PDF Market Design A Linear Programming Approach To Auctions and Matching Martin Bichler Ebook Full Chapter

Market Design A Linear Programming
Approach to Auctions and Matching

Martin Bichler
Visit to download the full and correct content document:
https://textbookfull.com/product/market-design-a-linear-programming-approach-to-auc
tions-and-matching-martin-bichler/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Groups Matrices and Vector Spaces A Group Theoretic

Approach to Linear Algebra James B. Carrell
https://textbookfull.com/product/groups-matrices-and-vector-
spaces-a-group-theoretic-approach-to-linear-algebra-james-b-
carrell/
Introduction to Linear Programming with MATLAB 1st

Edition Shashi Kant Mishra
https://textbookfull.com/product/introduction-to-linear-
programming-with-matlab-1st-edition-shashi-kant-mishra/
JavaScript Object Programming Rinehart Martin
https://textbookfull.com/product/javascript-object-programming-
rinehart-martin/
Beginning PyQt A Hands on Approach to GUI Programming

Willman Joshua M
https://textbookfull.com/product/beginning-pyqt-a-hands-on-
approach-to-gui-programming-willman-joshua-m/
Beginning PyQt A Hands on Approach to GUI Programming
Willman Joshua M
https://textbookfull.com/product/beginning-pyqt-a-hands-on-
approach-to-gui-programming-willman-joshua-m-2/
Medical Instrument Design and Development From

Requirements to Market Placements Claudio Becchetti
https://textbookfull.com/product/medical-instrument-design-and-
development-from-requirements-to-market-placements-claudio-
becchetti/
Vector Calculus Linear Algebra and Differential Forms A

Unified Approach 5th Edition John Hubbard
https://textbookfull.com/product/vector-calculus-linear-algebra-
and-differential-forms-a-unified-approach-5th-edition-john-
hubbard/
Statistics for the Social Sciences A General Linear

Model Approach Russell Warne
https://textbookfull.com/product/statistics-for-the-social-
sciences-a-general-linear-model-approach-russell-warne/
Quantum Computing Fundamentals From Basic Linear

Algebra to Quantum Programming 1st Edition Easttom Ii
https://textbookfull.com/product/quantum-computing-fundamentals-
from-basic-linear-algebra-to-quantum-programming-1st-edition-
easttom-ii/
Cambridge University Press
978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
Market Design
A Linear Programming Approach to Auctions and Matching
The digital economy has led to many new services where supply is matched with
demand for various types of goods and services. More and more people and organi-
zations are now in a position to design market rules that are being implemented in
software.
The design of markets is challenging as it is necessary to consider the strategic
behavior of market participants, psychological factors, and computational problems in
order to implement the objectives of a designer. Market models in economics have not
lost their importance, but recent years have led to many new insights and principles
for the design of markets which are beyond traditional economic theory. This book
introduces the fundamentals of market design, an engineering field concerned with the
design of real-world markets.
Martin Bichler is Professor of Informatics at the Technical University of Munich (TUM),

and a faculty member at the TUM School of Management. He is known for his aca-
demic work on market design, and he has acted as a consultant for private and public
organizations including regulators, telecoms, and procurement organizations. Projects
in which he is involved include the design of auctions for industrial procurement, logis-
tics, advertising, fishery access rights, and spectrum sales. His research addresses algo-
rithmic, game-theoretical, and behavioral questions and has appeared in leading journals
in computer science, economics, operations research, and management science. He is
currently Editor of Business and Information Systems Engineering and serves on the
editorial boards of several academic journals.
© in this web service Cambridge University Press www.cambridge.org

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
Market Design
A Linear Programming Approach to
Auctions and Matching
MARTIN BICHLER
Technical University of Munich

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
University Printing House, Cambridge CB2 8BS, United Kingdom

One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi - 110025, India
79 Anson Road, #06-04/06, Singapore 079906
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9781107173187
DOI: 10.1017/9781316779873
© Martin Bichler 2018
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2018
Printed in the United States of America by Sheridan Books, Inc.
A catalogue record for this publication is available from the British Library
ISBN 978-1-107-17318-7 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
To my wife Claudia and my daughters Mona and Sonja

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
Contents
1 Introduction page 1
1.1 Market Design and Mechanism Design 2
1.2 Market Design and Mathematical Optimization 3
1.3 Outline of the Book 4
1.3.1 Part I Microeconomic Fundamentals 5
1.3.2 Part II Multi-Object Auction Design 6
1.3.3 Part III Approximation and Matching Markets 7
1.3.4 Part IV Appendices: Mathematical Optimization 7
1.4 Acknowledgements 8
Part I Microeconomic Fundamentals
2 Game-Theoretical Basics 11
2.1 Normal-Form Games 11
2.1.1 Dominant-Strategy Equilibrium 13
2.1.2 Pareto Optimality 13
2.1.3 Nash Equilibrium 14
2.1.4 Correlated Equilibrium 17
2.1.5 Further Solution Concepts 19
2.2 Extensive-Form Games 20
2.3 Bayesian Games 22
2.3.1 Bayesian Nash Equilibrium 24
2.3.2 Ex Post Equilibrium 25
2.4 Games and Human Behavior 26
2.5 Summary 27
2.6 Comprehension Questions 28
2.7 Problems 28
3 Mechanism Design 30
3.1 Social Choice 31
3.1.1 Voting Rules 31
3.1.2 Arrow’s Impossibility 33

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
viii Contents
3.2 Utility Functions 35

3.3 Mechanism Design Theory 39
3.4 Quasi-Linear Mechanism Design 42
3.4.1 Quasi-Linear Utility Functions 42
3.4.2 The Vickrey–Clarke–Groves Mechanism 44
3.4.3 The Myerson–Satterthwaite Theorem 46
3.5 Summary 48
3.5.1 Robust Mechanism Design 48
3.5.2 Algorithmic Mechanism Design 49
3.5.3 Dynamic Mechanism Design 49
3.5.4 Non-Quasi-Linear Mechanism Design 50
3.7 Problems 51
4 Single-Object Auctions 53
4.1 Single-Object Auction Formats 53
4.2 Model Assumptions 54
4.3 Equilibrium Bidding Strategies in the IPV Model 55
4.3.1 Ascending Auctions 56
4.3.2 Second-Price Sealed-Bid Auctions 56
4.3.3 First-Price Sealed-Bid Auctions 57
4.3.4 Descending Auctions 60
4.4 Comparing Equilibrium Outcomes 60
4.5 Robustness of the Revenue Equivalence Theorem 62
4.5.1 Risk-Averse Bidders 62
4.5.2 Interdependent Values 64
4.5.3 Asymmetry of Bidders 65
4.5.4 Uncertainty about the Number of Bidders 65
4.6 Bidder Collusion 65
4.7 Optimal Auction Design 66
4.8 Selected Experimental Results 68
4.9 Summary 70
4.11 Problems 71
Part II Multi-Object Auction Design
5 An Overview of Multi-Object Auctions 75

5.1 General Equilibrium Models 75
5.2 Multi-Unit Auction Formats 77
5.2.1 Sealed-Bid Multi-Unit Auction Formats 77
5.2.2 Open Multi-Unit Auction Formats 78
5.2.3 Sequential Sales 80

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
Contents ix
5.3 Multi-Item Auction Formats 81

5.3.1 Simultaneous Auctions 81
5.3.2 Combinatorial Auctions 81
5.4 Online and Dynamic Auction Design 83
5.5 Summary 83
6 The Simultaneous Multi-Round Auction Format 85

6.1 SMRA Rules 85
6.2 Tactics in the SMRA 86
6.3 Strategic Situations in SMRA 88
6.3.1 War of Attrition 89
6.3.2 English Auction vs. War of Attrition 93
6.4 Summary 95
7 Sealed-Bid Multi-Object Auctions 96

7.1 Generic Bid Languages 96
7.2 The Winner Determination Problem 98
7.3 Payment Rules 101
7.3.1 Pay-as-Bid Payment Rules 101
7.3.2 Vickrey–Clarke–Groves Payment Rules 101
7.3.3 Bidder-Optimal Core Payment Rules 105
7.4 Equilibrium Bidding Strategies 107
7.4.1 First-Price Sealed-Bid Auctions 107
7.4.2 Bidder-Optimal Core-Selecting Auctions 111
7.5 Domain-Specific Compact Bid Languages 111
7.5.1 Procurement Markets with Economies of Scale
and Scope 112
7.5.2 Distributed Scheduling in TV Ad Markets 117
7.6 Combinatorial Double Auctions 120
7.7 Empirical Results 121
7.8 Summary 122
7.10 Problems 124
8 Open Multi-Object Auctions 126

8.1 Primal–Dual Auctions for Assignment Markets 127
8.1.1 Dual Prices and VCG Payments 128
8.1.2 An Ascending Auction for the Assignment Problem 131
8.2 Greedy Auctions and Matroids 139

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
x Contents
8.3 Models of Open Combinatorial Auctions 144

8.3.1 Limits of Linear Prices 144
8.3.2 Algorithmic Models of Ascending Auctions 150
8.3.3 Perfect Bayesian Equilibria with General Valuations 152
8.3.4 Large Markets with Non-Convexities 154
8.4 Overview of Open Combinatorial Auction Formats 155
8.4.1 Auction Formats with Non-Linear Prices 155
8.4.2 Auction Formats with Linear Ask Prices 158
8.5 Empirical Results 162
8.5.1 Experiments with Small Markets 162
8.5.2 Experiments with Larger Markets 163
8.6 Summary 164
8.8 Problems 165
9 The Combinatorial Clock Auction Formats 167

9.1 The Single-Stage Combinatorial Clock Auction 167
9.1.1 Auction Process 167
9.1.2 Efficiency of the SCCA 168
9.2 The Two-Stage Combinatorial Clock Auction Format 172
9.2.1 Auction Process 172
9.2.2 Activity Rules 173
9.2.3 A Note on Revealed Preference Theory 175
9.2.4 Strategies in the Two-Stage CCA 178
9.3 Experiments on the Two-Stage CCA 183
9.4 Summary 185
9.6 Problems 185
Part III Approximation and Matching Markets
10 Approximation Mechanisms 189

10.1 Approximation and Truthfulness 190
10.1.1 Deterministic Approximation Mechanisms 191
10.1.2 Randomized Approximation Mechanisms 191
10.2 Deterministic Mechanisms for Single-Minded Bidders 193
10.2.1 Greedy-Acceptance Auctions 194
10.2.2 Deferred-Acceptance Auctions 196
10.3 Randomized Mechanisms 198
10.3.1 The Relax-and-Round Framework 199
10.3.2 Combinatorial Auctions via Relax-and-Round 201
10.4 Summary 204

978-1-107-17318-7 — Market Design
Martin Bichler
Frontmatter
More Information
Contents xi
11 Matching Markets 206

11.1 Overview of Matching Problems 206
11.2 Two-Sided Matching 208
11.2.1 Definitions and Notation 209
11.2.2 The Gale–Shapley Student-Optimal Stable Mechanism 213
11.2.3 The Efficiency-Adjusted Deferred-Acceptance Mechanism 217
11.3 One-Sided Matching 219
11.3.1 The Top Trading Cycle Mechanism 220
11.3.2 The Probabilistic Serial Mechanism 223
11.4 One-Sided Matching with Complementarities 225
11.4.1 Multi-Unit and Combinatorial Assignments 225
11.4.2 Cardinal Assignment Problems with Complementarities 226
11.5 Applications and Empirical Results 231
11.6 Summary 233
11.6.1 Two-Sided Matching Mechanisms 233
11.6.2 One-Sided Matching Mechanisms 235
11.8 Problems 236
12 Outlook 238
12.1 Challenges in Market Design 239
12.2 Going Forward 240
Part IV Appendices: Mathematical Optimization
A Linear Optimization 245

A.1 Geometry of Linear Programs 245
A.2 Feasibility 249
A.3 Duality Theory 250
A.4 Integrality of Linear Programs 254
B Algorithms and Complexity 256

B.1 Computational Complexity 256
B.2 Algorithms for Linear Programs 258
B.3 Algorithms for Integer Linear Programs 258
B.4 Approximation Algorithms 262
B.4.1 Complexity Classes 263
B.4.2 LP-Based Approximation Algorithms 264
References 268
Index 281

1 Introduction
Market design is a kind of economic engineering, utilizing laboratory research, game theory,
algorithms, simulations, and more. Its challenges inspire us to rethink longstanding fundamentals
of economic theory.
Paul Milgrom, 2008
The digital economy has led to many new services where supply is matched with
demand for various types of goods and services. While only a few years ago academics
were mainly concerned with models describing markets, more and more people and
organizations are now in a position to design market rules that are being implemented
in software. The design of markets is challenging as it is necessary to consider the
strategic behavior of market participants, psychological and cognitive factors, and com-
putational problems in order to implement the objectives of a designer. Market models
in economics have not lost their importance, but recent years have led to many new
insights and principles for the design of markets which are beyond traditional economic
theory. In this book, we study market design, an engineering field at the intersection of
computer science, economics, information systems, and operations research concerned
with the design of real-world markets.
Consider a transportation market with multiple lanes on a transportation network,
one shipper and multiple carriers. The shipper has a set of truckloads to be shipped
from different origins to various destinations. The carriers are available to meet the
transportation demand, and they are invited by the shipper to submit sealed bids.
Carriers typically have preferences for bundles of lanes on a route such that there are
no empty lanes without a shipment, and they also submit respective bids on packages
of lanes. This simple logistics procurement example leads to challenges for the shipper.
First, the shipper needs to determine an optimal allocation of bundle bids such that his
costs are minimized. Cost minimization is a computationally hard optimization problem
in this setting. Second, carriers want to maximize their payoff and it is not necessarily in
their interest to reveal their costs truthfully. They might charge a high markup on some
lanes where they expect no competition or they might not bid on other lanes where
they expect high competition. However, strategic manipulation of this sort can lead to
suboptimal allocations of lanes to carriers, high procurement costs for the shipper, and
high bid preparation costs for the carriers, who would benefit from information about
their competitors. Ideally, a shipper would have an economic mechanism where carri-
ers have incentives to reveal their costs truthfully, and he or she can then determine
15:27:03
01
2 Introduction
the cost-minimal allocation optimally. In sales auctions the objectives are typically the
maximization of auctioneer revenue or overall welfare. The latter is also referred to as
(allocative) efficiency.
Overall, there are fundamental strategic and computational challenges in the design
of such multi-object auctions, which are at the core of market design. The 2012 Nobel
Memorial Prize in Economic Sciences for Alvin Roth and Lloyd Shapley honored
research in this field. Market design uses economic theory, mathematical optimization,
systems design, experiments, and empirical analysis to design market rules and insti-
tutions. Fundamentally, it asks how the design of the rules and regulations of a market
affects the functioning and outcomes of that market. The study includes auction mar-
kets but also markets without money such as matching markets, which are used in the
assignment of students to courses, in school choice programs, and in kidney exchanges.
This textbook focuses on the design and analysis of efficient multi-object market
mechanisms.
1.1 Market Design and Mechanism Design
Market design has theoretical foundations in mechanism design theory, an analytical

framework for thinking about what a given economic institution can achieve when
the information necessary to make decisions is dispersed and privately held. The 2007
Nobel Memorial Prize in Economic Sciences to Leonid Hurwicz, Eric S. Maskin, and
Roger B. Myerson was awarded in this field of economic theory. Mechanism design uses
an axiomatic method, deriving results from a number of basic assumptions about utility
functions or overall design goals of a mechanism. For example, mechanism design has
been used to characterize the utility functions of market participants and mechanisms for
which the truthful revelation of preferences is an equilibrium, i.e., a situation from which
no participant wants to deviate (aka incentive-compatible mechanisms). The mechanism
design literature shows that environments which allow for welfare maximization such
that participants cannot make a loss and have strong incentives to bid truthfully are
limited and require strong assumptions. The celebrated Vickrey–Clarke–Groves mech-
anism provides dominant strategies for agents to reveal valuations truthfully if they have
independent and private values and they can maximize their payoff. If these assumptions
are relaxed, so that bidders can have private budget constraints or valuations that are not
independent, such positive results no longer exist (unless preferences can be character-
ized by a single parameter only). Overall, environments where truthful bidding satisfies
a strong equilibrium solution concept are quite restricted. Often neither the bidders’
preferences in real markets satisfy the assumptions necessary for truthful mechanisms
nor is it possible to implement the required mechanisms for practical reasons. Actually,
the Vickrey–Clarke–Groves mechanism is rarely used in the field. Still, it is important
to understand the assumptions and mechanisms which would make truthful bidding an
equilibrium bidding strategy.
Market design starts from the requirements in the field, eliciting the preferences
and financial constraints of participants (which might be different from pure payoff
15:27:03
01
1.2 Market Design and Mathematical Optimization 3
maximization), information about competitors available to participants, objectives, and

constraints on the allocation problem and on the award process. In many cases these
requirements differ from those assumed in mechanism design theory, demanding the
development of new models that capture the specifics of a certain type of market. Exam-
ples are models of markets for procurement and logistics, for spectrum sales, or for dis-
play advertising. Often the specifics of these markets might not allow for mechanisms
which are incentive-compatible according to the established equilibrium solution con-
cepts. So, market design takes the market environment as given and derives designs that
satisfy some design goals (such as stability or budget balance of the outcome), while
relaxing others. In other words, it analyzes tradeoffs and aims for “satisficing” solu-
tions (Simon, 1996) to real types of market design problems rather than finding sets of
assumptions that allow for “optimal” designs according to some design desiderata.
Market design complements mechanism design in the level of detail that is specified
and analyzed. For example, in mechanism design valuations are typically described as
an abstract function of the objects to be allocated, while the parametric forms of the val-
uation or cost functions, of the utility functions, and of the corresponding bid languages
play a significant role in market design. Moreover, market designs almost always need
to consider allocation constraints such as restrictions on the number of winners or the
quantity awarded to individual bidders or groups of bidders. Therefore market design
typically starts out as a mathematical programming task concerned with the design of
an appropriate objective function, of appropriate constraints for the overall allocation,
and of activity rules and with the definition of a bid language that lets bidders express
their preferences in a succinct way. Milgrom (2017) provides an excellent and up-to-date
view on market design and how it relates to economic theory.
1.2 Market Design and Mathematical Optimization
One thesis of this book is that mathematical optimization, in particular linear and inte-
ger linear programming, plays a central role in the design and analysis of multi-object
markets and also provides a central foundation for theoretical models of multi-object
markets. Many techniques and models introduced in this book use the theory of linear
and integer linear programming in one way or another. Compared with algorithms
designed for specific problems, there is an amazing versatility in linear and integer
linear programming, which makes these techniques very important for market design.
Algorithms for the solution of integer linear programs have seen substantial progress
in the past two decades, allowing for totally new types of markets in different domains.
Before that, in 1975 Leonid Kantorovich and Tjalling Koopmans had already received
the Nobel Memorial Prize in Economic Sciences for their work on linear program-
ming and optimal resource allocation. However, the advances in recent years have
been staggering and have made many new markets possible that were not considered
tractable just a few years ago. The subtitle of this book is reminiscent of the seminal
book Mechanism Design: A Linear Programming Approach by Vohra (2011), who
emphasized the role of linear programming in mechanism design theory.
15:27:03
01
4 Introduction
However, market design is more than mathematical programming, as it takes into

consideration bidding strategies and human behavior. Ignoring bidding strategies and
different types of manipulation in market-based resource allocation is like optimizing
a problem with the wrong parameters. As indicated earlier, strong game-theoretical
solution concepts such as dominant-strategy or ex post incentive compatibility are
desirable but they might not always be attainable, as a number of impossibility results
in mechanism design show. These properties rest on assumptions about bidders’ utility
functions, which are often not given in the field. Still, in many markets it is possible
to devise market rules for which manipulation becomes hard, given the uncertainties
about other bidders and their valuations.
Overall, market design aims at design principles and rules for market institutions
which are robust against strategic manipulation and allow bidders to express their pref-
erences, so that the designer can aim for good or even optimal allocations. This is similar
to network security, where designers aim for secure network protocols that are hard to
tamper with, knowing that there is no absolute security. Note that “optimality” often
refers to social welfare maximization, but the market designer might have different
objectives or policy goals. In this sense, market design extends mathematical program-
ming to situations with multiple decision makers. It is a design science in the sense
of Herb Simon who received both the Nobel Memorial Prize in Economic Sciences
(1978) and the Turing Award in Computer Science (1975). Simon’s book The Sciences
of the Artificial (Simon, 1996) motivated the development of systematic and formalized
design methodologies relevant to many design disciplines, for example architecture,
engineering, urban planning, medicine, computer science, and management science.
Market design is therefore a very suitable name for the study of principles and methods
in designing markets in the spirit of design science.
1.3 Outline of the Book
This textbook is intended to provide a one-semester course on market design. I am pri-

marily targeting students with a background in computer science, information systems,
mathematics, and management science. Hence, I will first introduce in Part I necessary
concepts from game theory and mechanism design, as these students typically have not
had the respective introductory courses. Parts II and III cover material which is more
recent and is often not covered in microeconomics.
One prerequisite for this book is a familiarity with linear and integer linear program-
ming and an introductory course in calculus and probability theory. There are many
introductory textbooks on these subjects which would provide an excellent start for the
topics discussed throughout this book. The appendices summarize important results
from mathematical optimization and should serve as a convenient reference for the
reader and a brief introduction for those who have not studied the respective courses.
As outlined earlier, this book focuses on the allocation of multiple objects with dis-
tributed decision makers. While single-object auctions are fairly well understood, multi-
object auctions provide many more design options and they are more challenging to
15:27:03
01
design and analyze. The goal of this book is to introduce important models and princi-
ples from game theory, mechanism design, and single-object auction theory for which
there are several excellent textbooks. After basic concepts have been introduced, the
book focuses on multi-object markets and their properties.
This leads to the three parts of the book: Part I (Microeconomic Fundamentals)
discusses basic concepts from game theory, mechanism design, and single-object
auction theory. The goal of this part is not to discuss each of these topics in breadth but
to introduce the terminology and theory required for Parts II and III. The above topics
are typically not taught in computer science and management science, and this makes it
hard for students from these fields to understand the literature in market design. Part II
(Multi-Object Auction Design) introduces the more recent theory and specific designs
for multi-object markets. Part III (Approximation and Matching Markets) analyzes
approximation mechanisms which have been designed for markets where the designer
cannot hope to solve the allocation problem optimally. Then the design of matching
markets where monetary transfers are not possible is discussed. Approximation and
randomization play an important role in the recent matching literature. Part IV (Appen-
dices: Mathematical Optimization) summarizes the main results from linear and integer
linear optimization, on which I draw in Parts II and III.
There is always a tradeoff between the breadth and the depth of a textbook, and it is
a matter of choice how much space is devoted to each single topic. If all the various
literature streams in computer science, economics, and operations research were cov-
ered then a book on market design could easily cover thousands of pages, and it would
clearly be beyond a one-semester course. This book provides a fair amount of detail in
Part II and also in Part III, but Parts I and IV are limited to a necessary minimum for
readers who have not taken the respective introductory courses. Selected references to
the original literature are included in the text for those interested in learning more about
a specific topic and in reading the original literature.
1.3.1 Part I Microeconomic Fundamentals

Chapter 2 introduces basic game-theoretical notions and solution concepts relevant
for market design problems. Solution concepts for normal-form, extensive-form, and
Bayesian games are introduced. There are many excellent textbooks on game theory.
For example, Shoham and Leyton-Brown (2011) provides a comprehensive introduc-
tion for computer scientists, while I cover only a subset of the topics relevant for later
chapters.
In Chapter 3 I discuss mechanism design theory, also known as inverse game theory.
While the rules of a game are given in game theory, mechanism design tries to design
rules such that certain goals and solution concepts are achieved. Market design prob-
lems can be considered as games where participants should have incentives to reveal
their preferences for objects truthfully. If participants were truthful then the market
designer would only need to solve an optimal allocation problem. The first section shows
that aggregating general and unrestricted preferences is hard and that simple truthful
mechanisms for general preferences are impossible. Then preferences are restricted to
15:27:03
01
6 Introduction
independent and private valuations and quasi-linear utility functions, where agents max-
imize payoff and monetary transfers are allowed. This restriction of preferences allows
for a truthful mechanism with a dominant-strategy equilibrium. This means that bid-
ders do not need information about competitors to bid optimally. The Vickrey–Clarke–
Groves (VCG) mechanism is a celebrated result which is pivotal for much of the later
discussion on auction design.
Chapter 4 introduces the traditional literature on single-object auctions. This litera-
ture provides a consistent framework for thinking about phenomena in auction markets,
and it formalizes various assumptions. Equilibrium bidding strategies are discussed in
a number of standard auction formats, before the outcomes of these auctions in equilib-
rium are compared. Krishna (2009) provides an excellent textbook which covers auction
theory with a focus on Bayesian equilibrium analysis. This chapter is limited to key con-
cepts providing a framework for the analysis of multi-object auctions later on.
1.3.2 Part II Multi-Object Auction Design

Chapter 5 introduces a taxonomy of multi-object auctions. Multi-unit auctions (for
homogeneous objects) and multi-item auctions (for heterogeneous objects) are distin-
guished. For both types of auction, sealed-bid and open auction formats have been
designed. Open auctions include those that are continuous or organized iteratively in
rounds. They can be ascending or descending but always reveal some information about
competitors throughout the process. Multi-item auctions are ubiquitous in industrial
procurement, in logistics, and also in the public sector (Sandholm, 2012). Much of the
remaining discussion is focused on multi-item auctions.
Chapter 6 is devoted to a simple and widespread multi-item auction format, the
simultaneous multi-round auction (SMRA). It is a practical auction design and pro-
vides an excellent way to introduce some problems that arise in multi-item auctions.
The strategic problems in SMRAs led to the development of combinatorial auctions.
Combinatorial auctions allow bidders to express all types of preferences including
complements and substitutes and can therefore be considered as the most general class
of multi-object auction designs.
Chapter 7 introduces sealed-bid combinatorial auctions. The winner determination
problem and the respective bid languages are discussed, before the various different
payment rules are introduced. The VCG mechanism also leads to dominant-strategy
equilibria in combinatorial auctions with payoff-maximizing bidders. However, in
contrast with single-object auctions, this mechanism does not necessarily yield a
competitive equilibrium, i.e., a stable solution where nobody wants to deviate. Also,
the number of possible packages grows exponentially, so that a VCG mechanism
might not be practical. The chapter introduces alternative payment rules and illustrates
domain-specific bid languages addressing these issues. Combinatorial double auctions
are also discussed, but the literature in this field is scarce.
Chapter 8 focuses on open and iterative combinatorial auctions, where bidders
are able to outbid each other much as in an English auction. Assignment markets
are a restricted type of market, where each bidder wants to win at most one out of
15:27:03
01
several items. This restricted environment allows for ascending auctions with dominant
strategies, and, what is more, they provide an excellent environment to illustrate algo-
rithmic models of auctions. These types of model have received much recent attention
and complement the literature on sealed-bid auctions. Also, in open multi-item auctions
it is even more important to understand the interactions of bidders than it is in the
single-item counterparts of these auctions. Note that most high-stakes multi-item
auctions are open in format. Unfortunately, open combinatorial auctions cannot always
have dominant strategies if modeled as a Bayesian game, where bidders have only
distributional information about their competitors.
In Chapter 9 I discuss the single-stage and two-stage combinatorial clock auction
formats. These auction formats have been used for the sale of spectrum licenses by gov-
ernments around the world and provide an excellent example of practical market designs
addressing the complexities of multi-object auctions. I also consider some pitfalls and
problems that can arise in these auction formats.
1.3.3 Part III Approximation and Matching Markets

Chapter 10 analyzes market design problems in situations where the allocation problem
is not tractable. Actually, the allocation problem of many real-world market design prob-
lems can be described as a combinatorial optimization problem and so computationally
hard to solve. Unfortunately, the VCG mechanism provides dominant strategies only
if the allocation problem can be solved exactly. Approximation mechanisms solve the
allocation problem in polynomial time to give the objective function value of the under-
lying allocation problem only within a certain approximation ratio, but the mechanism
provides incentives for the truthful revelation of valuations.
Finally, Chapter 11 provides an introduction to matching markets, which do not allow
for monetary transfers. The assignment of students to courses at universities serves as
a widespread example. There are environments where both sides of the market have
preferences (e.g., students and course organizers) and others where only one side has
preferences (e.g., the students). Interestingly, there are mechanisms, for both types of
matching markets, where the truthful revelation of preferences is a dominant strategy
for one side of the market. For example, the well-known deferred acceptance algorithm
by Gale and Shapley provides dominant strategies for one side (e.g., the students) in a
two-sided matching market, and the outcomes are stable, i.e., in our course allocation
example there is no pair of students and course organizers who would want to switch.
There are many connections to the theory of multi-object auction markets discussed in
Chapter 8 and to approximation mechanisms as they are described in Chapter 10.
Chapter 12 provides an outlook on current trends and frontiers in market design.
1.3.4 Part IV Appendices: Mathematical Optimization

The appendices are intended as a reference to key results in linear and integer linear
programming that are relevant to multi-object market design in Parts II and III of the
book. Linear and integer linear programming is a vast field. I have assumed that readers
15:27:03
01
8 Introduction
have had an introduction to linear programming, and the appendices should serve only
as a reference or refresher. A textbook introduction to linear programming can be found
in Chvatal (1983) and to integer linear programming in Nemhauser and Wolsey (1988).
In the book I will often take a perspective including the reader and therefore will use
“we” instead of “I” throughout.
1.4 Acknowledgements
Some sections of the book overlap with parts of papers that have been co-authored
with students and colleagues, but I have combined them using a uniform notation
with examples geared towards a textbook-style introduction to the field. These papers
include Bichler (2010), Bichler et al. (2011a, 2013a, 2017b), Diebold et al. (2014),
Kroemer et al. (2014). In preparing this text, I have benefitted from support and advice
from many sources. I would like to express my gratitude to my current and former
students Franz Diebold, Salman Fadaei, Andor Goetzendorff, Zhen Hao, Dennis Kraft,
Richard Littmann, Stefan Mayer, Sören Merting, Per Paulsen, Ioannis Petrakis, Alexan-
der Pikovsky, Stefan Schneider, Thomas Setzer, Pasha Shabalin, Florian Stallmann,
Stefan Waldherr, Bernhard Waltl, Jürgen Wolf, and Georg Ziegler. I am thankful to
co-authors and colleagues such as Gedas Adomavicius, Larry Ausubel, Haris Aziz,
Oleg Baranov, Felix Brandt, Peter Cramton, Vitali Gretschko, Kemal Guler, Alok
Gupta, Karla Hoffman, Maarten Janssen, Jayant Kalagnanam, Wolf Ketter, Ramayya
Krishnan, Axel Ockenfels, Tuomas Sandholm, Achim Wambach, Christof Weinhardt,
and Richard Steinberg for joint work and inspiring discussions. Lauren Cowles, Esther
Miguéliz, and the team from Cambridge University Press provided outstanding support
for the publication of the book.
I am particularly grateful to Paul Milgrom for all his kindness, advice, and guidance
during my sabbatical at the Stanford Department of Economics. Special thanks also go
to Dirk Bergemann, Bob Day, and Jacob Goeree for joint work, visits, and the many
insights they offered. Last, and certainly not least, I thank my family for supporting me
through this time-consuming project. I dedicate this book to them.
15:27:03
01
Part I
Microeconomic Fundamentals
15:28:07
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316779873
15:28:07
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316779873
2 Game-Theoretical Basics
Part I (Microeconomic Fundamentals) of this book provides a succinct introduction

to methods and models from microeconomics. In this chapter we introduce and define
basic concepts and terminology from game theory will be used throughout the rest of the
book. Game theory is the mathematical study of interactions among independent self-
interested agents or players in a game. We limit ourselves to non-cooperative game the-
ory and refer the interested reader to more extensive treatments such as Osborne (2004)
or Shoham and Leyton-Brown (2009), whose notation we share. Non-cooperative game
theory focuses on situations where self-interested agents have conflicting goals. This
chapter will introduce different types of games and central solution concepts, i.e., meth-
ods to predict the outcomes of a game played by rational agents. In some parts of the
book we will also draw on cooperative game theory, which focuses on predicting which
coalitions will form among a group of players, the joint actions that groups take, and the
resulting collective payoffs. A cooperative game is a game with competition between
groups of players or coalitions. However, we introduce the respective concepts later
where needed, in order to keep the chapter concise.
2.1 Normal-Form Games
Let us start with a basic type of game description. In normal-form games, players’
actions are simultaneous. In other types of games, called extensive-form games, actions
take place sequentially.
Definition 2.1.1 (Normal-form games) A finite normal-form game with n players can
be described as a tuple (I, A, u).
r I is a finite set of n players indexed by i.
r A = A1 × · · · × An , where Ai is a finite set of actions available to player i. A vector
a = (a1 , . . . , an ) ∈ A is referred to as an action profile.
r u = (u1 , · · · , un ), where ui : A → R is a payoff or utility function for player i.
In definition 2.1.1, finite means that there is a finite set of players and each has a finite
set of strategies. Typically, a utility function maps the set of outcomes of a game to
a real-valued utility or payoff. Here, the actions possible for an agent also describe
the outcomes, which is why we use A in the description of a normal-form game. The
15:27:34
02
Table 2.1 Payoff matrix for the

Prisoner’s Dilemma game
C D
C 4, 4 1, 5
D 5, 1 2, 2
Prisoner’s Dilemma is a well-known example of a normal-form game and is useful to

illustrate basic concepts in game theory. There are multiple versions of the game; one
version is as follows.
Example 2.1.1 (The Prisoner’s Dilemma) The players of the game are two prisoners
suspected of a crime. Each prisoner can either confess (C) or deny (D) the crime. If they
both deny, they serve a two-year prison term. If one of them confesses, his prison term
will be one year and for the other player five years. If they both confess, they will be
in prison for four years each (see table 2.1). The first entry in each cell denotes the row
player’s time in prison and the second entry denotes the column player’s time in prison.
Regarding the row player, he is always better off by confessing, independently of what
the column player does, and vice versa. So, any rational prisoner will adopt the strategy
of confessing the crime regardless of the other player’s decision. However, if they both
denied, they would both be better off.
One type of strategy available to a player i in a normal-form game is to select a

single action ai and play it. Such a strategy is called a pure strategy. In the Prisoner’s
Dilemma both players might choose the action “confess”. But a player is also able to
randomize over the set of available actions according to some probability distribution.
Such a strategy is called a mixed strategy.
Definition 2.1.2 (Mixed strategy) In a normal-form game (I, A, u) the set of mixed
strategies for player i is Si = (Ai ), where (Ai ) is the set of all probability distributions
(aka lotteries) over Ai .
The probability that an action ai is played in strategy si is denoted by si (ai ). When

players choose to play a mixed-strategy profile (s1 , . . . , sn ), the expected utility of player
i can be described as ui (s) = a∈A ui (a)( nj=1 s j (a j )). Notice that the term in parenthe-

n
ses, j=1 s j (a j ), is the probability that action profile a = (a1 , . . . , an ) will be played.
A mixed-strategy profile is the Cartesian product of the individual mixed-strategy sets,
S1 × · · · × S n .
We analyze games using solution concepts or more specifically equilibrium concepts.
Solution concepts can be seen as a way to predict the outcome assuming some behavior
of rational agents. An economic equilibrium is a state where economic forces such as
supply and demand are balanced and the values of economic variables will not change.
Solution concepts are principles according to which we identify interesting subsets of
the outcomes of a game. The two most fundamental solution concepts are Pareto opti-
mality and Nash equilibrium.
15:27:34
02
2.1.1 Dominant-Strategy Equilibrium

In the Prisoner’s Dilemma game in table 2.1 a dominant-strategy equilibrium results if
both players confess. This means that each rational player would confess, independently
of what the other player does. Let us define this more formally. We say that the strategy
si strongly dominates the strategy s′i if player i always does better with si than s′i , i.e.,
∀s1 , . . . , si−1 , si+1 , . . . , sn ,
ui (s1 , . . . , si−1 , si , si+1 , . . . , sn ) > ui (s1 , . . . , si−1 , s′i , si+1 , . . . , sn )
The strategy si weakly dominates s′i if player i never does worse with si than with s′i , and
there is at least one case where player i does better with si than s′i :
∀s1 , . . . , si−1 , si+1 , . . . , sn , ui (. . . , si , . . .) ≥ ui (. . . , s′i , . . .)
and
∃s1 , . . . , si−1 , si+1 , . . . , sn , ui (. . . , si , . . .) > ui (. . . , s′i , . . .)
The strategy si is a strongly (weakly) dominant strategy if it strongly (weakly) dominates
every s′i ∈ Si .
Definition 2.1.3 (Dominated strategy) A strategy si is strongly (weakly) dominated for
an agent i if some other strategy s′i strongly (weakly) dominates si .
In a dominant-strategy equilibrium a player i will do better using si rather than a
different strategy s′i , regardless of what strategies the other players use.
Definition 2.1.4 (Dominant-strategy equilibrium) A dominant-strategy equilibrium is
a set (s1 , . . . , sn ) such that si is dominant for each player i.
Dominant strategies are very useful, as they are simple for players, but they do not
always exist. Interestingly, in the standard game-theoretical models of auctions, there is
an auction format where the bidders have a simple dominant strategy, as we will see in
chapter 4.
2.1.2 Pareto Optimality

Pareto optimality also plays an important role in the Prisoner’s Dilemma. A strategy
profile is Pareto optimal if there does not exist another strategy profile that Pareto-
dominates it. A strategy profile is Pareto-dominated if some player can be made better
off without any other player being made worse off.
Definition 2.1.5 (Pareto optimality) Any strategy profile that is not Pareto-dominated is
Pareto optimal. A strategy profile s Pareto-dominates strategy profile s′ if, for all i ∈ I,
ui (s) ≥ ui (s′ ) and there exists a j ∈ I for which u j (s) > u j (s′ ).
Thus, Pareto optimality is a solution concept from the point of view of an outside
observer of the game. It gives a partial ordering over strategy profiles, but we cannot
generally identify an optimal strategy. Weak Pareto dominance means that at least one
15:27:34
02

Matching Pennies game
Heads Tails
Heads 1, −1 −1, 1
Tails −1, 1 1, −1
agent is better off, and no one is made worse off. Strong Pareto dominance means that all
agents gain. The Pareto-optimal solution in the Prisoner’s Dilemma is for both players
to deny but the dominant-strategy equilibrium is for both to confess. This is where the
dilemma comes from.
2.1.3 Nash Equilibrium

The Nash equilibrium is arguably the most significant solution concept in game theory.
In a Nash equilibrium we look at games from an agent’s point of view, as we did in the
case of dominant-strategy equilibria.
Definition 2.1.6 (Nash equilibrium) A strategy profile s = (s1 , . . . , sn ) is a Nash equi-

librium if si is a best response to s−i for all players i. Here s−i is the profile obtained after
removing si from the profile s. The best response to s−i is a strategy chosen by player i
which maximizes his utility given that players other than i play s−i .
Players do not want to deviate unilaterally from a Nash equilibrium. If a predicted

outcome was not a Nash equilibrium, it would mean that at least one individual would
have an incentive to deviate from the predicted outcome and increase his utility.
Example 2.1.2 (The Matching Pennies game) The Matching Pennies game describes
a zero-sum game where both players simultaneously put a penny on the table. If one
player wins, the other player loses and the payoff matrix for a row and a column player
are described in table 2.2. The game has a mixed-strategy Nash equilibrium, but no
pure-strategy Nash equilibrium. In a Nash equilibrium nobody would want to change
her strategy if she knew what strategies the other agents were following, and therefore it
describes a form of stability. If the column player plays heads with probability p and tails
with probability 1 − p then the row player must be indifferent between the two actions
available to him and so will play a mixed strategy. If she is not indifferent between the
two actions, she would play purely the action which gives a better utility by setting a
zero probability for the other action. Thus, the only Nash equilibrium is randomization
with equal probability over both actions. This is easy to check. For the row player, these
probabilities can be derived as follows:
E(uheads ) = E(utails ) ⇒ 1p + (−1)(1 − p) = −1p + 1(1 − p) ⇒ p = 1/2
Although computing a mixed Nash equilibrium for small games as above is easy, the
problem of finding the Nash equilibrium for a game in general is a computationally hard
problem. It was shown to be PPAD-complete (Daskalakis et al., 2009), which means in
15:27:34
02

Battle of Sexes game
F B
F 2, 1 0, 0
B 0, 0 1, 2
the worst case that computing an equilibrium is exponential in the number of actions of
each player, assuming P = PPAD i.e., that the classes P and PPAD describe different sets
of problems. Let’s take a look at another well-known example of a normal-form game.
Example 2.1.3 (The Battle of Sexes game) The Battle of Sexes game has the bi-matrix
representation shown in table 2.3. This game has two players who want to go either to
a football game (F) or to a ballet (B). The states (F, F) and (B, B) constitute pure Nash
equilibria with rewards (2, 1) and (1, 2), respectively.
We can compute the probabilities for a mixed Nash equilibrium with which the col-
umn player needs to play F and B. These probabilities p and 1 − p for the column
player need to be set such that the expected utility for the row player is the same for
both actions:
2p + 0(1 − p) = 0p + 1 − p ⇒ p = 1/3
The probabilities q and 1 − q for the row player are obtained analogously:
q + 0(1 − q) = 0q + 2(1 − q) ⇒ q = 2/3
Consequently, the expected payoff for each bidder is 2/3 as a result.
Note that we considered only the expected payoffs in the pure equilibria in the Bat-
tle of Sexes game, i.e., the expected payoff in (F, F) or (B, B). The price-of-anarchy
(PoA) has been defined as the ratio between the “worst equilibrium” and the optimal
“centralized” solution. It can be defined with respect to pure Nash equilibria or mixed
Nash equilibria, for example. In the Battle of Sexes game, the optimal solution would be
(F, F) or (B, B) with a total utility of 3 and PoA 3/(4/3) = 9/4. The price-of-anarchy
measures the decrease in the efficiency of a system due to selfish behavior of the players.
The concept has received significant attention as it also allows us to analyze the robust-
ness of simple auction formats in the presence of complex utility functions. The PoA
can be seen as an analogue of the “approximation ratio” in an approximation algorithm
or the “competitive ratio” in an online algorithm.
A strict Nash equilibrium is one where every agent’s strategy constitutes a unique best
response to the other agents’ strategies. If, instead, for some player multiple strategies
have the same payoff then this constitutes a weak Nash equilibrium. The example in
table 2.4, with actions U, M, B, L, N, has a unique and pure best response (B, N),
because it has the highest utility for both players.
John Nash showed that every game with a finite number of players and action profiles
has at least one Nash equilibrium. However, typically there are more than one, which
often makes the Nash equilibrium a poor prediction device; if there are multiple Nash
15:27:34
02
Table 2.4 Normal-form game with a strict

pure-strategy Nash equilibrium in (B, R)
L M N
U 0, 4 4, 0 5, 3
M 4, 0 0, 4 6, 3
B 3, 5 3, 6 7, 7
equilibria in a game, the players face an equilibrium selection problem. If some players
believe that the equilibrium ultimately played out is s while others believe that it is s′
then the profile of strategies actually selected will be a mixture of s and s′ , a disequilib-
rium. In our Battle of Sexes example, without coordination bidders might easily end up
in a disequilibrium with zero payoff.
Unfortunately, there can be many Nash equilibria. McKelvey and McLennan (1997)
showed that, fixing the number of pure strategies for each player, the maximal number of
totally mixed Nash equilibria is exponential in the minimum number of pure strategies
for any player. A Nash equilibrium is totally mixed if every pure strategy is assigned a
positive probability.
There have been several suggestions about how players deal with the equilibrium
selection problem. Payoff dominance and risk dominance are two well-known refine-
ments of the Nash equilibrium solution concept defined by John Harsanyi and Reinhard
Selten. A Nash equilibrium is considered payoff dominant if it is Pareto superior to all
other Nash equilibria in the game. Risk dominance is best illustrated in the Stag Hunt
game in table 2.5. A joint action (hunting) in this game yields a higher payoff if all
players combine their skills. However, if it is unknown whether the other agent will help
in hunting, gathering is the better strategy for each agent because it does not depend on
coordinating with the other agent. The strategy pair (Hunt, Hunt) in the Stag Hunt game
is payoff dominant since the payoffs are higher for both players compared with the other
pure Nash equilibrium (Gather, Gather).
On the other hand, (Gather, Gather) risk-dominates (Hunt, Hunt) since if an uncer-
tainty exists about the other player’s action then gathering will provide a higher expected
payoff. In this symmetric game it is easy to see this. If each agent assigns a proba-
bility 0.5 to Hunt and to Gather then (Gather, Gather) risk-dominates (Hunt, Hunt):
0.5 × 4 + 0.5 × 2 ≥ 0.5 × 5 + 0.5 × 0. Alternatively, one can look at the product of
the deviation losses. If the column player deviates from (Gather, Gather) to Hunt then
the deviation loss is 2. The same is true for the column player if the row player deviates.
So, the product of the deviation losses is 2 × 2 = 4; this is more than the product of
the deviation losses from deviations of the (Hunt, Hunt) equilibrium, which is 1. This
means that, in this Stag Hunt game, payoff and risk dominance lead to different results.
Table 2.5 Payoff matrix for the Stag

Hunt game
Hunt Gather
Hunt 5, 5 0, 4
Gather 4, 0 2, 2
15:27:34
02

Chicken game
C D
C 6, 6 2, 7
D 7, 2 0, 0
2.1.4 Correlated Equilibrium

The correlated equilibrium is an interesting extension as it generalizes the Nash equi-
librium (Aumann, 1987). Let’s reconsider the Battle of Sexes game in table 2.3. Now,
consider a situation where a “trusted” authority flips a fair coin and, depending on the
outcome of the coin toss, tells the players what they should do. For example, if the coin
shows heads, the column player is told to choose ballet and the row player is told that
as well. No individual player has an incentive to deviate from what they are told to do.
In this case, when the row player is told to choose B he knows that the column player
is told to choose B as well. So, the row player has no incentive to deviate and switch
to F as the payoff would be lower (0 compared with 1). The advantage of following
such a procedure is that the expected rewards are now higher (3/2, 3/2) compared with
the rewards (2/3, 2/3) from the mixed Nash equilibrium. A more interesting example of
correlated equilibria is the game of Chicken in table 2.6.
Example 2.1.4 (The Chicken game) Two drivers drive towards each other on a collision
course: one must swerve or both may die in the crash, but if one driver swerves and the
other does not then the one who swerved will be called a chicken, meaning a coward
(C). The worst outcome occurs when both players dare (D). The deterministic or pure
Nash equilibria are (D, C) and (C, D) with a payoff of (7, 2) or (2, 7), respectively.
A mixed Nash equilibrium has an expected payoff of 14/3 for each player.
In a correlated equilibrium a trusted third party tells each player what to do, on the
basis of the outcome of an experiment with the following probability distribution: (C,
D), (D, C), and (C, C) all have a probability of 1/3. If the trusted party tells the column
player to dare, then she has no incentive to deviate. She knows that the outcome must
have been (C, D) and the row player will obey the instruction to chicken. If the column
player is told to chicken then she knows that the outcome must have been either (D, C)
or (C, C), each happening with equal probability. The column player’s expected payoff
on playing C, conditioned on the fact that she is told to chicken, is 0.5 × 6 + 0.5 × 2 =
4. If the column player decides to deviate and play D when told to play C then the
expected payoff is 0.5 × 7 + 0.5 × 0 = 3.5 < 4. So, the expected payoff on deviating
is lower than the payoff on obeying the instruction of the trusted party. The game is
symmetric and the row player has the same incentives. Note that, in the case of the
correlated equilibrium, the expected reward for each player is 1/3 × 7 + 2/3 × 4 = 5.
This is higher than the expected payoff of 14/3 in the mixed Nash equilibrium.
Definition 2.1.7 (Correlated equilibrium) A correlated equilibrium is a probability dis-

tribution {ps } on the space of strategy profiles that obeys the following conditions. For
each player i, and every two different strategies b, b′ of i, conditioned on the event that a
15:27:34
02
strategy profile with b as a strategy was drawn from the distribution, the expected utility
of playing b is no smaller than that of playing b′ :

(uisb − uisb′ )psb ≥ 0
s∈S−i
By S−i we denote the set of strategy profiles of all other players except for i. If s ∈
S−i , sb denotes the strategy profile in which player i plays b and the others play s. The
inequalities show that if a strategy profile is drawn from the distribution {ps } and each
player is told, privately, his or her own component of the outcome, and if furthermore
all players assume that the others will follow the recommendation then the expected
profit of player i cannot be increased by switching to a different strategy b′ , so that the
recommendation is self-enforcing.
The inequalities in the definition of a correlated equilibrium for the Chicken game
can now be formulated as follows:
(6 − 7)pC,C + (2 − 0)pC,D ≥ 0
(7 − 6)pD,C + (0 − 2)pD,D ≥ 0
(6 − 7)pC,C + (2 − 0)pC,D ≥ 0
(7 − 6)pD,C + (0 − 2)pD,D ≥ 0

pi, j = 1
i, j∈{C,D}
0 ≤ pi, j ≤ 1
The first two equations state the optimality of the distribution for the row player by
comparing the payoffs for these cases when the column player chooses strategy C or
strategy D. In the first equation the recommendation for the row player is C but this
player deviates to D when the opponent either plays C (first term) or D (second term).
The second equation deals with a row player who is recommended D, but might deviate
to C for the two cases where the column player plays C or D. The third and fourth
equations are for the column player, whose behavior is symmetric.
For every Nash equilibrium there exists a corresponding correlated equilibrium.
Actually, there is an infinite number of correlated equilibria: an intuitive subset is the
convex hull of the pure Nash equilibria (D, C) and (C, D) in our example above. There
are also correlated equilibria which are not Nash equilibria. For example, if pD,C and
pC,D are selected with a probability of 0.5, then this is not a Nash equilibrium but a
correlated equilibrium. This means that Nash equilibria are a subset of all correlated
equilibria.
The mediation value is the ratio of the maximal welfare of a game obtained in a cor-
related equilibrium and the maximal welfare obtained in a mixed-strategy Nash equilib-
rium (Ashlagi et al., 2008). In the above example of the Chicken game with two players
the mediation value 10/9 can be computed as the maximum welfare of a correlated
equilibrium (30/3) divided by the maximum welfare of a Nash equilibrium, which is 9.
The price-of-anarchy for the example is 12/9 = 4/3.
15:27:34
02
Table 2.7 Normal-form game leading

to (C, E) with iterated dominance
D E F
A 0, 2 2, 1 1, 3
B 1, 4 2, 1 4, 1
C 2, 1 4, 4 3, 2
2.1.5 Further Solution Concepts

For the static solution concepts that we have considered so far, we can define a rela-
tionship among the solution concepts, such that dominant-strategy equilibria ⊂ pure
Nash equilibria ⊂ mixed Nash equilibria ⊂ correlated equilibria. There are, however,
many other solution concepts for normal-form games. In the following, we provide a
short description of two widely used concepts and discuss their relationship to the Nash
equilibrium.
2.1.5.1 Iterated Elimination of Dominated Strategies

The iterated elimination (or deletion) of dominated strategies is one common technique
for solving games that involves iteratively removing dominated strategies. If all players
have a dominant strategy then it is natural for them to choose the dominant strategies
and so we reach a dominant-strategy equilibrium. Of course, if a player has a domi-
nant strategy then all other strategies are dominated. But there may be cases where a
player does not have a dominant strategy and yet has dominated strategies (see section
2.1.1 for a discussion of domination). The iterated elimination of dominated strategies
is a common technique for solving games that involves iteratively removing dominated
strategies. This process is valid if it is assumed that rationality among players is com-
mon knowledge, that is, each player knows that the rest of the players are rational, and
each player knows that the rest of the players know that he knows that the rest of the
players are rational.
There are two versions of this process. One version involves eliminating only strongly
dominated strategies. If, after this procedure, there is only one strategy for each player
remaining, that strategy profile is the unique Nash equilibrium. Another version involves
eliminating both strongly and weakly dominated strategies. Every equilibrium in dom-
inant strategies is also a Nash equilibrium. However, the Nash equilibrium found by
eliminating weakly dominated strategies may not be the only Nash equilibrium.
The order of elimination doesn’t matter when we remove strictly dominated strate-
gies. However, the order of elimination does matter when we remove weakly dominated
strategies; we might actually discard one or more Nash equilibria of the game. If by iter-
ated elimination of dominated strategies there is only one strategy left for each player,
the game is called a dominance-solvable game.
Example 2.1.5 Table 2.7 shows an example, where we can first eliminate row A of the
row player, because it is strongly dominated by row C. For the remaining rows, we can
eliminate column F because it is dominated by the mixed strategy 0.5D + 0.5E. Now
we have a payoff matrix with two columns and two rows. Row B is dominated by row
15:27:34
02
C in this reduced matrix. Finally, in the remaining row, column D is strongly dominated
by E. The strategy profile with the column player playing E and the row player playing
C is a Nash equilibrium. It is also the unique Nash equilibrium of the game.
2.1.5.2 ε-Nash Equilibrium

In a Nash equilibrium, no player has an incentive to change his behavior. In an approx-
imate or ε-Nash equilibrium, this requirement is weakened to allow the possibility that
a player may have a small incentive to do something different. The idea is that players
are indifferent to sufficiently small gains, which seems plausible in many situations.
Definition 2.1.8 (ε-Nash equilibrium) A strategy profile s = (s1 , . . . , sn ) is called an

ε-Nash equilibrium if ui (si , s−i ) ≥ ui (s′i , s−i ) − ε for all players i and for all strategies
s′i = si .
The concept has a computational advantage compared with Nash equilibrium, as

polynomial-time algorithms are known for it (Daskalakis et al., 2009). Every Nash equi-
librium is surrounded by ε-Nash equilibria with ε > 0. Unfortunately, a player’s payoff
in ε-Nash can be arbitrarily less than in a Nash equilibrium of the game. This means
that an ε-Nash cannot be used to approximate agents’ payoffs in a Nash equilibrium.
2.2 Extensive-Form Games
The normal-form game does not incorporate any notion of sequence. In game the-
ory, a sequential or dynamic game is a game where one player chooses his or her
action before the others choose theirs. The extensive-form representation makes tem-
poral structure in the game explicit. We only mention extensive-form games briefly, for
the sake of completeness, because we will rarely draw on such models in subsequent
sections. This might appear surprising, because many auction formats are organized
in sequential rounds. In the next chapter, on mechanism design, we will discuss the
revelation principle, given in Theorem 3.3.1, which provides theoretical reasons why
auctions are typically modeled as games in which bids are submitted simultaneously.
We do use extensive-form games, however, to illustrate strategic problems as they arise
in the simultaneous multi-round auction in Chapter 6.
A perfect-information extensive-form game can be modeled as a tree, where each
node describes the choice of an agent between possible actions represented by edges.
The leaves represent the outcomes for which each player has a utility. In each stage of
the game, the players know the node they are in. Every perfect-information game can be
converted to an equivalent normal-form game. However, the transformation to a normal-
form game comes at the expense of exponentially many outcomes in the resulting game.
The reverse transformation, of normal-form games to perfect-information extensive-
form games, is not always possible. Every perfect-information game in extensive form
has a pure-strategy Nash equilibrium in normal-form representation. Yet there might
exist some unsatisfying Nash equilibria which are based on “non-credible threats”.
15:27:34
02
Another random document with
no related content on Scribd:
has mixed up some tradition of this ancient work in Nubia. At any
rate, whatever be the truth about Lake Mœris, his account proves
beyond all question that the idea of a Reservoir was familiar to the
ancient Egyptians.
The tradition of a Reservoir somewhere on the upper waters of
the Nile lingered long in Egypt. There is a curious reference to it in a
book of travel by F. Vansleb, a Dutchman who visited Egypt about
the year 1670. The fertility of the Nile flood is caused, he says, by a
fall of dew, which usually takes place on June 17, just after the
appearance of the ‘green’ water. This dew purifies the foul water, and
makes it swell by fermentation. ‘Some of the country,’ however, he
proceeds, ‘that are ignorant of the true causes of this increase,
imagine that it proceeds from a large pond in Ethiopia in the river
itself, which the Abyssins begin to open about June 12, and let the
water out by degrees, more and more till September 14, by which
time they begin to shut it again. But this is a foolish fancy of the
Copties.’
We have seen how the tradition of Lake Mœris fascinated
Mehemet Ali; but the methods of Haroun-al-Raschid were not suited
to solid engineering works, as the history of the Barrage too plainly
shows. None of his descendants, with the exception of the Khedive
Ismail, had the wit to conceive or the ability to execute such an
undertaking, and Ismail’s fantastic imagination was fully occupied in
other directions. Fortunately for Egypt, the project had to wait until
the success of the Barrage made the time ripe for its execution, and
until skilful brains and strong hands were ready to plan and carry it
out in the most efficient manner possible.
There were three problems to be faced: first, Where could such a
Reservoir be erected? second, What arrangements could be devised
to avoid the danger of large silt deposits, which would soon seriously
diminish the capacity of the Reservoir, and, if allowed to accumulate,
render it in no long time entirely useless? third, Supposing that the
difficulties of site and design could be overcome, where was the
money to be found? During the first years of the British occupation,
while Egypt was still painfully struggling upwards from the abyss of
bankruptcy into which she had been cast by the mad whirlwind of
extravagance in Ismail’s reign, it was no time for the inception of
original works on a grand scale. But in 1890 the matter became an
affair of practical politics, and was at last seriously taken in hand.
Meantime discussion had been raging as to the best locality for the
Reservoir. An American gentleman, Mr. Cope Whitehouse, took up
the case of the Wadi Rayan, a depression in the desert to the south-
west of the Fayoum. This, he maintained, was the real site of the
ancient Lake Mœris, and here the Reservoir ought to be. He had no
professional knowledge, and he was utterly wrong in his ideas; but
his vehement method of controversy kept the subject thoroughly
alive. The whole land was filled with his clamour, and every expert
was forced to give his own views in self-defence. The debate served
a useful purpose. Gradually it came to be recognised that the river-
bed itself was the proper place for storing the water by means of a
Dam. Authorities differ as to whom belongs the credit of first making
this suggestion, or, rather, of first reviving the tradition of the past;
but it seems pretty clear that Sir Samuel Baker suggested the
construction of a Dam at the first cataract at Assouan as far back as
1867.
However that may be, in 1890 the Government took the matter
up, and charged Mr. W. Willcocks with the task of examining the river
north of Wadi Halfa, reporting upon the best available site for the
Dam, and preparing a design for it. After a careful survey, his plans
were completed in 1894, and his design for a Reservoir at Assouan
was then submitted to an International Committee of Engineers,
consisting of Sir Benjamin Baker, M. Boulé, and Signor Torricelli. Mr.
Willcocks’ plans were, with some modifications, accepted by a
majority of the Commission, and to him belongs the honour of having
designed the Dam. The selection of the Assouan site solved the first
of the three difficulties. There is at this point an extensive outcrop of
granite clean across the valley of the Nile, which it was thought
would give sound rock everywhere at a very convenient level for the
foundations of the Dam. Moreover, the trough of the river above the
cataract and a long way south of it is exceptionally deep, and this
makes it possible for a greater amount of water to be stored up
behind the Dam. But the prime necessity was for a solid foundation.
Elsewhere in Egypt the bed of the Nile is composed of shifting
sands, on which it would have been impossible to build a Dam
capable of holding up so great a head of water.
Mr. Willcocks’ design solved the second difficulty, the problem of
constructing a Dam strong enough for the purpose, and yet of
avoiding the danger of filling up the Reservoir by too great
accumulations of silt. The other great Dams in the world, as, for
instance, that built by Sir Arthur Cotton on the Godavery River, in
India, are solid throughout. They are planned so that the rising flood
shall pass freely over the top of them. But the Assouan Dam is of a
type previously unknown, and its success ought to stimulate
perennial irrigation in many parts of the world where such projects
have hitherto proved failures. Its principle is that even the highest
flood shall pass, not over it, but through it. To this end it is pierced
with 180 openings, which are like tunnels in the great mass of
masonry. The openings are controlled by powerful sliding-gates
worked from above. During the months of the flood every gate will be
up, and the ‘red’ water, carrying all its heavy burden of silt, will pass
through without impediment. Later in the year, about the end of
November, when the flood has subsided and very much less matter
is carried in suspension, the sluice-gates begin to be gradually
closed, and by the end of February the Reservoir is full, without
having affected the normal discharge of the river in any appreciable
degree. From April to July the water thus stored up is let out by
degrees for employment, according to the state of the river and the
requirement of the crops. By the time the next flood begins to come
down all the stored water will have passed out, and every sluice will
be once more open to give free passage to the rising stream.
Although the Nile in December and January carries an insignificant
amount of sediment compared to that brought down in August and
September, it yet brings down a very considerable quantity, far
greater than most other rivers at any time, and quite enough to go a
long way towards silting up the bed of the Reservoir, if it was allowed
to remain. But for this the river provides its own remedy: every year
the force of the flood will act like a gigantic broom, sweeping the floor
of the Reservoir. The sluices, arranged in sets of five, are distributed
at different levels, according to the formation of the river-bed on the
upstream side, so as to facilitate this process to the utmost. During
the months of the inundation the Nile at Assouan pours down for
weeks together a volume of 10,000 tons of water per second, and
sometimes as much as 14,000 or 15,000 tons per second. The rush
of this stupendous mass is sufficient to assure us that there will be
no silting up of the Reservoir.
Two of the difficulties had been thus overcome, when, from a new
and unexpected quarter, a storm sprang up, which very nearly
brought to a standstill the rising fabric of Egyptian prosperity. The
project of the Reservoir would have raised the level of the water, and
held up the river above the Dam to a head of 100 feet; this would
have involved the temporary submersion every year of the island of
Philæ, with its famous Temple of Isis, Pharaoh’s Bed, and other
monuments. A terrific hubbub arose. Archæological and antiquarian
societies, which until then had sometimes belittled the monuments of
Philæ as belonging to an inferior period, poured in their protests.
People who had never heard of Philæ before, but who were none the
less influential for that, joined in the outcry. Diplomatists, whose one
desire was to embarrass our progress in Egypt, took up the cause of
Art with a will. These champions of humanity at large forgot the poor
fellaheen, to whom the extra water means all the difference between
misery and happiness; nothing would satisfy them but the complete
abandonment of the project. The engineers fought stoutly in the
interests of Egypt; they offered to raise the whole of the monuments
bodily, or to transport them to the neighbouring island of Bigeh; but,
though they saved the Dam, the original design was lost, and the
Dam to-day is 33 feet lower than it ought to have been. The
foundations of Philæ have been underpinned and strengthened, the
island will only be partially submerged, and the injury to Egypt can
only be faintly estimated.
Was the sacrifice worth it? The value of Philæ lies in its beauty
more than in its antiquarian interest. No one who has witnessed
night after night the glorious sunsets on the Nile, the mysterious
charm of the changing waters, the dark belt of palms reflected in the
river below and standing out in strong contrast against the sky, the
limestone cliffs of the desert clear-cut in the dry air, and flushing pink
in the radiance of the indescribable after-glow, no one who has seen
the Temples of Karnak, could hesitate to make so small a sacrifice,
in comparison, for the sake of the river-side people. Moreover, Egypt
is rich in treasures of the past, as yet undiscovered, and wanting
only money for their development, which the Reservoir would in time
supply. And how few people visit Philæ at all! Surely, even in a
country a thousand times poorer than Egypt in artistic and
archæological interests, the well-being of the living and of the unborn
should have prevailed.
If all those who joined to swell the uproar had been really
disinterested lovers of the beautiful, there would have been small
reason to complain of their insistence. Enthusiasts can hardly be
expected to listen to the voice of reason, and Philæ has charms to
soften the heart of the most savage utilitarian. Its fate is a mournful
necessity, but it is a necessity, for the question of the Reservoir had
come to be a question of existence for Egypt. Even the advantage
gained by the opposition in lowering the height of the Dam is only a
delay. On the pylon of the Temple of Isis at Philæ is carved a huge
representation of the Pharaoh of the day, one of the most
degenerate of the Ptolemies, catching his defeated enemies by the
hair of their heads with one hand, an uplifted sword in the other. The
whole is a copy of the work of his warlike ancestors, and even as a
copy it is a delusion and a sham; for he won no victories, defeated
no enemies, and, indeed, scarcely ventured outside the walls of his
harem. The apparent victory of these lovers of Philæ, to call them by
their more honourable title, was not less delusive. Philæ is doomed.
Between half drowned and wholly drowned there is not much
difference in the case of an island, certainly not a difference worth
fighting for, and the Dam will be raised to its full height, perhaps as
soon as Egypt is ready for the extra water.
The financial difficulty remained. In spite of the prosperity of
Egypt, she is, as everyone acquainted with her history is aware,
bound hand and foot by international fetters in matters financial. The
Caisse de la Dette, founded to protect Egyptian creditors against the
dangers of bad administration, has remained to be an obstacle to
any improvements that must benefit these interests. The practical
outcome of the system is that, if the Caisse be hostile—and hostile it
has often been—no public work like the Nile Reservoir can be
carried out without the imposition of extra taxation to the amount of
double the annual expenditure required.
Time passed, the need became more pressing, but the prospects
of the Reservoir seemed further off than ever. Besides the regulation
of her water-supply, Egypt had on her hands the question of the
Soudan. From every point of view the reconquest of that province
and the upper waters of the Nile was a prime necessity; no one
could tell how long the war might last, or how great the expense
might be. It seemed impossible that she could bear the cost of two
such enterprises simultaneously, and under such circumstances her
credit would not have been sufficient to raise the capital sum
required on anything like reasonable terms. Not only so, but by the
peculiar constitution of Egyptian finance it was illegal for her to raise
a loan without the consent of the Caisse, a consent which it was
impossible to obtain.
But, fortunately for Egypt, there were a few men with clearer
vision and more faith in the future, and chief among these was Lord
Cromer. The statesman who had controlled the tangled destinies of
Egypt through so many dark years, and baffled so many tortuous
intriguers, as well as more open foes, was not the man to despair in
such a situation. In 1897 the first negotiations were quietly opened
with Sir E. Cassel. Then came the vote of the majority of the Caisse
to grant £500,000 towards the Soudan Railway, and the successful
action taken in the Courts against that vote. Everyone knows how
this seeming defeat was turned to overwhelming victory by the
decision of the English Government to grant £750,000 for the railway
on certain conditions. The enemies of England in Egypt received a
staggering blow.
But the story was not yet complete. In April, 1898, Sir E. Cassel
arrived at Cairo; in one day the details of the arrangements to
finance the Dam were settled; all that night the lawyers drafted the
necessary documents; a Council of Ministers was hastily called in
the morning, and the contracts were signed. Sir E. Cassel was to
provide the necessary funds for the execution of the work,
£2,000,000; repayment by the Egyptian Government was to be
deferred altogether for five years, and then to be spread over a
period of thirty years. The first payment of about £78,000 is included
in the Budget for 1903.
Looking back now after five years of prosperity, when Egyptian
securities have actually increased in value, while Consols
themselves have so greatly declined, it is easy to see that the
statesman and the financier were justified in their faith. But in those
days it needed a clear vision and a stout heart to calculate thirty
years ahead—nay, even five—in a country so much the sport of
international politics. The Soudan Campaign was not yet ended;
behind the dervishes there loomed vague possibilities of worse
complications. The Egyptian Government made a good bargain then,
though it would doubtless make a better now. But it was then, and
not now, that the business had to be settled.
CHAPTER VIII
THE DAM AND THE NEW BARRAGES
Once the financial difficulty was settled, no time was lost in setting to
work. As soon as the flood of 1898 began to subside, Messrs. Aird
and Co., the contractors, were busy with the foundations of the Dam.
Five years was the period allowed by the contract, but a succession
of low Niles gave unusual facilities for the work, and everything was
completed before the flood of 1902, a year before the specified time.
From its vast proportions, the Dam is infinitely more impressive to
the imagination than any other of the irrigation works in Egypt. But
from an engineering point of view its construction was a plain,
straightforward business compared with the difficulties of building a
Barrage, where the river-bed offered no more solid foundation than
shifting sands. Still, there was a moment, on the first uncovering of
the river-bed, when its fate seemed to hang in the balance. The
Assouan site had been selected principally because the outcrop of
granite, there running clean across the valley, would give, it was
thought, solid foundation at a convenient level. It was found that in
some places the rock was rotten to a depth of 40 feet. It was an
anxious moment, both from an engineering and a financial point of
view. Every foot of rotten rock meant a considerable addition to the
calculated expense, besides modifying the building plan. Once more
Lord Cromer’s strong will saved the situation. On the financial side
he stood on firm ground, and he proved as good an engineer as he
had been a financier. Solid rock was reached, and the work went
steadily forward. Ten thousand men was the usual sum of those
employed, and of these 800 were Italian stone-cutters specially
brought over to deal with the tough granite of which the Dam is built.
Granite and Portland cement are the two great materials used for
welding the fetters of the Nile.
The Dam is about one mile and a quarter in length, and at its
deepest point it is 126 feet high. Sixty-five feet of water can be held
up when the reservoir is full, and it is capable of storing about
1,200,000,000 cubic metres of water—that is to say, about the same
amount of water as passes through Assouan in a single day when
the flood is at its height. The face of the wall is a slope on the
downstream side, and its width at the bottom corresponds
approximately to its height. Seven hundred and eighty thousand
cubic yards of masonry have been used. On the western side a
ladder of four locks gives passage to boat and steamer traffic at all
seasons.
All parts of Egypt are to benefit in a greater or less degree from
the extra summer supply. The original calculation assumed an
amount to be distributed of 1,065,000,000 cubic metres. It was
allotted as follows:
Cubic Metres.
South of Assiout 170,000,000
Assiout to Cairo with the Fayoum 510,000,000
Gizeh Province 85,000,000
Lower Egypt 300,000,000
This division meant that 52,000 acres could be reclaimed to

cultivation in the Fayoum, and 120,000 acres in the Delta. Further,
south of Assiout 200,000 acres could be converted to perennial
irrigation by means of pumps upon the Nile banks. In Middle Egypt
458,000 acres could be converted to perennial irrigation, and in
Gizeh Province 106,000.
A low Nile on the average comes about once in five years, but
assuming that it happened every year, these results may be
expressed in terms of money on the basis that conversion to
perennial irrigation increases the yield per acre by £2 annually, and
that reclaimed land produces a yield valued at £5 annually. This
gives an increase of value in the annual yield:
£E
South of Assiout 420,000
Middle Egypt and Fayoum 1,176,000
Gizeh Province 212,000
Lower Egypt 600,000
Total £E2,408,000
while the direct annual gain to the State Exchequer in rental and
taxation would amount to £E378,400.
This does not exhaust the full extent of the benefits of the
Reservoir on which a money value can be placed. For some time
Egypt had been living beyond her real resources in the matter of
water. Encouraged by a series of good years, the acreage of cotton
had been greatly extended. The cotton-plant is very hardy, and can
retain its vitality for a certain period on a very scanty and irregular
supply of water. Cultivators, especially in the Delta, had taken
advantage of this quality up to the very hilt, and the annual crop had
reached an amount of 6,000,000 kantars.
Taking the not very high price of 175 piastres per kantar (1 kantar
= nearly 100 pounds), the annual value reached £10,000,000. A
season like that of 1889, when the summer supply was very low, and
not the lowest on record, would mean the loss of at least one-tenth
of this amount, and even more, in spite of the most successful
working of the Barrage, and the most careful system of rotations.
Against such a loss the Reservoir is a complete insurance, and, as a
low year cannot safely be reckoned as occurring less than once in
five years, the annual value of such insurance must be set down as
£200,000. The figures give some idea of the value of the new supply.
The estimate does not err on the side of exaggeration. It was framed
in the most cautious and conservative manner possible, and, in fact,
it would be by no means rash to put the total annual value a good
deal higher.
According to the financial arrangement, the first payment towards
defraying the cost of the works was included in the budget for 1903,
but no fresh taxation for the purpose was to be imposed till 1904,
and even then the full amount of direct benefit to the Exchequer will
not be realized till 1910. Time is thus given for the full effects of the
change to be felt, and the Reservoir will be paid for out of its own
profits. The alteration of the basin lands to the new system must take
some time, and their cultivators will thus be given full opportunity to
familiarize themselves with the new methods of agriculture which
they will have to employ.
The scheme of distribution allotted nearly 50 per cent. of the
Reservoir supply to the Fayoum and the province between Assiout
and Cairo. Just south of Assiout is the head of the great Ibrahimiyah
Canal, which not only supplies these provinces, but also feeds the
Bahr Yusuf, which waters the Fayoum. It was necessary, in view of
the increased discharge, to widen the upper reaches of this canal,
and to provide it with a new regulating head. But more than this was
required to insure its receiving the proper proportion of water
whenever the sluice-gates of the Dam were opened. Accordingly, at
the same time that the foundations of the Dam at Assouan were laid,
a new Barrage was begun at Assiout just downstream of the head of
the Ibrahimiyah Canal.
In principle the Assiout Barrage is exactly the same as that at the
point of the Delta, and the difficulties of construction were also
exactly similar, for in both cases the foundations had to be laid on
the same shifting sands, and as each section of the work was
undertaken a portion of the river had to be diverted from its course
by means of temporary earthen dams. The shifting nature of the
river-bed, the almost personal malignity of the water, constantly
bursting through in countless springs, each of which had to be
separately dealt with and suppressed, called forth the highest
exercise of engineering skill. But the experience gained in the long
struggle with the imperfections of the earlier Barrage infallibly told its
tale. Every difficulty was successfully encountered, and the Assiout
Barrage was completed by the summer of 1902, and was able to
hold up the 10 feet of water required of it, without any failure, at the
first attempt.
The visible part of this Barrage, which is just over half a mile in
length, consists of a viaduct or bridge with 111 archways, each 16
feet 5 inches in width, closed by strong iron gates, working in
grooves made in the supporting piers, and raised or lowered from
above. In contrast to the Delta Barrage, it is built throughout of stone,
and not of brick. At the western end is a lock, the largest in Egypt,
through which the largest of the boats that ply upon the Nile can
easily pass. Below the water lies the strength of the structure. The
viaduct rests upon a solid platform of granite and cement, 10 feet
deep and 87 feet wide, set at a suitable depth below the bed of the
river. As a further precaution against the action of the water, there
are also below the platform two continuous lines of iron sheet-piles,
with hermetically sealed joints. With such a series of obstacles to
encounter, the danger of the water forcing its way through
underneath the Barrage is small indeed. The whole amount of
masonry used in this Barrage is 220,000 cubic yards.
The prosecution of these great works in Upper Egypt by no
means exhausted the activity of the Irrigation Department; indeed, it
would almost seem that the building of Barrages has become part of
its ordinary routine, for a third remains to be chronicled. This is the
Zifta Barrage on the Damietta branch of the Nile, halfway between
the point of the Delta and the sea. Because it lies in a district
unvisited by tourists, though very important commercially, its
construction has not been heralded by any blowing of trumpets. Yet
it is a work of the very first class, and deserves to be reckoned
among the greatest of the triumphs of the department. Built on the
same plan as the Assiout Barrage, and, like it, capable of holding up
10 feet of water, it is designed to secure a better distribution of the
supply north of the Delta Barrage. As the area of the cultivated land
extended gradually northward, it became apparent that the canal
system taking off from above the original Barrage was becoming too
long to admit of the water in times of pressure reaching the
northernmost parts of the country. There were some who held that a
second Barrage, with a new system of canals taking off from it,
should have been erected on the Damietta branch, even in
preference to the new stone weirs, which have increased the
strength and distributing power of the old one. The dispute has been
happily settled by the adoption of both projects, and with the Zifta
Barrage completed in time for the summer of 1903, and the new
supply from Assouan, the reclamation of the northern lands will go
steadily forward.
It has been already pointed out that it is almost as important to get
the water off the land as to get it on, and the proper drainage of the
Delta lands has been the necessary complement of all the new
schemes. Though eclipsed by the splendour of the Reservoir and
other creations more taking to the eye, the performance of the
engineers in this direction during the last few years has been
sufficient to make them very memorable in irrigation annals. Since
1896 about 1,000 kilometres of new drains have been dug, and
nearly as great a length of existing drains remodelled. It has cost the
Egyptian Government close upon a million of money, but that this
expenditure has not been thrown away is proved by the great rise in
the value of all the lands affected.
The Zifta Barrage cost about £500,000, the Assouan Dam and the
Assiout Barrage £3,200,000. Apart from the ordinary expenditure on
maintenance and the suppression of the corvée, the twenty years
ended 1902 have seen an expenditure of £9,000,000 devoted to
irrigation and drainage. There could be no greater proof of the
wisdom of those who have directed Egyptian policy during that
period. However pressed they have been at different times by
demands for immediate expenditure on other objects when
resources were low, they have always adhered steadily to a policy of
liberality towards public works likely to be of a remunerative
character. In no other country do economic laws work out their
results with greater directness and certainty. Reproductive
expenditure is really worthy of its name, and brings its visible and
tangible fruit almost without a moment’s delay. As they have sown,
so have they reaped. There can be no comparison between this
expenditure and the benefits it has conferred upon Egypt. Three
salient examples may be given to point the force of these remarks.
First, the works undertaken by Colonel Ross to improve the
system of basin irrigation in Upper Egypt. 1877 was a year of very
low flood, and nearly 1,000,000 acres were sharaki—that is, entirely
exempted from taxation owing to absence of irrigation. In 1899, a
worse year—in fact, the lowest flood of the century—the sharaki
lands were no more than 250,000 acres, a result mainly due to the
successful reforms carried out by Colonel Ross in 1888-89. 1902
was a year very similar to 1899, and the sharaki acreage was no
more than 140,000 acres.
Second, the case of the Assiout Barrage. The work was finished a
year before the time named in the contract. On August 15, 1902, the
usual date of filling the basins, the flood was exceptionally low, and it
was decided to lower the gates of the Barrage. By so doing the
water-level of the canals was raised by 1½ metres, an increase
which was more than sufficient to avert the threatened disaster. The
money value of the crops thus secured to the land-owners of the
Fayoum and Middle Egypt is estimated at not less than £600,000.
The cost of the new works at Assiout, including the new regulator on
the Ibrahimiyah Canal, was about £875,000. Thus, in the first year of
their existence they nearly repaid their whole cost, and that, too, as it
were, by a sidewind; for the Barrage’s real function is to hold up the
river in the low summer season, and not in the flood.
Third, the case of the Assouan Reservoir itself. The summer
supply in 1903 has been the lowest on record; the discharge at
Assouan has fallen to 200 cubic metres per second. By means of the
Reservoir this supply has been actually doubled. Had the flood been
late in coming down, it would have been impossible to distribute the
water at so liberal a rate. The Soudan gauges gave warning that an
early flood was to be expected, and thus the authorities were able to
calculate with certainty, and open the sluice-gates with much greater
freedom. But the calculation would have been a useless exercise
unless there had been a store to draw upon. At the lowest
computation the loss avoided may be reckoned at a couple of
millions.
The administrators of Egypt have had many difficulties and
obstacles in their path, but to be able to point to such results is a
great compensation; their efforts need no formal monument.
CHAPTER IX
THE INAUGURATION OF THE RESERVOIR
December 10, 1902, was the official date of the inauguration of the
Reservoir, a memorable day in the history of Egypt, and worthy to be
marked with red even in the unchanging Mohammedan calendar.
The making of the Dam has been a great time for Assouan. The
town has thriven and prospered beyond all knowledge since the
days—not so very long ago—when two British battalions occupied
the barracks on the hill overlooking the river to the south. The
barracks are crumbling to pieces now, and only one or two
blockhouses remain as memorials of the past state of siege and fear
of dervish raids. The tide of war has rolled far away and spent itself
utterly in remote corners. The whole of the Soudan lies between
Assouan and the frontier of any possible enemy. Even the yellow fort
is untenanted save by a few policemen. True, for four years an army
of 10,000 men has been marshalled here, but it was an invasion of
the arts of peace, creative and not destructive. Possibly the
inhabitants would have liked their occupation to go on for ever; but
even the best of times must have an end, and it was a good
occasion for a holiday. In every Nile village flags and bunting were
flying. For once the fields were deserted, and everywhere the people
crowded to the bank in the hope of catching a glimpse of the
Khedive and his distinguished guests. Here and there the gaffirs, or
local policemen, lined the shore, standing stiffly to attention, or
saluting with their Remingtons in the regulation attitude of European
soldiers, which contrasted quaintly with their loose, flowing robes
and white turbans.
At Assouan the faithful subjects of the Khedive surpassed
themselves. Sunny Assouan lends itself readily to a festal garb.
Situated where the Nile broadens out after emerging from the rocky
defile of the cataract, the town has a most picturesque aspect at all
times; its embanked river-front makes it the neatest of all the cities of
Upper Egypt. Elephantine and the Sirdar’s Island rise green and
smiling out of the broad bosom of the river; the perpetual blue sky
makes everything doubly attractive to the Northern visitor. Dressed
for the festival, it was a charming scene. Triumphal arches of the
sacred yellow and brown, gorgeous hangings and many-coloured
festoons, bore testimony to the Oriental love of vivid hues; steamers
and dahabiehs, moored in line along the shore as well as along
Elephantine Island, vied with each other in their decorations, and
numerous feluccas were plying to and fro, half hidden by their
burdens of flags and palm-leaves. At night thousands of lamps
decorated shore and river alike, and the whole scene resembled
nothing so much as a Henley in Regatta Week, with its illuminations
unrestrained by doubts of weather. To the ear, however, the voices of
the night told a very different tale. The crooning song of the Nubian
boatmen, ‘Great is the Prophet, praise be to him!’ accompanying the
creaking of their clumsy oars with monotonous persistency, sounded
weird and barbaric over the twinkling waters. The bustle in the town,
too, was no mere ordinary festal murmur; for this was the month of
Ramadan, and the feast of lamps meant a great deal to all faithful
Moslems. All day long they have abstained from food or drink, and
the going down of the sun is keenly welcomed as the end of one day
more of fasting.
The secret of the prosperity of Assouan lies in its granite. It is the
granite bed of the river at this point that makes the Reservoir
possible; here are the granite quarries from which the Dam was built,
and from which every ruler of Egypt who wished to raise a
monument for all time has drawn his supplies. Nothing that I have
seen in this country brings the past so near as these quarries. Here
lies a rough-hewn obelisk, just ready to be rolled away; here an
enormous block of stone half hollowed into a bath for an Emperor, or
a sarcophagus for an Apis bull, designed by some mighty ruler who
‘thought in continents,’ and recked little of the lives and labours of
thousands provided he gratified his whim. But suddenly death or
some other fate intervened, and a feebler or more merciful
generation has never taken up the work. You may see the marks of
the wedges on some great face of rock, as fresh as if it was only
yesterday that Pharaoh’s workmen had driven them in and poured
water on the wood till it swelled and burst the stone; down below is
the fallen piece still waiting for the mason who never came back to it.
Perhaps some of the very stones cut by Cheops or Rameses have
been smoothed and planed and set in the Great Dam.
Passing along the raised causeway, down which so many great
monuments have been rolled slowly to the river, I came through a
long stretch of burning desert to a spur on the northern extremity of
the granite hills. Here, unexpectedly, I found myself overlooking the
lake formed by the filling of the Reservoir. Graceful lines of palm-
trees showed where the banks of the river had once been. Philæ
was but an insignificant speck on the blue waters, overpowered by
the fantastic piles of granite boulders that hem in the valley. In the far
distance rose some lofty hills, crowned by dazzling sand that might
easily have been mistaken for snow. The Dam builders have been
accused of vandalism, but they have created a standing pool in the
wilderness of surpassing beauty. The view of the lake was not the
only attraction of the spot; at my feet lay a colossal statue of Osiris,
destined for some temple, but never moved from the spot where it
was hewn.
Ancient Egypt may well look on with scornful wonder at our pride
in our achievements. The Great Pyramid at Gizeh contains three
times as much solid masonry as the Dam, cut from these same
quarries. Every one of those huge blocks had to be dragged to the
river, and carried down 600 miles, before it was hoisted into its place.
As an achievement of mechanical power the Great Dam cannot
compare with the Great Pyramid; but when at last I climbed a little
hill hard by the river, below the Reservoir itself, and saw the whole
length of the great stone rampart, stretching right across the valley,
the contrast between the world of Pharaoh and our own came strong
upon me.
Pharaoh, to whom time and life were nothing, out of the misery of
the forced labour of his subjects, raised a perfectly useless
monument of his own folly; yet he achieved his object, and made his
tomb one of the wonders of the world. We, with the free labour of
voluntary workers—paid, fed, and cared for, instead of being driven
by the whip—have dared to harness Nile himself. It is a work vital to
the interests of millions of dwellers by the river. Yet who can say that
the fame of the Pyramid will not endure the longer? Hundreds of
years after the time of Cheops a mighty Dam was built in Southern
Arabia for a like purpose of irrigation; it lasted for eight centuries or
more, and its bursting in 100 A.D. is mentioned in the Koran. Its ruins
remain to this day, and show that it was larger than the Nile Dam.
Eight hundred years is a long life for a reservoir, but if this one lasts
a quarter of that period it will have repaid its cost many hundred
times over.
Certainly it looks strong enough to last as long as the Nile itself.
Strength, and nothing but strength, shows in every stone of it.
Square, solid, and massive, it runs from shore to shore in an
absolutely straight line, without the slightest attempt at any trace of
ornament or decoration. Clearly, effect has been the last thing
thought of. Even the sluice-gates have absolutely plain rectangular
openings, and it detracts from the symmetry of the design that,
owing to engineering exigencies, they are not all of one height, but
run in sections of five, some of them lower than others. The top of
the wall runs in a simple, unbroken level; there is nothing to catch
the eye as it travels up the steep face of masonry except the
slightest change in the angle of the slope to a nearly complete
perpendicular. The wall simply leaves off because its builders
thought it high enough for the present. Along the broad surface of
the summit runs a tramway, with plenty of room for a man to walk on
either side, flanked by perfectly plain, solid parapets as high as the
waist, and more than a yard thick, of a piece with the masonry below.
On the eastern side the Dam is unostentatiously built into the
living rocks; no arch or pylon marks its start. On the western side it is
flanked by a ladder of four immense locks, set in a mountainous
embankment. The gallows-like arms of the draw-bridge, hideous in
appearance, but a marvel of mechanical ingenuity, over the upper
gate of the highest lock are the only break in the long, unrelieved
level. The hard gray colour of the granite strengthens the general
impression, though in time every part exposed to the action of the
water will be coated with the shining black varnish which the Nile
mud always lays on granite, and it will look exactly the same as the
natural bed of the stream.
How different all this is from the prettiness of the Delta Barrage,
with its brickwork, originally designed and built under French
influence, adorned with archways and towers, of which the lovely
garden, with its flowers and shrubs, its green lawns and leafy trees,
on the tongue of land which it crosses, seems a natural and
appropriate part! The Assouan Dam is the work of a practical,
unimaginative race. Its builders have had before them the problem of
harnessing the great river with a yoke that cannot be broken; they
had to hold up a reservoir containing 100,000,000 tons of water, so
that for 140 miles the river is turned back upon itself; and they have
succeeded.
In December all but a few of the sluice-gates are shut, for the
reservoir has to be filled. But imagine it at the height of the flood,
when the collected rainfall of half a continent is crashing past at the
rate of nearly 1,000,000 tons a minute, and through each of the 180
openings shoots a solid cube of dark water to dash thundering in
clouds of white foam on the rocks below, and rush tumultuously
down the swirling slopes of the cataract. Think of the huge bulk of
water held up when the reservoir is full. Then you will understand
something of the difficulties of the work. Solidity and strength could
not but be the first and overpowering idea in the minds of the
builders. The longer you look, the more you are impressed. The vast
dimensions of the Dam grow upon you from moment to moment.
There is, after all, a fierce beauty in those uncompromising features,
grimly set in the determination to hold the river in bondage. The
massive structure is in harmony with the forces of Nature. Feeble
and puny it may be compared with even the least of their handiwork,
but it is impossible not to feel that its builders have been inspired
with a spark of the same creative power. They drew their plans, and
dug and built and strove their best to control the great river. They
have succeeded because a portion of that spirit of Nature against
which they struggled has passed into their work.
Is it too much to hope that a scheme of decoration may yet be
found in consonance with these ideas? Some day the Dam will be
raised to the full height of the original design, thus doubling the
present capacity of the Reservoir. Then will be the time to finish the
work magnificently, and make of it a stately monument, to be the
glory of Egypt as well as the foundation of her material prosperity.
The subject is worthy of a great artist. Only a scheme conceived on
grand lines, perfectly simple and bold, can have the least chance of
success. Anything else would be as ridiculous as a proposal to place
a statue of ordinary dimensions on the top of the Great Pyramid. The
difficulties are great, and so would be the expense, so great, indeed,
that it would be far better to avoid any attempt at decoration, unless
the results are to be admirable beyond all question. Are there no
possible successors to the architects of Karnak?
The ceremony of inauguration was, like the masonry of the Dam
itself, sensible and solid, but it was not impressive. Those who
arranged its details had forgotten the vastness of the theatre in
which it was performed. From the purely spectacular point of view it
was a failure. Egypt has no money yet to spend on functions.
Perhaps a better stage management might have made a better
display without any greater expense. The material benefits of English
rule would be appreciated none the less gratefully for a little gilding.
But the Englishman in Egypt has had other things to think about than
the organization of what the native would call ‘fantasias.’ So he fell
back on the established custom of his own country. Wherever he
goes he carries with him law and order and equity and righteousness
and commonsense, and he also carries a peculiar kind of public
ceremonial routine. Everybody knows it. The invited guests arrive in
special trains, and perspire in top-hats and frock-coats for an
interminable time, until Royalty arrives full of gracious smiles amid a
cheering crowd. The distinguished persons pass into a pen carpeted
with red baize, where all the notables are assembled. Somebody
makes a perfectly inaudible speech, which receives a gracious and
inaudible reply. A button is pressed here, a lever turned there, and
several extraordinary things begin to happen in consequence. Then
a number of good men and true receive some well-earned
decorations, Royalty graciously departs, and everybody presses
home as best he can, while the band plays the National Anthem. The
whole is accompanied by the clicking of innumerable cameras, and

PDF Market Design A Linear Programming Approach To Auctions and Matching Martin Bichler Ebook Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PDF Market Design A Linear Programming Approach To Auctions and Matching Martin Bichler Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

Market Design A Linear Programming

Approach to Auctions and Matching

Groups Matrices and Vector Spaces A Group Theoretic

Introduction to Linear Programming with MATLAB 1st

JavaScript Object Programming Rinehart Martin

Beginning PyQt A Hands on Approach to GUI Programming

Medical Instrument Design and Development From

Vector Calculus Linear Algebra and Differential Forms A

Statistics for the Social Sciences A General Linear

Quantum Computing Fundamentals From Basic Linear

Martin Bichler is Professor of Informatics at the Technical University of Munich (TUM),

© in this web service Cambridge University Press www.cambridge.org

© in this web service Cambridge University Press www.cambridge.org

University Printing House, Cambridge CB2 8BS, United Kingdom

Cambridge University Press is part of the University of Cambridge.

© in this web service Cambridge University Press www.cambridge.org

To my wife Claudia and my daughters Mona and Sonja

© in this web service Cambridge University Press www.cambridge.org

Part I Microeconomic Fundamentals

© in this web service Cambridge University Press www.cambridge.org

3.2 Utility Functions 35

Part II Multi-Object Auction Design

5 An Overview of Multi-Object Auctions 75

© in this web service Cambridge University Press www.cambridge.org

5.3 Multi-Item Auction Formats 81

6 The Simultaneous Multi-Round Auction Format 85

7 Sealed-Bid Multi-Object Auctions 96

8 Open Multi-Object Auctions 126

© in this web service Cambridge University Press www.cambridge.org

8.3 Models of Open Combinatorial Auctions 144

9 The Combinatorial Clock Auction Formats 167

Part III Approximation and Matching Markets

10 Approximation Mechanisms 189

© in this web service Cambridge University Press www.cambridge.org

11 Matching Markets 206

Part IV Appendices: Mathematical Optimization

A Linear Optimization 245

B Algorithms and Complexity 256

© in this web service Cambridge University Press www.cambridge.org

1.1 Market Design and Mechanism Design

Market design has theoretical foundations in mechanism design theory, an analytical

maximization), information about competitors available to participants, objectives, and

1.2 Market Design and Mathematical Optimization

However, market design is more than mathematical programming, as it takes into

1.3 Outline of the Book

This textbook is intended to provide a one-semester course on market design. I am pri-

1.3.1 Part I Microeconomic Fundamentals

1.3.2 Part II Multi-Object Auction Design

1.3.3 Part III Approximation and Matching Markets

1.3.4 Part IV Appendices: Mathematical Optimization

Part I (Microeconomic Fundamentals) of this book provides a succinct introduction

2.1 Normal-Form Games

Table 2.1 Payoff matrix for the

Prisoner’s Dilemma is a well-known example of a normal-form game and is useful to

One type of strategy available to a player i in a normal-form game is to select a

The probability that an action ai is played in strategy si is denoted by si (ai ). When

2.1.1 Dominant-Strategy Equilibrium

2.1.2 Pareto Optimality

Table 2.2 Payoff matrix for the

2.1.3 Nash Equilibrium

Definition 2.1.6 (Nash equilibrium) A strategy profile s = (s1 , . . . , sn ) is a Nash equi-

Players do not want to deviate unilaterally from a Nash equilibrium. If a predicted

E(uheads ) = E(utails ) ⇒ 1p + (−1)(1 − p) = −1p + 1(1 − p) ⇒ p = 1/2