusions will be that if the services
related to the firms input are contractible, it does not matter: The firm can write a contract with
either the separate upstream firm or integrate with the firm and write a contract with the
divisions management. In section 17.2, we will study in detail what contracts are efficient from
the firms point of view, that is, what contracts maximize (joint) profits. Section 17.3 contains a
discussion on how competition policy should deal with vertical integration and vertical
restraints, i.e., contracts between firms in the vertical chain. We conclude in section 17.4 by
discussing vertical restraints in the European car market.
17.1 Advantages and disadvantages of vertical integration

Firms in the vertical chain of production may prefer vertical integration over vertical separation
for several reasons. First of all, by integrating vertically, they may avoid double
marginalization. As we saw in chapter 7, firms in the vertical chain may suffer from the double
marginalization problem. In the case of a chain of two monopolies, both will add a profit margin
to marginal costs, which results in a higher price than if the two firms were integrated. Indeed,
the two firms can increase their joint profits by merging into one firm.
Second, vertical integration may improve coordination between firms in different layers of the
vertical chain. In some cases, a firm can decide what inputs to buy only on the basis of their
prices. For example, a cheese factory requires milk. Because milk is a fairly homogeneous good,
the firm can decide to buy milk from the supplier offering the lowest price. In other cases, firms
in the vertical chain need to coordinate on the specifications of the inputs, in particular if the
inputs are highly specialized. For instance, a car producer combines a large set of complex
components into a car. In the case of vertical separation, the car producer should coordinate
with several suppliers on the specifics of each component. In the case of vertical integration, the
firm can establish this coordination in-house. In the integrated organization, the responsible
decision makers can jointly plan, produce, and assemble the individual components of a car.
Third, vertical integration may be a solution to free-riding problems. All kinds of free-riding
problems can emerge in the vertical chain. Consider digital cameras. The producer of the
cameras may sell them both on the Internet and in brick-and-mortar stores. Suppose the
retailers are separate firms. The shop owner of a brick-and-mortar store may be reluctant to
spend much time informing his customers about the cameras. The reason is that his customers
may decide to buy the camera, not at the brick-and-mortar store, but on the Internet for a lower
price. In other words, the Internet firm free rides on the sales efforts of the brick-and-mortar
stores shop owner. The camera company may suffer from the lack of sales efforts. Another free-
riding problem emerges if the camera retailer sells several camera brands. The camera company
may train the retailer to fix broken cameras. However, other camera companies may free-ride on
181

lOMoARcPSD|1733731
the efforts of the camera company, because the retailer can now also repair their brand of
cameras. To mitigate those free-riding problems, the camera company may prefer to integrate
forward into the retail market for cameras. By doing so, it can guarantee that potential
customers get sufficient information about its cameras and it can decide not to sell competitors
cameras in its retail shops.
Additionally, a firm may vertically integrate to avoid hold-up problems. For example, suppose
that your university asks a software company to develop a new website for the Economics
department. In the absence of a contract, the software company may be hesitant to develop the
website for fear that the university would want to renegotiate the price. After the software
company develops the website, it has a weak bargaining position relative to the university
because the website is tailored to the Economics departments needs and cannot be used for
alternative purposes. Vertical integration could solve this: The university may let its own ICT
department develop the website. However, vertical integration may create other hold-up
problems as we will see below.
Vertical integration may also serve as a device to foreclose competitors, i.e., to cut them off
from suppliers or retailers. By doing so, the firm may prevent the entry or induce the exit from
competitors, which may increase its market power. When the firm controls its suppliers, it
prevents them from supplying to potential entrants so that it may become more difficult for
them to enter into the market. Similarly, when the company has a substantial market share in
the retail market, it may not allow its retailers to sell competing products, which makes entry
more costly for potential competitors.
While firms in a vertical chain may have strong reasons to integrate vertically, at the same time,
they may prefer to remain separate. Economies of scale are one important advantage of vertical
separation (compared to vertical integration). For many inputs, the scale of a firms use is too
small to justify producing it at the firm rather than buying them from an independent supplier.
This holds true for products such as chairs, coffee, cups, paper, pens, printers, and so on. In
contrast to the buying firm, an independent supplier enjoys economies of scale by catering to
many firms.
Another reason why firms may prefer to separate vertically is that by doing so, they can reduce
influence costs. Influence costs arise if the supplier spends time and effort trying to affect the
decisions made by the firms top management, which may also consume a lot of management
time. Time and effort spent in influencing management decision making may substitute time and
effort spent on productive activities. In the case of vertical separation, the possibilities to
influence the firms management decrease so that both the supplier and firms managements
influence costs decrease.
Finally, vertical separation may act as a commitment device to mitigate hold-up problems. For
example, consider a high-tech firm that relies heavily on the development of innovative products
(think of markets for biotech products or pharmaceuticals). The firm may have a commitment
problem when hiring researchers to develop those products if their inventions are not
contractible. The reason is that the firm owns the intellectual property developed by the
researchers. This gives the researchers a weak bargaining position, so that they do not
experience strong incentives to develop a path-breaking invention. In contrast, if the researchers
form a separate firm, they own their inventions, which creates a stronger bargaining position
182

lOMoARcPSD|1733731
vis-a-vis the buying firm and, in turn, stronger incentives to invent. This reasoning holds
particularly true if the researchers can threaten to sell their intellectual property to a competitor
of the buying firm.
How does a company balance the pros and cons of vertical integration and vertical separation?
One important component in the answer of this question is whether the suppliers services are
contractible, i.e., whether the supplier and the user are able to specify all relevant characteristics
of the suppliers services in a contract that is enforceable by a court. If they can, it does not
matter whether the supplier and user are integrated or not. The user can give the supplier the
right incentives to invest either in a labor contract or in a contract with an internal division in
the case of vertical integration, or in a contract with an independent supplier in the case of
vertical separation. Good examples of contractible services are those related to standard
equipment such as furniture, word processing software, and coffee machines. Some firms have
their own employees who are responsible for those services (such as an in-house ICT
department), others outsource them. For at least two reasons, services might be difficult to
contract. First, services may be extremely complex so that it is virtually impossible to write a
complete contract that specifies all desired attributes. Software is a good example. Second, the
desired services may be context-dependent so that it is virtually impossible to write down all
possible contingencies in a contract.
Another important factor is the frequency of interaction between the firm and its supplier. Even
if the suppliers services are not contractible, a desirable outcome can arise if the two interact
repeatedly. As we saw in chapter 11, in the case of repeated interaction, decision makers can
establish an efficient outcome using relational contracts. In the case of frequent interaction, it
may not matter whether the firm integrates vertically or not: It establishes a relational contract
with either the management of an internal division or an independent supplier.
If the suppliers services are not contractible and interaction between the firm and the supplier
is infrequent, the preferred level of integration is highly context-dependent. To find it, the above
pros and cons should be weighted. Vertical separation is more likely to be more efficient than
vertical integration if: (1) the upstream market is more competitive (so that the double
marginalization problem is less severe); (2) less coordination is required between the supplier
and user; (3) less severe free-riding problems exist at the downstream level (that cannot be
solved on the basis of contracts); (4) economies of scale are greater; (5) influence costs are
greater; and (6) vertical integration presents the more severe hold-up problems.
17.2 Vertical restraints

In the previous section, we established that if the relevant supplier services are contractible, the
firm is indifferent between buying (i.e., vertical separation) and making (i.e., vertical integration).
In this section, we zoom in on the kind of contracts that firms may sign to alleviate potential
problems in the vertical chain of production. We call such contracts vertical restraints
because they restrict parties in the vertical chain as to what decisions they can take. As we will
see now, such contracts can serve as valuable commitment devices.
We start by considering franchise contracts. Under a franchise contract, a downstream firm

pays an upstream firm for the right to sell the upstream firms products. Franchise contracts are
common for fast-food restaurants, supermarkets, and petrol stations. Usually, franchise
contracts specify many things, including the products the downstream firm must carry, the
183

lOMoARcPSD|1733731
uniforms worn by its staff, and the decoration of the shop. In return, the upstream firm can offer
nationwide advertising, personnel training, management advice, and so forth. A franchise
contract may serve many purposes, including dealing with the double marginalization problem.
Let us return to the hypothetical vertical chain of smartphones shown in figure 17.1 that we
introduced in chapter 7. Retailer Red Apple Store sells Strawberry smartphones to end
consumers. Both firms are monopolists in the market for smartphones. We assumed that
Strawberrys marginal costs of producing a smartphone are equal to
= 50.
Strawberry supplies its smartphones to the Red Apple Store at price and the Red Apple Store
resells to consumers for a price equal to . Consumer demand for smartphones is given by
() = 750 .
The corresponding inverse demand equals
() = 750 .
In chapter 7, we found that in the case of vertical separation, Red Apple Store sells smartphones
for a price of 575 to consumers, way above the price of 400 which is the optimal price from the
viewpoint of the two firms.
Figure 17.1: The vertical chain in a hypothetical market for smartphones
A franchise contract can solve the double marginalization problem. For example, if the contract
specifies that Red Apples store pays Strawberry = 50 for each smartphone, Red Apple Store
maximizes profits by selling smartphones for a price of 400. Red Apple Store pays Strawberry a
fixed franchise fee to let it share the profits. Note that this contract is efficient because Red Apple
Store becomes the residual claimant of the marginal profits made in the industry: It can keep any
cent above the marginal cost of producing a smartphone. This result is in line with what we saw
earlier in the principal-agent context where the efficient contract has the agent become the
residual claimant of the fruits of his efforts. Note that Strawberry sells at marginal costs. It still
can makes a profit if it charges Red Apple Store a franchise fee. The level of the franchise fee
does not matter for the efficiency of the franchise contract as long as both parties are willing to
184

lOMoARcPSD|1733731
accept the contract. The reason is that the fee constitutes a sunk cost from the viewpoint of Red
Apple Store.
Resale-price maintenance is another contractual agreement that can solve the double
marginalization problem. Under resale-price maintenance, the upstream firm specifies the price
the downstream firm must charge to its customers. In the Strawberry/Red Apple Store example,
Strawberry can oblige Red Apple Store to sells its smartphones for a price equal to 400, i.e., the
optimal price from the firms viewpoint. Now, Strawberry and Red Apple Store have to negotiate
the wholesale price to establish each firms share of the pie. Any wholesale price works as long
as it lies between 50 and 400.
The firm may also make agreements about quantity. For instance, Strawberry and Red Apple
Store may write a contract so that Red Apple Store is required to buy 350 smartphones from
Strawberry. Note that 350 is the monopoly quantity. As the contract obliges Red Apple Store to
buy the smartphones, it regards their costs as sunk so that it is willing to sell them at the market
clearing price of 400. Beforehand, the firms should agree on a trading price for the 350
smartphones. Any unit price between 50 and 400 is feasible.
An exclusive territories contract could solve free-riding problems in the vertical chain at the
retailer level. Such a contract specifies that the retailer obtains a monopoly position in a well-
defined region. The retailers monopoly position prevents other firms in the region from free-
riding on the efforts of the retailer to increase the suppliers demand. Examples include
advertising, customer advice, and post-sales services. However, note that an exclusive territories
contract mitigates competition between retailers in the region, which may have negative welfare
consequences.
A producer may also sign an exclusive dealing contract with a retailer if it fears that competing
producers can free-ride on the producers efforts. An exclusive dealing contract specifies that the
retailer is not allowed to sell competing producers products. Recall the free-riding problem that
may emerge if a camera retailer sells several camera brands. If the camera company trains the
retailer to fix broken cameras, other camera companies may free-ride on the efforts of the
camera company. An exclusive dealing can mitigate this problem. One potential downside of an
exclusive dealing contract is that it may serve as a foreclosure device because competitors
cannot use the retailer as a channel to sell their product.
17.3 Vertical restraints and competition policy

How should competition policy deal with vertical integration and vertical restraints? Our
analysis in this chapter suggests that it should be lenient in most cases. We concluded that
vertical integration, franchise contracts, resale-price maintenance, and agreements about
quantity can mitigate the double marginalization problem. In chapter 7, we observed that both
the firms and the consumers benefit if the double marginalization problem is mitigated, so that
welfare increases. Note that this conclusion is orthogonal to equivalent strategies established by
firms that are active on the same level of the vertical chain, such as horizontal mergers, price-
fixing agreements, and quantity-fixing agreements. Still, vertical integration and vertical
restraints can have negative consequences on welfare, in particular when their effect is that
suppliers or retailers foreclose competitors.
Is competition policy more lenient toward vertical integration and vertical restraints than
toward horizontal practices such as horizontal mergers and agreements between firms that are
185

lOMoARcPSD|1733731
active in the same market? In some cases, it is. For instance, in its guidelines on the assessment
of non-horizontal mergers, the European Commission states the following:
The Commission is unlikely to find concern in non-horizontal mergers, be it of a

coordinated or of a non-coordinated nature, where the market share post-merger of the
new entity in each of the markets concerned is below 30 % and the post-merger HHI is
below 2 000.
In contrast, a horizontal merger may be a concern for the European Commission if the post-
merger HHI is below 2,000, in particular if because of the merger, the HHI increases with 250 or
more.
Moreover, recall Paragraph 1 of Article 101 of the Treaty on the Functioning of the European
Union:
At first sight, the Commission does not seem to make a distinction between horizontal
agreements and vertical agreements. Paragraph 1(a) suggests that resale price maintenance is
prohibited because such an agreement directly limits the selling prices a retailer may charge. An
exclusive territories contract or an exclusive dealing contract may be interpreted as a violation
of Paragraph 1(d) because they exclude trading parties that are not part of the contract.
However, in the case of vertical restraints, the European Commission often makes exemptions
on the basis of Paragraph 3 of Article 101:
The provisions of paragraph 1 may, however, be declared inapplicable in the case of any
agreement [...] which contributes to improving the production or distribution of goods or to
promoting technical or economic progress [...]
In this parts case study, we will examine the welfare effects of such an exemption for the
European car market.
17.4 Case study: Vertical restraints in the European car market

According to the analysis in this chapter, vertical agreements between firms could fuel both
market efficiency gains and anti-competitive effects. As a consequence, competition authorities
face the difficult task of evaluating vertical restraints. The European Commission follows a
186

lOMoARcPSD|1733731
rather lenient approach with respect to vertical restraints as illustrated by the large number of
specific and block exemptions granted for vertical restraints in the European Union (EU).
A well-known example of a market for which the Commission has granted block exemptions in
vertical restraints is the car market. Since 1985, the car market in Europe has institutionalized a
distribution system in the downstream market that is both selective and exclusionary. Selectivity
refers to car manufacturers selecting authorized dealers through their own qualitative and
quantitative criteria. Manufacturer can protect the selective relationship by prohibiting its
dealers to sell cars to unauthorized dealers, for example to dealers in other countries. Exclusivity
refers to manufacturers assigning exclusive territories to its dealers. To protect the territorial
exclusivity, manufacturers typically do not allow appointed dealers to maintain branches
outside their own contract territory.
The car industry and other advocates of the system defended it on the basis of increased
efficiency in the form of, for example, incentives for sales and after-sales services. However,
critics of that system (especially consumer organizations) pointed at potential anti-competitive
effects and a conflict with the European integration objectives. According to them, the
distributional system created large and persistent price differentials among EU members since it
limited cross-border trade: Selectivity impeded independent resellers to buy cars in one country
and resell them in another and exclusivity hindered authorized dealers to set up branches in
other countries.
Since the expiration of the block exemption granted to the car market in September 2002, the
European Commission has started investigating the costs and benefits of less restrictive
distribution systems. In particular, the Commission considered allowing manufacturers either to
select their dealers or assign exclusive territories, but no longer the combination of the two. The
idea behind that proposal is that it would increase competition between dealers of the same
brand, thus eliminating price differences to some extent.
Katholieke Universiteit Leuven economists Randy Brenkers and Frank Verboven examined how
the proposed liberalization could promote national and international competition and quantified
the significance of such effects. To begin with, they claim that by removing selectivity or
exclusivity, dealers can either expand abroad by opening their own outlets or directly sell cars to
other non-authorized importers. They argue that this will facilitate cross-border trade and
reduce the likelihood of international price discrimination, i.e., the same car brand being sold for
different prices in different European countries. In addition, they claim that the new regulation
will reduce prices within a country through increased competition between dealers of the same
brand (intra-brand competition). This would, in turn, reduce or even eliminate retailer market
power and hence the double marginalization problem.
What is the total effect of the proposed liberalization on consumer surplus and welfare? In
chapter 3, we argued that generally, less market power is welfare enhancing, However, as we
saw in chapter 15, relaxing the possibilities for firms to price discriminate may have ambiguous
effects on welfare. On the basis of an empirical analysis, Brenkers and Verboven conclude that
the proposed liberalization of the European car market can significantly increase both consumer
and producer surplus and, as a result, total welfare. The order of magnitude of the welfare gains
depends on how competitive the market was before liberalization. In a conservative scenario
where liberalization does not induce more national intra-brand competition, the authors
187

lOMoARcPSD|1733731
estimate the total welfare increase to be between slightly positive and 3 billion per year. In
contrast, if the previous system effectively limits national intra-brand competition, the authors
estimate welfare gains up to 11 billion per year. Intuitively, the estimate is much larger in the
second scenario because liberalization has the combined effect of reducing international price
discrimination and eliminating potential double marginalization effects. The broader conclusion
is that vertical restraints can have substantial negative welfare effects.
For an overview of the advantages and disadvantages of vertical integration and separation you
can refer to the textbooks cited in previous industrial organization focused chapters. Original
sources include Muellers (1969) Theory of Conglomerate Mergers and Bonanno and Vickers
(1988) Vertical Separation.
A number of vertical restraints are examined in detail in Steiner (1985) and Lafontaine (1992,
1993). The literature on resale price maintenance is quite extensive, see, e.g., Pickering (1966),
Marvel and McCafferty (1984), Deneckere, Marvel and Peck (1997) and Matthewson and Winter
(1998). There is also a large literature on the non-price vertical restraints mentioned in this
chapter including territorial arrangements and exclusive dealing and selling among others. Refer
to Aghion and Bolton (1987) for a study of such contracts as a barrier to entry. Marvel (1982)
and Besanko and Perry (1994) study exclusive dealing. Finally, Rasmusen, Ramseyer and Wiley
(1991) and Rey and Stiglitz (1995) examine exclusive territories.
The case study of this chapter is based on Brenkers and Verbovens 2006 paper in the Journal of
the European Economic Association.
References
Aghion, P. and P. Bolton (1987). Contracts as a barrier to entry, American Economic Review,
77(3), 388-401.
Bain, J.S. (1956). Barriers to New Competition: Their character and Consequences in
Manufacturing Industries, Cambridge, MA: Harvard University Press.
Besanko, D. and M.K. Perry (1994). Exclusive dealing in a spatial model of retail competition,
International Journal of Industrial Organization, 12(3), 297-329.
Bonanno, G. and J. Vickers (1988). Vertical separation, Journal of Industrial Economics, 36(3),
257-265.
Brenkers, R. and F. Verboven (2006). Liberalizing a distribution system: the European car
market, Journal of the European Economic Association, 4(1), 216-251.
Deneckere, R., H.P. Marvel and J. Peck (1997). Demand uncertainty and price maintenance:
markdowns as destructive competition, American Economic Review, 87(4), 619-641.
Lafontaine, F. (1992). Agency theory and franchising: some empirical results, RAND Journal of
Economics, 23(2), 263-283.
Lafontaine, F. (1993). Contractual arrangements as signaling devices: evidence from franchising,
Journal of Law, Economics and Organization, 9(2), 256-289.
Marvel, H.P. (1982). Exclusive dealing, Journal of Law and Economics, 25(1), 1-25.
Marvel, H.P. and S. McCafferty (1984). Resale price maintenance and quality certification, RAND
Mathewson, F. (1998). The law and economics of resale price maintenance, Review of Industrial
Organization, 13(1-2), 57-84.
188

lOMoARcPSD|1733731
Mueller, D.C. (1969). A theory of conglomerate mergers, Quarterly Journal of Economics, 83(4),
643-659.
Pickering, J.F. (1966). Resale Price Maintenance in Practice, London: Allen and Unwin.
Rasmusen, E.B., J.M. Ramseyer and J.S. Wiley (1991). Naked exclusion, American Economic
Review, 81(5), 1137-1145.
Rey, P. and J. Stiglitz (1995). The role of exclusive territories in producers' competition, RAND
Salinger, M.A. (1988). Vertical mergers and market foreclosure, Quarterly Journal of Economics,
103(2), 345-356.
Steiner, R.L. (1985). The nature of vertical restraints, Antitrust Bulletin, 30(1), 143-197.
Exercises
Exercise 17.1 Make or Buy

The company DutchHats is preparing the production of orange hats for Queens Day in
Amsterdam. Currently, there are three suppliers in the upstream market (for simplicity:
1, 2, 3), which produce the fabric needed for the orange hats. In the downstream market,
DutchHats is the only company allowed to produce orange hats for Queens Day. The upstream
firms compete la Cournot, and they all have the same cost function () = 8 where
represents meters of fabric and = 1,2,3. DutchHats is able to produce one orange hat with one
meter of fabric, and its cost function is (, ) = (5 + ), where is the total quantity of
orange hats produced and is the (market) price of one meter of fabric. The aggregate demand
of orange hats is:
() = 225 5,
where is the price of one orange hat.
a) Compute the upstream market price ( ), the downstream market price (), and quantity ()
in equilibrium.
DutchHats could also acquire one of the upstream firms, producing fabric in-house for Queens
Day. The cost of the takeover would be = 300.
b) Will DutchHats make (i.e., acquire one of the upstream firms) or buy (i.e., buy fabric from
the upstream market)?
c) How does your answer in b) change if the upstream firms compete la Bertrand?
Exercise 17.2 Make or buy

CrunchySombrero is the only producer of tortillas in the state of Chihuahua, Mexico. The famous
Mexican chips are mainly made of corn, and in Chihuahua the only corn supplier is the company
TequilaCorn. Assume that Chihuahuas demand for tortillas can be represented as
() = 40 2,
where is the price of a bag of tortillas. In the downstream market, CrunchySombrero buys all
the corn needed for its production by TequilaCorn at the price , all its other costs are
189

lOMoARcPSD|1733731
negligible, and producing a bag of tortillas requires one kilogram of corn. In the upstream
market, TequilaCorns cost function is () = 10, where represents kilograms of corn.
a) Compute equilibrium price and quantity in both the upstream and downstream market.
TequilaCorn wants to stipulate a contract with CrunchySombrero in order to avoid the double
marginalization problem. The contract imposes a resale-price to the tortillas producer (),
which would get an annual fixed fee in return.
b) Which resale price will TequilaCorn impose on CrunchySombrero in order to maximize its
profits?
c) Which is the maximum fee amount stipulated by the contract? And the minimum? Explain.
TequilaCorn is facing a dilemma. It can either stipulate the abovementioned contract with a fee
of = 15, or it can acquire CrunchySombrero by paying = 30. If it chooses the latter case,
TequilaCorns marginal cost will decrease to 8 due to coordination gains.
d) Will TequilaCorn buy (i.e., sign the contract) or make (i.e., acquire CrunchySombrero)?
Exercise 17.3 Vertical Integration

The market of mugs is currently dominated by one incumbent (I), which is a monopolist. I buys
the clay from the monopolist of the upstream market (U) at the price per kilogram. All the
other costs for the incumbent are negligible, while U has a marginal cost of = 12 per
kilogram of clay. With one kilogram of clay, I is able to produce one set of mugs. The demand for
mugs is:
() = 324 3,
where is the price of a set of mugs. Instead of buying clay from U, the incumbent could acquire
the upstream firm by paying = 6000.
a) Will the incumbent acquire the upstream firm?
A new firm considers entering the market of the incumbent. The potential entrant (E) faces a
cost of entry = 12. If E enters, the competition in the downstream market will be la Cournot,
and U, the upstream firm, will sell clay to both downstream firms at the same price . However,
if I acquires the upstream firm, it will foreclose the potential entrant from buying clay,
preventing E from entering the market.
b) Will the incumbent acquire the upstream firm?
Exercise 17.4 Vertical mergers and franchise contracts

Peanut butter monopolist Calv supplies peanut butter to Albert Heijn in an isolated village. The
supermarket is a monopolist in the village. Demand for peanut butter is given by:
= 100
where denotes the price a consumer pays for peanut butter. The supermarkets costs consist of
marginal selling costs equal to 40 per unit plus the costs it pays Calv for the peanut butter.
Suppose Calvs marginal costs are equal to 12 and that its fixed costs are equal to zero.
a) Suppose Calv sells the peanut butter for a price of per unit to the supermarket. How much
190

lOMoARcPSD|1733731
peanut butter will Albert Heijn offer to the market as a function of ?

b) Determine the that maximizes Calvs profits. Also, calculate the welfare in this market if
Calv and Albert Heijn use the above strategy.
Suppose Calv and Albert Heijn merge.
c) How much peanut butter will the merged firm offer to the market?
Instead of the merger, assume that Calv and Albert Heijn sign a franchise contract. The contract
specifies a two-part tariff with a unit price and a fixed fee .
d) Which and maximize Calvs profits?
e) Does the above-mentioned contract or the merger increase welfare compared to the situation
under (b)?
191

lOMoARcPSD|1733731
Chapter 18: Predation

Predation is a business strategy that sacrifices part of the short-run profits to induce exit or
deter entry by rival firms. Committing to such a strategy can be attractive for a company if by so
doing it increases its market power in the long-run compared to a situation where the rival firms
would stay in or enter into the market. In section 18.1, we will discuss several examples of
predation. In section 18.2, we will stress the importance of the credibility of predatory strategies.
The predator must be willing to sacrifice profits, by charging low prices, for instance, as long as
its competitor(s) remain in the market or as soon as one enters. We will give an example of how
a firm can credibly commit to such an aggressive strategy: For one, it can build excessive
production capacity before competitors enter. Section 18.3 contains a discussion of how
competition policy deals and should deal with predation. Section 18.4 includes a case study on
Microsofts alleged predatory behavior in the market for web browsers in the late 1990s.
18.1 Examples of predation

Predatory strategies come in many shapes and sizes. The most obvious one is probably
predatory pricing. A company engages in predatory pricing if it charges very low prices with
the aim to drive competitors out of the market. The idea is simple: When a firm offers its product
for a very low price, its competitors can only stay in the market if they incur substantial losses.
Their financial situation may be such that they cannot sustain such losses so they are forced to
leave the market. As soon as rivals leave the market, the firm becomes a monopolist and it can
make a nice monopoly profit. This chapters case study provides a potential example of such a
strategy: Microsoft dwarfs competitor Netscape in the market for web browsers by giving
buyers of Windows a free copy of Internet Explorer.
Firms may also engage in activities that raise rivals costs so that it becomes prohibitively
costly for them to enter the market. For instance, an incumbent firm may engage in R&D
activities not to develop new products or production technologies that it will actually use but to
obtain patents on those products or production technologies so that potential entrants will not
be able to use them. By doing so, the incumbent firm increases the entry costs of potential
entrants and it may succeed in keeping them out of the market. Similarly, incumbent firms may
advertise their products aggressively to deter entry. As potential entrants will have to initiate an
extensive advertising campaign as well to convince consumers to buy their products it may be
prohibitively costly for them to enter. Indeed, it is hard to imagine a firm entering the market for
cola successfully without starting an advertising campaign that can match Coca-Cola and
PepsiCos.
Firms may also be able to cut off potential rivals from essential facilities so that it becomes
impossible for them to enter the market. For example, a company can bid in license auctions to
preemptively buy market licenses. Mobile telecommunications is a good example of markets in
which firms can only enter when they acquire licenses. Firms offer mobile telecommunications
services using radio frequencies. As radio spectrum is scarce, governments grant licenses to
firms that enter the market, often to the highest bidders in auctions. A firm can participate in
such auctions not because it needs more radio frequencies but to prevent potential entrants
from acquiring them. Similarly, a firm may integrate vertically or sign contracts with upstream
or downstream firms to get control over essential facilities.
192

lOMoARcPSD|1733731
18.2 Building overcapacity to deter entry

Building excessive production capacity is another strategy that incumbent firms may employ to
deter entry. When the incumbent has installed its production capacity, part of its production
costs are sunk so that the marginal cost of production is relatively low. As a consequence, the
incumbent commits itself to an aggressive strategy when a rival enters the market. Being aware
of this, the rival may decide not to enter the market because it will not be able to compensate its
entry costs.
To illustrate how firms can establish credible commitment by building overcapacity, recall the
model of the coffee market on your university campus that we discussed in chapter 6. We
assumed that two firms compete in this market, Coffee Bucks and Star Company. Total demand
() for coffee on the campus is given by
() = 960 240.
When studying competition in this market, we modeled the firms interaction in several ways. In
all cases, we assumed that firms made their strategic decisions simultaneously. Now, let us
assume that Star Company enters the market earlier than Coffee Bucks. Suppose that the main
strategic decision variable is a firms production capacity. Once Coffee Bucks enters, Star
Company has already installed its production capacity so that Coffee Bucks can take this into
account when deciding on its own production capacity. Again, for both firms, the marginal costs
per cup of coffee are constant and equal to
= 1.75.
The question is: Does Star Company enjoy a first-mover advantage? The German economist
Heinrich Freiherr von Stackelberg developed a model to answer this question. He published his
results in 1934, almost 100 years after Cournot published his. Stackelbergs model is the same as
the Cournot model, with the crucial difference that firms move sequentially instead of
simultaneously. More precisely, the Stackelberg model is a dynamic game consisting of two
stages:
1. Star Company decides on its production capacity .

2. Coffee Bucks decides on its production capacity after having observed Star Companys
choice .
Let us find the subgame perfect Nash equilibrium of this game using backward induction. We
start at the end of the game, which is the stage where Coffee Bucks decides on its production
capacity after having observed Star Companys production capacity. Recall that given Star
Companys production capacity , Coffee Bucks revenue equals
= = (960 ) /240 = (960 ) /240.
= (960 2 )/240 = = 1.75,
which results in the following quantity for Coffee Bucks:
= 270 /2.
193

lOMoARcPSD|1733731
Now, we move to the first stage of the game, in which Star Company decides on its production
capacity. Star Companys revenue is given by
= = (960 ) /240 = (960 ) /240.
Of course, Star Company will take into account Coffee Bucks reaction given above:
(960 )
=
240

(960 (270 2 ) )
=
240

(690 2 )
=
240
= (690 )/240 = = 1.75.
Solving this equation, we find that Star Company produces
= 270
cups in equilibrium. Coffee Bucks equilibrium production capacity equals
= 270 /2 = 135.
Now, we can see that Star Company has a first-mover advantage: It sells twice as many cups of
coffee as Coffee Bucks at the same price-cost margin.
Could Coffee Bucks further exploit its first-mover advantage by building so much production
capacity that Star Company will not enter the market at all so that Coffee Bucks can enjoy a
monopoly position? The answer to this question can be confirmative if Star Company faces fixed
costs of entry. To show why, let us extend the Stackelberg game by adding the assumption that
before Coffee Bucks enters the market it has to sink fixed entry costs. Those costs may relate to
furnishing the shop, advertising, personnel training, and so on. If Coffee Bucks decides not to
enter the market, it will not incur the entry costs. The extended Stackelberg game consists of
three stages:
1. Star Company decides on its production capacity .

2. After having observed Star Companys choice , Coffee Bucks decides whether or not to
enter the market. (Only) if it enters, does it incur entry costs.
3. If Coffee Bucks enters, it decides on its production capacity .
Figures 18.118.3 depict various potential subgame perfect Nash equilibrium outcomes of the
extended Stackelberg game depending on the fixed costs of entry. Figure 18.1 shows the case in
which Coffee Bucks entry costs are very high. From the figure, it becomes clear that Coffee
Bucks profits are decreasing in the quantity Star Company can produce. Star Company deters
entry if its production capacity exceeds because Coffee Bucks will make a loss if it enters. Is it
profitable for Star Company to deter entry? It is: the monopoly quantity is greater than .
194

lOMoARcPSD|1733731
Clearly, Star Company will install exactly that production capacity because this allows it to enjoy
monopoly profits. We call this situation blockaded entry.
Figure 18.1: Blockaded entry
Figure 18.2 represents a setting where Coffee Bucks entry costs are so low that it will enter if
Star Company installs the monopoly production capacity. To deter entry, Star Company has to
increase its production capacity above the monopoly level up to . It has a good reason to do so,
because its profits equal if its production capacity equals and Coffee Bucks does not enter,
which is more than Star Company can obtain when it installs less capacity and the market
becomes a duopoly because Coffee Bucks will enter. We speak about entry deterrence.
Note that the strategy of building overcapacity is only a successful predatory strategy if the
incumbent can credibly commit to actually produce in the case Coffee Bucks does enter.
Therefore, it is important that Star Company has sunk a substantial fraction of the cost of
producing before Coffee Bucks enters. Simply announcing that it will produce may not be
credible. Why not? To fix the idea, consider the extreme case where the marginal costs of
building production capacity is zero. Star Companys best response when Coffee Bucks enters is
not but . Knowing this, Coffee Bucks has all reasons to enter the market so that Star
Companys predatory strategy fails.
Figure 18.2: Entry deterrence
195

lOMoARcPSD|1733731
Figure 18.3: Entry accommodation
Finally, figure 18.3 contains a situation in which Coffee Bucks entry costs so low that it is no
longer attractive for Star Company to deter entry. Star Companys profits in the case of deterred
entry are lower than the highest duopoly profits ( ) it can obtain when Coffee Bucks enters.
Star Company accommodates entry in this case.
18.3 Predation and competition policy

Predation by a dominant firm is considered an abuse of a dominant position under Article 102 of
the Treaty on the Functioning of the European Union. This may not be as straightforward as it
sounds. In fact, predation is one of the most controversial topics in competition policy. For one
thing, the welfare consequences of predatory behavior are ambiguous. Consider predatory
pricing. Initially, predatory pricing may increase welfare compared to normal prices because
the predator charges very low prices. (Of course, as soon as a firm leaves the market prices go up,
which may have negative welfare consequences.) Welfare may also improve when fewer firms
enter the market because it prevents wasteful duplication of entry costs.
196

lOMoARcPSD|1733731
Another reason why predation is controversial is that it is difficult to distinguish predatory

strategies from normal ones. When are prices predatory and when are they the consequence of
healthy competition? For example, when easyJet entered the London-Amsterdam route in 1996,
the incumbent Dutch carrier KLM responded with a price promotion called EasyChoice in
which it reduced its fares substantially to match easyJets. Although KLMs aggressive response
nearly bankrupted easyJet, KLM may well have only decreased the price as a response to
increased competition on the route without having the intention to drive easyJet out of the
market.
The Areeda-Turner test is supposed to distinguish predatory prices from competitive ones.
The test considers prices to be predatory if they are lower than the short-run marginal costs.
According to our analysis in chapter 3, if the price is below marginal costs, a profit-maximizing
firm has a good reason to reduce its output. So, the logic behind the Areeda-Turner test is that
the firm charges very low prices with the intent to drive competitors out of the market. The test
is even on the safe side by only considering the short-run marginal costs, i.e., it ignores part of
the marginal costs the firm has already sunk. As marginal costs are difficult to measure for
outsiders, usually the short-run average variable costs are taken as an approximation of short-
run marginal costs.
While intuitive, the Areeda-Turner test is not without criticism. Prices below short-run marginal
costs may very well be part of a healthy business strategy that firms employ without predatory
intent. For example, low prices can be part of a marketing campaign. To give an extreme example,
many firms give away free samples to acquaint the public with new products. A price of zero is
clearly below the short-run marginal costs, so the Areeda-Turner would falsely identify it as
predatory. Similarly, firms have a good reason to charge prices below short-run marginal costs
when introducing products that exhibit strong network externalities. A network externality
exists if a consumer purchasing the product has an impact on the products value for other
consumers. Telephony is a classic example: The greater the number of people owning a phone,
the more value there is attached to the phone. For the market of such goods to gain momentum,
many people should own one. Therefore, firms have a strong incentive to sell the products for
low prices to early adaptors. Yet another example is a firm charging prices below short-run
marginal costs to allow it to quickly gain experience with the goods production. Learning by
doing, the firm can lower the marginal costs substantially in the long run.
18.4 Case study: Predatory bundling in the market for web browsers
What most of us do as soon as we turn on our computer is open a web browser and start surfing
on the Internet. Nowadays there are several browsers we can choose from: Internet Explorer,
Mozilla Firefox, Google Chrome, Safari, to mention some of the most commonly used browsers.
However, it is only over the past few years that we have had such a wide choice of browsers.
Until 2006, Microsofts Internet Explorer dominated the market with about 90% market share,
and it was only in 2009 that Internet Explorer started its sharp decline. Before, Netscape
Navigator (the grandpa of Mozilla Firefox) was the main actor in the browsers market.
Developed in 1994, Netscape rapidly became the standard web browser among professional
users and home users, reaching its peak one year later with a market share above 90%. Netscape
Communications, the company behind the browser, doubled its revenues every quarter in 1995.
This fast and seemingly unstoppable growth coupled with an extraordinary initial public
offering on the stock exchange, granted Marc Andreessen, co-founder of Netscapes company,
197

lOMoARcPSD|1733731
the cover of Time Magazine with the title The Golden Geeks. However, its fall was as disastrous
as its rise was glorious. Three years later Netscapes market share dropped to 50%, and by the
end of 2004 roughly 1% of Internet users adopted Netscape. What happened? The birth of
Internet Explorer.
Microsoft saw the Netscape browser as a threat to its primary business operating systems
where the company held over 90% of market share. Netscape was available for several
operating systems, granting a similar interface and browsing experience regardless of the
operating system. With the perspective of the Internet becoming more and more important in
the computer business, the spread of such a browser would have determined a decrease in the
consumers switching costs to move from Microsoft Windows to another operating system. So
Microsoft launched Internet Explorer in 1995, and the browser war began.
At first, Internet Explorer could not compete with Netscape in terms of quality; Netscape had
more features and was overall more stable than Internet Explorer. Microsoft relied on two
strategies to fill the gap: unlike Netscape, Internet Explorer was sold for free to both Windows
and Mac users, and, most importantly, Microsoft bundled Internet Explorer with Windows, the
major operating system at the time, refusing to sell the operating systems without Internet
Explorer and preventing the users from uninstalling it. The tactic paid off, and in three years
following the launch, Internet Explorer acquired half of the browser market.
By bundling Internet Explorer with Windows, Microsoft achieved two objectives simultaneously.
On the one hand, it protected its monopoly in the operating systems market, further enhancing
the entry barriers by imposing Internet Explorer as the standard browser. On the other hand, it
used its monopoly power in the operating systems market to drive the competitor in the
browser market out of the market. This strategy is often called monopoly leverage, whereby a
firm with monopoly power in one market can use bundling as a leverage to monopolize the
bundled good market.
According to some economists, monopoly leverage can be seen as a predatory practice. With this
allegation, Microsoft was sued by the US Department of Justice in 1998. According to the
competition authority, bundling Internet Explorer with Windows was an anti-competitive move
purely designed to push Netscape out of the browser business. Microsoft and its CEO Bill Gates
defended themselves, underlining how the bundle was beneficial to consumers by giving them
more choices at lower prices. In addition, they claimed that Internet Explorer was an integrated
part of Windows, without which the operating system would not work properly. The case soon
became public debate, dividing public opinion and even academics. Several well-known
Industrial Organization economists took part in the debate, favoring either the Department of
Justice or Microsoft. (Most notably, two professors of MIT, Richard Schmalensee and Franklin
Fisher, took part in the trial as expert witnesses: the former in favor of Microsoft, the latter in
favor of the Department of Justice.) The trial ended with a conviction in the first stance, and a
settlement in appeal. However, it was too late for Netscape, which saw its market share being
eroded day after day.
The Industrial Organization textbooks cited in previous chapters all include a comprehensive
review of predatory conduct. For extensions and empirical applications of entry deterrence and
pricing policies of predatory conduct, you can study Spence (1977), Ghemawat (1984), Hall
198

lOMoARcPSD|1733731
(1990) and Baldwin (1995). The literature on predatory pricing is also extensive. Dixit (1980),
Milgrom and Roberts (1982), Burns (1986) and Bolton and Scharfstein (1990) will give you an
overview of the fundamental practices. Although the articles date almost three decades back,
they are still influential in up-to-date research and provide frameworks able to explain recent
empirical observations. If you want to read more about the Areeda-Turner test you can refer to
Areeda and Turner (1975, 1976). There you can find a formal exposition of the intuitive
summary of their test that was presented in this chapter.
Gilbert and Katz (2001), Klein (2001), and Whinston (2001) discuss the Microsoft case in a
symposium in the Journal of Economic Perspectives devoted to the case.
References
Areeda, P. and D.F. Turner (1975). Predatory pricing and related practices under section 2 of the
Sherman Act, Harvard Law Review, 88(4), 697-733.
Areeda, P. and D.F. Turner (1976). Scherer on predatory pricing: a reply, Harvard Law Review,
89(5), 891-900.
Baldwin, J.R. (1995). The Dynamics of Industrial Competition: A North American Perspective,
Cambridge: Cambridge University Press.
Bolton, P. and D.S. Scharfstein (1990). A theory of predation based on agency problems in
financial contracting, American Economic Review, 80, 93-106.
Burns, M.R. (1986). Predatory pricing and the acquisition cost of competitors, Journal of Political
Economy, 94(2), 266-296.
Dixit, A. (1980). The role of investment in entry-deterrence, Economic Journal, 90(357), 95-106.
Gilbert, R.J. and M.L. Katz (2001). An economist's guide to U.S. vs. Microsoft, Journal of Economic
Perspectives, 15(2), 25-44.
Ghemawat, P. (1984). Capacity expansion in the titanium dioxide industry, Journal of Industrial
Economics, 33(2), 145-163.
Hall, E.A. (1990). An analysis of preemptive behavior in the titanium dioxide industry,
International Journal of Industrial Organization, 8(3), 469-484.
Klein, B. (2001). The Microsoft case: what can a dominant firm do to defend its market position?,
Journal of Economic Perspectives, 15(2), 45-62.
Milgrom, P. and J. Roberts (1982). Limit pricing and entry under incomplete information: an
equilibrium analysis, Econometrica, 50(2), 443-459.
Spence, A.M. (1977). Entry, capacity, investment and oligopolistic pricing, Bell Journal of
Economics, 8(2), 534-544.
Whinston, M.D. (2001). Exclusivity and tying in U.S. vs. Microsoft: what we know and dont know,
Journal of Economic Perspectives, 15(2), 63-80.
Exercises
Exercise 18.1 Stackelberg

Three producers of espresso coffee machines want to enter the Chinese coffee market, which is
currently undiscovered by any other company. The three firms (for simplicity, let us call them
firm 1, firm 2 and firm 3) have the same cost functions ( ) = 20 , where is the number of
espresso machines each company produces, with = 1,2,3. After a preliminary investigation of
the market, the analysts report that the Chinese demand for coffee machines can be described as:
199

lOMoARcPSD|1733731
1
() = 460 ,
2
where is the price of a coffee machine. However, entry may have some delays due to
bureaucracy procedures, hence the three companies may not enter the market simultaneously.
Assume that firm 1 is the first to enter the market, followed by firm 2 and lastly firm 3. They
compete la Stackelberg (that is, the firms who entered earlier can anticipate the behavior of the
followers).
a) Compute equilibrium quantities, price, and profits.
Now assume that firm 1 is the first to enter the market, followed by firm 2 and firm 3 which
enter simultaneously.
b) Compute again equilibrium quantities, price, and profits.
c) Calculate the equilibrium quantity, price, and profits assuming that firm 1 is a monopolist.
d) What do you conclude with respect to the quantity firm 1 produces under a), b) and c)?
Exercise 18.2 Stackelberg

The Italian city of Turin gave birth to the famous chocolate named the gianduiotto. Invented by
the firm Caffarel in 1865, the gianduiotto is made with chocolate and cream of crushed, toasted
hazelnuts. The idea of mixing cocoa with hazelnuts was in response to Napoleons Continental
Blockade in the early 1800s, which drastically increased the price of raw cocoa. Assume that in
the late 1800s another firm, called Vecchio, entered the market of gianduiotti and competed with
the incumbent Caffarel. Vecchio used another recipe than Caffarel, and it used a different
wrapping. Therefore, Caffarels and Vecchios gianduiotti were only imperfect substitutes and
they received two different demands, each function of the prices of both firms:
(, ) = 180 10( )
and , = , for Caffarel and Vecchio respectively, and represent the prices of the two
firms for a box of gianduiotti. Due to the special form of the demands (a negative function for the
own price, but a positive function of the competitors price), Caffarel and Vecchio competed on
prices. Assume that both firms do not incur any cost of production.
a) Write down Vecchios reaction function.
b) Calculate the equilibrium quantities, price, and profits that result from Stackelberg
competition on prices, i.e., when Caffarel fixes its price first, followed by Vecchio.
c) What is the difference in terms of outcome with the regular Stackelberg competition on
quantities?
Exercise 18.3 Building overcapacity

Consider a homogeneous product industry with inverse market demand given by = 1100
2. There is currently one incumbent firm (firm 1) and one potential competitor (firm 2). Entry
into this industry implies a fixed cost of 450. Each firm has constant marginal cost of 300.
200

lOMoARcPSD|1733731
a) Calculate firm 1's profit-maximizing output and profits in the absence of potential
competition.
b) Determine the Nash equilibrium in the case that firms 1 and 2 compete la Cournot.
Now, suppose that firms 1 and 2 compete la Stackelberg: Firm 1 decides on its production
capacity first, followed by firm 2.
c) Calculate the output and profits of both firms in case firm 1 accommodates entry.
d) Does firm 1 enjoy a first-mover advantage, i.e., are its profits greater under Stackelberg
competition than under Cournot competition?
e) Calculate the output firm 1 should set to deter entry.
f) Calculate consumer surplus in the case of entry deterrence and in the case of entry
accommodation. Which situation leads to the largest consumer surplus?
g) In the case of entry accommodation, how far below 300 would firm 2's marginal cost have to
fall so that it achieved the same market share as firm 1? In answering the question, assume that
firm 1 knows this marginal cost when deciding on its own quantity. Furthermore assume that
the marginal cost of the leader is still 300.
Exercise 18.4 Building overcapacity

Woody has finally become the monopolist of the Amsterdams oranges market. However, he is
now facing the potential entry of James, a rampant new oranges seller. Both have the same cost
function () = 8, where represents tons of oranges. Since Woody is the incumbent in the
market, the competition is la Stackelberg. In addition, James faces a fixed cost when entering
the market. The demand for oranges can be summarized as follows:
() = 120 3,
where is the price for a ton of oranges.
a) Show that James profits are decreasing in , the quantity produced by Woody.
b) How many tons of oranges should Woody produce in order to deter entry? Compute it as a
function of .
c) Assume that = 27. Will Woody deter entry?
d) Assume now that = 12. Will Woody deter entry?
Imagine that one period has passed, and Woody is able to rapidly change the quantity he
produces without incurring any costs. In other words, the quantity choice decided in the
previous period does not represent a strong commitment over time.
e) How would your answer in part c) and d) change?
f) Find the minimum entry cost () that would block James entry. (Hint: Recall what blockaded
entry implies for the incumbent).
201

Very Extensive Lectures and The Book Economics of Markets and Organizations

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Very Extensive Lectures and The Book Economics of Markets and Organizations

Hochgeladen von

Copyright:

Verfügbare Formate

lOMoARcPSD|1733731

Economics of Markets and Organizations

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

4.2 Nash equilibrium ................................................................................................................................................ 46

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

13.1 The adverse selection problem ............................................................................................................... 142

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Chapter 17: Make or buy ........................................................................................................................................... 181

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Chapter 1: Organizations and efficiency

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Figure 1.1: Inverse demand for Alfa Benz

= = ( ) = 3 (20 16) = 12.

Welfare is the sum of consumer surplus and producer surplus, i.e.,

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Figure 1.2: Welfare generated by Alfa Benz

which is indeed higher than welfare at a price of 20.

1.2 General equilibrium

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

The fundamental theorem of welfare economics: An efficient allocation of goods

1.3 Perfect competition

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Figure 1.3: Towards an equilibrium in a perfectly competitive market

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

1.4 Market failures

Information asymmetry is another source of market failure. Information asymmetry emerges

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

1.5 Set-up and research methods

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

1.6 Case study: Apple Inc.

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

The fundamental theorem of welfare economics was mathematically demonstrated principally

Exercise 1.1 Welfare

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Exercise 1.2 Utility and consumer surplus

Exercise 1.3 Perfect competition

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Exercise 1.4 Perfect competition and entry decisions

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Chapter 2: Organizational Economics

2.1 Hiring decisions

Table 2.1: Berts and Ernies yearly production and salary

Education Yearly production Yearly salary

Profit per Profit per unit Number of Employee

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

2.2 The Principal-Agent problem

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

2.3 Critical assumptions in the Principal-Agent model

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)

Exercise 2.1 Shirking workers and the efficiency wage

Exercise 2.2 Linear and non-linear incentive contracts

Exercise 2.3 The optimal bonus

Distributing prohibited | Downloaded by Vlad SURDEA-HERNEA (surdea-hernea.vlad@fspub.unibuc.ro)