SHEPSLE, Keneth - Analyzing Politics - Rationality, Behavior and Institutions PDF

Additional Praise for Analyzing Politics
(continued from back cover)
"I regularly use this book in my classes in development and

comparative politics. The students find it accessible and enter-
taining. It's a winner." - Robert Bates, Harvard University
"With its clear presentation of concepts and many examples,

Analyzing Politics has become the standard text for teaching
the basics of rational choice theory in political science."
- Rebecca Mor ton, New York University
"Wonderfully written with powerful and well-chosen exam-

ples, and very easy to teach to students with little or no math-
ematical background."
-John Ferejohn, Stanford Univer sity
"This concise but thorough introduction to rational choice is

an invaluable classroom tool. Ken Shepsle is one of the most
important social scientists working today, and this book pro-
vides guidance on how to read, use, and assess the most
sophisticated analytic tools available."
-Margaret Levi, University of Washington
"The ideal place for students to cut their teeth on one of the
fields at the heart of modern political science."
- Michael Laver, New York University
The New Institutionalism in
American Politics Series
Series Editor, Kenneth A. Shepsle
Analyzing Politics is the cornerstone of The New Institutional-

ism in American Politics series. The series applies analytical
techniques to questions about political processes, actors, and
institutions. Eaclh book provides students with the tools for
understanding the intricate relationships between political
theory and practice, and uses those tools to explore an abun-
dance of real-world examples.
Also available
Analyzing Congress by Charles Stewart III

Analyzing Elections by Rebecca B. Morton
Analyzing interest Groups by Scott H. Ainsworth
Analyzing Policy by Michael C. Munger
Analyzing Politics
SECOND EDITION
'
•
Analyzing Politics
Rationality, Beh avior, and Institution s
S ECOND ED ITION
Kenneth A. Shepsle
HARVARD UN IVERSITY
'
IMt
W · W • Norton & Company
New York • London
In memory of my parents,
Philip a nd Edythe Shepsle
Copyright IC> 2010, 1997 by W. W. Norton & Company, Inc.
All rights reserved

Printed in the United States of America
Second Edition
Manufacturing by Courier Companies, inc.

Book design by Jack Meserole
Production managers: Chris Granville and Christine D'Antonio
Library of Congress Cataloging-in-Publication Data

Shepsle, Kenneth A.
Analyzing politics : rationality, behavior, and institutions I •
Kenneth A. Shepsle.- 2nd ed.
p. cm.
Includes bibliographica l references a nd index.
I SBN 97S-0-393-93507-3 (pbk.)
1. Political science. I. Title.
JA71.S435 2010
324.072- dc22 20100 14477
W. W. Norton & Company, Inc., 500 Fifth Avenue, New York, N.Y. 10110
www.wwnorton.com
W. W. Norton & Company Ltd., Castle House,
75176 Wells Street, London WIT 3Q1'
3 4 5 6 7 8 9 0
Contents
.
Acknowledgments lX
Part I
INTRODUCTION
1 It Isn't Rocket Science, but ... 3

2 Rationality: The Model of Choice 13
Part II
GROUP CHOICE
3 Getting Started with Group Choice Analysis 41

4 Group Choice and Majority Rule 53
5 Spatial Models of Majority Rule 90
6 Strategic Behavior: Sophistication,
Misr epresentation, and Manipulation 156
7 Voting Methods and Electoral Systems 191
vu
Vlll Contents
Part III
COOPERATION, COLLECTIVE ACTION,
AND PUBLIC GOODS
8 Cooperation 231
9 Collective Action 262
10 P ublic Goods, Externalities, and the Commons 306
Part IV
INSTITUTIONS
11 Institutions: General Remarks 355

12 Legislatures 369
13 Bureaucracy and Intergovernmental Relations 407
14 Leadership 445
15 Courts and Judges 473
16 Cabinet Government and Parliamentary
Democracy 503
17 Final Lessons 530
Index 535
Acknowledgments for the
Seeond Edition
The first edition of Analyzing Politics appeared more than a

decade ago. It is an elaboration of the lectures I give in Social
Analysis 46 at Harvard University, a course still offered all
these many years later. I owe a debt to the many students
over the years who have caused me to think hard about ana-
lyzing politics a nd about how best to encourage others to do
so. I hope the fruits of this reconsi~eration make this new edi-
tion not only new but improved as well.
I also owe a debt to the teaching fellows who have collabo-
rated with me in Social Analysis 46. Many are now promi-
nent scholars in their own right. They include Christopher
Afendulis, Alison Alter, Matthew Blackwell, Mark Bonchek,
Ethan Bueno de Mesquita, Eric Dickson, Sven Feldmann,
James Fowler, Sean Ingham, Miriam J orgensen, Michael
Kellermann , Steve Kafka, Iain Osgood, Shana Rose, Alexan-
der Schuessler, Gilles Serra, Melissa Thomas, and Robert Van
Houweling. Iain Osgood has done a splendid job in helping me
add discussion questions and problem sets to chapters of this
new edition. A password-protected instructor website provides
an answer key (wwnorton.com/instructors). In addition he and
• I have worked together to produce the "Experimental Corner"
sections found in many of the chapters of this edition.
Let me also thank the many readers who have contacted
me over the years with questions and suggestions, many of
which have found their way into the present volume. I have
lX
x Acknowledgments for the Second Edition
tried to acknowledge them in footnotes at the relevant places.

Harvard's Department of Government and its Institute for
Quantitative Social Science have been my homes for more
than two decades. In addition to being intellectually stimulat-
ing settings, filled with intelligent, ambitious, and supportive
people, they have provided me with many instances and occa-
sions in which to "analyze politics" as a participant-observer
(especially when I served as chair of Harvard's Department of
Government). I've learned much about politics (sometimes the
hard way) by being here.
My wife, Rise, has always been by my side. Her support
and love are the essential ingredients of my life, allowing me
the privilege to write, think, and teach, for which I am ever
grateful.
- Kenneth A. Shepsle, Harvard University
Analyzing Politics
SECOND EDITION
Part I
INTRODUCTION
1
It Isn't Rocket Science, but ...
This book pretends to be modest, but it really isn't. Although

the prose will be sprinkled with qualification and "on the one
hand this, but on the other hand that," I shall in fact be de-
scribing a sea change in the study of politics. Since the end of
World War II, changes in political studies have been numer-
ous and many-splendored, but I shall concentrate on two of
the most significant shifts of emphasis-from describing to ex-
plaining, and from judging to analyzing.
A S CIENCE OF P OLITICS?
Consider this illustration. We are all familiar with reproaches

in the popular press and in everyday coffee-break conversa-
tions about politicians. Their sins are routinely depicted; their
persons are often held in contempt; and their actions are reg-
ularly alleged to border on the venal, the immoral, and the
disgusting. In nearly every culture, politicians are taken as
scoundrels of one sort or another-sometimes charming, even
enchanting; necessary evils at best, but scoundrels nonethe-
less.1 These characterizations, in both oral and tabloid tradi-
tions, are rich in description and unforgiving in judgment.
1 Why else was Profiles in Courage, John F. Kennedy's book about politicians
who sacrificed their own personal well-being for a greater good, so short?
3
4 Analyzing Politics
Rarely, however, are they more than exercises in storytelling

and hand-wringing.
Political science at the end of World War II was much more
than this. It consisted, first and foremost, of detailed contem-
porary description and political history writing. Any of the
major books on the U.S. Congress, for example, was both a
compendium of facts about current legislative practice and an
account of how that practice had evolved over the entire his-
tory of the institution. The same could be said about scholarly
tomes on political parties, elections, the presidency, the courts,
interest groups, state and local institutions-in short, about
studies of nearly every facet of American political life. The an-
thropologist Clifford Geertz refers to this mode of analysis as
thick description. 2 Thick description is typically careful, de-
tailed to a fault, and comprehensive, in contrast to "mere"
storytelling. But it is also indiscriminate, and the accumula-
tion of detail does not always add up to much more than a pile
of facts.3
Postwar political science is also more than normative
hand-wringing. The evaluative emphasis of postwar political
science consisted chiefly of reformist sentiments. Thus, in the
very same thickly descriptive books on Congress there typi-
cally were objections to the overbearing nature of the Speaker
of the House, concerns about the dominance of congressional
committees, and general unhappiness with the antidemocratic
nature of the filibuster in the Senate, the inordinate influence
of lobbyists, and the intolerance of the majority party for di-
versity of opinion within its ranks (not to speak of the outright
2 Thick descript ion and political history writing, it should be added , were not
unique to the study of America.n politics of the t ime; they were the primary
methods of political analysis of politics in nearly every country in the world.
3 This is true of quantitative data as well as of qualitative facts. A detailed
census of some particular population is just a bunch of numbers until it is

analyzed. The numbers do not speak for themselves. Indeed, a humorist
once described statist ics as a tool of analysis in which the numbers are
grabbed by the throat and beseeched, "Speak to me! Speak to me!tt
It Isn't Rocket Science, but ... 5
contempt of the majority for opinions of the minority party).

Reforms were advocated to cure the legislature of these ills.
But in these descriptions, judgments, and reform sugges-
tions, attention was given by scholars and commentators nei-
ther to explaining the evils they described nor to defending
how their alleg,e d reforms would play out. A codification of his-
torical minutiae about congressional committees, for example,
was not accompanied by any consideration of why committees
were used in legislatures in the first place. That is, why is
ther e a division of labor in legislative bodies, and why does it
take the particular form that it does? Likewise, the advocates
of reform for various disapproved-of practices paid scant at-
tention to why these practices existed in the first place, why
an existing legislature would be moved to adopt the proposed
reform, why the proposed r eform would cure th e problems,
and why the r eformed institutional arrangement would itself
not be subject to further tinkering (or even regression to its
pre-reform condition).
In short, the political science that a college student at the
end of World War II might have encountered was primarily
descriptive and judgmental. It was much less oriented toward
explanation and analysis. Over the next twenty years, politi-
cal scientists got even better at description. They learned
data-collection skills, more precise measurement, and statisti-
cal techniques allowing more pr ecise inferences about causal
relationships. But it was not until the 1960s that systematic
attention began to focus on questions of "why." "Why?" is the
principal interrogative of science; answers to it are explana-
tions, and getting to explanations requires analysis.
The transformation of the study of politics from stories and
anecdotes, first to thick description and history writing, then to
systematic measurement, and more recently to explanation
and analysis, constitutes significant movement along a scien-
tific trajectory. Stories and anecdotes are part of an oral tradi-
tion most commonly found in the communities of journalists,
public-affairs commentators, and politics junkies. There is

often little presumption of anything more than the wisdom that
attaches (sometimes) to those who accumulate juicy detail.
While almost everything is anecdotal and idiosyncratic for the
storyteller, those who measure carefully and describe systemat-
ically are engaged in activity essential to the conduct of scien-
tific inquiry. From systematic description comes the possibility
of identifying empirical regularities. This requires separating
wheat from chaff, so to speak, but this cannot be done until
measurement and description have been carefully conducted.
It is empirical regularities, especially those that are robust
in the sense that they seem to recur often and under a vari-
ety of different circumstances, that pique our curiosity. Why
do incumbent legislators seem to get reelected so frequently
(known as the "incumbency effect")'? In modern democracies,
why do countries that use proportional representation have so
many political parties, while those that use alternative voting
arrangements (for example, district-based systems in which
the candidate with the most votes is elected) have fewer par-
ties (known as Duverger's Law)? Why do democratic states
rarely go to war against other democracies (known as the
"democratic peace" hypothesis)? Why have the countries of the
Pacific Rim developed their economies more successfully than
those of Africa or Latin America in the absence of resource ad-
vantages or other conspicuous differences?
Obviously I will not answer here the wide range of ques-
tions associated with the empirical regularities identified by a
generation of careful measurement and description. But I will
definitely focus quite intently on the "why" questions and on
how to think through to answers. This is one of the major
ways political science is practiced in the twenty-first century,
and is a harbinger of a maturing social science. Political sci-
ence isn't rocket science, but an emphasis on explanation and
analysis moves it closer in form to the physical and natural
sciences than was the case in an earlier era.
M ODELS AND THEORIES
Let's pause now and see where we are heading. The main pur-
pose of this book is not to expose students to specific details
about political life-details you undoubtedly study in other
courses in history, political science, and sociology- but rather
to introduce you to some theoretical tools you will find useful
in making sense of these details. We will be intent especially
on becoming acquainted with elementary models. These are
stylizations meant to approximate in very crude fashion some
real situation. Models are purposely stripped-down versions of
the real thing. Events in the real world are complex bundles of
characteristics, often far too complicated to understand di-
rectly. We depend on a stylized model to provide us with in-
sights and guidance that will shape our analysis of these
events. 4
In the next chapter, for example, we construct a simple
model of human choice. It contains no more than a shadow of
any real flesh-and-blood human being. Instead, in an utterly
simple way, it considers a person exclusively in terms of the
things he or she wants and the things he or she believes. We
want to get a feel for how a person makes choices when con-
fronted with alternatives. Since political behavior is often
about making choices, our model will provide us with hunches
and intuitions about how a generic or representative individ-
ual confronts these circumstances in the abstract.
In Chapters 3 and 4, to take another example, we expand
our focus from the individual to a group of individuals. Al-
though politics is often about making choices, only in: the
world of Robinson Crusoe do individuals make choices entirely
in isolation from others, and that is hardly a very political sit-
4 I have described here how I use the term model in this book. A model of
something is (or should be) recognizable as a highly stylized simplification of
the real thing. A sphere to represent a planet or th e nucleus of an atom is a
good example.
uation. We thus expand upon our ele mentary model of individ-

ua l choice by constructing a more elaborate model of a group
decision setting. ln the remaining chapters of Part II, we em-
bellish this group setting, taking into account more and more
features of the decision-making context. Our model starts out
being simple but begins to take on, a step at a time, some of
the complexity of real groups and the decisions they confront.
Along the way our intuitions about ''how the world really
works" become increasingly sophisticated.
Part III r equires some new, but still basic, models. Whereas
the groups we study in Part II make their decisions, a t least
part of the time, by voting (we may think of t hese groups as
committees or legislative bodies), other kinds of groups oper-
ate in different ways. A group of farmers, for instance, faced
with the problem of a mosquito-infested marsh adjacent to
I.heir respective properties, may decide to cooperate with one
another in order to rid themselves of the nuisance. The ques-
tion that arises here is why any farmer would cooperate. If the
others manage to er adicate the mosquito population, then the
noncooperator ben efits as well (and doesn't have to pay any of
the costs, to boot); if, on the other hand, the others don't man-
age to solve the problem, then only if the one presently not co-
operating would make a difference would he or she bear any of
the cost (and maybe not even then). The issue to be studied
here hinges less on how the group makes decisions- whether
by voting or by some other method-and more on the mecha-
nisms by which these individuals capture the dividends of co-
operation by creating the group in the first place. We need a
model that a llows us to study the logic of participa tion and col-
lective action. I elaborate on this in the chapters of Part III.
I endeavor throughout these parts of the book to enliven
things a bit with cases in point drawn from the real world.
Sometimes the cases are simply concrete instances of what
we're talking about. Thus, in Chapter 2, I illustrate the role of
uncertainty in individual choice by talking briefly about the
risky career choices of politicians. On other occasions, I play a

bit fast and loose with history in order to give an account of
some real historical occurrence to illustrate an important
idea. Thus, in Chapter 3, I provide some concrete examples of
tho manipulation of group choices by looking at how the
strategic maneuverings of congressional leaders influenced
tax policy during the Civil War, the Great Depression, and the
Reagan years. These cases would hardly pass a historian's
muster, since I have quite consciously abbreviated the presen-
tation and stripped the stories of all but what is essential to il-
lustrate a specific theoretical point. Remember, it is not thick
description we're after but theoretical principles.
These stripped-down historical cases come close to trans-
forming a specific model into a theory of something real. A the-
ory, as I plan to use the term, is an embellishment of a model
in which features that are abstract in the model are made
more concrete and specific in the theory. To move from an ab-
stract formulation of group choice to a more concrete applica-
tion involving a specific group (the U.S. Congress) at a specific
time (1862 or 1932 or 1986) on a specific issue (raising na-
tional revenue) requires us to nail our colors to the mast on a
whole variety of matters. What is the specific size of the
group? How many members does it take to pass a motion? Can
the motion be amended? Who gets to make motions? If a mo-
tion is defeated, can a new proposal be made? A theory, then,
is a specialized elaboration of a model intended for a specific
application. 5
We get even closer to t heory in the chapters of Part rv.
There we look at institutions, often in the abstract but occa-
sionally in specific detail. The discussion of legislatures and
their relationships with bureaucratic agents, for example, is
& Model a nd theory are terms used in various ways depending upon which
philosopher of science you wish to consult. Since there is no unifoi·mity of
usage, I can do no more than s tate clearly my own practice, in no sense
claiming superiority or seeking converts.
10 A nalyzing Politics
based on the conduct of intergovernmental relations between

the U.S. Congress and its committees on the one hand, and
between Congress and executive branch officials and regula-
tory bodies on the other. Likewise, the discussion of cabinet
government is based on real-world experiences, mostly in con-
tinental Europe, with multiparty parliaments and coalition
governments.
In this new edition, I have added two features that bring
some of the abstr action associated with models and theories a
bit more to life. The first is problem sets and discussion ques-
tions. In most chapters, I provide a series of problems or ques-
tions to give the reader some experience at putting the
principles of the chapter to work-sometimes on real-life mat-
ters, other times on interesting puzzles of a more abstr act
quality. 6 The second is the Experimental Corner sections. In
many of the chapters, I describe in some detail a social science
experiment, drawn from the new but growing literature in ex-
perimental economics and political science, that seeks either
to discover how real people (e.g., college sophomores) behave,
or to test some proposed theory a bout individual or group be-
havior against data generated experimentally. I hope you find
that these additions enrich the chapters of this book.
P OLITICS
Now that you know where I'm heading, it's time to get on. My
purpose in writing this book is to provide some tools to enable
you to conduct your own analysis of the political events t hat
affect your life. I believe that an understanding and applica-
tion of the concepts contained in this book will help you to pre-
dict a nd explain political events. My job is to present the
concepts clearly and to communicate how they can be a pplied
6 An answer key is provided to instructors at wwnorton.com/nrl.

It Isn't Rocket Science, but . .. 11
to real-world situations. Your job is to engage the material,

understand the concepts, observe how they may be applied to
real-world sitUJations, and try them out in coming to terms
with events in your own life a nd the world around you. If we
do our respective jobs properly, I think you will begin to see
the world differently. You will understand why certain groups
have difficulty cooperating or reaching decisions. You will un-
derstand why people grumble about problems but don't do
anything to solve them. You will understand why political can-
didates and political leaders do some of the crazy things they
do. And you will begin to appreciate why some problems may
be solvable and others not. In short, you will be able to ana-
lyze politics.
I haven't yet hazarded a definition of exactly what I am an-
alyzing when I analyze politics. As a final preliminary, then, I
need to de marcate our subject somewhat. In one of the most
famous definitions, David Easton defined politics as "the au-
thoritative allocation of values for a society." 7 This useful defi-
nition has been around for more than fifty years, but it leaves
out more than I would like. I magine an exhausted breadwin-
ner returning home worn out by a dreary day of "office poli-
tics." There is nothing "authoritative" about, say, the fact that
Smith was trying to impress the boss in order to improve his
chances of getting the regional manager post that would soon
be opening up. Moreover, the intrigues of the workplace surely
do not e mbrace the allocation of values for an entire society.
Office politics, university politics, church politics, union pol-
itics, clubhous·e politics, family politics, and many other
instances besides, all seem to involve what, in ordinary par-
lance, I include under the rubric of "politics," yet seemingly
are excluded by Easton's definition. For the purposes of our
discussion, I will take politics to be utterly indistinguishable
from the phenomena of group life generally. I t consists of indi-
7 David Easton, The Political System (New York: Knopf, 1953).

viduals interacting, maneuvering, dissembling, strategizing,

cooperating, and much else besides, as they pursue whatever
it is they pursue in group life.
One of the real benefits of attaching politics to all facets of
group life is to demystify politics. Our enterprise concerns not
only the "capital P" politics that takes place in the White
House, the Kremlin, 10 Downing Street, Capitol Hill, White-
hall, the Supreme Court, and other places of official activity. It
also includes the "small p" politics of the workplace, the fac-
ulty meeting, the student government committee, the union
hall, the kitchen table, the corporate boardroom, the gather-
ing of church elders, and other less formal group settings.
This hardly defines our subject comprehensively, but I am
content to leave it at this. If you can live with this bit of ambi-
guity, then I invite you to read on.
2
Rationality: The Model of Choice
I n a nalyzing politics I shall take what b as come to be known

as the rational choice approach. It ~so goes by other names:
formal political theory, positive political theory, political econ-
omy. Indeed, in a genuine (but failed) attempt at intellectu al
imperialism, some economists like to think of it as the eco-
nomic approach to politics.1 They are right in one sense. The
rationality assumption has been used most extensively and
has seen its fullest flowering in economics. But there is noth-
ing distinctly economic about rational behavior, as we shall
see. 2
The term rationality has a long history and, in ordinary
language, often means something entirely different from what
I have in mind. If a friend of yours does something that you
would not have done were you in the friend's shoes-say, go to
the movies the night before a final exam-you might say,
"Jeez, that's really irrational." By that you might mean: Given
what your friend wants, that is not the best way to go about
getting it. Or perhaps you mean something different: Given
1 For an excellent essay on positive political theory as a failure of economic
imperialism, see Peter C. Ordeshook, "The Emerging Discipline of Political
Economy,~ in James E. Alt and Kenneth A. Shepsle, eds., Perspectives on
Positive Political Economy (New York: Cambridge University Press, 1990).
pp. 9-31.
2 Even in economics, rationality is undergoing revision under the rubric of
"behavioral economics." For a fine review of this revisionst interpretation,

see Norman Frohlich and J oe A. Oppenheimer, "Skat ing on Thin Ice: Cracks
in the Public Choice Foundation," Journal of Theoretical Politics 18 (2006):
235-66.
13
what I want, I would not do what she is doing (and she ought
to want what I want). In either case, you are claiming that
what your friend is doing is crazy. Crazy it may well be, but I
shall reserve irrationality for something quite specific.
The term rationality as I shall use it does not mean bril-
liant or all-knowing. The inen and women whose behavior we
wish to understand are not gods, so we certainly do not want
to characterize any deviation from omniscient, godlike behav-
ior as irrational (for then nearly all behavior would fall in this
category). The people we model are neither all-knowing nor
worldly wise; they are ordinary folks. As such they have wants
and beliefs, both of which affect their behavior.
PRELIMINARIES
Individual wants, which I refer to as preferences, can be in-

spired by any number of different sources. Clearly we humans
come hardwired with a number of wants related to survival
and reproduction: food, protection from the elements, sexual
desires. Other wants may be socially acquired and only indi-
rectly related to such large and weighty matters as survival of
the species-a preference for the latest fashion in jeans or the
most recent jazz CD. Modern man and woman are economic
and social animals. While one cannot deny the strong influ-
ence of material, economic wants on individual preferences,
additional important sources of preference include religious
values, moral precepts, ideological dispositions, altruistic im-
pulses, and a sense of common destiny with a family, clan,
tribe, ethnic group, or other community.
The individuals who populate our model world are as-
sumed to have preferences derived from any and all of these
various sources. We do not pretend to know why people want
what they want-we leave that to evolutionary biologists, psy-
chologists, and sociologists. Nor do we need to know why in
Rationality: The Model of Choice 15
order to proceed. For us, preferences are one of the givens of a

situation and, for purposes of analysis, we assume that they
don't change much in the short run. In short, we take people
as we find them. 3
I shall occasionally say that people who act in accord with
their preferences are self-interested. Hindmoor has cleverly
noted that "most of us have no difficulty in accepting that
some people are self-interested all the time and that everyone
is self-interested some of the time, but we balk at the notion
that everyone is self-interested all of the time."4 As already
noted, however, I do not require a pinched view that people
are selfish in the ordinary sense of that word, but rather that
people are selfish only in a Jess self-absorbed sense: People
pursue the things they regard as important, to be sure, but
this may includle empathy for family, friends, whales, trees, or
random strangers. An individual's conception of self is re-
flected in his or her preferences and priorities. Pursuit of
those preferences and priorities is self-interest in this weaker
sense at work.
The world of preferences and priorities is an interior world.
Indeed, because a person does not wear her preferences on her
forehead, and sometimes, for subtle reasons, may not be all
that she seems. we often have to make assumptions about her
preferences. That is, in trying to figure out what someone
might do, we have to start somewhere, and entertaining
hunches and intuitions about that person's motives is often a
useful point of departure.
3 Methodologically, this is very similar to the approach of economists, who

take tastes for goods, services, labor, and leisure as fixed in the short run
and determined outside the boundaries of their inquiry. However, let us
reemphasize that preferences in our discussion are construed more broadly
than in conven tional economic models- they should not be equated with ma-
terial well-being. And we should leave room for preference change arising
from learning, experience, persuasion, or deliberation.
4 Andrew Hindmoo:r, Rational Choice (London: Palgrave-Macmillan, 2006),
p. 5.
---------------------
But preferences, tastes, and values are not all there is to

rational behavior. Complementing this interior world is an ex-
ternal environment in which people find themselves. This en-
vironment is filled with uncertainty- about how things work,
about the preferences of others, about random events over
which individuals have neither control nor sometimes even
knowledge. This uncertainty is of interest to us because it af-
fects the way people express their preferences. Individuals
have preferences, as already stated, and I assume they have a
behavioral repertoire or behavioral portfolio available as well.
They may do any of a number of things in pursuit of whatever
it is they want (things like going to the movies or studying on
the night before a final). They often cannot choose the thing
they want directly (like getting an A on the final), but instead
must choose an instrument-something available in their
portfolio of behaviors. If each instrument leads directly to
some distinct outcome, then the job of the rational person is
simple: choose the instrument that leads to the outcome pre-
ferred the most. If you want a n A on the final, and studying
the night before produces it while going to the movies does
not, then by all means study.
Enter uncertainty. More often than not, individuals may
not have an exact sense of how an instrument or behavior
they might adopt relates to the outcomes they value. That is,
they may have only the vaguest sense of "how the world
works," may not quite appreciate how the choices of others in-
fluence the final outcome, and may not be able to anticipate
random events (like the virus that arbitrarily picks you on the
morning of the exam). Consequently, the effectiveness of be-
havioral instruments for the things an individual wants is
only imperfectly known. Personal knowledge and wisdom take
one only so far. But one must use what he or she has avail-
able. We describe the hunches an individual has concerning
the efficacy of a given instrument or behavior for obtaining
something he or she wants as that person's beliefs. Beliefs con-
nect instruments to outcomes. Acting in accord both with one's

preferences and one's beliefs is called instrumental rationality.
Beliefs, like preferences, come from a variety of sources,
and we need not resolve their origins in order to take them as
part of what defines a n individual at any moment in time. In-
deed, beliefs may change as the individual acquires experi-
ence in his or her external environment. Learning takes place,
causing the individual to revise initial opinions about the ef-
fectiveness of a specific instrument for achieving some partic-
ular objective. To be on the "steep" part of a learning curve
means for you to be in a relatively novel situation of high un-
certainty in which each new bit of experience causes you to r e-
vise your views about how the world works in this situation.
As bits of experience accumulate, your beliefs begin to settle
down, your opinions begin to firm up, and you revise your
opinions less frequently and dramatically; you are in the "flat"
part of the learning curve-you've learned most of what there
is to know and have squeezed out most of the uncertainty
(that is squeezable) .
I've done a fair bit of throat-clearing to this point. 'lb sum
up this preliminary discussion, the conception of rationality
employed in this book incorporates both preferences and be-
liefs. A rational individual is one who combines his or her be-
liefs about the external environment and preferences about
things in that environment in a consistent manner. Since we
have no time to spare, I can only note in passing that the ra-
tional choice approach is a form of methodological indiuidual-
ism. 5 The individual is taken as the basic unit of analysis. In
contrast, many sociological theories take the group as the
basic building block. Marxist approaches begin with economic
classes as the actors in their models. Most theories of interna-
5 For r eaders interested in pursuing this theme further, see Geoffrey Brenna n
and Michael Gillespie, eds., "Special Issue: Homo Economicus and Homo
Politicus," Public Choice 137 (2008): 429-524.
tional relations aggregate all the way up to the nation-state as

the unit of analysis. Indeed, even some economic theories
treat aggregates like a firm or an entire industry as the unit of
analysis. The most important thing to know about method-
ological individualism is that it is taken as fundamental that
individuals have beliefs and preferences. These things are
the stuff of human cognition and motivation. Groups, classes,
firms, and nation-states do not have minds, and thus cannot
be said to have preferences or hold beliefs.
Now it is time to make these ideas more precise.
MOTIVATION
To motivate a rational model of political behavior, let's begin

with a glimpse of how economists practice their craft. I sim-
plify shamelessly in advancing the view that economics is con-
cerned mainly with how four different classes of actors choose
to allocate what's theirs. For the consumer the choice is one of
how to spend his or her monetary endowment so as to achieve
a maximum of contentment (or utility, as the economist likes
to say). Producers, on the other hand, possess various produc-
tive inputs and must determine how best to combine them so
as to maximize their profits. The endowment of a worker con-
sists of time. To keep things simple, suppose that workers toil
at a fixed wage rate so that once they decide how much time to
spend at work, both their total wages (and hence monetary
endowment from which they derive contentment when they
transform themselves into consumers) and the amount of time
left over for leisure are determined. Workers, then, pick an
amount of time to work in order to acquire purchasing power
and leisure time, each of which contributes to their content-
1nent. Finally, investors are providers of capital. They allocate
their wealth across alternative investment opportunities with
DISPLAY 2.1
Actor E ndowment Objective
consumers budget contentment
producers inputs profits
workers time purchasing power/leisure
investors wealt h long-run return
an eye on the overall long-term financial return. This is laid

out in Display 2.1.
Now, there is surely a mbiguity in each of these ideas,
but it is fair to say that economists, in one fashion or another,
firmly commit t hemselves to what it is that animates vari·
ous economic actors. This is not because they think their
assumptions- and Display 2.1 is one very simple set of as-
sumptions-are verifiable a s descriptive statements. Start
asking some obvious questions and you will quickly determine
that these assumptions are seriously flawed as descriptive
statements. Don't consumers care about anything except their
own consumption? Are producers driven entirely by the profit
motive, or do th ey give some weight to other things, like the
welfare of their workers or the quality of their products, even
if these sometimes come at the expense of profits? Don't work-
ers get any satisfaction from the job itself, or do they only care
about how much effort they must expend (and indirectly about
their wages)? Is investing for long-run return the only thing
an investor can do with her wealth? Can't she devote some of
it to finance a tropica l vacation in the dead of winter, or a new
wing on the local children's hospital?
It is evident that descriptive accuracy is not the point or
purpose of the economist's assumptions. The reason is scien-
tific, not substantive. The idea is this: Can we explain varia·
tions and regularities in economic performance, outcomes, and
behavior with a simple set of assumptions? The modern the-

ory of economics is a grand intellectual edifice precisely be-
cause it has succeeded, as no other social science has, in
constructing explanations logically, rigorously, and in empiri-
cally meaningful ways. At the foundation of this edifice is a
scientific commitment to explanation, not description. 6
This does not mean there is no controversy in economics.
What it does mean is that over the past two centuries, a cor-
pus of scientific knowledge has accumulated, a corpus different
from either an encyclopedia of descriptive detail or even a
body of wisdom (by which we mean serviceable commonsense
notions). It is, instead, a logically integrated collection of prin-
ciples, a set of tools of inquiry-a methodology, if you will-for
prediction and explanation. Of great importance is the fact
that this scientific knowledge is cumulative, something that
distinguishes it from wisdom, which is intuitive, implicit, and
often nontransferable (it dies with its possessor).
Is it possible, in a manner precisely analogous to what has
occurred in economics, to create a science of politics? That is,
is it possible to begin with a simple set of premises or assump-
tions and, from these, derive principles of political perfor-
mance, outcomes, and behavior? This is a daunting challenge,
but it is the objective that has motivated the body of work in
positive political theory that is the focus of this book.
THE SIMPLE LOGIC OF

PREFERENCE AND CHOICE
Our first building block is the notion of preference. We must

begin by defining terms, explaining notation, and making as-
sumptions. Since the machine we are building must serve in a
6 For an elaboration of this issue as a philosophical matter, see Frank Lovett,
"Rational Choice Theory and Explanation," Rationality and Society 18
(2006): 237- 72.
variety of contexts, our building blocks must be developed in

an abstract and general fashion (for which I beg the reader's
indulgence). To give the reader something concrete to hold
onto, however, consider the dilemma that Claire McCaskill,
state auditor of Missouri, confronted after the 2004 election.
CASE 2.1
CLAIRE McCASKILL'S E LECTORAL OPTIONS
Claire McCaskill graduated from law school in 19.78 and

went on to become a very successful state politician in Mis-
souri. After several years of judicial clerking and private
practice, she was a local county prosecutor, a county repre-
sentative, a state representative, and, in 1998, was elected
statewide as a111ditor, a position to which she was reelected
in 2002. Along the way she broke down barriers for women
in politically conservative Missouri. In 2004 she defeated
Governor Bob Holden in the Democratic gubernatorial pri-
mary, becoming the first person in Missouri history to de-
feat a sitting governor in a primary election. She lost the
general election to Secretary of State Matt Blunt, a Repub-
lican, with 48 percent of the vote to his 51 percent. This was
her first losing effort in a twenty-year political career.
What would her next career step be? Talking heads and
political insiders assumed that her close loss for governor in
a year when ciircumstances favored Republicans would pro-
pel her candidacy for governor four years later-a rematch
against Blunt, the only politician who had ever beaten her.
However, there was the possibility of 2006. Jim Talent, the
incumbent Republican senator was up for reelection. He
had come to the Senate in the 2000 election cycle in a most
peculiar way. Mel Carnahan, then Democratic governor of
Missouri, died in an airplane crash while campaigning for
the Senate seat only days before the election. His name re-
mained on the ballot and he actually won the election! A

special election was then called, and Talent defeated Car-
nahan's widow, Jean.
McCaskill had a choice to make. She could challenge the
incumbent Talent for the Senate seat in 2006, or wait and
challenge the incumbent Blunt for the governorship in 2008.
If she ran and won the Senate seat in 2006, she would not
enter the governor's contest in 2008. If she ran and lost the
Senate seat in 2006, her prospects, as a two-time loser, of
succeeding in 2008 would have been rather dim.
Thus, the career outcomes facing McCaskill were these:
x: a term as senator
y: a term as governor
z: out of politics for the near term
We can be reasonably confident that McCaskill preferred x

toy and y to z (and, as will be shown later in this chapter, if
she weren't "incoherent'' she would also certainly have pre-
ferred x to z). But she could not literally choose from {x,y,z}.
Her behavioral options were "run for senator," "run for gov-
ernor," and "run for both." Since I've assumed she would not
choose to run for governor if she won the Senate seat, and
she could not expect success if she ran for governor after
losing the Senate race, it is reasonable to suppose that Mc-
Caskill focused only on {"run for senator in 2006," "run for
governor in 2008"}.
We can think of her decision problem as that of choosing
between two lotteries. If McCaskill chose to run for the Sen-
ate in 2006, she would obtain outcome x with some proba-
bility p and outcome z with probability 1-p. If she held off
to run for governor in 2008 on the other hand, she could ob-
tain outcome y with probability q and outcome z with prob-
ability 1-q. The keys to her decision are how much she
prefers x to y a nd how good are her chances for the former
(p) as opposed to the latter (q).
McCaskill ended up running for (and winning, as it hap-

pens) a Senate seat in 2006 because her outright preference
for x over y was reinforced by a belief about her chances of
victory. In either case she would be running against an in-
cumbent, so that factor was more or less a wash. The rea-
son she thought 2006 would be a better year to run than
2008 was her fear that Hillary Clinton would win the
Democratic presidential nomination in 2008. (Recall that
back in 2005 and 2006, people believed that Clinton as the
Democratic nominee was a fait accompli.) In her view, no
Democrat would win statewide office in relatively conserva-
tive Niissouri with Ms. Clinton at the head of the ticket.*
*For a wonderful essay on Claire McCaskill's decision making, see Jeffrey
Goldberg, "Letter from Washington- Central Casting: The Democrats
Think About Who Can Win in the Midterms-and in 2008," The New
Yorker (May 29, 2006).
In the remainder of this section I abstract from the specific

features of this case in order to develop a general logic of ra-
tional choice. We begin with a situation in which there are
three objects over which a typical actor, named Mr. i, has pref-
erences. We call the objects alternatives, and label them x, y,
and z. Mr. i, in .a manner we make precise below, has the ca-
pacity to make statements like, "I prefer x toy," or "I am indif-
ferent between y and z." The alternatives may be career paths
(as in McCaskill's choice problem), or political candidates, or
potential marriage partners, or laptop computers. It does not
matter, for our purposes, what comprises the choice situation
or the set of alternatives. Nor does it matter how Mr. i arrived
at his preferences. What does matter is that Mr. i is rational
in the sense that his preferences have coherence and that his
ultimate choice bears a logical relationship to his preferences.
Symbolically, "xP;y'' means Mr. i (whose name appears as a
subscript) prefers x toy. In words, the symbols in quotation
marks state that "xis better than y according to Mr. i's prefer-
ences." Similarly, "xliy" means Mr. i is indifferent between

x and y. Thus Pi is i's strict preference relation and Ii is i's in-
difference relation.7
If Mr. i is given the opportunity to choose among x, y, and
z, then we say that his choice is rational if it is in accord with
his preferences. Thus, a choice is rational if the object chosen
is at least as good as any other available object according to
the chooser's preferences. Put differently but equivalently, an
object is a rational choice if no other available object is better
according to the chooser's preferences.
So far this is pretty straightforward and, once you get used
to the notation, pretty commonsensical. Now we must deter-
mine what must be true about the preference and indifference
relations just described so that choosing in conformity with
them accords with our intuitions about rational choice. What
we are seeking, in effect, are properties of preference relations
that allow the chooser to order the alternatives in terms of
preference (and enable him, being a rational soul, to choose
the top-ranked alternative in the ordering). It turns out that
two underlying properties capture the commonsensical notion
of rationality as ordering things in terms of preference:
Property 1: Comparability (Completeness). Alternatives are
said to be comparable in terms of preference (and the
preference relation complete) if, for any two possible al-
ternatives (say, x and y), either xP,y, y~x, or xliy. That
is, the alternatives are comparable if, for any pair of
them, the chooser either prefers the first to the second,
the second to the first, or is indifferent between them. 8
7
Putting strict preference and indifference together yields i's weak preference
relation, R;, so that "x R,y'' means that Mr. i either strictly prefers x toy or
is indifferent between them. In words, "xis at least as good as y according to
Mr. i's preferences."
8 Equivalently, the alternatives are comparable if, for any pair of alternatives
like x and y, either x R,y or y R, x or both. In words, a person has complete

preferences if either xis at least as good as y, or y is at least as good as x, or
both (that is, each is as good as the other). If the latter, then x l ;Y·
Property 2: Transitivity. The strict preference relation is

said to be transitive if, for a ny three possible alternatives
(say, x, y, and z), if xP;Y a nd yPiz, then xPiz. That is, if
Mr. i strictly prefers x to y and y to z, then he strictly
prefers x to z. Likewise, the indifference relation is tran-
sitive if xliy and yliz imply xl iz (if i is indifferent be-
tween x and y and between y and z, then he is indifferent
between x and z, too). 9
As Case 2.1 makes clear, Senator McCa skill possessed
complete and transitive preferences over the alternatives
{x, y, z}. She preferred a Senate seat (x) to the governorship
(y); a term as governor (y) to being out of politics altogether
(z); a nd, of course, the Senate seat (x) to the political wilder-
ness (z).
If i's preferences satisfy comparability and transitivity,
then i is said to possess a preference ordering. As noted, the
rational choice is the alternative at the top of the ordering.
Note that Pi and Ji are exactly like > (greater tha n) and =
(equal to), respectively, as applied to real number s. For real
numbers x and y , either x > y, or y > x, or x = y; hence, they are
comparable. Similarly, for any three numbers, x, y, a nd z, if
x > y and y > z,, then x > z; and if x = y and y = z, then x - z;
hence the relations are transitive. In consequence, real num-
bers can be ordered in terms of magnitude. (The reader can
check for himself or herself that the weak preference r elation,
R i, is analogous to ~ [greater tha n or equal to] as applied to
real numbers.)
This is all pretty simple. Preferences that permit rational
choices are, in effect, ordering principles. They are personal-
P1is Mr. i's particular way of ordering alternatives, which may
differ from Pi' Ms. j's way of ordering the alternatives. They
allow comparisons of alternatives a pair at a time (compara-
q Finally, the weak preference relation is trans itive ii x R,y and y R1z imply
x R,z.
bility). And the comparison they permit are internally consis-

tent (transitivity).
Before concluding that all is well and moving on, however,
we must satisfy ourselves about exactly what we are assum-
ing. We need to ask if all relations satisfy properties 1 and 2.
If so, then we haven't made very hard demands at all. If not,
then we need to know precisely what is excluded from consid-
eration by our assumptions.
In fact, not all relations are complete or transitive (or
both). Some relations satisfy transitivity but not complete-
ness.10 Others satisfy completeness but not transitivity.11 And
still others satisfy neither. 12
lO The relation "is the brother of" applied to the set of all males satisfies
transitivity but not completeness. It violates completeness since neither
"John is the brother of Bob" nor "Bob is the brother of John" is true if they
are not brothers! However, if John is the brother of Bob, and Bob is the
brother of Charles, then John and Charles are also brothers, so the relation
is transitive.
11 Suppose Ms. i prefers Bill Clinton to George H. W. Bush (C P; B), Bush to
Ross Perot (B Pi P), and Perot to Clinton (PP, C) in t he 1992 presidential

election. The alternatives clearly satisfy comparability, but they violate
transitivity. You may think Ms. i quite daffy in this case, but we know peo-
ple like her and expect you may, too. For example, whenever i thinks about
the Clinton-Bush comparison, domestic policy issues are triggered in her
mind ("It's the economy, stupid!" was the Clinton campaign war chant in
1992, after all), and she prefers the Democratic candidate on these issues.
Whenever she thinks about the Bush-Perot comparison, foreign policy is-
sues loom large and she worries about the ship of state in the hands of a
businessman with n o diplomatic or political experience, like Perot. Finally,
whenever Ms. i makes the Perot-Clinton comparison, she can't help thinking
about the character issue on which the businessman with no political skele-
tons in his closet dominates someone who has been nothing but a politician
his entire a dult life. Intransitivity or inconsistency may arise when different
criteria are used for different pairings. When this happens, it is not possible
to order all three alternatives in terms of preference. Ms. i ranks Clinton
ahead of Bush, Bush ahead of Perot, and Perot ahead of Clinton.
12 The relation "is the father of" satisfies neither completeness nor transitiv-
ity. Suppose we take the population of males and draw two at random. It is
entirely possible, indeed highly probable, that neither one is the father of
the other; thus, not all pairs of alternatives are comparable according to
this relation. On th.e other hand, even if, for three selected males, the first
is the father of the second and the second is the father of the third, it is ob-
vious to any five -year-old that the first is not the father of the third, but
rather is the grandfather. That eliminates transitivity.
So I have actu ally said something of substance when I as-

sume properties 1 and 2. The issue now is whether I can de-
fend what's been said. Regarding comparability, clearly you
could push things far enough so that making a comparison in
terms of preferences would be absurd. Sophie's Choice is au-
thor Willia1n Styron's literary invention for this absurdity. I n
his novel, a concentration camp prisoner in Poland is permit-
ted to save one of her two children from the gas chamber, but
she must choose whom to save; in the absence of a choice, both
will die. It is a horrible, inhuman choice. Nevertheless, Sophie
does indeed choose (for not to do so is far worse), even though
she does not regard her children as comparable. Horrible
choices may be painful to make, and so1ne of us may ulti-
mately lack the courage to do what we must. But, as in So-
phie's case, even the failure to choose is a choice with its own
consequences.
The real problem for the comparability property comes in
situations in which the comparison doesn't make sense to the
chooser. If objects do not connect up in the mind of the chooser
as competing alt-ernatives, then you are likely to get shrugs,
puzzled looks, and, if given the option, a response of "don't
know." If pollsters, in late 2007 or early 2008, were to have
asked a random sample of voters whether they preferred John
McCain or Barack Obama in the 2008 presidential contest,
they would have obtained many don't-know responses, for
that far in advance of an election, most candidate pairings re-
ally don't connect in the mind of the average voter. This is not
a critique of rationality-based models so much as a warning
label advising appropriate use. Choices must have meaning to
the choosers if they are to be guided by principled considera-
tions such as those associated with rationality. 13
13 Although peripheral to the main line of argument, it is nevertheless inter-

esting to ask what it means for someone to say "don't know" when con-
fronted by a pollster with one of these puzzling choices. It could either mean
"this comparison is loony and I cannot make a choice," or "th e a lternatives
Transitivity requires that the chooser not be confused in a

different sense. It requires consistency, something in short
supply at times. Psychology professors have, since time imme-
morial, imposed saline-solution and shades-of-gray experi-
ments on captive sophomores in introductory psychology
classes. The typical experiment begins with ten bottles of
water of varying salt content (or ten pictures of a triangle col-
ored white, black, or some shade of gray). A student is asked,
when presented with two bottles (or triangles), which tastes
saltier (or is darker). Her answer is recorded and then a differ-
ent pair is presented. This continues for some time as alterna-
tive pairs are presented and answers recorded (there are
forty-five distinct pairs). Because the different saline solutions
shade into one another (as do the gray triangles) , invariably
the student, sometime during the experiment, answers that
bottle 2 is saltier than bottle 9, that bottle 9 is saltier than
bottle 7, but that bottle 7 is saltier than bottle 2-a clear vio-
lation of transitivity. It is hard to be consistent in the manner
property 2 requires when the comparisons are so difficult,
when there are potentially consequential random events for
which the experimenter does not control (such as how thirs.ty
the subject is or how much sunlight is coming into the room),
when so little is at stake, and when the answers of a particu-
lar subject aren't likely to make much difference.
This, too, is less a critique of rationality than a warning
about the domain over which it is likely to be more or less rel-
evant and useful. When the stakes are low, uncertainty is
high, and individual choices are of little consequence to the
chooser, then inconsistencies are likely to be common. Behav-
ior is likely to be more random than rational, more arbitrary
are so close in my r.nind in terms of preference that it is a matter of indiffer-

ence to me." Which one it is is a judgment call that the researcher needs to
make. In terms of predicting behavior, however, it may not make any differ-
ence. Whether a person is indifferent or confused, if a choice is forced, his or
her behavior ii! likely to be random.
than principled. But when the choices matter to the chooser,

he or sh e is likely to be more inte nt on being consistent. As in
the case of comparability, whether transitivity is appropriate
or not is a judgment call to be made by the investigator. The
kind of consistency required by this property is demanding, to
be sure, even in more significant situations . But we need it to
get on with our business and must content ourselves with the
knowledge that, as in other sciences, simplifying assumptions
are necessar y in order to make progress.14
T HE MAxlMIZATION P ARADIGM
The assumptions of comparability (completeness) and transi-

tivity yield an "ordering principle"-they permit an individ-
ual to take a set of objects and place them in an order, from
highest to lowest (with ties permitted), that reflects personal
tastes and values. Rationality is associated with both this ca-
pacity to order and an aptitude to choose from the top of the
order.
The very existence of a "top" to a preference ordering, and
individuals with sufficient sense to choose it if given half the
chance, is the reason that most of us working in this tradition
think of rationality as consisting of maximizing behavior. In-
dividuals in social situations are thought to be seeking some
goal, pursuing some objective, aiming to do the best they can
according to their own ligh ts. Indeed, instead of describing an
individual in terms of his or her preferences, we may write
down the principle that led the individual to order alterna-
14 Transitivity strikes me as an assumption like that of perfectly spherical

atomic particles in particle physics, perfectly spherical planets in astron·
omy, and frictionless planes in mechanics. All were known to be contrary to
fact, even as they continued to be used; all never theless proved essential to
move the science forward; all ultimately were r elaxed as later generations
of scientists subsequently saw bow to strip away the offending parts.
tives as he or she did. We may, in other words, state what it is

that the person is seeking to achieve or trying to maximize.
The earlier economic example (Display 2.1), in fact, did
this. Consumers are interested in maximizing contentment,
producers want to maximize profits, workers want the best di-
vision of their time between labor and leisure, and investors
want the highest long-term return on investment. In the vari-
ous political models that are examined in the next three parts
of the book, political actors are similarly intent on maximiz-
ing. Elected politicians, for example, are interested in maxi-
mizing their votes at the next election. Legislators seek to
maximize the amount of pork and other policy satisfaction
they can deliver to the folks back home. Bureaucrats are in-
terested in maximizing their budgets or their turf. The lan-
guage in the remainder of this book will often reflect this
maximizing perspective.
E NVIRONMENTAL UNCERTAINTY AND BELIEFS
Rational individuals choose from the top of a set well ordered

according to preference. In many circumstances, however, the
individual doesn't get to choose outcomes directly, but rather
chooses an instrument that affects what outcome actually oc-
curs. Claire McCaskill (Case 2.1), for example, could not sim-
ply choose to be the senator from Missouri in 2006. All she
could choose was the option to run for that office. So we should
revise our idea of rationality, saying now that a rational indi-
vidual chooses the instrument or action he or she believes will
lead to the best outcome.
I slipped the word "believes" into the reformulated defini-
tion of rationality. Just as I was precise about preferences ear-
lier, I need now to be precise about beliefs. A belief is a
probability statement relating the effectiveness of a specific
action (or instrument) for achieving various outcomes. If an
individual is highly confident that he knows what will happen

if he does some particular thing (for example, if I turn the
handle the door will open), then he is operating under condi-
tions of certainty. An incumbent politician's choice to seek
reelection against a "sacrificial" opponent is made under con-
ditions of (near) certainty. If, on the other hand, a person is
not so confident that she knows what will happen, but never-
theless has a pretty clear sense of the possibilities and their
likelihoods (if I turn the handle, there is a fifty-fifty chance
that the door will open or be locked), she is operating under
conditions of risk. Thus, McCaskill's choice to oppose the in·
cumbent Jim Talent was a gamble made under conditions of
risk. Finally, if in the mind of the chooser the relationship be-
tween actions and outcomes is so imprecise that it is not
possible to assign likelihoods, then she is operating under con-
ditions of uncertainty.
To see what is meant by certainty, risk, and uncertainty,
consider the following example in which there are three possi-
ble outcomes-x, y, and z-and three actions- A, B, and C.
Our chooser has preferences over the outcomes; suppose she
ranks x first, then y, then z-written xyz, but she must make a
choice from among the actions. If she knew for certain that
C led toy, that Bled to z, and that A led to x, then her decision
is one of cer tainty (and, as the reader can ascertain, a pretty
simple one-choose A). If, on the other hand, she knew that A
led to a fifty-fifty chance of x or z, that B led to a fifty-fifty
chance of y or z, and that C led to an even chance of x, y, or z ,
then the choice involves risk (and is a bit more complicated).
Finally, if she weren't sure how to put probabilities on the
odds of various outcomes from specific actions, then she would
be uncertain (and, without further analysis, the appropriate
choice would be quite illusive).
When there is certainty, rational behavior is pretty appar-
ent: Simply pick the action or instrument that leads to your
highest-ranked alternative. When beliefs about action-outcome
relationships are more complex, the principle of rational be-

havior requires more explanation. You need to assign a nu-
merical value to each outcome, called a utility number. The
utility numbers for x, y, and z, respectively, are u(x), u(y), and
u(z); they reflect the relative value you associate with each
outcome. If you like x a whole lot better than y and z, and
there is not much difference between the latter two in your
mind, then u(x) will be a much larger number than u(y) and
u(z), and the latter two numbers will be close in magnitude-
for example, u(x) = 1, u(y) = 0.2, u(z) = 0. On the other hand, if
x is only barely your first choice, with z trailing badly, then the
utility numbers would be on the order of u(x) = 1, u(y) = 0.9,
and u(z) = 0.
In effect, we have "quantified" preferences by moving from
ordered preference information to numerical preference infor-
mation. There is nothing magical about the particular num-
bers we wrote down-they are gauged, by you, to best reflect
your relative valuations of the alternatives. 15 Now let's do the
same for beliefs. For each action or instrument, we can write
down the probability that it will lead to one of the final out-
comes. In the example above, action A led to a fifty-fifty
chance of x or z; that is, PrA(x) = 1/2, PrA(y) = 0, and PrA(z) =
1/2. The probability numbers must all be between zero and
one, and they must add up to one. As you can see, these beliefs
about action A effectively make A a lottery--0ne in which y is
15
It is the relative numerical values, not their absolute values, that convey
this kind of information. Consequently, it is typical to "normalize" the util-
ity numbers, setting your most·preferred alternative to a utility value of
one, your least· preferred to a value of zero, and intermediate alternatives at
utility levels between zero and one. It would have done just as well to set
most-preferred and least-preferred alternatives at 100 and 0, respectively,
or 1000 and - 1000, respectively. The normalization values are arbitrary. We
report on all this only for the rare reader who wishes to delve more deeply.
A standard, accessible reference for further details is Howard Raiffa, Deci-
sion Analysis (Reading, Mass.: Addison-Wesley, 1968). Readers will not
need very much detail to digest the materials in the remainder of our book,
so breathe easy!
an impossibility and x and z are equiprobable. We can write

A = (1/2 x, 0 y, 112 z). Each of the other actions is a different
lottery over final outcomes.
Making a decision under conditions of risk involves choos-
ing from a mong alternative lotteries. A rational choice en-
tails choosing th e "best" lottery. The rule of rational choice
is known as the Principle of Expected Utility. It provides a
method for assigning a single number to each action-lottery
and then choosing the one with the largest number. The ex-
pected utility of action A of the previous few paragraphs is
EU(A) = PrA(x) • u(x) + PrA(y) • u(y) + PrA(z) • u(z)
That is, the expected utility of action A is simply the sum of
the utilities of all the outcomes that could result from A,
weighted by the likelihood that each outcome will happen. If
we make the same calculation for actions B and C, then we
have a basis for comparing them. Rationality requires a
chooser to select the action that maximizes expected utility.
Under conditions of uncertainty, a chooser is sufficiently
confused that he or she cannot even figur e out the likelihoods
of various outcomes associated with each action. Needless to
say, it is hard to be rational, however you might define it, if
you are utterly confused. It turns out, however, that many
people do have hunches about likelihoods that they can associ-
ate with various actions. So, if pushed a bit, they can give
some quantitative precision to their beliefs. They, too, can
be treated as if the expected utility principle covered their
behavior. 16
16 There are many theories of decision under uncertainty that cover the cir·
cumstanccs in whlch choosers cannot assign probabilities of outcomes to
alternative actions. We do not review them here. Still one of the best pre·
sentations of this material is to be found in R. Duncan Luce and Howard
Raiffa, Games and Decisior1s (New York: Wiley, 1957), Chapter 13.
CONCLUSION
I have covered q uitc a bit of ground in this chapter and the

last. But everything can be summarized with a few simply
stated ideas. First, our general enterprise is that of explaining
social and political events and phenomena. Second, the indi-
vidual is our basic explanatory building block. Third, because
we are interested in prediction and explanation rather than
description, we characterize individuals in a very abbreviated
form, namely in terms of their preferences and beliefs. Fourth,
the individuals in our analysis are rational. This means that
they act in accor d with their preferences for final outcomes
and their beliefs about the effectiveness of various actions
available to them. The cause-and-effect relationships between
actions and outcomes may be well defined (certainty), proba-
bilistic (risk), or only crudely known (uncertainty). Fifth,
acting rationally requires ranking final outcomes, assigning
utility numbers to them if necessary, determining the ex-
pected utility of actions by weighing outcome utilities by ac-
tion probabilities, and then selecting the action that has the
highest expected utility. Sixth, and perhaps most controver-
sial of all, rational political choices-whether career paths
chosen by politicans, candidates chosen by voters, decisions to
go to war made by kings or presidents, or something as ba-
nal as pizza toppings chosen by a group of friends-all are
premised on the same comparability-and-transitivity founda-
tion. Aristotle, Hobbes, Rousseau, and other great political
thinkers and philosophers have suggested that there is some-
thing special about politics-that the collective choices for a
nation, for example, are altogether different from choosing
pizza toppings. Perhaps. Indeed, certainly this must be the
case. But the process of choosing rationally bears characteris-
tic markings in all these contexts and so may be analyzed
with the same intellectual framework.
In what follows, this rationality machinery is used repeat-

edly while keeping technical matters to an absolute minimum.
So, having covered the preliminaries, let's move on to the
study of groups and their politics.
PROBLE~1S AND D ISCUSSION Q UESTIONSa
1. How is rationality defined in this chapter? Answer with r ef-

erence to both preferences and behavior, and then concoct an
example of a violation of each of these aspects of the rational
actor model, explaining carefully which assumption has been
violated.
2. Rational choice models generally start with a well-defined

set of actors (N = {l, 2, ... , n}), a number outcomes over which
actors hold preferences (X = {x, y, z, ...}), a set of behaviors
or instruments with which to achieve preferred outcomes
(I = {A, B, C, ...}), and some rule which links actors' instru-
mental choices t o outcomes (R). For example, each of n voters
may vote for A , vote for B , or abstain. A candidate wins if he
or she gets more votes than any other. Thus, N = {l, 2, ... , n},
X = {A wins, B wins, tie}, and I= {vote A, vote B , abstain}. The
rule, R, is plurality rule, implying that a vote for A (or B) in-
creases the likelihood A (B, respectively) wins. Give simple
characterization of each of these model foundations for the fol-
lowing political actors: campaign contributors, political ac-
tivists, and candidates. In what forms do these actors confront
uncertainty in making their behavioral choices?
• In this and succeeding chapters, I provide some problems and discussion

questio ns to elabor ate ideas in the chapter and to allow th e student to test
his or her mastery. Difficult questions are marked with an asterisk.
3. Rational choice is a methodology defined by instrumental

action towar d a goal, where the goal itself is deter mined by in-
dividual values. Given this definition, is it possible for rational
individuals to un dertake altruistic acts? Provide an affirma-
tive response that a dyed-in-the-wool rat choicer would offer,
as well as the per spective of a critic.
4. Mr. i holds the following preferences over outcomes w, x, y,

and z: xPw, xPy, zPx, yPz, wPy, and wPz. When presented
with a choice over any subset of these outcomes (e.g., x, y, and
z; or all four outcomes), for which subsets can Mr. i identify
his most-preferred choice? Do any of those subsets contain a
preference intransitivity among all outcomes in the subset?
Now consider Ms . j, who holds preferences: xly, xPz, xPw, yPz,
yPw, wlz. Answer the same questions as before. What does
this exercise suggest about the relationship between transi-
tive preferences and maximizing behavior?
5. In November 2008, a couple of weeks after the election of

Barack Obama, Hillary Clinton was offered the job of Secre-
tary of State of the United States. It was generally assumed
that she faced the following trade-off: joining the new admin-
istration, in perh aps the highest-profile cabinet position,
which offered the chance of enhanced prestige and policy-
making clout in the executive branch, or continuing in the
Senate, an option that promised less power (she would still be
only one of a hundred) but greater autonomy. The other wrin-
kle was that most commentators assumed that taking an ad-
ministration job would preclude a primary challenge against
Barack Obama in 2012, and thus meant giving up on a life-
long dream to be president of the United States. Thus, Hillary
Clinton faced three possibilities: Remain in the Congress and
not win the presidency in 2012 (C), remain in the Congress
and win the presidency in 2012 (P), or join the administration
as secretary of state (S). State what you think Hillary Clin-
ton's preference ranking was at that time. If the probability of

winning the White House in 2012 if she had remained in the
Senate is p, then use an expected utility argument to deter-
mine the smallest p that would have induced Clinton to re-
main in the Senate in order to run in 2012. In your opinion,
did Hillary Clinton's decision make sense?
*6. Imagine that you are confronted with two pairs of lotteries
over the following three outcomes: x = $2.5 million, y = $.5 mil-
lion, and z ~ $0. The first pair pits P 1 against P 2 , where P 1 =
(p 1(x),p 1(y),p 1 (2)) .. (0, l, O) (i.e., you are certain to win
$500,000) and P 2 = (p 2 (x),p 2 (y),p 2 (z)) = (.10, .89, .01). The sec-
ond pair is a choice between P3 = (p3(x),p 3(y),p3(z)) = (O, .11,
.89) and P4 = (pix),p 4 (y),piz)) = (.10, 0, .90). Empirically, most
individuals express a strict preference for P 1 to P 2 , and P 4 to
P 3 . Is this behavior consistent with the theory of expected
utility? In order to solve this problem, rephrase each of the
expressed opinions in terms of expected utility (e.g., EU(Pa) =
. llu(x) + .89u(z)) and then use basic operations on the re-
sulting inequalities to see if a contradiction emerges. No
knowledge of t!he actual utility function is necessary to solve
this problem.
Part II
GROUP CHOICE
3
Getting Started with
Group Choice Analysis
AWARM-UP EXERCISE
Andrew, Bonnie, and Chuck are friends who have decided to
cut class on a pleasant spring afternoon in Boston. Andrew, an
intellectual snob, suggests going to see a fabulous collection of
Impressionist paintings at the M useum of F ine Arts. Bonnie,
a bit more political, wants to go down to the Boston Common
to attend a rally to raise funds and consciousness in support of
preserving Walden P ond, which is threatened by commercial
development. Chuck, a jock, thinks an afternoon at Fenway
P ark watching t he Red Sox would be just fine. Display 3.1
supplies each group member's rank order of the alternatives.
In terms of the notation in Chapter 2, Andrew's preference or-
dering is MFA P_4 WP PA RS. Similar expressions may be writ-
ten down for Bonnie and Chuck. Each member of the group,
{A, B , C}, has preferences over the alternatives, {MFA , WP,
RS}, satisfying completeness a nd tr ansitivity-properties 1
a nd 2 of Chapter 2.
It becomes e vident after the briefest of times that this
group of friends does not have an obvious course of action. It
suffers from a common group affli ction, an affliction with no
known cure short of authoritarian measures-preference
41
DISPLAY 3.1
A B c
MFA WP RS
WP RS WP
RS MFA MFA
diuersity. 1 If Andrew, Bonnie, and Chuck are going to hang out

together as a group, they must come to a choice in spite of their
heterogeneous preferences. Not realizing that the decision on
how a group should arrive at collective decisions is something
over which much blood has been spilled throughout human
history, our three friends rather casually decide to take "a" vote
and let "the" majority rule. (We put the indefinite and definite
articles of the previous sentence in quotation marks, since two
of the main points of this chapter are that there are many
ways to take a vote and there are many different majorities.)
Our group has a problem, it seems, because the three
friends do not unanimously share the same first preference; if
they did, there would be no problem. In the present circum-
stance, then, one commonsense way to "take a vote" is to poll
the group members on whether a majority shares a first pref-
erence in common. A quick inspection of the preference
orderings in Display 3.1 reveals that this method fails to re-
solve the group's problem. In terms of first preferences, our
group is heterogeneous in the extreme. As a fallback, then, a
more bothersome but still commonsensical plan for resolving
problems of group choice is to conduct a round-robin tourna-
ment. Each alternative is pitted against each other alternative
1 Although collections of individuals often consort together because of shared
interests, this does not mean that they share identical preferences on all
manner of things. Moreover, many collections of individuals of interest to po·
litical scientists consort together not because of shared interests, but more
often for exactly the opposite reason- they are representative bodies reflect·
ing the diversity of a larger population.
Getting Started with Group Choice Analysis 43
D ISPLAY 3.2
MFA vs. WP: WP wins 2-1 {B, C}

MFA vs. RS: RS wins 2-1 {B,C}
WP vs. RS: WP wins 2-1 {A,B}
and, if one is preferred by a majority to all the others, then it

is declared the group choice. (If this condition fails to materi-
alize, then we go back to the drawing board.) Consulting the
preference orderings above, we see that this tournament pro-
duces the results of Display 3.2.
Several things are wor th noting about this little exercise.
First of all, going to the Common for the Walden Pond political
r ally is the majority preference of this group: WP wins the
round-robin tournament by beating all the other alternatives
in pairwise contests. The vote in each case is 2-1, and the par-
ticular friends on the winning side in each instance are listed
in brackets in the display.
But notice that different majorities prefer WP to each of
the other alternatives; Bonnie is the only common member.
So, second, groups are composed of many majorities: {A,B}
{A,C} {B,C} and {A,B,C} are all majorities of our group of
friends. Letting "the" majority rule is not unambiguous and,
as we shall see, can get you into trouble.
Third, we have interpreted the taking of a vote here as
having each individual reveal his or her preference honestly.
When confronted with a pair of alternatives, each group mem-
ber voted for the one that he or she ranked h igher. This is
known as sincere prefer ence revelation. It is entirely possible,
of course, for an individual to vote contrary to preference, per-
haps because by doing so a person paradoxically makes out
better than if he or she had voted sincerely. This latter ma-
neuver, known as strategic or sophisticated preference revela-
tion, does seem a bit like cheating, especially among friends,
and involves some disingenuous, if not dishonest, behavior.
'
We will explore strategic behavior more systematically
in Chapter 6, but let's briefly and casually examine the
possibilities of strategic behavior in this example. What we
want to determine is whether any of the friends has an incen-
tive to misrepresent his or her preferences by voting strategi-
cally. A person might consider doing this if he or she could
produce a more preferred final outcome. In our example this
is a very real prospect because each outcome in the round-
robin was decided by a single vote. Thus, there are two things
to determine-feasibility and desirability: (1) Can someone
shift the outcome by shifting his or her vote? and (2) Given
the possibility, would such a person want to shift the out-
come?
It is clear that any one in the majority in each pairing
above could, by misrepresenting preferences, change the out-
come in that pairing. This answers the first question about
feasibility. As to desirability, one thing is clear: Bonnie has ab-
solutely no incentive to change her vote in any pairing since
she is the great beneficiary of sincere voting by members of
the group-her first preference wins the round-robin. Poor An-
drew is pivotal only in the comparison between Walden Pond
and Red Sox and, if he were to change his vote to favor RS,
then he would make the latter the overall winner of the round-
robin. But the latter is the worst thing for him, so he certainly
has no incentive to switch votes. This leaves Chuck as the
only one with a possible motive to behave strategically. It is
clear that he could change the result of the pairing between
MFA and RS, but this would not change the overall outcome
so there is not much point to his doing that. 2 However, what if
he voted against WP (his second preference) and for MFA (his
last preference) in the first pairing in Display 3.2? Then the
2 He could simultaneously change his votes in each of the first two ballots, in
which case MFA wo1Uld be the round-robin winner. But this is Chuck's least-
preferred outcome, so he hardly has an incentive to behave strategically in
this fashion.
round-r obin tournament would have no winner. Before we can

determine whe ther it is rational for Chuck to vote strategi-
cally in the firs t ballot above, we would need to know (and cer-
tainly he would need to know) what transpires in the event of
no winner (something I neglected to a rrange for in describing
this tournament above). 3 I shall not pursue this here, though I
promise to take it up more systematically in the chapter after
next. It has been sufficient to have demonstrated that one of
the friends may have an incentive to behave strategically.
Several lessons emerge from this little exercise. The first is
that there are multiple majorities, rather than "the" majority.
A second lesson is that there are multiple forms of preference
revelation-at the time of balloting a person may vote sin-
cerely or strategically. A third lesson of this example, one I
pursue a bit further now, is that there are multiple ways for
groups to decide by voting.
We have already seen that rule by unanimity is one way for
a group to proceed, but that in this particular case it fails to
produce a solut ion to our group's problem since preferences
are just too heterogeneous. Likewise, we have determined
tha t first-preference majority rule, in which each person votes
3 We actually have been purposeful in our "neglect," since this allows us to

point out that larger societies often neglect to arrange for alJ possible con-
tingencies in their constitutional deliberations. Sometimes the neglect is
done knowingly; some events are r egarded as so improbable or farfetched
that it simply isn't worth designing arrangements for them. onen, however,
the neglect is not by design, in which case a group may find it has to free-
lance. Kids playing baseball in a wooded area, for instance, add to the rules
of the game as unanticipated events occur, e.g., a ball hits an overhanging
branch. Grown-ups playing baseball, to take another example, add to the
rules of the game (for instance, establishing a "designated hitter" rule for
pitchers) in response to unanticipated market circumstances. (Major League
Baseball team owners, I am informed, worried thirty-some years ago that
baseball was getting too dull for the go-go American public; some of the
owners-those of the American League-decided to jazz the game up. This
incredibly stupid move, supported only by people ignorant of t he s ubtleties
of the game, and by the players' union in its effort to save the jobs of aging
sluggers, is a matter of great debate best left to another forum. Needless to
say, the author is a fan of the National League.)
sincerely for h is or her first preference, fails to resolve the

issue of what the group should do, and for the same reason-
too much preference diversity. Finally, we have discovered
that the preferences of individuals in the group do lend them-
selves to majority rule by round-robin tournament, at least as
long as everyone votes sincerely. Fortunately for our friends,
there were only three of them and only three alternatives so
that, given their preferences, a round-robin tournament was
both easy to administer and decisive. Had there been more
friends, more alternatives, or more diver se preferences, a
round-robin tournament may not have suited the needs of the
group as well as some alternative method. Indeed, there are
many ways in which to decide by voting. Often the institu-
tional features of the voting system will be absolutely essen-
tial in determining which alternative wins.
A R EVISED E XAMPLE
This will not be the last time you hear me say, "Institutions
matter." In this case I want to claim that the institutional pro-
cedure for conducting a vote to resolve a problem of group
choice dramatically affects that group choice. 1b show this, I
alter our "warm-up exercise" slightly. Suppose that Andrew,
Bonnie, and Chuck have the same preferences with which
they were endowed in Display 3.1. But now let us suppose
that, before any votes are taken, there is an intervening stage
of debate a nd deliberation (otherwise known as arguing).
While debate and deliberation may often seem like window
dressing, it is entirely possible that, from time to time at least,
some persuasion, reconsideration, conceivably even coercion,
takes place that r esults in someone changing preferences.
Thus, while our friends are deliberating, suppose Chuck
becomes convinced that a trip to the museum might be brief
and he might, at the very least, catch the last few innings of
D ISPLAY 3.3
A B c
MFA WP RS
WP RS MFA
RS MFA WP
the Red Sox game on the tube afterward. The Walden Pond
rally, on the other hand, would go on all afternoon. As a conse·
quence, suppose he elevates MFA in his preference ordering
(and lowers WP). Thus, before the voting but after delibera·
tion, the preferences of the group members now are as given
in Display 3.3. A round-robin majority rule tournament does
not produce a winner in this new situation (as shown in Dis-
play 3.4). Each alternative is beaten by one of the other alter-
natives: MFA loses to RS, which loses to WP, which loses to
MFA. If we were to write the group majority preference rela-
tion as Pc, then we would have RS Pc MFA Pc WP Pc RS.
But this doesn't look like a preference ordering at all. Indeed,
it is not! 4 Whereas the individuals of the group possessed
coherent-that is, transitive-preferences, the group does not.
The group preference relation is intransitive or, to put it more
colorfully, the group's preferences cycle, with a different ma-
jority coalition supporting the winner in each pairwise com-
panson.
Since round-robin voting doesn't solve our group's problem
in this circumstance, we need to think about other institu-
tional schemes. One arrange1nent that is found in official in-
stitutions is called the agenda procedure. 5 For a given set of
alternatives, some individual (or subcommittee)-called the
agenda setter-is charged with assembling an order of voting
• Notice that RS is simultaneously (!) at the top and bottom of this list-
something not characteristic of an ordering.
6 This procedure is also used in many sports competitions, under the rubric of
single-elimination tournament.
DISPLAY 8.4
Contest Winne r S upporte rs

MFA vs. WP MFA {A,C}
WP vs. RS WP {A, B}
RS vs. MFA RS {B,C}
for the larger group. The alternatives, themselves, are usually

generated by the larger group, although in some organizations
an especially powerful agenda setter both proposes items for
the agenda and puts them into a voting order. Once this
agenda is formed, the group votes on the items a pair at a
time. Specifically, the first two items on the agenda are voted
on by majority rule, with the losing item eliminated and the
winning item paired with the next item on the agenda. A ma-
jority vote between these two follows along the same lines.
This procedure is repeated as often as necessary to work
through the agenda. Whichever alternative survives the en-
tire gauntlet is th e winner.
Returning to our example, suppose Andrew, because he is
the oldest member of the group, is charged with proposing an
agenda. What are his options? Generally speaking, for a set of
k agenda items, there are (k) x (k - 1) x (k - 2) x ... x 3 x 2 x 1
ways to order an agenda. In the example, where k = 3, there
are 3 x 2 x 1, or six, orderings. But really there are only three
substantively distinct agendas, since what is really being cho-
sen is which of th e three alternatives goes last and which pair,
as a consequence, is voted on first. 6 So, Andrew may choose
one of the agendlas in Display 3.5. If he chose agenda I, for in-
stance, then MFA would be voted against WP with the winner
then voted against RS and the survivor declared the group
choice.
6 That is, in this circumstance it really doesn't matter which of the two items
in the initial pair is first and which is second.
DISPLAY 3.5
Agenda I Agenda II Agenda III
MFA RS WP
WP MFA RS
RS WP MFA
If Andrew knows his friends' preferences and is prepared

to believe that they will vote honestly once he chooses an
agenda (both of which we will assume here for the sake of ar-
gument), then he can actually figure out what will tran spire
for each agenda selection he makes. All he needs to do is con-
sult Display 3.4. If he chooses agenda I, then MFA is paired
against WP. From Display 3.4, MFA wins and advances to the
next round of voting, pitted aga]nst RS. RS prevails in this
pairing and thus in the entire contest, so Agenda I = RS. In a
similar fashion, he determines that agenda II = WP and
=
agenda III lv!FA. Thus, by choosing agenda III, Andrew can
produce his most-preferred alternative as the outcome of
group choice. Agenda power is powerful indeed! And the insti-
tutional norm that says, "If there is no round-robin winner,
then let the oldest in the group select an agenda," sure makes
a difference, too. Had the· norm given that power to the tallest
(Chuck) or the lightest (Bonnie), and each of them had gone
through th e same exercise, then agenda I and agenda II, re-
spectively, would have been chosen with altogether different
group choices.
SUMMARY
I've wandered a bit astray, but I hope the reader has been sen-
sitized to the fact that even when individuals honestly reveal
their preferences, it is nevertheless entirely possible for a
group's preferences to be "badly behaved" (read: intransitive)

in comparison to those of the individuals who comprise it. This
is an instance of what the political philosophers Brian Barry
and Russell Hardin call "rational man and irrational society."7
As a consequence, what is best for the group, or even what a
majority thinks is best for the group, is not at all evident.
Even more important, the precise institutional procedures by
which the group determines what it shall do are absolutely
critical in making that choice.
In the next several chapters I make these matters, and
more besides, a bit more precise. In Chapter 4 the focus will
be on the method of majority rule and the problem of group
preference intransitivity. Chapter 5 continues on this theme
by investigating the method of majority rule from the perspec-
tive of the spatial model of group choice. Chapter 6 turns at-
tention to the issue of manipulation, in terms of both the
misrepresentation of preferences and agenda stratagems. Fi-
nally, in Chapter 7 the theme of alternative ways for groups to
make choices is taken up.
P ROBLEMS AND D ISCUSSION Q UESTIONS
1. Five members of a committee are voting over four proposed

spending plans, abbreviated as A, B, C, and D (you can as-
sume that the committee members vote sincerely). Two of the
committee members have the preference ABCD (A is preferred
to B is preferred to C ... and so on). The other three have
preferences: BCDA, DBAC, and CBDA. Does any spending
plan win under a system of plurality rule, in which the out-
come with the most first-place votes prevails? Which plan
7 Rational Man and Irrational Society? (Beverly Hills, Calif.: Sage, 1982).
would win in a round-robin tournament? Suppose that the

committee members determine that there is a fatal flaw in
plan B, and only vote over A, C, and D. Would there be a clear
winner under the round-robin tournament method? Why or
why not?
2. Three individuals, i, j, and k, are voting over four outcomes

q, r, s, and t. Their preference orderings are
If they vote honestly using a round-robin tournament, are

there any group preference cycles? Now k changes his mind,
and switches his preferences to tPksPkrPkq. Are there any
group preference cycles? How does this problem illustrate the
idea of "rational man, irrational society?"
3. Using the same preferences as above (before k changes his

mind), can i fashion a sequential agenda (i.e., each outcome
introduced sequentially and only retained if it beats the exist-
ing winner) such that her top choice wins? What about after k
changes his mind? Then, identify agendas that j and k could
design to secure their top choices (for both before and after k
changes his mind, if possible). In general, can an agenda be
fashioned that leads to the defeat of an alternative favored by
a majority over each other alternative?
*4. Now using the preferences after k has changed his mind,
show that if player j proposes the agenda tsrq,a then player i
has an incentive to strategically misrepresent her vote if she
assumes the otlhers vote honestly. Now suppose k proposes the
• First t and s are voted on, then the winner of that contest faces r, then the
winner of that contest faces q. The winner of that final contest is the overall
victor.
agenda rqst. Can j do better than the expected outcome under

honest voting by misrepresenting his preferences?
5. In Chapter 2, we argued that complete and transitive pref-

erences at the individual level were a fundamental part of a
rational-actor model of politics. Why are transitive prefer-
ences so important? Are transitive preferences similarly im-
portant at the level of group decision making? Why or why
not?
4
Group Choice and Majority Rule
In the previous chapter I introduced three subjects, though

only in the most casual of fashions: cycling majority prefer-
ences, manipulation of agendas and of the way preferences
are revealed, and alternative voting methods for making
group choices. In the next several chapters I take these sub-
jects up one at a time and in considerably more detail. By the
end of our intellectual tour, I hope the reader will have come
to appreciate (if not admire) the expertise politicians must ac-
quire in order to master the arcane procedural details of group
choice.
CYCLICAL MAJORITIES
Condorcet's Paradox
We saw in the last chapter that a group of rational individuals
can collectively produce irrational results. Even though each
individual in the group has preferences that are consistent
(complete and tr ansitive), this need not be true of the group's
preferences. This puzzle has come to be known as Condorcet's
paradox, named after the eminent scientist, philosopher, and
mathematician of late eighteenth-century France who (re)dis-
53
D ISPLAY 4 .1
CYCLICAL MA.JORITY PnEFE REN CES
1 2 3
a b c
b c a
c a b
covered it. 1 Although not actually a paradox in the strict logi-

cal sense, the disjuncture between group preferences and the
preferences of individuals always seems to surprise and puz-
zle students when they first encounter it. Its general format is
given in Display 4.1, where a group G ~ {1, 2, 3} must choose
by majority rule from among the three alternatives, {a, b, c},
which could be political candidates, public policies, or places
to go in Boston on a sunny spring afternoon. A majority, {l, 3},
prefers a to b; another majority, {l, 2}, prefers b to c; but (con-
trary to trans itivity) still another majority, {2, 3}, pr efers c to
a. For members of a group with these preferences, majority
rule produces a lternatives that are said to cycle. More for-
mally, a group pre ference relation, P0 , is said to be cyclical if it
violates transitivity (property 2 of Chapter 2). In the example
given in Display 4.1 ,
which violates transitivity. (A transitive group preferring a to

b and b to c would prefer a to c.)
1 Por the longest time, Condorcet was credited with inventing this voting para-
dox. Only recently it has come to light that, in fact, he had rediscovered
something that had been known five hundred year s earlier. For a general
historical overview of this subject, see fain McLean and Arnold B. Urken ,
eds., Classics of Social Choice (Ann Arbor: University of Michigan Press,
1993).
Group Choice and Majority Rule 55
DISPLAY 4.2
ALTERNATIVE PREFERENCE ORDERINGS
OF THREE ALTERNATIVES
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
a a b b e e ab ae be a b e
b e a c a b abe
e b e a b a e b a be ac ab
This raises both a normative and a positive question-

what should the group G do? and what will the group G do? In
more general contexts, in which G is a legislature, a town
meeting, or indeed possibly an entire electorate or society,
these questions take on a broad significance.
However, before we get up a head of steam on ''broadly sig-
nificant" questions, and with all due respect to Monsieur Con-
dorcet, a prior question naturally arises: just how important is
this puzzle of gr oup intransitivity? Is it merely an arcane log-
ical possibility, a trick foisted upon the unknowing student by
professors, philosophers, and textbook writers? Or is it a pro-
found discovery, the stuff from which important insights about
political philosophy and social life are made? In my opinion,
the answer lies much closer to the latter.
The general issues raised in the preceding paragraph may
be approached by thinking first about the likelihood of Con-
dorcet's paradox in the simplest of all settings, namely, the
three individuals and three alternatives given in Display 4.1.
Any one individual may rank order the three alternatives in
thirteen different ways (see Display 4.2). Preference orderings
(1) through (6) involve no indifference and are said to be strong.
Orderings (7) through (12) involve some indifference, while
(13) represents total indifference; these latter orderings are
weak. Each member of the group thus may adopt any one of
thirteen orderings, so that there are 13 x 13 x 13, or 2197, com-
binations of three individuals with preferences over three alter·
DISPLAY 4.3
ANOTHER SBT OF C YCLI CAL M AJORITY P REFEHENCES
l 2 3
c a b
b c a
a b c
natives. That is, there are 2197 different "societies." 'lb keep
things simpler still, let's focus on the 6 x 6 x 6, or 216, societies
of three persons with strong preferences (preference ordering
1 through 6 in Display 4.2). It is now possible to calculate how
many of these societies are afflicted with Condorcet's paradox.
Notice in Display 4.1 that the cyclical group preferences are
produced by a situation in which each alternative is ranked
first by exactly one person, second by exactly one person, and
third by exactly one person. This produces the "forward cycle"
a Pc b P0 c P0 a. (Recall that Pc means "is preferred by a ma-
jority of the group.") There are actually six different ways to
produce this forward cycle, since Mr. l's preference ordering in
Display 4.1 could be held by any one of the three individuals,
Ms. 2's by any one of the two remaining, and Mr. 3's by who-
ever is left. There arc also six ways to produce the "backward
cycle," c P0 b P0 a P0 c, generated by the individual orderings
given in Display 4.3, as well as by any reassignment of them
among group members. 2 So, taking forward and backward cy-
cles together, there are exactly 12 of 216 (strong-preference)
societies that generate group preference cycles. 3
2 Display 4.3 is Display 4.1 with everyone's preference ordering reversed.

'1 I am claiming here that only the preferences given in Displays 4.1 and 4.3,
and their reassignments, produce group preference cycles. These twelve
preference configurations are the only ones in which each alternative ap·
pears exactly once at each rank level. lt is relatively easy to show that if any
alternative showi:; up at the same rank for more than one individual, then
lhe group preference> will not cycle. The l'Cnder may like to try his or her
hand at establishing this fact.
So we now know that in the world of three-person groups

choosing a mong three alternatives, the odds are extremely
good that majority rule will work smoothly. In 204 of the
216 possible configurations there will be a Condorcet winner
rather than a Condorcet cycle.4 If each of the preference con-
figurations (societies) that cause majority rule cycles is no
more likely than each of the 204 for which majority rule works
smoothly, then majority rule will generate consistent group
preferences most of the time. In short, majority rule is nor-
mally a splendid way to conduct business, when the number of
individuals is small and the alternatives ar e few.
If only life were so simple! Alas, it is not. And as soon as we
start complicating things, the case fo r majority rule needs re-
examination. There ar e r eally only two ways in which to make
this pure majority rule setting more complicated: increase the
number of individuals (n) or increase the number of alterna-
tives (m). (Of course, we can increase both at the same time,
too.) We are interested in deriving a probability or proportion
that gives the 1ikelihood of a majority rule preference cycle,
given the number m of alternatives and the number n of indi-
viduals in the group. We write this probability as Pr(m,n). We
already know, for example, that Pr(3,3) = 12/216, or .056. Gen-
erally,
Pr(m,n) - [#of "problem" preference confi.gurations]/(m!)0
This formula states that the probability of intransitivity in the
majority preferences of a group of size n voting on m alterna-
tives is the ratio of t wo numbers. The numerator is the num-
ber of "societies" with cycling group preferences, like those
presented in Displays 4.1 and 4.3. This number is 12 for m = 3
and n = 3, as we saw in the discussion surrounding those
4 A Condorcet winner is the alternative that can defeat all others in pairwise
majority contests. If three alternatives do not cycle, then it must be the case
that one of them is a Condorcet winner. I t is the one at th e top of the major·
ity preference relation, P0 .
TABLE 4.1
P ROBABILITY OF A CYCLICAL MAJORITY, P r(m,n)
Number of Voters (n)
Number of 3 5 7 9 11 limit
Alternatives (m)
3 .056 .069 .075 .078 .080 .088
4 .111 .139 .150 .156 .160 .176
5 .160 .200 .215 .251
6 .202 .315
limit ::::1.000 ::::1.000 :::1.000 :::::1.000 :::1.000 :::::1.000

SOURCE: William H. Riker, Liberalism Against Populism (San Francisco: Free·
man, 1982), p. 122
dis plays. The denomina tor gives the total number of possible
societies, computed as follows: With m alternatives, any in-
dividual in the group may choose any one of m x (m - 1) x
(m - 2) x ... x 3 x ~ x 1 (or m! in mathematical symbols) dif-
ferent ways to order his or her preferences over the alterna-
tives. Since there are n individuals, this means there are
m! x m! x ... x m! (n times), or (m!) 0 different societies. This
number, form= n = 3, is (3 x 2 x 1) 3 • 6 x 6 x 6 = 216, as we also
saw in the discussion of Display 4.2. Fortunately for us, com-
putationally talented scholars have determined Pr(m ,n) from
the formula given above for various values of m and n. A par-
tial summary of their calculations is presented in Table 4.1.5
The columns of this table give groups of different sizes,
ranging from three to some extremely large number (which I
call the "limit"). The rows give sets of alternatives of different
sizes, again ranging from three to a limiting (very large) num-
5 A very accessible discussion of this entire subject is found in William H .

Riker, Liberalism Against Populism (San Francisco: Freeman, 1982). Chap-
ter 5. The literature on estimating Pr(m,n) is cited t here in footnote 3.
ber. The entries of the table give the probability that majority
preferences cycle, Pr(m ,n). Thus, if we look at the first row
(m - 3 alternatives), the probability of a cyclical majority rises
slowly from the 12/216 - .056 computed above for three-
member groups to .088 in the limit as the number of group
members becomes very large. Increase the number of alterna-
tives to four and the probability of a cyclical majority roughly
doubles for each group size; that is, it starts at a higher level
and smoothly increases to a limiting probability of .176. As the
number of alternatives grows very large, Pr(m,n) approaches
1.0-it becomes nearly certain that there will be preference
cycles among majorities.
So, the good news of the s mall-group/few-alternatives situ-
ation does not extend to more general situations. As the num-
ber of group members increases, and especially as the number
of alternatives increases, the probability of badly behaved
majority preferences-that is, cycles-grows, becoming nearly
certain as we approach the limit. In general, then, we cannot
rely on the method of majority rule to produce a coherent6
sense of what the group wants, especially if there ar e no
institutional mechanisms for keeping participation restricted
(thereby keeping n small) or weeding out some of the alterna-
tives (thereby keeping m s mall).
This is a troubling state of affairs for anyone trying to ana-
lyze politics. We have just concluded that, most of the time,
when we employ majority rule, we must tolerate either group
incoherence, a highly compressed franchise (small n), or
highly restricted agenda access (small m). That is the gist of
Table 4.1. There is, however, an important qualification.
In computing the entries of Table 4.1, a very specific as-
sumption was made about the likelihood of various preference
configurations. We assumed that, for any size group (n), each
6 'l'hroughout this text, we shall use "coherent," "consistent," and "transitive"

interchangeably.
of the strong preference orderings is as likely as any other to

characterize the preferences of an individual. Moreover, one
person's "selection" of a preference ordering is entirely inde-
pendent of some other person's. Thus, for m = 5 for example,
there are 5 x 4 x 3 x 2 x 1 = 120 ways to strongly order the five
alternatives. Each of the group members is assumed to have
his or her preferences represented by any one of these 120
preference orderings with equal probability. Thus, for n = 7
persons for example, there are (120) equally likely seven-
person societies. And, as the appropriate entry in Table 4.1
reveals (the one form= 5 and n • 7), 21.5 percent of these so-
cieties generate cyclical majority preferences. That is, more
than one time out of every five, a group of seven choosing
among five alternatives by majority voting will produce group
incoherence.
However, almost any real conception of society is bound
to be more all-embracing than a collection of equiprobable
preference-orderings; indeed, most conceptions of society em-
phasize interdependence rather than independence among in-
dividuals. Individuals often choose to join groups, for example,
precisely because they have preferences in common with other
group members . 1'his would lead one to expect correlation, not
independence, between the preferences of group members.
The conception of society as a collection of independently cho-
sen preference orderings provides no more than a baseline as-
sessment of majority-rule methods. So, our concerns about
cycles in majority rule as reflected in Table 4.1 are probably
exaggerated- but only somewhat. Even if the feasib le "soci-
eties" are not equally likely, as long as either nor (especially)
m is large, the odds of majority preferences cycling is suffi.
ciently large to be of concern. Moreover, in other circum-
stances quite common in politics-circumstances described
momentarily as "distributive politics"-rnajority cycles are in-
evitable. Condorcet's paradox cannot be lightly dismissed. I
Group Choice and Majority R ule 61
hammer these points home in concluding this section, first

with one last abstract example and then with a couple of real-
world cases of cyclical majorities.
Cyclical Majorities and ''Divide the Dollars"

So many interesting political fights involve a group deciding
how to share something. Usually it is something desirable and
the fighting is about getting the most favorable distribution.
Manna from heaven is wonderful, but how should it be divided
up? The same logic, however, applies to things people want to
avoid. For the past decade or so, reducing the federal deficit
has been of great political import in the United States. Every-
one agrees that hundreds of billions of dollars of public expen-
ditures need to be cut (or revenues raised), but from where
and from whom? Divvying up program cuts or tax burdens,
just like sharing the revenues from newly discovered oil or
some other windfall, involves group conflict over distribution.
Suppose a small town has lucked into a windfall of $1,000,
because the state had made an earlier error of overcollecting
fees from the town. The town's three-person board of select-
men 7 must decide how to spend this "found" money, this
manna from heaven. The politicians on the board represent
the E ast, Central, and West districts of town, respectively,
and, like most representatives, they want to hang on to their
jobs by taking care of their respective constituents. By this it
is meant that each politician- whom we shall name E, C, and
W- believes that the more money he or she can land for the
district, the better his or her chances are for reelection. The
board operates by simple majority rule. Is there a division of
the spoils that a board majority prefers to any other division?
7 "Boardof selectmen" seems to be a New England thing. In other parts of the
United States, the local legislature is ca lied a city or town council or board of
aldermen.
That is, is there a Condorcet-winning sharing scheme or, in-

stead, do the sharing schemes cycle?
Let us write a share of the $1000 for each district as s(E),
s(C), and s(W), respectively. A sharing scheme, [s(E), s(C),
s(W)], is feasible if (1) each of its components is nonnegative
(you can't give one of the districts a negative share of $1000)-
s(E), s(C), s(W) 2' O; and (2) if the components sum to no more
than $1000-s(E) + s(C) + s(Vl) s $1000. I n this feasible set,
(1000, 0, O) is the most-preferred distribution for E, (0, 1000,
0) is most-preferred by C, and (0, 0, 1000) is W's first choice.
Generally speaking, a selectman prefers one distribution to
another if and only if his or her component is larger in the one
distribution than in the other.
My claim is that "divide the dollars" is a game that pro-
duces cyclical majorities. No distribution is preferred by
a majority to every other distribution; there is no Condorcet
winner. Before showing this generally, let's consider some
cases. Consider first what many would consider the fair distri-
bution-(333'13, 333'13, 3331/J]. A majority consisting of E and C
prefer (500, 500, O] to it; another majority consisting of E and
W prefer (500, 0, 500) to it; and, finally, the coalition of C
and W prefer [O, 500, 500] to it. Jn each of these instances two
of the three selectmen do better than with the fair distribution
in the sense that they bring home more revenue to their con-
stituents. The fair distribution is thus vulnerable to some
majority coalition of selectmen ganging up on the excluded se-
lectman.
But what about these latter distributions? It turns out that
they, too. are vulnerable. And the distributions to which they
are vulnerable are also vulnerable. In fact, majority prefer-
ences over various distributions cycle. To see this, consider the
distribution [500, 500, O] that E and C prefer to the fair distri-
bution (since 500 > 3331/J for both E and C). Against [500, 500,
OJ, E and W prefer [700, 0, 300] (since 700 > 500 for E and 300
> 0 for W); and then C and W favor the fair distribution to
DISPLAY 4.4
MAJORITY CYCLE I N "DMDE THE DOLLARS" GAME
Majority Coalition
Distribution 1 Distribution 2 Preferring 2 to 1
[3331/J, 3331/3, 3331/3] (500, 500, 0) {E,C}
[500, 500, OJ (700, 0, 300) {E,W}
(700, o. 300) [333'h, 333'h, 333'A] {C,W}
[700, 0, 300] (since 333V;i > 0 for C and 333V;i > 300 for W),
thereby producing a majority cycle (Display 4.4).8
I have claimed that "divide the dollars" represents a
generic kind of politics. I have just proved that sharing out
benefits and burdens, or what is known as "distributive poli-
tics,"9 is inherently cyclical in majoritarian settings. Any
proposed distribution is open to amendment as different ma-
jorities jostle with one another for advantage. Final outcomes,
whatever they happen to be, are extremely sensitive to other
institutional features of the group decision-making setting.
They will depend, for example, on someone's exercising
agenda power, on some arbitrary time limit on deliberation,
8 Generally speaking, consider an arbitrary feasible distribution, d • [x, y, z].

This can be any distribution whose components are greater than zero and
which sum to no more than $1000. Take two small positive amounts, II a nd E
(where ll + E s z), and reallocate them away from W to E a nd C. This new dis-
tribution, <£ - [x + ll, y + e, z - II - e], is also feasible as long as we don't take
too much a way from W (which is what the inequality in the parenthesis
above guarantees). Now, since II is positive, E prefers d' to d and, since e is
positive, C prefe:ra d: to d. (Of course, for exactly these same reasons, W
prefers d to d'.) Since a majority of the selectmen prefer d: to d, the latter is
not a Condorcet-winning sharing scheme. But d was an arbitrary scheme. It
could have been any distribution and the same logic would have applied. So,
what we have proved is that no sharing scheme is a Condorcet winner, since
a ny scheme is subject to reallocations like the one constructed above.
9 This term was popularized by Theodore J. Lowi in his classic paper "Ameri·
can Business, Puhlic Policy, Case-Studies, and Political Science," World Pol-
itics (July 1964): 677- 715, and developed further in his book, 1'he End of
Liberalism (New York: Norton, 1969).
on procedural features like who is permitted to make motions,

or whatever. In su m, in this very important class of political
activity, the only way to avoid preference cycles like the one in
Display 4.4 is to impose some form of antimajoritarian restric-
tion. This is the principal content of Arrow's Theorem, the sub-
ject of the next section. Before turning to that, I illustrate my
main point one more time with some illustrations drawn from
American political history (Case 4.1).
CASE 4.1
CIVIL W AR T AXES, G REAT D EPRESSION
T AXES, 1980s T AX R EFORM
Tax politics is distributive politics par excellence; it is cer-

tainly a good example of "divide the dollars." Various social
groups want to avoid paying more taxes, and this leads to
unstable coalition s and preference cycles. The question to be
asked, then, is how we in the United States have been able
to pass tax reform legislation at all (reforms either to in-
crease taxes, decrease taxes, or redistribute the burden). In
this case I have chosen three episodes to demonstrate the
dynamics of preference cycles as they play out in the U.S.
Congress. In the first episode, uncertainty played a critical
role. In the second, institutional features restricted the abil-
ity of groups to offer amendments to a proposed bill.* In the
final episode, organizational problems prevented a coalition
from uniting to block proposed legislation.t
The very first income tax in (what remained of) the
·The institutional features emphasized here are procedural rules that are
common in most legislatures. They will be discussed more systematically
in the next chapter.
t The organizational problems, known as collective action problems, will be
more fully analyzed in Chapter 9.
United States was passed in 1861 as part of the effort to

finance the Civil War. The interesting thing about the in-
come tax is that it resulted from a preference cycle (and
was not even part of the original suggestion for raising rev-
enue). There was a motion in the House of Representatives
to raise federal revenue by taxing wealth. To this motion
was offered an amendment to raise revenue instead by tax-
ing land. And, of course, if neither the original motion nor
the amended version passed, the status quo of no taxes
would prevail. Different majorities preferred the land tax to
the wealth tax, the wealth tax to no tax, and no tax to the
land tax. There was much confusion and to-and-fro during
the debate as t his majority preference cycle wreaked havoc.
Finally, someone introduced the idea of taxing income. This
swept to victory prin1arily because, unlike each of the other
taxes (in which politicians knew exactly whose ox would be
gored), there was much uncertainty about how an income
tax would impact various constituencies. Politicians pre-
ferred the "lottery" of an income tax to no tax at all or a tax
on either land or wealth.:j:
A legislative bill to raise, lower, or redistribute taxes is a
proposal to alter the status quo in some fashion. One way to
prevent preference cycles is to restrict the right of anyone
to amend a proposal. Then there are, in effect, only two al-
ternatives-the proposal and the status quo-and a major-
ity prefers either the proposal to the status quo or the
status quo to the proposal, or there is a tie (in which case,
by convention in nearly all legislatures, the status quo pre-
vails). At various times throughout its history, and espe-
cially in the first half of the 20th century, Congress has
restricted the rights of legislators to offer amendments to
tax bills. When it has not, preference cycles often emerged.
t The details are provided in the fine essay by James E. Alt, '"l'he Evolution
of Tax Structures," Public Choice 41 (1983): 181-223.
The Revenue Acts of 1932 and 1938, for example, were

pieces of legislation in which Congress did permit members
to amend the legislation on the floor, and scholars have
identified the majority preference cycles that resulted from
this activity. From experiences like these, members of Con-
gress have often agreed in advance to impose institutional
restrictions on one another's legislative rights-in this in-
stance, a restriction on the right to offer amendments-in
order to avoid preference cycles.§
The Tax Reform Act of 1986 provides still another
episode in which preference cycles were overcome. What
happened? Against the status quo of doing nothing-always
an option, of course-Senator Bill Bradley (D-N.J.) took ad-
vantage of various "supply-side economics" arguments to
fashion a proposal that attracted a majority from both par-
ties. As initially introduced by Bradley, the bill eliminated
hundreds of billions of dollars in tax breaks in return for a
lower individual tax rate. At this point it was expected that
so-called special interests (whose tax breaks were being
eliminated) would band together to offer a proposal (essen-
tially an amended bill) that would woo Republicans away
from the Bradl,e y bill. This proposal, while protecting
special-interest tax breaks and "bribing'' Republicans in
various ways, would nevertheless be sufficiently offensive
to the majority that they would rather have no bill at all.
That is, the so-called special-interest bill would defeat the
Bradley proposal but then would itself be defeated by the
status quo. The result would be no bill at all, even though
the Bradley bill was preferred to the status quo. In short,
the special-interest proposal would generate a preference
cycle that would have the effect of killing tax reform alto-
§The argument that Depression-era revenue bills were haunted by major-

ity preference cycles is found in John C. Blydenburgh, "The Closed Rule
and the Paradox of Voting," Journal of Politics 33 (1971): 57 71.
gether. But this special-interest proposal did not material-

ize. Like hogs around a trough, the individual groups in the
special-interest camp were so focused on preserving their
own tax breaks that they failed to coalesce to produce the
proposal to split the Bradley coalition. Deputy Treasury
Secretary Rich ard Darinan, the Reagan administration's
tax reform strategist, had said at the time that an organ-
ized effort by the special interests would have been a "killer
coalition." Darman went on to say that the lobbyists were
''brought down by the narrowness of their vision. Precisely
because they defined themselves as representatives of sin-
gle special interests, they failed to notice their collective
power."**
~The politics of the 1986 tax reform process, and the specific quote from
Darman, are found in Alan S. Murray and Jeffrey H. Birnbaum, "Law·
makers, Lobbyists and the Unlikely Triumph of Tax Reform," Congress
and the Presidency 15 (1988).
AR.Row's T HEOREM
I have looked both theoretically and practically at the puzzle
of Condorcet's paradox. The choices of rational individuals
(based on complete and transitive preferences) do not neces-
sar ily tran slate through majority voting into a well-defined
group preference. Condorcet's paradox is problematic for ma-
jority rule in theory because its probability in natural groups
is not trivial and grows both with the size of the group and
with the number of possible alternatives for choice. It is prob-
lematic for majority voting in practice because in very r eal po-
litical settings, especially those dealing with the distribution
of a fixed pie, voting cycles do emer ge.
Nevertheless, it may be that we aren't looking at the prob-
lem properly. Maybe the problem of group incoherence is a pe-
culiarity of round-robin tournaments, or of specific features of
majority rule, but not of voting more generally. If we employed

some alternative way of arriving at a group choice, the prob-
lem might be less severe, or might possibly even go away
altogether. That is, it might be possible to overcome the
theoretical and p ractical problems of group incoherence by
structuring the institutional arrangements of group choice dif-
ferently. Arrow's Theorem, one of the most profound insights
of twentieth-century political thought (and one that won its
originator a Nobel Prize in economic sciences), indicates that
these hunches, these hopes that the problem of group incoher-
ence will go away if only we think about it in the right way,
are wrong. Arrow's Theorem asserts that Condorcet's paradox
is a problem for any reasonable method of aggregating indi-
vidual preferences into group preferences.
Arrows Theorem: Assumptions
Kenneth Arrow, in his seminal Social Choice and Individual
Values ,to assembled a set of general conditions that he claimed
any reasonable method for aggregating preferences would sat-
isfy. He did not equate these conditions with "reasonableness."
Rather, in a much more powerful argument, he suggested only
that his conditions were minimal; any reader is free to add ad-
ditional reasonable requirements that the method by which a
group makes choices should satisfy. Arrow concludes that no
method of aggregating individual preferences into a coherent
group preference can simultaneously satisfy even his minimal
conditions. 11
Arrow sets the problem up in an abstract fashion. There is
a group of individuals, G = {1, 2, ... , n}, where n is at least
three. (Rather than naming group members, I shall simply
call them by numbers.) There is also a set of alternatives, A •
10 (New York: Wiley, 1951).

11 Therefore, the conclusion extends to any collection of conditions beyond the
minimal ones proposed by Arrow: If no method satisfies the minimal condi-
tions, then certainly it will not satisfy any expanded set of conditions.
{1, 2, ... , m} where m is also at least three. A typical individ-

ual is called Mr. or Ms. i. Likewise, alternative h or j or k iden-
tifies a typical alternative. All of this will be apparent in
context. 12 The individuals in G are assumed to possess prefer-
ences over the alternatives of A, written R; for i E G, which
satisfy:
Rationality assumption. R ; is complete and transitive.
This is just property 1 and property 2 from Chapter 2. Com-
pleteness is the assumption that Mr. i is capable of saying, for
any pair of alternatives, which he prefers (or that he is indif-
ferent between them). Transitivity is the requirement that Mr.
i's preferences are coherent-that if he prefers h to j andj to k,
then he prefers h to k .
To the rationality assumption for group members, Arrow
adds four conditions to be taken as minimal requirements for
the method by which the group makes choices:
Condition U (Universal Doma in). Each i E G may
adopt any strong or weak complete and transitive prefer-
ence ordering over the alternatives in A .
Condition P (Pareto Optimality or Unanimity) . If
every member of G prefers j to k (or is indifferent between
them), then the group preference must reflect a preference
for j over k •(or an indifference between them).
Condition I (Independenc e of Irrelevant Alte rna-
tive s). If alternatives j and k stand in a particular re-
lationship to one another in each group member's
preferences, and this relationship does not change, then
neither may the group preference between j and k. This is
12 For the reader unfamiliar with mathematical notation, I will occasionally

make use of the set-theoretic symbol E, which means "i.s an element of."
Thus, i E G means "Mr. or Ms. i is a member of group G." Similarly, h, j, k
E A means "h, j, and k' are all alternatives in A "
true even if individual preferences over other (irrelevant)

alternatives in A change.
Condition D (Nondictatorship). There is no distin-
guished individual i* E G whose own preferences dictate
the group preference, independent of the other members
of G.
Condition U makes sense if we are interested in designing
a mechanism of group choice that responds to the preferences
of group members. Although we assume that the individuals
in G are rational, we do not want to restrict their preferences
in any other manner. Rather, we want the mechanism to be
universally applicable. Thus, if A= {a, b, c}, then Ms. i may se·
lect as her preferences any one of the thirteen preference or·
derings given in Display 4.2.
The rationale for condition Pis also driven by a concern for
linking group preference to individual preferences. Surely if
the group preference ordering ranked alternative k ahead of
alternative j, even though euery member of G had the opposite
preference, then we would be correct in describing the choice
mechanism as perverse. It would certainly be the case that
the group preference ordering was not a "positive" reflection of
individual preferences, and this Arrow wanted to prevent.
Condition I states that the relative positions of alterna-
tives j and k in the group preference ordering should depend
only upon their relative positions in individual orderings.
Suppose an expert group of American historians in 2010
sought to rank American presidents. Some in this group
ranked Thomas Jefferson ahead of Andrew Jackson while oth-
ers had the opposite view. The decision rule-whatever it
happened to b~ombined these various views into a group
preference, say for Jefferson over Jackson. Suppose all ini-
tially had Barack Obama below these two. Suppose, however,
that Obama's brilliant leadership during the financial crisis of
2009 caused some members of .the group to elevate Obama in
their respective preference orderings. Condition I states that

it still should be the group's assessment that J efferson ranks
ahead of J ackson- that in the comparison between J efferson
and Jackson, the group's (changed) assessment of Obama is ir-
relevant and therefore should not affect this comparison.
Condition D is an extremely minimal fairness condition. It
says that if j is preferred to k by some specific person-say
Ms. i*- and if k is preferred to j by everyone else, then it can-
not be that the group preference is j P0 k. There cannot be
some privileged person in the group (Ms. i*) whose prefer-
ences become the group's preferences, no matter what the
preferences of the other members of the group, even if this
person is an expert, a philosopher-king, or a megalomaniac.
As I noted above, these four conditions are rather sparse
and minimalist. There is a range of other things, both proce-
dural and substantive, that many, thinkers would want to in-
elude as additional "reasonable" conditions on the mechanism
for group choi~. As well, the four conditions are stated in an
especially weak form. For example, condition D only precludes
the most extreme form of dictator--0ne who gets her way
against any opposition; it does not preclude weaker forms of
social and political inequality--0ligarchies, power elites, ex-
clusive committees, and so on. The theorem I am about to
state, as applied to the minimalist conditions, will apply with
a vengeance to any more elaborate set of conditions one might
want to impose on the method for making group decisions.
The Theorem and Its Meaning

I have emphasized how weak, reasonable, minimalist, and
sparse the Arrc;,w conditions are because Arrow's Theorem is of
the "even if" form: Even if we insist only on conditions as
weak, reasonable, minimalist, and sparse as these, something
horrible may still happen. The theorem, known as an impossi-
bility result, follows:
Arrow's The orem. There exist s no mechanis m for translat-

ing the prefe rences of rational individuals into a coh erent
group preference that s imultaneous ly satis fie s conditions
U, P, I, a nd D.
Put more dramatically, any scheme for producing a group

choice that satisfies U, P, and I is either dictatorial or incoher-
ent: the group is either dominated by a single distinguished
me mber or has intransitive preferences. This restates the
Arrow Impossibility Theorem in terms of the great trade-off it
implies: There is, in social life, a trade-off between social ra-
tionality and the concentration of power. Social organizations
that concentrate power provide for the prospect of social
coherence: the dictator knows her own mind and can act ra-
tionally in pu rsuit of whatever it is she prefers. Social organi-
zations in which power is dispersed, on the other hand, have
less promising prospects for making coherent choices. Though
these organizations may appear fairer and more democratic to
the person in the street, they may also be more likely to be
tongue-tied or inconsistent in ordering the alternatives under
consideration. 13
Does this mean that a ny pa rticular mechanism for aggre-
gating preferences is always either inconsistent or unfair?
Absolutely not. Earlier, for example, we saw that in the three-
voter/three-alternative situation, the method of majority rule
yielded coherent g roup preferences in 204 of 216 configura-
tions. It is easy to see (and I shall show this more formally
below) that majority rule satisfies conditions U, P, I , and D. It
is just that this 1nethod cannot guarantee group coherence in
all situations (as the twelve "troublesome" configurations give
testimony to). That is, the Rationality Assumption is violated
on some occas10ns.
rn An alternative interpretation of the trade-off is that we cannot insist on

coherence universally- that the Rationality Assumption and condition U
are in tension. As we shall discuss later in this chapter, the te ns ion goes
away if we relax our insistence on a universal domain.
Perhaps, as discussed earlier, we are overracting to this

problem. The doubtful reader might say, "Sure, coherence and
fairness in preference aggregation cannot be guaranteed, but
perhaps this conflict only arises occasionally. Nothing is per-
fect, after all." This is an overly optimistic view. The very fact
that some social situations produce either incoherence or un-
fairness means that it will be possible for clever, manipula-
tive, strategic individuals to exploit this fact.
Before turning to an illustration of how this theorem
should affect the way we think about social life, let me make
one last abstract observation. Although I have claimed that
Arrow's conditions of reasonableness are rather weak and un-
exceptional (and that, if anything, one might wish to impose
additional and more demanding conditions on the method for
producing group preferences), some may claim that Arrow's
conditions are already too demanding. That's fair enough. But
I can report that in the sixty years since Arrow's Theorem first
appeared, social choice theory (as the field created by Arrow
has come to be called) has become something of an academic
light industry. Somewhere close to 10,000 books and articles
have been written on Arrow's Theorem. 14 Scores of new
theorems-variations on Arrow's original result-have been
proffered in which 'o ne or more of Arrow's conditions has been
weakened or altered. 15 Short of actually eliminating one of the
fairness conditions-for example, by permitting dictators- the
Arrow result does not evaporate. Fairness and consistency in
decision makin;g by social groups must be traded off. This
14
Google Scholar r~ported in mid-2009 that Arrow's book had been cited nearly
7,500 times. Two other books on closely related themes had each been cited
nearly 2,500 times. See Amartya Sen, Collective Choice and Social Welfare
(San Francisco: Holden-Day, 1970), and Duncan Black, The Theory of Com-
mittees and Elections (Cambridge: Cambridge University Press, 1958).
16
Several surveys cover this broad literature. An accessible point of entry is
provided by Riker, Liberalism Against Populi.sm , especially Chapter 5. The
more advanced reader may consult Jerry S. Kelly, Arrow Impossibility The-
orems (New York: Academic Press, 1978) in addition to those cited in the
previous footnote.
forces us to think about political life in new ways, as Case 4.2

illustrates.
CASE 4.2
L EGISLATIVE I NTENT
A piece of legislation cannot possibly cover all conceivable

contingencies for which it might be relevant. So, in any
specific instance a bureaucrat, judge, or lawyer must deter-
mine whether a specific statute is applicable in a given
situation. Because differences of opinion can arise over
whether or in what manner a statute is applicable, the ap-
pellate courts a:rre often called upon to render a judgment.
Judges, lawyers, and legal scholars often give priority to the
following question in making this determination: What did
Congress intend in passing this law? In discovering con-
gressional intent , appellate courts hope to discern the class
of circumstances covered by a statute, even if not explicitly
mentioned in the statute.
Often the specific instance in question is a novel circum-
stance that could not possibly have been anticipated in ad-
vance in legislative deliberations. For example, do the laws
from the 1930s affecting and regulating the propagation of
radio waves also apply to television, satellite, or cellular
telephone transmissions? To answer in the negative is to
say that because the statute neither explicitly addressed
nor could conceivably have anticipated these novel develop-
ments, the statute does not apply. According to this view,
congressional intent can only be discovered in the "plain
meaning" of the language in the statute. To answer in the
affirmative, on the other hand, is to acknowledge that law-
making is a costly undertaking, that a legislature cannot be
all-knowing, and thus that legislation should be interpreted
broadly, and reasoning by analogy should be encouraged, in

order to minimize the occasions in which the legislature has
to revisit subjects.
Especially in the context of New Deal politics in the
1930s and civil rights politics in the 1960s, 1970s, and
1980s, liberals have tried to define legislative intent broadly
so as to permit the federal courts to expand the domain over
which congressional statutes applied. By arguing against a
broad interpretation of intent and in favor of the "plain
meaning" doctrine, conservatives have sought to limit what
they regard as judicial imperialism-a judiciary that, in ef-
fect, uses its power to interpret laws in order to extend or
rewrite them. Thus, liberals have developed principles of
statutory interpretation to enable broad meaning to be read
into acts of Congress, whereas conservatives have insisted
on canons of interpretation that require judges to stick to
the plain meaning of the statutory language.*
Who is right here? Short of appealing 'to our own per-
sonal prejudices and policy preferences, we can provide an
analytical perspective by means of Arrow's Theorem. The
theorem cautions against assigning individual properties to
groups. Individuals are rational, but a group is not, since it
may not even have transitively ordered preferences. If this
is true, then how can one make reference to the intent of a
group? That is, a legislator may have intentions; a legisla-
ture does not.. Indeed, in passing a statute, there may be as
many different intentions as there are legislators voting for
the bill. And some of these may not even be consistent with
one another. Former Senator John Danforth (R-Mo.) has
· While it is often the case that liberals seek broad interpretive principles
and conservatives prefer narrow ones, it does not al ways play out that
way. On matters of "police power," for example, conservatives often in-
cline toward a more permissive reading of a statute to justify greater
state activity.
said, "Any judge who tries to make legislative history out of

the free-for-all that takes place on the floor of the Senate is
on very dangerous grounds .... It is a muddle."t
I thus sympathize with the advice given by one of the
most eminent jurists in American legal history, Justice
Oliver Wendell Holmes: "The job of this Court is not to ask
what the legislature intended, but rather what the statute
means." This may not require an extremely literal reading
of all legislation (an extremely restrictive notion of "plain
meaning"), but it leaves no room for legislative intent. In-
deed, because groups differ from individuals and thus may
be incoherent, legislative intent, like jumbo shrimp and
student-athletes, is an oxymoron! Arrow's Theorem warns
us not to attribute individual characteristics, like rational-
ity, to groups.t
t Consequently, Antonin Scalia, associate justice of the Supreme Court,

has declared. "We are governed by laws, not by the intentions of legisla-
tors." The Danforth and Scalia quotes are both found in Joan Biskupic,
"Scalia Sees No Justice in Trying to Judge Intent of Congress on a Law,"
Washington Post, May 11, 1993, A4.
t The connection between legislative intent and Arrow's Theorem is devel-
oped in more detail in Kenneth A. Shepsle, "Congress Is a 'They,' Not an
'It': Legislative Intent as Oxymoron," International Review of Law and
Economics 12 (1992): 239-257.
ARROW'S THEOREM AND MAJORITY R ULE
I have been careful to state Arrow's Theorem as applying to

any process or mechanism by which individual preferences are
combined, or aggregated, or added up to produce a group pref-
erence ordering. From the perspective of democratic theory,
however, we are most interested in the applicability of Arrow's
Theorem to the method of majority rule. In this section I ex-
plicitly link the two.
First, I need to define terms. The method of majority rule
(MMR) requires that, for any pair of alternatives, j and k, j is

preferred by the group to k (written: j Pa k) if and only if the
number of group members who prefer j to k exceeds the num-
ber of who prefer k to j. We can characterize this method ex-
plicitly and then see that it is clearly a special instance of the
methods satisfying Arrow's conditions. That is, I first show
that MMR is actually composed of several essential building
blocks or properties. Then I show that these properties are all
special cases of the general conditions in Arrow's Theorem.
Consider, then, some additional (so-called) "reasonable"
conditions on preference aggregation methods:
Condition A (Anon ymity). Social preferences depend
only on the collection of individual preferences, not on who
has which preference.
Condition N (Neutrality). Interchanging the ranks of al-
ternatives j and k in each group member's preference or-
dering has the effect of interchanging the ranks of j and k
in the group preference ordering.
Condition M (Monotonicity). If an alternative j beats or
ties another alternative k-that is, j R a k- and j rises in
some group member's preferences from below k to the
same or a higher rank than k, then j now strictly beats k-
that is, j Pa k .
Like Arrow's conditions, conditions A, N, and M embody
notions of fairness in a prospective method of group choice.
Anonymity is just what it sounds like. It is a condition that re-
quires only that we know what an individual's preferences
are, not who tbe individual is holding them. Thus, if, in a
group setting like the one involving the group of college
friends described in Chapter 3, Andrew and Bonnie swap pref-
erence orderings, condition A requires that the group's choices
are unaffected by this. Condition A requires that each individ-
ual's preferences be fed into the group decision-making ma-
chinery with his or her name omitted-as, for example, is

done with a secret ballot.
Neutrality is for alternatives what anonymity is for indi-
viduals. Condition N says that it does not matter how we label
alternatives; all t hat matters is the alternatives' respective
ranks in individual preference orderings.
Finally, monotonicity requires that the method of group
choice respond "nonperversely" to changes in individual pref-
erences; moreover, it requires that the method satisfy a very
specific "knife-edge" property. The first feature of condition M
requires that if one alternative is strictly preferred by the
group to another (j Pc k), and then rises in someone's prefer-
ences, it still is strictly preferred; that is, the method of prefer-
ence aggregation docs not respond to this change in a perverse
or negative manner. The second feature of condition M states
that if two alternatives, j and k, are judged to be "socially
indifferent" (j 10 h), and j then rises in an individual's
preferences from below k to above k, j would now be strictly
preferred to k by the group--j P0 k; in effect, this says that the
decision procedure must be sensitive to changes in individual
prefer ences (a knife-edge property).
We can now report two results. The first characterizes the
method of majority rule (MMR):
May's Theorem. A method of prefe rence a ggregation
over a pair of a lternatives satis fies conditions U, A,
N, and M if and only if it is MMR.
Kenneth May, a mathematician inter ested in social choice is-
sues at about the time Arrow proved his famous theorem, es-
tablished this especially clean and clear way of describing
MMR. The theorem states, in essence, that if a group uses
"counting noses" as its method of deciding between any pair of
alternatives, then it necessarily satisfies the four conditions of
May's Theorem. T hese four conditions are the method of ma-
jority rule.
Consequently, if in some circumstance you believe MMR is
inappropriate, then it must be because you think one of May's

conditions should not hold. Should grades in the class for
which you are r eading this book be deter mined by a majority
vote among the students? Should amendments to the U.S.
Constitution be decided by majority rule? Should the captain
of a pirate ship be elected by a majority of the pirates? 16
Should my breakfast cereal tomorrow morning be decided by a
majority vote of my neighbors? Should family decisions be put
to a majority vote? In each of these cases, there is bound to be
some diversity of opinion. What May's Theorem permits each
of us to do is to defend our opinion by saying why (if in the af-
firm ative) we believe the four conditions in the theorem are
apt or why (if in the negative) we believe at least one of the
four conditions is unsuitable. Tell us why, for example, most of
you believe MMR is inappropriate as a means for selecting an
individual's morning repast.
There is more to the story, as given in the next theorem.
Display 4.5 will allow you to keep track of the conditions upon
which the theorems of Arrow and May are based (which by this
point are probably overwhelming you if you never encountered
them before). The theorem below shows that May's conditions,
which are equivalent to the simple majority-rule group deci-
sion process, are but a special case of Arrow's conditions. This
allows us to draw an obvious conclusion- majority rule must
be a process vulnerable to the afflictions described by Arrow.
May's Corollary. 17 The conditions of May's Theorem
are s pec ial cases of those of Arrow's Theorem:
--
May Condition Arrow Condition
A D
-
N I
M p
16
Apparently this was the case on many pirate ships. Arrow's Theorem would
lead one to believe either that there were many revolts aga inst captains, or
that an existing captain managed to control the agenda of s ubsequent votes.
17
The"- " symbol means "logically implies."
DISPLAY 4.5
CONDTTTO NS C HARACTERIZTNG GROUP
D ECISION-MAKING M ET HODS
Arrow May
U (Universal Domain) U (Universal Domain)
P (Unanimity) A (Anonymity)
I (Independence from Irrele- N (Neutrality)
vant Alternatives)
D (Nondictatorship) M (Monotonicity)
It is pretty easy t.o see why the first implication in the corol-
lary is true. If there were a dictator, then the condition of
anonymity could hardly hold, since the group preference is
produced by an identifiable individual; consequently if some
procedure is anonymous, then it must be nondictatorial.
The second and third implications are a little trickier to estab-
lish and so I implore the reader either to take the claims
on faith or to consult May's original paper or Riker's discus-
sion of it. 18
Stringing together May's Theorem, Arrow's Theorem, and
May's Corollary :yields the result that MMR is but a special
case of the aggregation mechanisms covered by the Arrow con-
ditions and, therefore, is subject to the same vulnerabilities.
Symbolically, we have:
MMR ++ {U, A, N, ~I} (May's Theorem)
{U, A, N, M} ~ {U, P, I, D} (May's Corollary)
{U, 1, P, D} ~ P 0 violates the Rationality
Assumption (Arrow's Theorem).
1
~ T<enneU1 0. May, "A Set of Indep<'11dcnt Necessary and Sufficient Conditions
for Simple Majority Decision," Econometrica 20 (1952): 680- 84; Riker, Lib-
eralism Against Populism, chapter 3.
MMR is "equivalent" 19 to {U, A, N, M} from May's ~'heore1n;

these four conditions, in turn, imply the Arrow conditions
fro m May's Corollary; and these latter conditions imply inco-
herent group preferences from Arrow's Theorem. Hence, MMR
cannot assure coherent group choice-something we have al-
ready seen in practice several times.
R ESTRICTIONS ON THE ARROW CONDITIONS
What is to be done? I have claimed that the Arrow conditions

and the May conditions-which are special cases of the Arrow
condjtions- are mild and innocuous requirements of fairness.
But it may be somewhat misleading to suggest that all the
conditions are, in fact, criteria of fairness. Condition P cer-
tainly is, since its claim is that it would be unfair if PG failed
to reflect whatever (unanimous) consensus exists among
group members. If everyone thoughtj was better thank, then
shouldn't the social preference bej PGk? Likewise, condition D
is a fairness requirement-alJowing for a dictator is unfair
prima facie. Condition I is intuitively less clearly a fairness re-
quirement. I t effectively requires that th e group preference
between any pair of alternatives, say j and k, depends only on
the individual preferences between j and k. It is claimed that
it would be "inappropriate" if individual preferences for irrele-
vant alternatives like h affected how j and k were ranked. In
short, condition I says that the only "sensible" way to deter-
mine whether a group prefers j to k, k to j , or is indifferent be-
tween them is to ask each group member what his or her
preference is between j and k. So, condition I is perhaps more
fittingly thought of as a criterion of appropriateness or sen-
sibleness, rather than of fairness. Nevertheless, it qualifies as
a procedural requirement.
19
That's what the do uble arrow means. That is,"-" means both "- " (implies)
and"-" (is implied by).
It is difficult, then, to alter condition I, P, or D without at

least having to present a compelling value-laden argument as
to why a fair or procedurally appropriate criterion should be
changed. Condition U, on the other hand, is an entirely dif-
ferent kind of condition. It is not a fairness criterion, nor a
criterion of appropriateness or sensibleness. It is a domain re-
quirement, and an especially wishful one at that. Essentially,
it expresses the desire that the group decision mechanism
work in all conceivable environments-that the mechanism
have the widest possible domain. This is certainly desirable.
But if we insist on this, then Arrow's Theorem tells us we will
inevitably trade off fairness for consistency. Maybe we can do
better by not insisting on condition U. That is, it may be possi-
ble to obtain both fairness and consistency, but in a restricted
domain of circumstances. This insight, at any rate, provided
the basis for several very interesting variations on the other-
wise pessimistic conclusion of the Arrow result.
Single-Peakedness
This most famous domain restriction was invented, even
before Arrow's Theorem, by the Scottish economist Duncan
Black. He believed that minimal forms of consensus well short
of unanimity might be sufficient to produce coherent group
°
choice.2 For example, consider the abortion issue in American
politics over the last four decades. Opinion is quite polarized
on a woman's options and rights. A sizable bloc of citizens,
calling themselves pro-life, believes abortion should be prohib-
ited in all circumstances (L). Another sizable group of citizens,
20 Black was unable to serve in the British Army during World War II, so he
took on voluntary night-watch duty in Scotland. Spending long nights in a
bunker watching for German aircraft, his curious mind toyed with various
geometric conditions that would enable majority rule to function smoothly.
These ideas appeared in a series of papel's at the end of the war. Their
fullest statement is found in Black, The Theory of Committees and
Elections.
calling themselves pro-choice, believes a woman has an ab-

solute right of choice (C). In 1973 the U.S. Supreme Court ar-
ticulated, in its famous Roe v. Wade decision, a compromise
permitting abortions that are not late term (R). A third group
of citizens is comfortable with this compromise position. Pro-
life and pro-choice citizens have diametrically opposed prefer-
ences- L P, R P, C and C Pi R P, L , respectively. Citizens who
do not identify with the two extreme positions have either of
two preference orderings- R P, L P, C or R P, C P, L . Notice
that neither the two extreme groups nor the one in the middle
regard R as the worst alternative; the extreme groups rank it
second, while the moderates rank it first. Now, this sort of
"consensus" is hardly earth-shattering, but in its most general
form it is sufficient to assure that a group preference based on
majority rule wi ll be transitive.
Black 's Single-P eak e dness Theorem. Consider a set
A of alterna tives from which a group G of individu-
als must make a c h o ice. If, fo r ever y s ubset of three
a lter n a tives in A, one of th ese alte rnat ives is n e ve r
worst a m o n g the three for any group m embe r , t h e n
this is suffi,c ie nt c onsen s u s so tha t t h e m ethod of m a-
j o rity rule y ie lds group pre fe r e n ces P G t h at are tran-
s it ive.
The condition that some alternative from every collection of
three alternatives in A is "not wor st'' for all group members is
called the single-peakedness condition, because it means that
there is a way to plot a preference curve for each group mem-
ber that has a :.single peak in it. I will show this in the next
chapter. Essentially the theorem says that majority rule
works perfectly well, even when group me1nbers hold wildly
divergent views on what the group ought to do, as long as a
minimal degree of consensus, captured by single-peakedness,
obtains.
Value Restriction
In a brilliant insight, the economist Amartya Sen asked,
"What's so special about being 'not worst'?" 21 Consider the set
of alternatives A • {a , b, c, d , e}. What if, for {a, b, c}, all the
members of a group agreed that b was "not best"?22 What if
they agreed that c was "not middling"? 23 Indeed, what if for
{a, b, c} there was consensus that alternative a was "not
worst," whereas for {b, c, d} all group members agreed that
d was "not best," and in {c, d, e} the group members agreed
that e was "not middling''? Sen refers to this as the condition
of value restriction. A group's preferences are value restricted
if, for every collection of three alternatives under considera-
tion, all members of the group agree that one of the alterna-
tives in this collection either is not best, not worst, or not
middling (with all members agreeing on which quality the al-
ternative in question was not). He proved a remarkable result,
generalizing Black's Theorem:
Sen,s Value-Restrict ion Theorem. The method of
majority rule yields coherent group preferences if
individual pre fere nces are value restricted .
Both single-peakedness and value restriction circumscribe
Arrow's universal domain condition, U. True enough, majority
rule won't work in all situations. But in a surprisingly large
number of such situations (204 of 216 in the three-person/
three-alternative situation, recall), the kind of consensus re-
quired may, in fact, exist.
21 "A Possibility Theor em on Majority Decision," Econometrica 34 (1966): 491-

99.
22 ln terms of Display 4.2, agreement by group members that b was "not best"
means that no member of the group had preference ordering 3, 4, 7, 9, or 11.
23 Likewise, agreement by the group that c was "not middling" means that
no group member had preference ordering 2 or 4 in Display 4.2.
CONCLUSION: THE I LLUSIVENESS OF

COLLECTIVE CLARITY
There is much to digest in this chapter, and you should not be

particularly ala rmed if much of this 1naterial remains some-
what alien and unfamiliar to you. Since Arrow's famous theo-
rem, social choice theory has become a technical language and
style of analysis with which to explore features of group deci-
sion making wit h great care and precision. It also permits the
careful consideration of significant facets of political philoso-
phy concerning what democratic, majoritarian, ins titutional
arrangements are capable of, as well as what meaning should
attach to the outcomes they produce. In short, the literature
on social choice is quite sophisticated and covers, in an en-
tirely more analytical style, much of the same ground as the
more qualitative work on democratic political philosophy.
As a practical matter, I hope the reader now appreciates
the fact- probably not at all obvious or transparent before you
read this chapter- that combining individual preferences into
a group choice, by majority rule or some other method, is not a
str aightforwardl undertaking. One thing should be clear from
even a quick and dirty read of the chapter: No method, no pro-
cedure, and no institutional arrangement that is fair in t he
most spartan, minimalist sense of this term is capable all the
time of manufacturing the silk purse of group coherence from
the sow's ear of individual coherence. Rationality may inhere
in the tastes and values of individuals, but there is no magic
wand that transforms this individual clarity about prefer-
ences into a collective clarity, especially when the group size
is large, when individual preferences are heterogeneous, or
when there is a large number of alternatives for group
members to consider. This is the content and import of Ar-
row's Theorem. However, certain kinds of consensus-single-
peakedness (Black) and value restriction (Sen)-lubricate the

institutional gears of group choice processes.
These are not the last words on these subjects in this book.
Indeed, they are only the first. Although this brief summary
hardly covers the territory of social choice, don't hang up yet;
the conversation continues in the next chapter.
1. Rational individual preferences can aggregate into irra-

tional group preferences characterized by majority preference
cycles. Are these preference cycles likely to be common accord -
ing to social choice theory? If so, under what circumstances?
Are such cycles frequent and visible occurrences in practice, or
are there features of decision-making institutions that have
reduced the apparent prevalence of group preference cycles?
Describe some of these institutions and explain how they
might provide "order" to an otherwise chaotic group decision-
making process.
2. Consider the following two sets of individuals and their

group preference rankings, aggregated using the same voting
rule.
1. individual preferences: x>y>z>w, y>z>w>x, and
z>w>x>y
group preferences: x>0 y, z> 0 x, w>0 x, y>0 w, y>0 z,
and z> 0 w
2. individual preferences: y>z>x>w, y>w>x>z, and
y>w>z>x
group preferences: y>0 x, y>0 w, z> 0 y, x>0 w, z> 0 x, and
z> 0 w
Which of Arrow's conditions (P, D, I, or Transitivity) is violated

by their group preferences? (Hint: checking I requires
comparing the outcomes in the two different groups to find a
violation.)
3. For each of the following voting rules determine if it vio-

lates one of Arrow's conditions (P, D, or I): the method of
majority rule, a round-robin tournament, the unanimity rule
(x P0 y if and only if at least one member of society strictly
prefers x toy, and no members of society strictly prefer y to x),
and the lexicogrraphic rule (society adopts the preferences of
the individual whose name comes first alphabetically). If not,
give a simple example to show that it can generate preference
cycles.
4. May's theorem suggests that any deviation from majority

rule must be justified by a reasonable departure from one of
four conditions: U, A, N, or M. For each of the following cases,
explain which of these conditions is violated by the electoral
rule, and suggest a possible justification: (a) the proposal of an
amendment to the U.S. Constitution requires two-thirds sup-
port in each chamber of Congress; (b) the International Mone-
tary Fund (IMF) uses a system of weighted voting where
weights are det ermined by contributions to IMF operating
funds. The United States also holds a veto in some circum-
stances; (c) a guilty verdict in a criminal case usually requires
unanimity, or a large supermajority on the jury; (d) the
French president is elected under t wo-stage majority rule. In
the first stage, all parties' candidates compete against one an- •
other. A second stage takes place between the top two vote get-
ters from stage one only if no candidate secured an outright
majority in stage one.
*5. For each of the following societies, state whether the pref-
erences satisfy Sen's value-restriction criterion, that is, that
for any three outcomes, all voters agree that at least one of
the outcomes is not first, middle, or last. If not, identify the
tuple(s) of preferences that violate value-restricted prefer-
ences. Assuming that the voting rule is majority rule, are the
group preferences in societies without value-restricted prefer-
ences transitive or intransitive?
Society 1:
yP 1 xP 1 zP 1 w
wP 2 yP 2 xP2 z
z P3 y P3 wP 3 x
Society 2:
y pl w pl z pl x
w P2 x P2 y P2 z
z P3 w P3 y P3 x
Society 3:
y p l w p l z pl x
zP 2 xP 2 yP 2 w
xP 3 yP 3 wP3 z
6. Explain the relevance of Arrow's Theorem for the following

individuals: a social scientist trying to predict how a legisla-
ture will vote to divide up some new tax revenue; a committee
trying to identify the perfect voting rule to avoid committee
cycling and deadlock; a political theorist who defends democ-
racy on the grounds that it is the only way to determine soci-
ety's true preferences; and a jurist attempting to interpret the
"legislative intent" of the coalition which passed a particular
bill in Congress.
7. Arrow's Theorem suggests that there is a trade-off between

the concentration of power and collective rationality. Explain
this trade-off, and then provide a brief example using three in-
dividuals voting over three outcomes to illustrate. An alterna-
tive formulation is that Arrow's Theorem suggests a trade-off
between fairness and consistency. Explain this trade-off, care-
fully describing the meaning of fairness and consistency under
these circumstances.
5
Spatial Models of Majority Rule
The story line to this point has emphasized a trade-off in

group decision making between the coherence of group choices
on the one hand and the fairness of the method of decision
making on the other. If we consider a limited domain of cir-
cumstances, then we may be able to avoid the pain of this
trade-off. Put somewhat differently, if individual preferences
happen to arrange themselves in particular ways- that reflect
a consensus of a specific sort-then group decisions (certainly
those made by majority rule) work out quite nicely. In the pre-
vious chapter, I described single-peaked preferences as one
kind of consensus that facilitated coherence in majority-rule
decision making. In this chapter I want to give an intuitive
geometric characterization of this condition.
Frankly, however, all this gets pretty boring pretty quickly.
The author, and perhaps some of the readers, may enjoy tech-
nical riffs and philosophical discourses, but most readers are
more impatient and anxious to see some payoff. I think this
chapter constitutes an important investment. Once I give
single-peakedness a geometric representation I will be able to
apply it to some interesting political situations-namely, two-
party electoral competition and legislative committee decision
making.
90
Spatial Models of Majority Rule 91
SPATIAL FORMULATION
The Simple Geometry of Majority Rule
Suppose a group's problem is, in effect, to pick a point on a
line: the group must select some single numerical parameter.
For example, a bank's board of directors must decide each
week on the week's interest rate for thirty-year home mort-
gages. In effect, the relevant interest rates are points on a
line, one endpoint being 0 percent and the other being some
positive number, say 10 percent. This interval is written as
[O, 10). In this and other circumstances, I want the reader to
imagine a group of individuals each of whom has a most-
preferred point on the line and preferences that decline as
points further away in either direction are taken up.
In Figure 5.1 the preferences of the five-person board of
bank directors, G = {l, 2, 3, 4, 5}, are displayed. The board is
meeting on Monday morning to decide the interest rate to
FIGURE 5.1
UTILITY
0 2 10
charge for home mortgages this coming week. Each individual

i E G has a most-preferred point (also called bliss point or
ideal point), labeled xi, located on the [O, 10] interval (drawn
as the horizontal a xis), representing his or her most-preferred
interest rate.1 Thus, director 1 h as a most-preferred interest
rate (x1) of just less th an 4 percent, director 2's (x2 ) is just
more than 4 percent, and so on. On the vertical axis I have
written the label "utility" to measur e preferences. For each in-
dividual I have graphed a utility function, which represents
the director's prefer ences for various interest rate levels in the
[0, 10) interval. Naturally, the utility function , labeled ui for
Mr. or Ms. i, is highest for i's most-preferr ed alternative, xi,
and declines as more distant points are consider ed. Thus, Ms.
5 most prefers an interest rate a little higher tha n 8 percent,
wi th her preference declining either for higher or lower r ates.
For obvious reasons (just look at the graphs) the preferences
of these individuals are single-peaked, which is defined as
follows:
Single-P eakedness Condition. The prefe r e n ces of
group m embers a re s aid t o be single-pea k ed if t he al-
ternatives under consid e ration can b e rep res e nte d
as points on a line a nd each of the utility func tio n s
representing prefe rences over these a ltern a tiv es
has a maximum a t some p oint on the line and slopes
a way from this m a ximu m on either side.
Is there any connection between this definition of single-
peakedness and Black's definition given in the pr evious
chapter? That is, do utility functions with a single peak, as
displayed in Figure 5.1, have anything to do with all voters
agreeing that some alternative is "not worst"? You bet! Take
any three inter est rates displayed in Figure 5.1-say, 3 per-
1
Recall that "i E G" means "the element i in the set G," where i stands for any
one of the five bank directors.
FIGURE 5.2
UTILITY
Ps(Y)
0 4 6 y= 7 8 10
X5 = 8.25
cent, 5 percent, and 9 percent. It is pretty easy to see that

5 percent is not the worst among these three rates for any of
the five members of the group. Indeed, for any three interest
rate levels the reader chooses, one of those is not worst for any
of the five bankers. That's what single-peakedness means!
In order to develop some tools that will be used in subse-
quent analysis, let's look at one of these individual bankers in
isolation (by which we really mean let's look at an isolated
utility function). In Figure 5.2 we show the most mean-
spirited of the bank's directors, Ms. 5, who most prefers a
fairly high inteirest rate: x 5 = 8.25%. Consider an alternative
rate, y = 7%. The set of points Ms. 5 prefers toy is described by
the set labeled P 5(y) in Figure 5.2. This is Ms. 5's preferred-
to-y set: if y wer e on offer, then P5 (y) describes all the points
she would prefer to it, given her preferences. As the figure
shows, P 5(y) is computed by determining the utility level for y
FIGURE 5.3
UTILITY
U5
U4
U3
U2
U1
0 2 4 6 8 10
I
X1 X2 y X3 X4 X5
'----y---1 '--.,--'
P2(Y) P3(y)
P1(Y) P4(y)
Ps(Y)
and then identifying all the interest rates on the horizontal

axis with utility levels greater than the utility for y. 2
In Figure 5.3 I display the preferred-to-y sets of all five
bank directors (note that y, in this figur e, is just below 6 per-
cent). Notice that these sets overlap to some degree-there are
points in common to Piy) and P5 (y), for example. This means
that there are specific points that both Mr. 4 and Ms. 5 prefer
to y. 3
Of great interest to us is the set of points a majority
prefers toy. This is called the majority winset of y, written as
W(y) . We define it as follows. Let M be the set of majorities in
our group of bankers, G; it is the collection of three-person
coalitions (there are ten such coalitions), four-person coali-
2
The endpoints of the preferred-t.o-y set are included even though, technically
speaking, th e group me mber ranks these endpoints at the same utility level
asy.
3
In set-theoretic notation, we can write these common points as the intersec-
tion of the two preferred-to-y sets: P 4(y)nP6 (y). (n is the intersection symbol,
so that AnB means "the points in both set A and set B .")
Spatial Models of Majority R ule 95
DISPLAY 5.1
THE MAJORITY COALITIONS OF G= {l,2,3,4,5}
S ize of Coalition Coalitions
3 {1,2,3}, {1,2,4}, {l,2,5},
{1,3,4}, {l,3,5}, {l,4,5}
{2,3,4}, {2,3,5}, {2,4,5}
{3,4,5}
4 {1,2,3,4}, {1,2,3,5}, {1,2,4,5}
{l,3,4,5}, {2,3,4,5}
5 {1,2,3,4,5}
tions (there are five of these), and the coalition-of-the-whole.

So, there are sixteen different majority coalitions in M; they
are listed in Display 5.1. For each of these sixteen majority
coalitions, consider the common intersection of preferred-to-y
sets (if there is any); these are the points that this particular
majority prefera to y. Thus, the members of the majority
{3,4,5} in Figure 5.3 share points each prefers toy. Determine
this set for each of the majority coalitions. Then take the
union of these sixteen sets. This is W( y). 4
It is now rather straightforward to describe the coherent
choices of groups. If some alternative, x, has an empty winset
(written: W(x) - 0, where 0 means "empty'' in set notation),
then it is a clear candidate for the group choice. Why? Simply
because W(x) • 0 means there is no other alternative that any
of the sixteen majority coalitions prefers to x. It's hard to deny
choosing x if there is nothing any majority agrees on in its
• In Figure 5.3 it turns out that members of only one of the sixteen majorities,
{3,4,5}, has overlapping Pl(y) sets. Members of the r emaining fifteen major·
ities (listed in Display 5.1) cannot agree on any points they jointly prefer to
y. For any one of those fifteen, say, {l,2,4}, some members prefer only points
to the left of y while others prefer only points to the right of y. As a group
they cannot agree. Thus W(y) - · P 3(y) n P 4(y) n P 6(y).
place. On the other hand, if the winset of xis not empty (W(x)
"' 0), then it is hard to justify the choice of x. How can you
choose x when some majority of the group clearly wants some
other specific alternative? And if the winset is nonempty for
euery alternative, we have a problem: the group's preferences
are incoherent, since some majority prefers something to
every alternative available.
The question of the moment is whether, or in what circum-
stances, an x possessing an empty winset exists. If any com-
plete and transitive preferences may be held by the
individuals in 0-Arrow's "universal domain" condition-
then, as we have seen, the answer is "not necessarily." Why?
Because under Arrow's condition U, it is possible for majority
preferences to cycle, in which case W(x) = 0 for no alternative.
But if preferences are restricted, then a different answer is
possible.
Black's Median-Voter Theorem. If members of g roup
G have s ing le-p eaked prefe r ences, then the ideal
point of the median voter has an e mpty w inset.
One such group consisting of individuals with single-peaked
preferences is pictured in Figure 5.3'. The median voter ideal
point in this group is x3 of Mr. 3. 5 The claim of Black's Theo-
rem (th·e same Duncan Black, by the way, as in the previous
chapter) is that W(xJ = 0, and that x 3 is the majority choice.
We can prove this theorem using the example of the five
bank board mcmlber s. Consider any arbitrary point in the fea-
sible set of interest rates, [O, 10], to the left of x3-say the
point labeled a i n Figure 5.3'. Notice that a is preferred to
x3 by members 1 and 2. since x3 is not in either P/a) or P/a),
but x3 is preferred to a by members 3, 4 and 5. Thus, x3 is
6
The median of a set ordered from left to right is the point such Lhat at least
half the points are aL or to its right and at least half the points are at or to
its left.
FIGURE 5.3'
UTILITY
0 2 4 6 8 10
I
x, X2 Ct X3 X4 X5
P2(et) P3(<X)
,------'----, ,-J--.,
Psets --------------------------------------~
P 1( a) P4{ a)
P5(<X)
majority-preferred to a. But a is any arbitrary point to the left

of x 3 . For any such point, we know at the very least that mem-
bers 3, 4, and 5 will prefer x3 to it. (It is possible that some of
the remaining members will share this preference, too.) Next,
consider any arbitrary point to the right of x3 (not pictured).
Members 4 and 5 may prefer it to x3, but members 1, 2, and 3
hold the opposi te preference, so that x3 is majority-preferred.
The argument is exactly the same as with a above, since we
selected an arbitrary alternative to the right of x3 . To sum up,
we now know that the ideal point of the median voter is pre-
ferred by a majority to any arbitrary point to the right or to
the left of it, that is, to all remaining points. Hence, it has an
empty winset and is the majority choice.
Before complicating this key result, I should mention that
there are three hidden assumptions, and probably more be-
sides, that warrant some discussion. First, in the example, the
group G of bankers is odd in number. Thus, in Figures 5.3 and
5.3' the displays of five group members possessed a unique

median-x 3 . What if the size of the group were even? Suppose,
for instance, that Ms. 5 in either figure is ignored and the
focus is instead on the truncated group G' = {l, 2, 3, 4}. Now
members 2 and 3 are both medians. Moreover, since it takes
three votes to constitute a majority, it is true that W(x2) ~ 0
and that W(x3) = 0, too. Indeed, the winset of any point in the
interval between the two, [xzi x:i], is empty. Technically, then,
Black's Median-Voter Theorem is true whether the group size
is odd or even. But, as has just been shown, when a group has
an even number of members, then more than one alternative
can have the property that it cannot be beaten. This follows
because of the possibility of tie votes. For example, suppose x
and y each defeat every other alternative in {x, y, z, w, u, v} by
a simple majority, but tie when paired against each other.
Then neither x nor y can be beaten. This absence of a unique
winner is a pain in the neck. It is a bit like more than one pre-
tender to the throne, or more than one person claiming to be
king of the mountain. It is for this reason that groups estab-
lish some procedure for breaking ties well in advance of any
substantive deliberations or, better yet, that they make sure
that the group is odd in number. 6
Second, full participation is assumed. Everyone with the
franchise is assumed to exercise it. Of course, in any particu-
lar instance of group choice this need not happen. If bank
board members 4 and 5 oversleep one week, then Ms. 2 be-
comes the median voter of the now reduced three-person
board; if members 1 and 3 are out of town the following week,
then Mr. 4 becomes the median. In each of these cases, as well
6 For example, the U.S. Constitution requires the Senate to have an even
number but establishes a tiebreaking procedure. The vice president of the
United States, sitting as the president of the Senate, is allowed to vote only
in case of a tic. Likew~se, the standing rules of the House of Representatives,
which has an odd number, provide a tiebreaking rule, asserting that a mo-
tion fails if it obtains no more yeas than nays-it fails on a tie.
as in the case of the full board, the median-voter result ap-

plies . Who the median is, however, depends upon who the par-
ticipants in the group are. We may forecast the group decision
if we make assumptions about participation (for example:
assume everyone votes), or we may make forecasts contingent
on participation (for example, if 4 and 5 oversleep, then we
predict x 2 will be the group's choice; but if everyone votes, then
x 3 is the predicted choice).
Third, it is assumed that those exercising the franchise do
so sincerely. But as we have seen in earlier chapters, group
members will have occasion and incentive to misrepresent
their preferences and not reveal them honestly. This is a sub-
ject of great interest that we take up on its own in Chapter 6.
The (Sli;ghtly) More Complicated Geometry

of Majority Rule
One-dimensional models of choice with the single-peakedness
condition permit rather sophisticated ways to think about real
politics. They generate very crisp expectations about how pol-
itics in these settings gets played out. But many social situa-
tions cannot be reduced to one-dimensional affairs.
Recall the game of "divide the dollars." If the game were
played by a group of three individuals, then it is necessary to
have two dimensions in which to represent outcomes. The first
dimension gives the amount that player 1 receives, while the
second dimension gives the amount that player 2 receives.
(Subtract the sum of these two numbers from the total num-
ber of dollars to be divided and you get the amount that player
3 receives. 7 ) I hope I convinced the reader earlier that games
of division, like "divide the dollars," are commonplace in polit-
ical life. So it must be conceded that as crisp and as sophisti-
7 Generally, when dividing a fixed pie among n categories (or people), we need
only n- 1 dimensions to display all outcomes.
FIGURE 5.4
UTILITY
u1 BUTIER
,'
-------------~~=-
,, ..
,
9 1Y "---- - ---+
GUNS
cated as the one-dimensional models are, they are special

cases of a more general multidimensional arrangement. We
need to see what this more general arrangement is like.
Most of what needs to be said is covered by focusing on a
two-dimensional circumstance, like that pictured in Fig-
ure 5.4. (There are actually three dimensions in this figure,
but this will be clarified shortly.) Let's consider a problem in
budgeting, in which a group of legislators, perhaps an appro-
priations committee, must decide how to divide expenditures
between "guns" and "butter" (symbolizing the competition be-
tween defense and other domestic programs). Outcomes, then,
are described by two numbers: dollars spent on butter and dol-
lars spent on guns. The set of outcomes, or simply the "policy
space," is two-dimensional, and this is the domain over which
preferences are expressed. 8 The third dimension of Figure 5.4,

marked "utility," permits us to draw three-dimensional graphs
of legislator preferences. As earlier, a legislator is assumed to
have an ideal point in the policy space. His preference func-
tion, or utility function, is at a maximum over this point. It is
assumed further (in most of the applications) that preferences
decline with "distance" from the legislator's ideal point. A typ-
ical legislator, with a typical ideal point and preference func-
tion, is displayed in Figure 5.4. The legislator's preference
function is a "hump" that reaches its highest utility level just
over his ideal point in which bt dollars are spent on butter and
g 1 dollars on guns. This ideal point, (bv g 1), is located in the
plane of the butter-guns policy space.
A more convenient way to represent precisely this same in-
formation, however, is given in Figure 5.5. In this figure the
reader is looking down directly onto the plane of the butter-
guns policy space. It is as though you are hovering in a heli-
copter above th e peak of the preference hump in Figure 5.4.
The location of our typical legislator's ideal point is exactly the
same as in Figure 5.4. But instead of adding a third dimension
(coming out of the page toward you) in order to graph his pref-
erence furtction, we instead overlay "slices" of his utility func-
tion onto the policy space, producing the set of nested circles
called indifference curves. Each circle is a slice of the policy
hump in Figure 5.4. It is a locus of policy outcomes among
which the legislator is indifferent (since all the points on a cir-
cle lie on the same slice and hence at the same height on the
utility function of Figure 5.4). Since distance from an ideal
point is a measure of preference, points on a circle centered on
8 If the size of the budget were fixed in advance, then by the argument in the
previous paragraph and footnote, only one dimension would be needed, say,
dollars for defense; once this is established, dollars for domestic programs is
strictly determined-it's what is left over. When the budget is not deter-
mined in advance, then we need both dimensions to display all outcomes.
FIGURE 5.5
GUNS
0 BUTTER
her ideal, being equidistant from that ideal, are equally pre-
ferred by her. The logic is the same in comparing a point on
one circle to that on a nother. A legislator prefers a point on a
circle with a smalkr radius to one on a circle with a larger ra-
dius, because this 1neans the form er point is closer to her ideal
than is the latter point.9
Notice the point labeled yin Figure 5.5. The circle through
9 In this simplest of multidimensional setups, in which t he policy space is

two-dimensional and preference is measured by distance, indifference
curves will be circles centered on the legislator's ideal point. In more than
t wo dimensions, the indi fference "contours" will be spheres or (in four or
more dimensions) hyp<!rspheres. A second sort of complication, which applies
in the (simplest) two· dimensional as well as higher-dimensional situations,
is to allow preferences to be related to di st ance, but in a more complicated
way. One dimension of policy may be "more important" to a legislator than
another dimension. Thus, movement away from her ideal point along one di-
mension v.'111 have a greater impact on utility than an identical movement
FIGURE 5.6
-W(y)
y centered on our legislator's ideal point, as we just deter-

mined above, contains all the points of legislator indifference
to y. This means that all the points inside the circle, being
closer to her ideal, are actually preferred by h er toy. That is,
we can call t he points inside the circle our legislator's
preferred-to-y set, a natural gener alization of that sa me con-
cept in the one-dimensional development ear lier in this chap-
ter. Figure 5.6 displays three legislator ideal points and each
legislator's indifference curve through y (the curve plus all
along the other dimension. Put differently, preference is said in this instance
to decline with weighted distance from the ideal (where the weights reflect
the salience of each dimension to the legislator). In this instance, indiffer·
ence contours will no longer be circles, but will be ellipses instead. I wiJl
stick with the most basic formulation.
points inside it comprising each legislator's preferred-to-y set,

labeled P;(y)). The shaded intersection P 1(y)nP2(y) gives the
points preferred by both legislators I and 2 to y; P 1( y)nPa( y)
are the points preferred by 1 and 3 to y; a nd, finally,
P/y)nP3(y) give those points for 2 and 3. Since two out of
three is a majority, the union of these three "petals" is the
winset of y, W(y). Each petal gives the points that a specific
majority coalition prefers toy.
If we move the ideal points of the three legislators in Fig-
ure 5.6 around, so that they line up in a row, we have a situa-
tion like that depicted in Figure 5.7. Can you determine the
winset of the middle legislator's ideal, W(x2)? The steps are
laid out in that figure. Mr. 2's preferred-to-x2 set is empty (how
could he prefer anything to his most-preferred point?). Ms. l's
and Ms. 3's indifference curves through x 2 and centered on
their respective ideal points are tangent to one another. They
do not overla p at all. Hence W(x2) .. 0, since there are no
FIGURE 5.7
,,
,,
points preferred to x 2 by majority {1,2}, {1,3}, {2,3}, or {1,2,3}.

That is, the members of no majority coalition have preferred-
to-x2 sets that intersect. Thus, x 2 is the majority choice.
Another look at Figure 5.7 should show why this hap-
pened. Consider the bold line through x 2 perpendicular to the
dashed line. Oo the x1 side of this bold line, for any selected
point off the dashed line, x 2 is closer to x3 than the selected
point is. So a majority, {2,3}, prefers x2 to any such point. From
a precisely parallel argument, a majority, {1,2}, prefers x 2 to
any point off the dashed line on the x3 side of the bold line. So,
the only points that remain are those on the dashed line. That
is, even though the group choice problem is actually two-
dimensional, individual preferences line up so as to make the
problem effectively one-dimensional. On this line individual
legislators have single-peaked preferences (the reader should
convince herself of this), with x 2 the median ideal point.
Hence, Black's Median-Voter Theorem applies, which is pre-
cisely what Figure 5.7 demonstrates.
Having the legislator ideal points line up is pretty conve-
nient, isn't it? Pretty unlikely, too. Certainly it seems more un-
likely than arbitrary configurations such as that in Figure 5.6.
Thus, while we have tools like Black's Median-Voter Theorem
with which to analyze majority rule in one-dimensional set-
tings, it is probably fair to say that many interesting political
circumstances are genuinely multidimensional. Can we say
anything about the prospects for a majority choice in multiple
dimensions? The answer is yes, but the news is not very good.
The highly unlikely distribution of individual preferences
in Figure 5.7 pr ovides a basis for generalization. What allows
the ideal point of Mr. 2 to emerge as the majority choice is the
fact that the ideals of the others are "symmetrically" distrib-
uted about 2's ideal. From Mr. 2's ideal, any movement away
from it is obviously opposed by Mr. 2 himself; but it is also al-
ways opposed by at least one of the other guys. In fact, as we
saw when considering the bold line through x2 perpendicular
to the dashed line containing all three ideal points, any point
on Ms. 3's side of that line is less preferred than x 2 by 1 a nd 2,
and any point on Ms. l's side of the line is less preferred than
x2 by 2 and 3.
Now let's add two more voters to the picture that are sym-
metrical i.n precisely the same way (Figure 5.8). Voters 2, 4,
and 5 lie on a line , just as 1, 2, and 3 do. It is still the case that
W(x 2) - 0, because any departure from x 2 is opposed by at
least three of the :five voters. You may test this proposition out
for yourself by laying a straightedge through x 2 at any a ngle.
There a re always two voters who would like to move to some
point on one side of the straightedge, two who would like to
move to points on the other side of the straightedge, and
one (Mr. 2) perfectly content to stay at x 2 . Since no majority
favors moving in any direction (there are always three votes
against), the winset of x 2 is empty. Something about distribut-
ing voters symmetrically around a common point seems to be
producing a coherent majority choice.
Indeed, we can be very specific here. Let us consider a set
of m voters (where mis any number, which we will take to be
odd to si mplify the presentation), whose ideal poin ts are xi' x2'
... , x n.. These m ideal points are in a multidimensional policy
space, like the one pictured in Figure 5.8 (although the results
presented below apply to policy spaces of more than two di-
mensions as well). These ideal points are distributed in a radi-
ally symmetric fashion if the following conditions hold:
(1) There is a distinguished ideal point, labeled x*; (2) the n-1
remaining ideal points can be divided into pairs (since n is
odd, n-1 is even and this is possible); and (3) t he two ideal
points in any pair, say X; and xi' plus x* all lie on a line with x*
"between" x; and x;yj. In Figure 5.8, x 2 is the distinguished
point, x 1-x3 and x 4 -x" are t he pairs of remaining ideal points,
and x 2 lies on a line " between" the ideal points in each pair.
Notice that radial symmetry does not require the two ideal
FIGURE 5.8
points of a pair to be equidistant from the distinguished point

(x3 is closer to x 2 than x 1 is); they must simply line up.
The economist Charles Plott noticed that radial symmetry
of ideal points captured in higher dimensions a property that
single-peaked preferences possess in one-dimensional policy
spaces. In a famous paper in 1967,10 he established the follow-
ing result:
Plott's Theorem. If voter s possess distance-base d
spatial preferences, a nd if their ideal points are dis-
tribute d in a radially symmetric fashion with x* the
distinguished ideal point, and the number of voters
is odd, then W(x*) = 0.
°Charles R. Plott, "A Notion of Equilibrium and Its Possibility under Major-
1
ity Rule," Amer~can Economic Review 57 (1967): 787-806.

Plott's Theorem generalizes Black's Median-Voter Theorem to

more than one dimension. 11 It provides the conditions for the
centripetal ("center-seeking") tendency observed in the one-
dimensional setting studied by Black. Thus, in Figure 5.8 dis-
playing the five legislators whose ideal points satisfy radial
symmetry, x 2 is the majority choice. It is preferred by a major-
ity to any other point in the policy space.
That's the good news. The bad news is that Plott's Theorem
does not provide very general conditions. Return to Figure 5.8
and, in your mind's eye, move x 4 a small distance to the north-
ea st (say half an inch). Now lay your straightedge along a line
through x 2 but just below x 5 . Since x3, x 4 , and x 5 now all lie
above this line, this means that there is a majority-{3,4,5}-
that prefers moving away from x 2 • Specifically, there are
points on the northeast side of this straightedge through x 2
that voters 3, 4, and 5 prefer to it, so that W(x 2) .,, 0. That is,
Plott's condition of radial symmetry is highly sensitive to
small perturbations of voter ideal point locations. This can be
put in an especially dramatic form. Imagine the ideal points of
1,000,000 voters radially distributed around the ideal point of
the 1,000,00lst voter. So, W(x 1.000.001) ~ 0, in accord with
Plott's Theorem. Now suppose two new voters move into the
com munity, and their ideal points are not radially symmetric
about x 1.ooo.ooi· This small perturbation in the voting situa-
tion-after all, how much effect can the introduction of two
new voters have in a voting population of more than one mil-
lion?-completely destroys the previous equilibrium. 12
11 One result is said to '"generalize" another when the latter is a special case of
the former. Thus, Black's median-voter result is the one-dimensional ver-
sion of Plott's Theorem, in which the median voter 's ideal is the distin-
guished point and pairs of voter ideal points, one from each side of the
median, are distributed arou nd it in a radially sym metric fashion.
12 The sensitivity is not quite so severe wh en the number of voters is even. In
this case the disting uished point is not a voter ideal point. Some shifts in
voter ideal points are possible without disturbing the empty winset prop-
erty of this distinguis hed point.
If departures from radial symmetry were relatively un-

usual events, then t his sensitivity to ideal point distributions
in Plott's Theorem would not really be bad news. But, as the
reader may grasp intuitively, the requirement of radial sym-
metry is actually quite restrictive; one would not expect groups
"naturally" to have their preferences distributed in so elegant
and uniform a manner as this. So departures from this condi-
tion take on a greater significance. In what is one of the most
remarkable theoretical statements in this entire field, Richard
McKelvey demonstrated exactly how significant these depar-
tures from radial symmetry are.
Mc Kelvey's Chaos Theorem. 13 In multidimens iona l
spatia l settings, except in the case of a rare distri-
bution of ideal points (like radia l symmetry) that
h ardly ever occurs naturally, there will b e no major-
ity rule empty-winset point. Instead t h ere will b e
chaos-no Condorcet winner, anything can happen,
and whoever controls the order of voting can deter-
mine t h e fin al outcome.
I started out by seeking ways to restrict Arrow's universal-
domain condition to see if there were narrower domains in
which majority rule worked tolerably well. In one-dimensional
choice situations, we saw that single-peakedness is sufficient.
In multidimensional situations, a radially symmetric distribu-
tion of ideal points is sufficient. But small departures from the
latter throw everything into chaos. No point is the ''king of the
mountain" in the sense that it is preferred by a majority to all
contenders, so it is difficult to justify any particular group
choice (since for' any proposed choice there is some alternative
a majority prefers to it). This, in turn, means that there will
always be majority cycles.
13 Richard D. McKelvey, "Jntransitivities in Multidimensional Voting Models,"

Journal of Economic Theory 12 (1976): 472-82.
Indeed, McKelvey establishes that all the points are in one

great big cycle. Wh at this means, practically speaking, is that
the situation is ripe for manipulation by whoever controls the
agenda. What McKeJvey shows is this: Pick a ny two points in
the policy space-call them s (starting point) and t (terminat-
ing point) . Then there is a sequence of points- zl' z2' ... , z;zk
(for some finite number, k) s uch that z 1 Pas, z 2P r,Z" z 3P cz2 , . . . ,
zk Pc k · 1, and t Pc zk. That is, from a ny s tarting point, there is
a sequence of votes by which a majority will move the outcome
to any terminal point (including, say, the ideal point of the
agenda setter) .
This is illustrated in Figure 5.9 for a three-person legisla-
ture. The ideal points of the three legislators are x" x2' and x:{·
The point s is the status quo ante. If Mr. 3 were the agenda
setter empowered to make motions a nd order them in a voting
sequence, then he could, in a s mall number of steps-in fact,
in only three steps-drive the outcome to x 3, his ideal point.
First he proposes z l' which both Mr. 1 and Ms. 2 prefer to s. So
z 1Pas. Then he proposes z 2, which both he and Mr. 1 prefer to
z1 ; so z 2P cz1. Then, in the final step, he proposes his ideal
point, x 3, which both he and Ms. 2 prefer to z2 . Voila! He has
driven the legislative process, by artfully choosing the alterna-
tives upon which to vote, to a terminal outcome located at his
ideal policy: t = x:i. 14
APPLICATIONS
Applications of the spatial model are so plentiful and rich that

it is hard to know where to start. I begin at the beginning, so
14 It should he noted that the members of each majority coalition in this exam-
ple bl indly vote their prefere nces, like la mbs following the judas goat to
slaughte r. The legislators seem like putty in th e hands of the wily agenda
setter, Mr. 3. In the n ext chapter, we will endow "followers" with some so·
phistication by which they might be able to control their "leader."
FIGURE 5.9
to speak, with Downs's model of electoral competition. An-

thony Downs was one of the first scholars to use the spatial
model for political analysis. This application also demon-
strates both the strengths and weaknesses of the simplifying
assumption that the political world can be modeled as one-
dimensional. Then we will turn our spotlight on institutional
analysis, looking at both a one-dimep.sional and multidimen-
sional analysis of legislative politics.
Spatial Elections
The real origins of the spatial model are found in a famous
paper written in 1929 by Harold Hotelling. 15 An economist in-
15 "Stability in Competition," Economic Journal 39 (1929): 41-57.

terested in the locational decisions made by firms, Hotelling

was especially fascinated by the stylized fact-true then, and
still true today- that competitor firms regularly locate their
retail shops next door to or just across the street from one an-
other. Gasoline stations are found on opposing corners of an
intersection, a pair of major department stores "anchor" a sub-
urban shopping mall, Starbucks and Peet's Coffee & Tea are
practically neighbors, and, in small -town America in the good
ol' days, competing "five and dime" stores like Woolworth's
and Kresge's were located just opposite one another. Why
would nominal competitors, who have a great big geographic
market to divide up among themselves, locate in such close
proximity? 16
I leave the economic location question to one side, but not
without noting that Hotelling mused that political parties
seemed to behave in much the same fashion as economic
competitors. This musing became the major focus of the now
classic study by Anthony Downs, An Economic Theory of
Democracy, where he gave the "spatial model of electoral com-
petition" its fullest development and exposure. L7
The "spatial" part of Downs's spatial model consists of a
one-dimensional ideological continu um , [O, 100]. The contin-
uum is scaled by the proportion of economic activity left in the
hands of the private sector, so that the left endpoint reflects a
fully socialized economy, while the right endpoint is identified
with a totally private-enterprise economy. While political com-
petition in real life consists of taking positions on and articu-
lating visions about a host of political issues, Downs supposes
that, when all is said and done, political debate boils down to
ideology-do you want some good, service, or purpose provided
16 Even more spectacular in many cities is a small stretch of a major highway

along which dozens of automobile dealerships locate.
17 Anthony Downs, An Economic Theory of Democracy (New York: Harper &
Row, 1957).
by government or by the private sector? Political competition,

then, is a contest between politicians intent upon capturing
control of govern1nent by appealing to voters with offers of al-
ternative plans, platforms, programs- indeed visions. These
appeals are identified with different points on the left-right
ideological continuum.
As a first approximation for the hurly-burly of campaign-
ing, electioneering, and voting, this is not a bad one. Politi-
cians are conceived of as single-minded seekers of election.
They are graduates, so to speak, of the Vince Lombardi School
of Politics, whose motto is, ·~winning isn't everything; it's the
only thing." 18 Downs assumed politicians seek to maximize
votes, although in variations on his model, politicians alterna-
tively maximize their vote plurality (the difference between
their vote and that of their closest competitor) or their proba-
bility of winning. In any event, most early spatial models of
electoral competition took votes to be the coin of the realm, re-
garded politicians as focused exclusively on winning elections,
and suggested that they did so by promising policies, plat-
forms, and programs that attracted voters. In this spatial con-
text, a candidate is represented by some location on the
ideological continuum, some point in the [O, 100) interval.
This is his or her political position.
Voters, Downs assumed, were single-mindedly interested
in policy: the goods and services produced by government (or
left to the private sector); the form and content of government
regulation of the private sector; the distribution of tax, unem-
18 For those too young to remember, Vince Lombardi was the legendary foot-
ball coach of the Green Bay Packers and, at the end of his career, the Wash-
ington Redskins . To Lombardi, nothing was more important than winning.
While not exactly an uplifting imperative, Lombardi's maxim is a pretty
good first approximation for what it takes to succeed in the national politics
of most countries, the business world, and the Na tional Football League.
(By the way, though associated with Lombardi, it is not clear h e really ever
said this!)
FIGURE 5.10
1 (125) 3 (125) 4 (125) 5 (125)
0 X1 Xz l 100
ployment, and inflation burdens; government policies on social

issues like abortion and divorce; and matters of war and
peace. Voters care mightily about these matters and base their
assessments of candidates accordingly. Voters, however, are
heterogeneous in their tastes so, just as there are left-wing
and right-wing politicians, there are left-wing and right-wing
voters. Specifically, each voter is identified with some point in
the (0, 100] ideological space-the voter's ideal point-and his
or her preferences are assumed to decline for points more and
more distant from this ideal. That is, the set of voters may be
represented by single-peaked preferences. Figure 5.10 dis-
plays an electorate of 625 voters (actually, five different voter
"types" with 125 voters of each type). A voter of type i (i =
1,2,3,4,5) has ideal point xi, and preferences declining in dis-
tance from x;.
The most famous version of the Downsian model involves
two-candidate competition. The question Downs asks is: Given
a distribution of voters like that in Figure 5.10, where will two
single-minded seekers of election locate themselves? We can
gain some insight into this question by fixing the position of
one of the candidates. Let's fix the position of L, the leftie can-
didate, at l as shown in the figure. What position, r, should R,
the rightie candidate, adopt so as to maximize his votes? To
answer this question we need a rule of calculation. The Down-
sian rule is that each voter votes for the candidate whose loca-
tion is closest to his or her ideal point. 19
We can now answer the questions posed in the previous
paragraph. Candidate R should snuggle up infinitesimally
close to the right-hand side of L. That way, R gets all the votes
to the right of l and, since l is to the left of the midpoint of the
voter distribution, that means that R gets more than half of
all the votes. 20 That is, R gets the 375 votes from voters of
types 3, 4, and 5; L gets the 250 votes of types 1 and 2. Put
more generally, L's location divides the electorate into two
groups: those with ideals less than l and those with ideals
greater than l. K s optimal response is an r just next to l on
the side of the larger group. We have thus figured out how R
will respond to any move made by L . L thus knows that her
position will divide the electorate into two groups and she will
get the smaller group. Given that she, too, wants to maximize
votes, she should try to make this smaller group as large as
possible. She can do this, the reader may have guessed, by set-
ting l equal to the ideal point of the median voter, since the
groups to the left and right would then be equal in size. If l
and r just straddle the ideal point of the median voter, x3, then
each location is optimal against the other's and the election
ends in a virtual tie.
We draw precisely the same conclusion if we fix K s posi-
tion first and let L respond optimally. For any r chosen by R, L
will set l just next to r on the side of the larger group. Under
19
For any two candidate positions, say, a and ti in [O, 100), where a is to the
left of~. the midpoint is (a + ~)/2. The candidate located at a receives the
votes of all voters with ideals to the left of this midpoint, whereas the can·
didate located at ~ gets the votes of all voters with ideals to the right of this
midpoint. Voters at the midpoint are indifferent between the two candi-
dates, since their positions are an identical distance from these voters'
ideals; these vote·rs tlip coins to decide for whom to vote.
20 If l happened to be to the right of the midpoint of the voter distribution,
then R would maximize his votes by squeezing up against Lon its left side,
thereby getting a majority of the votes.
these circumstances, the best R can do is to "move to the

median."
Finally, suppose L a nd R n1ust announce their policy plat-
forms simultaneously. Once a policy is announced, if a candi-
date is "stuck" with the position for the duration of the
campaign, then he or she is likely to worry that his or her po-
sition is vulnerable. A position is vulnerable if the opponent's
position lies between it and the median of the voter distribu-
tion, since, by the Downsian rule of calculation, the opponent
will then get more than half the votes. 2 1 The only position that
cannot be vulnerable is one that actually is at the median
ideal. If, on the other hand, candidates are not stuck with
their announced positions but can revise their policy plat-
forms during the course of a campaign, one of two patterns
will be observed. If both initial announcements are on the
same side of the median ideal, then there will be a "leapfrog-
ging" converging pattern as the vulnerable position (as just
defined) leapfrogs over her opponent's position in order to be
closer to the median, that position in turn is leapfrogged over
by the now vulnerable opponent, and so on until there is no
more leapfrogging to do-namely when both positions have
converged upon the median. If, on the other hand, initial an-
nouncements are on opposite sides of the median ideal point,
then there will be a homing in on the median from each side
as the one more distant moves closer.22
In all of these circumstances, each a slightly different mod-
eling assumption about the sequence in which various events
take place in the course of a campaign, there is a common con-
vergence on the ideal point of the median voter. And this cen-
tripetal tendency is precisely what is predicted by Black's
21 Indeed, a position is vulnerable if the opponent's position is closer to the

median. The position closest to the median wins more than half the votes.
22 Of course, if politicians are perfectly informed about voter preferences, they
won't need to proceed tentatively toward the median, whether by leapfrog-
ging or homing in; instead, they can move directly to the median.
Median-Voter Theorem. In effect, Downs's model provides a

rationale for why majoritarian politics is centripetal. 23
The logic of Black's theorem, as elaborated in the elector al
context by Downs, reminds me of those occasions when some-
one says something very intelligent and quite obvious (once it
is said!), causing you to reflect, "Now why didn't I think of
that?" Downs was motivated by the fact that so many foreign
observers of American life had, since practically the beginning
of the Republic, noted how similar America's political parties
were: "Tweedledum and Tweedledee," empty bottles differing
only in their labels. 24 More recently, observers of British poli-
tics have begun to notice that the losing party ultimately
transforms itself to look at least a bit like its more successful
opponent. Thus, in the postwar period, British Turies have ac-
cepted a good dleal of the welfare state championed by the
Labour Party, whereas, in the latter part of the twentieth cen-
tury, Labour has trimmed its more socialist sails in order to
look to voters a bit more like Margaret Thatcher's and John
Major's Conservative Party. In the twenty-first century, both
parties struggle toward the center, neither bearing the ex-
tremist trappings that once described them.
The centripetal forces Downs identified are certainly plau-
23 Notice that the rationale is not that the middle is "where the votes are."
Certainly this may be true; in many circumstances the middle of the spec·
trum is where most persons' preferences lie, with the numbers getting
smaller as one moves toward the more extremist tails of the distribution.
But go back to Figure 5.10 and suppose that the extremists are the more
plentiful. That is, suppose types 1 and 5 have 250 voters each, types 2 and
4 have 62 voters each, and type 3 consists of a single voter. Will the Down·
sian logic recounted above be any different here? No. The centripetal pull is
the same, even though the "center" is least populated wilh voters!
24 It might interest the reader to know that Downs's book originated as a
doctoral dissertation in economics at Stanford University, where a member

of Downs's dissertaion committee was Kenneth Arrow. So, Downs had both
cycles and instability a la Arrow's Theorem on one side and their opposite-
stylized facts about stable party configurations-on the other. His research
sought to make sense of these seemingly incompatible matters. Single-
peakedness did the job.
FIGURE 5.1 1
T gets x votes
L, R get y/2 votes each
T L, R
1 2 3 4 5
o ---~~~~~~~~~~"-~~~~~~~~~~~
100
x m· y
sible, yet it is clear that parties do not converge all the time.
Why n1ight this be? Downs's spatial model is quite user -
friendly as a "discovery tool," so we can vary some of its as-
sumptions and see what happens. Suppose, for example, we
do not foreordain that there are two candidates. What if Leftie
(L) and Rightie (R) are not the only two kids on the block?
There is a third candidate-call her Trey (T)-who may enter
the race if she thinks she has a chance. Well, if L and R locate
at the median (call it m*)- l ~ r ~ m*-an d if, when there are
more than two candidates, the one with the most votes (not
necessarily a m ajority) wins the election, then T certainly
does have a cha nce. She can locate close on one side or the
other of the median, win nearly all the votes on that side, and
thus defeat Land R, who end up splitting the remaining votes
(Figure 5.11 ). On the other hand, if the positions of L and R
are sufficiently widely dispersed. then T can enter between
them at some position t. She will get the votes of voters whose
ideal points lie in the interval [(l + t)/2, (t + r)/2]. T he left
boundary of this interval is the midpoint between the posi-
tions of Land T, whereas the right boundary is the midpoint
between the positions of T and R. By the sa me Downsian r ule
of calculation, L gets all the voters in the interval (0, (l + t)/2],
and R gets all the voters in [(t + r)/2, 100). If there are more
voters in the first interval than in the second or in the third,
then Twins. So, when there is the possibility of entry, L and R
can locate neither too closely together nor too far apart.
In fact, there may be a set of entry-deterrence locations for
T and R, with tlhese two getting roughly the same number of
votes, and no third candidate able to locate in any place that
would give her a victory (thereby discouraging her from enter-
ing at all). The point here is that when we broaden Downs's
initial model to take account of some factor he had omitted-
the possibility of entry by a third candidate-we discover that
there may be occasions and circumstances in which the estab-
lished parties (T and R) are ill-advised to converge toward the
median.
Research has, in fact, been conducted on precisely the issue
of Downsian candidate competition with (prospective) entry. 25
As noted, it is clearly one extension of the original Downsian
assumptions that produces the possibility of nonconvergent
candidate locations. But there are other possibilities. Candi-
dates, for instance, may have their own policy preferences,
ones often known to the voters. Thus, suppose T and R have
their own policy ideal points at l* and r*, respectively (shown
in Figure 5.12). They may declare policy programs at other lo-
cations, say, l .. l* and r .. r*. But why should the voters
believe these policy declarations? It's not necessary to be alto-
gether cynical to believe that once one of them wins, she or he
will be sorely tempted to implement her or his preferred policy
(l* or r*), not declared policy (l or r); politicians cannot be
trusted to do what they say when they have preferences of
their own. Effectively, then, candidates once again will not
converge, this time because there is no point in doing so (they
25 The interested reader may consult Kenneth A. Shepsle, Models of Multi-

party Electoral Competition (Chur, Switzerland: Harwood, 1991), for a sum-
mary of much of this research.
FIGURE 5.12
/" I m· r r·
won't be believed by voters), even if they were willing to im-

plement what they promised. The "commitment technology" is
simply not up to the task.
What if it were? What if candidates had policy prefer-
ences, as in the previous paragraph, but had available to
them means of making promises stick. Perhaps all they need
to say is "Cross my heart," and the voters will believe them.
Perhaps voters believe policy promises because they know
that politicians know that a reputa tion for deception and mis-
representation is a serious electoral obstacle in future elec-
toral campaigns. So, for any of a n umber of reasons, suppose
tha t candidates' promises are credible on the one hand, but
that candidates still care about what policies are implemented
on the other hand. What will the candidates do in this circum-
sta nce? In a lovely pa per, Randall Calvert26 demonstrates
that, just as in the case where candidates didn't care a whit
about policy, these two candidates will converge to the median
voter's ideal. Referring again to Figure 5.12, L wants a n out-
come closest to l* and R wants t he final policy to be closest
to r*. If these two points happen to be equidistant from m*,
~i; Randall Calvert, "Robustness of M ultidimcnsional Voting Models: Candi -
da Le Motivations, Unce rtai nty, and Convergence," American Journal of
Political Science 29 Cl 985): 69-95.
and if each candidate (credibly) announced his or her ideal

policy, respectively, then the election would end in a tie (and,
presumably, the winner would be determined by something
like a flip of a fair coin). But L, by moving just a tad toward
the center, could win the election outright at a very small cost
to herself in terms of policy. But this would be terrible for R-
not that he lost the election but that a policy near l* is so
awful. He could avoid all this by moving in toward the center
a bit more than L had, which, in turn, encourages L to move
in a bit more, a nd so on. In the end, even though both candi-
dates had policy preferences and, in fact, did not care at all
about who won the election but only about what policy would
be implemented, they converge to the median voter's ideal
anyhow.
Needless to say, we could play with Downs's model in a va-
riety of interesting ways. Many have. 27 What has been shown
in this section is that the stripped-down spatial model of
Downs, with competition on an ideological dimension between
two election-oriented candidates, leads to policy convergence.
The policy that emerges from the competitive forces captured
by this model is the ideal point of the median voter. This re-
sult, to the casual observer, describes what often happens in
real elections, as candidates try to smooth down their more
extremist edges in order to curry favor with voters in the cen-
ter of things. Thus, once Bill Clinton vanquished his liberal
opponents within the Democratic Party in 1992 (Jesse Jack-
son and Mario Cuomo), he headed toward the ideological cen-
ter, running in the general election as a more conservative
27
For two eady summaries of the extensive literature the interested reader
may turn to James M. Enelow and Melvin J. Hinich, The Spatial Theory of
Voting (New York: Cambridge University Press, 1984); and James M.
Enelow and Melvin J. Hinich, eds., Advances in the Spatial Theory of Voting
(New York: Cambridge University Press, 1990). A more recent review is
Thrun Dewan and Kenneth A. Shepsle, "Economic Models of Elections," An-
nual &view of Political Science, forthcoming 2011.
"new Democrat." The incumbent president, George H. W. Bush,

on the other hand, tried to shed some of his hardline conserva-
tive attributes, also moving toward the center as he compro-
mised on his "no new taxes" pledge. In 2000, the Republican,
George W. Bush ran as a "compassionate conservative" against
the centrist Democrat Al Gore. In 2008, the Democrat, Barack
Obama moderated his views in the general election, after se-
curing the nomination in a close contest with Hillary Clinton;
the Republican Party nominated its most 1noderate candidate,
John McCain. In many other elections one sees a similar
dynamic-partisan candidates of the left and the right hedg-
ing, qualifying, and compromising in order to appear more
centrist.
This convergence is not always complete, however. Some-
times a candidate applies brakes on convergence for fear of
alienating his or her base, or even stimulating a third-party
entrant. Thus, civil rights activists, unions, and government
workers-elements of the Democratic base-made it virtually
impossible for Walter Mondale to converge toward the center
as a candidate in the 1984 presidential election. Elements of
the conservative movement kept Ronald Reagan ideologically
true in that same election. Third-party candidates entered the
presidential races of 1968, 1980, and 1992 (George Wallace,
John Anderson, and Ross Perot, respectively), sometimes be-
cause the candidates were thought to have converged too
much, 28 sometimes because they were thought to have stayed
too close to their more extremist supporters. 29 Thus, both too
much convergence and too little convergence may provide the
impetus for a third-party challenge.
28 In 1968, Wallace entered on the right, thinking "there's not a dime's worth
of difference" between the Democrat Hubert Humphrey and the Republican
Richard Nixon.
29 Both Anderson and Perot sought to capture the center, which they believed
had been conceded by Carter and Reagan in 1980 and Clinton and Bush in
1992, respectively.
I have clearly only scratched the surface of Downs's spatial

model of party competition a nd only covered some of the many
mechanisms a nd rationales according to which competitors
converge toward the median voter's ideal policy on the one
hand, or maintain distinctive policy profiles on the other. This,
in sum, suggests the richness of Downs's approach.
Electoral phenomena, however, are not the only focuses of
the spatial model. A twin enterprise, a kind of "elections writ
small," has employed the spatial model to study the selection
of policy in legislative settings. I turn to those now, and exam-
ine both one-dimensional and multidimensional versions of
the spatial model.
Spatial Models of Legislatures

There will be a much more thorough look at legislatures
in Part IV, so here I am interested primarily in seeing what
the spatial mode l can do. It turns out to be quite a powerful
analytical tool for r epresenting the ways in which preference-
based (rational) behavior and structural features of institu-
tions interact to produce final outcomes. It suggests that
legislative outcomes depend in essential ways not only on
what legislators want but also on how they conduct business
in the legislature.
To keep things as simple as possible, the legislature is
taken to be a set of n individuals, where n is an odd number,
and where everyone casts a vote. It makes decisions by major-
ity rule. The most elementary situation, one that is examined
first, is th e unidimensional case in which the legislature must
choose a point on a line. Each legislator, i, has an ideal point
xi, and single-peaked preferences. The median voter is legisla-
tor m with ideal point xm. We know in this circumstance that
xm can defeat any other point on the dimension in a majority
contest (Black's Theorem). Perhaps more amazing is the fact
that the median preferences prevail in a comparison between
any two alternatives, so that if m prefers x to y then so does a

majority for any x and y.3o
In addition to the preferences of the median legislator, xm,
two other distinguishing features of the situation are impor-
tant. Whenever a legislature faces a decision-makin g opportu-
nity, there is always a status quo in place, labeled xO. This is
the current policy at the time of legislative choice. It remains
in place if the legislature chooses not to change it. 31 The sec-
ond feature of interest common to most legislatures is a
division-of-labor arrangement known as a committee system.
In such a system, a committee is a subset of the n legislators
(momentarily, we describe some of its specific powers). The
median ideal point of the committee members is labeled xc.
Just as majority preferences in the entire legislature are
identical to the preferences of the legislature's median voter,
majority preferences inside a committee are a copy of the pref-
erences of the committee's median member. Because of these
identities, much of our analysis need only consider x!', xm, and
xc. In what follows, then, I put the spatial model through
30 This may be proved as follows. Suppose x , x, and y are all points in the di-
mension. and that xms.xsy. Legislator m clearly prefers x to y. But then so
does every legislator t-0 the left of m. Together these legislators constitute a
majority, so xis pl'.'eferred by a majority toy. Likewise, by the same reason-
ing, if ysxsx.,, then both m and a majority (all the legislators to the right of'
m) prefer x to y. So, it has been shown that whenever x and y are on the
same side of the median, a majority a lways agrees with the preferences of
the median voter. Suppose, then, that xsxmsy. and that m prefers x to y.
Consider the legislator just to the left of m. Her ideal is closer to x and far-
ther from y than was m's ideal; so if m prefers x to y, surely she does, too.
But then. so do all the other voters to the left of m's ideal, and once again
they jointly comprise a majority. This establishes that a majority always
agrees with the preferences of the median voter.
31 In some circumstances, the status quo policy does not remain in place, un-
less the legislature takes positive action to keep it in place. If the legisla-
ture fails to do anything, then the status quo reverts to some specific policy
(known, naturally enough, as the reuersion point). This is true, for instance,
in statutes that possess sunset prouisions, an example explored further in
Case 5.1.
its paces in examining the making of policy choices by an

n-member legislature possessing a committee system.
Three decision-making regimes, or institutional arrange-
ments, are identified. The first is pure majority rule. There
is a status quo, and any legislator can offer a motion to change
it. A motion, once proposed, is pitted against the status quo. If
it wins it becomes the new status quo; if it loses it goes to the
place where all losing proposals go (a sort of elephants' burial
ground). The floor is once again open for some new motion
. (against the old status quo, if it survived, or the new status
quo, if the previous proposal prevailed). This procedure of mo-
tion making and voting continues until no member of the leg-
islature wishes to make a new motion. 32 ·
The second regime is the c losed-rule committee sy stem.
In this system, a (previously appointed) committee first gets
to decide whether the legislature will consider changes in the
status quo; that is, it has gatekeeping agenda power and can
decide whether to open the gates to enable policy change or
not. Second, if the gates are opened, only the committee gets
to make a proposal (monopoly proposal power). Third, the par-
ent legislature may vote the committee's proposal either up or
down. If it passes, then it becomes the new status quo; if it
fails, then the o~d status quo prevails. The proposal is closed
to amendments. Hence, the proposal is said to be considered
under a closed rule, and the committee is said to offer its par-
ent body a take-ii-or-leave-it proposal.
The third regime is the open-rule committee system.
This system is identical to the one described in the previous
paragraph, except for the third feature. Under an open rule,
once the committee has made a proposal, the parent legisla-
ture may open the :floor to amendments to the committee's pro-
82
A variation on this "stopping rule" is to allow a motion to be in order at any
time to close the floor to new motions (in effect, a motion to take a final vote
and then to adjourn the legislature, at least on the subject matter at hand).
posal. Once the committee has opened the gates and made a
proposal, it concedes its monopoly access to the agenda.
Each of these systems is explored in both the one-
dimensional and the multidimensional setting in order to de-
termine whether there is anything regular or routine that we
can expect from these alternative majority-rule regimes. Some
brief comparative observations on these regimes are offered,
leaving a full-blown consideration for chapters 11 and 12,
where institutions are taken up more systematically.
PURI<: MAJORITY RULE I start with a legislative choice in a one-

dimensional spatial setting. In the pure majority-rule regime
there are no committees, so we need to know only the loca-
tions of xm and xP. A typical situation is given in the top panel
of Figure 5.13. From the median ideal and the status quo, we
determine the median legislator's preferred-to-xO set, P m(x0). 33
Since we've established that the median's preferences are the
same as a majority's preferences, this interval is the set of mo-
tions that would prevail over x 0 in a majority contest. So, if
someone is recognized and makes a motion outside this set, it
will go down in flames, whereas any motion inside this set will
be victorious and become the new status quo. It is evident that
the political process defined this way will produce outcomes
that either leave the status quo unchanged or move it closer to
xm (since every point in P m(x 0) is closer than xO to xm). As the
process of motion making and voting is repeated, the winning
alternative will ultimately converge on xm. Moreover, it will
not depart once it reaches xm (since, as shown above, the sta-
tus quo cannot move further from xm in any vote). So, just as
in the Downsian model of electoral competition, there is a cen-
tripetal tendency in the pure majority-rule legislative regime.
33 Since we assume t hat legislative preferences a1·e distance-based, we know

that legislator m prefers to :xfJ all points closer than it to xm. This determines
the set Pm(:xfJ) pictured in Figure 5.13.
FIGURE 5.13
0 1
1
EQUILIBRIUM
OUTCOME
It is for this reason that we think of pure majority-rule legisla -

tive choice as an "election writ small."
The great utility of these spatial models of legislative
choice is they permit the analyst to do what in economics is
called comparative statics-we can ask "what if" questions.
Having derived an equilibrium outcome from our basic setup,
as I did in the previous paragraph, we may now ask how
that equilibrium changes as relevant parameters change. We
have already seen that there are really only two relevant
parameters- x0 and xm. Holding the latter fixed, we first' ask
what happens if the former changes-that is, what would hap-

pen to the final outcome if the status quo were closer to or fur-
ther from the median legislator's ideal? The answer is:
nothing! Although different locations for :A,.o will affect P m(.xO),
and hence what motions can succeed at any point in time, we
know that ultimately the result converges to xm, no matter
what .xO is. So one interesting conclusion we draw from the
pure majority rule model is that it does not possess a con-
servative bias, weighting past decisions unduly. The past (as
reflected in the status quo) will influence the "path" (by re-
stricting what motions can succeed at different stages of the
process) but will not affect the ultimate destination.
Reversing emphasis and holding .xO fixed, we now ask what
happens if xm were different. The answer is graphed in the
lower panel of Figure 5.13 for xm located on a line between 0
and 1. The location of xm in the [O, 1] interval is given on the
horizontal axis and the equilibrium outcome in this interval is
given on the vertical axis. The graph is a 45° line showing (for
any .xO) that the equilibrium outcome perfectly tracks the iden-
tity of the median ideal point. This is centripetality in the ex-
treme! Not only does pure majority rule legislative choice
converge to the median ideal, but if that median should
change, then so will the equilibrium outcome. So a second
interesting conclusion we can draw from pure majority rule
is that it is perfectly responsive to central tendencies: The
median legislator's ideal is, by definition, the central point
in the distribution of preferences; pure majority rule produces
an outcome at this point; and, were this point to change (as a
result, say, of an election), the legislative outcome would
"track" it.
CLOSED-RULE COMMI'ITEE SYSTEM Most legislatures are not

pure majority-rule institutions. Even town meetings and other
approximations of pure majority rule about which observers
occasionally wax romantic require some mechanism to deter-
mine the content of agenda items and the order in which they
will be taken up. Some legislatures establish a single agenda
committee to decide these matters. However, most legislatures
(certainly in the United States) employ a division-of-labor
comn1ittee system that divides up agenda power by policy
area. Subsets of legislators have disproportionate influence
over the agenda in specific policy jurisdictions. The committee
serves, in its jurisdiction, as a n agenda agent for its parent
legislature.
I will have much more to say about these things in Part IV.
For now, I need focu s only on the fact that what distinguishes
the closed-rule regime from pure majority rule is that ther e is,
in addition to xfJ and xm, a third pa ra meter of interest, namely
the median ideal point of an agenda-setting committee, xc.a4
Many of th e conclusions drawn about this regime depend on
the relative locations of :xfJ, xc, and xm.
The decision-making procedure, as suggested earlier, is for
the committee either to make no proposal at all, in which case
xO remains in place, or to make a motion to change the status
quo, which the parent body must accept or reject as is. What
will such a committee do? To answer this question, we once
again determine P m(x'Y), as in the top panel of Figure 5.13.
This is a set whose boundary points are xfJ itself and x*; it con-
tains the only points a legislative majority prefers to xfJ. The
committee, as personified by its median voter, c, treats these
points as its "opportunity set," picking its favorite as the mo-
tion it makes (if it makes a ny motion at all). We look at three
orderings of the relevant parameters (there are six orderings
in all, but the omitted ones are simply mirror images of the
ones we conside r):
34 Since the one-dimensional model is being elaborated here, we are concerned
with the medjan ideal of only a single committee. In multiilimensional con·
texts, where there arc many jurisdictions into which the dimensions of t he
policy s pace are arranged, we will need to know the policy preferences of
different committees, each responsible for its own bundle of policy dimen-
sions. More on this will be developed in Part IV.
CAS£ 1 (.x0sxmsxc). Here the median legislator is between

the status quo and the median committee member. In this
case x* is the right boundary of P 111 (x0), just as shown in
Figure 5.13. If xcsx*, then the committee will propose its
median ideal point, which then will be approved by a leg-
islative majodty (since it lies inside P m(.xO)). If, on the other
hand, x*s xc, then the best the committee can do is to pro-
pose x*, which is approved by a legislative majority.:i5 In ei-
ther case, both committee and parent legislature wish to
move away from the status quo in the same direction. The
final outcome will move x 0 in that direction, further than
the median voter would want, but not always as far as the
committee median wants.
CASE 2 (x0 sx1.sx,,,). Here the median committee member is
between the st atus quo and the median of the whole legis-
lature. In this case xcEPm(xO) automatically. So the com-
mittee can get majority legislative approval for xc, the
committee's best outcome.
C ASE 3 (xmsxOsx.J. In this last setting the status quo is be-
tween the two medians. This is a particularly interesting
case because committee and legislative majority are at log-
gerheads . The committee wishes to move right, while a
majority of the parent legislature wants to move left. The
committee's gatekeeping authority pays off for it in a big
way here because it will choose simply to keep the gates
closed. 36
35 Actually, a legislative majority is indifferent between x 0 and x*. I assume

that an indifferent voter votes for the motion on the floor. (Alternatively. the
committee could propose a point just to the left of x*. which secures a ma·
jority outright.)
36 The committee could move a proposal (some point to the right of l-'°). but it
would be defeated. So it might as well not bother and s im ply keep the gates
closed (especially if the bother were at all costly). On the other hand, if out-
So, the first thing we learn about the closed-rule regime is

that only a very limited number of things can happen- three
things, in particular. If xe is interior to the legislative median's
preferred-to-x<> set, then the outcome is xe. If it is not, then ei-
ther of the two endpoints of Pm(x<>) are possible-x<> if commit-
tee and legislative median are at loggerheads; x* otherwise. In
the pure majority rule regime, in contrast, only one thing can
happen: xm, something that never happens under the closed-
rule regime (unless, by coincidence, xm=xc or xm=x0). This sug-
gests that endo·wing a privileged group with agenda power is
not without its consequences: agenda power discourages cen-
tripetal outcomes as it tugs the process in the direction of the
privileged group.
A variety of comparative statics exercises exists that one
might do. I focus on one: For a fixed legislative median and
committee median, what happens as x0 changes? (Whoever
asks this question doesn't literally mean that the status quo
suddenly changes. Rather, it is a question of what would hap-
pen if the status quo were more or less extreme.) In case 1, for
example, if x<> were further to the left, then Pm (.xO) would get
bigger (x* moves to the right). At some point it contains xc (if it
doesn't already). So, as x<> moves away from the chamber me-
dian, there will be a discontinuity when xc jumps from being
outside m's preferred-to-x<> set to inside that set. Put crudely,
the worse the status quo, from m's perspective, the more likely
c can get her way.37 The same pattern prevails as x0 moves to
the right. At first it moves toward xm, so m's preferred-to-x0 set
side interests took heart in the fact that the committee was at least putting
up a good fight and rewarded the committee accordingly, then the commit-
tee might wish to ''bother" (though the result would be uncbanged-xO
would stay in place).
37
The classic statement of this result, plus a derivation of some of the political
consequences of it, is found in Thomas Romer and Howard Rosenthal, "Po-
litical Resource Allocation, Controlled Agendas, and the Status Quo," Public
Choice 33 (1978): 27-43.
contracts. Once it "passes" xm, the preferred-to set begins ex-

panding again. 38
CASE 5.1
S UNSET P ROVISIONS AND
Z ERO -BASED B UDGETING
In the 1970s public policy analysts developed two ideas as

an attempt to counter rising budget pressures. The first
idea was called a sunset provision. The identified problem
was the persistence of expenditures that might have out-
lived their usefulness. It seemed that once a project was on
the books, it never went away. With a sunset provision as
part of the enabling legislation, the project would have to
be renewed after a specified time period in order to extend
its life. In other words, the sun would automatically set on
a project unless the legislature took further action.
The second idea was called zero-based budgeting. Also
associated with the problem of expenditures that were
growing out of control, this concept required bureaucratic
agencies to justify a project budget "from zero," rather than
merely justifying the growth in proposed expenditures over
the previous year's budget. It was alleged that this proce-
dure would reduce accumulating and persisting inefficien -
cies in agency budgets.
For our purposes, sunset provisions and zero-based
budgeting are similar because they create situations in
which the status quo alternative to a proposal is zero. Con-
sider the case of zero-based budgeting for an agency. The
legislative median is at xm and last year's agency budget is
at B. Now we let the agency make a proposal for next year's
aa The reader might try to see what happens as .t.o changes in cases 2 and 3
above.
budget, a proposal which the legislature may accept or re-

ject by majority rule. Under ordinary procedures, we as-
sume that legislative rejection of the agency proposal
results in last year's budget, B, continuing in place next
year. Under zero-based budgeting, on the other hand, we
assume that legislative rejection leads to a zeroing out of
the agency budget altogether.
At first glance, it would seem that the zero-based budg-
eting procedure is pretty tough on the agency. That's the
whole idea, since this method was designed to limit an
agency's power in budget negotiations. However, zero-based
budgeting actually increases agency discretion. This is seen
in the figure below. If, under ordinary procedures, Bis the
reversion outcome (the outcome if the legislature rejects the
agency proposal), then as long as the agency proposes a
budget in the gray region, the legislature will approve it
(any such proposal is in P"'(B)); that is, under ordinary pro-
cedures, the agency could extract a budget as large as B*.
If, on the other hand, the zero-based budgeting procedure
were in effect, then the agency could get a budget as large
as N (since N is in Pm(O)). In their pure form, both sunset
provisions and zero-based budgeting provide perverse in-
centives. Their whole rationale was to discipline "out-of-
control" agencies and budgets, not empower them. (The
reader might explore the consequences of a zero-based
budgeting regime in which agency proposals may be
amended by th e legislature.)
0 B x,,. B* N
In concludi ng this brief treatment of the closed-rule

regime, let me reemphasize the fact that the key parameters
are xf>, xc, and x,,.. An electoral earthquake that fails to change
relationships among these parameters will not change policy
outcomes (a fact that may puzzle those not equipped with the
theory developed here). If, for instance, a legislative election
caused massive turnover in incumbents but did so symmetri-
cally so as to leave xm unchanged, then "the more things
change, the more they stay the same." Likewise, if before an
election legislators c and m are at loggerheads, as defined in
case 3 above, then electoral change, no matter how massive,
that leaves the (possibly newly determined) c and m at logger-
heads will simply maintain the status quo ante. The institu-
tional impediments implicit in the closed-rule regime stand in
stark contrast to the hypersensitivity of pure majority rule.
OPEN-RULE COMMIT'l'EE SYSTEM We've seen thus far that al-

though there is an entire continuum of possible final outcomes,
only one thing (x111) can occur under the pure majority-rule
regime, and only one of three things (xc, ,_..o, or x*) can possibly
happen under the closed-rule regime. In the following treat-
ment of the open-rule regime, it will be seen that only two pos-
sibilities exist. Eit her the gates remain closed and xfJ prevails
or the gates are opened and xm is the final outcome. Nothing
else is possible. We will consider all the cases a s we did in the
. .
preVIous regime.
ln the open-rule regime the committee once again has the
first move. If it makes no motion, then xfJ persists. If it makes
a motion, then that motion is open to amendment (hence the
term open rule). It is assumed here that alternative amend-
ments continue to be offered until no legislator wishes to offer
another. So the committee proposal is initially pitted against
the first a mend ment, the winner of that against the next
amendment, and so on until all the amendments have been
taken up; the survivor of that sequence is then pitted against
the status quo (this last vote is often ca lled the "vote on final
passage"). 39
:s9 An alternative procedure would be to allow, after a committee motion, an
amendment that is directly voted on. The winner stays on the floor and is
FIGURE 5.14
This procedure looks very much like the pure majority rule
regime, except that the committee has the first move. Once it
opens the gates, we're in the world of pure majority rule. This
means that once a proposal is made, it will be amended and
amended again, with successful amendments converging the
process toward xm. Indeed, it doesn't even matter what the ini-
tial committee proposal is. The reality is:
open the gates :::::> xm
keep gates closed :::::> :xfl
The commit.tee decision is really pretty simple. If the com-
mittee median voter, Ms. e, prefers x"' to x 0 , then she makes a
motion (any motion); if Ms. c prefers xD to xm, then the commit-
tee keeps the gates closed. Thus, all we need to inspect is Ms.
e's preferred-to-x0 set, Pc(:xfl), to see whether xm is in it or not.
Recall th e t hree possible cases in the preceding section.
For the parameter ordering of case 1 (:xflsx"'sx), the commit-
tee clearly prefers xm to x 0 , so it will open the gates. For case 3
(xms:xflsx), the committee clearly has the opposite preference,
so it will keep the gates closed. It is the case 2 ordering
(:xflsxcsxm) that is the interesting one. If e's ideal policy is less
than halfway between x 0 and xm, then she keeps the gates
closed; if it is more than halfway, then she makes a motion.
The first of these case 2 situations is shown in Figure 5.14.
What makes this especially interesting is that it represents a
subject to another amendment. The process continues until no more amend-
ments are forthcoming, after 'which there is a final vote between the alter-
native left standing on the floor and the status quo.
very frustrating situation. The committee, because it prefers

the status quo to the median legislator's ideal, will keep the
gates closed. But both a committee majority and a chamber
majority prefer every point in Pc(x0) to :>!>. That is, the open-
rule environment, which at first blush appears to give a leg-
islative majority potent authority, in fact penalizes both
committee and legislative majorities. It gives the chamber too
much authority-the right to amend whenever it wants. Its
strength is its weakness, because it cannot promise not to use
its authority; yet, it would be better off if it could credibly
promise not to amend some proposal in P/xfJ) made by the
committee (for, if it could precommit in this fashion, then the
committee would be prepared to open the gates).
CASE 5.2
THE I MPORTANCE OF COMPROMISE AND
STRATEGIC THINKING
The discussion of the closed-rule and open-rule regimes

addresses the general question of how politicians and in-
terested others think about legislative possibilities. In
one-dimensional situations, as we have seen, politicians lo-
cate a proposal on the policy dimension relative to their
own preferences and the status quo that will otherwise pre-
vail if the proposed change is rejected. When faced with a
choice between the proposal and the status quo, the politi-
cian votes for the alternative closer to his or her ideal. With
the open-rule regime, once a proposal is made, the dynamic
of amendment activity leads inexorably to a unique out-
come-the ideal point of the median legislator. With the
closed rule, a proposal wins only if it is closer than the sta-
tus quo to the median voter's ideal.
Failure to recognize these dynamics can lead to disap-
•
pointment for principled-that is, stubborn- lobbyists. Ad-

vocates of proposed legislation must take into consideration
the preferences of the decision maker(s), the rules of proce-
dure in effect, and the relative location of the status quo. An
unwillingness to compromise in light of these strategic rea)-
ities can keep the status quo in place, even though the pos-
sibility exists to defeat it with results satisfactory to the
lobbyist. This is illustrated in the figure below, where the
legislative median is xm, the status quo is :xfJ, and a powerful
lobbyist's ideal policy is I.
c I
Without going into any of the specifics concerning the
way powerful lobbyists exercise their power, suppose that
lobbyist I is in a position to undermine any change in the
status quo if it finds the change not to its liking (perhaps by
"bribing" influential legislators- that is, contributing to
their campaign committees). Under the closed rule, the best
I could hope for is the compromise point, C-a policy just a
little bit closer than :xfJ to xm. If the lobbyist stubbornly
refuses to accept C by seeking something . more extreme,
it loses. Under the open rule, it must be prepared to ac-
cept xm, for Uris is where the process of amendment will
drive the final result. In either of these cases, the lobbyist
must be able to anticipate the best deal it can cut and settle
for it. In particular, it must be especially sensitive to the
fact that "the best deal it can cut" depends upon the proce-
dural rules for amendments. Even though it is powerful
enough to undermine proposed changes in :xfJ, it cannot im-
pose its own "'·ill. It needs a little help from its (legislative)
friends.
Some observers have cited the absence of such strategic
thinking as a reason for the failure of the Equal Rights
Amendment* ahd Proposition 174 in California, which

would have implemented school choice as a voucher sys-
tem.t Unwilling to compromise, lobbyists unwittingly kept
their proposals further from the status quo than the com-
promise point required by the strategic realities. Politicians
or voters voted against their proposals when more moder-
ate versions very probably would have passed. The impor-
tance of such strategic thinking is obvious after the fact but
not always in the heat of battle.
*Jane Mansbridge, Why We Lost the ERA (Chicago: University of Chicago

Press, 1986).
t Jack Anderson and Michael Binstein, "'School Choice' Pitfalls in Califor-
nia," Washington Post, November 1, 1993.
MULTIDIMENSIONAL Once multiple dimensions

EXTENSIONS
come into play, matters get a bit more dicey. In a pure
majority-rule regime, the results of the McKelvey Chaos The-
orem loom large. Putting to one side the highly unlikely cir-
cumstance that legislator preferences are distributed in a
radial symmetric manner, we know that W(x)"'0 for any x in
the policy space. Anything can be beaten. In particular, any
status quo, x 0 , has a nonempty winset, W(x0). As long as a mo-
tion is made from that set, the status quo will be replaced. But
then x 1 E W(xO), in turn, has a nonempty winset of its own,
W(x1). A motion x 2 E W(x 1) will replace x 1. Under the assump-
tions made about legislative voting, 40 an existing status quo is
continually replaced.
Suppose the setup is a ltered ever so slightly. The condition
from pure majority rule still holds that anyone is free to make
a motion to change the status quo. But let's assume that deci-
sion making takes place one dimension at a time in some pre-
set order. The first person recognized to make a motion on the
40 Namely, that everyone votes his or her preference rather than voting strate-
gically (which is taken up in the next chapter).
FIGURE 5.15
xc2 - - - - - - --- -- ----- - -- -- -- - -- - • ...xc

'
'
Xa2 ~ -------------- -------------- .J,f."! ________ ____ ~a
x b2 ~ - - -- - - - - - -- ... xb
''
'
xc 1 x a1 1
initially designated dimension states his or her amendment to

xD; this amendment can only change xD on the dimension cur-
rently under consideration. The group continues to focus on
amending the status quo on this dimension until no more
amendments are offered. Once it completes its task, the group
turns its attention to the next dimension. It continues in this
manner until th ere is no dimension left on which any legisla-
tor wishes to alter the status quo level.
It is easy to see in Figure 5.15 that this multidimensional
version of pure majority rule mimics the result of the one-
dimensional setting. There are three legislators with ideal
points xa=(x0 1, x 0 2), xb=(x1,1, xb2), andxc-(x/, x/). For any status
quo (not pictured), x0 =(x0 1, x 02), motions are entertained, first
on dimension 1 and then on dimension 2. At the end of the
day, when all motion-making and voting are said and done,
the final outcome is the multidimensional median, xm=(x/ ,
x0 2), since legislator c is median on the first dimension and
legislator a is median on the second. This does not mean
that W(xm)=0. There are points to its northeast, for instance,
that both a and c prefer to xm. Rather, it means that on any
dimension-say, the first-holding policy fixed on the other di-
mension, no movement away from xc 1 would be supported by a
majority. (For any point on the horizontal line through xm-
the points in which policy on the first dimension changes but
remains fixed on the second-two of the three legislators al-
ways prefer xm to it. The same holds for points on the vertical
line through xm. The only points preferred by a majority to xm
require changes on both dimensions at once.)
Thus, the multidimensional version of pure majority rule
yields one of two possible conclusions, depending upon
whether there is additional institutional structure or not. In
the pure case, the status quo is continuously vulnerable to
change. The group's choices are never very durable, since it is
always in someone's interest to introduce a motion to change
it, and it is always in some majority's interest to comply. In
the case of institutional structure in the form of dimension-by-
dimension decision making, the result is both predictable and
centripetal. The median ideal point on each dimension pre-
vails under the procedure described above (although it need
not be the same median voter on each dimension, of course).
Since I will take up the multidimensional versions of the
open-rule and closed-rule regimes in the chapter on legisla-
tures in Part IV, the discussion will be especially brief on this
subject now. Imagine, in Figure 5.16 (a reproduction of the
spatial positions in Figure 5.13), that Ms. c is an agenda setter
and the status quo is :xfJ. If her proposals are subject to amend-
ment by the parent legislature, then we are back to the wild-
and-woolly open-rule majority system. Under a closed rule,
however, she can make a take-it-or-leave-it proposal, one that
Spatial lvfodels of Majority Rule 141
FIGURE 5.16
~W(x°}
is not subject to amendment but only to an up-or-down vote.

The petal-shaped shaded regions comprise e's opportunity
set-it is W(x-0). The circles centered on xc are various of e's in-
difference contours. Her objective is to move the final policy
outcome onto the indifference contour of smallest radius
(hence closest to her ideal point) that still lies in W(Xl). The
point in this figure (a big black dot!) at the tangency between
one of the petals of W(xO) and the smallest indifference curve
of Ms. c is the proposal she will make, which a majority (a and
c) will then support. With the closed rule, then, a monopoly
agenda setter has considerable power, though constrained by
majority prefer ences.
CONCLUSION
The spatial mode l will be used time and ti1ne again in the
analyses of the remainder of the book. I've put forward the
basic ingredients in this chapter and briefly explored majority
rule in electoral and small group settings. But a good deal of
na·ivete characterizes voter/legislators (apparently, only candi-
dates and agenda setters a re wily). I want to relax this unre-
alistic feature in Chapter 6. In Chapter 7 t he a nalysis moves
beyond majority rule to the multitude of ways human creativ-
ity has manifested itself in devising sometimes bizarre and in-
tricate ways for groups to arrive at decisions.
What is exciting about the spatial model to an analytical
political scientist is the opportunity it affords to capture many
of the interesting details of political competition-whether be-
tween candidates in an election or between alternative mo-
tions in a legislative assembly-and, at the same time, to do
so in a fairly clean and simple manner. We appreciate that the
reader may not a gree entirely with the last sentiment, since
this chapter has required your undivided attention and careful
reading. Nevertheless, political scientists over the past fifty or
so years have found the model to serve as the principal build-
ing block for the analysis of political rivalries of all stripes.
A single chapter in a book, of course, can only portray the
spatial model at its simplest. But even the simple formulation
possesses nonobvious implications. In the context of one-
dimensional pure majority rule with single-peaked prefer-
ences, for example, whether in two-party electoral competition
or legislative policy choice, the magnetic attraction of the me-
dian participant's ideal point is powerful. Majoritarian politics
is subjected to centripetal forces, producing outcomes that ob-
servers describe with words such as "compromise," "moder-
ate," or "centrist." At the very least, then, the simple spatial
model provides a rationale or expla nation for the inexorable
movement of majority-rule competition towa rd the center of
.
participant preferences. Its surprise value lies in the fact that

this centripetal dynamic is not because "that's where the votes
are." As we demonstrated earlier, movement toward the me-
dian voter's ideal is an equilibrium tendency in a pure major-
ity rule arrangement even if there are very few voters at the
center of things.
In the legislative realm, where rules of procedure and
agenda-setting committees constrain the operation of pure ma-
jority rule, there were other surprises. At least in the world of
one-dimensional politics, only a limited number of things are
possible. A committee system operating under the open rule
can produce only one of two possible results. It can, by "closing
the gates" and not permitting a motion to be proposed, keep
policy at the status quo. Or, if it should make a motion, the se-
quence of amendments permitted under the open rule will
drive the outcome to the median legislator's ideal. These are
the only possibilities. Under a closed rule, there are three pos-
sibilities. If the gates are kept closed (possibly because the
committee and legislature are at loggerheads, with the status
quo between their respective median ideals), the status quo
remains intact. If the committee median's ideal lies between
the status quo and the median legislator's ideal, then the com-
mittee's ideal will be proposed and will pass. Finally, if the
median legislator's ideal lies between the status quo and the
committee's median ideal, then the outcome is the point clos-
est to the committee's median that leaves the median legisla-
tor of the full legislature just indifferent between it and the
status quo. The details a re found in this chapter. The surprise,
however, is found in the conclusions: first, that only a small
number of items are possible under various legislative proce-
dural regimes and, second, that these s mall numbers of things
differ from regime to regime. P ut differently, institutional
arrangements-the political ways of doing business- matter
profoundly for the outcomes that emerge from the political
process.
The spatial model also allows us to begi n to assemble ex-

planations for wh y convergence to the center is not, always
complete. The centripetal tendency is always present, to be
sure, but, there 1nay be countervailing tendencies as well. In
the context of Downs's model of elections, for example, we
noted that a politician may fear he will lose his extremist sup-
port (to a bstention or to a third-party entrant) if he converges
too much toward his opponent. In the legislative arena, pow-
erful agenda setters may, through their control of motions and
amendments. prevent the process from converging on the me-
dian legislator's ideal policy, either because the agenda setter
can propose and get passed something she likes bette r or be-
cause she chooses to keep the gates closed.
Thus, the great advantages of the spatial model are its
(relative) simplicity, its analytical power, and the "surpr ises"
it produces. Not only do we begin to understand things that
we may have long appreciated in an intuitive fashion (like the
tendency toward moderation in majority-rule systems), we
develop a sophisticated understanding of new things. While
surely not a perfect explanatory tool, it's a pretty good start.
One of the matters that I touched on only briefly a nd un-
systematically was strategic thinking. Many a pplications, es-
pecially early in the history of spatial modeling, a ssume that
voters and legislators are "honest" in their voting behavior.
When confronted with two alternatives, they simply vote for
their favorite, doing so without regard for subsequent conse-
quences. Yet there are many circumstances in which a rational
person will think things through in a more sophisticated fash-
ion, sometimes coming to the conclusion that, in a particular
voting opport unity, she should not vote for her favorite alter-
native. This is sophisticated or strategic behavior, the subject
of the next chapter.
EXPERIMENTAL CORNER
Group Choice under Majority Rule
Imagine a group that must make a decision. For example,
suppose a group of five individuals must choose a point in a
two-dimensional space, one that looks like that in Figure
5.8, where each individual values the points in this space
differently. Ms. 1 most likes points in the southwestern part
of the space; Mr. 3 favors points to the northeast; others
prefer points in other regions of the space. In short, there is
conflict among the preferences of the members of this
group. There are many theories of group choice that at-
tempt to analyze and explain situations like this. In a clas-
sic paper, Morris Fiorina and Charles Plott set out to assess
these theories by running experiments with real, live
groups (mainly of college students, but others have experi-
mented with university employees, business people, towns-
people, etc.).8 In this "Experimental Corner," I will focus on
one of their experiments. This is actually a slight variation
that I run with my own students.
Begin with an odd number of group members. Each is
given experimental preferences by the experimenters. For a
five-member group as portrayed in Figure 5.8, member 1
has an ideal or most-preferred point located at x1, member
2 at x 2, and so on. Their preferences are Euclidean, which
means that they prefer the final outcome to be a point
closer to their ideal point than one further away. Thus, 3,
for example, is indifferent among all the points on a circle
center ed on her ideal point (because they are equidistant
from x3) and prefers any point on a circle of smaller radius
to one on a larger radius (because the former is closer to x3
• Morris P. Fiorina and Charles R. Plott, "Committee Decisions under Ma·

jority Rule: An Experimental Study," American Political Science Review
72 (1978): 575-98.
than the latter is). In the experiment, the horizontal and

vertical dimensions range from zero to 100. A status quo, x0 ,
is set at (100, 100) in the extreme northeast of the two-
dimensional space. Subjects are paid cash depending on the
final outcome-a higher payment for a final outcome closer
to their ideal point and a lower payment for an outcome fur-
ther away. Each subject knows his or her preferences and
payment schedule but not those of other subjects.
The procedure for decision making is as follows. A sub-
ject is randomly selected by the experimenter to make a
proposal. For example, if 1 is recognized, she might say, "I
move we change the status quo from (100, 100) to (25, 50)."
This motion is put to a majority vote. If a majority ap-
proves, then (25, 50) becomes the new status quo; if it fails
(with ties constituting failure), then (100, 100) remains the
status quo. A new person is recognized to make another mo-
tion. This process is repeated, with subjects recognized ran-
domly to make proposals. At any time there is a privileged
motion: Any subject can move to end the session. This is
voted on immediately, and if it passes, the session ends and
each subject is paid the value to him or her of the sta tus
quo pr evailing at that time. Otherwise, motions continue to
be made until no subject wants to make a new proposal.
(Even if subjects still have motions to make, if a time limit
is reached, then the session automatically ends.)
What happens? My experience in running this experi-
ment in classes over nearly two decades is that the spatial
theory of majority rule works quite well. To illustrate, re-
turn to Figure 5.8, where preferences as displayed there
satisfy the Plott Theorem (distance-based preferences, an
odd number of subjects, ideal points distributed in a radi-
ally symmetric fashion). The theory of majority rule implies
that the final outcome should be the ideal point of subject 2,
since x 2 can defeat any other point in a majority compari-
son. As long as someone proposes this point, it will become
the new status quo and then it cannot be dislodged by any

subsequent proposal. In actual experiments, subjects start
off tentatively, partly because they are only just figuring
out the setting and partly because they don't know the pref-
erences of other subjects. But once they gain familiarity,
things move along at a rapid clip. Subjects seek recognition
to make motions; motions are made and approved that
move the status quo from the extreme northeast point into
the "center of things." Each s uccessive motion, as the
theory would pr edict, moves the status quo closer to xv the
point the Plott Theorem identifies as the final outcome
(the equilibrium of the majority-rule process). Logically,
ther e is always a majority that will prefer a point closer to
x 2 to one furth er away, so motions that move a status quo
closer to x 2 should prevail over a more distant status quo,
and motions that move t he status quo further away should
fail. This typically happens in the experiments I've run,
though occasionally a mistake is made. But even if this
should happen, it is quickly corrected by a new motion.
There are intermittent efforts to bring the proceedings to a
close (especially when a status quo lands close to some sub-
ject's ideal point), but typically a majority r eject s this until
x 2 is reached. Once x 2 is reached, any subsequent motion is
defeated a nd, ultimately, subjects tire of more motions and
finally approve a motion to end the session. A typical exper-
imental session draws to a close quite rapidly; rarely is it
ended because it reaches the time limit.
If, however. a distribution of ideal points is not like that
of Figure 5.8. violating radial symmetry instead, then
majority rule does not have an equilibrium like x 2 . The
McKelvy Theorem applies. What happens in this experi-
mental condition? My experience in running the experi-
ment in this context is that McKelvey's Theorem captur es
the situation. Majorities approve proposals moving the s ta-
tus quo around but never settling on any specific point.
More often than in the previous setting, the time limit de-
termines when the session will end. The final outcome is
somewhere in the middle of the distribution of preferences,
but it does not hone in on a specific point as it did when the
Plott conditions were satisfied. lt is not exactly "chaos," but
outcomes do not display the same regularity they do when
preferences are distributed in a radially symmetric fashion.
As a final exercise, I have added a new twist to the orig-
inal Fiorina-Plott experimental design. In some sessions I
tell subjects not only their own ideal points but also those of
the other subjects. Does this make a difference? Theory
says it shouldn't-either there is a Plott equilibrium point
or there is the wandering around of the McKelvey Theorem.
The additional information does not alter these facts. And,
experimentally, the results are about the same as in the
limited-information context, though the chatter among sub-
jects during the experiments does reveal envy and competi-
tiveness when some approved status quo looks to benefit
the person sitting across the table.
At least in this rather carefully controlled setting, pure
majority rule (with only the most elementary of institu-
tional features) pretty much works the way the theorems of
Plott and McKelvy suggest. Indeed, against a large number
of theoretical competitors, Plott and Fiorina conclude that
this theory of majority rule is superior.
1. Suppose that a society consists of three individual s, 1, 2 and

3, who must choose one among three proposed budgets, x, y,
and z. Their preferences over these three possible budgets are
as follows: xP 1yP 1z, yP'J?PzX, and zP3xPY'·
• Write down the majority preference relation for this pro-

file of preferences (e.g., indicate which alternatives
would beat which in two-way contests).
• Does Black's Median-Voter Theorem support a predic-
tion about which policy will be chosen if the group uses
simple majority rule? Why or why not?
• Suppose that the group is going to use a voting agenda u
= (y, x, z) to select the budget, where this notation
means that the group first votes over y and x, and then
votes over the winner of this contest and z, where the
winner of this second vote is chosen as the budget. If
each individual votes sincerely at each stage of the
agenda, what would the outcome be? What would the
outcome be if u' - (z, x, y) or u" - (z, y, x)?
2. The Senate Finance committee, in debating a health care

reform bill, contains some members who demanded a full
government-run insurance scheme ("the public option"), oth-
ers who were strongly opposed, and a third group who favored
a compromise health care cooperative, a kind of government-
affiliated nonprofit. Both those in favor of a full public option
and those opposed agree that the "co-op" is the second-best
compromise outcome. What can you say about this group's
preferences? Could you predict the outcome of a vote if you
knew the number of members who held each of the three pref-
erence profiles?
Some of the committee's more liberal members start to take

political flak from interest groups strongly in favor of a full
public option, who convince them that it would be better to
vote against the "co-op"-and wait for a more propitious
time-if the full public option is unachievable. What can you
say about this new set of preferences? Could you always pre-
dict the outcome of a vote if you knew the number of members
who held each preference profile?
3. A famous example of preferences which violate single-

peakedness was first presented in Verba et al.'s "Public Opin-
ion and the War in Vietnam."" This study found that most
individuals identified a specific policy as their ideal point on a
one-dimensional policy space varying from a reduction or end
to U.S. engagement at one end, continuation of U.S. engage-
ment in the middle. and expansion of U.S. commitment at the
other end. Moreover, most stated that policies further away
from their ideal points were increasingly less desirable. How-
ever, there was a small segment of the population who favored
either a U.S. withdrawal or an expansion of the U.S. commit-
ment to a continuation of the present policy. What property of
a society's preference profile is not respected in this example?
How widespread must non-single-peakedness get before group
transitivity is violated? Illustrate this graphically, and then
explain whether preferences over policy in the Vietnam War
might better be represented using two or more issue dimen-
sions. What might those di1nensions include?
4. A seven-member governance committee of a corporation is

charged with allocating funds for end-of-year bonuses, and a
special subcommittee is appoinled to research and then pro-
pose to the full governance committee a total value of all
bonuses. The corporation's rules state that the subcommittee
may bring a proposal before the full committee under a prede-
termined amendment procedure (either an open or closed
rule). All subcommittee and committee decisions are made
using majority rule; the minimum the committee may allocate
is $0 and the maximum is $12,000. The preferences of each
members of the seven-member governance committee over al-
ternative total amounts to allocate for the bonuses are single-
peaked and symmetric about each member's ideal points, and
are arrayed as follows:
" Sidney Verba et al., 1967, "Public Opinion and the War in Vietn11m," Ameri-
can Political Science Reuiew 61, no. 2 (1967): 317-33.
Bobby and Emma: O; Amy: 1500; Cathy: 6000; Frank: 7500; Geri :
10,000; and David: 12,000.
Answer the following questions, illustrating pictorially where

helpful.
• The subcommittee is composed of Frank, Geri, and

David. What is the subcommittee's most-preferred level
of funding?
• Last year the governance committee allocated $3000 for
bonuses, a nd under committee rules this is the status
quo level of funding (it will be the decision this year, too,
if the full committee makes no change). Assuming the
subcommittee brings a proposal to the full committee
under a closed rule, what is the subcommittee's proposal
and what will be the outcome? Why?
• Now assume the subcommittee brings its proposal to the
full committee under an open rule. Will the subcommit-
tee "open the gates" and bring a proposal to the floor? If
so, what will it propose? What will be the outcome?
Explain.
• Now assume th at the governance committee follows
"zero-based budgeting" in which each year's allocation is
initially set to zero, and the committee then proceeds to
vote on any new proposals made by the subcommittee.
In this case, what will the outcome of the voting be
under a closed rule? Under an open rule? Comment on
the implications of zero-based budgeting.
5. A legislature is going to vote on a policy tha t is well repre-

sented by a sin gle-issue dimension, on a scale of zero to one.
The initial policy proposal will be supplied by a committee
(whose ideal point for this exercise can be any point along the
x-axis) to a legislature with median ideal point M. The status
quo is currently at point SQ. Draw a line showing the equilib-

rium outcomes for any committee ideal point, when the com-
mittee's proposal is considered under a closed rule. On a
separate set of axes, perform the sa me exercise when the com-
mittees proposal is considered under an open rule. Then, do
the same exercise for both rules (using a dashed line on the
two axes already created), assuming that the legislature oper-
ates on the principal of zero-based budgeting, so if no bill is
passed in the current session, the policy reverts to ZB = 0.
What is the impact of zero-based budgeting under a closed

rule? Are equilibrium outcomes closer or further from the legis-
lature's median, on average? What about under the open rule?
FIGURES.A
1
w
:E
0
()
I-
::i
0
:E
::i
a: M
en
...I
::i
0
w
SQ
ZB=O
ZB=O SQ M 1
COMMITTEE IDEAL POINT, C
6. A legislature with three groups similar to that shown in

Figure 5.6 is attempting to reach a bargain on some issue
which has two policy dimensions. 'l'he legislature uses major-
ity rule, and the indifference contours are all circular. Suppose
that groups 2 and 3 reach an agreement on some deal where 3
gets its ideal policy on the horizontal dimension and the out-
come on the vertical dimension is one-third of the distance
from 2's ideal point moving toward 3's. Draw the preferred-to
sets for 2 and 3 and show that there are other bargains be-
tween 2 and 3 that could leave them better off. Consider the
set of bargains that 2 and 3 could make that cannot be altered
without making either group worse off; what does this set look
like? '!'his is called the "contract curve" for groups 2 and 3.
Draw in the contract curve for all other possible majority
coalitions. Are any points on these contract curves empty win-
set points under majority rule?
.Now, suppose the legislature operates under a unanimity rule.

If the status qu o is at the agreement point between 2 and 3
described above, will the three sides be able to find a bill that
makes all of them better off? What if the status quo is at 2's
ideal point in the horizontal dimension, and halfway between
2 and 3's ideal points in the vertical dimension? What bilJs
have empty winsets when the unanimity rule is employed? Do
the existence of these empty winsets contradict McKelvy's
Chaos Theorem'?
*7. Assume five members of a voting body A, B, C, D, and E.

They have ideal points in a two-dimensional issue space,
given by A: (1,4), B: (4,4), C: (2,2), D: (1, 1), and E: (2, 1), where
the first coordinate gives a person's ideal policy on the X di-
mension and the second coordinate gives the ideal policy on
the Y dimension. Assume further that each member prefers
an outcome closer to his or her ideal point to one further away.
All votes are taken over X- Y policy bundles, so all proposals
are in the form of ordered pairs (x,y). Is there an equilibrium

proposal if the body uses majority rule?
Now suppose that player B has been appointed the sole

agenda setter. If the status quo is at C's ideal point, can B con-
struct an agenda that will lead eventually to her own ideal
point being approved by the entire legislative body? If so, con-
struct such an agenda (you may assume sincere voting, and
that the agenda setter can always persuade legislators indif-
ferent between two policy bundles to vote for the bundle .that
she prefers). What theorem does this illustrate?
8. Perhaps the most famous political aphorism belongs to Otto

von Bismarck: "Politics is the art of the possible." Comment,
with reference to Black's Median -Voter Theorem, McKelvey's
Chaos Theorem, and the discussion of legislative rules (espe-
cially Case 5.2).
9. This question asks you to consider the links between

Arrow's Theorem and McKelvey's Chaos Theorem. Do the
group preference intransitivities that motivated Arrow's ideas
occur in McKelvey's multidimensional, majority-rule settings?
Does the "anything can happen" message of McKelvey res-
onate with any of the lessons of Arrow's Theorem? Does
agenda control bestow an ability to determine the outcome in
the Arrowian world? In all cases? Develop some small three-
person, three-outcome examples to show that the perverse
outcomes of the multidimensional spatial model can, but do
not always, occur in the discrete world of Arrow.
10. Anthony Downs employed the logic of unidimensional deci-

sion making first articulated by Duncan Black in order to un-
derstand politics in the largest "committee" imaginable-the
electorate. He pointed out that if there were only two parties,
and if politicians with extreme views had little inclination to
Spatial Models of Majority R ule 155
form parties of their own, then the platforms and campaign

promises of the two parties would be very similar.
• Why is that? How does his conclusion follow from

Black's Median-Voter Theorem?
Downs takes politicians to be interested only in winning
office. Does a different result (i.e., something other than
convergence) arise when politicians have strong policy
preferences of their own? Under what circumstances?
• Which other of Downs's assumptions (about the number
of candidates, voting behavior, or the dimensionality of
the issue space) might explain the apparent differences
in party platforms in the United States, and elsewhere?
6
Strategic Behavior:
Sophistication, Misrepresentation,
and Manipulation
In models of social choice and spatial decision making, voters

vote their preferences. To some, this is the essence of rational
behavior. To others, however, rationality is more subtle and
nuanced. It entails doing the best one can with what one's got,
and this sometimes requires making strategic maneuvers, in-
vestments, sacrifices, and retreats. I t is embodied, for exam-
ple, in the Protestant work ethic, which encourages deferred
gratification in order to harvest later returns. Some might
claim that Protestants, in some perverse sort of way, like de-
ferred gratification, but the alternative view is that their ethic
(an ethic common among As ians, Jews, and others, too) re-
flects the strategic decision to maximize over t he long haul by
r esisting enslavement to short-term preference satisfaction.
Models need to reflect this possibility, for a failure to take
strategic capabilities into account may result in disaster (as
Case 6.1 demonstrates). This chapter is devoted to elaborating
on the multiplicity of ways strategy r ears its head.
156
Strategic Behavior 157
CASE 6.1
"NEED-BLIND" COLLEGE ADMISSIONS
A prestigious university (which shall remain unnamed), in

a spirit of generosity a nd genuine concern, instituted a
need-blind admissions policy for its graduate school. It
sought to attract the finest graduate students, independent
of their ability to pay the cost of graduate education. But
universities, like any other organizations, are faced with a
large number of possibilities and a limited number of re·
sources with which to pursue them. Each academic depart-
ment's entering class had to be limited in number and, to
accomplish that, each department was given a fixed budget
to spend on graduate students. As we shall see, financial
scarcity and n eed-blindness in admissions compete at the
margin, sometimes with perverse consequences.
The admissions procedure works as follows. First, each
applicant gives financial data to the central office of the
graduate school. Simultaneously, each applicant submits
admissions materials (GRE scores, undergraduate grade
transcripts, teacher recommendations, essays, and so on) to
the academic department in which he or she is interested,
but not the financial data (which was given to the graduate
school). Thus, the department, in ignorance of candidate fi.
nancial need, constructs a preliminary rank-order of the ap-
plicants, along with a cutoff line below which rejected
applicants fall. This preliminary list is forwarded to the
graduate school, which attaches next to each candidate a
minimum financial stipend the department must offer. De-
partments, of course, can make larger offers; but they can-
not admit a student and make a s maller offer- that is the
rule. This minimu,m stipend is need-blindness in action-it
indicates how much subsidy is required to bring a candidate
up to a financial status that would permit him or her to
attend graduate school. These marching orders arc trans-

mitted back to the department, which now gets a second
crack. Given its limited budget, and its desire to have as
fine a class of e ntering students as its budget permits, it
may have second thoughts about those it has admitted, now
that it knows the minimum price it must pay from its scarce
resources. For e xample, an especially needy student ranked
25 may require, by graduate school dictates, $20,000 of sub-
sidy. Candidates 12 and 13 may be less needy, but, the de-
partment reasons, if we take the $20,000 from 25 (and
therefore choose not to admit that person after all), and use
it to "top up" the offers we've made to candidates 12 and 13,
we may have Ii better chance at attracting two, higher-
ranked students.
The need-blind admissions procedure inadvertently dis-
criminates agai nst financially needy students who are not
highly ranked (though clearly above the cutoff line). Astra-
tegic department will have strong incentives to reallocate
what would have to be spent on these students toward
higher-ranked (possibly unneedy) candidates to enhance
the prospects of the latter accepting an offer.
Need-blind admissions is a noble endeavor; so, too, is pro-

ducing excellence in graduate education. In Case 6.1, univer-
sity administrators should have devised a scheme to achieve
need-blind admissions that took into account the possibility
that others on whom they depend- namely, professors on de-
partmental admissions committees-may not share their pur-
poses to the same degree. The scheme described above clearly
did not prevent strategic maneuvers that had the effect of un-
dercutting noble purposes. Someone wasn't looking ahead;
someone was failing to anticipate.
R ATIONAL FORESIGHT
In politics it is essential to look ahead, to anticipate, to exer-

cise prudence and foresight. The political world is full of pur-
poses, some noble and some ignoble, some competing and
some complementary. Rat ional actors, seeking to enhance the
prospects of the purposes they pursue, must think str ategi-
cally. And one of the fundamental principles for thinking
str ategically is looking before you leap. 1
Thinking strategically is not always so easy. Consider a
three-person legislature, each of whose members would like a
pay raise. 2 But each legislator realizes that constituents will
not be pleased with a r epresenta tive voting to increase his or
her own salary. So, the best of all possible worlds for legisla-
tor i is for the other two legislators to vote in favor of the
pay raise, thereby causing it to pass, with legislator i voting
against. In this case, he does not displease his constituents
but receives the extra cash anyhow! The worst of all worlds,
of course, would be for i to vote yea but for the motion to raise
legislative pay to fail. The other two possibilities fall in be-
tween, and all three legislators have precisely the same feel -
ings on this issue. That is, they rank-order the outcomes (pass
and vote nay) P (pass and vote yea) P (fail and vote nay) P (fail
and vote yea).
Now then, the roll is a bout to be called. Each legislator will
be asked for a public declaration on the motion to raise pay.
And, being a roll, it lists legislators in alphabetical order-i
first, then j, then k. Suppose you are legisla tor i. How should
you vote? Thin k about your answer, and write it down on a
piece of scratch paper before proceeding.
1 An outstanding book with exactly this title is Avinash Dixit and Barry Nale-
buff, Thinking Strategically (New York: Norton, 1991). While I can devote
only a few paragTaphs to this topic in my book, the Dix:it-Nalebuff volume is
a superb extended discussion of strategic behavior.
2 This example is drawn from Peter Ordeshook, Game Theory and Political
Theory (New York: Cambridge University Press, 1986).
The author has tried this out on many student audiences.

The typical response is that legislator i ought to vote yea. The
following reasoning is offered:
I don't know how j and k will subsequently vote. And since I don't
know I figure it's equally likely that each votes one way or the
other. This means there's a 50 percent chance that they will split
their votes, making my vote the deciding one. So I'd better vote
yea, securing my second-choice outcome, rather than voting nay
and ending up with my third-best outcome.
This is not a well-thought-out position. The student whose

reasoning we've reproduced has not thought ahead and has
not taken into account that legislator i's actual behavior will
affect whatj and k do. 3
The way to think about this is to put yourself at the end of
the process first- in k's shoes- and work backward. There are
really only three circumstances for k to consider (2 nay votes,
2 yea votes, or a split in the vote), and in each her preferences
provide her with clear counsel on how to proceed. If two nays
have preceded, then she will definitely vote nay. If two yeas
have preceded, then she will again definitely vote nay. If there
is a split vote, then her vote is deciding and she will definitely
vote yea. In each contingency we know definitely how k will
vote.
Specifically, legislator j (who chooses next to last) can fore-
cast with certainty how k will vote (as long as k does what she
should!). Suppose, when it comes time for him to vote, i has al-
ready voted yea. Then j knows that if he votes yea the bill will
pass (and, from the above, k, whose vote will have no effect,
3 What I mean to say here is that most students implicitly assume that j and
k have already made up their own minds about how to vote. Students do not
appreciate, in general, that the choices j and k will actually make are reac·
tions to i's initial choice, so that i can, in fact, affect the thinking of j and k.
The true strategist, in contrast, appreciates his or her power to inRuence the
thinking of subsequent movers.
Strategic B ehavior 161
will vote nay and this will be the best outcome for her); on the
other hand, if j votes nay, the bill will still pass because k will
be forced to vote yea, and this will be best for j. What we have
figured out so far, then, is that if i votes yea, then j can afford
to vote nay, forcing k to vote yea; the bill will pass, i and k will
get their second-preference outcome, while j gets his first
choice.
But what if i decides to vote nay? He can think ahead,
since he has deduced how the others will react. Legislator i
knows that, foUowing his nay vote, if j votes nay, the bill fails
(no matter what k does). On the other hand, if j votes yea this
will produce a split vote forcing k to vote yea, too. Legislator j
is snookered here, since he will reluctantly vote yea to get his
second choice r ather than vote nay, killing the bill and getting
his third choice. We now know what i should do. By voting
nay, he forces both the other legislators to carry the burden of
passing the pay-raise bill.
CASE 6.2
CONGRESSIONAL PAY R AISE DILEMMAS
An examination of actual cases of legislative pay raises pro-

vides anecdotal evidence of a logic much like that described
in this section. The basic idea is that legislators want to
vote no if their vote is irrelevant to the outcome and yes
only if they are critical to the outcome. In 1990, the U.S.
Sen ate was debating a major piece of legislation to reform
congressional pay. Senator Daniel Patrick Moynihan (D-
N .Y.) intr oduced an a mendment to the bill to increase pay.
One report recounted the scene this way:
As opponents nearly defeated the amendment on a 50-50 t ie,
Moynihan stood in the center aisle of the chamber, hands out-
stretched. "One more vote!" he pleaded. Laughing, Alaska Re-
publican Frank H. Murkowski switched his vote to support

Moynihan, bringing the final tally to 51-49.*
Although we have no direct evidence, we suspect that in

the early voting Senator Murkowski understood that voting
no would force others to get the bill passed, and thus that
he could have his cake and eat it, too. Only when his vote
became essential did he switch.t
A similar situation took place in 1992 when the Senate
was considering an amendment to transfer funding from
military projects to breast cancer research. Senators were
reluctant to break the "firewall" between defense and do-
mestic spending, but they also feared being caught on the
losing side of an issue like breast cancer, given the salience
of women's issues in 1992. If the breast cancer amendment
were going to pass anyway, or if it were certain to fail, a
senator would want to be recorded as yes. Only if he or she
were the decisive vote would an alternative behavior be
considered. This created a "pay-raise dilemma." The scene
on the Senator floor, as described by the Washington Post,
demonstrates the argument:
When votes for the project were safely in hand [that is, the
amendment was certain to pass), senators began streaming
onto the floor to change their 'no' votes to 'yes.' Minority
Leader Robert J. Dole (R-Kan.) was overheard urging the Re-
publican cloak r oom to alert absent senators to the stampede.
An unofficial tally showed 28 senators, including . . . Dole,
switching their votes. Several others waited until the last
minute before casting votes in favor of the proposal.t
• "Lawmakers' Pay Raised, Fees Curbed," Congressional Quarterly Al-

manac (1990): 74.
t Why he didn't stick to his guns and make some other no voter bear the
cost of switching isn't answered here. The reader might wish to specu·
late. I will discuss such phenomena, known as the "free-rider problem,"
in Part III.
i Helen Dewar, "Senate's New Sensitivity," Washington Post, Septem-
ber 24, 1992, p. Al.
These two cases are by no means unique. Watch C-SPAN

for other instances of last-minute vote switches as repre-
sentatives fight to stay on the right side of an issue, giving
new meaning to the term "politically correct."
Strategic thinking and rational foresight are general terms

for the kind of calculations I've described. They entail a logic
that takes advantage of the sequential structure to the deci-
sion making. They involve, oddly enough, thinking fo rward by
reasoning backward. Indeed, the method is called backward
induction, and works as follows. Take a generic sequential sit-
ua tion, like the one pictured in Figure 6.1. There are three in-
dividuals who must make a collective decision (not unlike the
legislative pay-raise example above), each by revealing an in-
dividual choice in turn. Tu complicate things a bit, let's sup-
pose that Mr. I and Mr. III each have three options available
in their "action sets"- {al' av a 3} and {cl' c2, cg}, respectively-
whereas Ms. II has only two options in hers-{bl' b2}. In addi-
tion there is a fictional fourth player, an androgynous 0 known
as Chance. In Figure 6.1 Chance moves first, his/her only re-
sponsibility to select the order in which the other players
move. Of the six possible choices Mr./Ms. 0 could make, we
suppose he/she chooses the order: I, III, II. The "game" now
starts. 4
An outcome of this social circumstan ce results from the
three participants making choices from their action sets.
•In fact, we are describing what is known as an extensive form game in game
theory. Game-theoretic ideas run all throu gh this volume, though I don't al-
ways take the time to point this out. Political science models are increas-
ingly game-theoretic in spirit if not in fact. A good place to encounter game
theory in the context of politics is the Ordeshook volume cited in footnote 2
above. For an outstandingly clean a nd clear (and brief!) presentation of the
essentials of game theory, nothing competes with David M. Kreps, Game
Theory and Ecol'IOmic Modelling (Oxford, U.K.: Oxford University Press,
1990).
FIGURE 6.1
Mr. I Mr. Ill Ms. II Outcome Payoff (I, II, Ill)
!~ b2
CD
2
(3, 3, 2)
(7, 1, 5)
b, 3 (1, 6, 10)
§1 © (8, 9, 3)
b,
=~
(3, 3, 9)
-= Final out come
~ j(a, 4, s>I
@!] 0 {10, 6, 4)
b2 8 (5, 1, 2)
-
@!]
b2
:®10
(7, 7, 6)
(4, 2, 4)
b, 11 (1, 4, 2)
~ @ (5, 10, 1)
b, 13 (9, 3, 6)
~ ® (3, 8, 2)
....
@!]
b2
:@16
(3, 10, 8)
(7, 4, 2)
@!] @ (6, 2, 5)
b2 18 (10, 1, 8)
There are eighteen different possibilities (resulting from the

product of the t hree possible choices of Mr. I, the two possible
choices of Ms. II, and the three possible choices of Mr. III) . The
method of backward induction begins with the player who
moves last- Ms. II in this case. When it actually comes her
turn to choose, she will find herself at one of nine choice
points. (These are the nine nodes in the game tree in Figure
6.1 that line up below Ms . II's name.) At each node she may
decide between two final outcomes (which are determined
once she chooses between the two options in her action set).
There is nothing fancy for Ms. II to do: once she learns which
of the nine nodes the process has arrived at (as determined by
choices by the other players), she will choose the option that
leads to the outcome she most prefers. 5 At the top node, for ex-
ample, she will select b 1 or b2 depending upon whether she
likes outcome 1 or 2, respectively, best. The same holds for
each of the other eight nodes at which she may have a choice
to make. Without bothering to write down complete preference
orderings for all three players, I've simply indicated the pay-
offs to the players in Figure 6.1. For example, if outcome 1
prevails, then Mr. I receives a payoff of 3, Ms. II a payoff of 3,
and Mr. III a payoff of 2. We assume, and this is very impor-
tant, that the three players know one another's payoffs.
For each combination of choices by I and III, Ms. II has a
choice between two outcomes, depending upon whether she
chooses b 1 or b2 . We have put a box around the maximizing
choice she will, in fact, make, and circled the outcome in each
pair that will, in fact, be realized.
Mr. III is the next-to-last chooser. He knows that he will be
at one of three choice nodes, depending upon Mr. I's prior
choice. Using his foresight-that is, his knowledge of how
~ This is actually an important general principle that l will highlight later in

this chapter- namely, that the person moving last behaves pretty much ac·
cording t.o her preferences. There are no fancy stratagems at this ultimate
stage of the game.
Ms. II will choose at each of her subsequent choice opportuni-

ties given her preferences over final outcomes- he can deter-
mine how he shou ld choose, given his own preferences over
final outcomes. Tn effect, he can now erase Ms. II from his
mind, replacing her nominal discretion after he chooses with
the choice he knows she actually will make (the ones I've al-
ready boxed) . Mr. III treats the boxed choices of Ms. II as
strategic equivalents for those nodes. So, at Mr. Ill's top node
he knows the implication in terms of final outcomes of each of
the three options available to him there. If he chooses cl' then
outcome 1 is sure to occur; if he chooses c2' then it will be out-
come 4; if he chooses c3, outcome 6 will be realized. The same
will be the case at each of the other two choice nodes for III.
He has prefer ences in each of these instances and, without
writing them down explicitly, I simply box his optimal choice
at each of his three choice nodes. The reader can check that
the boxed choice at each of Ill's possible decision nodes is opti-
mal for him, taking account of !I's subsequent optimal choice.
Finally, it comes time for Mr. I to choose. He is now able to
suppress the discretion of both Mr. III a nd Ms. II. He knows
exactly what they will do in each contingency in which they
might find themselves. Consequently, he knows the effect of
each of the three choices he could make. So for him it is simply
a ma tter of deciding which of the three outcomes he likes
best.6
Working up the game tree via the method of backward in-
duction, we have determined precisely how strategic actor s
will behave. Each will take into account what has, in fact, pre-
ceded their choices, as well as what they forecast will happen
after they ha ve made t heir choices. The final outcome of this
6 If he chooses a" he knows Mr. Ill will follow with c3 and Ms. II with b2 . Thus
a choice of a 1 yields outcome 6 with a payoff of 8 to Mr. I. If h e chooses a 2 ,
then the other players will choose c~ and b., respectively, yielding Mr. I a
payoff of 7 from outcome 9. Finally, if he chooses a 3 , outcome 15 with a pay-
off of 3 for him results. Of these he prefers outcome 6 and thus chooses a 1•
entire exercise is called the strategic outcome or sophisticated

outcome. 7
This concept is explored in more detail in the remainder of
this chapter. But before moving on, I should point out that we
have already seen at least one instance of strategic behavior.
In the last chapter, a committee chair operating under an open
rule in a one-dimensional legislature ''backward-inducted." He
asked himself what would happen if he opened the gates and
made a proposal, concluding that rational behavior on the floor
would lead to amendments ultimately producing xm. On the
other hand, the final outcome would remain at :xfJ if he kept the
gates closed. His decision on opening the gates, then, de-
pended on his preferences between xm and :xfJ. Thus, his own
agenda-setting choice was predicated on foresight. We have
also seen a failure to exercise this sort of foresight. The gradu-
ate school designers of the need-blind admissions policy
described in Case 6.1 failed to exercise foresight, blithely as-
suming straightforward compliance of departments with their
objectives. They failed to appreciate that there are competing
objectives pursued by strategic actors.
CASE 6.3
P RESIDENTIAL V ETO STORIES
The ability of the president to veto legislation passed by

Congress provides an excellent opportunity to examine
strategic behavior and backward induction. Some political
observers have taken the infrequency with which presi-
dents have used the veto as an indicator of cooperation be-
7
When the choice is from a fixed sequence of votes, like a legislative agenda,
most authors refer to "sophisticated" voting. When the choice is made in a
one-shot circumstance, like switching your vote in an election because you
believe your favorite candidate is out of contention, it is called "strategic"
voting. I will tend to use these terms interchangeably.
tween the legislative and executive branches. In this view,

infrequent use of the veto corresponds to a high degree of
harmony between the branches. Thus, in his 1994 State of
the Union speech, President Bill Clinton proudly an-
nounced the end of gridlock and the beginning of a new pe-
riod of interbranch harmony by pointing to the fact that not
once had he vetoed a piece of congressional legislation dur-
ing 1993. Our backward induction argument suggests that
President Clinton did not have logic on the side of his argu-
ment. The reason is simple: Rational foresight tells us that
Congress should often change its behavior in order to avoid
stimulating a ve~o; to do so is a measure of congressional
foresight, not an indication of harmony between the
branches. For example, in 1992 the Senate was reported to
have dropped family planning money from a foreign aid bill
in order to avoid a veto from President Bush.
Why would we ever see a presidential veto, if Congress
can anticipate this veto and trim its sails accordingly? One
answer is procedural: There is the possibility of an override.
Congress can nullify the presidential veto if each chamber,
subsequent to a veto, re-approves the bill by more than a
two-thirds majority. In these circumstances, Congress can
push ahead, knowing it can subsequently overcome presi-
dential obstacles if it needs to. But this just pushes the
question back a stage, namely, Why didn't the president ex-
ercise rational foresight in anticipating the override and
therefore desist in using the veto? A second possibility,
then, is uncertainty. Congress may not know how the presi-
dent will respond to the legislation; the president may not
correctly guess that his veto will be overridden.
Despite the veto possibility and even without a likely
override, it may still be rational for Congress to send a bill
to the president. Symbolic or constituent politics is one rea-
son. During the Bush administration public pressure led
Congress to consider a series of campaign-finance reform
bills. The optimal outcome for each congressperson was to
vote for the bills, thereby appeasing constituents, but some-

how to avoid actually having to implement the reforms. A
promised veto by President Bush gave the legislators what
they wan ted. According to the Washington Post, Congress
was "secure in the knowledge that President Bush would
veto [the bills] and the veto could not be overridden; the vir-
tuous vote was free."* In this case, ironically e nough, the
presidential veto was a sign of cooperation between Con-
gress and the White House, not conflict.
Party politics may also lead Congress to send a bill to the
president knowing it will be vetoed. In 1992 Democratic
Party desires to portray th e Republican White House as the
source of governmental gridlock led to a series of bills on a
variety of subjects-cable TV r egulation, family leave policy,
most-favored-nation status for China-all of which Presi-
dent Bush , because of previous commitments, was forced to
veto. Because these bills wer e popular with the general pub-
lic, and because the vetoes displayed an executive at logger-
heads with the legislature, the president's actions came at a
considerable cost to him. In this case the presidential vetoes
were a sign of conflict, not cooperation, with Congress.t
It should be apparent that very little can be inferred
about the degree of conflict or cooperation in a situation
just by observing an outcome. A presidential veto may re-
flect miscalculation (by Congress if the veto is not overrid-
den, by the president if it is), interbranch harmony, efforts
by one branch to embarrass the other, and many other
things besides. Only by a car eful analysis of the strategies,
alternatives, payoffs, incentives, and circumstances can we
form an accura te interpretation of the event.
·"Campaign Refo:rm Anyone?" Washington Post, February 7, ' 1993, edito-

rial page.
t Ann Devroy, "Congress Pitching, President Vetoing," Washington Post,
September 24, 1992, p. Al.
In what follows, I take strategic behavior to consist of an

extended sort of r a tionality. An individual does not merely as-
sess options in front of her nose, choosing the one that seems
best in terms of her preferences. Rather, she takes account of
the fact that the choice before her is but one of a sequence of
choices in an ongoing process. Strategic behavior requires that
she look beyond her nose; the immediate choice before her,
therefore, is not merely a one-shot, myopic decision, but instead
one with longer-term effects. This leads, in each of the con-
texts we are about to explore, to the possibility of anomalies in
which, at a specific choice opportunity, one ends up choosing
what would appear to an outsider as a less-preferred option, if
taken as a myopic choice. Strategic behavior, in short, taking
the full horizon of a process into account, may r equire individ-
uals to make seemingly less-than-ideal choices at some points
in order to secure superior outcomes at the end of the trail.
Below, we see this manifested in (1) sophisticated voting a nd
agenda maneuverings in legislative settings, and (2) strategic
voting and issue manipulation in electoral settings. First,
however, I take up the general issue of strategic manipulation.
MANIPULATION
Before we go any further, it might be worth taking up an issue

that a few doubters may be thinking about. To some, the idea
of strategic behavior may be alien and, possibly, morally con-
temptible. Whether one admires brilliant strategic maneuvers
or not, however, we need to know about them and where they
are likely to be encountered. One thing we might want to
know is whether there a re decision-making procedures that
encourage only h onest, nonstrategic behavior- procedures
that are basically strategy-proof. This is a n abstract question
that can be tackled at a fairly general level.
Suppose there are n group members, G = {l, 2, ... , n}, who
must choose frQm a set of m alternatives, A= {al' azi ... , am}.

I won't be very specific about the way in which the group
makes its choice, except to say that the final choice will be one
of the alternatives in A on the one hand, and that it will de-
pend (somehow) on the preferences expressed by the group
members on the other hand. That is, the social choice is writ-
ten as F(Q1, Q2 , . . . , Qn, A) EA. In this expression, A is the set
of alternatives defined above, Qi is a preference ordering of the
alternatives in A expressed by member i , and Fis some deci-
sion process that transforms these expressed preferences into
an outcome in .A. Enter strategy.
I italicized the word "expressed" throughout the last para-
graph for a reason. In most group choice situations we are not
in possession of a fancy "preference meter" that reads people's
minds. The group choice procedure, F, can only take the
preferences that individuals choose to reveal. And where is it
written that people are always honest in their preference rev-
elation? We have already seen situations where people vote in a
deceptive manner. More generally: Might an individual be able
to secure a better outcome (according to her true preferences)
by revealing untrue preferences-by behaving strategically?
Let's suppose that the true preference orderings of the
members of Gover the alternatives in A are P 1 , P 2 , . . . , Pn, al-
though no outside observer has any way of knowing or vali-
dating this. We say that Ms. i is sincere only if, in the group
decision setting where she is asked to reveal her preferences,
her revealed preference (Q,) is identical to her true preference
(P); if Qi .,e Pi, then she is said to be sophisticated. So P, re-
flects her true tastes, Q, is what she chooses to reveal, and she
is sincere only if the two are the same.8 A sophisticated indi-
8 I am being a little cagey here because it is quite possible that, sometimes,
honesty is the best policy. Thus, QFPi is a necessary but not a sufficient con-
dition for someone to be said to be sincere. Confused? Don't worry about it.
Interested? Then take a look at the piece by David Austen-Smith, "Sophisti-
cated Sincerity: Voting over Endogenous Agendas," American Political Sci-
ence Review 81 (1987): 1323-30.
vidual is someone who may misrepresent her true preferences

and, when she does so, she is said to .manipulate F, the social-
choice procedure. Given all this, I now repor t the bad news:
Gibbar d-Satte rthwaite Theorem. Assu me a group G
of a t least three indiv iduals and a set A of at leas t
thre e alt ernat iv e s. Also a ssume that any m e mber of
G may have, as his or her true p refe r ences, any pref-
erence ordering ove r A (universal d omain). Then
every nondicta torial s ocial-choic e p rocedure, F , is
manipulable for some distribution of prefere nces.
Allan Gibbard, a philosopher, and Mark Satterthwaite, an
economist, simultaneously established this result about so-
phisticated behavior in the mid-1970s. 9 For any group (of at
least three) and any decision setting (of at least three things
to choose among), if the way in which decisions are made does
not allow one member of the group to dictate the choice no
matter how others feel (Arrow's Condition D), and if individu-
als are free to have whatever preferences they wish (Arrow's
Condition U), then it is entirely possible for circumstances to
arise in which at [east one individual has an incentive to re-
veal his preferences strategically. No method of group choice is
immune from manipulation.
Tvvo brief examples will illustrate what is meant here and
preview the rest of the chapter. In legislative politics, there is
something known as a killer amendment. It is an amendment
to a bill which, if successfully attached to the bill, will cause
the bill to be defeated, even though the bill would have passed
if it had not been amended. Discovering such amendments,
and engineering them through the legislative thicket, thereby
9 Allan Gibbard, ''Manipulation of Voting Schemes: A General Result," Econo-

metrica 41 (1973): 587-601; Mark A. Satterthwaite, "Strategy-proofness and
Arrow's Conditions: Existence and Correspondence Theorems for Voting Pro-
cedures and Social Welfare Functions," Journal of Economic Theory 10
(1975): 187-217.
snatching defeat from the jaws of victory (often by someone

purposely seeking to defeat the bill), is a political gift found in
only the most talented politicians (some of whom are profiled
in an entertaining little book by the late William Riker, The
Art of Political Manipulation). 10 Such amendments often re-
quire sophisticated voting in which an enemy of the original
bill votes for the killer amendment, even though she doesn't
like the amendment per se. She does so because she appreci-
ates that the now amended bill will be defeated, whereas the
unamended bill would have passed. Thus, conventional leg-
islative decision making is, as the Gibbard-Satterthwaite The-
orem suggests, often vulnerable to manipulation.
So, too, is electoral politics. In the section after next, I ex-
amine the case of strategic voting, a brief example of which
will have to suffice for now. It is well known that Anglo-
American electoral arrangements, known as plurality voting
systems, are fertile soil for two-party politics (a subject pur-
sued in Chapter 7). In this electoral order, any number of par-
ties may compete for a single office, with the candidate of the
party winning th e most votes (not necessarily a majority of the
votes) declared the winner. The reason this arrangement
nearly always reduces to two-party competition is that indi-
vidual voters arre loath to waste their votes, individual con-
tributors are loath to waste their campaign resources, and
individual political managers are loath to waste their election-
eering skills on !hopeless candidacies. They tend to desert such
candidacies, even if they would rather see that candidacy suc-
ceed because it is preferable in their view. That is, people who
want to make the most of their strategic endowments (votes,
dollars, organizational skills) will prudently deploy them
where they t hink they might make a difference (say, in help-
ing to choose the lesser of evils), rather than deploying them
where they serv-e only to express a preference but have no ef-
10 (New Haven: Yale University Press, 1986).

feet on the final outcome. In the history of the United States,

with some rare but important exceptions, third parties are
victimized by strategic voting. Their final vote count underes-
timates their actual support in the electorate, reflecting misre-
vealed preferences- manipulation- by strategic actors (many
of whom are ordinary voters).
SOPHISTICATED V OTING
Sophisticated voting takes the form of voting against one's

true preferences at one stage of the legislative process in order
to achieve an even better outcome (according to one's true
preferences) at the end of the process. There is no law written
anywhere that says that a legislator must vote his or her true
preferences. Consider the history of the Powell Amendment
(Case 6.4).
CASE 6.4
AID TO EDUCATION AND THE POWELL
AMENDMENT
Probably the most famous and most often reported case of

sophisticated voting surrounds the efforts of the Democratic
majority in the House of Representatives in the 1nid- 1950s
to pass legislation enabling the federal government to pro-
vide financial support to local public school districts.* The
status quo (x0) at that time provided for virtually no federal
role in public education from kindergarten through the
• Fuller versions of this story may be found in William H. Riker, Liberal-

ism Against Populism (San Francisco: Freeman, 1982); Riker, The Art of
Political Manipulation; and Arthur Denzau, William Riker, and Kenneth
Shepsle, "Farquharson and Fenno: Sophisticated Voting and Home
Style," American Political Science Review 79 (1985): 1117-35.
twelfth grade. The Education and Labor Committee of the

Democrat-dominated House of Representatives introduced
a bill, B, to authorize the federal government to subsidize
educational efforts by the states. Adam Clayton Powell
(D-N.Y.), the second-ranking Democrat on this education
committee and perhaps the most prominent black politician
in the country at the time, moved an amendment (A), now
known as the Powell Amendment. If amended, the bill
would subsidize elementary and secondary educational ef-
forts by the states, but would restrict any federal funds
from flowing to a school district that practiced segregation
of the races.
The rules of procedure in the House required that the
original bill a nd the amendment be pitted against each
other (B vs. A). The winner was then subjected to a vote on
final passage (effectively, winner vs. xll). The two votes are
displayed in the accompanying table. Reading its last col-
umn, it may be seen that the Powell Amendment passed,
229 to 197. From the last row, however, it may be seen that
the amended bill failed when pitted against the status quo,
199 to 227. What went wrong?
While I cannot give a complete account here, several
things may be r emarked upon. The 132 yea-yea voters in
the upper left corner of the table are the quintessence of
sincerity. They were mostly northern liberal Democrats
who both abhorred racial segregation and supported federal
aid to education. In the end, as we shall see, they were the
ones whose na'ivete (if that's what it was) was exploited.
The 67 nay-yea voters in the lower left corner are an inter-
esting mix. These are the sophisticated Democrats. They fa-
vored school aid, on the one hand, and they saw Powell's
motion as a killer amendment, on the other hand. Some
may have favored the substance of the amendment, but
they voted against it nonetheless. They preferred half a loaf
to none. The 97 upper right corner yea-nay voters were
crafty Republicans who opposed school aid. They were simply

delighted to support Mr. Powell's effort to eradicate racial
segregation in the South, not because they cared about that
issue but becaus-e they saw that it would sink school aid (by
peeling off southerners in the vote on final passage who
would have supported an unamended school aid bill).
F I NAL P ASSAGE
Yea Nay Total
P OWELL Yea 132 97 229
AMENDMENT Na y 67 130 197
Total 199 227 426
In voting against their nominal preferences in the con-

test pitting B vs. A, the Republicans assured themselves of
a more preferred final outcome; their sophistication paid
off. Had the 132 legislators in favor of school aid been will-
ing to settle for half a loaf (in which school aid would not
have been denied to segregated districts), they could have
voted sophisticatedly against Powell's amendment, even
though they preferred it, thereby assuring that B would
have passed. Why didn't they behave strategically here?
Denzau, Riker, and Shepsle speculate that many northern
liberals feared explaining to their black constituents that
they voted against Powell for "strategic" reasons.
S TRATEGIC VOTING
We have just seen that a voter with a sophisticated capacity

exercises rational foresight by looking ahead- that is, "down"
an agenda that is fixed in advance. He or she may choose, at
some stage or other, to vote against his or her nominal prefer-
ences for strategic reasons. Sometimes, however, a decision
maker exercises rational foresight, though not in quite this
same manner (Case 6.5).
Strategic Beha1,,•ior 177
CASE 6.5
NOT W ASTING O NE'S VOTE
The 1968 national election in the United States found H u-

bert H umphrey (D) pitted against Richard Nixon (R) for the
presidency. There was a third candida te in the r ace, George
Wallace, who ran on the ticket of a new third party, the
American Independent P arty. 1'hroughout the campaign
during t he fall of 1968, Humphrey and Nixon ran neck and
neck, with polls showing their support in the 40 percent
range. Wa llace trailed badly, t hough still acknowledged by
nearly 20 percent of the electorate as their first preference.
In the "only poll that counts," as politicians like to refer to
the actual election, Nixon and Humphrey each captured
43.5 percent of the popular vote, with Wallace coming in
third with 13 percent. It appeared that Wallace's strength
flagged in the final weeks of the campaign; he a pparently
lost mor e than a quarter of his support.
Let's assume that all the Nixon supporter s stuck with
their man, as did all the H umphrey s upporter s. Moreover,
let's assume th.at the Wallace supporters- indeed, the e n-
tire nation- felt the election was going to be very close, but
that their man was out of the running. There are actually
three different t ypes of voters preferring Wallace: (1) those
who ranked H umphrey second [W-H-N] ; (2) those who
ranked Nixon second [W-N-H] ; a nd (3) those who were in-
different between the two [W - (N,H)]. (In the type (1) and
(2) categories, respectively, we include those who had
Wallace tied with Humphrey or Nixon, respectively.) The
Wallace campaign tried to transform all Wallace supporters
into type (3) voters. His slogan emphasized that there was
"not a dime's worth of difference between the major party
candidates." Undoubtedly, however, some of those in the
first and second category, in the privacy of the voting booth,
decided not to waste their votes on a hopeless candidate, in-

stead switching to their second preference. As a result, the
actual vote totals of both Humphrey and Nixon grew rela-
tive to late poll data, whereas Wallace's shrank.
Fast-forward a quarter century to the 1992 presidential
election. Once again a winning candidate, Bill Clinton, re-
ceived only 43 percent of the popular vote. Once again there
was a popular third-party candidate, H. Ross Perot. And
once again, the third-party candidate's support hovered
around the 20 percent mark during most of the campaign.
But something was different this time. Perot's str ength did
not diminish at the end (he actually finished with 19 per-
cent). Why did Perot preferrers not desert their candidate
as Wallace preferrers had twenty-four years earlier?
Consider the three types of Perot suporters: (1) [P-C-B];
(2) [P-B-C] ; and (3) [P-(C,B)]. The wasted-vote argument
has clout only with voter types (1) and (2)- that is, with
voters who have a decided preference between the major-
party candidates, Clinton and Bush. Strategic voting has no
allure for type (3) voters. If a higher proportion of Perot
preferrers than Wallace preferrers were type (3), then there
would be less possibility of falloff in Perot support. An in-
teresting research project (I haven't done it, and don't know
of any work in this area at this writing) would be to com-
pare the different voter types for Perot and Wallace to de-
termine whether a preference-distribution argument could
account for the different falloff rates.
There is a second argument meriting investigation. In
1968, the election was seen as "too close to call." In 1992, in
contrast, by the last week of the election, Clinton was per-
ceived as pulling away from the incumbent president,
George H. W. Bush (who ended up with less than 38 percent
of the vote). A Perot supporter, even a type (1) or type
(2) supporter, could hardly be accused of wasting her vote
by casting it for Perot. Neither second-preference candidate
would benefit from a Perot supporter's switch if the election

weren't close-Clinton didn't need the help and it would be
too little too late for Bush. An alternative explanation,
then, is not that Perot supporters exercised less foresight
than Wallace supporters, nor even that they were differen-
tially distributed across preference types, but that, having
exercised foresight, a Perot supporter concluded that a vote
for Perot was apparently not a wasteful use of resources.
I have distinguished sophisticated voting from strategic

voting, although each is an instance of rational foresight. So-
phisticated voting is made possible by backward induction on
a fixed agenda. In the three-person presidential contest, on
the other hand, the issue is not one of voting contrary to pref-
erence at one node of a fixed agenda in order to achieve a more
desirable outcome at a later point; it is one of deciding
whether supporting your first choice is a hopeless undertak-
ing. Put differently, strategic voting involves weighing two dif-
ferent lotteries. The first lottery (in which you vote for a
Wallace or a Perot) gives, in comparison to the second lottery
(vote for your second choice), a slightly higher chance of your
first choice winning, along with a slightly higher chance of
your last choice winning, too. The second lottery gives a
higher chance of your middle alternative winning, reducing
the chances of either your most-preferred or least-preferred
outcome. A strategic voter, in effect, concedes that discretion is
the better part of valor.
H ERESTHETIC
I conclude the discussion of strategic behavior by claiming

that the sort of strategizing just described is "strategizing in
the small." For a clearly defined political situation, whether a
sequence of votes in a legislature or a national election, ma-
nipulation takes the form of not voting for the alternative
most highly ranked in terms of preferences. There's no doubt
that this is an important form of strategic behavior, as the
cases above suggest. But it is a restrictive view of strategic
possibilities, because it takes the situation confronting group
members as fixed and given. For instance, it does not ask
where the agenda came from in the first place. Or, how did the
election get shaped the way it did? Asking these questions
opens tip the possibility of "strategizing in the large," or what
William Riker called heresthetic.
You will not find this term in a dictionary, for Riker coined
it himself. He views heresthetic as the companion to rhetoric.
The latter-the art of designing an argumen~was a stan-
dard part of a young man's education in ancient times. Heres-
thetic-for Riker, the art of designing situations-is a word
made up of parts of appropriate Greek words for "choosing"
and "electing." Riker felt it should also have been part of that
ancient education, for making arguments without attending to
the larger strategic context is to strategize in the small, but
not in the large.
We have already seen the heresthetician at work in the
treatment of agenda-setting bodies, such as wily committee
chairs in legislatures. Their jobs consist of structuring the
content and sequence of voting-through proposing bills and
amendments-so that the result turns out as the agenda set-
ter would like.
Often, heresthetical maneuvers entail making something
seem other than what it really is. This is not so much a decep-
tion as a "redefinition" of a situation. For example, Senator
Warren Magnuson (D-Wash.), for relatively obvious reasons,

sought to get the U.S. Senate to block the Defense Depart-
ment from transporting potentially lethal nerve gas canisters
across his home state.11 Fearing that his opposition to this na-
tional defense activity would be construed as reflecting merely
parochial concerns-concerns that paled in significance to the
urgency of removing these dangerous military assets from the
post-Vietna m Pacific theater-he pursued a strategic tack
that did not recount all the potential dangers to which his con-
stituents might oo exposed during transit. Indeed, his argu-
ment steered clear altogether of the substance in dispute.
Instead, h e suggested that the issue at hand was really about
the constitutional powers of the Senate in foreign relations.
He suggested that the decommissioning of the nerve gas, with
its subsequent transit across the Northwest, was part of a
larger matter in which the president had failed to consult
with the U.S. Senate as was his constitutional obligation.
What was at stake, suggested Magnuson, was the very au-
thority of and respect for the U.S. Senate. This r edefinition of
the issue, as it happened, contributed to Magnuson's securing
an outcome he preferred. The important point here is not so
much that the senator won, but that he had the wit to see
that reinterpretation was a viable strategic maneuver that
promised the possibility of victory, where simply articulating
an argument based on his concerns for the welfare of his
Washington State constituents would surely have been dis-
counted, even ignored.
The Magnuson ma neuver bordered on the rhetorical, be-
cause it involved formulating an argument in order to per-
suade a small number of people on a well-defined issue.
Another form of heresthetic involves r edefinition on a grander
stage. Riker writes extensively about the (strategic) develop-
11 The entire episode is recounted in William H. Riker, The Art of Political Ma-
nipulation, Chapter 10.
ment of the slavery issue as an electoral heresthetic. 12 Briefly

stated, for much of the first half of the nineteenth century,
American national politics were dominated by the Jeffersonian-
Jacksonian coalition. Certainly by 1820, after the Federalist
Party had disappeared, this coalition was virtually unop-
posed. The coalition was united principally by the issue of
agrarian expansionism and found its greatest strength in the
middle Atlantic states, the South, and the states of the North-
west Territory. Opposition politicians, men like Henry Clay,
who were ambitious for themselves and their causes, searched
and searched for issues that might split this governing coali-
tion. Their substantive opposition to the agrarian expansion-
ism of the Jeffersonian-Jacksonians consisted in their desire
for public policy to encourage commercial development. But
this electoral contest between agrarian expansion and com-
mercial development had already been fought out over the
previous generation, with Jefferson (later Jackson) and his al-
lies winning big. No, the opposition would not win by simply
repeating the old arguments and fighting the old battles. It
needed to find a new issue that would split the currently dom-
inant governing coalition, one that would divide Mid-Atlantic
from Rim South, Northwest from Deep South. And the slavery
issue was the answer.
Riker makes the argument that slavery worked not be-
cause of its moral content (although large numbers of Ameri-
cans in the mid-nineteenth century found slavery abhorrent),
nor even because so many people were animated by abolition-
ist agitation. There are many morally significant issues float-
ing around at any particular time, but they do not necessarily
bring ruling coalitions down. Slavery worked as a strategic
maneuver because it divided the members of an existing
winning coalition, some of whom tolerated slavery and some
of whom opposed it. Once the northern elements of the
12 This is developed m Riker, Libera lism against Populism, Chapter 9.

Jeffersonian-Jacksonian coalition came to fear that support of

slavery on which their southern coalition partners depended
would be their own personal undoing, the coalition could no
longer hold. Subsequent events about which the historians
wrote-the Kansas-Nebraska Act of 1854, the Dred Scott deci-
sion in 1857, and ultimately civil war itself- put this coalition
to a n end. But it was the heresthetical maneuverings of losing
politicians looking for ways to become winners that set a ll this
in motion. 13
CONCLUSION
This chapter h as covered a number of subtleties of group be-

havior. It serves as something of an antidote, however, to ear-
lier discussions of group choice, because here individuals are
endowed with a capacity to consider the broader implications
of their actions .. Although I have referred to this as strategic
behavior and have occasionally characterized it in emotionally
13 'rhe logic of here:sthetic can be understood as the introduction of a new

issue, or the redefinition of an old one, in order to destroy a currently win-
ning coalition and replace it with some other. Students of politics should not
think that heresthetic is either rare or purely of historical interest. Issues
capable of splitting winning coalitions arise all the time. It only takes a
master beresthetician (like Warren Magnuson or Henry Clay) to use the
issue as a wedge to divide t he opposition. Modern issues exhibiting heres-
thetical traces include the gun-control movement's proposal to ban "cop-
killer" bullets and assault weapons in order to split the coalition between
the law enforcement community and the National Rifle Association; Ronald
Reagan's appeal to anticommunism and conservative social values to create
"Reagan Democrats" out of a portio n of his former opposition; the u se of t he
abortio n iss ue by Democrats to woo pro-choice Republican women; use of
that same issue by pro-life activists to induce Christian fundamentalists to
desert the Democratic Party; and finally, the exploitation by antismoking
forces of conflict within the tobacco industry between farmers and manufac-
turers over the matter of tobacco impo rts. Heresthetical maneuvers do not
always succeed. But they constitute the set of activities that those currently
out of power employ in an effort to get back on top . See Kenneth A. Shepsle,
"Losers in Politics (and How They Sometimes Become Winners): William
Riker's Heresthetic," Perspectives on Politics 1 (2003): 307-15.
charged terms (misrepresentation, manipulation, and so on),

all I have really done is to acknowledge the individual's capac-
ity to look beyond his or her nose, the individual's proficiency
in taking a longer-term view of things-in short, the individ-
ual's talent for behaving deliberatively and exercising fore-
sight.
Foresight comes in many shapes, and I have covered some
of them here. Sophisticated behavior, especially in the context
of legislative settings, is the capacity to make voting decisions
in a sequential process with an eye to final results. Sometimes
this entails voting contrary to nominal preferences- for exam-
ple, voting against an amendment you like because you know
it will damage the chances for the whole bill to survive. For
this reason, I have alluded to individuals misrepresenting
their preferences. Really, though, they are just taking care of
business in the most sensible fashion available to them. Like-
wise, in the electoral setting, voters who elect not to vote for
their favorite candidate because he doesn't have much chance
of winning are clearly behaving strategically; but here, too,
citizens are simply engaging in a perfectly legitimate activity,
namely, using the instruments at their disposal (their votes at
the very least) to effect outcomes in a direction they prefer.
Sophisticated behavior is also associated with activities
other than voting. The committee chair's judgment call on
whether to open the gates or not is one manifestation of this
kind of exercise in foresight. The opposition politician's injec-
tion of new issues into an electoral campaign is another. In
each case politicians use the resources at their disposal (con-
trol of the legislative agenda and influence over public opin-
ion, respectively) to accomplish goals- policy goals in the case
of the legislative chair, electoral goals in the case of the oppo-
sition politician. Sophistication resides in their ability to use
the assets at their disposal instrumentally.
To appreciate fully the strategic options available to indi-
viduals, then, it is clear that we must understand the context
in which they operate, for it is the context that provides them

with opportunities to deploy their resources instrumentally.
This is no more apparent than in the world of electoral poli-
tics, where different electoral arrangement effectively consti-
tute altogether different contexts in which to deploy resources.
This is precisely our agenda for the next chapter.
E XPERIMENTAL CORNER
Agenda Setting and Group Choice

The strategic behavior put on display in the material in this '
chapter is multifaceted and many-splendored. Most of the
instances covered here involve voters or legislators misrep-
resentating their "honest" preferences by casting strategic
or sophisticated votes. Likewise, in Case 6.1, a university
department misrepresents its honest evaluation of gradu-
ate admissions candidates in order to make the most of its
limited resources. In all of these circumstances the actors
take the alternatives on offer as given and, based on what's
available, figure out how they can best accomplish their
personal objectives. In this experimental corner I describe a
different kind of strategic behavior-setting the agenda
from which choices will ultimately be taken. This is the
focus of a wonderful paper by Plott and Levine.a
The motivatiion for this experiment came from a situa-
tion of personal significance to Plott and Levine: "As a prac-
tical matter, we were involved in an important and complex
committee decision. A large flying club in which we held
membership was meeting to vote upon the size and compo-
"Charles R. Plott and Michael E. Levine, "A Model of Agenda Influence on

Committee Decisions," American Economic Review 68 (1978): 14~0 .
Also see William H. Riker, The Art of Political Manipulation (New
Haven: Yale University Press, 1986), Chapter 3.
sition of the aircraft fleet which would be available to the

membership for flying. As members we had preferences
about the fleet available to us and an opportunity to shape
the agenda. Preliminary discussions and meetings had nar-
rowed the range of possibilities greatly. ... Over these re-
maining possibilities, however, there were conflicting and
strongly held opinions. The group was to meet once and de-
cide the issue by majority vote . . .. The meeting was held.
The group used our agenda. The decision was the one we
predicted" (p. 146).
Plott and Levine wondered whether their success in ma-
nipulating the group's choice by selecting an agenda that
yielded the result they wanted was an accidental piece of
good fortune or something more general. They set up an ex-
periment to test their hunches. Their experiment involved
partitioning alternatives into subsets and having subjects
select a subset. The winning subset, in turn, is further par-
titioned and a subsequent choice is taken. This continues
until a unique alternative remains. For example, suppose
the issue were one of planning a dinner party in which two
choices had to be made: cuisine (French or Mexican) and
dress (formal or informal). The four possible outcomes are
{French formal, French informal, Mexican formal, Mexican
informal}. One agenda requires the choice of cuisine to be
taken first and dress next. Consequently, subjects choose
between two subsets: {French formal, French informal} and
{Mexican formal, Mexican informal}. The winner- say, the
first subset-then becomes the set of alternatives on offer
when choosing dress. Another agenda would have dress
chosen first, then cuisine: {French formal, Mexican formal}
versus {French informal, Mexican informal}. Other agendas
are possible, though they entail choices across these two
categories, for example, {French formal, Mexican informal}
versus {French informal, Mexican formal}. Thus, the
agenda allows group members to cast votes, but limits the
items up for a vote and their order. The experimenter gets

to select the agenda (with an eye to "manipulating" the
group into choosing the final outcome he or she likes best).
Plott and Levine focus on three voting strategies that
participants might adopt: (1) sincere voting-at each parti-
tion, vote for the subset that contains the participant's
most-preferred alternative (from among those alternatives
still alive); (2) avoid-the-worst-at each partition, vote for
the subset that does not contain the least-preferred alterna-
tive among those still possible; and (3) average value-at
each partition, treat the alternatives in a subset as a lottery
with each having equal probability and vote for the subset
with the highest expected utility. Plott and Levine do not
know, for any experimental subject, the particular decision
rule he or she is using.
Experimental subjects were students from Caltech,
UCLA, and USC. They were gathered in a classroom, given
a group decision to select a letter from a subset of the al-
phabet, and were provided with a payoff sheet indicating
their particular monetary payoff depending upon which let-
ter was chosen. Thus, each subject had induced preferences
over the letters. Decisions were made by majority rule.
The induced preferences of the experimental subjects
were such that, from the set {A,B,C,D,E} of alternatives,
each of th e first four alternatives was preferred to E; A was
preferred by a majority to every other alternative; and B, C,
and D were part of a majority rule cycle (B preferred to C, C
preferred to D, and D preferred to B). Agendas were de-
signed by the experimenters so that if the subjects were of a
particular type-(1), (2), or (3) above-and behaved accord-
ing to the theory, then a specific outcome would prevail.
(Actually, Plott and Levine have a more complicated theory
in which they assume probabilities for the different types
and design an agenda in which the expected outcome can be
derived from their theory.) For example, con sider the parti-
tion {A,B,E} versus {C,D}. If everyone is type (1), a majority

will select the first subset, since a majority prefers A to any
of the other alternatives in the other subset. If the partition
of this winner is {A} versus {B,E}, then A will prevail. If
everyone is type (2), then a majority will select {C,D}, since
E is worst for a majority, and then, from among these, C
will prevail (since C is preferred by a majority to D). If
everyone is type (3), then if their payoff for E is made espe-
cially bad, they will opt for {C,D} and then choose C; but if
E is not sufficiently bad, then they might opt for {A,B,E}
and then ultimately for A. More generally, the experi-
menters can develop expectations for which outcome they
can induce by their strategic choice of agenda.b
Running this experiment across many groups of experi-
mental subjects, the experimenters find that their expec-
tations are extraordinarily accurate. Space precludes a
detailed discussion of the results, so readers should consult
the paper on their own. Plott and Levine conclude that, at
least in the laboratory, "the agenda can indeed be used to
influence the outcome of a committee decision" (p. 156).
This provides some empirical support that agenda power is
a manipulable strategic resource-something that legisla-
tive committee chairs, academic department heads, and
those who lead meetings undoubtedly discover. Rank-and-
file members, on the other hand, will want to avail them-
selves of parliamentary protection to reduce the degree to
which they can be exploited by clever agenda setters.
b If, instead of hypothesizing that all group members are of a specific type,
the experimenters assume some distribution of types across the group
members, they can still deduce probabilistic expectations associated with
particular agenda choices.
1. Suppose that strategy c3 is unavailable to Mr. III in the game

displayed in Figure 6.1. Use backward induction to solve this
amended game. Now suppose that c3 is again available to Mr.
III, but Mr. I can no longer play a l. What is the final outcome?
2. Why would supporter s of a particular bill ever vote in favor

of a killer amendment? Some things to consider might include
how constituents evaluate their legislators and uncertainty
about whether an amended bill will pass.
*3. For this question we return to the setup of Problem 1 in

Chapter 5 to see what happens when individuals in an agenda
setup vote sophisticatedly.
• It is the last round of a three-item agenda, v, so the soci-

ety is voting over the option that won the first round and
the final option on the agenda. Will any player wish to
misrepresent her true preferences? Try out some specific
head-to-head matchups (e.g., x vs. y, or y vs. z) to build
up your intuition.
• With the agenda v = (y,x,z), first determine what hap-
pens in the final round depending on whether y or x
wins the first round. Based on this, can player 2 ever do
better by supporting x in the first round, contrary to her
nominal preferences? What about 3 voting against x?
Knowing this, should 1 misrepresent his preferences by
playing y in round 1?
• (Bonus) Identify the outcomes (and which player acts
strategically) if the agenda is u' = (z, x, y) and v" = (z, y,
x). Compare these answers with the honest voting out-
comes.
4. Why do some voters "waste" their vote by supporting third-

party candidates who have no chance of winning office? Why
do other voters, who may prefer a third-party candidate,
nonetheless vote for someone else? Discuss the phenomenon of
strategic voting in plurality elections and illustrate your argu-
ments with reference to illustrious third-party candidates in
the United States- for example, Ralph Nader (2000), Ross
Perot (1992), and George Wallace (1968). Why have the expe-
riences of third-party candidates (in terms of election day
dropoffs in support) been so different?
5. In the study of legislatures, political scientists often rely on

voting scores to measure the preferences of legislators on pol-
icy issues. For instance, the League of Conservation Voters
compiles a list of key votes on environmental matters during a
session of Congress and then ranks members of Congress
based on how often they took the "pro-environment" side. Will
we get meaningful results from these types of scores? Under
what circumstances? In answering this question, consider the
role of the closed rule (which allows committees to make take-
it-or-leave-it offers), strategic voting. and sequential voting/
minimum winning coalitions (discussed in Case 6.2).
6. Is an assumption of sincere voting ever suitable for analyz-

ing politics? In a legislature? Among voters in an election? On
a small decision-making committee?
7. Explain the meaning of the Gibbard-Satterthwaite Theo-

rem in your own words, being careful to define terms like
strategy-proof, manipulable, and sophisticated voting. What
are the implications of the theorem for the normative argu-
ments in favor of democracy? Of what significance is the theo-
rem for social scientists trying to make predictions about
political outcomes?
7
Voting Methods and
Electoral Systems
In the last few chapters an implicit theme has emerged: It is

nearly impossible to arr ange for the making of fair and coher-
ent group choices. Preference cycles, agenda manipulation,
strategic misrepresentation of preferences, heresthetical ma-
neuvers, and so on frustrate our best attempts. The coup de
grace, developed in this chapter, is that "popular sovereignty"-
by which we mean any method for allowing individuals in a
group to affect their own fates through voting-is not unam-
biguous either. There are lots of different ways to cast and
count votes or "do" majority rule, for instance. If all these meth-
ods differed only in the details but not in the final result, then
we could relegate the matter of details to politics junkies to
chat about. Alas, the devil is in the details. In this chapter,
therefore, I explore the procedural context of voting-the rules
by which small committees and large electorates make choices.
Tbe discussion is partitioned into two sections according to
what it is the group is choosing. The first part of the discus-
sion focuses on how relatively small groups-a set of friends, a
club, a committe~hoose some alternative from a set of
available alternatives. I call these arrangements uoting meth-
ods. The second part of the discussion emphasizes how rela-
tively large groups (called electorates) choose a specific thing
(called a legislature). I call these arrangements electoral sys-
tems. In each part of this discussion, I am not so much inter -
191
ested in conveying specific details about the mJl·iad ways of

making group choices as in hammering home the simpler
facts that there are myriad ways of making group choices, that
each has merit in some circumstances, but that the rationale
for none seems decisive or compelling in all situations.
V OTING M ETHODS
The Problem with Methods of Voting-They Matter!
Getting down to business, suppose we have a group of 55 indi-
viduals, choosing among five alternatives, {a, b, c, d, e}. 1 Of the
120 possible complete and transitive strict preference order-
ings an individual might adopt as his or her own, 2 there are
only six distinct orderings, or "opinions," represented in this
particular group. They are listed in Table 7.1, along with the
number of group members holding each. (The underlining in
Table 7.1 will be explained later.) For the sake of discussion,
six different "reasonable" ways for the group to arrive at a
choice among the five alternatives are considered. The reader
may well be able to devise others and should rest assured
that, in human history, a multitude of alternative methods
have been devised. 3
1
This absolutely evi l example was invented by Joseph Malkevitch and is dis-
played in Figure 2 of his article, "Mathematical Theory of E lections," Annals
of the New York Academy of Sciences 607 (1990): 89-97. For another exa mple
that shows some of the outcome variation that arises in moving from one
voting system to another, even with preferences held fixed, see Donald G.
Saari, "Chaos, But in Voting and Apportionments?" Proceedings of the Na-
tional Academy of Sciences 96 (September 1999): 10568-71.
2 Recall that there are five possible first-preference alternatives, four remain-
ing possibilities for second preferences, and so on-or, 5 x 4 x 3 x 2 x 1 • 120
ways to strictly order five alternatives.
3 A considerably more systematic treatment of alternative voting methods is
found in William H. Riker, Liberalism Against Populism (San Francisco:
Freeman, 1982), Chapter 4. Somewhat more technical, yet quite insightful,
is Peter Fishburn, "A Comparative Analysis of Group Decision Methods," Be-
hauioral Science 16 (1971): 38-44.
Voting Methods and Electoral Systems 193
TABLE 7.1
AN EVIL EXAMPLE
I II III IV v VI
(18) (12) (10) (9) (4) (2)
a b c d e e
d e b c b c
e d e e d d
c c d b c b
b a a a a a
SOURCF.: Joseph Malkevitch, "Mathematical Theory of Elections," Annals
of the New York Academy of Sciences 607 (1990): 89-97 (Figure 2)
Consider, as a first method, the simplest of them all:

simple plurality voting. Each voter casts a single vote for a
single alternative, and the alternative with the most votes
wins. This is one of the hallmarks of the Anglo-American sys-
tem for electing legislators. 4
A slight variation is the plurality runoff, in which each
voter casts a single vote for a single alternative, and the two
alternatives with the most votes move to a second stage in
which the balloting is repeated between these two survivors
according to simple plurality voting.
An even more intricate and general form of runoff is the
sequential runoff. Here each voter casts a single vote for a
single a lternative, the alternative with the fewest votes is
eliminated, and the balloting is repeated. This procedure con-
tinues until there is a single alternative left.
The fourth method we examine allows voters to express
preferences about all the alternatives. According to the Borda
count, a scoring system much like that used in international
track competitions, each voter expresses personal preferences
•When discussed in the latter h alf of t his chapter as an electoral system, it

wi II be called first pa.~t the po~t.
over the five alternatives by awarding four poi nts to his or her
first choice, three to the second choice, two to the third, one to
th e fourth , and none to the fifth. These points a re totaled and
the alternative with the most points wins.
Fifth, the Condorcet p rocedu re seeks to determine
whether there is some specific alternative that can secure a
majority agains t each of the others in a pairwise round-robin
tournament. If so, t hat is the winner. If not, then we will need
to provide some a lternative procedure (perhaps one of the
others).
Finally, consider the method of approval voting, invented
by the political scientist Steven Brams and the operations re-
search scholar Peter Fishburn. It puts no limit on the number
of votes an individual can cast. Each individual casts votes for
all those alternat ives she "approves of." This means that if she
wishes, she may cast votes for all the alternatives, none of
them, or any number in between. The winner is the alterna-
tive that receives the most approval votes. 5 All of the methods
are listed in Display 7.1.
With the data of Table 7.1 we can determine how the vari-
ous forms of popular sovereignty listed in Display 7.1 perform.
Voters are assumed to vote sincerely. Display 7 ..2 provides the
results, with the winning alternative given in bold.
The simple plurality method produces a victory for a, since
it has the most first-preference supporters, though a peculiar
victory it is, given that all but the first group hate this alter-
native. This is made all too apparent when we look at the plu-
rality runoff procedure in which b triumphs; indeed, any
alternative that made it to the "finals" against a would have
5
An individual who votes for all the alternatives states that all are "above
threshold," or acceptable. The impact on the final outcome is exactly the
same as the voter who abstains (or, equivalently, states that she approves of
no alternative). What matters in approval voting is that an alternative do
well relative to its competitors; its absolute vote total is less important.
DISP LAY 7.1

S OME VOTING M ETHODS
Me thod Descr iption

Simple plurality voting Alternative with most votes (plurality)
wins.
Plurality runoff 'lbp two vote getters move to a second
round; new balloting determines
second-round winner by simple
plurality voting.
Sequential runoff Alternative with fewest votes is
eliminated and balloting repeated;
elimination procedure continues until
one alternative remains.
Borda count Alternatives assigned points in accord
with voter rank-orders. The alternative
with the largest sum of points wins.
Condorcet procedure Pairwise round-robin tournament
determines if one alternative defeats
each of its rivals.
Approval voting Each voter casts votes for any
alternative he or she approves of. The
alternative with the most votes wins.
beaten it. The sequential runoff procedure produces c as the

final outcome. The Borda count gives the victory to d. And the
Condorcet procedure shows that alternative e receives a ma-
jority (28 or more of 55 votes) against every other alternative.
In short, each of the first five preference-based methods of
group choice yielded a different winner. The sixth, approval
voting, yielded a tie between d and e.
So we must conclude that the rules of preference aggrega-
tion matter, and sometimes (as in this example) they matter a
lot. It is evident in this example that whoever chooses the
method of counting noses determines, finally and decisively,
DIS PLAY 7.2

ELECTROAL SYSTEM R ESULTS
Simple plurality: a=l8 b-12 c-10 d-9 e-6
Plurality runoff'
round 1 a=l8 b=l2 c-1 d=9 e=6
round 2 a-18 b=37
Sequential runoff-
round 1 a=l 8 b=l2 c=lO d=9 es6
round 2 a=18 b=l 6. c=l2 d-9
round 3 a=18 b-16 c=21
round 4 a-18 c=37
Borda count: a-4(18) + 0(12) + 0(10) + 0(9) + 0(4) + 0(20)- 72
b-0(18) + 4(12) + 3(10) + 1(9) + 3(4) + 1(2)-101
c•l(18) + 1(12) + 4(10) + 3(9) + 1(4) + 3(2)-107
d-3(18) + 2(12) + 1(10) + 4(9) + 2(4) + 2(2)-136
c-2(18) + 3(12) + 2(10) + 2(9) + 4(4) + 4(2)-134
Condorcet:*
a b c d e
a 18 18 18 18
b 37 16 26 22
c 37 39 12 19
d 37 29 43 27
e 37 33 36 28
Approval:t
a 18+0+0+0+0+0 -18
b 0+12+10+0+4+0 -26
c 0+0+10+9+0+2 -21
d 18+12+10+9+4+2 =55
e 18+12+10+9+4+2 =55
* Reading across each row in the matrix give;. the number of votes that
each row alternative gets when paired against each of the column alter·
natives. Thus, reading across the second row, alternative b gets 37, 16, 26,
and 22 votes out of 55 against a, c, d, and e, respectively.
t Each group is assumed to cast votes for all the alternatives above the
line in Table 7.1. Thus, every group but the third votes for three of the five
alternatives; the third group approves of four out of the five.
the final outcome. Put somewhat differently, whatever proce-

dure is in place for choosing the method determines the final
outcome. But putting procedures in place is the business of
constitutions; they either say directly what method will apply
or indicate who (or what body) gets to decide. No wonder con-
stitutional politics are such struggles! So much is at stake.
And while the example above provides something of a worst-
case scenario (what with each method producing a uniquely
different o,utcome), the reader should understand that it is not
an altogether extreme instance. As the section title says, the
problem with voting methods is that they matter!
Thinking about Voting Methods

At present there is no generally accepted way to think about
voting methods. There are so many different ways to vote, and
so many potentially useful criteria to bring to bear on alter-
native systems, that it is easy to become quickly confused.
Here I want only to suggest a couple of directions for thought.
Generally speaking, a voting method may be thought of in
terms of (1) the inputs required, (2) what the procedure does
to those inputs, and (3) the output or outcome produced. That
is, the final outcome is a function of the inputs (written: out-
come - F(inputs)), and we can think about each of the three
italicized components separately.
1. In terms of inputs, plurality voting makes the simplest
demands on voters (and perhaps for this reason, it is a com-
monly used method of group choice); each person must simply
name an alternative (his or her first preference if a sincere
voter, something else otherwise). On the other hand, the de-
mand on voters in a runoff plurality election depends upon
how it is administered. If voters are expected to show up for
two separate rounds, then less information at each round is
required (but the cost of showing up both times is greater). In
round one, the same data as in a simple plurality contest is re-
quired. In round two, relative preference between the two

highest vote getters in round one is needed. If voters wish to
economize on trips to the polls. showing up only once, then
they must provide more information on that one occasion. Spe-
cifically, they must provide information on every possible pair-
wise comparison (if there are five a lternatives, then there are
ten comparisons). since it is not known in advance which pair
will advance to the second round. H owever, all this compari-
son data is contained in a voter's preference ordering, so that's
all a voter need provide at the outset. This is precisely the
same data required for a sequential runoff election, a Borda
count procedure, and a Condorcet procedure, too. Approval
voting requires a s much information a bout preferences as
each voter wants to reveal. 6
Allowing these six methods to stand for the many hun-
dreds of voting methods that have sprung from human cre-
ativity, the point here is that methods differ as to what they
require-a single alternative, a subset (of whatever size each
voter wishes), or an entire preference ordering. There may be
grounds for preferring one method over a nother, quite apart
from the particular result each may yield, based on the ease of
administering it or on the desire to economize on the burden
of the voter. On the other hand, the necessary inputs may de-
pend on what you want to get out of the group choice-some-
thing discussed below. Wherever one stands on these or a host
of other criteria for thinking about inputs, it is patently clear,
on the basis of required inputs alone, that democratic voting,
broadly understood, takes on a multiplicity of forms.
2. I won't spend much time on the procedures themselves
and what they do, since I described a few of them already in
Display 7.1. I do want to point out, however, that these proce-
6 That is, the voter can either submit a subset of the full set of alternatives
(the "approved of" alternativeis) or, as in Table 7.1, he or she ca n hand in a
preference ordering with a line drawn below the "approved of" alternatives.
dures (and any others you can think up) have their peculiari-
ties. Plurality rule, for instance, is especially odd. Alternative
a was the plurality winner in the example above, yet it loses to
every other alternative in pairwise comparison. Additionally,
the Condorcet ·.vinner, e, which many would take as a strong
normative candidate for the group choice, actually got the
fewest votes in the plurality contest. Runoffs, whether simple
or sequential, have the perverse possibility of eliminating an
alternative that can beat every other in a pairwise contest (e
never made it very far in these runoffs). The Borda count
method (indeed, this is true of all the methods) is very vulner-
able to strategic behavior. Notice that the twelve voters of
group II or th e ten voters of group III in Table 7.1, who prefer
e to d, can actually give e a victory by misrepresenting their
preference ranking for d (pretending it is lower in their rank-
ing).7 The Condorcet procedure does not always produce a
winner- and th.en what do you do?
Finally, there is an issue that applies to each of the voting
methods we have described, but we will ruscuss it in terms of
approval voting, since its proponents seem so unperturbed by
it. 8 In the example above there are five alternatives. Those al-
ternatives might be various motions (say, what movie the fra-
ternity house should rent this evening) or candidates (say,
which of the sorority sisters should be the representatives on
the Greek Council). However, which motions are moved or
which candidacies are activated depends intimately on the
voting method, F. You might figure, for example, that The
Hurt Locker would get a lot of second-choice votes from your
7 Borda, a member of the French Academy of Sciences in the late eighteenth

century, was informed by one of its other members, Condorcet, that the
Borda count could be manipulated. Borda is reported to have sniffed, "My
method is only for honorable men."
8 See, es pecially, the various writings of Brams and Fishburn on the merits of
approval voting, including "Approval Voting," American Political Science Re-
view 72 (1978): 831- 47, and their book Approval Voting (Boston: Birkhau-
. ser, 1983).
frat brothers, and thus have a good chance of winning if the

voting method were the Borda count. But you also believe it
wouldn't get many first-preference votes, so you probably
wouldn't even bother proposing it if the decision rule were plu-
rality voting. To look at a fixed set of a lternatives and compare
the outcome under plurality against that of approval or the
outcome under Borda against that of Condorcet, for example,
misses this point. In the jargon of the field, the set of alterna-
tives or candidates is endogenous-that is, highly dependent
on the method of counting heads. In this regard, my hunch is
that approval voting encourages a larger number of alterna-
tives to be brought forward (so to speak) than many other vot-
ing methods. Candidates know they do not have to be the top
choice of a voter, but merely among the "approved of" set, and
thus may find it easier to rationalize their prospects of victory.
Likewise, a motion need not be the favorite of many voters but
only among the favorites to prosper under approval voting, a
fact that may give encouragement to potential motion makers.
So, the question comes down to whether it is better or worse
for a group to have a rich set of items from which to choose or
a more spartan set. Is more always better than less? This
question will take on an interesting political significance in
the second part of this chapter when we examine proportional
representation versus other systems for electing legislatures.
3. It may seem odd that we even need to discuss the output
of a voting method, since it is no more than the thing that is
chosen. But exactly what is that "thing''? We have somewhat
abstractly described alternatives by letters, suggesting that
the thing the group must choose is some unitary entity, some
element of the set {a, b, c, d, e}. But we can quickly complicate
matters quite a bit. Shortly, for instance, I will talk about
ways of choosing members of a legislature. The "thin g" here
could be a single legislator, a group of legislators from a multi-
member district, or the legislature in its entirety. Or, to give
an example with a different emphasis, imagine that what we,
as a group, are doing is choosing instructions to give our

agent. Imagine that our agent must choose among the five let-
ters for us, but she will not know in advance (nor will we)
whether all five are available or only some subset thereof. She
will need to know more than the group's favorite, since that
alternative may turn out to be unavailable. Thus, the group,
in this instance, needs to choose a collective preference order-
ing to guide its agent. The point here is that one set of criteria
to evaluate a voting method is relevant when all the group
needs to do is choose a letter, but an entirely different set may
come into play •nhen the group needs to come up with a full
preference ordering. The nature of the output, then, should af-
fect the way we think about voting methods.
This entire discussion produces a serious philosophical
puzzle. If the "wish of the group," or the "collective will," or the
"public interest'' -whatever you want to call the output of
group deliberation-is to be ascertained from the inputs that
the individual group members bring to the voting method (and
those inputs vary from method to method), then how are we to
give meaning to "wish of the group," "collective will," or "pub-
lic interest''? Using one method may yield one conclusion
about what the group wants, while using a different method
yields a different conclusion because it operates on some new
set of alternatives. This is crazy! But it is even worse. Sup-
pose, as in the example associated with Table 7.1 , the alterna-
tives under consideration r emain constant, and that voter
preferences have also remained fixed. Then, with alternatives
and preferences fixed, it seems only natural to presume the
wish of the group is well defined- it is whatever it is. The only
thing that might change is the way in which those wishes are
revealed or ascertained by the voting method. Yet we have
seen that the outcome does change (most evilly displayed in
the example at th e beginning of this chapter). But surely it
seems perverse to conclude that the group wish has changed
just because the method we have used to ascertain it-and
only the method-has changed. Are we driven to this conclu-

sion? We ask the reader to think hard about this question, for
it has motivated much of the discussion of the last five chap-
ters. I know of no definitive answer to it, although I shall ex-
amine some in the summary that follows this chapter.
This part of our discussion is in a woefully incomplete
state. Our principal purpose, however, has been less to pro-
vide a broad summary of the myriad methods of conducting
group voting than to convey by illustration a sense of fragility
in group life. The decisions a group reaches, as the last few
chapters have su ggested, depend not only upon the options
made available, not only on the order in which some agenda
setter presents them, not only on the degree to which group
members reveal or misreveal their preferences, but also on the
way they conduct the actual decision making. And all those
other things, likewise, are influenced by the voting method we
adopt. A group decision surely reflects member preferences.
But it also reflects much more, a theme to which I will return
in the summary.
ELECTORAL SYSTEMS
Just as there are many voting methods, there also is an in-

credible variety of electoral systems. I restrict consideration to
systems for electing legislatures, using these institutions to
represent a broad class of elected governance arrangements. 9
I claim here that electoral systems may be thought of in terms
of the degree to which their "core value" is representation or
governance. By the former I mean an electoral arrangement
that places priority on the degree to which the elected reflect
9 In a sense, a legislature is a generic elected institution. For example, a pres-

ident, governor, or mayor may be thought of as a one-person legislature. an
elected court as a one· (or more) person legislature, and so on.
(or represent) the beliefs and preferences of the electors. By

the latter I mean an arrangement yielding elected representa-
tives capable of acting decisively, of governing. Obviously, both
of these purposes are noble. Yet, they often operate at cross-
purposes because an arrangement that emphasizes represen-
tativeness may make governance more difficult, and vice
versa. 10
Associated with the end of the spectrum giving priority to
representation is the broad family of electoral methods known
as proportional representation (PR). At the other end of the
spectrum is the family of plurality voting methods (referred to
ear lier in this chapter). The discussion begins with plurality
methods that are common in the United States a nd Great
Britain. This will be fo1lowed by a treatment of the mor e ex-
otic PR methods found in continental Europe. Finally, argu-
ments are offered on why these broad classes of electoral
system are seen as either representation-oriented or gover-
nance-oriented.
General Remarks
One of the leading contemporary students of the theory of
electoral systems, Gary Cox, has defined an electoral system
in terms of five bits of information. 11 For Cox, as for us, the
critical separation is between plurality and proportional sys-
tems, but five bits of information can be used to characterize
10 For an elaboration on this theme, with special reference to the U.S. Con·
gress, see Kenneth A. Shepsle, ''Representation and Governance: The Great
Legislative Tradeoff," Political Science Quarterly 103 (1988): 461-84.
11 'l\vo especially important papers by Cox are mathematically advanced but
well worth examining, if only to get a feel for the kinds of analysis scholars
like Cox are able to do. See his "Electoral Equilibrium under Alternative
Voting Institutions," American Journal of Political Science 31 (1987):
82- 109; and "Centripetal and Centrifugal Incentives in Voti ng Systems,"
American Journal of Political Science 34 (1990): 903-36. These are su mma·
rized and extended in Gary C. Cox, Making \!Otes Count (New York: Cam·
bridge University Press, 1997).
each. These five describe the resources given to voters, what

the voters can do with their resources, and finally, how the
electoral for mula produces a final outcome-in effect, the in-
puts, procedures, and outputs discussed earlier. Generally
speaking, Cox maintains the following distinction between
plurality and proportional arrangements. 12
By a plurality formula, I mean one in which voters cast votes for
individuals (rather than party lists) and the top ... vote-getters
win seats .. .. Proportional formulas, on the other hand, are those
in which voters vote for parties, and seats are allocated in propor-
tion to the vote polled by each party.
The five bits of information are v (number of votes per

voter); p (if u > 1, whether voters must cast all u votes, or may
partially abstain); c (if u > 1, whether voters may cumulate
their votes, or must distribute them); k (the number of legisla-
tors to be elected per district, known as the district magni-
tude); and f (the electoral formula). Electoral systems can be
represented by this information, and Display 7.3 lists some
common plurality types.
Plurality Systems: First (or More) Past the Post

The most famous of the plurality systems is single-member
districts, first-past-the-post (FPP). As Display 7.3 describes,
each voter gets one vote, may cast it for any candidate he or
she pleases, and the single candidate with the most votes (not
necessarily a majority) is elected. The legislature thus consists
of legislators elected from separate districts in this manner.
This is the electoral system found in Great Britain and many
of its former dependencies (including, of course, the United
States).
The key feature, it seems to me, is that each district, or
constituency, gets but a single representative. This may
12
"Centripetal and Centrifugal Incentives in Voting Systems," 905-906.
DISPLAY 7.3
ALTERNATIVE ELECTORAL SYSTEM S
v p c k f
First Past the !Post 1 no no 1 Plurality
(FPP)
Single 1 no no k>l Plurality
Nontransferable
Vote (SNTV)
Limited Vote (LV) <.k yes no k Plurality
Cumulative Vote (CV) s.k yes yes k>l Plurality
mean-and thi.s is in fact a common complaint with the sys-

tem-that the winner is not particularly representative of the
district in which he or she is elected. The district may be
60 percent male and 40 percent female; whoever is elected will
not represent (in the sense of "reflect") a rather sizable chunk
of the electorate. Another (melting pot) district may be 25 per-
cent Roman Catholic, 23 percent Greek Orthodox, 16 percent
Jewish, 15 percent Baptist, 11 percent Episcopal, and 10 per-
cent agnostic. Its representative will not share religious and
cultural traditions with at least three-fourths of the con-
stituency. If the district magnitude were more generous-if k
were larger- then it would be possible to represent more of a
district's heterogeneity. But it would also mean a larger legis-
lature, for one thing, and one almost certainly with a greater
heterogeneity of views. This might make it more difficult for
representatives to govern- to debate, deliberate, form coali-
tions and compromises, and ultimately come to some conclu-
sion on public policy issues facing the society. This is the
"great trade-off" between representativeness and governance.
Oddly enough, an ethnically homogeneous society, Japan,
until very recent ly employed a plurality system with a larger
district magnitude. It is identified in Display 7.3 as the method
of single nontransferable vote (SNTV). In this arrangement
each voter is still endowed with only one vote, but now the k
highest vote getters are elected from the district, where h is
given in advance as the district's magnitude.
Just as SNTV is a small alteration of FPP (namely, a
change in di strict magnitude from 1 to some k > 1), the
method of limited vote (LV) is only a slight alteration of SNTV.
Specifically, LV endows each voter with more than one vote
(but fewer than h), still allowing for k winners. If, under FPP,
one representative is to be elected from a district and each
voter casts one vote, and, under SNTV, more than one repre-
sentative is to be elected from a district and each voter (still)
casts one vote, then, under LV, more than one representative is
elected from a district and each voter may cast multiple votes.
Thus, a district m ay elect four legislators under LV by giving
each voter, say, two votes. This method, however, offers an ad-
ditional strategic maneuver to voters-"plumping" (as the
English called it in the eighteenth century), or voting only for
your favorite candidate. 13 Consider the district with k = 4 and
v = 2 as just mentioned. Suppose I am is considering casting
my votes for my two favorites in the field of candidates. The
latest public opinion poll showed my second-choice candidate
running in fourth place and my favorite candidate in fifth. If I
proceed to vote for both candidates, my second choice may
win, but just possibly at the expense of my first choice, who
will finish just ou t of the running. I might be better advised to
cast only a single vote for my first choice, thus foregoing the
support I had planned to give to my second choice, but just
13 In Sa/ire's Political Dictionary (New York: Random House, 1978), the follow-
ing entry is found und er plump:
"One of the English election phrases for which there is no equivalent. in
the United States," wrote the New York 'Jribune in 1880. "is 'plumping.'
Whenever [an English] constituency returns two memhers, each voter ca n
give one vote each to any two ca ndidates but he cannot. give his two votes
to any one candidate. If he chooses he can give one vote to only one can-
didate, and this is t ermed 'plumping."'
Vbting Methoas and Electoral Systems 207
possibly helping to elevate my favorite into fourth place. The

LV method permits this sort of partial abstention.
Cumulative voting (CV) goes one step further, permitting
voters not only partially to abstain, but also to cumulate their
votes. 14 In the example of the previous paragraph, where the
voter has two votes and there are four candidates to be elected,
but the method is now CV, he could cast both of his votes for
his first-preference candidate. The state of Illinois, until 1980,
elected state legislators from multimember districts via CV.
Similarly, nearly half of the states permit and more than one-
fourth actually require CV in elections for boards of directors
of publicly traded corporations. The idea behind CV is that
well-organized minorities, by cumulating their votes, can as-
sure themselves a modicum of representation. Cumulative
voting in multimember districts, for example, has become one
of the "methods of choice" for a number of civil rights activists
for electing county commissioners, school boards, and other
local officials, primarily in the South, in order to assure mi-
nority representation. 15
Equilibrium in Plurality Systems

Scientists and engineers can usually make intelligent remarks
about and comparisons among alternative engines. The same
is true here regarding the electoral engines just described;
thus far, however, man has been much more creative in in-
venting electoral engines than in understanding their operat-
ing characteristics.
14
Safire, in the entry referred to in the previous footnote, goes on to indicate
that the American usage of the term "plump" differs from the English. In
the United States, "to plump" means to cumulate your votes for a particular
candidate.
15
For various points of view, see Lani Guinier, "The Representation of Minor-
ity Interests," and Kenneth Benoit and Kenneth Shepsle, "Electoral Sys-
tems and Minority Representation," both in Paul Peterson, ed., Classifying
by Race (Princeton, N.J.: Princeton University Press, 1995), pp. 21- 49 and
50-84.
-- ----------------------------
The key to understanding the equilibrium tendencies of the

alternative plurality systems we have described is the number
of candidates in a district. But .before developing this thesis, I
first must say what I mean by equilibrium in this context. Re-
stricting analysis to one-dimensional spatial representations
(like those described in the first part of Chapter 5), each voter
is assumed to have a unique ideal point and single-peaked
preferences on the one dimension. Candidates compete for
votes by identifying with specific locations on the dimension.
Qualitatively, two equilibrium tendencies may be identified. A
central tendency is one in which the candidates tend to con-
verge on the median voter's ideal point location. A dispersed
tendency is one in which the candidates tend to distribute
themselves along the dimension, adopting distinctive policy
positions. The former is dominated by centripetal forces in
which electoral competition drives candidates toward one an-
other, while the latter is dominated by centrifugal forces that
drive candidates away from one another in order to differenti-
ate themselves.
Cox shows how the number of competing candidates is a
key parameter in determining whether centripetal or centrifu-
gal incentives dominate. Cox cuts the cases according to
whether the system permits cumulation of votes or not (see
the column labeled c in Display 7.3). He then shows that when
cumulation is not allowed (c =no), "if the number of candidates
competing for election is small enough relative to the number
of votes per voter [v in Display 7.3], then centripetal forces will
dominate (in the sense that equilibria will be centrist); but if
the number of candidates is large enough, then centrifugal
forces become strong enough to create a certain amount of dis-
persal in equilibrium. When cumulation is allowed (c = yes),
then "centrifugal forces will always dominate." 16
16
"Centripetal and Centrifugal Incentives of Electoral Systems," 912. Cox
makes "small enough" and "large enough" precise.
In principle, the reader could take a copy of Cox's paper (in

which more carefully and precisely stated propositions may be
found) and a governmental handbook describing a specific
electoral system and, on the basis of these two sources of in-
. formation, make forecasts about how candidates will actually
distribute themselves politically in election contests. Cox illus-
trates this for the case of Japan, a country whose electoral
system (at the time Cox wrote) was described by SNTV:
Most of the 511 members of the Japanese House of Representa-

tives are elected in districts of magnitude 3, 4, or 5 in which each
voter has a single nontransferable vote.... In all of Japan's multi-
member districts [the number of candidates is large relative to
the number of votes per voter]. This means that there is never a
median voter or central clustering result predicted for Japan. In-
stead, the dispersion result applies and predicts that, if there is
an equilibrium, then candidates will not be bunched together
anywhere along the left-right spectrum; and some candidates will
adopt [extreme positions]. 17
Social scientists are not rocket scientists- that's already

been conceded. But we are growing increasingly sophisticated
about how various procedures of social choice actually work.
In this section][ have done no more than to illustrate a small
fraction of the rich class of pluralitylike electoral systems and
how they may be analyzed. Electoral systems can be boiled
down to a relatively small number of parameters and equilib-
rium analysis conducted on them. Depending upon the re-
sources given a voter (particularly, the number of votes), how
those resources may be deployed (particularly, whether they
can be cumulated and whether they may be only partially
used), and the nature of the task (particularly, the number of
legislators to be elected), it is possible to determine whether
candidates have incentives to cluster centrally or disperse
17
"Centripetal and Centrifugal Incentives of Electoral Systems," 915.
themselves. This, in turn, tells us something about the kinds

of legislatures these electoral systems produce.
We should remember, however, that not all electoral sys-
tems are horse races among individual candidates . A large
class of systems operates at a more highly aggregated level-
one in which parties, rather than individual candidates, are
the strategic players. Legislative representation in these sys-
tems is determined by the proportion of the popular vote each
party receives. We shall look at these systems before compar-
ing them to the horse-race variety. Even before doing that,
however, we will look at one last interesting pluralitylike sys-
tem that would seem to defy Cox's classification scheme.
A Most Unusual Plurality System:

The Single Transferable Voter (STV)
The single-transferable-vote system, sometimes called the
Hare system after one of its early students, differs from other
multimember plurality systems in that each voter essentially
reports his or her entire preference ordering over the candi-
dates. Riker describes it as follows:
The rule for the single transferable vote method is: For districts
with S seats and m candidates (m ;;,; S), the voters, V in number,
mark ballots for first choice, second choice, ... , and mth choice. A
quota, q, is calculated thus:
q = V/(S + 1) + 1
and q is rounded down to the largest integer contained in it. If a

candiqate receives at least q first-place votes, he or she wins, and
any surplus votes (i.e., the number of first-place votes in excess of
q) are transferred to nonwinning candidates in proportion to the
appearance of these candidates in next place on all ballots of the
initial winner. Another candidate who then h as q first-place
[plus] reassigned votes wins, and his or her surplus is transferred
to the next nonwinning candidate on his or her supporters' ballots
(again in proportion to their appearance in next place) and so on
until all seats are filled. If at any point in the process (including
the beginning) no candidate has q first-place and reassigned
votes, the candidate with the fewest first-place and reass igned
votes is eliminated and all the ballots for her or him are trans-
ferred to candidates in the second (or next) place on those ballots;
and this is repeated until some candidate has q votes. 18
To make all this concrete, suppose there were 100 voters
(V - 100) in a district charged with electing 3 representatives
(S = 3). The quota- known as the Droop quota (Mr. Droop was
a friend of Mr. Hare)-is q = 100/(3 + 1) + 1 "' 26. That is, any
candidate receiving 26 votes can be assured that no more than
two other candidates can get as many as she. 19 If, in fact, a
candidate got in excess of 26 first-preference votes, and all the
remaining candidates did not, then the preference orderings of
the excess voters are consulted for the second preference llsted
and those votes are distributed to them. If this pushes some
other candidate over the 26-vote quota, then he or she is
deemed elected. 'I'his continues until all three candidates have
been elected or until fewer than three have been elected and
no remaining candidate has the quota. In this case, the
process starts eliminating candidates, starting with the one
with the fewest total votes. All of that candidate's votes are
distributed to t he candidates named second on each ballot.
This continues until all seats are filled.
The STV method is used to elect the parliament in Ireland
(called the Dail) and the city council of Cambridge, Massachu-
18
Riker, Liberalism Against Populism, p. 49. Duncan Black refers to STV as a
system of proportional representation because it tends to approximate the
representativeness that many PR systems display. See his famous Theory of
Committees and Electioni; (Cambridge, U.K.: Cambridge University Press,
1958). But it is decidedly a plurality system in that election is dependent on
getting more (first-preference plus transfer) votes than other candidates.
19
lf three other candidates got at least 26 votes, then they would jointly have
78 votes, which would mean that, together with the candidate who already
has 26 votes, 104 votes had been cast. But this cannot be, since only
100 votes can be cast. Therefore, it also cannot be that more than three can·
didates receive 26 or more votes each.
setts. In Ireland there are 41 districts from which 166 mem-

bers of parliament are elected; typical district magnitudes are
3, 4 , and 5. The City of Cambridge elects a city council (nine
members) from one at-large district. In both locations, there is
much local lore about strategic behavior, as candida tes cam-
paign not only among their own supporters but also among op-
ponents' s upporters. The purpose for a candidate is both to
energize his own supporters and to try to get listed high up in
the preference or derings of voters supporting other candi-
dates, the latter in order to benefit from potential "r eassigned"
votes. Cambridge has refused to compu terize its operations, so
it takes nearly two weeks following an election to allow tens of
thousands of paper ballots to be counted and recounted; many
a local political junkie socializes at election headquarters dur-
ing this period, cheering as one candidate or another sur-
passes the Droop quota and claims a council seat. 20
CASE 7.1
O SCAR NOMINATIONS
The Academy of Motion Picture Arts and Sciences organizes

the nomination process for Academy Awards. Each section
of the academy is eligible to participate in the nomination
processes for its particular award. According to the New
York Times reporter Tom O'Neill, here's how it works (with
my occasional kibitzing in brackets):
In the best-director category, for example, there are about 300
Academy of Motion Picture Arts and Sciences members who de-
20 For those interested in learning more about this system, they should con-
sult Gideon Doron, "The Hare Voting System Is Inconsistent," Political
Studies 27 (1979): 283-86; and Gideon Doron and Richard Kronick, "Single
Transferable Vote: An Example of a Perverse Social Choice Function,"
American Journal of Political Science 21 (1977): 303-11.
cide the nominees by listing their favorites from 1 to 5. When

the ballots are received by PricewaterhouseCoopers, they are
put into piles based upon who is listed as the No. 1 choice, said
Greg Garrison, one of the eight accountants who spend a week
determing the nomination lineup. "We divide the number of bal-
lots [V s 300] by six [S + 1] to determine how many No. 1 votes
are needed to establish a nominee," he said. So a director with
50 or more No. 1 votes is a utomatically a nominee. ''Let's say
only two directors have that many," he continued. "We take all
of their ballots and set them aside. Then we conduct a second
pass-through. We start with the smallest stacks of ballots, dis-
card the No. 1 choice and redistribute them according to who's
listed in second place. We work through all the stacks that way,
from smallest to largest, redistributing ballots until we have
five stacks with more than 50 of the same names in each one."*
As a reader can see, this is very close to STV as described in
the last few pages. The only apparent difference is that "ex-
cess" votes of those who qualify as nominees aren't redis-
tributed.
*Tum O'Neill, "Oscar Watch: Winner Takes All," New York Times, January
9, 2005, Arts & Leisure section.
We conclude by noting that a version of this method is used

to elect members of the parliament in Australia. In that coun-
try, however, districts have a magnitude of only one, so the
method is given a different name just to confuse everyone (nat-
urally); it is known there as the alternate uote. With S "" 1, the
quota formula above becomes q - V/2 + 1. With 100 voters, a
candidate needs 51 first-preference plus reassigned votes to be
declared the winner. 21
21 STV is also used to elect members of the Faculty Council of Harvard Uni-
versity. (It is something of a historical irony that it was recommended to the
university by Kenneth Art"OW while he was a member of the faculty!) It runs
smoothly, virtually without controversy, and certainly with no strategic be-
havior, because so few faculty members offer themselves as candidates.
Proportional Representation
I will be analytically less precise about PR systems because,
quite frankly, not much work has been done on them. The pur-
pose of a proportional system is to produce a legislature that
mirrors, in some fashion, the larger society. If, for example,
the main cleavages in society are ethnic (as in many of the
democracies in Africa and the Middle East), or religious
(Northern Ireland), or linguistic (Belgium), then a PR system
will tend to reproduce them inside the elected legislature.
Under most such systetns, no stratum, unless especially small
or poorly organized (or just plain stupid) is highly underrepre-
sented.
One would think that the design of a PR system is straight-
forward. Let each citizen cast a single vote for his or her fa-
vorite party (or any other, for that matter). Add up the votes
for each party. Give each party a proportion of legislative seats
exactly equal to its proporLion of the total popular vote. Voila!
Not so fast! First of all, the legislature's size is typically
fixed in advance. For most "reasonably" sized legislatures, it is
typically not possible to translate electoral proportions evenly
into seat proportions. Suppose the Beer Lovers Party (an ac-
tual party in Poland) captured 1 percent of the popular vote.
How many seats should it receive in the 450-seat Sejm (the
Polish House of Representatives)? It cannot be given the 4.5
seats to which it is entitled according to its electoral propor-
tion. Most PR schemes-and there are actually quite a large
number of them-differ primarily on how they handle the
problem of allocating these "fractional seats."
A second issue involves exactly who should get elected to
the legislature in the first place. If the Polish Beer Lovers
Party receives one percent of the popular vote, as in the previ-
ous paragraph, should it get any representation at all? If the
answer is in the affirmative. then at what point does the an-
swer change to negative-0.5 percent? 0.25 percent? 0.10 per-
cent? Where do you draw the line? PR systems differ quite

dramatically on this matter. In practice most require that a
party exceed a specific minimum popular vote proportion be-
fore it is entitled to any parliamentary representation. This
parameter, known as the threshold, varies dramatically from
country to country. Poland (in its first few democratic elec-
tions) and Israel (throughout its democratic history) have had
very low thresholds. For the 450-seat Sejm in the 1990s, a Pol-
ish party needed to obtain 1/450th, or 0.22 percent, of the pop-
ular vote to be awarded a seat. For the 120-seat Knesset, an
Israeli party (until recently) needed 1/12Qth, or 0.83 percent,
to win a seat. Germany, on the other hand, has a very high
threshold; a party must obtain 5 percent of the vote before it
qualifies for seats in the Bundestag. Thus, in the 1987 elec-
tions, the Green P arty won 8. 7 percent of the popular vote and
was awarded 42 seats in the 497-seat B undestag (8.45 percent
of the seats); in the 1991 election their popular vote percent-
age fell just a hair below 5 percent and they lost their entire
legislative representation. Clearly, high thresholds make for
more disproportionality. 22
Representation versus Governance: PR v. Plurality

There are few results to report on equilibrium properties of
various PR systems, so I turn in this concluding section to a
brief comparison of the two broad families of electoral systems
I have been discussing. In all our discussion I have been vague
22 For a full discussion of this and related issues, the reader may wish to con·
suit Rein Taagepera and Matthew Soberg Shugart, &ats and 'W>tes: The Ef-
fects and Determinants of Electoral Systems (New Haven: Yale University
Press, 1989). For .an interesting discussion on PR, consult the debate be-
tween Cox and Ga llagher in the pages of the leading journal on electoral
systems: Michael Gallagher, "Proportionality, Disproportionality, and Elec-
toral Systems," Electoral Studies 10 (1991): 33-51; and Gary W. Cox, "Com-
ment on Gallagher's ' Proportionality, Disproportionality, and Electoral
Systems,"' Electoral Studies 10 (1991): 348--52. For recent information on
PR systems, see J osep M. Colomer, ed. The Handbook of Electoral System
Choice (London: Palgrave-Macmillan, 2004).
about the actual number of candidates who compete for leg-

islative seats. Indeed, I have not ventured a conjecture about
whether many or few candidates compete and, more signifi-
cantly, whether or not the electoral system has anything to do
with this.
ln fact, there is a long literature on this very subject, the
most famous proposition of which is known as Duverger s Law,
named after the famous French political scientist who dared
to call it an empir ical regularity. Duverger 's Law comes in two
parts. The first, for which there is both strong argument and
evidence, states that first-past-the-post, single-member-dis-
trict systems are strongly associated with two-party (or two-
candidate) competition. The idea here is that third parties and
third candidates (or both) will ordinarily be loath to enter the
race because they have so little chance of winning; in turn,
they have so little chance of winning because neither voters,
nor campaign consultants, nor campaign contributors are
likely to waste th ei r votes, time, and money, respectively, on
hopeless candidacies. 23 The second part of Duverger's Law, for
which there is ample e mpirical support but less compelling
analytical argument, states that PR systems are associated
with multiparty competition.
As an analytical claim, it seems to me that the kernel of
truth here is that districts in which there are, by design, a
very small number of winners-only one in first-past-the-
post; exactly kin k-past-the-post-discourage independent po-
litical entry and encourage cooperation, coordination, coali-
tion, and merger-like political activity before elections. In
districts where there are many possible winners, as in most
PR systems (especially those with a very low threshold) and
even in those k-past-the-post systems where k is quite large,
independent political entry is encouraged and various forms of
cooperation, coordination, coalition , and mergerlike political
23 See Case 6.5.

activity are either discouraged altogether or deferred until

after elections. First-past-the-post systems typically, and
other "small" k-past-the-post systems often, resolve many con-
flicts before legislative politics commences. PR and "large" k-
past-the-post systems, on the other hand, defer this kind of
conflict resolution until the legislature convenes. Thus, parJia-
mentary political conflict tends to be more muted and centrist
in legislatures elected by FPP; indeed, there is typically a sin-
gle majority party that can get on with the business of imple-
menting its agenda. Legislatures elected by PR refl,ect rather
than resolve political conflict in advance, depending upon
post-election parliamentary politics-coalition government,
for example-to discover the means for resolution.
It should not be surprising, then, that a number of demo-
cratic regimes seek to obtain the best of both worlds by imple-
menting a "mixed" method. Accordingly, a certain proportion
of legislators is elected from districts according to plurality
rule, with the remaining legislators elected by PR based on
the national vote proportions received by parties. In 1993,
both Italy and New Zealand underwent changes in electoral
law. Like ships not quite passing in the night, New Zealand
deserted plurality for an approximation of the mixed method,
while Italy deserted PR for the mixed method. 24 The mixed
method appears: to enhance governance by keeping the num-
ber of parties relatively small, on the one hand, while main-
taining a modicum of representativeness, on the other.
EXPERIMENTAL CORNER
\tbting Rules and Jury Verdicts
Choices over voting methods don't affect only electoral poli-
tics. Indeed, one of the most consequential decisions on a
24 In 2006 Ila ly returned to full PR.
voting method that any society can make concerns voting

rules for juries, which are responsible for determining guilt
and innocence in criminal proceedings in dozens of coun-
tries. For example, for criminal trials in the United States,
generally the jury must unanimously agree to convict, oth-
erwise the defendant is set free. This very high standard of
agreement is regarded by most as an important procedural
safeguard, reducing the possibility of the conviction and
incarceration of innocent individuals accused of serious
crimes. So imagine the surprise provoked by two game
theorists, Timothy Feddersen and Wolfgang Pesendorfer
(hereafter, FP), when they claimed in a 1998 article that
unanimity rule may actually lead to more false convictions
of innocent defendants than less stringent alternatives like
simple majority rule.a
How could this be? First, some background. The prevail-
ing formal models of jury decision making before FP in-
volved each juror independently assessing the evidence and
reaching a conclusion. But due to psychological, cognitive,
or some other differences across jurors, it is possible that
some might conclude that the defendant was innocent,
and others that the defendant was guilty. Assuming that ju-
rors vote sincerely (see Chapter 6), based on their best
guess given the evidence, unanimity rule would then be far
less likely than a less demanding procedure to produce a
false conviction, because it requires that each and every
individual-no matter their differences-be truly per-
suaded of the defendant's guilt.
FP turned this logic on its head, however, asking
whether it might be optimal for jurors to vote sophisticat-
edly. Consider the following scenario: You are a single juror
• Timothy Feddersen and Wolfgang Pesendorfer, ''The Inferiority of Unani·

mous Jury Verdicts under Strategic Voting," American Political Science
Review 92 (1998): 23-35.
on a twelve-person jury operating under unanimity rule,

which is about to take an anonymous ballot. Your best
guess is that the defendant is innocent, but you are com-
pletely unsure what the other jurors think. Two arrange-
ments are worth considering. First, in the unanimity-rule
situation it is possible, if unlikely, that all eleven other ju-
rors are voting to convict and so if you vote sincerely, you
will be freeing an individual who in all likelihood is guilty
(or at least all of your fellow jurors seem to think so!). How-
ever, if fewer than eleven jurors are convinced of the defen-
dant's guilt, then misrepresenting your instincts and voting
guilty will have no impact on the outcome (since at least
one other individual also votes innocent). That is, voting to
convict is, in the language of game theory, a dominant strat-
egy. In this sense, there may be subtle pressures on jury
members to vote for conviction-even when they believe the
defendant is innocent-to avoid freeing a guilty party.
When these pressures exist across all twelve jury members,
the likelihood of false conviction rises markedly.
In a majority-rule situation, however, these pressures
would be far weaker, as long as the juror's perceptions of
guilt and innocence are fairly good on average, because any
one vote is less likely to be determinative of the final out-
come. Put a nother way, the harm from a single member in-
correctly pegging a defendant as innocent is much less in
the majority rule case than under unanimity rule.
This discussion suggests two fundamental questions
about jury behavior- Do jurors vote sophisticatedly? Is
unanimity rule associated with more improper convictions
of innocent defendants?-which Serena Guarnaschelli,
Richard McKelvey, a nd Thomas Palfrey put to the test in a
series of experiments.b Here's the setup: Each member of
b Serena Guarnascbelli, Richard McKolvey, and Thomas Palfrey, "An Ex-
perimental Study of Jury Decision Rules," American Political Science Re-
view 94 (2000): 407- 23.
their experimental juries was asked (while blindfolded) to

choose a ball from a jar, which either has predo1ninantly
red balls or predominantly blue balls. They were told to
keep the color of their ball a secret from the other jurors
and then to cast a vote about whether the jar they drew
from was either one of the "mostly red'' jars or one of the
"mostly blue" jars. If the group voted correctly, then every-
one shared a large prize. This experiment was run with two
voting rules, however. For unanirnity rule, the jurors were
told that correctly identifying the red jar required that all
members vote for red, and if a single member voted for
blue, then the group's vote would be blue. The majority rule
was undertaken as usual, however, with ties going to an
overall vote for blue. Thus, the experiment was meant to
recreate the circumstances of FP's original article, with a
series of individuals each possessing a private signal about
the true state of the world and using group decision making
under two different rules to cast their collective ballot. The
red jar and red balls stand in for "guilty" here, while "blue"
stands for innocent; the experimental subjects were given
no inkling that they were contributing to research that cuts
to the core of our judicial system.
The first major result of the experiment is that jurors
most assuredly do vote sophisticatedly under unanimity
rule. In a small jury of three individuals operating under
majority rule, only about 6 percent of individuals voted con-
trary to their signal if they drew a blue ball, while in the
case of unanimity rule, 36 percent did so. (It is worth noting
in passing that relatively few-but nonetheless between
3 percent and 10 percent-voted against their signal if they
drew a red ball.) Why this sharp divergence? Recall the
logic of FP's argument. Voting blue is a precarious act
under unanimity rule. If the true state of the world is a
"mostly red" jar, then you've lost your chances at winning
the prize. A logical question is whether the high rate of so-
voting Methods and Electoral Systems 221
phisticated voting observed in the experiment under una-

nimity might be tamped down if the jury were larger. In
fact, exactly the opposite occurred. More than 47 percent of
those drawing a blue ball voted for the "mostly red" jar, a
result that is in fact consistent with the theory articulated
by FP. Why? More voters, in a sense, gives you more cover
to vote contrary to your signal. If everyone else has drawn a
red ball, than you're most likely wrong, even though you
drew a blue ball, and your incorrect vote would be decisive.
If the true state of the world is a "mostly blue" jar, then in
all likelihood several others will have drawn blue and will
vote accordingly, so no harm is caused by voting red ... un-
less, of course, others think similarly.
While the premise of FP's approach to jury voting (ju-
rors are sophisticated voters) was confirmed, their conclu-
sion (unanimity rule leads to more false convictions) was
not. For six-person juries drawing from the "mostly blue"
jar, almost 30 percent falsely voted for a red jar under ma-
jority rule, while only 3 percent voted for a red jar using
unanimity rule-exactly the opposite of what FP's logic
would lead you to expect. (The rates of false conviction were
roughly the same, at about 18 percent, for three-person ju-
ries.) Why did this happen? The authors hypothesize that
certain voting rules are not very resistant to inaccurate or
erroneous voting, especially unanimity rule which requires
that all individuals vote one way. Recall from above that
3 percent to 10 percent of voters voted against their signal
if they drew a red ball, and in fact that 10 percent was for
six-person juries using unanimity rule. The effect of just
every tenth individual receiving a red ball voting for blue-
a choice that is difficult to defend on rational grounds- was
more than enough to counteract the sophisticated voting
that was prediicated to lead to false convictions. In other
words, unanimity rule led to fewer false "convictions,"
mostly because erroneously cast ballots by red ball holders
prevented juries otherwise dead set on conviction from se-

curing that end.
So can we safely draw from this experiment the conclu-
sion that unanimity rule is after all an effective bulwark
against falsely convicting the innocent? Probably not. It's
hard to believe that voters on a real jury, knowing that a
human being's fate lies in the balance, would vote as casu-
ally as in a laboratory experiment involving balls in a jar.
Moreover, voter error hardly seems like a sound basis on
which to rest as important an issue as the credibility of jury
verdicts. At the same time, it remains an open question
whether or not jury members in actual criminal trials-out-
side the arid confines of the lab- truly vote sophisticatedly.
These experiments certainly suggest that this is a possibil-
ity, however, and this is clearly a topic that merits much
further research.
1. A small society of nine people holds the following prefer-

ences: w lzxy (3 people); xzy I w (4 people) and y lzwx (2 peo-
ple), where all outcomes to the left of the vertical line are
"approval-worthy." Which outcome will win if the society em-
ploys simple plurality voting? What about if it employs plural-
ity runoff, approval, or Borda count voting? Would the society
select a clear winner if it used the Condorcet procedure?
2. The election of 1844 featured two major-party candidates

(the Democrats nominated James K. Polk and the Whigs,
Henry Clay), with a final electoral vote count of 170 for Polk
and 105 for Clay. Polk was a strong supporter of the entry of
new slave states into the union, while Clay (himself a slave
owner) opposed the addition of new slave states. James Birney
entered the race as a committed abolitionist for the Liberty

Party and managed to secur e 2.3 percent of the popular vote
overall. Her e are the results for the State of New York (with
36 electoral votes) in the election of 1844: Polk, 48.8 percent of
the popular vote; Clay, 47.85 percent; and Birney, 3.25 per-
cent. In the following analysis, assume that any Birney voters
strictly preferred Clay to Polk.a
a. Suppose that New York's electoral votes were allocated ac-

cording to sequential runoff. Who would then have won the
election? Explain why.
b. Suppose that New York's electoral votes were allocated ac-
cording to approval voting. Suggest a scenario in which Clay
wins the U.S. presidential election. How plausible do you
think your scenario is?
c. Social choice theorists have suggested that "independence of
entry of clones" (IEC) might be a useful criterion for judging
voting rules. IEC states that the addition of a candidate iden-
tical (or very similar) to one of the current candidates should
not cause the mnner of the election to change. Does plurality
rule satisfy IEC? What about approval voting? Do you think
this is a reasonable criterion for a voting rule to satisfy?
3a. Suppose that candidates a and b are in a lopsided race, so

58 percent of the population feels that b > a, and 42 percent
that a> b. Candidate a, who is to t he left of b, contemplates
paying the conservative spoiler candidate c to enter the race
to siphon off the approximately 17 percent of the electorate
with the preference ranking c > b > a. Would this be a sound
investment under either plurality rule or sequential runoff?
Assume sincere voting.
•This example and many others are contained in William Poundstone, Cam ·
ing the W>te: Why Elections Aren't Fair (and What We Can Do About It) (New
York: Farrar, Straus and Giroux, 2008).
b. Now consider a different three-way race where opinion polls

report the following: 40 percent feel a> b > c; 31 percent pre-
fer c > b >a; and 29 percent b >a> c. Who will be the winner
if an election were held today under SR? Suppose that by cam-
paigning hard, a is able to steal 3 percent of e's voters. Who
would win the election now? Which of the properties e mployed
in May's Theorem has been violated in this example?
c. Based on your answers to the first two parts of this ques-
tion, does sequential runoff have desirable normative proper-
ties? What do you think about sequential runoff as a voting
system on practical grounds?
*4. This question asks you to again consider equilibria in plu-

rality systems, and specifically Cox's (1990) claim that if the
number of candidates (let's call it m) is less than two times
the number of votes per voter, then a centrist tendency is pre-
dominant. Assume that preferences are single-peaked and
voters are honest. What is a stable equilibrium for a first-
past-the-post system when m = 2? What voting model does
this result reiterate? Is that same value an equilibrium when
m = 3 or 4?
Now suppose the same setup except now each voter has two
votes ( v • 2) which are not cumulable (c = no) (this is one form
of a Limited Vote system). If m ... 3, what is a stable equilib-
rium? Is that same value an equilibrium when m - 4 or 5?
5. Here are the first- and second-ranked preferences for sev-

eral groups of voters in a thirteen-member voting body that
uses STV with the Droop quota: b > d (4 members); b > e
(3 members); c > e (2); e > c (2); d > b (l); and a > d (1). The
group wishes to select three outcomes from a, b, c, d, e. Which
three do they select?
6. Duverger's first hypothesis is that FPP, single-member

district systems should be conducive to two-party or two-
candidate competition. Do you suppose that this hypothesis is

more accurate at the national level or at the level of electoral
districts? Does the observed outcome occur because FPP de-
ters candidate entry or because it changes voters' behavior? In
answering this second question, consider the importance of
campaign contributions, organization, and endorsements, as
well as voters, in producing a winning electoral campaign.
7. The world's democracies may be broadly divided into those

that choose their legisla tures by proportional representation
(like much of Europe) and those that elect legislators from
local districts according to plurality rule (like the British
House of Commons or the U.S. House of Representatives).
What differences are likely to be observed between FPP,
single- member district systems a nd proportional representa-
tion systems? You might consider the kinds of politicians each
approach attracts, the kinds of party systems each approach
encourages, the kinds of legislatures associated with each ap-
proach, or the kinds of public policies legislatures of each type
are more likely to produce. In light of the answers you have
given, which system do you feel better embodies democratic
ideals?
8. Presidents, legislators, judges, and voters often justify their

political choices with an appeal to "the public interest." Mark
Twain cast a skeptical eye on this concept, however, writing
"no public interest is anything other or nobler than a massed
accumulation of private interests." Having now read Part II
on the peculiarities associated with transforming rational in -
dividual preferences into a group choice, what do you think of
the notion of "the public interest"? What effects do preference
cycles, chaos, all sorts of strategic behavior and manipul ation,
and a lternative methods for aggregating individual prefer -
ences have on the idea of a "public interest"?
Summary of Part II
Nominally, in the previous five chapters I have written about

group choice- ways to think about it, and ways to analyze it.
Having said that, the second thing to be said is that I have
barely scratched the surface. There are large and fascinating
literatures on every facet of group choice, and there is much I
have not touched upon here. I have focused on group choice as
achieved through voting, and within that sphere focused pri-
marily on majority rule. Little ink has been devoted to super-
majorities (e.g., two-thirds voting), simple yes-no voting (in
which a single alternative is posed and the group votes
thumbs up or thumbs down), or bicameralism (in which ma-
jorities in two bodies must come to some common agreement).
Then, of course, there is participatory democracy (New En-
gland town meetings), and recent suggestions of an electronic
town hall democracy (in which each of us sits at our computer
console voting on the great issues of the day). Any or all of
these topics would make for great group discussions, short es-
says, or longer term papers.
Arrow's Theorem, with which we began our discussion, ap-
plies to many of these forms of group choice (indeed, just how
it applies would make an interesting research topic). And in
fact, the material in Part II has conveyed an important gen-
eral message that is applicable to the broad range of group
choice: Group decision making may depend upon individual
preferences and may reflect individual preferences, but it de-
pends upon and reflects much more besides. First, as I have
226
Summary of Part II 227
mentioned throughout these chapters, individual preferences

do not announce themselves. They are not transparent or self-
evident. Rather, they depend upon the disposition of each indi-
vidual to reveal preferences sincerely or strategically. Second,
even if the disposition to report preferences honestly or not
were of no consequence, the fact remains that there are many
procedures by which to reveal preferences and combine them
into social outcomes-procedures that produce profoundly dif-
ferent social outcomes.
To these considerations I must add one more: Collectivities
are unlike individuals in the sense that their "preferences"
rarely add up in a coherent fashion. For nearly any method of
group decision making that we would find minimally accept-
able on grounds of fairness, the group outcome often violates
the central notion of coherence (transitivity). In important
ways, the actual outcome of group choice is arbitrary. So much
depends upon t;he frictions of institutional minutiae-the
order of voting, who gets to make motions, and who gets to de-
cide when enough motions have been made.
All of this, in turn, causes a certain a mount of philosophi-
cal distress. As much as we would like to anthropomorphize
the group by endowing it with a will, an interest, a preference,
group choices can hardly be conceptualized in this fashion.
They simply lack the coherence of individual choices, and de-
pend much on idiosyncracy and ephemera.
We might even become dubious about the idea of a pub-
lic interest. A public has no identifiable interest if its prefer-
ences are either incoherent or overly idiosyncratic. An elec-
torate's decision, a legislature's policy, a (multimember) court's
opinion-all these things are decide_dJy outcomes. But the ma-
teria ls of the previous five chapters might give us pause before
imbuing these decisions with anything more normatively con-
sequential. Like Rousseau's general will, the public interest is
a normative ideal that cannot be given concreteness in most
real settings. In real settings, there are usually too many ma-
jorities, too many methods of deciding, too much strategizing,

too much incoherence. The models of Part II, then, provide the
means for analyzing politics and, at the same time, short-
circuit the tendency to judge politics. At the very least, they
force us to come to terms with our weakness for anthropomor-
phizing groups and institutions. We must understand, when
we judge a political outcome, that it is often the result of split-
second coordination by some temporary majority that exhib-
ited coherence for a nanosecond before "morphing" into some
new political entity- hardly a firm foundation on which to
build a philosophy of public interest.
This is actually something quite constructive. In raising
questions about the general will and in casting doubt about
such a thing as the public interest, the materials of Part II
have laid a revisionist foundation. They have caused us to sus-
pend momentarily, at the very least, ways of thinking about
politics that many of us had probably taken for granted.
Groups cannot be legitimately personified-they may make
decisions, but those decisions hardly constitute mandates or
other reflections of a collective consciousness. The philosophi-
cal vacuum left by Arrow's Theorem and these other materials
only suggests that there is plenty of work left to do.
The materials of Part II are constructive in another, per-
haps more important, sense. They provide tools of analysis for
understanding, explaining, and perhaps predicting politics.
The theorems and formulations I have presented are tools of
discovery and inquiry. They will help us to think systemati-
cally about political institutions, the task of Part IV. But first,
I turn to "politics writ large": the politics of collective action
and mass political behavior.
Part III
COOPERATION,
COLLECTIVE ACTION,
AND
PUBLIC GOODS
8
Cooperation
In this and the next two chapters I focus quite intently on

what people do as members of groups. Clearly, they vote, at
least some of tbe time. But I want to extend the conversation
beyond this procedural feature of group life and look at what
people actually do in a substantive sense. This chapter focuses
on the activity of cooperating, a microlevel phenomenon in
which individuals have to decide whether to be naughty or
nice to their friends, colleagues, roommates, spouses, cowork-
ers, coconspirators, allies, partners, or fellow club members.
In Chapter 9 the problem is examined in a somewhat larger
context: If cooperation is a form of group activity "in the
small," then collective action is its analog "in the large." Fi-
nally, Chapter 10 each of these considerations is linked to the
social production of what economists call public goods. In as-
sembling these chapters, I have sought to find a middle
ground between the extremes of the "isolated individualism"
of psychology and the "groupthink" of sociology. Along with
economics, the study of politics is the study of individual ra-
tionality and social interdependence.
231
A WORLD WITH No COOPERATION:

W lIAT WOULD IT B E LIKE?
In nineteenth- and early twentieth-century America, much

was made of the virtue of "rugged individualism." People were
thought to be virtuous if they were self-reliant-if they devel-
oped the requisite coping skills and other forms of "human
capital" to enable them to survive and prosper in a world full
of opportunities, to be sure, but full of pitfalls and hazards as
well. This ideology assumed mythic proportions, personified in
the famous Horatio Alger stories about a young lad who suc-
ceeded in a cruel world by dint of individual effort and
cunning.
The ideology of rugged individualism, however, was quali-
fied by an abiding faith in two forms of community. First, it
was not really individuals, per se, who were to be rugged and
self-reliant, but families. Responsibility for, and cooperation
within, families (sometimes of a rather extended sort) were
values held dear in an earlier era, putting to shame the pur-
ported "family values" of contemporary political debate. Sec-
ond, the idea of neighborliness apparently sat comfortably
alongside that of self-reliance. Groups of neighbors engaged
collectively in activities ranging from helping one another at
harvest time to barn-building parties to volunteer fire depart-
ments to taking the law into their own hands (posses and vig-
ilantism).
In contrast, the modern urban landscape, with its combi-
nation of social isolation (despite physical proximity). alien-
ation, and individual surliness, is often portrayed as a world
devoid of the cooperative spirit of that earlier era. (Try flag-
ging a motorist down to help you with your disabled vehicle
during the evening rush hour in New York or Los Angeles.) It
is, however, only a pale approximation of the quintessential
world of no cooperation- the fictive "state of nature" invented
Cooperation 233
by the seventeenth-century English philosopher Thomas

Hobbes. Hobbes described the life of the individual human be-
fore the advent of civil society as one that was "solitary, poor,
nasty, brutish, and short." In this world, individuals had to
scrape and scratch for survival not only against the natural el-
ements-hunting and gathering food, providing for shelter
and clothing, and so on- but also against other humans. Indi-
viduals, that is, had to make provision not only for the haz-
ards that nature furnished but also for the predation and
thievery that other humans inflicted upon them. These were
not happy campers!
As Hobbes and many commentators after him observed,
human effort devoted to protection against assault from oth-
ers was necessary, to be sure, hut was also quite wasteful in
either of two circumstances. If humans would restrain them-
selves from preying on others (moral principles) on the one
hand, or if social mechanisms of some sort could be put into
place to provide the restraints (civil society) on the other, then
individual energy devoted to protection would become unnec-
essary, and that effort could instead be redirected toward pro-
ductive activities.
Much energy has been devoted down thl"ough the ages to
creating systems of values, both philosophical and religious,
which, if internalized, would release human resources from
otherwise wasteful protection activities. But philosophers like
Hobbes have not been optimistic about this prospect. For one
thing, systems. of values have, throughout human history,
often come into conflict with one another and have probably
killed more people than they have saved. Nearly all crusades,
holy wars, and ideologically inspired conquests have had, at
root, a philosophical or religious foundation; they have rarely
produced anything resembling civility, much less utopia. For
another thing, humans are hardwired with various wants and
needs that cannot always, because of scarcity, be simultane-
ously provided to all. Scarcity thus breeds conflict. While reli-
gious principles and moral dicta may have a partially re-

straining effect-the music that soothes the savage breast, so
to speak-the history of humankind suggests they are insuffi-
cient to the task.
Most, therefore, have placed their bets on the creation of a
civil society (in which individuals are restrained from taking
advantage of their fellow "citizens") as the way to liberate
human energy for productive uses. A command-for example,
"Thou shalt not steal or otherwise prey on your neighbors"-
backed by a capacity to detect violations and to punish viola-
tors is what we are talking about here. Hobbes named the
entity with this capacity to command, detect, and punish
Leviathan, and saw it as humankind's solution to its "problem
of order."
There is no doubt that various aspects of civil society have
come to provide order to the lives of many people, and much of
this book is devoted to studying the institutions of civil society
for precisely this reason. But before I jump into an analysis of
institutions (reserved for Part IV), let me first examine the
possibility that a false dichotomy has been posed. It was sug-
gested above that problems of protection from predation-
what we called the "problem of order"-can be solved either by
causing individuals to internalize pacific attitudes and inten-
tions, in the form of altruism, religion, or other moral princi-
ples, or by endowing Leviathan with the power to root out
predatory behavior and otherwise regulate social life for
peaceful ends. Might there not, however, be a third alterna-
tive? The remainder of this chapter will provide an affirmative
answer to this question. Cooperation may emerge and be
maintained, even though people have not internalized pacific
or otherwise touchy-feely attitudes, and even though there is
not the heavy sword of Leviathan hanging over them. This is
no conjurer's trick; it is ruthless self-interest at work.
Cooperation 235
THE SIMPLEST CASE:

T wo -PERSON COOPERATION
Individual behavior typically involves bearing some costs in

order to secur·e some benefits. The student studies hard in
school in order to secure a good job after graduation. The sub-
urban homeowner devotes weekend time and energy to her
garden in the spring in order to enjoy its beauties in the sum-
mer. The consumer exchanges some of his hard-earned cash
for a new Audi or a tube of toothpaste. In each of these situa-
tions, a rational individual weighs benefits against costs. The
former are enjoyed exclusively by her; the latter are borne ex-
clusively by her. It is an individual optimization problem for
which the kind of decision theory briefly surveyed in P art I
(and covered in any standard economics course) is relevant.
What about group situations in which a collection of indi-
'viduals is pursuing some objective? Individual group members
must bear the burdens-club dues, effort, investments of time,
perhaps-but the benefits are often not exclusively private.
(Indeed, some of the so-called private situations in the previ-
ous paragraph may have consequences beyond the individual
taking the action; the homeowner's beautiful summer garden
provides pleasure for her neighbors, for example.)
The classic illustration of a group interaction of this sort is
provided by another British philosopher, David Hume. Hume
tells the story of two farmers whose respective fields abut a
common marshland. If the marsh were drained, common ben-
efits would be generated-for instance, the destruction of a
mosquito habitat. Farmer A's individual effort in draining the
marsh, itself a burden, would produce this benefit not only for
himself, but also for Farmer B. Each farmer is certainly de-
sirous of the benefit, but is loath to pay the price-especially if
he can get the other guy to do all the heavy lifting!
In this circumstance the key is what game theorists refer
to as strategic interdependence. We can analyze the situation

systematically as follows. Suppose each of Hume's farmers
valued the drained marsh at 2 utiles. 1 If either were to take
the project on by himself, the cost to him (in terms of the
things he would have to forgo in order to take on the bother of
the job) would be 3 utiles. Thus, if there were only one farmer
available to take on the task, then it certainly would not be
worth his while. Suppose, however, that if each farmer worked
"cooperatively" with the other, then it would cost each only
1 utile. 2 In this case each farmer would enjoy 2 utiles' worth of
drained marsh at a cost of but a single utile-a pretty good
deal. Still, though, the best deal of all would be for the marsh
to be drained entirely by the other farmer. This can be seen in
Display 8.1. 3
If both choose to drain the marsh (top left cell), then each
gets 2 utiles of benefit at 1 utile of cost for a net payoff of l. If
neither chooses to drain (bottom right cell), then with nothing
ventured there is nothing gained: the payoff is 0. If one farmer
does all the work (either of the off-diagonal cells), then he gets
2 utiles of benefit for 3 utiles of cost-a net payoff of - 1-while
the nonworking farmer also gets the 2 utiles of benefit but at
no cost-a net payoff of 2. How would a ruggedly individual
(read: rational) farmer choose?
l A utile is a made-up unit of value or utility. All that matters for our purpose
is that more utiles means more value to a person. If it makes it easier, you
may think of the units as thousands of dollars, so that each farmer values
the drained marsh at $2000.
2 That is, there are increasing returns to marsh-draining effort. One person,
working alone, would end up expending 3 utiles of energy, whereas two
working together would jointly expend only 2 utiles.
a Some readers may recognize this payoff matrix. In game theory it is known
as the Prisoners' Dilemma. 1\vo petty criminals are arrested for a burglary.
Tf both keep quiet, the district attorney has to release them (payoff of 1
utile). If both squeal they get time in the slammer, but with a plea-bargained
reduction (payoff of 0 utiles). But if one squeals and the other keeps quiet,
theu the squealer is given a reward and the book is thrown at the "squealee"
(payoffs of 2 and -1, respectively). This story yields the same payoff matrix
as Display 8.1, so our analysis will be the same.
Cooperation 237
DIS PLAY 8.1

H UME'S MARSH-DRAINING GAME*
Farmer B's Choice

Drain marsh Do not dra in marsh
(Cooperate) (Do not cooperate)
Farmer A's Choice
Drain marsh l, 1 -1,2
(Cooperate)
Do not drain m arsh 2, -1 0,0
(Do not cooperate)
*The first number in each cell is the net payoff in utiles to Farmer A ; the
second number is the payoff in utiles to Farmer B.
Suppose you are Farmer A (recall that your payoffs are the
left-most number in each cell). If Farmer B chooses to drain
(so we're looking at the left-most column of Display 8.1), then
you get 1 utile if you drain and 2 utiles if you do not. If Far-
mer B chooses not to drain (right column of Display 8.1), then
you get -1 utile if you drain and 0 utiles if you do not. No mat-
ter what Farmer B does, Farmer A always gets a higher payoff
if he chooses not to drain. The reasoning is precisely the same
if you are Farmer B: No matter what Farmer A does, Farmer B
always gets a higher payoff if he chooses not to drain.
From the perspective of either farmer, there is a double
reason to choose not to drain. First, you do better by not drain-
ing no matter what your counterpart chooses to do. But sec-
ond, and perhaps more psychologically compelling (since you
never trusted your neighbor very much anyhow), precisely be-
cause the same payoff profile holds for your counterpart he is
likely not to drain, making it clearly in your interest not to do
so either. That is, the other guy's incentives reinforce your
own inclination not to drain, and vice versa, ad infinitum.
Each farmer has a "dominant" strategy to be uncooperative,
not because either is mean-spirited, but rather because nei-

ther has an incentive to cooperate and neither wants to be
taken advantage of. This is the paradox of cooperation: Nei-
ther farmer lifts a finger, the mosquitos flouri sh so each
farmer gets a 0 payoff, yet each could have gotten a payoff of
1 if only the two had cooperated. The result is that ruggedly
individualistic rational behavior has produced a state of
affairs less preferred by both farmers than an available alter-
native. This is the rational individual/irrational society conun-
drum in another form. Another example, also drawn from
Hume, is perhaps a more poignant illustration of the tragedy
arising from the failure to cooperate. Hume writes of two corn
farmers:
Your corn is ripe today: mine will be so tomorrow. 'Tis profitable
for us both that I shou'd labour with you today, and that you
shou'd aid me tomorrow. I have no kindness for you, and know
that you have as little for me. I will not. therefore, take any pains
on your account; and should I labour with you on my account, I
know I shou'd be disappointed, and that I shou'd in vain depend
upon your gratitude. Here then I leave you to labour alone: You
treat me in the same manner. The seasons change; and both of us
lose our harvest for want of mutual confidence and security. 4
This is not the end of the story, though Hume's example

should be taken seriously because it stands as a metaphor for
a host of social situations having a similar incentive profile. A
quick glimpse of this prospect is found in Case 8.1.
4 See David Hme, A Treatise of Human Nature. ed. L.A. Selby-Bigge and P.H.
Nidditch (Oxford: Clarendon Press, 1975 (1737)). I thank Dr. Mark Yellin for
bringing this to my attention.
Cooperation 239
C ASE 8.1
THE PARADOX OF COOPERATION:
N UCLEAR DISARMAMENT IN THE COLD
WAR AND CONGRESSIONAL
PORK-BARRELING
Examples of the Humean marsh-draining, corn-growing

farmers abound in politics. As we have seen, the coopera-
tion game can be characterized as "I would, if you would,
but I can't trust you, so I won't." When self-interest out-
weighs trust, the outcome is less satisfactory than it might
be for both parties. The Cold War between t he former So-
viet Union and the United States provides an excellent ex-
ample of the two-player cooperation game. Both countries
kept enormous arsenals of nuclear weapons pointed at each
others' major cities and defense sites. Both countries in-
curred substantial economic and psychological costs main-
taining these arsenals. Since neither country was able to
generate a clear advantage in its nuclear thr eat, and since
use of these weapons was suicida l, an outcome superior to
this "standoff" was for both countries to get rid of their
weapons. If they had, then neither side would suffer strate-
gic harm, yet both would save the maintenance costs they
currently incurred. As the nearby payoff matrix displays,
this desirable outcome could not be achieved. If both dis-
arm, each receives a payoff of 10; if each maintains its
weapons, the payoff is O; if one disarms and the other
doesn't, then the now superior player gets a payoff of 100
and the now inferior player gets a payoff of - 100.
SOVIET U NION
Maintain Disarm
U NITED Maintain 0,0 100, -100
STATES Disarm - 100, 100 10, 10
Given their incentives, both countries preferred to main-
tain their forces rather than disarm regardless of the other's
actions. Each side considered unilateral disarmament
equivalent to surrender. The specter of this event, indicated
by the -100 payoff above, prevented the mutually preferred
outcome of bilateral disarmament. The dynamics of the co-
operation game kept the nuclear weapons in place despite
intentions and preferences to the contrary. *
Though substantively quite different, the politics of
pork-barreling in the U.S. Congress is theoretically another
instance of the paradox of cooper ation. The term "pork-
barrel politics" r efers to the appropriation of federal funds
for inefficient projects that benefit individual congressional
districts but offer little benefit to the nation as a whole. The
incentive to engage in pork-barrel politics is the opportu-
nity it affords for legislators to claim credit at election time
for prominent, federally subsidized projects in their dis-
tricts. Pork-barrel politics often centers on agricultural sub-
sidies, defense contracts, and transportation projects. A
$180-million-a-year wool and mohair subsidy, the $31-
hillion NASA space station, the mass-transit system in
downtown Buffalo, and the "Big Dig" harbor tunnel con-
necting the city of Boston to its airport have all been ac-
cused of being "pork." Pork-barrel politics has come under
close scrutiny recently as budget pressures force politicians
to reexamine their budget expenditures.
• Interestingly, disarmament has occurred over the last two decades, but it
has not been unilateral and it hail not been all or nothing. It has pro·
ceeded, in a sense, in baby step~: "You get rid of some of your arsenal and
we'll get rid of some of ours." The result has been a reduction in nuclear
weapon stockpiles but not their elimination.
Cooperation 241
The phenomenon of pork-barrel politics can be under-

stood as arising from the paradox of cooperation. Since a
pork-barrel project benefits only the district or geographic
area that receives it, with the costs to all taxpayers in the
country far outweighing this benefit, all legislators would
be better off if there were no pork-barreling at all. But each
legislator nonetheless has a strong incentive to continue
trying to get projects for his or her own district. Thus,
everyone kno\VS that despite everyone being better off in a
pork-free world, everyone will continue to push for projects
for his or her OWh districts. That is, the "cooperative divi-
dend" of no pork-barrel projects-in which a district loses
its own project but more than makes up for this by not hav-
ing to shell out its share of cash to finance projects every-
where else-is not stable, pecause politicians continue to
have an incentive to use whatever influence they can
muster to continue targeting projects for their states or dis-
tricts.t
t For an analysis of contemporary pork barreling, known as earmarking,
see Kenneth A. Sbepsle, Robert P. Van Houweling, Samuel J . Abrams,
and Peter C. Hanson, "The Senate Electoral Cycle and Bicameral Appro·
priations Politics," American Journal of Political Science 53 (2009):
343-59.
COMPLICATING THE SIMPLE CASE:

Two- P ERSON COOPERATION WITH
REPEAT P LAY
In Hume's exa mple, the occasion for cooperation between the

two farmers involves jointly working to drain a marsh. The
opportunity for a net benefit for each exists only if both farm-
ers put their backs into it, so to speak. That is, there is a coop-
eration dividend to be had for this two-person society, if only
its members can structure r elationships appropriately to cap·
ture it. But the relationship, as portrayed in Display 8.1, is

not appropriate for capturing the dividend. Were this the en-
tirety of the relationship between Farmer A and Farmer B,
the sad fact of the matter is that the cooperation dividend
would remain uncaptured and life for each farmer would be
slightly more impoverished than it might otherwise have
been.
Hume's example, then, is best thought of as a self-
contained situation in which there is the prospect of a cooper-
ation dividend in this one circumstance. It is a one-shot deal.
Most societies, however, including the one consisting of
Hume's two farmers, are more enduring. They do not usually
materialize for that one opportunity of securing a cooperation
dividend; nor do they immediately disintegrate thereafter.
Rather, this week it's the marsh that needs draining, next
week it's the common fence between the farmers' fields that
needs patching, the week after there is the two-man job of re-
placing a roof on one farmer's barn, and the week after that
it's the other farmer's pond that needs to be sealed. In short,
societies consist of a series of repeated (or even continuous)
encounters, not one-shot plays of a game (as suggested in the
second Hume example of farmers who have a sequence of op-
portunities to help with one another's harvest).
This fact of repetition changes things dramatically, but
only if some other conditions are satisfied. To see this, imagine
that the strategic interaction portrayed in Display 8.1 is
played not once, but twice-exactly twice-and both farmers
know it. Suppose, for example, there are two marshes that
need draining. If this is the case, then each farmer will know
that the second play of the strategic interaction will be the
last-it will be a one-shot affair. So, in that last interaction
each farmer will, for all the reasons given a moment ago, play
his noncooper ative strategy. But then, backing up to the play
before that (the first play), the farmers will realize that, effec-
tively, that one is the last play, since the second play will be
Cooperation 243
determined no matter what happens in the first play. So once

again, each will rationally play his "do not cooperate" strategy.
More generally, even if a strategic interaction like Hume's
marsh-draining game is played repeatedly, if the number of
repeat plays is finite and commonly known to the members of
the society, then each encounter will be played out as though
it were a one-shot affair. Repetition in this instance is no more
than a string of one-shot games, and the cooperative dividend
in each case is lost. 5
The idea of a known finite number of repetitions, however,
see1ns almost as artificial as the one-shot example with which
we began. Societies are ongoing and continuous. Farmers A
and B may not live forever, but they don't know when their
microsociety will come to an ~nd. Thus, they don't know when
the last play for a cooperation dividend will arise. Conse-
quently, they might as well proceed as though their society is
unending. It is this form of repeat play that allows for the cap-
ture of cooperation dividends ... sometimes.
If each farmer assumes the string of opportunities for coop-
eration will be very long, then each may be willing, on the first
occasion, to take a chance. The worst that could happen is
that he will be exploited that one time, learn his lesson, and
simply refuse to cooperate on subsequent occasions. Given the
symmetry of the situation, both farmers may take a chance in
that first encounter, resulting in an outcome in the top left cell
of Display 8 .1 and a payoff of 1 utile. On the next occasion,
each farmer will remember that the previous encounter had
6 This paragraph gives the theoretical answer to the question of what will
happen when a commonly known finite number of repetitions of the game in
Display 8.1. occurs.. If that finite number is small, then the logic conveyed in
the paragraph is, in my opinion, quite compelling. If, on the other hand, the
number is very large, even if it is still finite and still is commonly known by
alJ members of society, the logic becomes less compelling. There is a strong
incentive, it seems to me, for the members of society to seek out some means
for pretending that their own rationality has been disabled, thereby allow-
ing, at least for a while, for some of the cooperation dividends to be captured.
This is a very complicated question in game theory.
elicited cooperation from the other, thereby encouraging each

to try it again. In short, a little positive reinforcement may
well set them on a "cooperation path" for quite some time. It is
the "shadow of the future"-the prospect of cooperation divi-
dends not just now but stretching out over the longer haul-
that make cooperative moves look very attractive.
In his famous work on this subject, Robert Axelrod 6 calls
the behavioral strategy of "being nice" the first time and then,
on each succeeding occasion, doing what the other guy did the
time before the tit-for-tat strategy. The first time, cooperate.
The next time, cooperate if your colleague cooperated the last
time. But don't cooperate if he didn't the last time, and don't
cooperate again until he changes his wicked ways. That is, co-
operate conditionally after the first play of the game .
. It is but the tiniest of steps from the observation of each
farmer playing his tit-for-tat strategy in the repeat play of
Hume's marsh-draining game (or, as we called it in note 3, the
Prisoners' Dilemma) to the claim that a norm of reciprocity ex-
ists in this society. The farmers have not internalized a reli-
gious principle (like the Golden Rule), although their behavior
seems to exhibit it. Nor is there a sword-wielding Leviathan
making the two cooperate. Instead, each of two ruggedly indi-
vidualistic, rational egoists has, by virtue of being embedded
in an ongoing social relationship, found it in his interest to co-
operate with his counterpart.
Before breaking out the champagne, let us hasten to note
that there is an "evil twin" to the norm of reciprocal coopera-
tion. If the relationship had gotten off to a bad start-with one
or more of the farmers not being "nice" at the outset-then tit-
for-tat would echo this misfortune. At each play each farmer
will "punish" the other for failing to cooperate the time before.
6 Robert Alex.rod, The Evolution of Cooperation (New York: Basic Books, 1984).
An earlier and highly influential analysis is Michael Taylor, Anarchy and
Cooperation {New York: Wiley, 1976).
Cooperation 245
This social interaction would look more like a blood feud or a

civil war than a love fest. 7 And surely the world is full of eth-
nic, tribal, racial, and interpersonal hostilities that look like
tit-for-tat gone mad.
The happier point of our exercise, however, is to demon-
strate that the dividends of cooperation may be captured as
a sensible and rational response by individuals to the circum-
stances in which they find themselves. Religious and philo-
sophical dogma, not to speak of external enforcers, may well
reinforce this sort of thing. But we believe that these latter al-
ternatives would have a much harder time if they could not
rely on the self-interest of the cooperators. What we have
shown is that there are circumstances in which this self.
interest exists.
ALTERNATIVE M ECHANISMS
I NDUCING COOPERATION
I. Internalized Values
As strongly as I believe that rational responses to ongoing
relationships are responsible for quite a lot of the cooper-
ative dividends most of us ,realize in everyday life, they
clearly aren't the only thing going. Let us look briefly at some
alternatives.
People do, in fact, internalize values that dispose them to
cooperate, if only to cause them to be "nice" at a first en-
counter so that reciprocal norms might develop. But this does
not account for why one set of moral or religious principles
rather than another is internalized. Moreover, we observe
1
This situation is nicely analyzed in Avino.sh Dixit and Barry Nalebuff, Think·
ing Strategically (New York: Norton, 1991), Chapter 9. For the interested
reader, this volume is perhaps the most accessible, and certainly the most
delightful, book on game theory available.
that, with unfortunate frequency as I noted earlier, these

principles often fail, as the best of Christians or Muslims or
Jews manage to slaughter one another in the name of their fa-
vorite religion. But I can comment, in a superficial sort of way,
about the mechanism by which such internalized principles
might operate.
To do so, let us return to Display 8.1 but rename it the
Prisoners' Dilemma (the reader should r eread note 3). Sup-
pose the two prisoners are Mafia soldiers who have sworn
omerta (conspiracy of silence) in dealing with anyone outside
the family. 8 The payoff matrix in the display does not seem to
capture this internalized value very well. Failing to cooperate
in this situation-namely, squealing on the other guy- is not
looked upon kindly by the family. Indeed, the squealer, if dis-
covered by the family, typically will be found lying in a dark
alley with a slit throat and a canary stuffed in his mouth. I
think the payoffs of Display 8.1 are wrong! They should be
those given in Display 8.2.
l have transformed Display 8.1 into Display 8.2 by chang-
ing one payoff in each of the off-diagonal cells. Thus, if one of
the Mafia soldiers implicates his partner, but the other does
not, then his payoff is now -~a very large and nasty nega-
tive number. A game-theoretic analysis of this situation sug-
gests that there are two possible outcomes for this game.
Either both will squeal and receive 0 utiles each, or neither
will squeal and receive 1 utile apiece. Each of these is an
"equilibrium point" in the sense that in each of these outcomes
neither player has an incentive to change his strategy if he be-
lieves the other guy isn't going to change his. For example, if
mafiosos A and B are both planning to squeal (the lower right
cell), then mafioso A hardly has an incentive to clam up (not
8 Clearly, the larger society would be better served if the two prisoners were
not able to capture the dividends of cooperation. But that is not our concern
here; we want to remain neutral for the time being about whether the coop-
eration we're investigating is good or bad in some broader sense.
Cooperation 247
DISPLAY 8.2
THE MAFIOSO (NON) DILEMMA*
Mafioso B's Choice

Don't squeal Squeal
(Cooperate) (Do not cooperate)
Mafioso A's Choice
Don't squeal 1, 1 - 1 ,-~
(Cooperate)
Squeal -~. -1 0, 0
(Do not cooperate)
* The first number in each cell is the net payoff in utiles to mafioso A; the
second number is the payoff in utiles to mafioso B. The symbol ~. which
stands for "bad," is a very large number!
squeal), since this would change his payoff from 0 utiles to -1

utile; and likewise for mafioso B. Alternatively, if both are co-
operating by keeping quiet (upper left cell), then neither
would be so foolish as to squeal, changing his fortunes from 1
utile to -B utiles.
But the game theoretics, in my view, do not give sufficient
psychological weight to that big negative payoff of -~. I sus-
pect that neither of the soldiers would take a chance on being
the only one to g,queal (no matter how good the Federal Wit-
ness Protection Program is). Omerta frequently wins out, the
Mafia soldiers cooperate, and many a prospective .conviction
eludes an ambitious district attorney. Internalized values,
then, can produce cooperation, even in one-shot games. But
they do so by changing the game.
II. External Enforcement

In the previous discussion of internalized values, noncoopera-
tive choices were "punished." Something (your conscience) or
someone (the Don) transforms the payoff of the original game
for the circumstance in which you behave inappropriately.

When it is very explicitly someone else fiddling with the payoff
matrix, then it is probably more appropriate to think of this as
a case of external enforcement, the subject of this section. 9
The idea of third-party enforcement of agreements reached
by two contracting (or cooperating) individuals is the principal
mechanism upon which the entirety of neoclassical economics
rests. Tn economic contexts it is normally :issumed that con-
tracting individuals are assured that their contracts will be
enforced by a judge, court, or sheriff and, moreover, that these
agreements are enforced costlessly and in an error-free fash·
ion. Third-party enforcement, in this instance, is precisely the
kind of assurance that is required in order to consummate
many trades (that is, to capture the dividends of cooperation).
I magine a seller whose product is sufficiently complicated
that a prospective buyer could not tell, by just looking at it or
kicking the tires, whether the product were any good or
whether it would last very long. Although there are risk tak-
ers (and suckers) in any crowd, it will be difficult for this
seller to consummate a sale, even with a buyer who might oth-
erwise be interested in the product if it lives up to expecta-
tions. Yet there is a cooperation dividend to be shared between
buyer and seller if the seller can convince the buyer that the
product is as she represents it. To facilitate this, suppose the
seller announces, "I guarantee it. If the product does not meet
your satisfaction within one week, l will return your money in
its entirety. Should the product fail to perform for a full year
as represented, I will give you a prorated return of your
money." The problem is, What is this guarantee worth? The
seller may disappear (as happens with fl y-by-night opera-
9 In a sense, t.hough, there may not be a clear distinction between these two
mech!\nisms all the time. In the presence of external enforcement, individ-
uals will often take this fact on hoard and act as though they had internalized
some value. In effect, they have anticipated the external sanction that would
be forthcoming and have avoided it in advance· by displaying proper behavior.
Cooperation 249
tions) or she may claim that the guarantee does not cover
what the buyer complains about. ("In the fine print we said
only the left widget wouldn't fail, not the right widget.") But if
the incentives for the seller to renege are strong, then the
buyer will probably anticipate that the guarantee is "not
worth the paper it's written on." Thus, guarantees, by them-
selves, may not do the trick.
If, however, a guarantee were enforceable because there
existed a third party prepared to make the guarantor deliver
on her promise, then a buyer might well be prepared to make
the purchase. Both the buyer and the seller would be pleased
by the existence of this enforcement institution. Coerci ng her
to deliver on her promise makes the seller's promise credible.
It is the credibility of her promise that induces the buyer to
buy, after all. So, the institution of thirdparty enforcement al-
lows cooperation (Buyer: "I'll buy your product." Seller: "I'll
guarantee its quality."). The absence of this exogenous en-
forcement institution makes both buyer and seller worse off. 10
In the Mafioso Dilemma, for example, imagine the payoffs
are as they originally were in Display 8.1. Mafioso A says to
mafioso B, "I won't squeal." Mafioso B says to mafioso A, "Nei-
ther will I." These promises are credible because there is a
thjrd-party enforcer, Don Corleone, who imposes sanctions on
those who break their promises. The presence of Don Corleone
effectively transforms Display 8.1 into Display 8.2. What is es-
pecially interesting in this instance is that as long as mafioso
A and mafioso B believe they are playing the game in Display
8.2, the Don never has to display his might. Indeed, to those
ignorant of the Don's existence, the two prisoners might be
thought simply to be honest men who keep their promises! Al-
10 A buyer may still buy under these circumstances, but because the seller
cannot credibly commit to honor b.er guarantee, the buyer will require a
break on the price. This differenc~ between the price a buyer would pay
under a fully enfo:rceable guarantee a nd one that is not enforceable is, in
effect, the insurance premium the buyer requires.
~· -·-=-=-~-=--·~--~
'
ternatively, it may be thought that there is "honor among
thieves" (a moral principle) .
By introducing a third-party enforcer, much like Hobbes did
with his invention of Leviathan, we have effectively coerced
people to behave in a manner that yields the m a cooperation
dividend. But our analysis would be woefully incomplete if we
failed to inquire further into the nature of this enforcer. To be
precise, I need to take up three matters: costly enforcement,
- ------
let his brother-in-law off for speeding but nail the town's "radi-
cal lawyer" who has been such a nuisance to the police depart-
ment? Who would doubt that the rabbi, the third-party
enforcer in the little Polish villages that the Nobel laureate
Isaac Bashevis Singer wrote about, would make sure that his
own son got small advantages in village life? What is to pre-
vent a future black government in South Africa from stifling
dissent within the white community? In short, while many
communities-ranging from small-town America to villages in
eastern Europe to the national community in South Africa-
rely on institutions of third-party enforcement, all are vulnera-
ble to difficulties arising from inappropriate incentives. The
traffic officer above responds not only to his official responsibil-
ities but also wants to keep his wife out of his hair and his bud-
dies in the department off his case. We all look after our own.
That third-party enforcers may march to their own drum-
mers is especially troubling when it is offered, as it was by
Hobbes, as a solution to disorder in the state of nature. Third-
party enforcement is often offered as a rationale for the very
existence of the state. The state is seen as the community's
mechanism, first and foremost, for allowing its citizens to
avoid wasting their resources on their own protection and,
more generally, for permitting cooperation to occur. The state,
with its monopoly of force, empowers state officials not only
to provide the assurances that permit cooperation to take
place among citizens but also to use that force for their own
purposes. 12 But, as the saying goes, " Who will guard the
guardians?" Until one can be satisfied that the incentive prob-
Cooperation 253
P RELIMINARY CONCLUSION
At the end of this chapter I offer only the most preliminary of

conclusions, because the discussion continues in the next two.
I have set out the problem of cooperation and examined vari-
ous manifestations of cooperation in the simplest of societies-
the world of two persons. We have seen that this world is
not always a happy one, inasmuch as there may be no effec-
tive means for capturing prospective dividends of cooperation.
Internalized value systems and thjrd-party enforcers offer
some promise but are not without their dangers. Repetition of
social interactions may allow for cooperation to develop, be-
cause the prospects of ongoing, long-term relationships may
be too valuable to jeopardize by cheating at any one opportu-
nity. But repetition has its darker side, with jealousy, feuding,
and revenge as by-products. Cooperation is a complicated
business, and it is no surprise that humankind has only
slowly lifted itself out of the state of nature, and done so only
imperfectly at best.
The next chapter expands the discussion. A society of two,
as I have considered it in the current chapter, is pretty artifi-
cial and should be thought of only as a building block or mod-
ule for a larger society. It is this larger society to which we
turn next, and th e topics of n-person cooperation and collec-
tive action.
E XPERIMENTAL C ORNER
Altruism and Trust in the Prisoners' Dilemma

The social ·circumstances modeled in the Prisoners' Di-
lemma (PD) have proven extraordinarily adaptable to a
variety of political situations- everything from the decision
of political opponents to engage in negative attack ads to

international environmental agreements. However, this
adaptability to real-life situations is only part of the game's
appeal. The PD also touches on core themes in the emo-
tional life of social beings, and humans, after all, are social
animals. Perhaps most important among these are the
questions of altruism and trust that this ubiquitous game
raises. Thus, a great number of experimental social scien-
tists have attempted to assess the extent to which these
social virtues contribute to cooperative play, some with re-
sults that might surprise you.
First, some background: The Prisoners' Dilemma was
among the first games to be tested in laboratory experi-
ments by social scientists and, owing to the widespread ap-
plication of the game, was considered an important test of
the theory of equilibrium behavior. Recall that the sole
equilibrium in the usual one-shot PD is for both players to
defect, leaving everyone worse off than if both had somehow
managed to cooperate. Is this prediction supported by be-
havior in the lab setting? In a 1960 test, Minas et al.a found
that among their experimental subjects, 38 percent of play-
ers cooperated in any given play of the game, while 62 per-
cent defected, leading to mutual cooperation in only about
16 percent of plays. Results with a similar flavor-although
adjusted up or down depending on a variety of factors, in-
cluding payoffs, number of rounds, participant communica-
tion, and gender and nationality of participants- were
replicated in laboratories across the world. The obvious con -
clusion from these experiments is that rates of cooperation
in one-shot and finitely repeated PD games are consistently
higher than zero (contrary to predictions of equilibrium
• Sayer Minas, et al., "Some Descriptive Aspects of Two-Person Non-Zero-

Sum Games," Journal of Conflict Resolution 4 (1960): 193-97.
Cooperation 255
analysis) but consistently much lower than 1; the obvious

question is, why? Why would a player cooperate with an
opponent-especially a nameless, faceless opponent-in a
one-shot Prisoners' Dilemma?
Of course, there are in principle a variety of personal,
cultural, and situational factors involved, but systematic
thinking about cooperation in the ensuing years increas-
ingly focused on two distinct social values that the PD com-
bines to exquisite effect: altruism and trust. Why altruism?
Suppose in the one-shot version of the PD that a player
feels with near certainty that his partner will cooperate. If
he decides to respond in kind by cooper ating, one interpre-
tation of his behavior is that he has in some sense internal-
ized the desires of the other player and takes into account
the harm he would have caused that individual by playing
defect, even though it would leave him personally better off.
One immediate question this raises is how an altruistic
player might balance his own payoffs relatives to others'.
But a related and more cutting question concerns the
player's level of certainty that his partner will cooperate in
the first place. which raises the issue of trust. A player
might be quite altruistic and therefore absolutely unwilling
to stick his partner with the sucker's payoff. But unless
you're a saint, at some point altruism will fall victim to a
lack of trust, and even the kindest player won't be willing to
sacrifice his own well-being for a partner who might-but
probably won't-meet him halfway.
Unfortunately, because the PD combines these two is-
sues in such an interesting way, experimental tests of this
game are the wrong way to ascertain the relative impor-
tance of these factors in the interactions of humans who
know one another only imperfectly. If we observe in one
round of the Prsioners' Dilemma that one player has de-
fected on her cooperating partner, should we conclude that
that player is a heartless competitor or that she simply felt

insecure about her partner's trustworthiness? We could
never answer su ch a question in this context without peer-
ing inside her brain, and while magnetic resonance imaging
may someday make this a possibility, in the meantime ex-
perimentalists h ave innovated two new games to test sepa-
rately for the presence of these attributes: trust and
altruism. They have now become classics in the experimen-
tal literature of games.
In order to ascertain the extent of altruistic behavior,
game theorists set to work innovating a model of social in-
teraction that generates an opportunity for altruistic be-
havior with no corresponding need for trust from the
partner. The solution, an elegant little social interaction
called the Dictator Game, works as follows: Player A is
given a small budget and told that she can give as much as
she would like to Player B and the rest she can keep for
herself. Player B does nothing but get handed an envelope
with the amount A deigned to share (A is the dictator, after
all!). Forsythe et al.b were among the first to test this model
systematically in the lab and found that when Player A was
given $10 to allocate, on average she gave $2.33 to B. More-
over, only 21 percent of those in the Player A role gave $0;
on the other hand, 21 percent split the pot evenly, giving $5.
The rest fell somewhere between $1 and $4. A variety of
other tests of the Dictator Game with different experimen-
tal conditions, including raising the stakes considerably
and varying the level of participant anonymity vis-a-vis the
experimenter and the other player, have confirmed that, on
average, dictators give between 15 percent and 30 percent
of the pot to their partner, but a hard core of selfish types
b Robert Forsythe ct al., "Fairness in Simple Bargaining Experiments,"

Games and Economic Behavior 6 (1994): 3-17- 69.
Cooperation 257
(giving nothing) and generous types (giving lots) exist

across societies.c
These experiments confirm that altruism is prevalent al-
though far from complete in many societies. But what form
does this altruism take? Clearly, very few are willing to sac-
rifice everything for a fellow participant. At the same time
the reasoning behind giving 20 percent versus 40 percent is
not entirely clear either. Andreoni and Millerd attempted to
untangle varieties of altruism in the following way. As in
the usual dictator experiment, one subject gets a budget to
allocate; however, he is told that wha tever money he gives
to the other participant will be multiplied by a factor p. For
Andreoni a nd Miller's experiments, p varied between one-
third and three depending on the pair. What does this do for
us? Well, a subject whose altruism is based on notions of
fairness would be likely to increase his allocation to his
partner if he knows she has a low p, a nd decrease it if she
has a high p. Alternatively, a subject who is concerned with
maxmizing collective welfare would likely keep almost
everything if p is less than 1 and give away almost every-
thing if p is greater than 1. Of course, some people are just
plain selfish no matter what. Andreoni and Miller in fact
found that for different subjects, different forms of altruism
were practiced, and that a healthy majority of subjects be-
haved consistently across rounds. About 40 percent of sub-
c He nrich et al., in a s uperb comparison of these types of games across fif-

teen different societies worldwide (some industrial, some agricultural,
and some foraging; some sedentary and some nomadic), note Lhat only
among their American college student participants was allocating 0 per-
cent the most common outcome; among the Orma, a tribe in Kenya, the
most common offer was 50 percent. See Joseph Henrich e t al., "In Search
of Homo Economicus: Behavioral Experiments in 15 Small-Scale Soci-
eties," American Economic Review 91 (2001): 73-78.
d James Andreoni and John Miller, "Giving According to CARP: An Exper-
imental Test of Consistency of Preferences for Altruism," Econometrica
70 (2002): 737- 53.
jects could be squarely placed in the "just plain selfish" cat-

egory, usually taking everything or just about everything for
themselves. Another 25 percent or so corresponded more to
the fairness model, carefully calculating an even distribu-
tion of payouts to ensure rough parity in winnings between
the dictator a nd his partner. Anot her 11 percent seemed to
be ma ximizing overall social welfare, and the final 24 per-
cen t acted idiosyncratically from round to round. Of course,
this is just a small sliver of an interesting and diverse liter-
ature, but I hope it has illustrated t hat altruism is present,
but not omnipresent, in social inter actions .
So getting back to our other question, how might a par-
ticipant in a Prisoners' Dilemma decide whether or not his
partner can be trusted to play nice? Another literature in
experimental social science has emerged to measure the ex-
tent of trust among anonymous players, using another sim-
ple game. Player A is granted a sum of money and told that
he can keep it all or invest some portion with Player B. Any
money that is given to Player B will grow by some percent-
age i Oike an interest rate). Then B can decide unilaterally
how much she will return to A. The measure of trust here is
the proportion of A's funds he gives to B. In a world of per-
fect utility maximizers, you might think that A would give
nothing to B, knowing full well that if given the chance, B
will run off with all of A's money. In a series of international
experiments, Croson and Buchan° found that an average of
67 percent of the initial allocation was invested/ while ex-
periments conducted in the United States have generally
found lower rates of trust, hovering nearer to 50 percent of
wealth invested on average.
• Rachel Croson and Nancy Buchan, "Gender and Culture: International
Experimental Evidence from Trust Games," American Economic Review
Papers and Proceedings 89 (1999): 386-91.
r Interestingly, th ey a lso found no significant differences between men and
women in level of trust, but significant differences in trustworthiness.
Women on average returned more of the investment than men.
Cooperation 259
How do these two strands of research h elp us under·

stand the Prisoner s' Dilemma? Well, it turns out that a
great many people we interact with are fairly giving, at
least when it comes to one-on-one interactions with relative
strangers. Of course, a fair number are not quite so gener·
ous, so we certainly don't expect to see cooperation exclu·
sively in experimental tests of the Prisoners' Dilemma. This
manifestation of altruism explains half of our puzzle-why
we might see individuals cooperate, even though they could
make more by defecting on their gullible partners. But how
do folks r each the point where they believe their partner
will cooperate in the first place? Well, for whatever reason,
it appears that trust is quite widespread, even trust in
anonymous strangers in labora tory settings. Thus, a good
number of individuals enter these experiments fully believ-
ing that their partner will cooperate, and some, due to their
altruistic natures, even reciprocate in kind, leading to
higher rates of cooperation in the Prisoners' Dilemma than
predicted by pure theory alone (although still only a measly
16 percent!).
P ROBLEMS AND DISCUSSION Q UESTIONS
1. Explain how the insights of Hume's marsh-draining game

might be used to understand the following situations: (1) two
politicians competing for the same office must decide whether
or not to use negative "attack ads"; (2) two opposing interest
groups consider whether or not to contribute to a senator's re-
election campaign; and (3) the major industrialized countries
decide on the extent to which they will reduce greenhouse gas
em1ss10ns.
2. This chapter proposed repeat play as a plausible solution to

cooperative dilemmas. Explain the intuition behind this pro-
posed solution and then give some features of a cooperation
game (e.g., particular payoffs, number of repetitions, value of
present versus future payoffs) that you think might be con-
ducive to cooperation. Is cooperation always achieved in such
repeated settings, in theory or in practice?
*3. An alternative vision of the problem of social cooperation is

provided by the Stag Hunt game, in which two hunters can co-
operate in hunting a large deer or can each individually bag a
small hare. As Rousseau noted, the crux of the problem is that
if two hunters have partnered to hunt a deer, one partner
might be tempted to abandon the hunt to grab a passing hare
for himself, leaving his companion in the lurch. The payoffs
are as follows: R
Hunter B:
Hunter A: Stag Hare
Stag 3, 3 0, 1
Hare 1, 0 1, 1
What is the most-preferred outcome, and is there another out-

come in which neither player has an incentive to alter his
strategy (assuming the other player's strategy stays fixed)?
Does either player end up doing better playing either Stag or
Hare no matter what his partner chooses to do, as in the
marsh-draining game? How certain must A be that B will be
playing Stag to do the same? Explain how this game illus-
trates the role of trust in cooperation.
4. Hume's marsh-draining game involved a problem of cooper-

ation, whereas the following game is generally thought to in-
• A's payoff is the left number in each cell; B's is the right number.
Cooperation 261
volve a problem of coordination. The setup is that two friends

have a disagreement about whether to go to the movies or
play basketball, though neither wants to be on his own, even
doing his preferred activity:
Friend B:
Friend A: Basketball Movies
Basketball
Movies
I 3. 1
0,0
I o.1, 30 I
Does either player end up doing better with either Basketball
or Movies no matter what his friend chooses to do, as in the
marsh-draining game? What are the outcomes in which nei-
ther player has an incentive to alter his strategy (assuming
the other player's strategy stays fixed)? Explain why this
game illustrates the problem of coordination.
*5. An alternative to punishing defection, commonly employed

by criminal gangs in the United States, is rewarding coopera-
tion, for example, by looking after an individual's family while
he is in prison. Suppose, using the payoffs given in Display
8.1, that a bonus of f3cv is given to a criminal who cooperates
but whose partner defects, while a payoff of f3cc is given to a
criminal who cooperates and whose partner also cooperates.
Rewrite the payoff matrix, suitably updated for the new payoff
regime. For what values of f3cv and f3cc is cooperation an equi-
librium? For what values is it the only equilibrium?
6. This chapter has discussed several solutions to the coopera-

tive dilemma which variously find their forms in the institu-
tions of state, civil society, a nd religion. Discuss how each of
these social institutions has evolved features to facilitate coop-
eration, with r eference to external enforcement, internal en-
forcement, and repeated interaction.
9
Collective Action
To many readers of this book, the decade of the 1960s seems a

long time ago-ancient history. Besides the Beatles, bell-
bottom pants, granny glasses, afro hair styles, droopy mous-
taches, and love beads, this decade is perhaps remembered
most vividly (through that haze of imperfect memory, for those
of us old enough, or from documentary-film collages for
younger folks) as a period of intense group protest and collec-
tive action. Early in the decade, first at the University of Cal-
ifornia at Berkeley and then throughout the country, students
protested against the seemingly arbitrary and capricious au-
thority of universit y officials. The Free Speech Movement and
the Movement to Save People's Park were two major protests
aimed at carving out both personal and physical space within
which to "do one's own thing." ("Movement'' was the collective
noun of choice at the time.)
This was not the silliness many elders portrayed it as; it
was serious politics with a serious political agenda. Indeed, as
the decade wore on, mass action moved beyond the college
campuses and the agenda of protest became more explicitly
political:
• protest on behalf of civil rights, culminating in a march

on Washington that involved hundreds of thousands of
participants, and ultimately producing landmark civil
and voting rights legislation;
262
Collective Action 263
• protest against the war in Vietnam, culminating in

a march on the Pentagon (also involving hundreds of
thousands of people, as brilliantly described in Nor-
man Mailer's Armies of the Night), and ultimately
causing an incumbent president to decline to seek re-
election;
• protest against environmental degradation, culminating
in a mass rally in Washington on Earth Day, and ulti-
mately in landmark environmental-protection regula-
tions and the creation of a cabinet-level agency devoted
to the env-ironment.
The seeds of mass protest were cultivated with loving care

during this period, producing offshoots of mass protest in the
seventies, eighties, nineties, and into the twenty-first century,
concerning women's rights, gay and lesbian discrimination,
and continuing environmental causes (global warming, ozone
depletion, destruction of rain forests, strip mining, nuclear en-
ergy, toxic waste disposal, and endangered species, to name a
fe w). Perhaps the most salient of these today involves both
sides of the abortion issue-the pro-choice and the pro-life
movements.
Having become used to reading about mass action on our
campuses and in the streets of our cities, we may have taken
it as customary and failed to appreciate that it is, in fact, a
puzzling phenomenon. Mass action involves huge numbers of
individuals deciding to participate. Yet what possible differ-
ence could any one person make to a final outcome by attend-
ing one of these rallies? His or her individual contribution
is bound to be minute, while the cost of taking this action
is far from trivial. Participation involves time, possibly ex-
pense, and perhaps risk to life and limb. Putting these to-
gether, it would appear that the instrumental benefit is small
and the potential cost large. Participation sounds crazy,
doesn't it? I will explore this question in the next few

pages. 1
I will also examine other manifestations of collective action
to understand why it occurs. Why do individuals in a commu-
nity seem to follow conventions-like everyone driving on the
right side of the road (except the English and Japanese, who
drive on the left)? How are large numbers of people able to co-
ordinate their behaviors- as when the members of a large
symphony orchestra all manage to play together? How do we
make sure that people do their fair share in collective under-
takings-like cleaning up the common room in the dormitory
or the TV room in the frat house? In short, much of what
we do in life we do in groups, so much so that we often com-
mit the fallacy of false personification and talk as though
these groups had a life of their own. In fact, for much of this
century, the study of politics was typically cast as the study of
groups.
TH E GROUP B ASIS OF P OLITICS
It is no accident that much political discourse is conducted in

terms of groups. In any large society it is simply impossible to
think in highly disaggregated terms. Instead, we think about
an issue in terms of the categories of people taking an interest
in it, and a conflict in terms of the groups that line up on one
side or the other of the divide. Farmers lobby both for price
supports for their crops and for high tariffs to keep crops from
other nations out of their domestic market; consumers, of
course, are on the other side of these issues. Labor unions
1 An exceptionally subUe treatment of instrumental behavior and collective

action is found in Richard Tuck, Free Riding (Cambridge, Mass.: Harvard
Univers ity Press. 2008). Tuck, a political philosopher, treats these topics in
terms of both philosophy and the history of ideas.
push for mandated wage increases and improvements in

fringe benefits, which employer groups oppose. Minority
groups urge passage of civil rights legislation, while those
whose competitive advantage is eroded by it are in opposition.
Economic producer organizations seek various protections
from market competition, again at the expense of consumers.
Professional groups want licensing authority, while those who
use professional services want this authority regulated and
market power of professionals restrained. Associations of col-
leges and universities seek large student aid allocations and
research budgets from governments, only to be opposed by
those who want those limited budget allocations devoted to
their own activities. The list is endless.
In a pluralistic political system these groups are known as
"lobbies," "interest groups," or ,more pejoratively as "pressure
groups." At the highest level of aggregation these groups are
actually collections of groups- so-called peak associations
ranging from the AFL-CIO (whose members are unions), to
the National Association of Manufacturers and the Business
Roundtable (whose members are corporations), to the Farm
Bureau Federation (whose members are local fa1·m bureaus).
At less august levels groups simply represent aggregations of
individuals sharing a common interest- the Possum Hollow
Rod and Gun Club, the Harvard-MIT Apple User Group, the
Boston Policemen's Benevolent Society, the Massachusetts
Federation of High School Basketball Coaches, or the New
England Political Science Association.
The very ubiquitousness of groups in pluralistic political
systems explains why, for most of the first half of the twenti-
eth century, the study of politics was the study of groups. Po-
litical outcomes were seen as the result of struggles among
groups. Indeed, in its most famous representation, Arthur
Bentley wrote almost like a physicist about the "parallelo-
gram of forces" that constituted group interactions and in-
fighting. 2 He thought of the status quo in any policy domain as

a point on a page. Change occurs as this point gets pushed
around by various forces impinging on it. Bentley treated each
group in this political fray as a vector consisting of a "direc-
tion" and a "magnitude." Direction indicated what changes in
the status quo a group wanted; magnitude measured the
group's political strength. One group might "push" the status
quo-that point on the page-far to the east; another to the
southwest, but a shorter distance (reflecting its weaker politi-
cal clout); still a third due north with yet another magnitude.
The net effect of this pushing and shoving-the resultant of
the various group forces applied to the existing status quo-is
a new policy status quo. Politics, in this view, becomes
physics, each group is a "force vector," and the political out-
come of a struggle is simply the mechanical resultant of the
various forces at play.
The pull and tug of group infighting was, to nineteenth-
century observers like Alexis de Tocqueville, the definitive fea-
ture of American political life. Indeed, they admired the
voluntaristic political pluralism that was absent in less liberal
societies. 3 In either its pluralistic or less liberal form, however,
the group-based formulation of politics took groups as funda-
mental and assumed their existence. The essential axiom was
this: Common interest, however defined and however arrived
at, leads naturally to organizations coherently motivated to
pursue that common interest; politics is all about how these co-
herently motivated organizations support and oppose one an-
other.
2
Arthur Bentley, The Process of Government (Evanston, m.: Principia Press,
1908). Its famous midcentury companions were V. 0. Key, Politics, Parties,
and Pressure Groups, 3rd ed. (New York: Crowell, 1952), and David B. Tru-
man, The Governmental Process (New York: Knopf, 1951).
8 Even in these less liberal societies, conflicts were often portrayed in group
terms as struggles between large social aggregates: bourgeosie versus prole-
tariat; aristocracy versus merchant class; various racial, religious, linguistic,
or regional groups in opposition to one another; and so on.
C ASE 9.1
WHO l s R EPRESENTED?
The pluralists believe that groups for m naturally out of

shared interests. If this belief is correct, groups should form
roughly in proportion to people's interests. We should find a
greater number of organizations around interests shared by
a greater numiber of people. The evidence for this pluralist
hypothesis is quite weak. Kay Schlozman and John Tierney
examined interest groups that represent people's occupa-
tions and economic roles.* Using census data and listings of
interest groups, they compared how many people in the
United States have particular economic roles and how
many organizations represent those roles in Washington.
For example, they found that (in the mid-1980s) 4 percent
of the population was looking for work, but only a handful
of organizations actually represented the unemployed in
Washington.t
There is a considerable disparity in Washington repre-
sentation across categories of individuals in the population,
as the table below suggests. Schlozman and Tierney note,
for example, that there are at least a dozen groups repre-
senting senior citizens, but none for the middle-aged. Ducks
Unlimited is an organization dedicated to the preservation
of ducks and their habitats; turkeys, on the other hand,
have no one working on their behalf. The pluralist's inabil-
ity to explain why groups form around some interests and
not others led some scholars to investigate the dynamics of
• Organized Interests and American Democracy (New York: Harper & Row,
1986), pp. 69-71.
t Of course, the number of organizations is at best only a rough measure of
the extent to which various categories of citizen are represented in the in-
terest group world of Washington.
collective action. Mancur Olson's work, di scussed later in

this chapter, is the best-known challenge to the pluralists.
Economic Role % of U.S. % of Type of Org. in Ratio of

of Individual Adults Orgs. Washingto n. O.C. OrgsJ
Adults
Managerial/ 7 71.0 Business 10. 10
Administrative association
Professional/ 9 17.0 P rofessional 1.90
Technical association
Studentl 4 4.0 Educational 1.00
Teacher organization
Farm Workers 2 1.5 Agricultural 0.75
organization
Unable to 2 0.6 Handicapped 0.30
Work organization
Other Non· 41 4.0 Union 0.10
Farm Workers
At Home 19 1.8 Women's 0.09
organization
Retired 12 0.8 Senior citizens 0.07
organization
Looking for 4 0.1 Unemployment 0.03
Work orgamution
COOPERATION ACCOUNTS
Collective Action as Multiperson Cooperation

Groups of individuals pursuing some common interest or
shared objective-maintenance of a hunting and fishing habi-
tat, creation of a network for sharing computer software, lob-
bying for favorable legislation, playing a Beethoven symphony,
or whatever-consist of individuals who bear some cost or
make some contribution on behalf of the joint goal. Each mem-
ber of the Possum Hollow Rod and Gun Club, for example,
pays annual dues and devotes one weekend a year to cleaning
up the rivers and forests of the club-owned game preserve.
We can think of this in an analytical fashion, somewhat re-

moved from any of these specific examples, as an instance of
two-person cooperation writ large. Accordingly, each of a very
large number of individuals has, in the simplest situation, two
options in his or her behavioral repertoire: "contribute" or
"don't contribute" to the group enterprise. 4 If the number of
contributors is sufficiently large, then a group goal is ob-
tained. However, just as in Hume's marsh-draining game of
the previous chapter, there is a twist. If the group goal is ob-
tained, then every member of the group enjoys its benefits,
whether or not he or she contributed to its achievement. As long
as the marsh is drained, neither farmer is bothered by mos-
quitoes, a benefit completely detached from contribution lev-
els. In the next chapter I refer to a goal like this as a public
good, since once it is provided it becomes available to all,
whether or not they participated in its provision.5
Although this may seem like a simple extension of the two-
person cooperation problem of the previous chapter, there are
several complications on which we must dwell. First, we need
to designate whether the situation is dichotomous, in which
case the group goal is either attained or it is not, or continu-
ous, in which case the outcome varies quantitatively with the
number of contributors (or the amount of contribution). It is
•A more complicated version of this situation, one not pursued here, enriches
each person's behavioral repertoire. Instead of a twofold choice, there is
some continuously variable input, like effort or money, that the individual
may choose to contribute.
6 Surely there are situations in which individuals may actually be denied the
benefits of a group if they do not contribute to its production. If the benefit is
highly valued, then this capacity to deny it makes it easier for the group to
elicit contributions. If you want to ride on the Massachusetts Turnpike, for
example, you must make a contribution, called a toll. If you want to fish on
the Possum Hollow River, then you must pay your club dues. If you want to
hear the Boston Symphony Orchestra play a Beethoven symphony, then you
must purchase a ticket. These activities are not quite public goods, and l
defer further consideration of them to the next chapter. In the present dis-
cussion I take up the polar case of a group good or goal, once produced, being
available to all.
almost always possible to transform a dichotomous situation

into a continuous one, though it is often easier to think in
terms of a dichotomous outcome. For example, consider al-
liance behavior during wartime (or team behavior during the
Super Bowl). Allies, depending upon the number of individual
contributions (relative to contributions obtained by the other
side), may either win the war or lose it (dichotomous out-
come); but surely one can think of the "victory" as varying
from total success to one barely better than a draw (or, even
worse, a Pyrrhic victory), and "defeat" ranging from a scant
last-minute loss to total humiliation. In the remainder of this
chapter, to keep things sufficiently straightforward, I will
stick with the dichotomous circumstance: The group goal is ei-
ther achieved or not.
Second, in a multiperson situation, we need to specify how
many contributors are necessary to attain the group goal. Put
differently, we need to specify the relationship between each of
the individual contribute/don't contribute choices and the final
outcome (what economists call the production function). At
one extreme, unanimous participation is required: unless
every person in the group contributes, the group goal is not
achieved. This is a particularly interesting situation to ana-
lyze. Suppose each individual in t he group evaluated the
group goal at some level, say B utiles (where B stands for
"benefit"); assume that B > 0. 6 Suppose further that the utility
value of contributing to the group project is -C utiles (C
stands for "cost"), where C > 0, too. If C > B, then no one will
contribute, no matter how many other s do. It's just not worth
the candle, since the benefit net of costs, B - C, is negative. On
6
I could complicate the story by individualizing evaluations, so that Mr. i
values the group goal at B(i) utiles, while Ms. j values it at B(j), where
B(i),. B(j). But this enrichment does not change the thrust of this story, so I
will not employ it here. Conflict of interest among group members-that is,
differential evaluation of group goals-will be explored below.
the other hand, if B > C, then every member of the group will
contribute. Why?
Actually, there are two possible outcomes when B > C, but
we argue that only the one in which everyone contributes
makes sense. Suppose everyone has chosen not to contribute,
and, of course, the goal is not attained. Would any group mem-
ber be so unwise as to reconsider her choice? Hardly, for if,
say, Ms. j, decided to contribute, then her payoff would be -C
(the cost she bears) and there would still be no compensating
benefit (since the latter is produced only if everyone con-
tributes). So, it is possible for a group to be stuck in an "equi-
librium trap" in which no one is contributing, even though
everyone would be better off if everyone contributed.
But this outcome is not very likely. Everyone will realize
that everyone else benefits from achieving the group goal, and
that the only way for this to happen is if everyone contributes.
So, Ivls. j, like every other group member, makes the following
calculation:
If I don't contribute, then I get a payoff of 0. If I contribute, and so

does everyone else, then I get a payoff of B-C > 0. If I contribute,
but someone else does not, then I get -C. Everyone else makes
the same calculation. Everyone will realize that everyone in the
group appreciates his or her own essential status to achieving the
group goal on the one hand, and that there is nothing to be
gained by not contributing (aside from avoiding putting oneself at
'risk) on the other hand.
That is, individluals will often manage to coordinate on con-

tributing, because there is a net benefit to doing so relative to
the equilibrium-trap outcome, because every person's partici-
pation is absolutely essential, and because everyone knows
this and knows that everyone else knows this. In short, there
are a number of reinforcing factors causing individuals to see
contribution as the sensible thing to do in this circumstance
and to believe that everyone else will see it as the sensible

thing, too. The "everyone contribute" event is what Thomas
Schelling has called a focal point. 7
Suppose, now, that a unanimous contribution rate is no
longer necessary. Indeed, of the n persons in our group, sup-
pose that only k contributors are necessary for the group objec-
tive to be obtained (where k is a number greater than zero but
less than n). Ms. j's calculation is considerably different in this
case, as may be seen in Display 9.1. If fewer than k-1 of her
colleagues are con tributing, or more than k-1, then it doesn't
pay for Ms. j to contribute. In either of these cases she garners
a higher payoff by not contributing (0 instead of - C and B in-
stead of B - C, respectively). Only if exactly k-l others are con-
tributing is Ms. j's contribution essential, in which case she
would obtain B - C > 0 instead of 0 by contributing. Ms. j does
not have a clearcut course of action, because she does not know
in advance whether less than k-1, mor e than k-1, or exactly
k-l of her colleagues are going to contribute.
D IS PLAY 9.1
Ms. J'S N UMBER OF OTHER G ROUP
CHOI CE MEMBERS CONTRIBUTING
less t han k - 1 e xactly k - 1 k or mor e
Cont ribute -C B-C B-C
Do not con tribute 0 0 B
This development suggests th at t here are two rational, or

equilibrium, outcomes possible. Either no one contributes or
exactly k do. In either of these cases, no group member will
have reason to reconsider his or her action. If no one is con-
tributing, then Ms. j (or any other member, for that matter)
7
This idea was first developed in his famous book, The Strategy of Conflict
(Cambridge, Mass. : Harvard University Press, 1960).
would be foolish to decide to contribute. On the other hand, if

exactly k people are contributing, then those not contributing
have no need to contribute (and would be ill-advised to do so,
since they enjoy the group benefit even though they haven't
contributed), while the k that are contributing know that each
and every one of them is absolutely essential.
1'hus, for the group objective to be accomplished, it is clear
that something of a "knife-edge" condition must hold for suffi-
cient contribution to occur. Exactly h individuals will need to
believe that they, and only they, are likely to contribute. I
claim that this poses a much more complicated problem for
group members than the case where unanimity was required.
Indeed, the "tipping point," k, is a crucial determinant of
whether or not this group is able to get its act together.
Let's take a concrete example. Suppose n = 100. I have
already analyzed the case of k = 100, the requirement of unan-
imous contribution, and concluded that unanimous contribu-
tion is a likely occurrence. What if k were a relatively large
number but less than 100, say, k = 95. Most group members,
Ms. j included, are bound to think that if they are going to
achieve the group goal, then an awful lot of them are going
to have to contribute. The group cannot stand too many defec-
tors without los.ing the goal a ltogether. It see.m s to me that the
likelihood between the outcome of no one contributing and k of
them contributing will be a tendency for k (or possibly even
more than k) to end up contributing. What about k = 85?
Again, there will be considerable psychological pressure on
people to contribute, although not so much as when k equals
95 or 100. Ask gets smaller, the pressure is reduced.
Consider the case of k - 25. In this instance, there are
bound to be an awful lot of group members who calculate that
their contribution is just not needed. It's not (necessarily) that
they are mean-spirited but rather that they feel liberated to
make alternative uses of their effort without risking the
achievement of the group goal. Instead of going out with
twenty-four other me1nbers of the Possum Hollow Rod and

Gun Club to clean up litter along the Possum Hollow River,
for example, Ms. j may conclude that that particular Saturday
afternoon is best devoted to taking her children shopping for
spring clothes. Cooperation is especially hard to elicit from
members of a group when it is patently evident that the mem-
bers' cooperation is probably inessential; this evidence mounts
as k becomes small relative to n.
When k gets very small, nearly everyone in the group will
bail out, devoting their energies to nongroup activities. The
"rational" forecast is that the group objective will simply not
be accomplished. Yet other factors may come into play in these
circumstances, fa ctors not incorporated into the analysis
given in Display 9.1. In every group there are always some
people who "do the right thing" no matter what. For one
thing, they may secure utiles directly from the participation
itself (it feels good to tramp up and down the Possum Hollow
River on a beautiful spring afternoon). For another, as noted
in the previous chapter on two-person cooperation, they may
have internalized a value system that encourages contribution
to group life. Surely, as the psychological/strategic pressures
to which I've alluded above decline, people will drop out of
participation- but perhaps not all, and if it takes only a few to
secure the group goal, then it may well be achieved.
The general conclusion of this analysis is that the combina-
tion of strategic and psychological pressures that encourage
contribution rise as k gets large relative to n. Holding n fixed
(as I did in the development above), the likelihood that ther e
will be sufficient contribution declines as k declines-certainly
a nonobvious conclusion, inasmuch as it would appear that as
k declines it gets "easier" to secure the group goal; as a quali-
fication, however, I suggested that for very small k there may
be an uptick in the likelihood of group success, since there is
often a "dependable few" who will contribute for nonstrategic
reasons.
Collectiue Action 275
These general tendencies, I claim, get more pronounced as

n gets large, as C gets large, or as B - C gets large, holding k
fixed. As n gets large, the general conclusion reported in the
previous paragraph holds- it is practically equivalent to hold-
ing n fixed and letting k get smaller. But not quite. As n gets
large it seems to us that the psychological identification with
the group, an identification that may well affect the benefit
utiles one enjoys upon achieving a group goal, becomes more
tenuous. It's hard to feel part of a group of ten million others.
Sure, all the residents of Greater New York City have an obli-
gation to keep its streets and parks litter-free, but it is hard to
appeal to in-group feeling in urging folks to come out on a
beautiful Saturday afternoon to clean up the shores of the
Hudson River.
As C gets large, holding n , k, and B - C fixed, there will be
a tendency for the group to fail to secure its objective. The "no
one contributes" equilibrium seems specially compelling in the
case of large C, even if B - C remains fixed, because the psy-
chological risk of contributing (paying the cost of -C) but the
group falling short of the k contributors is inhibiting.
Finally, as B - C gets large, holding everything else fixed,
the importance of the group goal grows, and people are pre-
pared to take psychological risks in these circumstances. For
any size group n, and decisive contribution level k, the
prospect of the group obtaining enough contributors grows as
B - C grows. 8
8 As a first cut this analysis is plausible. But it is not satisfactory as a fully
strategic analysis. For example, I claimed in the text that as B - C grows,
the prospects for group provision grow for every group size, n, and decisive
contribution level, k. Surely it is corr ect to believe that, for any specific
group member, the pressure to contribute grows as the net benefit of group
provision (B - C) gets large. On the other hand, that same individual will
just as surely a ppreciate that tbe pressure will also be growing on all her
group colleagues. Thus, the large net benefit tugs her toward contributing,
but the certain knowledge that others will also feel this way may allow her
to rationalize letting them do all the heavy lifting. Weighing the relative
force of these countervailing effects is the sort of consideration that is part of
a deeper game-theoretic treatment.
Multiperson Cooperation and Coordination

In a sense I have just taken up the easiest kind of multiperson
cooperation, since there is only one thing the group seeks-
w hat we have been calling the group "goal" or "objective"-
and everyone in the group shares an interest in achieving it.
Into this analytical stew I now want to add a tablespoon of
nonuniqueness and a pinch of conflict of interest.
NONUNIQUENESS Let us suppose that there are many things a

group might like but that the group members are indifferent
among those many things as long as they obtain one of them.
A social club, for example, might enjoy an evening at the bal-
let or a night at the opera. They are indifferent as to which, so
long as they do it together. Suppose then that a typical group
member, Ms. j again, has two choices: go to the opera or go to
the ballet. If every group member makes the same choice,
then they each realize a benefit of B utiles. (Everyone's atten-
dance is required for those utiles because the group qualifies
for a group ticket price and gets premium seating.) Failure to
coincide on a common choice yields a positive payoff smaller
than B, say b utiles (where b < B ), though with an exception.
If everyone else coordinates on a common choice, but Ms. j
does not, then her payoff is 0. The situation facing Ms. j is
given in Display 9.2.
DISP LAY 9.2

C HOICES OF OTHERS
Ms. J'S Everyone Everyone No con -
CHOICE else else sen s us
ch ooses chooses
o pera ballet
Opera B 0 b
Ballet 0 B b
The last column of the display is irrelevant for Ms. j; her

payoff is the sam e no matter what she does, since there is al-
ready division among her prospective companions. So, for ana-
lytical purposes, it is only necessary to look at the two-by-two
part of the display consisting of the first two columns and
rows. Game-theoretic analysis tells us the obvious-that there
are two "rational" or "equilibrium" outcomes to this group ac-
tivity. Thus, abou t the only thing game theory can tell us is
that the group should coordinate. It doesn't tell us how, nor
what Ms. j should do. To be more precise, it tells us no more
than that each person should "flip a coin."
In this situation, then, the nonuniqueness of desirable
goals for the group means that it will be problematic for the
group actually t.o achieve one of them. If there were only two
people, for example, each of whom flipped a coin, then there
would be a fifty-fifty chance that they would coordinate
choices-one chance in four that both would go to the opera,
one chance in four that both would go to the ballet, and two
chances in four that they would go their separate ways. For
three people, the chances of all coordinat ing drop to one in
four, and for four people, to one in eight. Generally, for an
n-person group, the chances that all will choose ballet or all
will choose opera is (Y2) 0 - 1, a number that rapidly a pproaches
zero as n gets large. (If n = 10, for example, the probability
of successful coordination is .00195.) Thus, it is difficult for
groups to obtaiin the dividends of cooperation when there are
alternative directions in which its members may head, absent
some sort of coordination mechanism.
Although I will not develop the analysis at length, it would
appear that rep eat play of this kind of group interaction may
have the same sort of salutary effect it had in the two-person
world. P eople would probably coordinate on which side of the
road to drive, even in the absence of a traffic cop, just by
virtue of doing it over and over again. If an alien plopped
down in the middle of Beacon Street in Boston, I expect it
would learn to stay to the right quite quickly (and there would
be all manner of Boston drivers to let it know if it did other-
wise). Similarly, I could imagine our social group, after some
trial and error, implicitly coordinating-for example, going to
the opera in even-numbered months and to the ballet in odd-
numbered ones. 9
DIS PLAY 9.3

C HOICES OF OTHERS
Ms. .1's Everyone Everyone No con-
CHOICE else else sen s u s
c hooses c h ooses
o per·a ballet
Opera B,, B2' ... , Bn 0 b
Ballet 0 PI' P2' · · · ' Pn b
CONFLICT OF INTEREST Now let's consider the situation, proba-

bly the typical one, in which there are several different goals a
group might pursue and members of the group value the dif-
ferent goals differently. This situation is portrayed in Display
9.3. Ms. j, the typical group member, gets Bj utiles if everyone
9 r have said nothing about communication, and the curious reader mny
have begun to wonder why the members of this allegedJy social group
seemed incapable of talking to one another! Communication surely is an op-
tion, but it adds an additional layer of complexity that I'd just as soon not
tackle. For one thing, communication may be costly. For another, it may be
asymmetric (some people having the power to transmit, others to receive,
still others to do both, and finally some poor sods who can do neither). For
yet another, it opens up all kinds of strategic possibilities (one person leav-
ing a voice-mail message for everyone else that he's going to the opera and
will not be able to receive any incoming calls from this point on- a fait ac-
compl i). A rigorous treatment of strategic communication may be found in
Jeffrey Banks, Signaling Games in Political Science (Chur, Switzerland:
Harwood, 1991).
in the group coordinates on going to the opera and ~i utiles if

they all go to the ballet. These two payoffs need not be the
same and, depending on their relative magnitudes, Ms. j will
have very distinct preferences as to which venue she would
prefer her colleagues to coordinate on, even if both Bi and ~i
are bigger than b. Of course, if every group member ranks the
same alternative highest, then there will be no problem, but
then there will be no conflict of interest. However, in the inter-
esting case where there is conflict of interest-where Bi> ~i
but B k < ~k for group members j and k, respectively-coopera-
tion among group members is exacerbated as members not
only have to overcome the normal difficulties of coordination
associated with nonuniqueness (as discussed just previously)
but also must overcome inherent intragroup differences of
opinion about where to go.
It is conceivable that, in some specific situations, the con-
flict of interest is resolvable in a relatively painless way. If, for
instance, five of the six members of a group strongly prefer
opera to ballet, then it doesn't take much figuring by the one
ballet lover that, if the group is to obtain either of its goals, it's
going to end up at the opera. But what about the situation in
which a fifteen-person group is hopelessly divided into five in-
tense opera lovers, five intense ballet lovers, three mildly
opera-oriented members, and two mildly ballet-oriented mem-
bers. This situation is sufficiently complicated that it requires
some sort of institutional solution, as indeed are most situa-
tions that involve both nonuniqueness and conflict of interest.
I am not saying that groups, even large groups, don't manage
to overcome these difficulties in interesting, often idio-
syncratic, ways. Repeat play, for example, might help here
as it has in other situations I have examined. But the reality
is that, rather than depending upon the problem to resolve
itself in some fashion, a group will typically institutionalize
a solution. This, I believe, is the important insight of Man-
cur Olson in his seminal book, The Logic of Collective

Action. 10
O LSON'S L OGIC OF C OLLECTIVE A CTION
Olson, writing in 1965, essentially took on the political science

establishment. He noted that the pluralist assumption of the
time, that common interests among individuals are automati-
cally transformed into group organization and collective ac-
tion, was problematic. I ndividuals are tempted to "free-ride"
on the efforts of others, have difficulty coordinating on multi-
ple objectives (nonuniqueness), and may even have differences
of opinion about which common interest to pursue (conflict of
interest). In short, the group basis of politics is a foundation of
J ell-0: One cannot mer ely assume that groups arise and are
maintained; rather, formation and maintenance are the cen-
tral problems of b>Toup life and politics generally.
Olson is at his most persuasive when talking about large
groups and mass collective action. I n these situations, like
many of the demonstrations and rallies of the 1960s, n is very
large and k is relatively small. This is another way of saying
that no one individual is very significant, much less essential,
to the achievement of a group goal. In these circumstances,
the world of politics is a bit like Hume's marsh-draining game
writ large, where each individual has a dominant strategy of
not contributing.
Olson claims that this difficulty is severest in large groups,
for three reasons. First, large groups tend to be anonymous.
Each household in a city is a taxpaying un it and may share
10 (Cambridge, Mass.: Harvru·d University Press, 1965). For an excellent and

highly accessible game-theoretic treatment of collective action, the reader is
encouraged to consultAvinash Oixit, Susan Skcath and David H. Reiley Jr.,
Games of Strategy, 3rd ed. (New York: Norton. 2009), Chapter 12.
the group objective to lower property taxes, but beyond that it

is difficult to forge a group identity on such a basis. Second, in
the anonymity of the large-group context, it is especially plau-
sible to claim t hat no one individual's contribution makes
much difference. Should the head of a household kill the bet-
ter part of a morning writing a letter to his city council mem-
ber in support of lower property taxes? Will it make much
difference? If hardly anyone else writes, then the council
member is unlikely to pay much heed to this one letter; on the
other hand, if tlhe council member is inundated with letters,
would one more have a significant additional effect? Finally,
there is the problem of enforcement. In a large group, are
other group members going to punish a slacker? By definition,
they cannot prevent the slacker from receiving the benefits of
collective action, should those benefits materialize. (Every
property owner's taxes will be lowered if anyone's are.) But
more to the point, in a large, anonymous group it is often hard
to know who has and who has not contributed, and, because
there is only the most limited sort of group identity, it is hard
for contributors to identify, much less take action against,
slackers. As a consequence, many large groups that share
common interests fail to mobilize at all-they remain latent.
This same problem plagues small groups, too, as Hume's
marsh-draining game in Display 8.1 reveals. But Olson ar-
gues, much as I have in this text, that small groups manage to
overcome the problem of collective action more frequently and
to a greater extent. Small groups are more personal; there-
fore, their members are more vulnerable to interpersonal per-
suasion. In small groups, individual contributions may make
a more noticeable difference (k is large relative to n) so that
individuals feel that their contributions are more essential.
Contributors in small groups, moreover, often know who they
are and who the slackers are. Thus, punishment, ranging
from subtle judgmental pressure to social ostracism, is easier
to effectuate. Fina lly, small groups often engage in repeat play

and therefore can employ tit-for-tat strategies to induce
contributions.
In contrast to large groups that often remain latent, Olson
calls these small groups privileged because of their advantage
in overcoming the free-riding, coordination, and confl.ict-of-
interest problems of collective action. It is for these perhaps
ironic and counterintuitive reasons that small groups often
prevail over, or enjoy privileges relative to, larger groups: pro-
ducers over consumers, owners of capital over owners of labor,
a party's elites over its mass members.
Olson elaborates on this asymmetry between large and
small in suggesting that it applies within groups as well. If
group members are unequal in important ways, these inequal-
ities may actually help the group achieve its goals. But in
doing so, it often leads to inequality at the level of contribu-
tion, with the larger or more powerful members being ex-
ploited by their smaller, weaker colleagues. Because a large,
powerful group member is likely to "make a difference" in
many situations, he or she will be under intense pressure to
contribute. The ,e nterprise may succeed even if one of the
weaker members opts not to contribute, but it is more apt to
fail if one of the powerful members does not contribute. Could
the United States free-ride on its NATO allies, or the former
Soviet Union on the former Warsaw Pact? There is evidence,
in fact, that both the United States and the Soviet Union paid
more than their "fair share" to support their respective al-
liances, and that many of their smaller partners paid much
less than their fair share.
C ASE 9.2
T HE LARGE AND THE SMALL
In the case of NATO and the Warsaw Pact, the large partic-
ipants, the United States and the Soviet Union, respec·
tively, subsidized the other members of their alliances.*
Their large size and their belief in the necessity of the al-
liance structure handicapped them in negotiating with
their allies over a more equitable division of alliance costs.
David Marsh's research on corporations in Great Britain
belonging to the Confederation of British Industry (CBI)
suggests a similar exploitation of the large by the small.t
Large firms are more dependent on the group and small
firms are more likely to free-ride. The managing director of
one large firm said, "If the CBI didn't exist we would need
to create it. We: need someone to stand up and talk for in-
dustry" (p. 264). An executive at a small firm said, ''Whether
we were members of the CBI or not we would derive some
benefit from the part it plays for industry. You get that type
of benefit whether you are a member or not" (p. 264). In
s upport of Olson's theory, smaller group members often free
ride on the efforts of larger members whose participation is
more critical to the group's existence.
• Mancur Olson and Richa1·d Zeckhauser, "An Economic Theory of Al·
liances," Review of Economics an.d Statistics 48 (1966): 266-79.
t "On Joining Inter est Groups: An Empirical Consideration of the Work of
Mancur Olson, J r.," British Journal of Political Science 6(1976): 257-71.
Olson s By-Product Theory and

Large-Scale Collective Action
The politically powerful conclusions of Olson's argument are,
first, that groups are difficult to create and maintain, and sec-
ond, that smaller groups seem less afflicted by these difficul-
ties than larger groups. In rejecting the group basis of politics

that much of twentieth-century political science took as its
foundation, Olson provides a cogent explanation for the power
of small numbers, even in political democracies that nomi-
nally count noses and reward size.
Let us not, however, cast a blind eye to the reality of large-
scale collective action. The free speech movement, the civil
rights movement, and other mass-action activities do, in fact,
arise, and sometimes change the course of history. Olson
would not want to deny these powerful facts; he does, how-
ever, claim that they require explanation. The existence of
groups cannot merely be assumed. Olson's explanation for this
large-scale collective action is known as the by-product theory.
Olson argues that large groups are able to elicit contribu-
tions from rnembers by providing them more than just the
successful achievement of group objectives. Remember that
individuals enjoy successfully achieved group goals, whether
they contribute or not. So, the prospect of the goal itself is
often not sufficiently compelling to induce contribution. But
what if members are given other things, conditional on
whether they contribute? That is, what if one of the reasons
members join groups is for the private things that they get
only by making their contributions? If this is the case-if the
motivational force for contribution is the private goodies in
addition to the collective goals-then the goals achieved are,
in a manner of speaking, achieved as by-products. The contri-
butions to large groups are forthcoming because of the bene-
fits that contributors, and only contributors, enjoy. These
benefits operate as selective incentives to contribute.
Why, for example, would minimum-wage earners con-
tribute time and money to lobby the state legislature to raise
the minimum wage? If the wage is raised, then all benefit,
whether they contribute or not; and no single contribution
makes much difference anyhow. The problem for a large group
with a common interest is that it cannot deny successes to

noncontributors. Suppose, however, that only contributors
(who contributed. by joining the group and paying annual
dues) received such selective benefits as low-cost term life in-
surance, a group rate for membership in a health club, special
airfares to Florida in the winter, drug and alcohol counseling
services, and a Vlednesday night bowling league. These selec-
tive incentives might be sufficient to induce contribution. Over
and above these tangible benefits that might be denied to non-
contributors, any individual member may assert further that
what he or she really values is the solidarity that being part of
the group allows-another selective incentive.
In sum, the by-product theory suggests that a group that
provides only collective group goals may have a hard time. Es-
pecially if the group is large and anonymous, it simply may
not be possible for the goal itself to elicit enough contribution
from prospective members to permit the goal to be achieved.
The group, in effect, remains latent. So, groups search for
things separate from the main group mission that can be
withheld from those who do not contribute:
• worker groups, nominally organized to raise wages and

improve working conditions, offer Wednesday night
bowling;
those organizing donation campaigns for National Pub-
lic Radio give donors coffee mugs and sweatshirts identi-
fying them as contributors to public radio;
• trade associations, officially organized to lobby Congress
for policies beneficial to their entire industry, offer mem-
bers (contributing companies in the industry, in this
case) specialized access to trade statistics and other
industry-relevant data;
• environmental groups, campaigning for public action in
behalf of the environment, offer contributors discounts
in the purchase of camping equipment, inexpensive

fares to exotic environmentally unique destinations, and
reduced prices on books about the environment.
This list could be multiplied endlessly, which only gives

further testimony to the plausibility and persuasiveness of
Olson's by-product theory. Most mass associations and organi-
zations do precisely these sorts of things to attract and retain
members. And yet, the by-product theory seems incomplete in
several respects. To some, the by-product theory does not take
ample account of the role played by leaders. To others, it
strikes a rather cynical chord in its failure to incorporate gen-
uinely nonmaterialistic motivations. Don't some people join
groups and make contributions because they believe in the
group's cause and require nothing more than the good feeling
that they have made a contribution to it? In the next two sec-
tions I take up each of these in turn.
Political Entrepreneurs
It is very unusual in the academic world for a book review to
become an important part of the literature on a subject. But
this is precisely what happened to Richard Wagner's review of
Olson's book. 11 Wagner noticed that Olson's arguments about
groups and politics in general, and his by-product theory in
particular, had very little to say about the internal workings
of groups. In Wagner's experience, however, groups often come
into being and then are maintained in good working order not
only because of selective incentives but also because of the ex-
traordinary efforts of specific individuals-leaders, in ordi-
nary language, or political entrepreneurs in Wagner's more
colorful expression.
Wagner was motivated to raise the issue of group leaders
11
Richard Wagner, "Pressure Groups and Political Entrepreneurs," Papers on
Non-Market Decision Making I (1966): 161-70.
because, in his view, Olson's theory was too pessimistic. In the

real world, labor unions, consumer associations, senior-citizen
groups, environmental organizations, and so on all exist, some
persisting and prospering over long periods. Likewise, mass
activities like those described at the beginning of the chapter
seem to get jump-started somehow in the real world. Wagner
suggests that a special kind of by-product theory is called for.
Specifically, he argues that certain selective benefits may ac-
crue to those who organize and maintain otherwise latent
groups.
Senator Robert Wagner (no relation) in the 1930s and Con-
gressman Claude Pepper in the 1970s each had private
reasons--electoral incentives-to try to organize laborers and
the elderly, respectively. Wagner, a Democrat from New York,
had a large constituency of working men and women who
would reward him by reelecting him-a private, conditional
payment-if he bore the cost of organizing (or at least of facil-
itating the organization of) workers. And this he did. The law
that bears his n ame, the Wagner Act of 1935, made it much
easier for unions to organize in the industrial north. 12 Like-
wise, Claude Pepper, a Democratic congressman with a large
number of elderly constituents in his South Miamj district,
saw it as serving his own electoral interests to provide the ini-
tial investment of effort for the organization of the elderly as a
political force.
In general, a political entrepreneur is someone who sees a
prospective cooperation dividend t hat is currently not being
enjoyed. This is another way of saying that there is a latent
group that, if it were to become manifest, would enjoy the
fruits of collective action. For a price, whether in votes (as in
12 The Wagner Act made it possible for unfons to organize by legalizing the so-
called closed shop. If a worker took a job in a closed shop or plant, he or she
was required to jo~n the union there. "Do not contribute" was no longer an
option, so workers in closed shops could not free-ride on the efforts made by
others to improve wages and working conditions.
the cases of Wagner and Pepper) or a percentage of the divi-

dend, or the nonmaterial glory and other perks enjoyed by
leaders, the entrepreneur bears the costs of organizing, ex-
pends effort to monitor individuals for slacker behavior, and
sometimes even imposes punishment on slackers (such as ex-
pelling them from the group and denying them any of its se-
lective benefits).
To illustrate this phenomenon, there is the (no doubt apoc-
ryphal) story about a proper Britis h lady who visited China in
the late nineteenth century. She was shocked a nd appalled
upon noticing teams of men pulling barges along th e Yangtze
River, overseen by whip-wielding master s. She remarked to
her guide that such an uncivilized state of affairs would never
be tolerated in modern societies like those in the West. The
guide, anxious to please in any event but concerned in the
present circumstance that his employer had come to a wildly
erroneous conclusion, hastily responded, " Madam , I think you
misunderstand. The man carrying the whip is employed by
those pulling the barge. He noticed that it is generally diffi-
cult, if you are pulling your weight along a tow path, to detect
whether any of your team members are pulling their s or, in-
stead, whether they are free-riding on your labors. He con-
vinced the workers that his entrepreneurial services were
required and that they should hire him. For an agreed-upon
compensation he monitors each team member's effort level,
whipping those who shirk in their responsibilities. Notice,
madam, that he rarely ever uses the whip. His mere presence
is sufficient to get the group to accomplish the task."
Thus, political entrepreneurs may be thought of as comple-
ments to Olsonian selective incentives in motivating groups to
accomplish collective objectives. Both are helpful, and some-
times both are needed to initiate a nd maintain collective ac-
tion. In this respect, groups that manage, perha ps on their
own, to get themselves organized at a low level of activity
often take the next step of creating leaders and leadership in-
stitutions in order to increase the activity level and resulting

cooperation dividends. Wagner, in other words, took Olson's
by-product theory and suggested an alternative explanation,
one that made room for institutional solutions to the problem
of collective action.
Political Ideology and Belief Systems

I noted just before the previous section, as I noted in our ear-
lier discussion of multiperson cooperation, that some people
contribute to collective undertakings without either selective
incentives or leaders staring over their shoulders. Some, that
is, have internalized a value system that makes contributing
to group life a priority, whether or not it is accompanied by
material incentives or overseers. This value system is often
referred to as an ideology or belief system.
Rational-choice explanations of group phenomena-of
which my ana[ysis of multiperson cooperation, Olson's by-
product theory, and Wagner's theory of political entrepreneurs
are standard instances- tend to give short shrift to ideological
explanations, principally because they ignore the question of
where a particular ideology originates. That is, why would
an ideology or belief system that disposes one toward
cooperative/contributory behavior survive in a population and
be sufficiently numerous to overcome all the problems of
multiperson cooperation that I have discussed? These are im-
portant issues, so important that an ideological explanation,
in my view, is bound to fall short unless it can satisfactorily
account for them.
Nevertheless, I should point out (and will do so more ex-
tensively in the next section) that behavior may be thought of
in either of two ways. Most rational analysis takes behavior to
be instrumental-to be motivated by and directed toward
some purpose or objective. But behavior may also be experien-
tial. People do t hings, on this account, because they like doing
them-they feel good inside, they feel free of guilt, they take
pleasure in the activity for its own sake. We maintain that this
second view of behavior is entirely compatible with rational
accounts. Instrumental behavior may be thought of as invest-
ment activity, whereas experiential behavior may be thought
of as consumption activity. I still will not have answered the
question of where such beliefs and values originate nor why
they survive. But economists do not tell us where consumer
tastes originate either, and yet they make central use of those
tastes in constructing their theory of price. The key thing to
appreciate here is that we can still make precise statements
about how dispositions toward both investment and consump-
tion affect the prospects of collective action. I shall do this in
more detail when discussing voting, our next topic.
But before leaving this section, let me briefly note one other
aspect of experientially oriented behavior: It is the behavior
itself that generates utility, rather than the consequences pro-
duced by the behavior. To take a specific illustration of collec-
tive action, many people certainly attended the 1963 March
on Washington because they cared about civil rights. But it is
unlikely that many deluded themselves into thinking their in-
dividual participation made a large difference to the fate of
the civil rights legislation in support of which the march was
organized. Rather, they attended because they wanted to be a
part of a social movement, to hear Martin Luther King Jr.
speak, and to identify with the hundreds of thousands of oth-
ers who felt the same way. Also-and this should not be mini-
mized-they participated because they anticipated that the
march would be fun, an adventure of sorts.
So, experiential behavior is consumption-oriented activity
predicated on the belief that the activity in question is fulfill-
ing apart from (or in addition to) its consequences. Individuals,
complicated beings that we are, are bound to be animated both
by the consumption value of a particular behavior that I just
described and its instrumental value, the rational (investment)
explanation that has been used throughout this book. To insist

on only one of these complementary forms of rationality and to
exclude the other is to provide only a partial explanation. This
is no more apparent than in the activity of voting, the center-
piece of democratic politics.
CASE 9.3
W HAT D OES THE EVIDENCE SAY?
What does the evidence say about these different explana-
tions of collective action: Olson's by-product theory, Wag-
ner's theory of political entrepreneurs, and the rationality
of ideology and experiential behavior? In various studies
group members have been surveyed to determine why they
become members. The survey results indicate that people
join for a combination of reasons. In support of Olson's the-
ory, members of economic groups are more likely to join for
selective, material benefits than for collective benefits. Eco-
nomic groups include unions, farm groups, and business as-
sociations. Members of these groups often disagree with the
political goals of the group, suggesting that the latter are
not the chief reason for joining.*
In opposition to the by-product theory, some studies
have found that members of noneconomic groups are moti-
vated primarily by collective benefits. Individuals often join
noneconomic groups such as Common Cause or the Sierra
Club primarily because they agree with the group's political
goals.t Overall, the evidence indicates that the motivation
for membership in interest groups is a combination of selec-
tive and collective benefits, differing slightly for economic
· Terry Moc, The Organization of Interests (Chicago: University of Chicago

Press, 1980).
t John Mark Hansen, "The Political Economy of Group Membership,"
American Political Science Review 79 (1985): 79-96.
and noneconomic groups. Members of economic groups join

primarily for the selective benefits (instrumental behavior),
while members of noneconomic groups join primarily for
the collective benefits (experiential behavior).
Whether they join for the selective or the collective ben-
efits, me mbers appear to be behaving rationally. A study of
members in thirty-five national organizations in the United
States found that "members are attracted to, or seek out,
those inducements that are most closely related to their
central interests in an organization.":!: The study is impor-
tant because it shows that ideology can be incorporated into
a rational choice account of political behavior. Individuals
committed to saving endangered species, an instance of a
collective benefit, may "shop around" among different
groups in order to find the one they believe will best serve
their objectives.
Wagner's theory of political entrepreneurs is also sup-
ported by social science research. Two studies have found
that outside support is often vital to forming and maintain-
ing a group. In addition to Hansen, cited above, Walker
found that most political action is supported by large,
wealthy institutions, such as charitable foundations.§
Thus, the collective action problem for some interest groups
is overcome by political entrepreneurs in the form of patron
institutions. But what about these latter institutions- how
do they solve their collective action problem? Robert H. Sal-
isbury points out that these institutions are, in Olson's
terms, privileged groups small enough to overcome the col-
lective action obstacles they face.** If we expand the con-
cept of political entrepreneur to include patron institutions,
Wagner's argument holds up quite well.
l David Knoke, "Incentives in Collective Action Organizations," American
Sociological Review 53 (1988): 311 29.
I Jack L. Walker, Jr., Mobilizing Interest Groups in America (Ann Arbor:
University of Michigan Press. 1991).
•• Interests and Jnst1tutions (Pittsburgh: University of Pittsburgh Press,
1992).
T HEORIES OF VOTING AND

COLLECTIVE ACTION
The kind of collective action with which nearly all citizens of

democracies are most familiar is that of choosing leaders. In -
deed, Americans elect more officials a t all levels of government
than any other de mocracy in the world. Rarely a year goes by
for the typical American eligible to participate in elections
without there bein g a race for the House or Senate, a presi-
dential election, a state legislative or gubernatorial race, or
even a contest for the proverbial town dog catcher. Involve-
ment in the electoral process, whether attending campaign
rallies, contributing money to a favorite candidate or distrib-
uting her literature door to door, helping to "get out the vote,"
or voting itself, is collective action par excellence. Participa-
tion in election activity, then, like any other group activity,
cannot be assumed, but rather must be explained. Why partic-
ipate? That is t.he question. A strictly instrumental analysis is
the appropriate starting point but, as we shall see, does not
provide the last word. In one of the most fa mous articles on
voting in the political science literature, William Riker and
Peter Ordeshook supply the following analysis. 13
Suppose there are two candidates for a public office, Jack-
son and Kendall. A typical citizen eligible to vote in this elec-
tion, Ms. j again, must decide whether to vote and, if so, for
whom to vote. The act of voting costs Ms. j C utiles, reflecting
the time and energy, and perhaps the financial expense, of in-
forming herself and actually going to the polls. Suppose, with-
out loss of generality, that Ms. j prefer s the election of J ackson
to that of Kendall. That is, u(J) > u(K), where u is Ms. j's util-
ity function and J a nd K stand for the election of Jackson and
Kendall, respectively. Put equivalently, if J ackson should win
13 William H. Riker and Peter C. Ordeshook, "A Theory of the Calculus ofVot·
ing," American Political Science Reuiew 62 (1968): 25-42.
rather than Kendall, then Ms. j gets a benefit B = u(J) -u(K)

> 0. If Ms. j were the only voter, then the answer to the ques-
tion of participation is straightforward. If the benefit of pick-
ing the winner, B, exceeds the costs of doing so, C, then she
should do so, picking Jackson; if, on the other hand, C > B,
then she shouldn't bother (allowing the choice to be made ran-
domly instead), s ince the utility difference between the candi-
dates is not worth the cost Ms. j would bear to make the
choice. But of course Ms. j is not the only voter, so she must
take the intentions and capabilities of others into account.
Ms. j, we shall suppose, lives in a district in which there
are n eligible voters in total, each of whom has a preference
between the two candidates, bears costs if he or she exercises
the franchise , and hence must go through essentially the
same kind of analysis as Ms. j. The task must seem daunting,
even for moderately sized n, if one must try to scope out, for
each and every other eligible voter, who is going to vote and
for which candidate. In fact, however, the task simplifies quite
naturally. There are really only five circumstances that Ms. j
(or any other voter) needs to consider. These involve how the
other n-1 voters (excluding Ms. j) behave in the aggregate,
and may be partitioned into five "states of the world" (labeled
S 1 through $ 5 ). These "states" are the outcomes that would
transpire if Ms. j abstained.
8 1 : Jackson loses to Kendall by more than one vote.

8 2 : Jackson loses to Kendall by exactly one vote.
8 3 : Jackson and Kendall tie.
8 4 : Jackson beats Kendall by exactly one vote.
8 5 : Jackson beats Kendall by more than one vote.
If Ms. j believes 8 1 prevails, then her vote will have no effect

on the final result, no matter how she casts it, since Kendall
wins in any case. If she believes that 8 2 is the prevailing state,
then she knows that she can cast a vote for Jackson that pro-
duces a dead h eat. With S 3 she can break what would other -
wise be a tie. In state S 4 her vote for Kendall would produce a
tie (though why she would ever want to do this is beyond us,
since she prefers J ackson). Finally, if S 5 is the prevailing
state, then, like in S., her vote, however she casts it, will have
no effect since J ackson wins regardless. Display 9.4 gives the
complete picture.
Each cell of the display gives the utility payoff to Ms. j,
which depends both on the state of the world (what everyone
else is doing) and her own choice. Notice that if Ms. j votes,
then her utility for the outcome, whatever it is, is always re-
duced by C, the utility cost of her participation. Of course, if
she abstains, then she does not pay this cost. The only term in
this display requiring further explanation is L , which enter s
once in each row. L stands for " lottery," and reflects the fact
that the election ends in a tie. In this case, we assume that
some random device is used to determine the winner; the lot-
tery is a fifty-fifty chance of either J ackson or Kendall. The ex-
pected utility theorem (see Chapter 2) implies that u(L) •
112 u(J) + 1/2u(K).
A glance at Display 9.4 reveals that Ms. j should never vote
for Kendall, since in each state of the world the payoff to her
in the "vote for Jackson" row is at least as big as its counter-
DISPLAY 9.4
Ms. J'S STATE OF THE W ORLD
CHOICE
81 82 83 8, 86
Vote for u(K)-C u(L )-C u(J)-C u(J)-C u(J)-C
J ackson
Vote for u(K)-C u(K )-C u(K)-C u(L)-C u(J)-C
Kendall
Abstain u(K) u(K ) u(L) u(J) u(J)
part in the "vote for Kendall" row. 14 Undoubtedly the reader is

thinking, "I didn't need a fancy analysis to tell me that!" In-
deed, it should be obvious that when there are only two candi-
dates, you either vote for your preferred candidate or don't
bother voting at all. There's never anything to be gained by
voting for your less-preferred candidate.
The analysis will have something to tell us that's not so ob-
vious when we ask the fundamental question: Does the payoff
from voting for Jackson exceed the payoff from abstention?
This requires us to compare the first and third rows of the dis-
play. Unlike the comparison of rows 1 and 2, however, in some
states (columns) "vote for Jackson" gives the larger payoff
while in others "abstain" is more attractive. In order to sort
this out, we must incorporate into the analysis Ms. j's beliefs
about the likelihoods of the various states. Then we must use
some simple algebra to figure out what Ms. j should do.
To simplify things, let us set u(J) - 1 and u(K) = 0. 15 More-
over, let us represent Ms. j's beliefs by probability numbers.
In particular, Ms. j believes 8 1 occurs with probability pl'S2
with probability p 2 , 8 3 with probability p 3 , S 4 with probability
p 4 , and 8 5 with probability p 5 (wher e each probability number
is 0 or larger and together they sum to 1). This information
is contained in Display 9.5, which is simply Display 9.4 with
the "vote for Kendall" row deleted and the above revisions
14 In the first and fiftlh states, the payoffs are identical. In the third state,
since u(J ) > u(K ). the statement in the text holds. The only places in which
there may be some confusion is when L is involved. In the second state, a
fifty·fifty chance of getting your preferred candidate is surely better than
the certainty of getting your worst choice. ln t,he fourth stat,e, the certRinty
of getting your best candidate is surely better than a lottery in which your
chances of getting him are only fifty.fifty. So, the conclusion holds that
Ms. j docs at leas t as well voting for Jackson as she does voting for Kendall.
15
In our discussion of utility funcnon s in Chapter 2, we mentioned that it is
often convenient to "normalize" the analysis, without doing any logical dam-
age, by setting the utility of th<.' most-prefened alternative to unity and
that of the least-preferred to 0.
DISP LAY 9.5

M s. J'S STATE OF T KE W ORLD
CHOICE (PROBABlLITY)
sI 82 Sa 84 86
(p l ) (p 2) (P 3) (p 4) (p 5)
Vote for -C ~-C 1-C 1-C 1-C
J ackson
Abstain 0 0 Yi 1 1
incorporated-u(J) set to 1, u(K) set to 0, and u(L) set to

1/2. 16
We apply the expected utility theorem from Chapter 2 di-
rectly to this display as follows:
EU (vote for J ackson) = p 1 (-C) + p 2 (1/2- C) +Pa (1-C) + P4
(1-C) + p 5 (1-C)
1/2 P2 +Pa+ P4 + Ps - C
=
EU (abstain) = p 1 (0) + p 2 (0) +Pa (1/2) + P4 + P 5
= 1/2 P 3 + P4 + Ps
Ms. j should vote for J ackson rather than abstaining if and

only if EU (J ackson) > E U (abstain). With a little more alge-
bra, this means that Ms. j should vote for Jackson if a nd only
if p 2 + p 3 > 2C. In words, she should vote if the sum of the
probabilities that she either makes a tie (voting for J ackson in
8 2) or breaks a tie (voting for J ackson in Sa) exceeds twice the
cost of voting.
1.'his is certainly not obvious, and is a good deal more com-
plicated than the reader might have thought at the outset .
What does this implication for Ms. j's participation mean?
16 The lottery givin.g a fifty-fifty chance of Jackson or Kendall is, given the
normalization, a fifty-fifty chance of getting a utility of 1 or of 0. Thus, the
expected u tility of th is lottery is equal to 1/2.
This deduction is actually quite rich in implications. First, it

says that the costlier it is to participate, the less likely Ms. j
will be to do so. This follows because the inequality of p 2 + p 3 >
2C is more difficult to satisfy as C gets large. Indeed, if C is
sufficiently large (specifically if C :2: 112), she should never par-
ticipate (since probability numbers could not satisfy the in-
equality). Second, it says that Ms. j should be most disposed to
participate if she believes the election is going to be close. This
follows because p 2 + p 3- the likelihood of "making or breaking
a tie"-is a serviceable definition of a close election; the larger
those probabilities get, the more likely it is that the inequality
will be satisfied.
In sum, the Riker-Ordeshook calculus of voting provides a
rationale for participation based, first, on the cost of partici-
pating and, second, on the likelihood a prospective participa-
tor will make a difference. But there is a third, most
disturbing, implication of this analysis. Suppose C were very
small relative to u(Jackson) - u(Kendall) = 1; say, C • 111000.
This is not an unreasonable supposition in important elections
like presidential contests, since it asserts that the cost of vot-
ing to the average citizen is quite small relative to the differ-
ence it makes to them whether, say, Democrat or Republican
wins. So the inequality p 2 + p 3 > 2C now says that Ms. j should
vote if and only if her probability of making or breaking a tie
(p 2 + p 3 ) were larger than 2/1000. What is the likelihood of
making or breaking a tie in a mass election in the United
States, where there are approximately 125,000,000 prospec-
tive voters? It is infinitesimal, and certainly much smaller
than 211000. So it is very unlikely that a sensible person, like
Ms. j, will believe that p 2 + p 3 exceeds 211000. Thus, when all
is said and done, most sensible persons, according to this
analysis, will conclude that they should not vote.
One interpretation of this strictly instrumental analysis of
Riker and Ordesh ook is that instrumental calculations are
simply insufficient to induce participation in large-scale elec-
tions of the sorrt that occur in most modern democracies. The

fact that in the real world of mass elections there is consider-
able participation proves embarrassing. Noting this, Riker
and Ordeshook salvage their own analysis by conceding that
there is more to a calculus of voting than computing the conse-
quences of voting versus abstaining. They suggest, though not
in the words I use, that there is an experiential as well as an
instrumental basis for voting-that voting has consumption as
well as investment value.
For one thing, individuals in democratic societies possess a
belief system or ideology in which great stock is placed in par-
ticipation. Abst ention is frowned upon-by one's neighbors, by
one's spouse, even by one's children.17
For another thing, punishments are often inflicted on non-
participators. In some societies, a fine is imposed. In others
there are "watchdogs": the neighbor who goes door to door or
leans over her fence, imploring her neighbors to vote; the shop
steward who makes it clear to the men on the shop floor that
they had better make time at lunch hour to vote; the party ac-
tivists who, late in the day, check with pollwatchers to see who
has not yet voted.
Finally, it must be said that individuals participate in elec-
toral activities not only to avoid feelings of guilt or to dodge
the "punishments" inflicted by others but also because it can
be fun. A voter may find satisfaction in standing in line at the
polls chatting with neighbors. One enjoys the next-day conver-
sations at the office, or over coffee with neighbors, not only
about the election but about one's own participation in it.
Riker and Ordeshook account for this experiential source of
utility by alterirlg the payoffs in Displays 9.4 and 9.5. In the par-
ticipation rows- "vote for J ackson" and "vote for Kendall"-
17 The author remembers all too keenly how painful it was to face his children
afer returning home from work on the evening of an election and confessing
that he "just didn't have the time to vote." That experience, seared in his
memory, has factored into all subsequent participation decisions for him!
but not in the "abstain" row, an additional consumption payoff

is added. That is, while there surely are costs associated with
participation (measured by the -C), there are also benefits,
like those just discussed above. These benefits may be suffi-
ciently large to induce some who would not otherwise partici-
pate on purely instrumental grounds. 18
There is now a large literature on rational theories of vot-
ing. 19 Since my primary interest in voting here is to display it
as an instance of collective action, I will not pursue the subject
in further detail-except to note one feature that the percep-
tive reader may have already discerned. In my treatment of
collective action, I have mainly considered a single group in
isolation faced with prospective dividends from cooperation.
Indeed, I went further by focusing on a generic individual in
isolation, seeking to determine the conditions under which
that individual would choose to contribute to a group enter-
prise. The example of voting, however, suggests that many col-
lective action situations pit groups against each other. Ms. j
may well be a Jackson supporter, and the question for her is
whether to contribute, along with other Jackson supporters, to
the Jackson cause. But somewhere out there is lurking a
Mr. k, a Kendall supporter, with a similar problem. Yet, their
respective problems are interdependent and should not be
treated in isolation.
That is to say, many collective action problems are not only
18 I will not burden the reader further with an analytical demonstration. In-
terested readers may consult the original analysis by Riker and Ordeshook.
For a broad discussion of the issue, including a critique of the Riker and Or-
deshook approach, the reader is encouraged to examine Brian Barry, Sociol-
ogists, Economists, and Democracy (Chicago: University of Chicago Press,
1970).
19 The ambitious reader is directed to two companion volumes: James M.
Enelow and Melvin J. Hinich, The Spatial Theory of Voting: An Introduction

(New York: Cambridge University Press, 1984), and James M. Enelow and
Melvin J. Hinich, eds., Advances in the Spatial Theory of Voting (New York:
Cambridge University Press, 1990). Also see Samuel Merrill III and
Bernard Grofman, A Unified Theory of Voting (New York: Cambridge Uni-
versity Press, 1999).
problems of individuals sharing a common objective and seek-

ing to overcome incentives that discourage contribution. They
also involve strategic interactions among competing groups.
The fate of groups in securing their respective objectives often
depend not only upon own-group success in encouraging par-
ticipation but also upon other-group success in collective ac-
tion. In short, life is complicated, and this analysis has only
scratched the surface. In order to treat this more complicated
manifestation of collective action, I would have to take up is-
sues of strategy in considerable detail, entering the domain of
game theory. Of course, I cannot do everything in a single
textbook, but I encourage the student to pursue this matter
independently.20
1. One vision of the thrust-and-parry of democratic politics is

that it is essentially about the interaction of interest groups.
So, if we wish to understand the outcome of some policy de-
bate, we should focus primarily on the political groups with an
interest in the issue at hand, and their power vis-a-vis one an-
other. Can we take the existence of organized interest groups
for granted in aittempting to understand politics? Why or why
not? What variables might we consider in attempting to qua n-
tify the power of political interest groups? For example, what
role does the size of a group (latent or active), the structure of
interests within the group, or the quality of its organization
and leadership play in determining its strength?
20
Excellent starting points are two books to which I have referred readers be·
fore: Avinash Dixit and Barry Nalebuff, Thinking Strategically and Avinash
Dixit, Susan Skeath, and David H. Reiley, Jr., Games of Strategy.
2. Four friends live together in a college apartment and must

work together to ciean the common areas before their parents
arrive for Parents' Weekend. For simplicity, suppose that this
outcome is dichotomous (the apartment is either fully cleaned
or it isn't) and has all the features of the collective action prob-
lem described in Display 9.1 and the surrounding text. As-
sume B > C for each individual.
• What are the possible equilibrium outcomes whe n all

four friends must contribute to clean the apartment?
Which do you think is likely?
• What are the possible equilibrium outcomes when only
two of the four friends must contribute to clean the
apartment (i.e., k = 2)? Can you predict which outcome
will occur without further information?
How might your prediction change when k = 2 if B in-
creases, or C decreases? What about if Bis different for
different members of the group?
3. Most industries (e.g., steel, sugar, or automobiles) organize

one or more trade associations to lobby Congress on matters of
regulatory, trade, and tax policy, among many others. For each
of the following situations, describe the likelihood of success-
ful organization for lobbying on policies favorable to that in-
dustry, the fruits of which are enjoyed by all individuals in the
industry whether they contributed or not. If there is a clear
answer, explain the thinking behind that prediction.
• One industr y has a large number of small firms produc-

ing while another has only a few very large firm s.
Both industries have a medium number of firms, but in
one, successful lobbying will require nearly universal
support of the lobbying effort, while in another, success-
ful lobbying will require only about 50 percent of firms

to contribute.
• All firms within an industry agree on policy priorities,
while in a second industry there is considerable dis-
agreement over targets for lobbying efforts.
One industry contains firms of relatively equal size,
while another industry contains a few very large firms
(whose contributions to lobbying are proportionally effi-
cacious) and a large number of small firms.
4. AARP is a political organization that represents the inter-

ests of people over fifty, through lobbying, negotiation of spe-
cial discounts for members, and research on issues of interest
to its members. Members pay small yearly dues and in ex-
change are eligible for the above-mentioned discounts as well
as AARP-affiliated insurance plans. In fact, AARP's insurance
recommendations ended up being highly profitable for AARP
co-founder Leonard Davis, who left the organization in 1979
over discontent with the nonprofit organization's links to the
insurer Colonial Penn.8 Today, AARP has roughly 40 million
members, is one of the most powerful congressional lobby
groups in Washington, and is a major player in all legislation
of interest to senior citizens. Their lobbying activity is backed
up by impressive registration and get-out-the-vote efforts
every election cycle, although AARP itself is nonpartisan.
Based on this vignette, how has AARP been so successful in

organizing senior citizens, securing massive membership and
high rates of political engagement? Use the ideas presented in
this chapter to make your case.
• "Leonard Davis; Helped Start AARP and Geronotology Programs at USC,"

Los Angeles 'IImes, January 23, 2001.
5. What is the paradox of participation? Explain why voting

can be considered a collective action problem, and why voting
behavior based strictly on the utility of possible outcomes
raises this paradox.
Now consider a further wrinkle: every district has at least

some voters, but participation is much higher in some places
than in others, a nd at some times more than at others. What
might account for this variation in participation rates? Does
the Riker-Ordeshook analysis suggest any possible implica-
tions, or are other explanations more compelling? In answer-
ing this question it might be useful to consider "motor-voter"
laws, election day registration, mail-in ballots, and other pro-
gra ms designed to reduce C, as well experiential benefits of
voting.
*6. This question is based on a model discussed in McCarty

and Meirowitz (2007), which is in turn based on Palfrey and
Rosenthal (1984).b Suppose that there are n individuals who
desire a collective good that yields benefit B to all n individu-
als. Provision of the good requires only one individual (k = 1)
to expend C to provide it (B > C).
• Show that ther e n possible sets of pure strategy equilib-

ria (each player i plays "contribute" or "don't" with prob-
ability l).c
• Now suppose that all players are playing an identical
mixed strategy; that is, they probabilistically choose
b :-lolan McCarty a nd Adam Meirowitz, Political Game Theory: An Introduc-
tion (Cambridge, U.K .: Cambridge University Press, 2007), a nd Thomas Pal-
frey and Howard Rosenthal, "Participation and the Provis ion of Discrete
Public Goods: A Strategic Analysis," Journal of Public Economics 23 (1983):
171-93.
r Recall the definition of equilibrium th11t we have been using thus far: Taking
all other players' strategies as fixed, each player has no incentive to alter his
strategy. Thus, an e quilibrium is a point of stability, in which no one wishes
to change his or her "play."
whether to play C or D. Call p the probability that any

one player plays C. First, show that fo r any player i, if
he doesn't contribute, the probability that the good is
supplied by someone else is 1 - (1 - p )n- 1 • Second, equate
the expected utility for i of contributing with the ex-
pected utility of not contributing. Note that when these
expected utilities are equal, i is happy to play C with
probability p , just like the others. Third, solve the ex-
pression you just found for p. Fourth , show that p is de-
creasing in C, increasi ng in B, a nd decreasing inn.
7. Aristotle wrote in Politics (Book II, Chapter 3):

For that which is common to the greatest number has the least
care bestowed upon it. Every one thinks chiefly of his own, hardly
at a ll of the common interest; and only when he is himself con-
cerned as an individual. For besides other considerations, every-
body is more inclined to neglect the duty which he expects
a nother to fulfill; as in families many attendants are often less
useful than a few.
In other words "everybody's business is nobody's business."
How does this phenomenon affect groups with political inter-
ests, a nd why is it especially problematic for what Olson calls
''large, latent groups"?
8. Chapter 2 articulated a relatively "thin" vision of rational-

ity, defined in terms of an ability to discern among possible
outcomes, and th en purposive action to secure those preferred
outcomes. From this perspective, is voting irrational (or para-
doxical)? Does the idea of experiential benefits provide a satis-
fying rational choice explanation of voting? Does the paradox
of voting call into question the enterprise of explaining politi-
cal behavior with a rational actor model? Give arguments pro
and con for this final question.
10
Public Goods, Externalities,
and the Commons
The social dilemmas that arise from properties of goods, and

the manner in which they are produced or consumed, bear a
close relationship to the establishment a nd maintenance of co-
operation and collective action. This may sound like economics
but, in fact, it's politics through and through. Markets are best
thought of as hun1an constructions, not as elements of some
natural order. They require that political understandings and
institutions come into being and persist. Surely, politics can
break markets, as economists are wont to remind us in their
critiques of overly zealous regulation; but politics makes mar-
kets, too.
DEFINING T ERMS
Most goods exchanged in economic markets are called private

because the "owner" has full control over their use. If you buy
a tube of toothpaste it belongs to you in the sense that its use
is entirely under your control. 1 Specifically, you can exclude
others from enjoying its use. This is an especially important
1
Of course, there are (politically imposed) limits to your discretion. In most
societies there are laws against squeezing toothpaste into someone else's
iPod without their permission, for example.
306
Public Goods, Externalities, and the Commons 307
form of control inasmuch as toothpaste gets "used up"; if the

purchaser could not exclude others, then it would hardly be
worth it for her to make the purchase in the first place. Conse-
quently, private goods are defined by two properties: exclud-
ability (the owner may exclude others from enjoying the good)
and solitary supply (use depletes the availability of the good). 2
Goods lacking in both of these properties are called public
goods. They are nonexcludable- anyone can e njoy them
whether she has paid for that privilege or not-and are jointly
supplied (nonrivalrous)-one person's use does not diminish
the supply available to others. Classic examples of public
goods include national defense and lighthouse services.
Consider the former. If defense services are extended over
a broad territory; then anyone living in that territory is a ben-
eficiary. Suppose, for example, that a feudal lord's castle, can-
non, and knights in fourteenth -century England effectively
protect a territory extending, say, twenty miles in any direc-
tion from the castle. The significance of this is that bands of
robbers are discouraged from practicing their trade within the
lord's jurisdiction. 3 Thus, anyone living within the boundary
enjoys relative peace-and that enjoyment is not contingent
on whether the person pays general feudal dues to the lord,
pays specific user charges for protection, or is an ally of the
lord. Just by being a resident of the castle's territory, one may
"consume" the lord's protective services.
What, really, is this protective service? In effect, it is deter-
rence. The lord's might discourages predators, and this dis-
couragement cannot be parceled out very effectively to some
in the territory a nd denied to others. Thus, the entire territory
2 Alternatively, thesegoods are sometimes called riualrous.

3 This does not mean there is no robbing and thievery. Robber bands may hide
out in, say, Sherwood Forest a nd make raids on the lord's territory. The pres-
ence of the lord's defense forces, however, by raising the costs of thieving,
discourages the frequency and intensity of the activity. Citizens, conse-
quently, may not be spared entirely but are spared more so than if the lord's
defense forces did not exist at all.
enjoys it. In this instance, we say that defense is a public good

because it is nonexcludable (it is available to everyone if it is
provided to anyone) and jointly supplied (one person's enjoy-
ment of this deterrence does not diminish its availability for
enjoyment by others in the territory).
In between the ideal types of private and public goods are
mixtures of the two. Some goods are jointly supplied, but ex-
cludable. A Madonna concert comes to mind in which a high
wall, a limited number of entry points, and turnstiles serve to
exclude those without tickets, even though, within limits, Ma-
donna's music is jointly supplied. 4 Other goods are nonexclud-
able but not jointly supplied. What is ideal about a Cape Cod
beach in August is enjoyment of the sun and surf without feel-
ing like you are crowded together with others like sardines.
The ideal beach, then, shares with private goods the attribute
of solitary (or at least limited) supply-beyond some level of
density its enjoyment by an additional family diminishes its
pleasure for other families. (This density threshold is often
surpassed on Cape Cod in August!) But it does not possess the
other private-good attribute of excludability. To the contrary,
nonexcludability, at least as regards public beaches, means
that on the hottest days of summer you are cheek-by-jowl with
loads of others. All these distinctions are illustrated in Display
10.1. I shall focus mostly on the extreme cases of an ideal-type
public good and an ideal-type private good.
•The limits have to do with the degree to which jointness of supply is compro-
mised by crowding. If greater crowding, beyond some limit, actually affects
the quality of the good, then we cannot claim that the good is jointly sup-
plied. We shall pursue this prospect in the next illustration.
DISPLAY 10.1
P UBLIC AND P RIVATE Goons
Excluda bility
Yes No
Yes :v1adonna concert Public goods: defense,
Joint ness lighthouse services
of
S upply No Priva te goods: Crowded Cape Cod
toothpaste, BMW 320i beach
P UBLIC Goons AND POLITICS
Politics rears its ugly head because, like cooperation (which is

often undersupplied) and collective action (which is often un-
derachieved), the provision of public goods is subject to socially
destructive incentives. Because a public good is nonexclud-
able, it may be enjoyed without paying a price for it. But a
producer will be loath to provide a good if he cannot elicit pay-
ment for it. And even if there were some imperfect method by
which a potential producer could extract a return from provid-
ing a public good, the amount supplied would likely be very
much less than it would be if payment could be extracted di-
rectly. As a result, everyone is worse off.
Peasant farmers in a feudal world may well be willing to
pay something for the lord's protection (certainly as much as
they would have to pay, in terms of time, energy, and lost op-
portunities, to guard against predators themselves); but if the
lord has no way of eliciting this payment from beneficiaries of
his protection, then he will be less disposed to provide it, or
very much of it at least, in the first instance. It was something
like this that supported the enforcement of feudal arrange-
ments in many parts of the world; accordingly, peasant fami-
lies were coerced into contributing hours of labor in the lord's
fields, a younger son to the lord's army, and a proportion of

their crops to the lord's granary in exchange for being kept
safe. Because protection is a public good, its supply by means
of ordinary market exchange is problematic, necessitating the
substitution of politics- the enforcement of coercive feudal in-
stitutions- for economic exchange. Political institutions, like
feudalism, arise to fill the economic vacuum.
Sometimes the political arrangement is at one remove. The
classic example of lighthouses illustrates this. The services of
a lighthouse con stitute a quintessential public good. If a light-
house is erected on high ground near a shipping hazard, the
warning it emits is available to every ship that passes (nonex-
cludability) a nd its use by one ship does not deny it to others
(jointness of supply). No ship will willingly pay for lighthouse
services, since nonpayment cannot lead to a refusal of service
to nonpayers-if the service is provided at all. But if a private
individual or firm cannot be compensated sufficiently to earn
a normal return, it will not be inclined to invest in the provi-
sion of lighthouse services. For this reason, lighthouses turn
up in introductory economics texts as t he classic instance of a
public goods problem- in which a public good is undersup-
plied owing to socially perverse incentives.
In a superb piece of economic detective work, the Nobel
laureate Ronald Coase revealed that generations of under-
graduates had been misled by the lighthouse example.5 In
England, at least, lighthouses we re quite commonly provided
along its western coastline by private enterpreneurs. But how
could s uch entrepreneurs obtain a return on their investment?
The ingenious answer Coa se provided is that lighthouses typ-
ically were positioned near harbors, allowing ships to enter
without crashing onto da ngerous shoals. The lighthouse was
primarily needed by precisely those ships coming into port.
5 Coase's essay first appeared in 1974 and has been reprinted in R. H. Coase,
The Firm, the Market, and the Law (Chicago: University of ChicaRo Press,
1988), pp. 187- 215.
Ships not intending to put ashore would typically travel

somewhat farther out to sea, thus not especially requiring the
services of a lighthouse. Consequently, there was a way to dis-
criminate between most users and nonusers. Was there a
method for converting this capacity to discriminate into a ca-
pacity to extract payment? If a monopolist controlled the wa-
terfront of the harbor, then he could price lighthouse services
jointly with docking privileges in a manner that captured a re-
turn for the former. Lighthouse services, then, were part of a
tie-in sale; if a ship owner wanted to use wharf a nd warehous-
ing facilities of the port, h e would be r equired to pay for the
lighthouse services he consumed as well.
The monopoly position of the entrepreneur is crucial her e.
If the lighthouse provider was only one of many owners of
wharfs and warehouses, he could not charge extra because of
competition for customers. Other wharf and warehouse own-
ers could charge a price for their services lower than the tie-in
sale price. So, in order for a lighthouse to be provided, an en-
trepreneur must enjoy the political protection of his monopoly
position. 6
Consider one last illustration, this one of more contempo-
rary vintage. In th e 1970s the entire industrialized world was
held hostage by a cartel known as OPEC-the Organization of
Petroleum Exporting Countries. This organization, led by the
oil ministers of the member states, conspired to jack up the
price of petroleum by restricting the amount that would be
available for export. The logic according to which they oper-
8 Having just read a chapter on multiperson cooperation, the astute reader
might wonder whether there are alternatives to granting an entrepreneur
mo nopoly ri ghts. That is, even if there were several wharves and ware-
houses in port, their owners migh t arrange to jointly finance a lighthouse
and cover their costs through charges on their port services. This is an inter-
esting possibility that the reader might like to think through. Note, however,
that this type of cooperation is, in a modern context, regarded as collusion in
restraint of trade and thus a violation of the law, because it essentially en-
tails price-fixing. A waiver from antitrust prohibitions, like protection of mo-
nopoly power, is a political necessity.
ated was quite straightforward and well-known. From the

simple law of supply and demand, for a given level of world
demand for oil, if the supply were restricted, then its price
would rise. Suppose the competitive price for and quantity of a
barrel of oil- the ones that would emerge from competition
among oil producers in the absence of a cartel-are Pc and Qc,
respectively, with total revenue, Re= Pr x Qc. I f the cartel could
successfully restrict quantity to Qoprc' an amount less than Qr,
then the price would rise to Popec• an amount higher than Pc.
The new total revenue is Ropec =Popec x Qopec· Under conditions
prevailing in the 1970s, it was possible to find a Qopec and its
associated Poper that produced a larger total revenue, that is,
Ropec >Re. Thus, if the oil producers could agree on a system of
quantity-restricting production quotas-one for each exporter
- that added up to Qnpec' and could hang together by honoring
these quotas, th ey would thereby reduce the amount of oil
available on the world market and have a bigger revenue pie
to slice up among themselves.
The higher price that prevails because of this restriction on
oil supply is a public good for OPEC (and a public bad for
everyone else). Let's see how this works. I said that a public
good is, first of all, nonexcludable, and this is certainly true of
a prevailing price. Every oil exporter gets the prevailing
price-certainly those in the cartel, but even those that are
not members. Second, a public good is jointly supplied, and
this, too, is true of the prevailing price. One supplier selling
its product at that price does not deny that same price for
some other supplier.
The joint actions that sustain this price require each sup-
plier to stick to its production quota (so that the total amount
of oil for sale adds up to the optimal Qopec' thus generating the
optimal revenue, R 0 peJ· But providing this particular public
good, like the provision of public goods generally, is problem-
atic. Each supplier will be tempted to cheat on the cartel by
producing more than its quota. If the little bit extra is suffi.
ciently smal1 so as not to affect the prevailing price, then a

cheater can seJl more than its quota at the cartel-s upported
higher price than it would if it honored the quota. But if each
member of OPEC cheats on the cartel, then there will be more
oil on the market, the price will decline, revenues will drop,
and each member will have incentives to begin a further
round of cheating. In the end, like so many instances of collec-
tive action that we have already examined, everything unrav-
els and the cartel fails.
Indeed, this is what ultimately happened to OPEC in the
1970s. But it took quite a while for the cartel to break apart;
in the meantime, OPEC did very well while the r est of the
world suffered immense economic hardship. Why did the car-
tel last as long as it did? The a nswer, like our answers to the
provision of defe nse and lighthouses, is that a political under-
standing sustained OPEC's operation. In this case, one petro-
leum exporter, Saudi Arabia, was dramatically larger than
a ny of the other members of OPEC. Saudi Arabia vigilantly
enforced the cartel agreement by using various carrots and
sticks to induce compliance by its smaller cartel partners with
previously set production quotas. Saudi Arabia (which was in-
tent upon being the dominant state in the Arab world and, not
uncoincidentally, also had the most to gain from cartel pricing,
given its oil resources) took on the burdens of political leader-
ship to hold the cartel together. 7
We thus see- that public goods will go underproduced, if
produced at all, because individuals have private incentives at
odds with those required to support their production. Individ-
uals have private incentives to enjoy the benefits of defense
and lighthouses without paying for them . Potential producers
appreciate this prospect and, consequently, a re discouraged
7 A sophisticated strategic analysis of OPEC, with Saudi Arabia conceived of

as a dominant member seeking to preserve its reputation, is fou nd in J a mes
E. Alt, Randall Cahvert, and Brian Humes, "Reputation and Hegemonic Sta-
bility," American Political Science Review 82 (1988): 44fHi6.
from producing them unless they can find some means by

which to elicit contributions. Potential cartel partners often
forgo cartel formation because they can anticipate that their
various partners will cheat on the cartel, in effect seeking to
enjoy cartel benefits without paying for them. Again, the pub-
lic good-in this case a higher price for cartel products-will
be produced only if members can assure one another that the
behavior required to sustain the higher price will be forthcom-
ing. In all these cases the solution, like the solution to the
problems of cooperation and collective action reviewed in the
two preceding chapters, is political. Perhaps the most common
solution of all- the quintessential political solution-is the
public supply of public goods. This solution requires a section
all its own.
PUBLIC SUPPLY
I have argued that the provision of public goods is poorly han-

dled by ordinary market means. They are undersupplied rela·
tive to the levels that the members of society would prefer.
Absent some sort of political intervention, there is, as we have
just seen, too little protection from predators and too few
lighthouses. Politically enforced feudal arrangements and mo-
nopoly rights in ports, respectively, are solutions to these
problems (though not perfect solutions). An alternative is to
turn to the state for public goods provision. Let the govern·
ment build lighthouses and raise armies.
In many parts of the world, lighthouses, the protective
services of the police and army, judicial services, public utili- ,
ties like water, sewage, and power, and provision for public
health, roads, and other infrastructure are commonly pro-
vided by government. Telephone and television are also often
provided publicly in many countries. The argument is that,
because they are public goods (or at least "publiclike"), private
market actors will not provide them (at least not in sufficient
quantities) because they cannot be assured of adequate com-
pensation. The state, on the other hand, may use its authority
to require payment, either out of general revenue raised by
taxation or from user charges of vaxious sorts. 8
However, there is a paradox associated with the public pro-
vision of public goods. Public provision does not just happen.
Political pressure must be mobilized to encourage the institu-
tions of government to make this provision a matter of public
policy. Bills must be passed, appropriations enacted, and gov-
ernment agencies created. In short, political actors must be
persuaded to act. But if the provision of a public good distrib-
utes a benefit widely, and if the enjoyment of that benefit is
unrelated to whether a contribution has been made toward
.mobilizing politicians to act, then we may reasonably ask:
Why would any individual or interest group lobby the govern-
ment for public goods? Why wouldn't they, instead, free-ride
on the efforts of others, thereby freeing up thei!' own resources
either to lobby for some other private benefits or to deploy in
the private sector for private gain? That is, if many public
services are like public goods, then therr supply depends upon
individuals and groups successfully engaging in collective ac-
tion to get the government to provide them. Since magic
wands are not available, the "public supply" solution to the
provision of public goods becomes a problem in collective
action.
8 The reader should notice that public goods, as we have defined them, and
publicly provided goods may not be the same. The latter may be public goods,
like lighthouses and national defense; but the state provides lots of other
goodies-like mail delivery, for example-that are sufficiently like other pri-
vate goods that they undoubtedly could be provided reasonably well in the
marketplace. (Indeed, courier services, overnight mail delivery, and package
delivery are provided privately in direct competition with the U.S, Postal
Service.) Publicly provided goods and services- the activities in which gov-
ernments engage-refl ect the political advantages possessed by interests in
the political process t.hat are sufficient to induce the public sector to do their
bidding. Some of these things are public goods, but not all of them.
At this point, the reader may wish to return to Chapter 9

to review various conventional solutions. However, ·there is a
less conventional answer. First, we must distinguish between
the consumption of a public good and its production. 9 When
designating a good as public or private, we are really talking
about consumption properties- whether you can exclude oth-
ers from consuming the good or not and whether consumption
diminishes the availability of the good. I have not remarked at
all about production. In fact, in nearly every instance public
goods are produced with substantial private input.
When the U.S. government began constructing the mas-
sive interstate highway system in 1956, it was not intended
for the government to get into the concrete business, the paint
business, the sign- or guardrail-making business, or even the
highway construction business. The government would use
its taxing and borrowing powers to raise money on the one
hand, and its substantive political authority to make choices
about highway routes and road attributes on the other. But it
would then request proposals for building highways from pri-
vate contractor s subject to these specifications. Successful
contractors-actors from the market economy- would then
make the concrete, pour it according to design, paint the yel-
low lines in the middle, assemble guardrails, signage, over-
passes, and so on.
The highway system, surely public in consumption, is in
fact mostly private in production. Various aspects of the pro-
duction process can be divvied up among contractors. The con-
tract to provide concrete, for example, is excludable (the
winning contractor gets the contract and can bar losing bid-
ders from sharing in the associated profit) and is solitarily
supplied (giving the contract to A eliminates its availability to
9 This distinction is made persuasively by Peter Aranson and Peter Orde-

shook, "Public Interest, Private Interest, and the Democratic Polity," in
Roger Benjamin and Stephen Elkin, eds., The Democratic State (Lawrence:
University of Kansas Press, 1985).
B, C, D, ...). According to the criteria in Display 10.1, high-

way construction fits squarely in the private-good category.
So, who do you suppose lobbies for a highway system? On
the consumption side, as we have seen, there are collective-
action problems. By conventional means they may be overcome
to some extent. Thus, the American Automobile Association
and the American Truckers Association, representing different
segments of the consuming public, undoubtedly brought their
political muscle to bear on legislators and executive branch of-
ficials on behalf of a highway program. Similarly, there are
likely to have been political entrepreneurs taking up the
cause-for example, legislators representing districts contain-
ing large transshipment centers (Chicago, Denver). But those
most likely to gain directly and immediately (and less likely to
have been as plagued by collective-action problems as groups
on the consumption side) are those who would actually pro-
duce the public good. Concrete producers, highway contrac-
tors, makers of heavy equipment, manufacturers of guardrails
and steel supports for overpasses, owners of rights-of-way, and
many others all stood to make enormous sums of money from
this multibillion-dollar project. In short, the politics of public
supply is as much about the production of public goods as it is
about their consumption.
The lesson here is that the politics of public supply cannot
be adequately understood as a col1ective-action phenomenon
among those wishing to consume public goods. Consumers of
public goods like good highways, clean air, lighthouses, and
security from national defense certainly play a role in provid-
ing political pressure. They are, however, limited by the
collective-action obstacles with which the reader is now famil-
iar. In fact, their interests often never materialize into group
action; they remain latent.
On the other hand, for every reference to consumers of na-
tional defense, to take one of the most important public goods,
there are thousands of references to the "military-industrial
complex," those who profit directly from the production of na-

tional defense. They are Olson's "privileged groups" who have
the ability to s urmount their own collective-action problems
and the incentive to do so (profits). They are found in the com-
mittee rooms and hallways of the Capitol, testifying, lobbying,
and spreading campaign dollars around to any legislator who
will take up their cause. Weapon systems constitute an excel-
lent illustration of this phenomenon. In 2009, President
Barack Obama sought to eliminate the production of a num-
ber of F-22 fighter planes (at a saving of many billions of dol-
lars). Some legislators, like Senator (and former presidential
candidate) John McCain, supported the president, arguing
that these weapons were no longer necessary for the national
defense. But most legislators fought the president tooth and
nail. The reason: Businesses in more than forty states and
three hundred congressional districts had subcontracts for the
production of these planes. These businesses, along with la bor
unions representing workers and state and local officials con-
cerned with the business climate in their localities, consti-
tuted a lobbying force their legislators in the House and
Senate found extremely difficult to ignore. In trying to under-
stand the public supply of public goods, then, the astute ob-
server will look at the supply side as well as the demand side
of the "market."
Before concluding this discussion let me note several com-
plaints lodged against public supply. The major concern with
public supply as a solution to the problem of providing public
goods is that public-sector actors may not have "good" incen-
tives. In this version of the "who will guard the guardians"
problem, the question is not whether government is capable of
supplying public goods but rather how well it does the job.
A classic instance of this involves the production of scien-
tific knowledge. Many kinds of knowledge constitute public
goods to the extent that they cannot be patented or copy-
righted. Once it is known, for example, that e = mc2 , individu-
als cannot be excluded from this knowledge on the one hand,

and one person knowing it does not diminish its availability
on the other. Scientific knowledge belongs in the public goods
cell of Display 10.1.
The production of scientific knowledge is undertaken very
s ubstantially by the private sector- in places like California's
Silicon Valley, Boston's Route 128, and North Carolina's Re-
search Triangle. But this kind of research tends to be very ap-
plied, tied t_o specific product development, conducted secretly,
and often patentable (thereby preventing those who do not
"own" it from making use of it). Thus, applied scientific re-
search, to the degree that property rights may be assigned to
its products, is essentially a private good. However, basic or
fundamental research-research that often does not have im-
mediate application-is not patentable and thus cannot be
owned; it therefore tends to be underproduced by the private
sector for all the public-goods reasons mentioned earlier.
Consequently, the U.S. government, through various agen-
cies like the National Science Foundation, the National Insti-
tutes of Health, the National Aeronautics and Space
Administration, the Department of Energy, and the Depart-
ment of Defense, sponsor basic scientific research-a clear in-
stance of the public provision of a public good. Some of this
research is actually done in government laboratories. But
much is contract ed out to university scientists. Consider now
the incentives facing, first, the legislators who provide the fi-
nancial resources for and oversee the execution of this public
good and, second, the bureaucrats who actually administer
the programs.
As it happens, the universities that are best positioned to
compete for basic research grants are not randomly distrib-
uted throughouit the territorial United States. While many
locations have t!he capability, ther e are discernible concentra-
tions of excellence: the Bay Area, Los Angeles, and Seattle on
the West Coast; Chicago and Minneapolis in the Midwest;
Chapel Hill- Durham- Raleigh, Miami, and Atlanta in the

South; and Washington, New York, and Boston on the East
Coast-to name some of the most prominent. If the National
Science Foundation (NSF), for example, were to support re-
search proposals strictly on the basis of merit, a disproportion-
ate amount of its budget would be spent in these pockets
of excellence. 10 Institutions in a great majority of the legisla-
tive districts of the nation would do rather poorly in the com-
petition. And this, in turn, would not kindly dispose their
representatives toward NSF. In short, while legislators may
generally approve of producin g public goods like scientific
knowledge, they are much more focused on getting federal dol-
lars for citizens and institutions in their districts. A govern-
ment agency that flouted this concern of large numbers of
legislators would undoubtedly not fare well in the annual ap-
propriations process.
The administrators at NSF are not stupid. They can fore-
cast the profound budgetary problems their agency would en-
counter if it did not attend to the conditions of representative
government. So, they arrange for alternatives to the merit-
based allocation of their budget. Instead of earmarking their
entire budget for basic research-which would end up being
spent chiefly in a small number of pockets of research
excellence- they invent new categories and new programs in
which less-well-endowed parts of the country are competitive.
Research in science education (as opposed to pure science), for
example, may be quite competently conducted in many places
around the country, places that do not require advanced re-
search laboratories and cutting-edge scientists.
Constituency-oriented legislators and survival-oriented bu-
reaucrats and administrators, not philosopher kings, support,
10
If, instead, we were discussing the production of art and culture as financed
by grants from the Nation11l Endowment, for the Arts, merit-based concen-
tration would be even more extreme, with New York and Los Angeles secur-
ing the lion's share of support.
finance, and administer public programs that produce public

goods. Their incentives dispose them to move away from what
would be optimal if only the most effective production of pub-
lic goods were motivating them. Public provision, then, is wa-
tered down by these competing, indeed distracting, objectives.
Thus, while public provision may seem the best way to go in
correcting for the underproduction of public goods, it is not
without its shortcomings.
A second incentive distortion associated with public provi-
sion involves time horizons. Many scientific projects are years
in the making. The initial phases are often relatively inexpen-
sive and invisible, as ideas are examined, developed, and
tested in small ways. Only after these initial hurdles are
cleared are greater sums spent on large-scale testing and de-
velopment. It is the latter, however, that involve new labora-
tory facilities, expensive high-tech equjpment, or advanced
testing sites-the sorts of things to which the local legislator
can point with pride (and snip the ribbon at the dedication
ceremony hea...ily covered by the local merua). The political
pressures associated with public provision, as a result, involve
truncating the longer incubation and percolation process ide-
ally associated with scientific research into a much shorter
time horizon. 11
To sum up, the production of public goods is a problem for
communities because of the very nature of these products. Pri-
vate incentives are typically insufficient to e ncourage suffi-
cient production voluntarily. Some sort of political fix is
required, examples of which include grants of monopoly privi-
lege, waivers of antitrust laws, public subsidy of private pro-
duction, and outright public provision. None of these is ideal
because each entails the grant of extraordinary privilege or
authority to some individual- the lord of the manor, the firm
11 This argument is elaborated in Linda Cohen and Roger Noll, The Technology
Pork Barrel (Washington: Brookings Institution, 1991).
granted a monopoly, a public-sector bureaucrat-whose incen-

tives may not be aligned properly to the social objectives being
sought. The lord of the manor wants prestige and glory, not
public defense; the firm wants profits, not the optimal na-
tional highway system; the bureaucrat wants turf and budget
authority, not scientific discoveries. Though perhaps an overly
cynical view, the public good is the incidental by-product of,
not the motivation for, their behavior.
It is, therefore, not surprising that different communities
at different times experiment with alternative (imperfect) so-
lutions. In the past few decades, for example, we have wit-
nessed a tidal wave of change in which public sectors that
formerly provided public goods directly are abandoning these
activities. Under the rubric of privatization, both developing
political economies and already developed ones are selling off
state-owned assets to the private sector, hoping that, imper-
fect as they may be, private-sector incentives will be better
aligned to social objectives than under the former arrange-
ment of direct public provision. This may also entail techno-
logical enhancements that mitigate some of the "publicness" of
the good. 12
One of President Obama's early initiatives was a "cap-and-
trade" program to control the air pollution that contributes to
global warming. The idea is to set a pollution target, issue
"pollution permits" (either giving them away initially or a uc-
tioning them off), and then allow permit holders either to use
them (enabling them to employ production technologies that
have pollution as a by-product up to the limit allowed by the
pollution permit) or to trade them (sell them) for others to use.
12 For example, electronic lighthouses emit an electronic signal, rather than a

light, which is received only by those ships that purchase the special signal
detector. Cable television, likewise, requires a cable box and hookup that
permits exclusion (thereby privatizing a public good). A public water supply
may be metered at each household, thereby permitting user charges; so,
too, may a firm's effluent (via sewer or smokestack), thus allowing for the
pricing of its use of the environment as a dumping site.
The objective is to limit pollution, as well as to allow the

''right" to pollu te to gravitate, via a market for permits, to
those most needing to use polluting technologies. (For a good
source of information on these developments, see the blog of
environmental economist Robert Stavins, "An Economic View
of the Environment," at http://belfercenter.ksg.harvard.edu
/analysis/stavin , and, in particular, his May 29, 2009, entry,
"The Wonderful Politics of Cap-and-Trade.")
We also observe the related phenomenon of deregulation in
which heavy-h anded bureaucratic oversight, command, and
control are being relaxed or relinquished altogether. The im-
perfectness of any solution to the production of public goods
stimulates this experimentation; but politico-economic change
of this magnitude is, as we have emphasized, political through
and through, with winners and losers determined at the end
of the day in political arenas.
CASE 10.1
P UBLIC G OODS, P ROPERTY RIGHTS, AND
THE R ADIO S PECTRUM
An interesting example of a public good is the radio spec-

trum. As a public good, the radio spectrum is nonexcludable
and jointly supplied. In other words, anyone is physically
able to broadcast on the radio spectrum, and my broadcast
doesn't prevent you from broadcasting. The problem is that
my broadcast interferes with your broadcast if both are si-
multaneous and on the same (or a closely neighboring) fre-
quency. To make radio transmission coherent, there must
be some means for allocating the radio spectrum.
Frequencies on the radio spectrum in the United States
are allocated by the Federal Communications Commission
(FCC). Different frequency bands have differ ent uses, in-
eluding television, radio, cellular telephones, and radar. In-

dividuals and organizations are given exclusive rights to
particular frequencies. Broadcasting on a frequency for
which you do not have rights is illegal. (The movie Pirate
Radio (2009) depicted one such station.) By allocating prop-
erty rights to the radio spectrum, the public good is made
excludable and the crowding that might otherwise result
from joint supply is prevented.
The example of the radio spectrum demonstrates that
there are numerous approaches to allocating property
rights. Historically, the FCC has distributed licenses at no
charge, either through application or lottery. The Clinton
administration, seeing an opportunity to bring some rev-
enue into the federal coffers, explored the possibility of auc-
tioning radio licenses for new personal communications
technologies.* At one point the administration predicted
revenues of $4.4 billion over four years. What is common to
all of the arguments about plans like this one is their fun-
damental political nature. Who will benefit from the plan?
Who will be hurt? Is the plan fair? What are the values th at
determine how we manage our public resources? Although
discussions of topics like auctions, revenue sources, and ex-
ternalities often appear purely economic and technical in
their nature, it is important to remain conscious of the po-
litical issues that lie beneath the surface.
* Edmund L. Andrews, "Radio Rights: A Move to Auction Licenses that
Sell." New York TI.mes (March 21, 1993), p. E6.
E XTERNALITIES
An externality is a special kind of public good. It is typically

the unintended by-product of voluntary activity that is im-
posed on others. Thus, an externality is jointly supplied and,
because it cannot be easily avoided, nonexcludable (although
here we might more accurately say it is unavoidable). Some
externalities are valued-the scent and appearance of the
roses planted in a neighbor's garden; the freedom from infec-
tion others obtain when we innoculate our children against
communicable diseases; the protection provided to both part-
ners when one uses a condom in sex- so we call them positive
externalities. In each case someone else benefits, perhaps un-
intendedly, from an individual's action. Other externalities are
loathed-the effects from the burning of high-sulfur coal in a
manufacturer's boiler; litter in public parks; the loud music of
boom boxes in H arvard Square-thus we call them negative
externalities. Since externalities are special instances of pub-
lic goods ("bads"), we may deduce that the positive ones are
undersupplied and the negative ones oversupplied, relative to
what would be optimal for the community as a whole. Neither
the neighbor planting her roses nor the factory burning coal
takes our preferences into account. If, as we are often advised,
we "stop to smell the roses," we discover that there are too few
roses and too many other things to smell.
The phenomenon of externalities is nicely illustrated by an
experiment the author regularly runs in an undergraduate
class at Harvard University.13 An even number of students is
selected, half of whom are designated as "buyers" and half as
"sellers" in a make-believe market. Each buyer is given a
schedule informing him how much the experimenter will pay
L3 The experiment is described in great detail by its designer in Charles Plott,

"Externalities and Corrective Policies in Experimental Markets," The Eco·
nomic Journal 93 (1983): 106-27.
him at the end of the session for each unit of the product pur-
chased during the exper iment. For example, the experimenter
may pay 30 points for the first unit, 28 points for the second
unit, 25 for the third, and so on. A buyer, then, makes a profit
if his first purchase in the market is for less that 30 points,
the second purchase is for less than 28 points, the third for
less than 25, and so on. Each buyer wants to earn as many
points (as much profit) as possible, since these points will be
added to the scor·e of his midterm examination. Similarly, each
seller is given a schedule informing her of the cost of produc-
ing each unit. For example, the first unit may cost 12 points,
the second 15 points, and so on. A seller makes a profit if she
sells each unit for more than its cost (the first unit for more
than 12 points, the second for more than 15, and so on). She,
too, wants to earn as many points as possible--for the same
reason.
The buyers and sellers sit across a table from one another.
When the market opens, bargaining begins with buyers shout-
ing out "bids" and sellers shouting out "asks" in what is
known as a double oral auction. When a buyer and seller come
to an agreement on a price p, the sale of a unit is registered. If
the buyer with the schedule given in the preceding paragraph
is buying his first unit, then his profit is 30 - p; if the seller is
selling her second unit (having already sold one earlier), then
her profit is p - 15. (If p lies between 15 and 30, then both
make a profit.) The market remains open until no one can
agree on a price for consummating any further sales. Since
the experimenters have fixed the schedules so that the "ceil-
ing" on acceptable prices for a buyer gets lower and lower with
each purchase, and the "floor" on acceptable prices for a seller
gets higher and higher with each sale, there will always come
a time when it is no longer possible for a buyer and seller both
to make a profit. 14 The market closes at this point.
14 The seller's floor ultimately becomes higher than the buyer's ceiling.
As described, this experimental market is a model of the

trucking, bargaining, and haggling that goes on in a bazaar or
city market. It is well understood according to the law of sup-
ply and demand, and experimental results validate this law
quite impressively. But we are not interested in that, since
this experimental market setting has a twist. Every time a
sale is consummated, everyone in the market, both tbe buyer
and seller participating in the particular sale as well as all
those buyers and sellers not participating, is charged 1 point
each. In effect, the consummation of a sale generates a nega-
tive externality, harming participant and nonparticipant
alike. The particular buyer and seller can take this "damage"
on themselves into account. Factoring in the externality, the
buyer in the previous paragraph will figure his profit at 30 - p
- 1 (so that p will have to be less than 29, or "no sale"), while
the seller will figure her profit at p - 15 - 1 (so that p will have
to be greater than 16). The effect of the externality is to nar-
row the bargaining range for this buyer-seller pair. 15 But nei-
ther buyer nor seller has an incentive to take into account the
impact of the externality on others.
And they don't. Even though the experimenter provides
each participant with a table informing them of the impact on
the entire market of each sale they consummate (1 point of
"damage" on every buyer and seller per unit sold), the subjects
never take this information on board. The only things they
care about are their profit thresholds (ceiling and floor for
buyer and seller, respectively), the negotiated price p , and the
impact of the externality on each of them. Each participant is
intent on maxing out on points, thereby raising his or her
midterm examination grade (and, presumably, increasing the
chances of getting into law school). Nevertheless, on other oc-
casions these very same students are heard denouncing pol-
16 With no exernality, the bargaining range for pis 15 to 30. With the exernal-
ity, this range becomes 16 to 29.
luters of the atmosphere, destroyers of the ozone layer, litter-

bugs, and producers of secondhand smoke! The fact is that it
is easier to see the scoundrel in others than in ourselves.
P ublic policy economists, at least since Adam Smith, have
worried quite a lot about how externalities, both positive and
negative, might be taken into account by those who produce
them. I cannot review all these solutions here, but will men-
tion a few in passing. Probably the most popular and widely
used solutions are taxes and subsidies, the former to discour-
age negative externalities and the latter to encourage positive
externalities. In t he experiment above, suppose the experi-
menter informed the market participants that there would be
a sales tax, t, charged against the .seller each time a sale was
consummated. The seller two paragraphs back would now
earn a profit of p - 15 - 1 - t, effectively raising the minimum
price she must now secure to show a positive profit. LG From
Econ 101 it is well known that the effect of raising a price is
that fewer sales will be consummated (at higher prices), and
hence fewer externalities generated. We don't want to elimi-
nate sales altogether (unless the externality were so horrid as
to overwhelm the benefits from having this market in the first
place). But there is an "optimal" tax, one that internalizes the
full effect of externalities. The tax, in this case, implicitly
forces the buyers and sellers in a market to take account of
the external consequences of their actions, something they
were not willing to do unless coerced in this manner.
The argument is exactly analogous when positive external-
ities are involved. In place of a tax, a subsidy is given to one or
the other of the market participants in order to encourage
more sales (and more externalities) than would otherwise
transpire.
ir. The reader should not think that l am being unfair here in placing the tax
only on the seller, smce some of it will be passed on to the buyer in negoti·
ating a final purchase price.
Experimentally, taxes and subsidies work as this theoreti-

cal argument su ggests. And, in the real world, taxes on the
sulfur content of coal, on gasoline, and on solid-waste effluents
cause their users to internalize the negative social effects of
the pollution their activities are producing. Subsidies for in-
noculating against communicable diseases (available at less
than cost), secu1~ing a higher education (tuition never covers
costs), taking public transport (fares never cover costs), car-
pooling (designated commuter lanes), using solar power (tax
breaks), and moving to the frontier (free or cheap land), in
precisely the same fashion, reduce the costs of engaging in
these activities, thereby increasing their levels (and the posi-
tive externalities associated with them).
There are two major shortcomings associated with this
strategy for dealing with exter nalities. The first is that of set-
ting the appropriate level for taxes and subsidies. In the ex-
periment described above, each sale generated one point's
worth of "damage" for every market participant. If there were,
say, six buyers and six sellers, then twelve points of external-
ity would be generated per unit sold. As we have already seen,
two of these twelve points are taken account of- namely, the
one point of "damage" falling on each party to the exchange.
It's the ten points falling on those not party to the exchange
that are ignored. by the parties to the deal. Thus, by setting
the sales tax at ten points per transaction, the contracting
parties are forced to act as though they are considering the ex-
ternal effects of their actions. The equilibrium number of sales
in this market now is socially optimal. 17
The matter of setting the optimal tax rate is quite straight-
17 The reader should note that in taking account of the external effects of
transactions, this solution does not eliminate externalities altogether.
Rather, the damage done by the externality is balanced against the benefit
that accrues from allowing buyers and sellers to capture gains from ex-
change. In general, we typically do not, as a matter of public policy, wanL to
drive negative externalities to zero because this would mean passing up
profitable exchan ges.
forward when demand conditions, supply conditions, and ex-

ternality effects are known with quantitative precision, as in
our experimental world. In the real world, however, matters
are not so straightforward, since we rarely know everything
we need to know (that is, the things provided by the experi-
mental design). The consequence is that tax or subsidy rates
are often little better than educated guesses. They may im-
prove the situation, but they may also make matters worse. 18
The second, more serious drawback to the tax-or-subsidy
solution to externality problems involves the matter of exactly
what activities should be taxed or subsidized. If one were to
survey the activities that are taxed or subsidized in any place
at almost any time, it would be impossible to claim that con-
trol of externalities had much bearing on these policies.
Surely some goods are taxed or subsidized to deter negative or
encourage positive externalities, and I have given examples of
these in the preceding discussion. But many goods are taxed
or subsidized because political machinery for taxing and sub-
sidizing exists in the first place and comes under the influence
of those who benefit from its policies, quite independent of any
consideration of externalities. On the other side of the coin, so
many other goods are not taxed or subsidized, even though a
control-of-externalities case could be made, for much the same
reason-political influence. "Optimal" taxes and other ideas
from welfare econ omics theory, even if they might work in
principle, get steamrolled in the rough and tumble of politics.
A classic instance is found in America's experience with air
is Although I will not trouble the reader with details, in the running example
from the experiment., a tax of ten points per sale will still permit some sales
to be consummated, though a smaller number than in the absence of the
tax. If we had not been sure about the damage done by externalities, and
(incorrectly) guessed that instead of one unit per person the damage was
two units per person, the tax (now twenty units per sale) would have com-
pletely shut the market down. No sales would have occurred. Thus, mis-
taken guesses about the right tax or subsidy rate may make matters worse
than no tax or subsidy at all.
pollution. In the 1970s much pollution was created by station-

ary sources, like power plants, burning high-sulfur coal. When
coal with high. sulfur content is burned, sulfur compounds
spewing out of smokestacks combine with water in the atmos-
phere to produce "acid rain," which damages crops, forests,
wildlife habitats, and fresh water sources, not to mention
human lungs. Much of this dirty coal was (and still is) mined
in Pennsylvania, Kentucky, and West Virginia. A clean alter-
native exists in low-sulfur coal, mined in the western United
States. This was a clear circumstance for the imposition of a
tax. If coal were taxed in proportion to its sulfur content, then
stationary sources would find it in their interest to switch at
least some of their energy demand from eastern to western
coal: the higher the tax rate, the more the substitution of
clean for dirty coal.
Enter politics. The West Virginia coal industry had during
this period a very powerful protector- West Virginia senator
and the majority leader of the U.S. Senate, Robert Byrd. The
Senate is an institution in which well-positioned individuals
(especially committee and subcommittee chairs and party
leaders) can exercise significant veto power. It is relatively dif-
ficult to get a bill through the Senate, and it is considerably
easier to prevent a bill from passing. And this Byrd did. De-
spite a powerful environmental lobby, and an administration
sympathetic to its preferences, Byrd managed to thwart
sulfur-content taxes by acceding to a much milder policy of
requiring the installation of pollution scrubbers on smoke-
stacks.19
There are countless stories of this sort in which a powerful
politician uses his or her position to block either the imposi-
tion of taxes on key supporters or the reduction of their subsi-
dies. Only under the direst of fiscal circumstances (like the
19
The entire story is told in Bruce Ackerman and William Hassler, Clean
Air I Dirty Coal (New Haven: Yale University Press, 1981).
large federal deficits in the United States during the late

1980s), when the insatiable revenue requirements of govern-
ment cause it to raise taxes and scale back subsidies wherever
it can, is this protection insufficient. Tax-or-subsidy solutions
to externality problems are only occasionally effective, be-
cause politics constrains their proficiency when the shoe
pinches the wrong toes.
Two other categories of solution to externality problems
merit brief consideration. We have just seen that Senator Byrd
was able to replace what would have been an onerous tax on
his dirty-coal constituency with a more tolerable regulatory
regime. Regulation is a more hands-on approach to the control
of externalities. It typically entails the creation of a govern-
mental bureaucracy- an agency, bureau, or commission-
charged with setting standards, prices, fees, or practices in
consumption or production activities that generate externali-
ties. Statutory authority usually spells out the purposes to
which this bureaucratic control should be put and the discre-
tion the bureaucratic entity has in pursuing those purposes.
Through administrative procedures, or the civil and criminal
court system, the agency has an ability to enforce its com-
mands. Thus, a governmental entity such as the Environmen-
tal Protection Agency, with authority granted to it by a law
such as the Clean Air Act, can specify the kind of smokestack
scrubber required of a stationary-source polluter.
Alternatively, externalities can be contolled by a respecifi-
cation of property rights. Part of the quandary underlying ex-
ternality problems is poorly specified rights of ownership and
use. Since nobody "owns" the air, anyone can use it as a repos-
itory for dumping things (like sulfur-based particulates). To
take an example a little closer to home, since no one owns or
has responsibility for the common room in the dormitory, it is
forever a mess. In some situations, however, it is conceivable
that one could respecify property rights so that damage done
by externalities can be held in check. I pursue this alternative
Public Goods, .Externalities, and the Commons 333
in more detail in the next section, where I discuss "commons

problems." Here, however, I remind the reader of an interest-
ing property-rights solution to air pollution (noted earlier).
Partially in reaction to the poor performance of other meth-
ods for controlling externalities, some economists have sug-
gested that there may be a way to allow the atmosphere to be
"owned." By "owned" it is meant that someone has the right to
use the atmosphere as he or she sees fit on the one band,
and may sell or trade that right instead of using it if that is
preferred. This is accomplished by distributing marketable
pollution permits, each one entitling the holder to pollute the
atmosphere in some standardized quantity. This is the "cap
and trade" policy described earlier. A factory in Los Angeles,
for example, might hold a 10-dirt permit ("dirt" being a ficti-
tious unit of pollution). If its production process generated
only 5 dirts of pollution, then it could sell the remaining
5 dirts on its permit to some other user for cash, for the prom-
ise of an 8-dirt permit five years from now, or for something
else of value. The Environmental Protection Agency, for exam-
ple would set the overall quantity of permits available at any
one time, after which a market in permits would arise.
Pollution is suddenly costly to its producer, because he or
she must now devote dirts to it that have alternative uses
(like selling or trading them). The Los Angeles factory may
now determine whether it is worth its while to retrofit its pro-
duction process so that less pollution is generated; if the cost
of retrofitting is exceeded by the sale of the pollution permits
it currently owns or would otherwise have to purchase in the
market, this move makes sense. The result, then, of this mar-
ket for pollution permits is that polluters now have incentives
to reduce their pollution and that pollution rights will ft.ow to
those that value them the most. These latter polluters are
those for whom it is cheaper to buy pollution permits than it is
to reduce their emissions.
The key, of course, and the place where politics is central,
is the determination of the aggregate amount of pollution to

be permitted on the one hand, and the initial distribution of
pollution permits on the other. The first is a straightforward
political judgment call of the sort that our political institu-
tions are char ged with making all the time. The second is po-
litical dynamite, since so much is at stake. But a s long as the
n1arket works smoothly once an initial distribution is made,
the permits ultimately will flow to their highest-valued uses.
Even the judgment call on t he aggregate amount of pollution
to permit in the first place has a certain self-correcting quality
to it. If Friends of the Earth or the Audubon Society feel s that
the political authorities have set the aggregate pollution level
too high for, say, the Los Angeles metropolitan area, then
these environmental groups can jump into the dirt market
there and buy up pollution permits. These they can either per-
manently retire or resell in some other area of the country
with an ambie nt air quality that can absorb additional emis-
s10ns.
This section on externalities can be summed up by noting
that none of the solutions I have reviewed- taxes a nd subsi-
dies, regulatory regimes, redefined property rights- are with-
out problems. In each case there are practical or logistical
complications that must be overcome. Even putting these dif-
ficulties to one side, however, there is always the problem of
politics. Once the machinery to tax and subsidize, to regulate,
or to redefine ownership is put in place, it may be used or
abused. It is absolutely essential to be aware that the problem
of externalities is transformed into a problem of providing ap-
propriate incentives to those in charge of the externality-
control apparatus. Jt is the same "who will guard the
guardians" problem that we have encountered elsewhere in
this volume, a problem I shall examine very closely in Part IV
in our treatmen t of institutions.
THE P ROBLEM OF THE COMMONS
Sitting just outside the office in which these words are being
written is the Cambridge Common, a lovely urban public
place most famous for the fact that it was there that General
George Washington mustered the twelve hundred volunteers
of what became the Continental Army in 1776. Publicly owned
parks and land reserves are today the object of great passion
by those who place significant value on "green space." In an
earlier time in Europe (and still today in various parts of the
world), commons were valued for more practical reasons-
notably as places to graze cattle and to forage. 20 Today, exam-
ples of commons include not only green space and sites for
grazing and foraging but also bodies of water utilized for com-
mercial fishing, irrigation systems, urban water supplies, and,
indeed, even the earth's atmosphere. 21
A commons is, by definition, owned by everyone (in com-
mon), and therefore is the responsibility of no one. Consider a
field owned by a village and used by its residents' herds as a
grazing commons. Each villager gets to graze his or her cattle
"for free." If a villager is contemplating adding a head to his
herd, he will take into account his costs of doing so, but this
calculation will not include the cost of grazing. If the commons
is large, and the village demands on it minimal, this will not
pose serious problems. But even if demands on the commons
grow, no villager has an incentive to restrict his use of this
"free" resource, resulting in what Garrett Hardin called "the
20 Several hundred years ago in England and elsewhere, there were political
move ments t hat succeeded in enclosing common lands, that is, dividing
them into parcels and distributing or selling them to individuals as private
property. 'fhe modern co unterpart of this practice is the sale of state-owned
assets (privatization) in both socialist and capitalist economies. '
21 The most insightful discussion of "common pool problems," of which these
are examples, is Elinor Ostrom, Governing the Commons (New York: Cam-
bridge University Press, 1990). Os trom was awa1·ded the Nobel Prize in
Economic Sciences in 2009 for this work.
tragedy of the commons." 2:l The commons will be overgrazed

and ultimately destroyed, inasmuch as its capacity to regener-
ate itself will have been disabled.
Overgrazing the commons is a metaphor for a host of prob-
lems. large and small, in which lack of restraint in using the
commons leads to social catastrophe:
• The portion of the North Atlantic off the coast of New
England is a common habitat for lobsters. Overgrazing
in this instance takes the form of too many lobstermen
harvesting too many lobsters (especially small lobsters
that haven't reproduced).
The acquifer under Cape Cod is a commons constituting
the source of fresh water for that beautiful strip of land.
Overgrazing this commons occurs because of residential
and commercial development. With more people on the
Cape, pollutants seep into the acquifer, affecting its pu-
rity. Perhaps more profoundly, with more people draw-
ing more fresh water from the acquifer, salt water
penetration from Massachusetts Bay and the Atlantic
Ocean intensifies. Ultimately, rain- and spri ng-fed re-
newal will be insufficient, and the acquifer will be
destroyed.
• The earth's atmosphere is a commons into which pollu-
tants are dumped. It is replenished by oxygen created as
a by-product of photosynthesis. The destruction of vast
forests for development simultaneously increases the
production of pollution and reduces the atmosphere's ca-
pacity to replenish itself.
The problem of the commons, like the problems associated
with cooperation, the production of public goods, and the con-
trol of externalities, is a problem of private and social incen-
n Garrett Hardin. "The Tragedy of the Commons," Science 162 (1968): 1243-
48. This now cla;;sic paper is must reading for the interested student.
tives in conflict. A commons is a free lunch to its common own-

er s. Indeed, each possessor of rights to the commons has a
very strong incentive to use those rights. Let us r eturn to the
village with a common grazing field. A hundred villagers are
each grazing two cows on the commons, a number that the
commons can support and still regenerate itself. Each villager
considers a dding one cow to his herd, concluding that this
would be profitable, especially in light of the free grazing priv-
ileges. If any one villager were to proceed, the commons would
be damaged only marginally, indeed hardly a t all, since the
herd size will only have increa sed from 200 to 201. But if all
the villagers proceed, there will be a 50 percent increase in
grazing, an a mount exceeding the carr ying capacity of the
commons. So, if all proceed, each will be worse off, since they
will have destroyed their field. But if any one villager pro-
ceeds, he will be better off and a ll the others will hardly be af-
fected at all. Thus, individual incentives and social necessity
clash. Indeed, overgrazing the commons is, in many r espects,
the large-number analog of the Prisoners' Dilemma and
Hume's mar sh-draining game that I discussed in Chapter 8.
As the reader is now undoubtedly aware, preservation of
the commons is a public good. I will not rehearse again all of
the standard methods for its provision, leaving this to the
reader as the proverbia l homework assignment. I will, how-
ever, comment briefly on two aspects of this knotty problem.
First, it is well known that commons problems arise be-
cause of imperfectly specified property rights. If a single indi-
vidual rather than an en tire community owned the commons,
or if she and her fellow villagers each owned well-defined plots
within the commons, then she would have all the incentives
associated with ownership of a private good to preserve the
value of this asset. An individual would no more overgraze her
commons, or overharvest her forest, or overfish her pond than
she would abuse a ny other physical asset she owned. The en-
closure movement in England in a n earlier era and contempo-
rary experiments with marketable permits in rights to pollute

are instances of redefining property rights, reallocating own-
ership from the community to specific individuals.
Second, as I have emphasized throughout this chapter, po-
litical arrangements affect both the solutions selected to deal
with commons problems and the likelihood of success. In her
pathbreaking study of common pool problems, Governing the
Commons, Elinor Ostrom makes very clear that humankind
has been incredibly inventive over millennia in coping with-
and, indeed, sometimes avoiding-the tragedy of the com-
mons. These coping strategies are much like the constitutions
(both written and informal) by which communities govern
themselves. They involve mechanisms by which collective de-
cisions about the use of the commons are made, monitored,
enforced, and changed in an orderly manner. Ostrom provides
instances of both successful and unsuccessful "commons con-
stitutions," emphasizing that the successful ones are those
with design features possessing:
• clearly define d boundaries;
• congruence between rules for using the commons and
local needs and conditions;
• individual rights to formulate and revise the rules for
operating the commons;
monitoring arrangements in which the monitors are ul-
timately responsible to the community;
• graduated punishments for violation of rules;
• low-cost arenas for resolving disputes; and
• relative freedom of users of the commons from external
governmental authorities. 23
In short, Ostrom has found that the management of a com-
mons is a political problem. If rights over this commons can-
not be parceled up into private bundles-the property-rights
2r1 Ostrom, Gouerning the Commons, pp. 88-102.

solution-then, to encourage the cooper ation required to pre-

serve the commons and to discourage the practice of over-
utilization, the group of users must enter into a political
agreement-a form of self-enforcing self-restraint.
CASE 10.2
FISHING AND THE
T RAGEDY OF THE COMMONS
Fishing provides an excellent subject for inquiring into the

tragedy of the commons. As we have seen, the tragedy of
the commons is. a problem when individuals share a com-
mon, depletable resource. In their efforts to maximize their
individual gains, users often overuse the resource to the
detriment of all. Oceans, Jakes, and rivers have largely
been viewed by fishermen as a commons. The result is over-
use: two-thirds of all assessed fish stocks in 1994 were ei-
ther overexploited or fully to heavily exploited, according to
the United Nations Food and Agricultural Association.*
Overuse remains a problem today.
Governments have responded to the problem with a vari-
ety of approaches. Their responses provide evidence for Os-
trom's theory that a commons can be managed through the
allocation of proper ty rights or the evolution of self-enforcing
restraints. Several nations, including Iceland and New Zea-
land, have address.ed the problem of overfishing by allocat-
ing property rights. Their system of individual transferable
quotas (ITQs) divides up the catch within national waters
among commercial fishermen. Quota owners can either keep
or sell their fishing rights. The government sets the total
quota to maintain and conserve the resource over time.
*Mark Trumball, "Fisheries Crisis Stretches across the Globe," Christi-On

Science Monitor (July 6, 1994), p. 8.
----
There is considerable debate over the use of ITQs. l\llany

organizations and governmental authorities are seeking al-
ternative mechanisms. In New England, the government,
the fishermen, and the fisheries worked to address the
sharp decline in fish catches over the course of the 20th cen-
tury.t The importance of politics is apparent in the case of
New England fishing. Government programs and interest
group politics accelerated the overutilization of the com-
mons. In 1977, the United States banned foreign trawlers
from fishing within 200 miles of the U.S. coastline, partly to
prevent overfishing of New England waters. But a federal
loan program at the end of the 1970s and the beginning of
the 1980s led to a boom in domestic boat construction. Fish-
ermen organized to lobby against catch quotas and fishing
limits. In 1982 the New England Fishery Management
Council gave in to the pressure from fishermen and dropped
the quotas and limits. The result was a rapid increase in
overuse, and fish catches declined by over a third in four
years. The decline led many fishermen to see the connection
between their individual behavior and their collective fate.
The vice president of the Atlantic Offshore Fish Association
in Newport, Rhode Island, remarked, ''I used to be strongly
opposed to any kind of limited entry in fisheries. But I've
come to feel we have to have some way of rationally allocat-
ing fishery resources just as we do other resources.''
What happens when fishery resources can't be "ration-
ally allocated'"? In other words, what happens when the re-
source is not conducive to assignment of property rights?
Ostrom predicts that users will enter into a political agree-
ment involving self-enforcing self-restraint. We find sup-
port for this prediction in the case of "straddling stocks,"
fish that migrate between national and international wa-
1 Lawrence Ingrassia, "Overfishing Threatens to Wipe Out Species and

Crush Industry." Wall Street Journal (July 16, 1991). p. l.
ters. Straddling stocks account for abou t one-fifth of the fish

caught around the world each year. Their migration be-
tween national and international waters prevents countries
from declaring ownership of the stocks and assigning prop-
erty rights through fishing quotas. Countries have been
forced to work cooperatively on the problem. The United
Nations has been used as a forum for creating agreement
and addressing issues of monitoring and enforcement, just
as Ostrom's theory would predict.
The importance of enforcement can be seen in a "natural
experiment'' created by the fall of the Soviet Union.~ Ninety
percent of the world's sturgeon stocks are in the Caspian
Sea, which is bordered by Iran and the former Soviet Union.
For decades, the harvest of caviar from spawning sturgeon
was tightly regulated by the Soviet government. Quotas for
the annual sturgeon catch were established by the Ministry
of Fisheries in Moscow and enforced by armed inspectors
who kept the lid on poachers and illegal dealers. In 1992, the
birth of four new independent states and two new au-
tonomous regions along the spawning grounds of the stur-
geon, together with the breakdown of the chain of command
out of the Kremlin, led to a nlarked decline in the enforce-
ment of these quotas. The result was a rapid increase in
sturgeon fishing. The director of the Fish eries Resear ch In-
stitute in Astrakkan, at the mouth of the Volga River, said,
"Central authority has disappeared. People are living by the
law today: Catch whatever you can and don't care about to-
morrow. If things are allowed to go on like this, within three
to five years sturgeon stocks will be completely depleted."
The importance of reaching an agreement on how to use
the sturgeon resource was emphasized by Moscow's chief
fisheries inspector: "Either we agree on rules for catching
*·MichaelDobbs, "A Warning by the Sturgeon General," Washington Post

National Weekly Edition (June 8-14, 1992), p. 1.
sturgeon or we simply destroy the fish altogether. If we can

reach agreement with the United States on limiting produc-
tion of nuclear missiles, surely we can reach an agreement
with other [former Soviet] republics on catching sturgeon."
The reader may want to ruminate on this last remark.
Knowing what you now know about cooperation, collective
action, and problems of commons, is it really as easy to cut
a deal on sturgeon among various new states and au-
tonomous regions as it is for two superpowers to sign a bi-
lateral agreement?
CONCLUSION
The problems we have confronted in this chapter are, in many

respects, those we met in the previous two chapters. Too little
cooperation, too little collective action, too few public goods,
too many negative and too few positive externalities, and too
much use of common resources are all social dilemmas in
which individual incentives are in conflict with socially desir-
able outcomes. These are summarized in Display 10.2.
The problem of cooperation, as exemplified by the marsh-
draining game of Chapter 8, is one that pits the joint benefits
of cooperation against the individual motives to defect (since
"do not cooperate" is the individually advantageous option
whether the other guy cooperates or not). The problem of col-
lective action is the problem of coopers tion writ large, where
defection takes the form of free-riding on the effort of others.
This, in turn, is directly analogous to "not contributing" to the
provision of a public good; to defect, in this interpretation, is
to withhold payment for a public good, since, if it is provided,
noncontributors cannot be prevented from enjoying it. Like-
wise, paying no attention to the (positive or negative) external
effects of your actions is a bit like not controlling your produc-
tion of public "bads" or ignoring your production of public
DISPLAY 10.2
COOPERATION, COLLECTIVE ACTION, PUBLIC GOODS
SUPPLY, EXTERNALITY CONTROL, COMMONS
GOVERNANCE: COMPARIS ONS
Behavior to Be
Problem Controlled Illustrative Solu t ion
Cooperation Defection Repeat play
Collective Action Free-riding By-products,
political entrepreneur
Public Good Supply Noncontribution Public provision
Externality Control Inattention to Tax/subsidy scheme
external effects
Commons Governance Overutilization Property rights
regime, gover-
nance structure
goods; each is a by-product of your actions for which you shun

responsibility. Finally, overutilizing a common resource is
''antisocial" in the sense that this action fails to take account
of the damage your actions wreak on others.
For each of t hese social dilemmas, numerous solutions are
advocated (a representative one of which is listed in the last
column of Display 10.2), and a variety of human experience
with all of them. Rarely are the solutions, even those that work
tolerably well, ideal. (If any solution were, then we wouldn't be
spending so much time writing about them.) One thing is clear,
and bears repeating one last time. Solutions are political, both
in their advocacy and in their implementation. To understand
why they work or why they fail, the observer must come to
terms with the political ambitions and motives of the actors in-
volved, and with the institutional contexts in which these am-
bitions and motives get played out. In the concluding section of
this text, I turn to an analysis of political institutions.
EXPERIMENTAL C ORNER
Punishing Free-R iders
Public goods, as we have seen, are underproduced because
individual incentives to contribute to their production are
weak. Especially in large anonymous groups, individuals
are strongly tempted to free-ride on the contributions of
others, since they can enjoy whatever public good is pro-
duced while avoiding any of its costs. Indeed, if the group is
large enough and anonymous enough, their failure to con-
tribute will often go unnoticed. Of course, they risk being
punished for their antisocial behavior, but then punishment
itself is costly for others so it is unlikely to be much of a
detriment to free-riding. But what if, despite its cost, people
did punish "irresponsible" behavior?
In a wonderful experiment, Fehr and Gachter demon-
strate that "free riding generally causes very strong nega-
tive emotions among cooperators and that there is a
widespread willingness to punish the free riders ... even if
punishment is costly and does not provide any material
benefits for the punisher."a Moreover, they show that the
punishment increases in severity the more the free riding
deviates from cooperative levels, so that the opportunity
and inclination to punish will have the effect of diminishing
free riding.b
Their public goods experiment has a no-punishment and
a punishment treatment. In the former, theory tells us that
there should be widespread free-riding, possibly undermin-
• Ernst Fehr and Simon Gachter, "Cooperation and Punishment in Public

Goods Experiments,.. American Economic Review 90 (2000): 980-94. Quo-
tation is on p. 980.
b A closely related experiment that inspired Fehr and Gachter deals with
the problem of the commons described in the text. See Elinor Ostrom,
James Walker, and Roy Gardner, "Covenants with and without the
Sword: Self.Governance Is Possible," American Political Science Review
86 (1992): 404-17.
ing the provision of any of the public good. In the punish-

ment treatment, if punishing is costly, then again we
should expect massive free-riding. The possibility of pun-
ishment, so the argument goes, is irrelevant because no po-
tential free-rider will expect anyone to engage in this costly
behavior. Thus, the theoretical expectation is that the mix
of cooperation and free-riding should be the same in each
treatment. Surprisingly (from a theoretical perspective, at
least), they are not!
Fehr and Gachter's experiment, described below, is
quite subtle. They note that in repeated interactions it
might be argued that engaging in costly punishment now
may have material payoffs later-that is, that someone
could rationally develop a reputation for being willing to
punish so as to induce cooperation over the long run. I
noted in Chapter 8 that the prospect of punishment in in-
definite repeat play of the Prisoners' Dilemma game can
induce a positive level of cooperation through reputation-
building. Fehr and Gachter, therefore, are careful to remove
this material incentive in their experimental design. They
do this by having a stranger treatment and a partner treat-
ment. In the former, each subject plays a public goods game
several times with different, randomly selected, subjects; in
the latter, he or she plays several times with the same set of
subjects.
The experimental design thus has four treatments-the
stranger treatment with and without punishment oppor tu-
nities and the partner treatment with and without punish-
ment. In the partner treatments, it is a certainty that a
subject will be matched with the sam e other players. In the
stranger treatments, the probability is less than 5 percent
that an individual will be matched with someone with
whom he or she was matched in an earlier round.
In each round of the no-punishment treatment of the
game, a player is provided with y tokens and makes a
choice of how many tokens, gi for the i 1h player, to invest in

a public project, where 0 s C; s y. These choices are made si-
multaneously by the players, and the payoff for Ms. i is n; =
y - g; + a Ij Cp where 0 < a < 1. That is, in each period, Ms.
i pays in gi token from her endowment of y tokens but gets
back a proportion a of the total contributions made. In this
circumstance it is easy to see that Ms. i's dominant strategy
is to set gi - 0, thereby free-riding on the contributions of
others. Why? Because with a < 1 for each unit of contribu-
tion she makes, she gets only a units in return from her
own investments.
If punishment is available, then the game is the same as
in the preceding paragraph, with the following twist. At the
end of a round the experimenter announces individual con-
tributions. Subject j can punish subject i by assigning her
punishment points, P/- For each punishment point assigned
to i, her payoff, 1t;, is reduced by 10 percent (but not below
zero). The cost of punishment, c(p/). is strictly increasing in
P/. So, the revised payoff for Ms i is n;R - n; [1 - (1110) I.j~i P/1
- c(I.;..j p/)- that is, her original payoff reduced by 10 per-
cent of however many punishment points she receives and
also by the cost of her assigning punishment points to oth-
ers.
The experime ntal sessions are run in a computer labora-
tory. Subjects anonymously interact with each other, know-
ing that it will be the same set of people over multiple
sessions in the partners treatment and randomly selected
individuals each session in the strangers treatment. Sub-
jects were randomly assigned to the four treatment cate-
gories. The experimenter s looked at different parameter
settings for initial endowment (y), return from total contri-
butions (a), and the cost of punishing (c(pj~). All of these
details may be found in the original article.
What transpir ed? The results are quite dramatic. Even
in the strangers treatment, where an individual's reputa-
tion could not be developed because his or her cosubjects

changed each session, contributions to the public project
rose dramatically when the opportunity for punishment ex-
isted. Given an initial endowment of twenty tokens, the
mean contribution level in the no-punishment treatment is
3. 7, a fairly paltry amount. In the punishment treatment,
in contrast, the mean contribution level was 11.5, more
than triple the no-punishment level. Even in the last :pe-
riod, where everyone knew t here could be no future benefit
from punishing now, the no-punishment mean was 1. 9 to-
kens and the punishment mean was 12.3. 'This result quite
forcefully rejects t he hypothesis that there should be no dif-
ference (in the strangers treatment) between a punishment
and no-punishment condition. (1'he existence of punish-
ment opportunities in the partners setting also produces
much higher contribution levels than in the no-punishment
treatment.) Interestingly, comparing the partners and
strangers settings under punishment, in both cases contri-
butions are high and remain high, but in the partners set-
ting the contribution level approaches full cooperation, with
an average contribution of 17.0.
Fehr and Gachter were quite careful to remove, as much
as possible,, the possibility of any material gain from pun-
ishing. Especially in the strangers condition, where punish-
ing someone this time would not affect a subject's prospects
in the next round, costly punishment nevertheless oc-
curred. They conclude that even highly rational people have
a consequential emotional life in which there is a strong
propensity to punish others for inappropriate behavior,
even if the punisher must bear a cost to do so. Thus, as they
note, we observe drivers expressing their rage when some-
one butts into line, striking workers conveying strong dis-
approval of strike-breaking "scabs," students shunning
fellow dorm residents for not cleaning up after themselves
in common rooms, and so on.
At least two puzzles generated by these experiments are

worth pondering. First, why do people come, possibly hard-
wired, with the emotional responses leading them to punish
as described above? Might an evolutionary story of some
sort help explain this? The second puzzle is why people
don't free-ride on punishing- that is, let someone else pun-
ish a transgressor? If Ms. i is offended by Mr. j's inappropri-
ate behavior, why doesn't she avoid punishment costs by
letting k, l, and m do the punishing? The answer to both
puzzles .may reside in the neurobiological prospect that we
take emotional pleasure in expressing our disapproval of
those we believe violate expectations of cooperation- that
part of our identity is tied up in acting on these emotional
impulses.
PROBLEMS AND D ISCUSSION Q UESTIONS
l. Write out an empty 2x2 chart like that in Display 10.1, and
then for each square develop an original illustrative example
of the type of good, explaining why your example is either ex-
cludable (or not) or jointly supplied (or not). Then, explain
what "problems" arise from nonexcludability and from joint-
ness of supply (or lack thereof!).
2. The website You'fube.com hosted approximately 75 billion

videos in 2009, charging both uploaders and viewers nothing
for the service a nd placing no restrictions on the quantity of
content that it would host (although it does enforce some lim-
its on inappropriate content). You'fube expends hundreds of
millions of dollars on bandwidth, as well as infrastructure
and maintenance costs, and earns money primarily by hosting

ads on its pages. 'lb date, the company has not turned a
profit.a
Identify what type of good You'fube is offering based on the
categories outlined in Display 10.1, explaining carefully how
you reached your conclusion . Is it possible to supply profitably
the type of good you have identified? Are there alternative
modes of delivery that You'fube could adopt to deliver the
same service more profitably?
*3. Five civic-minded patrons of a public library contemplate

donations of time to its annual fundraiser. Each individual i
bases his or her decision of how much time to donate, x,, on
the following utility function:
U,(x) - q·25 - .25x,
where q - .f x . (the total amount of time given by all library pa·
J•l J
trons) and .25xi is the cost of losing xi of one's leisure time.
a. What is the socially optimal amount of time donated?

Determine this by summing the utility functions of five
individuals, and finding the q that maximizes this
function.
b. Now consider the case of an individual who assumes
that everyone else will contribute no time. How much
time will this individual donate, and is it at the socially
optimal level?
c. Now assume that each individual assumes that the
other four members will donate .8 units of time. Does
this individual donate more or less time than before?
Why is this?
• Malcolm Gladwell , "Priced to Sell," The New Yorker, July 6, 2009.

d. Why is a socially suboptimal amount of time donated in

parts b and c? Consider the balance between internal-
ized costs and internalized benefits.
e. [Bonus] Repeat the same exercise for the case of ten in-
dividuals. What proportion of the socially optimal
amount is donated in an equilibrium now? What does
this suggest about the difficulties of supplying quas1-
public goods as the number of people increases?
4. Perhaps the default solution for the provision of public

goods is public supply- by the state. Does public supply hap-
pen automatically where markets fail to provide public goods?
Explain your reasoning. If not, outline some of the ways in
which public supply is secured. What are some of the prob-
lems that recur in the state supply of public goods?
*5. A factory located in a small village produces a good with

increasing marginal costs, given by equation M C(q) = 12 + q,
so, for example, the first unit costs 13 to produce, the second
14, and so on. This firm can produce at most 15 units and can-
not produce fractional amounts (e.g., 1.5 units). The market
price for that good is currently at p = $20, and the firm's level
of production does not affect this price. Assume that the fac-
tory owner maximizes profit, and her utility is measured in
dollar terms. Profit is caJculated by summing up the differ-
ences between the price and the marginal cost of each unit
produced, so the fi rst unit earns a profit of 7, the second of 6,
and so on.
Unfortunately, it turns out that the factory is quite noisy and

interferes with the practice of a neighboring doctor. In fact, for
every extra unit produced by the factory the doctor loses $2
worth of profits. Just as for the factory owner, assume the doc-
tor's welfare depen ds only on his profits, which are 50 - 2q.
How many units of the good will the factory produce if it

ignores the externality imposed on the doctor in its
profit ma:·dmization? What will be the aggregate social
utility (sum the factory's and doctor's total profits)?
• Identify the level of production that is socially most pre-
ferred, that is, maximizes aggregate social utility.
• Propose a government taxation scheme that will lead to
the socially preferred outcome.
6. Taxes/subsidies, regulation, and respecification of property

rights are three canonical solutions to the problem of exter nal-
ities. Explain th e thinking behind each of these solutions and
discuss any difficulties with the implementation of such a so-
lution. Then, provide an example, real or of your own devising,
of each of these solutions applied to situations where external-
ities arise.
7. One interesting example of a positive externality is so-

called network effects: each additional user of some good or
service increases the value of th at good or service to all other
users. The New York Stock Exchange, telephones when they
were first introduced, and Facebook are often cited as exam-
ples. Using one of these examples, or one of your own, explain
why network effects fit the definition of an externality. Do net-
works then suffer from the problem of underprovision, as with
other positive externalities? at what stage in their develop-
ment? What are some possible solutions to overcoming these
problems?
Part IV
INSTITUTI0 NS
11
Institutions: General Remarks
I spent a good deal of time in Part III on the circumstances

and conditions in which political actors (who may be no more
than ordinary citizens) engage in cooperation and collective
action in order to solve problems related to public goods, ex-
ternalities, and the overutilization of commons. We would not
expect a political community to devote much effort to the solu-
tion of intermittent and relatively unimportant problems.
Most likely these problems would be dealt with in an ad hoc
fashion- sometimes successfully, other times not. A small
town in Mississippi, for example, would be unlikely to devote
much energy or many resources to the public good of snow re-
moval, even though once every other decade the town is shut
down by a freak snowstorm. A political community is much
less likely to treat recurring, consequential problems in an ad
hoc manner. Instead, it develops routines-standard ways of
doing things by organizations endowed with resources and au-
thority. Thus, the city of Cambridge, Massachusetts, has a De-
partment of Public Works with authority and an annual
budget for snow removal. In a word, r~sponses to regularly re-
curring problems are often institutionalized. Collective action
comes to pass in the political community because standard
procedures are established that provide political actors with
appropriate incentives to take the action necessary to provide
a public good or control an externality.
In the next several chapters I focus on institutions as
355
repositories of authority and resources to solve such problems.

I emphasize official institutions-executives, legislatures, bu-
reaucracies, courts-though I hasten to add that institutions
like these are not encountered only in the public sphere. De-
spite the fact that political scientists tend mostly to study
these public institutions, the form of analysis that is elabo-
rated below is applicable to a wide array of private-sector
organizations-families, firms, houses of worship, universi-
ties, charitable organizations, unions, and so on. Each of these
is a "political community" in its own right and institutional-
izes procedures to deal with recurring, important problems it
faces. The institutionalization of politics arises not only in
Washington, D.C., but also around the dinner table, in the
board room, in the College of Arts and Sciences, and in the
union hall. Politics is omnipresent in human society, and its
more routine aspects are dealt with by institutions.
The institutionalization of aspects of political life, re-
served, as I noted, mainly for recurrent, significant problems,
has important consequences. Political performance is now no
longer exclusively a function of the "quality" of political actors.
Cooperation in a political community, for example, does not
now depend only on the good fortune of that community to
have cooperatively inclined political agents in place. Collective
action, for another, is not held hostage to altruistic inclina-
tions or the charitable dispositions of its citizens. Rather, even
if "men are not angels," as Madison once observed, the meth-
ods for doing political business may have a salutary effect.
This was noted in the eighteenth century by the philosopher
David Hume. He observed that people
ought not to trust the future government of a state entirely to
chance, but ought to provide a system of laws [by which he means
institutions] to regulate the administration of puhlic affairs to the
latest posterity.... I W]ise regulations in any commonwealth are
the most valuable legacy that can be left to future ages. In the
Institutions: General Remarks 357
smallest court or office, the stated forms and methods [institu-

tions again], by which business must be conducted, are found to
be a considerable check on the natural depravity of mankind.
He goes on to comment that it may be commonly observed in

the history of governments that "one part of the same [govern-
ment] may be wisely conducted, and another weakly, by the
very same men, merely on account of the difference of the
forms and institutions, by which these parts are regulated." 1
The take-home message of Part IV of this book, emphasized
by Hume and emphasized again here, is that institutions, not
just individual preferences, matter for collective results.
A F RAMEWORK FOR S TUDYING I NSTITUTIONS

Before turning to specific institutions in subsequent chapters,
it will be useful to think about these matter s a bit more gener-
ally. While I want to convey loosely what I mean when I talk
about institutions, I do not want to get hung up on definitions.
(Students should put their highlighters away!) The framework
I elaborate consists of four components: division of labor and
regular procedure; specialization of labor; jurisdictions; and
delegation and monitoring.
Division of Labor and Regular Procedure

The tasks confronting actors in an institutional setting often
lend themselves to preliminary structuring. For example, any-
one who has ever attended a meeting of a club knows that its
proceedings are divided into a number of categories-officer
1 DavidHume, Essays (Indianapolis: Liberty Fund, 1985), Essay III. My

thanks to Richard Tuck for bringing this passage to my attention.
reports, old business, new business, and so on. A school board

seeking community input may begin its sessions with a "citi-
zens speak" slot, in which any resident may raise a school-
related issue. The presiding officer then assigns the topic to
the agenda of a subsequent board meeting. The U.S. House of
Representatives has a well-structured division of labor, which
is reflected in the various categories of its "calendar." Time is
allotted for representatives to make remarks on subjects near
and dear to their constituents, for committees to meet and
make reports, for debates and votes to take place, for legisla-
tion to be received from the Senate, and for messages to be de-
livered by the president.
There is nothing very surprising in all this. Actors meeting
regularly in an institutional context evolve procedures by
which the various bits of business are divided into manage-
able units and then sequenced into a specific order. They do
so, most obviously, to bring order to their deliberations. Econ-
omists would remark that this aspect of the division of labor is
an efficiency-enhancing aspect of an institution. Without some
procedural regularity, the proceedings would be more chaotic
than they need to be. Dividing up the business and proceeding
according to an institutional script economizes on the costs of
doing business (or what economists call transaction costs).
There are, however, additional purposes served by this di-
vision of labor. Procedures governing a division of labor not
only regularize a group's official affairs but also allow its indi-
vidual members to plan their own participation. This is be-
cause a set of procedures and a division of labor define a
strategic context in which individuals may think about
whether to participate at all, in what parts of the sequence to
participate, and, finally, precisely how to participate. More-
over, they may condition these choices on what will already
have transpired, as well as on what they know is coming up. A
member's preferences about what new undertaking the organ-
ization should pursue, for example, may well depend on how

the organization resolves some current undertaking. Thus, it
is necessary to dispatch "old business" before considering new
departures. An organization that allows deliberation and deci-
sion on "new business" before this old business is resolved
would prevent its individual members from conditioning their
behavior on the latter's resolution. Members would then have
to make judgments with insufficient information, raising the
prospect that the issue would have to be revisited (wasting
members' time) once the relevant information is available. In
this sense, procedures embodying a division of labor are strat-
egy enhancing.
Another purpose served by the division of labor and regu-
lar procedure is the empowerment of the organization's mem-
bers . Reliable and publicly known procedures provide a check
against arbitrary and capricious behavior by institutional
leaders. Leaders, as we will see a t several points in the next
few chapters, have considerable discretion that they can direct
toward their own ends. Partly this is a reward to leaders-a
selective incentive, in Mancur Olson's language-for shoulder-
ing organizational burdens. This reward, however, becomes a
blank check if leaders are not held responsible for observing
a modicum of regularity in the ways they proceed. Official pro-
cedures provide this regularity.
Finally, I s hould note that a particular divison of labor
or set of procedures is not carved in stone. There are, in the
constitutions, bylaws, or standing rules and orders of most in-
stitutions, mechanisms by which to change the way the insti-
tution's business is conducted. Two categories of departure
from regular procedure arise, depending upon whether the
change is permanent or temporary. Standard operating proce-
dures may be amended, so that some particular piece of busi-
ness, and al] subsequent business of a similar sort, is
conducted in a different manner. Or procedures may be tern-
porarily suspended to handle a specific matter, after which the

original procedure is rei nstated. The former represents a per-
manent change in the institutional status quo, assumed to
apply to future proceedings unless subsequently amended.
The latter is a "short-circuiting" device by which some specific
piece of business is conducted in a different manner, but the
rules governing future business of the same sort revert to the
earlier institutional status quo. Each of these techniques has
been employed by the two houses of Congress a nd is illus-
trated by a specific example.
Until the mid-1970s, a motion to close debate and bring a
matter to a vote in t he U.S. Senate, called cloture, required
the support of "two-thirds of those present and voting." This
rule empowered a minority of 34 senators (sometimes fewer if
fewer than 100 senators were "present and voting'') to fili-
buster, thereby preventing the Senate from bringing a motion
to a vote. The standing rules of the Senate were amended at
that time-mainly to break the stranglehold of southern sena-
tors preventing the Senate from voting on civil rights bills-so
that cloture could be voted by 60 senators. 2 This rule change
became part of the standing rules governing future Senate
business.
In contrast, the House of Repr esentatives often operates in
suspension-of-the-rules mode in order to avoid procedural com-
2
The Senate has always prided itself on being "the world's greatest delibera-
tive body," and some defenders of the earlier rule on cloture believed that it
faci litated debate and deliberation by preventing the body from taking a
vote until an extraordinary majority felt there was nothing further to delib·
erate on or debat.e. Detractors, on the other hand, felt that the old cloture
rule was simply a procedural device behind which a disciplined minority
could extract concessions from the majority, if not defeat the majority's objec·
tives altogether. During the civil rights revolution, detractors had the votes
to pull ofJ the procedural coup described in the text (though southerners ac-
tually filibustered duri ng the debate on whether to change the cloture rule!).
The House limited debate in its formal rules more than a century earlier,
corning to the conclusion that no large gro up of politicians could be expected
to keep quiet long enough to take a vote unless forced to by institutional pro-
cedures.
plexities that would make it difficult for the House to complete

its business. The rules of the House are contained in a rather
thick volume, and the precedents and interpretations about
how to apply these rules now comprise a library shelf full of
thick volumes. If the House actually had to follow its rules to
the letter each time it took up even the most trivial of matters,
it would give gridlock an altogether new meaning, making the
operation of the present House look sleek and efficient by com- ·
parison. In order to keep the rules intact-for there actually
are circumstances in which it is important for all the hoops to
be jumped through-yet facilitate the expeditious prosecution
of the people's business, it is always in order for a me mber of
Congress to move to suspend the rules and proceed directly to
a vote on the matter at hand without further delay. A com-
bined motion both to suspend the rules and to pass the bill be-
fore the House is approved if supported by two-thirds of those
present and voting; ordinary procedures to pass the bill would
only require a simple majority. So an extraordinary majority
may expedite business by temporarily bypassing standard op-
erating procedures, but this maneuver is successful only for
relatively uncontroversial legislation. Moreover, and in con-
trast to amending the rules, the suspension is only t emporary,
with procedure reverting to its standard form in subsequent
deliberations.
Specialization of Labor
In an institution, each actor is not a perfect copy of any other
actor. Instead, actors tend to have different talents, interests,
and preferences. Many institutions reflect this fact and make
good use of it, by specializing labor. Some professors, there-
fore, spend disproportionate amounts of time in the classroom;
others spend large amounts of time in the lab or library. Some
legislators specialize in particular policy areas, others special-
ize in different areas of policy, while still others remain gener-
alists. 3 By specializing labor in the sense of allowing different

members to do different things, in light of their different in-
terests and talents, an institution is able to capitalize on the
rich endowment of "human capital" contained in its member-
ship.
Not all institutions embrace specialization. Perhaps the
most prominent American national institution lacking a spe-
cialization of labor is the Supreme Court. All justices hear all
oral arguments, attend conferences with their colleagues,
write draft opinions, engage in deliberation and bargaining,
and ultimately sign a majority opinion or draft a concurring or
dissenting opinion. Unlike legislators, who typically serve on
a small number of committees handling specialized areas of
public policy, and thus who devote the lion's share of their en-
ergies to these special domains, justices are generalists par
excellence.4 • 5
It is not unusual for an institution to evolve from a rela-
tively unspecialized form to a more specialized form. In the
3
A wag (probably a member of the House of Representatives) once asked:
'What do a U.S. Senator and the mouth of the North Platte River have in
common? Both are a mile wide and an inch deep." Senators, because of t heir
larger, more heterogeneous constituencies, normally cannot afford to special·
ize in terms of policy as much as representatives can. They remain general·
ists rather than specialists (or, as the wag would have it, superficially
informed about nearly everything).
4
This is a slight (but only slight) overstatement. Justices tend to develop rep-
utations, owing to past training and personal interest, as experts in specific
areas of law. This expertise is reflected in the fact that specific justices tend
to pay extraordinary attention to specific areas-some focusing on capital
punishment cases, others on regulation, still others on individual liberties.
But each justice devotes time to every case that comes before the Court and
is not prepared to deifer even to expert colleagues. To our knowledge. satis-
factory explanations for the lack of specialization in the Supreme Court have
never been provided. Nevertheless. for a fascinating description of how the
Supreme Court operates as a relatively unspecialized political institution,
see H. W. Perry, Deciding to Decide (Cambridge: Harvard University Press,
1991).
5
This lack of specialization on the Court may reflect a deeper prohibition
against "trading across issues." The administration of justice, it may be be-
lieved, is compromised if it is the result of logrolling. I thank a reader, Jay
Hamilton, for this interpretation.
marketplace, for example, production often begins as an arti-

sanal activity or a household industr y in which a producer
makes the entire product and is paid according to the number
(and quality) of units produced. 6 This is known as the piece-
rate method of compensation, in contrast to the wage system,
in which compensation is in accord with a specific activity and
the amount of time spent on it. The latter system is often as-
sociated with mass production, in which each worker devotes
himself or herself to a specific part of the production process
in an assembly-line arrangement. Accordingly, in mass pro-
duction labor becomes specialized, allowing for the acquisition
of expertise that makes the production process more efficient.
The same evolutionary pattern is observed in the history of
the U .S. Congress, as well as in most of the state legislatures.
They began as relatively unspecialized assemblies with each
legislator (essentially like a contemporary Supreme Court jus-
tice) participating equally in every step of the legislative
process in all r·ealms of policy. This was certainly true for the
U.S. Congress in the first decade or so after the adoption of
the Constitution. By the time of the War of 1812, if not earlier,
Congress began employing a system of specialists (the stand-
ing com mittee system), as members with different interests
and talents wished to play disproportionate roles in some
areas of policymaking while ceding influence in other areas in
which they were less interested. 7
The great advantage of the specialization of labor is that
an institution benefits from the acquisition of expertise by its
6
The image ordinarily conjured up is of some nineteenth-century cottage in-
dustry, like weaving, in which the artisan is paid for each finished article of
clothing produced. Late-twentieth-century high-tech analogs include the fa .
mous story of Steve Jobs and Steve Wozniak assembling the first machine of
what was to become Apple Computer in Jobs's family's garage.
7
The story of the evolution of the standing committee system in the House
and Senate in the early part of the nineteenth century is told in Gerald
Gam m and Kenneth Shepsle, "Emergence of Legislative Institutions: Stand-
ing Committees in the House and Senate, 1810-1825," Legislative Studies
Quarterly 14 (198-9): 39-66.
members. By divvying up activities, giving disproportionate

influence in different areas to different members of the insti-
tution, incentives are created for members to "learn their
area," developing specialized knowledge and accumulating
relevant current information. This expertise has the prospect
of making the institution more effective in whatever activities
it pursues.
Jurisdictions
The incentive for those with special interests or talents actu-
ally to specialize in those activities is facilitated by parceling
activities into jurisdictions. Each jurisdiction is a bundle of ac-
tivities, and members of the institution assigned to a specific
jurisdiction become jurisdictional specialists. Universities,
firms, and legislatures are institutions that manifest this or-
ganizational feature most clearly.
In universities, for instance, the world of knowledge is di-
vided into disciplines, each of which is assigned to the juris-
diction of a department. The students of that department
specialize in mastering the knowledge of that discipline, while
the professors specialize in creating new knowledge in that
field and teaching it to others.
In exactly the same manner, corporations partition their ac-
tivities; indeed, there are alternative ways to bundle activities.
For example, firms may be organized in a functional manner,
with separate divisions devoted to separate functions- finance,
advertising, production, distribution, marketing, research and
development, and so on. Or firms may be organized according
to product line, with each division's jurisdiction consisting of
all the activities associated with developing, producing, and
bringing to market a specific product. In either case, members
of this institution are assigned a jurisdictional niche in which
their activities are specialized.
Legislatures, too, are organized into specialized jurisdic-
tions. The world of policy is partitioned into policy jurisdictions
that become the responsibility of committees. The members of

the Committee on Armed Services, for example, become spe-
cialists in all aspects of military affairs, the subject matter
defining their committee's jurisdiction. Committee members
tend to have disproportionate influence in their respective ju-
risdictions, not only because they have become the most know-
ledgeable members of the legislature in that area of policy, but
also because they are given the opportunity to exercise vari-
ous forms of agenda power-a subject I will develop further in
the next chapter.
An interesting consequence of the division and specializa-
tion of labor, institutionalized into jurisdictional arrange-
ments, is that members of an institution often become so
highly specialized that they know little about their institution
outside their own bailiwick. A Harvard political scientist typi-
cally knows mor e about what is happening in the Department
of Political Science at Stanford than she knows about develop-
ments in Harvard's Department of Chemistry. Likewise, a
congressman serving on the House Agriculture Committee
keeps abreast of political developments in his own committee
and perhaps those of the Senate Agriculture Committee, while
knowing little (and caring less) about the goings-on in the
House Committee on International Relations (except on mat-
ters of agricultural trade).
I n discussing these abstract features of institutions, I am
claiming that organizations arrange their procedures and
structures in specific ways, the most notable characteristics of
which are jurisdictional division and specialization. Some of
the push to do things in these ways comes from the outside
environment:
• Firms that organize themselves in ways that entail un-

necessary costs or other inefficiencies find it difficult to
survive in the competitive struggle.
Legislatures that fail to take advantage of specialization
find themselves outmaneuvered at every turn by the ex-

ecutive branch.
Universities that fail to foster the specialized discipli-
nary development of their faculty lose both professors
and students to those universities that do.
External pressures, however, are not the only conditioning

factors. Institutions-self-governing ones, at least- are to
some degree creatures of their members and thus are subject
to internal pressures as well. Even if Stanford did not exist,
Harvard would still be disposed to encourage the disciplinary
development of its faculty, because that's the way its faculty
wants things to be. Members of Congress want an organiza-
tional arrangement in which they can gravitate to, and have
disproportionate influence in, those areas of policy that are
most central to their own interests and those of their con-
stituency. The institutional profile that results, therefore, bal-
ances requirements of the external environment against the
desires of internal institutional actors.
Delegation and Monitoring
Dividing up institutional activities among jurisdictions, thus
encouraging participants to specialize, has its advantages.
But it has costs, too. If the research and development division
of a firm could do whatever it wanted in its jurisdiction, then
there is no guarantee that it would put its specialized talents
to work for the best interests of the entire firm. If the political
science department of a university could pursue whatever ac-
tivities its faculty members desired, then some central n1is-
sions of the university (advising undergraduates, for example)
might go wanting. If the Armed Services Committee of the
House of Representatives had no restraints, its members
would undoubtedly shower their own districts with military
facilities and contracts. I n short, the delegation of authority
and resources to specialist subunits exploits the advantages of
the division and specialization of labor but risks jeopardizing

collective objectives of the group as a whole.
The monitoring of subunit activities, thus, goes hand in
hand with delegation. Firms, for example, institute financial
controls to make sure that subunits devote resources to activ-
ities approved of by central management. The dean of faculty,
partly through resource allocation and partly through the
control of hiring and firing, influences the missions his or her
faculty pursues. A political majority of the full legislature, con-
trolling the assignment of members to committees, the assign-
ment of policy areas to a committee's jurisdiction, and the
final disposition of any legislation the committee approves,
keeps committees from pursuing a private agenda at the ex-
pense of the larger institution.
In sum, the division and specialization of labor, underwrit-
ten by the creation of jurisdiction-specific subunits, allows an
institution to decentralize its operations. This, in turn, facili-
tates the delegation of authority and resources to specialists
who, because they have disproportionate influence over events
in their r espective bailiwicks, also have incentives to develop
their expertise further. The very act of delegating, however,
generates a problem of control in which specialists may have
opportunities to pursue private objectives at odds with the
public purposes of the institution (what organization theorists
refer to as the moral hazard problem). It is probably impossi-
ble to solve this problem entirely, but institutions normally
institute mechanisms both to monitor subunit performance
and to control behavior wildly at odds with institutional
objectives. 8
8
I have made much of the phenomenon in which a parent body delegates
authority and resources to its subunits because this organizational format is
so pervasive in the real world. Presumably it reflects a gain in the capacity
of an organization to achieve its goals, relative to what it would otherwise
achieve in a more centralized format. Delegation and decentralization, it
should be noted, are instances of a more general principal-agent relation-
ship. I do not have t ime to go into this in detail. Suffice it to say that when-
CONCLUSION
This brief introduction to the world of institutions has been

intended to emphasize that, while institutional arrangements
come in different sizes and shapes, there is a common set of
considerations by which they may be analyzed and under-
stood. Without giving a formal definition, I have suggested
that an institution consists of a division of activities, a parti-
tioning of individuals, and the matching of activities with in-
dividuals so that a subgroup of individuals has jurisdiction
over a specific subset of activities. An institution also consists
of mechanisms of monitoring, control, and other incentives
that connect the jurisdiction-specific activities of subgroups to
the overall mission. I have pointed to the divisional structure
of firms, the departmental structure of universities, and the
committee structure of legislatures as quintessential illustra-
tions of these kinds of institutional arrangements. Indeed, it
is appropriate to think of an institution as a governance
structure, permitting individual ambition and organizational
purpose to be blended. Finally, I have emphasized that insti-
tutions are not static-that they change in response to their
external environment and internal pressures-so that they
are rarely born whole in the first instance and they rarely stay
undisturbed over long stretches of time. They go with the flow.
ever a delegator (the principal) selects a delegatee (the agent) whom he or

she entrusts with authority and/or resources to accomplish some purpose for
the delegator, ther·e is a problem of control. 'l'hus, th e problem that a con-
stituency has in controlling its representatives is a principal-agent problem,
as art> the problems followers have in controlling their leaders, faculties
have in controlling their departments, assemblies have in controlling their
committees, executives have in controlling their staff, and so on-the list is
endless. In each instance some form of monitoring accompanies the delega-
tion of authority.
12
Legislatures
There is a rather ambitious agenda for the next chapters-

namely, to squeeze into a very few pages analyses of specific
political institutions, each of which has preoccupied genera-
tions of scholars who have filled library shelves with their
studies. I certainly do not intend to r eplicate or even compete
with the work of these scholars. Rather, I intend to put the
analytical developments of the previous parts of the book on
display in specilfic institutional contexts. In the case of legisla-
tures, the subject of this chapter, the spatial analysis of the
operating properties of majority rule developed in Part IP is
combined with the analysis of cooperation and collective ac-
tion developed in P art III, 2 while keeping in mind the frame-
work for studying institutions developed in the previous
chapter.
COOPERATION AMONG L EGISLATORS
A popularly elected legislative assembly-the Boston city

council, the Massachusetts legislature, the U .S. Congress, the
French National Assembly, or the European Parliament-
1
The reader may wish to return to "Spatial Models of Legislatures" in Chap-
ter 5 for a quick review.
2 Chapters 8 and 9.
369
consists of politicians who harbor a variety of political objec-

tives. Since they got where they are by winning an election,
and many hope to stay where they are or possibly advance
their political careers, these politicians are intimately aware
of whom they must please to do so:
Because campaigns are expensive propos1t10ns, most
politicians are eager to please those who can supply re-
sources for the next campaign- financial "fat cats," po-
litical action committees, important endorsers, small
contributors, party officials, volunteer activists.
• The most recent campaign-one that the politician won
- provided h er with information about just why the vic-
tory was secured. It is sometimes quite difficult to sort
out the myriad of factors, but at the very least the politi-
cian has a good sense of what categories of voters sup-
ported her and may be prepared to support her again if
her performance is adequate.
• Many politicians aim to please not only campaign con-
tributors and voters; they also have an agenda of their
own. Whether for virtuous reasons or evil ones, for pri-
vate gain or public good, politicians ordinarily come to
the legislature with goals of personal importance.
We may think of all the people to whom the legislator is ac-
countable (including to himself or herself) as the legislator's
constituents. Different legislators will give different weight to
personal priorities and the things desired by campaign con-
tributors and past supporters. Some regard themselves as per-
fect agents of others; they have been elected to do the bidding
of those who sent theni to the legislature, and thus act as del-
egates. Other legislators conceive of themselves as having
been selected by their fellow citizens to do what the legislator
thinks is "right," and thus act as trustees. Most legislators are
mixes of the two ideal types.
A legislative assembly therefore consists of a motley crew
Legislatures 371
of legislator s, each motivated by a combination of desires-

wanting to please those who control his or her political future
and wanting to achieve persona] goals. Broadly speaking,
their instrumen tal objectives involve trying to secure things
for their constituents, however they might define the latter.
The first premise, then, is that legislators come to the legisla-
ture with political purposes that motivate them to want to
pursue specific p ublic policies (instrumental behavior).
The second premise is that in a representative democracy
the specific public policies that r epr esentatives wa nt to pursue
are heterogeneous (preference heterogeneity) . We mean this in
two respects. First, owing to their differen t constituencies, leg-
islators will give priority to different realms of public policy. A
Cape Cod congressman will be interested in shipping, fishing,
coastal preservation , harbor development, tourism, and ship-
building. A Philadelphia congresswoman may not care much
at all about any of those issues, focusing her attention instead
on welfare reform, civil rights policy, aid to inner-city school
systems, and job retraining programs. Montana's sole member
of Congress is probably not interested in coastal preservation,
nor in inner-city schools, but rather in issues of"ranching,
agriculture, mining, and public land use. An assembly con-
tains a melange of legislative priorities.
Legislative assemblies are also heterogeneous in the opin-
ions their members hold on any given issue. While some may
care passionatel y about the issue in question, and others not a
whit, there is bound to be conflict, both at the broad philo-
sophical or ideological level and a t the practical level, on how
to proceed. Thus, while interest in environmental protection
ranges from high priority among those who count many Sierra
Club members a mong their constituents to low priority among
those who have other fish to fry, once environmen tal protec-
tion is on the agenda there is a broad range of preferences
over specific environmental initiatives. Some want pollution
discharges carefully monitored and regulated by a r elatively
powerful environmental watchdog agency. Others believe that

more decentralized and less intrusive means, such as mar-
ketable pollution permits, are the way to go. Still others think
the entire issue is overblown, that any proposed cure is worse
than the disease, and that the republic would best be served
by leaving well enough alone. This preference heterogeneity is
modeled in Chapt er 5 in terms of a spatial model in which the
distribution of individual preference curves reflected a range
of ideal policies.
To the two premises of instrumental behavior by legisla-
tors and preference heterogeneity among them, I add a third
premise. Diversity in priorities and preferences among legis-
lators is sufficiently abundant that the view of no group of
legislators predominates (diversity). Legislative consensus
must be built-this is what legislative politics is all about.
Each legislator clamors to get her priority issue the attention
she believes it deserves, or to make sure that her position on a
given issue prevails. But neither effort is likely to succeed on
its own merits. Support must be assembled, compromises
made, deals consummated, and promises and threats utilized.
In a word, legislators intent upon achieving their objectives
must cooperate.
Cooperation, as we learned in P art III, may be assembled
separately on each occasion. But one-shot efforts at cooperat-
ing often run into insurmountable difficulties. Cooperating
parties are often suspicious, for one thing, and thus guard
against being taken advantage of. At the very same time, they
contemplate the pros and cons of taking advantage of others.
Finally, they loathe having to waste resources on securing
compliance each and every time. A more promising possibility
arises, it would seem, when cooperation can be spread across
recurrences of the same issue over time or across lots of differ-
ent issues at any specific point in time. I'll scratch your back
on this issue at this time; you scratch mine on that issue at
some other time. But still, there is the problem of one party
Legislatures 373
securing another's cooperation at one point in time, only to re-

nege on reciprocating as promised at some other time. Prob-
lems of compliance do not go away, even if cooperation is
spread across issues and time.
Cooperation, especially on recurring matters, is facilitated
by institutionalization. Indeed, we shall claim that many in-
stitutional practices in legislatures reflect the requirements of
facilitating cooper ation in an environment of preference het-
erogeneity, diversity, and instrumental behavior. This leads to
legislative effort being divided and specialized, legislative pro-
cedures regularized, legislative subjects partitioned into juris-
dictions, specific forms of agenda power and other distinctive
advantages delegated, and the interactions fostered by these
arrangements monitored to assure compliance with coopera-
tive objectives. All of these features arise as part of a coopera-
tive governance structure.
UNDERLYING PROBLEMS
Before we can understand why legislative assemblies select

particular ways to institutionalize their practices, we need a
finer appreciation of the underlying problems with which leg-
islators must grapple. We have encountered many of these
previously in this text in rather abstract form. Here we give
them more concreteness.
Majority Cycles
Preference heterogeneity and diversity mean, first and fore-
most, that no part icular view on public policy predominates in
the legislature. Were this not true-if, that is, a large propor-
tion of the members were more or less in agreement on how to
proceed- then the problem confronting a legislature would
merely be one of providing expedited procedures to enable this
consensual group to vote its program through and implement

it. But because there is typically preference heterogeneity and
diversity, no specific program of policy distinguishes itself as
an obvious course on which to proceed. Worse still, as Part II
made clear, if issues are multidimensional in nature (and
surely they are much of the time), then heterogeneity and di-
versity mean there is no equilibrium to majority voting. Any
status quo can be beaten, a ny policy alternative beating the
status quo can be beaten, any policy that beats the policy that
beats the status quo can also be beaten, and so on without end.
Procedures are required to cut through all this instability.
But procedures that give asymmetric advantages to some at
the expense of others would not be tolerated if they extended
across all issues. Putting such concentrated power to one side,
the best a legislator can hope for is that whatever procedural
practices are instituted to bring orderliness to legislative deci-
sion making allow him or her to have some say in those issue
areas he or she cares about most. Our Philadelphia legislator,
for example, would probably be content to trade away her in-
flu ence on mining or fishing issues to the Montana and Cape
Cod legislators, respectively, in exchange for having a dispro-
portionate say on aid to inner-city school districts.
Matching Influence and Interest

Legislatures are highly egalitaria n institutions. Each legisla-
tor has one vote on any issue coming before the body. Unlike a
consumer, who has a cash budget that she 1nay allocate in any
way she wishes over categories of consumer goods, a legislator
is not given a vote budget in quite the same sense. Instead, his
budget of votes is "earmarked"-one vote for each motion be-
fore the assembly. H e is a bit like a consumer who is given a
series of $1 bills, each designated for a different consumer
good category; he cannot aggregate the votes in his possession
and cast them all, or some large fraction of them, for a motion
Legislatures 375
on a subject near and dear to his h eart (or those of his con-
stituents). This is a source of frustration, since, as we have
noted, the premise of instrume ntal behavior mea ns that legis-
lators would, if they could, concentrate whatever resources
they commanded on those subjects of highest priority to them.
In principle, this frustration could be alleviated by a sys-
tem of vote trading. Votes cannot literally be traded, but
promises about casting the m in particular ways can. Our
Philadelphia legislator can promise her Montana colleague to
vote as the latter wishes on an upcoming mining bill if the lat-
ter, in turn, supports the Philadelphian's preference on an up-
coming jobs bill. There are two problems (at least) with this
idea, however. First, can you imagine the negotiating and
bookkeeping complexities of a system of vote trading in any
but the s1nallest of assemblies? It's one thing to think through
an isolated bilateral exchange of votes. The sheer cost of or-
ganizing and sustaining a "market" in votes involving 435
traders across a myriad of issues boggles the mind. Second,
vote-trading agreements, like other deals among politician s,
are not enforceable contracts. If the Philadelphia legislator
casts her vote for the Montanan's mining bill, but the Mon-
tanan then reneges on his promise to vote for her jobs bill,
what can she do? She can certainly refuse to do business with
him in the fu ture, but she can't sue in a court for nonperfor-
mance of a contractual obligation. For both of these reasons, a
market in votes entails very high costs of transacting, and
thus is a very costly way to match legislator influence with
legislator interests.
It is sometimes thought that party leaders act as liaisons
in vote-trading transactions, matching up legislators inter-
ested in making exchanges, and generally facilitating the
making of deals. The problem with this view, indeed with
nearly any highly centralized manner for organizing ex-
change, is the almost superhuman burden on a leader to know
who is willing to trade what at what price. Again, in s mall so-
cieties this is conceivable. And some legislative leaders, like

Lyndon J ohnson in the U.S. Senate in the 1950s and Robert
Byrd in that same body in the 1970s and 1980s, developed
enormous reputations for their cunning and insight concern-
ing what buttons to push on which members. But even here it
is unlikely that all the deals that could have been consum-
mated were, in fact, struck. Leaders will have a role to play in
the solution I suggest below, but that role is not primarily as
an intennediary in a market for votes.
Information
The refrain of many urban legislators in the last few decades,
like our Philadelphia congresswoman above, is "more jobs at a
living wage." This is a response both to the disappearance of
many jobs from most American cities (they gravitate to lower-
wage regions of the country or out of the country altogether to
lower-wage regions of the world) and, of those that remain,
the often unattractive wages, benefits, and career prospects
attached to them. Many legislative solutions to this serious
problem have been proposed. Some urge a higher minimum
wage; some mandate better fringe benefits-health care cover-
age, day care subsidies, pension benefits, parental leave poli-
cies, and so on; some underwrite training programs to
improve the productivity of workers; some advocate all these
things and more. What works? These are very complicated
matters; even those legislators for whom the problems are
most pressing are often quite unsure how to answer this
question.
If legislators voted directly for social outcomes, then this
wouldn't be a problem al all. The Philadelphia legislator could
simply offer a bill "mandating" more jobs at a better wage in
urban areas and, if it passed, then-abracadabral- the man-
dated effects would become a reality. Alas, legislators do not
vote for outcomes directly, but rather for instruments (or poli-
Legislatures 377
cies) whose effects produce outcomes. Thus, in order to vote

intelligently, legislators must know the connection between
the instruments they vote for and the effects they desire. In
short, they must have information and knowledge about how
the world works.
Few legislators-indeed, few people in general- know how
the world works in very many policy domains in all but the
most superficial of ways. Nearly everyone in the legislature
would benefit fr0m the production of valuable information- at
the very least, information that would allow legislators to elim-
inate policy instruments that make very little difference in
solving social problems, or even make matters worse. Produc-
ing such information, however, is not a trivial matter. Simply
to digest the knowledge that is being produced outside the leg-
islature by knowledge-industry specialists (academics, scien-
tists, interest groups) is a taxing task. In short, policy-relevant
information is a public good and we know, from Part III, that
public goods are undersupplied. Clearly, institutional arrange-
ments that provide incentives to some legislators to produce,
evaluate, and disseminate this knowledge for others will per-
mit public resources to be utilized more effectively.
Compliance
The legislature is not the only game in town. The promulga-
tion of public policies is a joint undertaking in which courts,
executives, bureaucrats, and others participate alongside leg-
islators . If the legislature develops no means to monitor what
happens after a bill becomes law, then it risks public policies
implemented in ways other than those stipulated in the law.
Cooperation, that is, does not end with the successful passage
of a law. If legislators wish to have an impact on the world
around them, especially on those matters to which their con-
stituents give priority, then it is necessary to attend to policy
implementation as well as policy formulation. But it is just not

practical for all 435 representatives and all 100 senators to
march down to this or that agency at the other end of Pennsyl-
vania Avenue to ensure appropriate implementation by the
executive bureaucracy. Compliance will not "just happen" and,
like the production and dissemination of reliable information
at the policy formulation stage, the need for oversight of the
executive bureaucracy is an extension of the cooperation th at
produced legislation in the first place. It, too, must be institu-
tionalized.
We have suggested that legislative politics, reflecting the
diversity and heterogeneity of the preferences of representa-
tives, requires various forms of cooperation that generally are
hard to manufacture. The fact that majorities cycle in their
preferences means that to break the policy indeterminacy, it is
necessary to institute procedural regularities that bring clo-
sure to legislative subjects. To do this, however, is to risk giv-
ing advantages to some at the expense of others, unless
asymmetric advantages can be held in check, on the one hand,
and subdivided and distributed widely, on the other. Related
to this, representatives are not uniformly interested in all
subjects that come before the legislature, since there is nor-
mally heterogeneity among their respective constituencies.
Yet they are endowed with one vote each on every subject that
comes to a vote. This imbalance between interests and influ-
ence provides a setting in which there are prospective gains
from exchange in order to improve the match between the
two. Finally, part and parcel of the cooper ation that is re-
quired to get anything accomplished in a legislative assembly
is the generation of information and the monitoring of results
in the wider political system. Tu produce these public goods,
incentives must be properly arranged.
Legislative institutions emerge and evolve in response to
these problems. T hey also respond to other problems, like
Legislatures 379
wartime emer gen cies or environmental crises; but those de-

scribed in the preceding paragraphs are the most enduring
ones to which i nstitutional solutions are matched. Institu-
tional solutions, moreover, are not static, one-time-only re-
sponses, both because the problems change from time to time
and because the solutions arrived at are not always the best
ones. Learning a nd adaptation to changes in the policy envi-
ronment surely occur in institutional life. Finally, I would be
remiss if I did not mention that institutions incorporate con-
flict. Legislatures are not only forums in which teams of kin-
dred spirits try to solve common problems. They are also
battlefields on which the deployment of public authority is de-
termined (and indeed, at its most ideological, on whether pub-
lic a uthority should be deployed at all) .
L EGISLATIVE STRUCTURE AND PROCEDURE
With the stage thus set, I can now apply our framework for
analyzing institutional politics to legislatures. Throughout I
assume the legis lature consists of n legislators (where n = 435
for the U.S. House, n = 100 for the U.S. Senate, n - 650 for the
British H ouse of Commons, and so on). I maintain as working
premises that legislators are instrumen tal in seeking policies
that will please various of their constituents, that legislator
preferences over policy options arc heterogeneous, and that no
legislator preference is sufficiently numerous in the legisla-
ture as to be decisive. Finally, I assume that legislative policy
may be represented by a spatial formulation, with each di-
mension reflecting a n aspect of public policy salient to at least
some legislators. With these features as background, I argue
that legislators choose a division and specialization of labor,
regular procedure, a jurisdictional arrangement, and delega-
tion a nd monitoring technologies in order to facilitate coopera-
tion and other gains from exchange. After making my case for
these claims, I will go on to analyze additional layers of insti-
tutional detail that constitute the operating features of real
legislatures.
Jurisdictional Division and

Specialization of Labor
All of the elements of my model of an institutionalized legisla-
ture fit together almost seamlessly. So it is somewhat arbi-
trary where to start the discussion. Perhaps it is best to begin
with the policy space. The dimensions of this space constitute
the parameters of public policy, which the legislature may set
and reset through legislation. By voting larger appropriations
for the military than in the past, for example, the legislature
moves a parameter of defense policy from a less hawkish to a
more hawkish setting. By granting broadly interventionist
statutory authority to the Environmental Protection Agency,
the legislature moves a parameter for regulatory policy from a
more restrained to a more activist setting. There are many
such dimensions of public policy on which the legislature
works its will. Legislators, we have assumed, have policy pref-
erences on these various dimensions (though any one legisla-
tor may care passionately about only a small subset of them).
Some prefer hawkish defense policies, while others are more
dovish; some want a proactive environmental protection pol-
icy, while others want less public intervention; and so on.
Suppose, for the sake of argument, we can describe public
policies in terms of some specific set of m policy dimensions,
D = {d1, d 2' . . • , dm}. Now let's group these m dimensions into
a set of jurisdictions. A jurisdictional arrangement, J, is a col-
lection of these groups-J = {JI' J 2 , . . . , J,}. That is, there are
t jurisdictions, each of which contains some relatively small
number of dimensions of public policy. We assume that each
policy dimension in D is in exactly one of the jurisdictions in
Legislatures 381
J, so that the jurisdictional arrangement partitions D into a

collection of mutually exclusive and collectively exhaustive
groupings. For example, if there are five policy dimensions,
D = {d1, d2' d 3, d 4 , dJ, then one of the possible jurisdictional
arrangements puts the first two dimensions in a jurisdiction,
the third dimension in a second jurisdiction, and the last two
dimensions in a third jurisdiction-J ~ {Jl' J 2, JJ, where J 1 -
{dl' d 2}, J 2 - {dJ, and J 3 - {d4 , dJ.
Now let's do the same thing to the members of the legisla-
ture, N • {l, 2, ... n}, grouping these n members into commit-
tees exactly equal in number to the number of jurisdictions.
Call C • {C1, C2 , . • • C1} a committee system, where each mem-
ber of N is on exactly one committee. 3 Finally, we connect C to
J, assigning each committee one of the t policy jurisdictions.
Thus, J I' consisting, say, of the various policy dimensions as-
sociated with agriculture, might be assigned to cl' thus iden-
tifying the legislators in that group as members of the
Agriculture Committee.
In this manner we have characterized a jurisdictional divi -
sion and specialization of labor in a legislative setting. The
policymaking business of the legislature is divided up into ju-
risdictional responsibilities. Legislative labor is specialized by
dividing up the membership into committees, each of which is
assigned one of the specialized jurisdictions.
Camouflaged beneath all this notation are two crucial po-
litical processes. The first involves the assignment of issues to
committees. In many, perhaps most, instances, the subject
matter of issues fall squarely into one or another committee
jurisdiction. However, new issues arise that often fit neatly
into no jurisdiction. Some, like the emerging issue of energy
supplies during the 1970s, are so multifaceted that bits and
3 Actually it is quite all right, and factually accurate in the case of the U.S.
Congress and most of the state legislatui·es, to allow members to serve on
more than one committee. What is normally forbidden is to give a member
no committee assignment.
pieces of them are spread across many committee jurisdic-

tions. Thus, the Com1nerce Com1nittee of the U.S. House of
Representatives had jurisdiction over the regulation of energy
prices, the Armed Services Committee dealt with military im-
plications, the Ways and Means Committee dealt with tax-
related energy aspects, the Science and Technology Commit-
tee claimed jurisdiction over energy research, the Agriculture
Committee dealt with grain-to-energy conversion matters,
and several other committees picked off still other pieces of
this hydra-headed issue. Other issues, like that of regulating
tobacco products, fall in the gray area claimed by several dif-
ferent committees-in this case the Commerce Committee,
with its traditional claim over health-related issues, fought
with the Agriculture Committee, whose traditional domain in-
cludes crops like tobacco, for jurisdiction over this issue. Turf
battles between committees of the U.S. Congress are notori-
ous.4 These battles, often extending over many years, involve
committee chairs, the Parliamentarian's Office , the political
leadership of the chamber, and, from time to time, selected
committees appointed to realign committee jurisdictions. All
in all, jurisdictional conflict is the raw stuff of politics, since,
as we shall see, committees with jurisdiction over issues have
significant leverage over their resolution.
The second J)<)litical process submerged beneath our ab-
stract description of legislative institutions is the committee
assignment process-the assignment of members to commit-
tees. The process determining who gets on what committees is
significant because it designates which members will have ex-
traordinary influence over the issues falling into the jurisdic-
tion of each committee. In most legislatures, this process is
conducted within the political parties. Each party has a cer-
4 An outstanding description and analysis of these battles is found in David C.

King, "The Nature of Congressional Committee Jurisdictions," American Po-
litical Science Revil'w 88 (1994): 48-{)3. See also his Turf Wars (Ann Arbor:
University of Michip:an Press, 1996).
-
Legislatures 383
tain number of vacancies per committee with which to work;

these are assigned to new and returning members (the latter
of whom typically retain the committee assignment they held
in the just-completed legislative session). It is a highly charged
political process, especially for the newcomers, having major
ramifications for policy making in that legislative session. 5
Delegation of J urisdictional Authority

In order to give operational significance to the division and
specialization of labor described above, it is necessary to as-
sign committees specific forms of authority in their respective
jurisdictions. If committees had no authority, or if whatever
they did had little impact on the policies ultimately selected
in the legislature or the outcomes in the larger political sys-
tem, then about the only thing a committee member could do
would be to list her membership on her resume, along with
memberships in other clubs and honorary societies. 6 Commit-
tees in the U.S. Congress, however, have considerable author-
ity, some of whiich I described briefly in Chapter 5.
It needs to b e said at the outset that committee jurisdic-
tions and authority, like nearly all other aspects of structure
and procedure, are created by the parent legislature.7 In the
U.S. Congress, the House and Senate give these things to, and
5 For an extensive discussion of the committee assignment process in the U.S.

House, see Kenneth A. Shepsle, The Giant Jigsaw Puzzle: Democratic Com-
mitt.ee Assignments in the Modern House (Chicago: University of Chicago
Press, 1978). An extremely comprehensive and up-to-date study is Scott A.
Frisch and Sean Q. Kelly, Committee Assignment Politics in the U.S. House
of Representatives (Norman: University of Oklahoma Press, 2006).
6 Many parliamentary democracies have committee systems in which the
committees actually have very little to do. The business of government in
various policy jurisdictions is dominated by cabinet ministers and senior
civil servants. Committees, for the most part, are window dressing.
7 In the United States, for example, the Constitution does designate certain
requirements for the two houses of Congress. But it also permits each house
to set its own rules, thereby giving the parent bodies tremendous discretion
to revise structural and procedural feat ures as they wish.
may take them away from, their subunits. In this view, com-
mittees may be thought of as agents of the parent body to
whom jurisdiction-specific authority is provisionally delegated.
Of what does this delegation consist? I describe committee au-
thority in terms of gatekeeping power, proposal power, inter-
chamber bargaining power, and oversight authority.
In all of the dimensions of public policy falling within a
committee's jurisdiction, there is always some status quo policy
in place. It may be that the issue is new, in which case "cur-
rently doing nothing" is the prevailing policy. On other issues
there are well-established laws keeping a bevy of bureaucrats
and other government officials busy at implementation and ad-
ministration. At any time on any of these dimensions, a com-
mittee may choose to "open the gates" by proposing new
legislation to change the status quo ante. Although a slight ex-
aggeration, it is fair to say that 'in many legislatures commit-
tees have, in their jurisdiction, exclusive gatekeeping authority.
That is, it is practically impossible for the full legislature to
consider changes in the status quo in a committee's jurisdic-
tion unless the committee consents to open the gates. This
makes a committee an agenda monopolist in its jurisdiction.
A committee's agenda monopoly should be thought of as
provisional. It is a standing decision by the parent legislature.
At any time, however, the parent legislature can revise that
standing decision. This is the parent legislature's "club behind
the door." A committee, in exercising agenda power, is always
at risk of inducing the parent legislature to revise its agenda
authority. A second mechanism by which the parent legisla-
ture can discipline a committee is the discharge petition. This
is a temporary procedure in which the legislature, upon a ma-
jority of its members signing a petition, removes a particular
bill from the committee and brings it directly to the floor.
Closely related to gatekeeping is a committee's proposal
power. Normally, any member of the legislature can submit a
bill calling for changes in the status quo in some policy area.
Legislatures 385
Almost automatically, this bill is assigned to the co.m mittee of

jurisdiction and, very nearly always, there it languishes. In a
typical year in the House of Representatives, nearly fiftee n
thousand bills are submitted. Fewer than a th ousand are
taken up by that committee and assigned to the appropriate
committee of jurisdiction. In effect, while any inember is enti-
tled to make proposals, the real proposal power inheres in
committee majorities. So, committees get to decide not only
whether to open tho gates but also what proposal emerges
through those gates. Committees, then, are lords of their ju-
risdictional domains, setting the table, so to speak, for their
parent chamber.8
Finally, a committee has responsibilities for bargaining
with the other chamber and for conducting oversight, or what
we call after-the-fact authority. Because many legislatures are
bicameral-the U.S. Congress, for instance, has a House and
a Senate-once one chamber passes a bill, it must be trans-
mitted for consideration to the other chamber. If the other
chamber passes a bill different from the one passed in the first
chamber, and t!he first chamber refuses to accept the changes
made, then the two chambers ordinarily call a conference in
which representatives from each chamber (called conferees)
meet to hammer out a compromise. In the wide majority of
cases, conferees are drawn from the committees that had orig-
inal jurisdiction over the bill. For example, in a sample of Con-
gresses in the 1980s, of the 1,388 House members who served
as conferees for various bills during this period, only seven
were not on the committee of original jurisdiction; similarly, in
the Senate on only seven of 1,180 occasions were conferees not
8 This clearly gives commitLee members extraordinary power in their respec-

tive jurisdictions. allowing them to push policy into line with their own pref-
erences--but only up to a point. If the abuse of their agenda power becomes
excessive, the parent body, as noted above, has structural and procedural
remedies available to counteract this-like stacking the committee with
more compliant members, deposing a particularly obstreperous committee
chair, or removing policies from a committee's jurisdiction.
drawn from the "right" committee.9 The committee's effective

authority to represent its chamber in conference committee
proceedings constitutes after-the-fact power that complements
its before-the-fact gatekeeping and proposal powers. 10
A second manifestation of after-the-fact committee author-
ity consists of the committee's primacy in legislative oversight
of policy impleme ntation by the executive bureaucracy. I will
have more to say about this in the next chapter. Suffice it to
say here that even after a bill becomes a law it is not always
(indeed, it is rarely) self-implementing. Executive agents-
bureaucrats in the career civil service, commissioners in regu-
latory agencies, political appointees in the executive branch-
march to their own drummers. Unless legislative actors hold
their feet to the fire, they may not do precisely what the law
requires (especially as s tatutes are often vague and ambigu-
ous). Given this possibility, the Legislative Reorganization Act
of 1946, a law that reformed and redefined how the House and
Senate have conducted their business through most of the late
twentieth century to the present day, instructed congressional
committees to be "continuously watchful" of the manner in
which legislation is implemented and administered. Commit-
tees of jurisdiction play this after-the-fact role by allocating
committee staff and resources to keep track of what the execu-
tive branch is doing and, from time to time, holding oversight
hearings in which particular policies and programs are given
intense scrutiny. Anticipating this surveillance, and knowing
9 See Kenneth A. Shepsle and Barry R. Weingast, "The Institutional Founda-

tions of Committee Power," American Political Science Review 81 (1987):
85-104. The evidence for the claim in the text is found in Tab le 1.
°
1
Forty-nine of the fifty U.S. states have bicameral legislatures and thus have
mechanisms for th e bicameral resolution of differences. Only Nebraska is
unicameral. Interestingly, to my knowledge there is only one city in the
United States with a bicameral legislature. Everett, Massaschusetts, has a
seven-member Board of Aldermen and an eighteen-member Common Coun-
cil, together comprising the City Council. Why more cities and towns do not
have bicameral arrangements is a puzzle, one deserving research and
thinking.
Legislatures 387
what grief a congr essional committee can cause an executive

branch agent found deviating from what legislators want,
these officials are very keen to keep their congressional "mas-
ter s" content. This, in turn, gives congressional committees an
additional source of leverage over policy in their jurisdictions.
CASE 12.1
CAMPAIGN CONTRIBUTIONS*
I have argued that committees are powerful institutions in

legislatures because of their gatekeeping power, proposal
power, interchamber bargaining power, and over sight au-
thority. The ways in which interest groups allocate their
campaign contributions provides evidence of this committee
influence.
Interest groups often contribute money to legislative
campaigns in an attempt to influence legislative outcomes.
Charles H. Keating Jr., the bank executive prosecuted for
his role in the savings and loan crisis of the 1980s, was
asked about th e $1.4 million in campaign contributions be
gave to five senators, all members of the Senate Banking
Committee, w·ho came to be known as "the Keating Five."
Keating replied:
One question, among the many r aised in recent weeks, had to
do with whether my financial support in any way influenced
several political figures to take up my cause. I want to say in
the most forceful way I can: I certainly hope so.t
•This chapter examines the effects of institutional arrangements on leg-

islative behavior. Interest groups are often accused of biasing this behav-
ior by "buying" legislators with campaign contributions. The two cases in
this chapter examine the role of campaign contributions and interest
group activities using the institutional approach to legislatures.
t Cited in James Ring Adams, The Big Fix: inside the S & L Scandal (New
York: Wiley, 1990), p. 254. Also see Dennis F. Thompson, Ethics in Con-
gress (Washington, D.C.: Brookings Institution, 1995).
Assuming that interest groups are rational and that

their primary goal is influencing legislation that affects
them , we would expect groups to allocate their contribu-
tions to legislators with the most influence. So where do in-
terest groups allocate their money? Based on the material
in this chapter, you shouldn't be surprised by the answer: to
legislators who sit on committees that have jurisdiction
over the groups. Banks give to the members of the banking
committees, food and tobacco producers to those sitting on
the agriculture committees, doctors and insurance compa-
nies to those serving on the health-related committees. The
power of committee chairs is also apparent from the alloca-
tion of campaign contributions. Of the legislators who sit on
the committees relevant to a particular interest group, com-
mittee chairs often receive a greater share of the campaign
contributions.+
i See Larry Sabato, PAC Power: Inside the World of Political Action Com-
mittees (New York: Norton, 1984). Also see Kevin B. Grier and Michael C.
Munger, "The Impact of Legislator Attributes on Interest Group Cam-
paign Contributions," Journal of Labor Research 7 (1986): 349-61.
Monitoring and Control

As should be apparent by now, committees are quite conse-
quential players in their respective policy jurisdictions. If
unchecked, committees could easily take advantage of their
authority. Indeed, what prevents committees from exploiting
their before-the-fact agenda power and their after-the-fact
bargaining and oversight authority? To answer this question
we have to see what "taking advantage" means. And this re-
quires us to think about which legislators become members of
which committees.
To make this clear, suppose a committee's jurisdiction were
one-dimensional. In the full legislature the policy that a leg-
islative majority prefers in this jurisdiction is, from Black's
Legislatures 389
Theorem in Chapter 5, the most-preferred policy of the median

legislator, xm. The policy most preferred by the committee, on
the other hand, is the most-preferred policy of the median
committee member, xe. Only by accident would these two ideal
points be the same (but accidents can happen, as I remark on
shortly). Thus, the committee wants to pull policy toward xe,
whereas the full legislature wants to pull policy toward xm. It
is this tension that pits legislature against committee, princi-
pal against agent, delegator against delegate.
Recalling our analysis in Chapter 5, the form this tension
takes depends partly on the order along the policy dimension
of committee median and legislative median in comparison to
the status quo, xfJ. If xfJ lies between the two medians, in which
case committee and chamber want to move policy in opposite
directions, the tension is caused by the committee's refusing to
open the gates. (Of course, if they did stupidly open the gates,
then any proposal they made to improve their lot would be
voted down or, to make matters worse, amended in a direction
away from the committee median.) If xm lies between the com-
mittee median and the status quo, or if xe lies between the
chamber median and the status quo, then it is possible for
both committee and chamber to improve their respective lots,
depending on the way in which committee bills are handled in
the full house (see the analysis in Chapter 5 for details). In ei-
ther of these latter cases, however, the tension does not go
away even under the best of circumstances, and it gets worse
the greater the distance is between Xe and Xm.
There are basically two "ordinary'' ways in which the par-
ent chamber keeps committees in check, thereby reducing
their frustration with potential committee abuses of author-
ity: committee assignments and amendment control rules. 11
11 In addition, an "extraordinary" way, much like dropping a bomb, entails

taking committee jurisdiction away or, the equivalent of capital punishment,
disbanding the committee and dispersing its jurisdiction to other commit-
tees. A prominent example of this form of capital punishment involved the
Regarding committee assignments, the key issue is the de-

gree to which a committee is representative of the entire
chamber. Representative committees may be thought of as "lit-
tle legislatures." Majorities there are likely to behave much in
accord with majorities in the larger chamber from which they
were drawn. Outlier committees, on the other hand, have a
composition significantly different from the parent chamber.
In these situations the distribution of preferences in the com-
mittee is very different from that of the full legislature. If a
committee's jurisdiction consists of one policy dimension, for
instance, the differences between these two types of commit-
tees is captured in the distance between the median commit-
tee ideal point, xc, and the median legislator in the full
legislature, xm. Roughly speaking, representative committees
are associated with small deviations between the two,
whereas it is larger for outlier committees.
Why, one might ask, doesn't the legislature ensure com-
mittee responsiveness to preferences in the chamber by ap-
pointing only representative committees? That way, so the
argument goes, the legislature would gain the benefit of a
small jurisdictionally specialized group, one producing public
goods like information and oversight of the executive, without
having to risk abuse of the asymmetric authority it possesses.
The reason legislatures typically do not do this follows
from our preference heterogeneity premise. Many committee
jurisdictions consist of policy a reas of interest to relatively
infamous House Un-American Activities Committee (HUAC), in 1969 re-

named the House Internal Security Committee (HISC), 1938-75. This com-
mittee, regardless of its name, embarrassed much of the House during the
postwar period by exaggerating fears that communists were infiltrating
American public and private institutions. Publicity· seeking members of the
committee held pubhc hearing after public hearing, pillorying guilty (of
which there were a few) and innocent (of which there were many) alike,
with low regard for the civil liberties or the presumption of innocence of
those appearing before it. By the early 1970s most members of the House
had had enough, and the committee was dispatched to the junk pile of
history.
Legislatures 391
small numbers of legislators. Congressmen and congress-

women from coastal districts, for instance, and virtually no
others, give priority to the jurisdiction of the Merchant Ma-
rine and Fisheries Committee. Fewer than a quarter of the
House of Representatives would kill to get on the Agriculture
Com mittee; ditto for the Banking and Housing Committee.
These committees deal with problems and issues of great
importance to rural districts and central city districts, respec-
tively, whose representatives constitute relatively small pro-
• portions of the entire chamber. It is chiefly legislators from the
Maryland and Virginia suburbs of Washington, D.C., who
want a seat on the Post Office and Civil Service Committee,
since so many of their constituents are employed by the fed-
eral governmen t. Contrast these committees with the Appro-
priations Committee, the Ways and Means Committee, the
Armed Services Committee, the Budget Committee. These lat-
ter committees deal with issues of general importance regard-
ing taxing, spending, budgeting, and defense. Nearly every
district and state is affected by their decisions; nearly every
member of the House of Representatives or Senate would de-
light in serving on one or another of these committees.
Most of the legislature is not particularly concerned about
what goes on in many of the specialized committees listed
early in the p1·eceding paragraph, and most are highly con-
cerned with decisions coming out of the more generalist com-
mittees listed toward the end of the paragraph. Thus, one
would expect an arrangement in which
• legislators were allowed freely to flow to the specialist

committees to which they give priority, even though
committees populated this way would look like outliers
to an outside observer; but
• the composition of generalist committees would be more
carefully monitored by the parent body, with greater ef-
fort expended to make these committees representative.
Evidence from the House of Representatives is consistent with

this expectation: on average, the difference Ix"' - xc I is larger
for specialist committees than for generalist committees. 12
The committee assignment process, while generally respon-
sive to requests for committee positions coming from the
chamber's members, is especially responsive if the request is
for a specialist committee whose jurisdiction is not a high pri-
ority to a large proportion of the membership. In these cases
there does not appear to be much concern if a committee is not
representative of the full chamber. Responsiveness, however,
is constrained by concerns for representativeness if the re-
quest is for a generalist committee whose policies generate ex-
ternal effects on committee member and nonmember alike. 13
There is a second, complementary tool by which the parent
chamber monitors and controls its subunits. Whenever a com-
mittee brings a pt·oposal to the floor, the legislative chamber
must decide how it will deliberate on the proposal. In many
legislatures there are predetermined standing rules governing
the disposition of legislation. In the U.S. House, however, it is
customary to craft a rule specific to the legislative proposal in
question. This rule, which is called an amendment control
rule, regulates the amount of time devoted to debate, how that
time is divided between proponents and opponents, and, most
12 There is still controversy among scholars concerning this issue, but it mainly
involves the very thorny problems of measuring preferences of real legisla-
tors. The most committed of legislative jocks who want more on this sub-
ject, may find it for the period through the mid-1970s in Shepsle, 'l'he Giant
Jigsaw Puzzle. A study covering most of the postwar period up through the
early 1990s is Gary W. Cox and Mnthew D. McCubbins, Legislative
Leviathan (Berkeley: University of California Press, 1993). Also see their
Setting the Agenda (New York: Cambridge University Press, 2005).
13
I remarked earlier that it would be quite accidental if xm • x, but quipped
that accidents do happen. We now know why. Because generalist commit-
tees have such pervasive policy impacts on the welfare of a large number of
different constituencies, the legislature explicitly stacks these committees
so that the committee median is relatively close to the chamber median.
This is what it means to make these committees "representative."
Legislatures 393
important, what amendments are in order, if any. Thus, while

committees inay have disproportionate agenda power, control-
ling whether the gates are opened in their respective jurisdic-
tions and, if so, what proposals come through them, they do
not dictate final outcomes in their jurisdictions. Their propos-
als may be modified by the parent chamber.
The parent chamber's capacity to "keep an eye on" commit-
tee activities tends to be the most exacting, as the reader
might guess, in the case of specialist committees. Their pro-
posals normally are governed by an open rule (see Chapter 5),
allowing for virt ually unrestricted amendment activity by the
parent chamber. Thus, even if such committees are outliers,
their products must pass muster in a full chamber armed with
an ability to make wholesale changes. Generalist committees,
on the other hand, often receive restrictive rules for their pro-
posals, limiting the chamber's ability to amend. The extreme
version of the latter is the closed rule (also discussed in Chap-
ter 5), according to which no amendments are in order.
The reader undoubtedly will concede that it makes sense
for the proposals of outlier committees to be subjected to close
scrutiny and control. Precisely because they are unrepresenta-
tive of the parent chamber, their efforts will need close atten-
tion.14 The reader will also concede that it makes sense for the
proposals of r epresentative committees to be subjected to
much more relaxed scrutiny, precisely because they are repre-
sentative of the parent body. What may remain puzzling is
14 The careful reader might note that because their jurisdictions are of rele·
vance to narrow constitutencies, specialist committees really don't need to
be monitored clasely since their policy effects are confined. While this is
true most of the t ime, it is not always true. In particular, whenever propos·
als en tail the expenditure of revenues, they have implications for the activ·
ities of every other committee, since these expenditures draw from the
same common revenue pool from which other committees hope to draw.
Most of the House would not be pleased with an Agriculture Committee, for
example, that proposed a level of price support for agricultural products
that consumed a large proportion of the annual public budget.
why some protection from wholesale change in the larger leg-

islature is (at least some of the time) afforded the proposals of
outlier-specialist committees while, on the other hand, the
chamber grants itself some ability to amend the proposals of
representative-generalist committees.
An answer to the first part of this puzzle is provided by
Keith Krehbiel in his masterly study Information and Legisla-
tive Organization. 15 Krehbiel argues that most legislative ac-
tivity takes place in a context of great uncertainty about the
relationship between proposed legislative solutions and social
problems. Information about these cause-and-effect relation-
ships helps to reduce this uncertainty. Thus, encouraging com-
mittee members to become expert specialists in the subject
matter of their jurisdictions and to disseminate this expertise
to their noncommittee colleagues enhances the legislature's
capacity to cut through the uncertainty that surrounds most
occasions for legislating. The promise of at least partial pro·
tection for their proposals is, in effect, the price the legislature
is prepared to pay in exchange for reliable information. This
will mean that the committee can, in fact, extract some advan-
tage from their prelegislative and postlegislative authority (by
having their proposals protected from wholesale revision on
the floor), but it will also mean that the legislation that ulti-
mately results is likely to have benefited from the expertise
which committee members have been encouraged to develop.
In short, the system of protection of legislative products is an
incentive system that encourages the development of legisla-
tive expertise.
The second part of the puzzle is why a chamber nearly al-
ways retains some ability to revise committee proposals, even
those of representative-generalist committees. At least part of
the answer lies in the fact that no committee can be represen-
1
~ (Ann Arbor: University of Michigan Press, 1991).
Legislatures 395
tative in every respect. A committee, for example, may have

slightly more westerners on it than are found in the full
legislature, or be slightly more liberal than the rest of the
chamber. The disparity between committee and chamber pref-
erences may not be as large in generalist committees as in
specialist committees, but it may not be zero either. And there
is one way in which any committee is unrepresentative. The
specific members of a committee represent specific geographic
districts, and represent none of the districts of members
not on the committee. On matters that involve distributing
goodies- ranging from the classic pork barrel of dredging
rivers and harbors, building post offices, and locating military
installations, to the more subtle forms of pork-barreling like
siting sexy scientific facilities, giving grants to favorite univer-
sities, or enriching particular contractors- members of a com-
mittee, if unmonitored and unchecked, will undoubtedly find
ways to distribute the lion's share of their goodies to their own
districts. 16 The chamber, by retaining rights to amend, deters
at least the more egregious instances of this kind of oppor-
tunistic behavior. 17
16 Evidence on this may be found in one of the classics of this literature, John
Ferejohn, Pork Barrel Politics (Stanford, Calif.: Stanfot·d University Press,
1974). For an alternative perspective, see Robert M. Stein and Kenneth N.
Bickers, Perpetuating the Pork Barrel (New York: Cambridge University
Press, 1995). For an excellent recent study, also see Diana Evans, Greasing
the Wheels (New York: Cambridge University Press, 2004).
17 Nevertheless, in the 1960s it was rumored that Charleston, South Carolina,
was sinking under the weight of all the naval installations the chair of the
House Armed Services Committee, Mendel Rivers, had managed to secure
for his district. J\.fore recently, there has been concern that the same fate is
being suffered by the state of West Virginia, whose powerful senior senator,
Robert Byrd, the chair of the Senate Appropriations Committee and former
majority leader, !has managed to entice numerous government agencies to
relocate to his state from Washington, D.C.
CASE 12.2
INTEREST GROUP I NFLUENCE
In the 2008-10 electoral cycle, many millions of dollars in

campaign contributions were given by interest groups and
political action committees (PACs) to incumbent legislators
and challengers. What does this money buy? A popular con-
ception is that campaign contributions buy votes. In this
view, legislators vote for whichever proposal favors the bulk
of their contributors. Although the vote-buying hypothesis
makes for good campaign rhetoric and newspaper editori-
als, it has little factual support. Empirical studies by politi-
cal scientists show little evidence that contributions from
large PACs influence legislative voting patterns.*
If contributions don't buy votes, then what do they buy?
l\1y claim is that campaign contributions influence legisla-
tive behavior in ways that are difficult for the public to
observe and for political scientists to measure. The institu-
tional structure of Congress provides opportunities for in-
terest groups to influence legislation outside the public eye.
Committee proposal power enables legislators, if they are
on the relevant committee, to introduce legislation that
favors contributing groups. Gatekeeping power enables
committee members to block legislation that harms con-
tributing groups. The fact that certain provisions are ex-
cluded from a bill is as much an indicator of PAC influence
as the fact that certain provisions are included. The differ-
ence is that it is hard to measure what you don't see. Com-
mittee oversight powers enable members to intervene in
bureaucratic decision making on behalf of contributing
·See the important early Janet M. Grenke, "PACs and the Congressional
Supermarket: The Currency Ts Complex," American Journal of Political
Science 33 (1989): 1- 24.
Legislatures 397
groups. In the case of Charles Keating Jr., described in

Case 12.1, for example, five senators on the Senate Banking
Committee used their oversight authority to induce bank
regulators to ease up on their supervision of Keating's sav-
ings and loan (which ultimately failed).
The point h ere is that voting on the floor, the alleged ob-
ject of campaign contributions according to the vote-buying
hypothesis, is a highly visible, highly public act, one that
could get a legislator in trouble with his or her broader elec-
toral constituency. The committee system, on the other
hand, provides loads of opportunities for legislators to de-
liver "services" to PAC contributors and other donors that
are more subtle and disguised from broader public view.
Thus, I suggest that the most appropriate places to look for
traces of campaign contribution influence on the legislative
process are in the ways committees deliberate, mark up
proposals, and block legislation from the floor; existing, as
they do, outside public view, these are the primary arenas
for interest group influence.
LEADERSHIP AND COORDINATION
I have argued that the committee system reflects the desire,

common in legislatures and most other organizations, to divide
up business and specialize labor to cut through indeterminacy,
to generate inforined decision making, and to guarantee that
the legislature has impact in the larger political system. An
important by-product of this arrangement (indeed, some
would say it is the main product) is that it permits legislators
to bring their interests and their authority more into line, al-
lowing members, subject to constraints on representation to
flow to those committees whose jurisdictions reflect their pri-
orities and interests. Thus, while legislatures are one-person,
one-vote institutions, most legislators have a larger, more in-

fluential voice in those policy areas of greatest concern to their
constituents.
Committees, in effect, are legislatures writ small; they
have the corresponding motive to divide business and special-
ize labor. Thus, the roughly twenty standing committees of
the U.S. House are, in turn, divided into about a hundred,
even more specialized subcommittees. With so many of the
Montana congressman's constituents involved in growing
wheat, for example, he could best serve them-and best posi-
tion himself to secure their votes in the next election- if he
were not only a member of the Agriculture Committee but also
on its Subcommittee on Feedgrains. These subcommittees
serve their full committees in precisely the same manner the
full committees serve the parent chamber. Thus, in their nar-
row jurisdictions, they have gatekeeping, proposal, intercham-
ber bargaining, and oversight power s. In order for a bill on
wheat to be taken up by the full Agriculture Committee, it has
first to clear the Feedgrains Subcommittee. All of the issues
involving assignments, jurisdictions, amendment control, and
monitoring that we discussed earlier regarding full commit-
tees apply at the subcommittee level as well.
Taking this all in, the reader may already have arrived at
the conclusion that institutions are not just collections of indi-
viduals; they also consist of specific structur al and procedural
arrangements for the conduct of bus iness. Obviously if these
additional arrangements generated more bother than benefit,
an organization might not adopt them at all. Families, for ex-
ample, while specialized to a certain extent, do not display the
fine-grained division and specialization of labor of, say, Micro-
soft. That these features of institutional governance carry
with them additional burdens for an organization should thus
not be underestimated. Institutionalization is not pure profit.
In the case of legislatures, the burden of layers of institu-
tional complexity is reflected in the need for coordination and
Legislatures 399
leadership. One cannot turn twenty committees with a hun-

dred subcommittees loose without some means of guarantee-
ing that their work is s upervised by the parent chamber. The
supervision, moreover, cannot merely involve one committee or
one subcommittee at a time. Someone must coordinate across
institutional subunits. Someone must lead. I discuss leader-
ship more generally in Chapter 14, so I can be brief here.
At the committee level the mantle of leadership falls on the
committli?e chair. He or she must orchestrate the proceedings
of the committee's staff, investigatory resources, a nd subcom-
mittee structure. 18 For many years the Congress followed a
rigid seniority rule for the selection of these chairs. Accord-
ingly, a person was elevated to the chair if he or she had the
longest continuous service on the committee. The benefits of
this rule are twofold. First, the chair will be occupied by some-
one knowledgeable in the committee's jurisdiction, familiar
with interest group and executive branch players in the wider
Washington community, and politically experienced. Second,
the larger institution will be spared leadership contests that
often reduce the legislative process to efforts in vote grubbing
by contenders. There are costs, however. Senior individuals
may well be knowledgeable, familiar, and experienced, as sug-
gested above; but they also may be unenergetic, out of touch,
even senile. Even when these liabilities do not appear, senior
me mber s may nevertheless be out of step with their commit-
tee and the parent chamber. From the time of the 1965 Voting
Rights Act to what the press called the Republican Revolution
of 1994 (when Republicans won majorities in the House and
Senate for the first time in nearly half a century), old-
fashioned southern conservatives have been a declining force
within the Democratic Party; even so, those who remained
benefited from a seniority system that elevated them to chair-
is Subcommittee chairs do essentially the same things in their narrower juris-

dictions, so I won't provide a separate discussion of them.
manships. I t was thus not at all unusual for a committee con-

sisting chiefly of northern and "New South" Democrats to be
run by a southern dinosaur who had been around for thirty
years.
Different legislatures make the trade-off differently be-
tween seniority-rule automatic elevation, with its profile of
benefits and costs, a nd leadership election, the main a lterna-
tive to seniority. The U.S. House operated according to a strict
seniority principle from about 1910 until the mid-1970s, when
most members felt the burdens of this arrangement were be-
ginning to outweigh its advantages. 19 With Democrats control-
ling the House since 2006, committee chairs are now elected
by the majority-party members of the full legislature, though
there remains a presumption (which may be rebutted, of
course) that the most senior committee member assumes the
chair. Thus, the House added leadership selection to the other
control mechanisms employed by the parent body over its sub-
units.
If committee chairs are the principal vehicles through
which the activities of subcommittees are coordinated, then
party and other institutional leaders from the full legislature
are the ones to coordinate the activities of committee chairs.
The Speaker of the House, Majority Leader, and Minority
Leader, all elected either by the full house or their respective
party caucuses, a re granted responsibilities and authority
aimed at making sensible use of the institution's resources.
They don't always have a completely free hand, just as com-
19
In 1972 the Democratic Caucus approved caucus· wide election of committee
chairs, though it was assumed at the time that no incumbent. previously el-
evated to a chair because of seniority, would lose his or her chairmanship.
This, in fact occurred, though John McMillan (D.S.C.), chair of the District
of Columbia Commi ttee and one of those crusty southerners referred to in
the text, attracted considerable opposition. After the 1974 election, however,
three committee ch airs were defeated (and a fourth resigned to avoid de·
feat). When the Republicans captured the House in the 1994 election, they
abandoned both seniority and election. effectively allowing their leader to
appoint chairs.
Legislatures 401
mittee chairs do not, but they are delegated responsibility

from the chamb ers and the parties to choreograph the pro-
ceedings; the electoral mechanism by which they hold on to
their jobs is designed to make sure they are responsive to rel-
evant institutional constituencies.
C ONCLUSION
Much of the original work reported in Part II on the theory of

majority rule was done by (political) economists. They thought
they understood the political world once they grasped how ab-
stract majority rule worked (or failed to work). Many years
ago one remark ed to the author in a private conversation,
"Now that we know all about median voters and that sort of
thing, we have a pretty good understanding of how the House
of Representatives works." Starting with this chapter, we hope
to convey the na1vete of that sentiment. It's one thing to think
of a legislature as n politicians taking votes by majority rule.
Indeed, it is just like thinking of a firm as personified by a
boss single-mindedly pursuing profits. Each of these is an ab-
straction that may be useful for studying legislatures in the
context of a larger political system or firms in the context of a
larger industry. But once one shifts the magnifying glass from
political system and industry to legislature and firm, these
particular abstractions are no longer nearly as serviceable.
Structure and procedure, an institution's lifeblood, become
n1uch more central to understanding life on the inside of an
institution. The divisional structure of the firm, or the com-
mittee structure of a legislature, abstracted away in more
general studies, take on a much greater descriptive and ana-
lytical salience. And once one opens his or her mind to these
features, issues of authority and control become important. In
short, the observations of some economists to the contrary
notwithstanding, there's no getting away from politics.
The characterization of a legislature thr oughout this chap-

ter, it is only fair to note, describes the practices of some of the
world's legislatures-particularly the U.S. House and Senate,
the legislatures of the U.S. states, and the national and re-
gional legislatures in parts of Latin America. The British
House of Commons and most of the continental European leg-
islatures have a much less internally elaborated structure.
They are more centrally organized, with a prime minister and
his or her cabinet calling most of the shots. Because this
arrangement is radically different from the American-style
legislature, I will take it up separately in Chapter 16.
1. Look at the preferences over issues x and y in the seven-

person legislature shown in Figure 12.1. The utility of legisla-
tor i is maximized at i's ideal point and decreases with dis-
tance from that point. Consider three common institutional
features of legisla tures: committees of members with proposal
power, specific jurisdictions attached to committees, and rules
of amendment once a committee has sent a bill to the full leg-
islative body. Suggest one possible partitioning of legislatures
into committees and structure of jurisdictions for those com-
mittees that would give rise to a stable, predictable equilib-
rium under a closed rule. What would the likely outcome be
under an open rule, assuming that a germaneness rule is in
effect (all proposed amendments must relate to the substance
of the original biH)?
Would a change in the committee membership invalidate the

equilibria you just found under the closed or open rule? Could
Legislatures 403
FIGURE 12.1
DIMENSION Y
•7
•4 •5 •6
•3
• •2
1
DIMENSION X
it invalidate . the existence of any equilibrium? Answer the

same questions for the jurisdictional setup and germaneness
rules. How does this exercise illustrate the link between insti-
tutional rules and stable decision making?
*2. Figure 12.2 represents a stylized legislature with three

blocs, 1, 2, and 3, whose preferences on a two-dimensional is-
sues space are shown. Any two of the three legislative blocs
constitute a majority for the purposes of passing legislation.
A legislative committee that consists solely of members of
group 3 has proposal power, but the legislature operates
under an open rule. Suppose that the status quo is at point q.
If the committee "opens the gates" and proposes a point like p
to the whole house, could it achieve final passage of that bill?
Would the committee be guaranteed final passage of a bill that
it prefers to q? If not, what might another plausible outcome
be? (Hint: Check the winsets of various point.)
FIGURE 12.2
DIMENSION Y
•2
•p
•1 •q •3
DIMENSION X
Now suppose that there is a rule which grants the members of

the committee, 3, an after-the-fact veto, allowing it to reject
any final bill passed and r einstate the st atus quo. If the com-
mittee "opens the gates," proposing p, would it be guaranteed
final passage of a bill that it prefers to q? Comparing the two
outcomes, what role does the institutionalization of committee
power play in matching legislator interest with influence?
*3. A per sistent feature of American political life is legislative

gridlock, an inability to change the status quo (even during
periods of u nified party government) that is not consistent
with the McKelvi an vision of political chaos and instability.
Keith Krehbiel8 proposed a novel explanation for gridlock
using a unidimensional spatial model combined with some
institutional details. Assume a liberal president with ideal
• Keith Krehbiel, Pivotal Politics: A Theory of U.S. Lawmaking (Chicago: Uni-

versity of Chicago Press, 2008), Chapter 2.
Legislat ures 405
FIGURE 12.3
0 p v c f 1
point p , and a moderate Congress controlled by the president's

party in which both houses have median ideal points c. Con-
gress can override a presidential veto if two-thirds of the
members vote to do so: The ideal point of the pivotal member
of Congress needed to override a veto lies at u. Finally, note
that most bills can only escape the Senate with a vote for clo-
ture (to close debate ending a filibuster). The ideal point of the
sixtieth Senate vote needed to close debate is located at f. Sup-
pose that these "pivot points" yield the spacing along the sin-
gle policy dimension provided in Figure 12.3.
If the status quo lies between c and f, can the Congress secure
implementation of any preferable law? What if it lies between
u and c, or p and u, or to the right off? How does this help us
understand the circumstances that generate gridlock?
4. The House and Senate delegate aspects of policy making

from the full chambers to a variety of committees and subcom-
mittees, some of whose members are quite unrepresentative of
the Congress as a whole. Why does the Congress delegate au-
thority to these specialized committees? What are the costs
associated with delegation, and what institutional tools has
the Congress developed to ensure that committees "stay in
line," acting in accord with the interests of the entire legisla-
tive body?
5. Reacting in part to a sharp decline in "party voting'' during

the 1960s and early '70s (i.e., the extent to which parties vote
as unified blocs on important pieces of legislation), the Demo-
cratic party, which controlled the House at the time, instituted

several policy changes to increase the strength of party lead-
ership. These included greater leadership role in determining
committee assignments, expanion of the system of whips
(tasked with vote counting and arm twisting of Democratic
caucus members) and Democratic leadership committees, and
increasing leadership control over committee referral and
amendment rules for bills.b
In what ways does cooperation among legislators suffer in

general from the cooperation and coordination problems dis-
cussed in Chapter 9? What role does party leadership play in
ameliorating these problems, and what did the decline in
"party voting" suggest about House leadership before the re-
forms? Why would representatives with little hope of rising to
the top of the leadership hierarchy go along with reforms
strengthening House leadership at the expense of committees
and committee ch airs?
"A thorough discussion of these changes and their implications is provided in

David W. Rohde, Parties and Leaders in the Postreform House (Chicago: Uni-
versity of Chicago Press, 1991), Chapter 4.
13
Bureaucracy and
Intergovernmental Relations
The term bureaucrat, a perfectly innocent word describing a

government official or civil servant (and also sometimes used
for clerks and managers in the private sector), has acquired
disagreeable connotations of heartlessness and sluggishness.
At the same time that we accept the pursuit of self-interest by
the butcher, baker, and candlestick maker, we expect our pub-
lic servants to be motivated by a higher calling and are merci-
less in our condemnation when they appear to let us down. We
appreciate that the people we interact with in the market-
place are, like the rest of us, trying to make an honest living,
and coping with mortgages, car payments, and college tuition
for the kids. While we hold some professionals in the private
sector to a stricter standard-priests, physicians, and thera-
pists come to mind (perhaps professors, too)- we rarely insist
on self-sacrifice to the exclusion of all else. Public servants, in
contrast, are expected to be on call, ready to serve expedi-
tiously and efficiently, and certainly not afflicted with any of
the personal burdens or ambitions the rest of us carry around.
This idea in the popular consciousness, that government
e mployees somehow are expected to shift moral and behav-
ioral gears when they are transformed from private citizen to
public servant, is unrealistic, to be sure, and unfair to those in
public service who probably do not perform with significantly
407
less devotion to their jobs than the r est of us. Nevertheless,

this view persists, as do the high expectations that are shat-
tered when civil servants fail to perform as demanded. More-
over, given its emphasis on judging performance, this view
has provided a largely useless basis for analyzing bureau-
cratic behavior.
In this chapter I want to unleash our rational choice en-
gine on this s ubject. In order to analyze bureaucratic behavior
properly, it is necessary to embed it in the larger poli tical sys-
tem. I begin by examining one model of bureaucratic behavior
carefully. It is probably the most famous such model, empha-
sizing the bureaucratic pursuit of resources controlled by
elected politician s. After that I examine, though briefly, some
alternative approaches to the bureaucrat-politician connec-
tion. Thus, this s tudy of bureaucracy is really a way to inves-
tigate the relationship between policy for mulators (primarily
legislators) and policy implementers (primarily bureaucrats),
or what we r efer to as intergovernmental relations.
I NTERGOVERNMENTAL R ELATIONS, I :
B UDGET-MAXIMIZING B UREAUCRATS AND
A P ASSIVE L EGISLATIVE SPONSOR
Motivational Considerations
The popular view of bureaucrats as failing to live up to the ex-
pectation that they will serve the public interest is judgmental
and emotionally laden, not analytical. In a now classic treat-
ment, the economist William Niskanen rejected this misbegot-
ten view entirely. 1 Instead, he proposed that we consider a
bureau or depart ment of government as analogous to a divi-
sion of a private firm, and conceive of the bureaucr at just as
1 Bureaucracy and Representative Gouernmerit (Chicago: Aldine, 1971).

Bureaucracy and Intergovernmental Relations 409
we would the manager who runs that division. In particular,

Niskanen stipulates for the purposes of modeling bureaucratic
behavior that a !bureau chief or department head be thought of
as a maximizer of his or her budget (just as the private-sector
counterpart is a maximizer of his or her division's profits).
There are quite a number of different motivational bases
on which bureaucratic budget maximizing might be justified.
A cynical (though some would say realistic) basis for budget
maximizing is that the bureaucrat's own compensation is
often tied to the size of his or her budget. Not only might bu-
reaus with large budgets have higher-salaried executives with
more elaborate fringe benefits, there may be enhanced oppor-
tunities for career advancement, travel, a poshly appointed of-
fice, possibly even a chauffeur-driven limousine.
A second, related motivation for large budgets is not mate-
rial compensation but nonmaterial personal gratification. An
individual quite understandably enjoys the prestige and re-
spect that comes from running a major enterprise. You can't
take these things to the bank or put them on your family's
dinner table, but your sense of esteem and stature are surely
buoyed by the conspicuous fact that your bureau or division
has a large budget. That you are also boss to a large number
of subordinates., made possible by a large bureau budget, is
another aspect of this sort of ego gratification.
But personal salary, "on-the-job consumption," and power-
tripping are not the only forces driving a bureaucrat toward
gaining as large a budget as possible. Some bureaucrats, per-
haps most, actually care about their missions. They initially
choose to go into public safety, or the military, or health care,
or social work, or education- as police officers, soldiers, hospi-
tal managers, social workers, and teachers, respectively-be-
cause they believe in the importance of helping people in their
co1nmunities. As they rise through the ranks of a public bu-
reaucracy into management r esponsibilities, they take this
mission orientation with them. Thus, as chief of detectives in
a big-city police department, as head of procurement in the

Air Force, as director of nursing services in a public hospital,
as supervisor of the social work division in a county welfare
department, or as assistant s uperintendent of a town school
system, individuals try to secure as large a budget as they can
in order to succeed in the missions to which they have devoted
their professional lives.
Whether from cynical, self-serving motives or for the no-
blest of public purposes, it is entirely plausible that individual
bureaucrats seek to persuade others (typically legislators or
taxpayers) to provide them with as many resources as possi-
ble. Indeed, it is sometimes difficult to distinguish the saint
from the sinner, since each sincerely argues that he or she
needs more in order to do more. This is one of the nice features
of Niskanen's assumption of budget maximizing: It doesn't re-
ally matter why a bureaucrat is interested in a big budget;
what does matter is that she wants more resources rather
than less.
Niskanen s Model
Imagine, then, a government bureau whose chief is interested
in eliciting as large a budget as his legislative overseers will
appropriate. The legislature, we assume, is inter ested in the
bureau's output-number of crimes solved, quality of weapons
procured, number of patients served, number of juveniles
counseled, number of high school students graduated. But,
like almost any other good or service, this preference for bu-
reau output increases at a decreasing rate (and may even turn
down after it reaches some level). In pecuniary terms, the leg-
islature (or more precisely, its median voter) is prepared to
pay more for the first unit of output than the tenth, and decid-
edly more for the first than the hundred and tenth. Its
willingness-to-pay curve (labeled B for ''budget") is graphed in
Figure 13.1. The quantity (Q) of the bureau's output is plotted
Bureaucracy and Intergo vernmental Relations 411
FIGURE 13.1
$
TC
"Profit"
L o· H Q
on the horizontal axis. The total budget in dollars that the leg-
islature is prepared to pay for various values of Q is on the
vertical axis. It is assumed that the bureaucrat knows this in-
formation about the legislature. Notice that the curve arcs
down, reflecting the marginally diminishing nature of the leg-
islature's preferences.2
A public bureaucracy, like a private firm, must institute a
production process to enable units of output to be produced. A
public hospital, for instance, looks just like a factory in many
2 In Figure 13.1 I have graphed B as turning down for values of Q greater

than Q*. The argument in the text does not require this; it only requires
that if B increases. it does so at a decreas ing rate (so that it has an arced·
down shape).
respects, combining such inputs as land, labor, and capital

(human and physical) into a complex organization in order to
mend broken limbs, operate on diseased hearts, innoculate
the sick, and otherwise care for the ill and prevent the well
from becoming sick. These inputs, as the raging debate on
health care of the last two decades emphatically underscores,
are not free. These are the costs of production for units of
health care. Niskanen assumes, quite conventionally, that the
production process in public bureaus displays diminishing re-
turns to scale, so that per-unit costs are increasing. Figure
13.1 shows such a cost curve (labeled TC for "total cost"),
arced up to reflect the fact that total cost increases at an in-
creasing rate. 3 F or every level of Q produced, TC gives the
total cost in real resources required. Niskanen assumes that
this is private information-that the bureaucrat knows his
own costs but the legislature does not, just as a firm knows its
production costs but the firm's customers do not.
It is assumed by Niskanen that a public bureau must al-
ways cover its costs. If it is allocated $10 million to produce its
output, and if its production process is such that it can only
produce 1000 units of output for this sum (as may be read off
its TC curve), then it cannot promise to deliver more than
1000 units. This means that the bureau can only agree to
produce values of Q in the range in which the legislature's
willingness-to-pay curve lies above the bureau's total cost
curve. In Figure 13.1, this is indicated by the interval [L, H]
(standing for "low" and "high" production, respectively). Thus,
a budget-maximizing bureaucrat, constrain ed to cover produc-
tion costs, will choose the value of Q that is associated with
the highest point on the B curve in the range [£, H] .4
3 Algebraically, we may write the equations for these curves as: B - aQ - bQ2
and TC - cQ + dQ2, where a, b, c, and d are positive constants.
4 ln Figure 13.1, Lis displayed there as zero; this is not necessary (as shown
in Figure 13.2). In terms of the equations in the previous note, a budget
maximizer will choose the quantity Q that maximizes aQ - bQ2 , subject to .B
"' TC, or aQ - bQ2 >= cQ + dQ2.
In Figure 13.1 the value of Q associated with the highest

point on the B curve is labeled Q*. Notice that it lies within
[£, H], and therefore satisfies the cost-covering constraint. 6
Notice also that this yields the bureau a "profit"-the differ-
ence between B and TC when Q* units are produced. Since
this profit depends upon the bureau's costs of production
(TC)-something the legislature is ignorant about- legisla-
tors have no way of determining the bureau's "profit margin."
However, because the bureau is a public entity there are no
"owners" who can pocket this profit. Instead, the bureau has
some "pin money' with which it can enable its managers to
"consume on the job" (nice offices, posh furnishings, more staff
to boss around, new desktop computers, resources to send
managers to conferences in pleasant locations). 6
In Figure 13.2 we have, owing to different param eter val-
ues for the willingness-to-pay (B) and production cost (TC)
equations (see footnote 3), a very different situation. The un-
constrained maximum of the willingness-to-pay function, Q*,
lies outside the [L, H] interval. Given this bureau's production
costs, it cannot secure the largest budget its legislative spon-
sor is prepared to provide and cover the costs required to pro-
duce Q* units. Its costs are simply rising too steeply. The best
a budget-maximizing bureau, which must cover its costs, can
do is given by Q** in that figure-it represents the production
level associated with the largest budget consistent with cost-
covering. 7 As is shown in that figure, Q* is well enough de-
fined; it simply lies outside the feasible set of production
possibilities available to the cost-constrained bureau.
5 The unconstrained maximum of B is derived by talring its first derivative
and setting it to 0. This yields a - 2bQ- 0. Solving for Q gives Q* - al2b. So,
the most the bureau will produce, given sponsor willingness-to-pay function
B, is a/2b units.
6 If there were a way to pocket this, through bonuses, extra raises, or extra
vacation time, theo no doubt the bureaucrats would find a way. This is typi·
cally precluded in 1:;he public service of most countries by civil service law.
7 This is the value of Q for which B - 'l'C. Substituting for B and 'l'C (as de·
fined in footnote 3) and solving this equality for Q yields Q** - (a - c)/(b +cf).
FIGURE 13.2
$
TC
L o0 H o· 0
<= o~>
Niskanen compares the demand-constrained level, Q*, and

the cost-constrained level, Q**, to what he defines as the so-
cially optimal level, QO. This production level is the one that
maximizes the difference between willingness to pay and total
cost (and is shown on both Figure 13.1 and Figure 13.2). Thus,
it is akin to the production level that would be chosen by a
profit-maximizing firm. On the assumption that "society" and
not the bureau keeps this profit, it represents the point of
maximum net social satisfaction. Niskanen demonstrates, as
the previous figures make clear, that the socially optimal level
of bureaucratic production is less than either the demand-
constrained level or the cost-constrained level-QO s Q** s Q*.
That is, bureaus with a propensity to seek budgets as large as
possible also have a propensity to produce too much.
Niskanen concludes, therefore, that public bureaucracies
are too big, their budgets too large, and their social output
Bureaucracy an.d Intergovernmental Relations 415
more than society wishes. Relative to the willingness of their

representatives to pay for units of Q, society is burdened,
according to Niskanen, with too many weapons; too many
police, welfare, and health services; too much money spent on
education, and so on. The government, driven by budget-max-
imizing bureaus and political overseers who (passively) reveal
their willingness to pay, is simply too big.
Variations on Niskanen
Niskanen's model of bureau behavior, although attractive for
a variety of reasons, is certainly not without controversy.
Whether or noit one "likes" the conclusions of a neoclassical
economist intrinsically suspicious of government, Niskanen,
to his credit, has laid all his assumptions out on the table
in full view. It is thus up to the critic to point to specific fea-
tures of the model that, if changed, would produce different
conclusions-not merely to criticize the conclusions. In this
way debate is focused on analytical or empirical features of a
theoretical argument, not on matters of taste and personal
ideology.
I will mention a few changes that might be considered,
though I hardly have the space to elaborate on forty years of
reaction to Niskanen's famous book. The first place to start is
with the matter of motives. I have defended Niskanen's
budget-maximization hypothesis from a variety of perspec-
tives, that should, at the very least, suggest to the reader that
it is not entirely off the mark. Nevertheless, there are alterna-
tives, and they are likely to produce very different views of bu-
reaucratic phenomena.
For example, suppose that bureaucrats cared not a whit for
their bureau's mission but mainly about the perquisites of
bureaucratic life they might enjoy-the so-called on-the-job
consumption that could be financed out of bureau revenues. If
they are able to camouflage their costs of production from the
legislators or taxpayers who provide their budget (something

Niskanen assumes), then instead of maximizing B, suppose
they seek to maximize B - TC (allowing them to spend the dif-
ference on making life on the job comfortable).8 This will lead
to a bureau output of Cf, precisely the socially optimal level of
production identified by Niskanen (but the "profit" is not re-
turned to the treasury). In this instance, the bureau operates
at the "right" level, but does so inefficiently-society, so to
speak, has to bribe the bureau to operate there by allowing
the bureau to keep its profits. Under this motivational sce-
nario, then, bureaus are not too large-they do not produce
"too much" output-but they are too expensive. This, in turn,
suggests that the root of the problem resides in the secrecy
that protects the bureau's costs of doing business, a theme to
which I return shortly.
Alternatively, suppose that bureaucrats are interested nei-
ther in operating at full throttle (maximizing B) nor in se-
curing maximal perquisites (maxi mizing B - TC). Instead,
suppose they are interested in "the quiet life," engaging in
what has been termed "minimal squawk behavior. 9 It is often
alleged in popular discussions, for example, that the personnel
composition of the public sector is a product of self-selection.
People seek out appointment to the civil service for its secu-
rity (twenty or thirty years of steady work, no heavy lifting, a
secure pension with the promise of an early and comfortable
retirement). They are not the competitive sort who wish to
expose themselves to the dog-eat-dog rivalry of the private
marketplace. Nor are they eager to rock any boat (especially
bureau managers closing in on retirement). "Steady as she
goes" is the watchword. Such people tend to be risk averse, in-
8 In the literature on bureaucratic performance, B - TC is known as organiza-

tional slack.
9 See Clare Leaver, "Bureaucratic Minimal Squawk Behavior: Theory and Ev-
idence from Regulatory Agencies," American Economic Review 99 (2009):
582- 607.
terested mainly in a modest pace, incremental change, and a

commitment to routines and standard operating procedures. 10
Although I have hardly put this description into analytical
form, it seems likely that bureaucrats defined in this way will
underproduce relative either to Niskanenian bureaucrats (Q*
or Q**) or to the social optimum (QO).
The point he re, and I won't dwell on it further, is that it
is quite possible to generate the conclusion that bureaus
overproduce relative to the social optimum, underproduce rel-
ative to the social optimum, or produce exactly at the social
optimum (but do so inefficiently). Analyzing politics in this
manner requires the analyst to t hink hard about which moti-
vational hypothesis is the more appropriate in any particular
application. 11
I NTERGOVERNMENTAL R ELATIONS, II:

B ILATERAL B ARGAINING
There is a second major variation on Niskanen's model of bu-

reau behavior, one that leaves the budget-maximizing motiva-
tion in place but calls into question two other features of his
formulation: the asymmetry in information between bureau
and legislature and the passivity of the legislature. Recall
that Ni skanen assumes that the willingness-to-pay schedule
of the legislature (as represented by its median member),
graphed as B in figures 13.l and 13.2, is commonly known.
Thus, the legislature, the only customer of the bureau's prod-
10 A good example is the Federal Trade Commission (F1'C) of the 1950s and
1960s. Today it is a ve1·y activist regulatory commission. But then it was
staffed by political cronies of the chairman of the congressional oversight
committee (the so·called "Tunnessee crowd") and, because it aimed mostly
to keep its head down and not make waves, the FTC was celled, among
other sobriquets, "the gentle matron of Pennsylvania Avenue."
11 Of course, it is likely that no single motivational hypothesis applies every·
where equally well.
uct, in essence tells the bureau how much it is willing to pay

for various production levels, Q. The chief critics of Niskanen
on this issue, Gary Miller and Terry Moe, suggest that this is
akin to a customer walking onto a used-car lot and telling the
salesman precisely how much she is maximally willing to
spend for each of the vehicles. 12 On the other hand, the bu-
reau's production cost schedule, graphed as TC in figures 13.1
and 13.2, is assumed by Niskanen to be private information,
known by the bureau but not by the legislature. To continue
the parallel, the u sed-car customer has no knowledge of what
the salesman paid for the various models on his lot. In effect,
Niskanen's approach can be characterized as a model of a mo-
nopoly bureau with private information and a passive sponsor.
We have seen in the preceding section what happens in
this situation. Whatever the bureau's motivations, it can ex-
ploit knowledge of its customer's preferences, the privacy of
the details of its own production, and the passivity of its cus-
tomer. But we have also seen, in the previous chapter, that
legislatures are able to anticipate implementation problems of
this sort. Surely it does not take long before legislators realize
that if they neither discover the bureau's production parame-
ters, nor disguise their own willingness to pay for bureau out-
put, nor counteract their own passivity, 13 then they will be
hoodwinked every time by the permanent bureaucracy. 14
In a representative democracy it may be difficult for the
legislature to keep silent on its own willingness to pay. The
bureau, at any rate, can do some research in order to judge
12
This and other related points are drawn from Gary J . Miller and Terry M.
Moe, "Bureaucrats, Legislators, and the Size of Government," American Po·
litical Science Review, 77 (1983): 297-323.
13
This takes the form in Niskanen's model of the bureau being permitted by
its legislative sponsor to select, from the latter's willingness-to-pay sched·
ule, a budget, B, And associated quantity, Q, of deliverable bureaucratic
product.
' ' Much like the parliamentary ministe r is manipulated by his wily senior
civil servants in the British television series, Yes, Minister.
the preferences of various legislators based on who their con-

stituents are. But legislators can do research, too. Indeed, I
suggested in the previous chapter that the collection, evalua-
tion, and dissemination of information- in this case, informa-
tion about the production costs of bureaucratic supply- are
precisely the things in which specialized legislative commit-
tees engage. Committees hold hearings, request documenta-
tion on production, a ssign investigatory staff to various
research tasks, and query bureau personnel on the veracity of
their data and on whether they employ lowest-cost technolo-
gies (making it more difficult for the bureau to disguise on-
the-job consumption). After the fact, the committees engage in
oversight, making sure that what the legisla ture was told at
the time authorization a nd appropriations were voted actually
holds in practice. In short, the legislature can be much more
proactive than Niskanen gives them cr edit for. And, in the real
world, it is.
Miller and Moe, for example, suggest that instead of an
instance of a monopoly bureau facing a passive legislative
sponsor-Niskanen's formulation- the reality in American
politics is much more that of bilateral monopoly: a single cus-
tomer (the legislature as represented by one of its specialized
committees) ba rgaining with a single supplier (the bureau),
each of which h as information (about willingness to pay a nd
production costs, respectively) that may or may not be known
to the other party. In effect, Miller and Moe are saying that
Niskanen failed to model the legislature as an active player in
the making of policy.
What happens in the circumstance in which the legislature
is mor e than passive depends upon the precise way it institu-
tionalizes interactions with the bureaucracy. Miller and Moe
show (the reader must consult their article for the details)
that it need not follow that the bureau is in the driver's seat.
An a uthorization or appropriation or oversight committee able
to discover, through the hearing process, the parameters of
bureau production, or able to disguise its own willingness to

pay, can force the bureau into revealing a menu of production
possibilities from which the legislature can then choose. In ef-
fect, Miller and :tvioe turn Niskanen on his head; instead of the
legislature revealing information from which the bureau
chooses in order to maximize budget size (or slack or quiet
life), Miller and l\1oe have the bureau forced into revealing the
intimate details of its operation, allowing the legislators to
make choices in pursuit of their own objectives. Whether,
under this alternative scenario, government bureaus are too
big-producing too many units of their product-or too
small-producing too few-will depend on a much more
complicated set of considerations than those suggested by
Niskanen.
I NTERGOVERNMENTAL R ELATIONS, III:

P RINCIPALS AND AGENTS
In both Niskanen's model of a budget-maximizing bureau with

a passive legislative sponsor, and Miller and Moe's model of a
budget-maximizing bureau with a proactive legislative spon-
sor, it is presumed that the structure of intergovernmental re-
lations is fixed in advance-as though it were some sort of life
form discovered by the Puritans when they arrived in the New
World. The reality, however, is quite different. In the United
States the executive bureaucracy, piece by piece, was created
by legislative enactment. Acts of Congress created the execu-
tive departments, regulatory agencies, intelligence agencies,
and auxiliary entities like the Office of Management and Bud-
get and Federal Reserve System. An act of Congress reorgan-
ized the Joint Chiefs of Staff. An act of Congress combined a
host of health agencies and bureaus into the National Insti-
tutes of Health. Acts of Congress directly or indirectly created
the alphabet soup agencies of the New Deal and World War II;
subsequent acts of Congress dismantled many of them.
So, while presidential signatures were required on these
congressional enactments and affirming judicial rulings were
needed to keep them alive, Congress can lay claim to pater-
nity (and maternity), birthing, and wet-nursing responsibili-
ties for the structure of the federal bureaucracy. Moreover,
Congress passed various pieces of legislation regarding both
personnel practices and administrative procedures, which
serve to regulate entry into, and advancement through, the
civil service on the one hand, and the official procedures by
which civil servants conduct bureaucratic business on the
other. Finally, as I emphasized in the previous chapter, Con-
gress maintains a constant presence in the affairs of the bu-
reaucracy through
• the drafting of laws revising and extending bureaucratic
authority;
• the approval of appropriations for the various divisions
of government; and
• the subsequent ("continuously watchful") oversight of the
uses to which this authority and these resources are put.
It is difficult to square Niskanen's view of a dominant bu-
reaucracy manipulating its passive legislative masters with
this litany of facts. The bureaucracy is created by Congress
and sustained by Congress. Although there are bound to be oc-
casions in which a bureau chief pulls the wool over a legisla-
tor's eyes, and t here are surely times in which a department's
budget people use "blue smoke and mirrors" to misdirect con-
gressional inspectors, it is unlikely, I believe, for these abuses
to persist for ve·r y long-certainly not without the implicit ap-
proval of congressional players in the bureau's jurisdiction.
In fact, it is even something of a strain to square with
these facts the much more balanced view of Miller and Moe in
which legislature and bureau bargain more or less as equals.

They restore a modicum of symmetry to intergovernmental re-
lations, allowing !for monopsony legislative consumer and mo-
nopoly bureaucratic producer to be in rough bargaining parity
(especially concerning information about preferences and pro-
duction capability). But in doing so, they suppress the very
consequential asymmetry described in the previous two para-
graphs-namely, the legislature's awesome power to create
and sustain, or destroy, bureaucratic organization.
Notice what these two approaches postulate. The first,
Niskanen's, asserts that it is the bureaucrat who is in the dri-
ver's seat; she is constrained by the legislature to be sure (she
must cover costs and accommodate herself to the legislature's
maximum willingness to pay), but she gets to choose Q from
the menu of alternatives presented to her in the legislature's
willingness-to-pay schedule. The second approach, Miller and
Moe's, claims that there is much greater balance in the rela-
tionship between bureau and legislature, so that rather than
one choosing subject to constraints, it is more akin to bilateral
bargaining (as, say, when Amalgamated Coal bargains with
the United Mine Workers over the price and quantity of
labor). There is, however, an alternative to both the Niskanen
and Miller-Moe formulations, one that emphasizes the subor-
dinate role of the bureaucracy and the superior position of
elected politicians. 15 To describe this it is necessary to digress
briefly on a general consideration of principals and agents.
15
Throughout, I will suppose that these elected politicians are legislators.
There is, in fact, an entire literature, known as the "Congressional Domi-
nance School," that explicitly stipulates that the bureaucracy is a creature
of legislative preferences. I would want to quali fy this, however, by empha-
sizing that other elected and appointed politicians-€specially the president
and judges-influence bureaucratic activity, too.
Principal-Agent Relationships
Most people hire a doctor, lawyer, or contractor to tend to their
health, to draw up their wills and other legal papers, or to
renovate their !kitchen, respectively. In principle we could do
these things our selves, but in practice we find it more satisfac-
tory to spend our time and energy on other things. This is just
another way of saying that most of us reject self-sufficiency
and accept the superiority of market exchange. In effect, we
retain agents to act in our interest, agents whose specialized
knowledge and skills make them more effective in doing our
bidding than we could ourselves. In these relationships, each
of us is a principal, and the problem we face is that of control-
ling our agents.
When we hire a kitchen contractor, agreeing to pay him a
particular sum of money in exchange for renovation services
(typically stipulated in pages and pages of detailed specifica-
tions), we face the problem that we don't know how good he is
and often aren't able to detect the quality of workmanship at
the time it is performed. Suppose he used aluminum wiring
rather than copper wiring, something we unhappily discover
after a fire. Suppose he said he would use copper piping but in
fact used plastic piping (obtained cheaply from his brother-in-
law and hidden out of sight in the walls), something detected
only years later when the less reliable plastic pipe springs a
leak. Suppose he claimed to know how to assemble and install
the extra-fancy gourmet cooking surface we purchased, only to
concede (after our kitchen had been torn apart) that it was
more complicated than he had figured.
There are two broad categories of control mechanism en-
abling a principal to guard against opportunistic or incompe-
tent agent behavior. The first is employed before the fact and
depends upon the reputation an agent possesses. One guards
against selecting an incompetent or corrupt agent (one who
cannot or will not perform with the principal's interests at
heart) by relying on various methods for authenticating the

promises made by the agent. These include advice from people
you trust (your neighbor who just had her kitchen remodeled);
certification by various official boards (county medical society,
bar society, association of kitchen contractor s); letters of rec-
ommendation and other testimonials, credentials (law or med-
ical degrees, specialized training programs); and interviews.
Before-the-fact protection relies upon the assumption that an
agent's r eputation is a valuable asset that he or she does not
want to see depreciated.
The second class of control mechanisms operates after the
fact. Payment may be made contingent on completion of vari-
ous tasks by specific dates, so that it may be withheld for non-
performance. Alternatively, financial incentives (for example,
bonuses) for early or on-time completion may be part of the
arrangement. The agent may be required to post a bond that
is forfeited for lack of performance. An inspection process,
after the work is completed, may lead to financial penalties or
bonuses or possibly even legal action. Of course, the principal
can always seek legal relief for breach of contract, either in
the form of an injunction that the agent comply or of an order
that the agent pay damages.
This brief discussion identifies principal-agent relation-
ships as arising because there are genuine a dvantages for in-
dividuals to specialize in their activities and trade with one
another (in-kind or for cash) rather than depend on self-
sufficiency. But balanced against these conspicuous advan-
tages (none of us would want to set her own broken arm or
install his new dishwasher) is the obvious fact that an agent
does not work out of the generosity of her spirit-she has bills
to pay, a mortgage to service, and kids to educate, too. She
works for herself a s well as working for you, and this can lead
to potential conflicts of interest. While control mechanisms
exist, they are not perfect.
In a principal-agent r elationship it is the principal who

stipulates what. he wants done, relying upon the agent's con-
cern for her reputation, appropriate incentives, and other
control mechanisms to secure compliance with his wishes.
Analogously, it may be argued that legislative principals es-
tablish bureaucratic agents- in departments, bureaus, agen-
cies, institutes, and commissions of the federal government-
to implement the policies promulgated by Congress and the
president. Like any principal with an ounce of intelligence,
the legislature would surely not create an agent in the image
of Niskanen, one that could exploit the legislature for the
agent's own benefit. Indeed, the legislature would not even
wish to establish a bureaucratic agent with equal standing in
a bilateral bargaining situation in the image of Miller and
Moe. The legislature wants a compliant agent and will do its
best to "hardwire" control mechanisms in the enabling legisla-
tion that creates the bureaucratic entity in the first place.
This is precisely the argument made in a series of impor-
tant papers by Mathew McCubbins, Roger Noll, a nd Barry
Weingast, known collectively as McNollgast. 16 For them, a
piece of legislation creating a new agency or assigning some
new mission to an existing agency creates a principal-agent
relationship betwee n an enacting coalition, consisting of legis-
lators in the two houses of Congress and the president (them-
selves agents for constituents), and a bureaucratic entity. The
enacting coalition has coordinated around a policy objective,
seeking not only its faithful implementation, but also an
arrangement possessing durability (extending beyond when
16 These three authors have now written a large number of papers. The two
that will get the interested reader started are "Administrative Procedures
as Instruments of Political Control," Journal of Law, Economics, a1id Orga-
nization 3 (1987): 243-79; and "Structure and Process, Politics and Policy:
Ad ministrative Arrangements and the Political Control of Agencies," Vir-
ginia Law Review 75 (1989): 431- 83.
the members of the particular enacting coalition have de-

parted from the scene.) 17 But this is not as easy as it seems, as
the following little argument suggests.
Suppose the original enabling legislation that created the
Environmental Protection Agency (EPA) required that new
legislation be passed after ten years renewing its existence
and mandate. The issue facing the House, the Senate, and the
president in their consideration of renewal revolves around
how much authority to give this agency and how much money
to permit it to spend. The House, quite conservative on envi-
ronmental issues, prefers limited authority and a limited
budget. The Senate wants the agency to have wide-ranging
authority but is prepared to give it only slightly more resources
than the House (because of its concern with the budget
deficit). The president is happy to split the difference between
House and Senate on the matter of authority but feels be-
holden to Sierra Club types in this election year and thus is
prepared to shower it with resources. Bureaucrats in the EPA,
for all the reasons we provided earlier in this chapter, want
more authority than even the Senate is prepared to condone
and more resources than even the president is willing to
grant. These preferences are displayed in Figure 13.3, where
H , S, and P represent the ideal points of the political princi-
pals, and B the ideal point of the bureaucratic agent. An
enacting coalition-relevant majorities in the House and Sen-
ate (including the support of relevant committees) and the
president-agrees on a policy, x, reflecting a compromise
among their various points of view-something close to what
17 Why, the reader might wonder, would poli ticians care about what happened
after they retired Crom political life? The reason is that the constituencies
they represent, and who reward them in the here and now for "good deeds,"
have long time horizons. Since they care about the long term, they are pre·
pared to reward politicians who attend to the long term. Thus, politicians
will do their bidding, thereby attempting to influence policy even after their
active political life.
FIGURE 13.3
RESOURCES
AUTHORITY
the president wants on the authority dimension and close to

what the Senate wants on the resources dimension.
The bureacrats are not particularly pleased with x, since it
gives them considerably less authority and many fewer dollars
than they had hoped for. If they flout the wishes of the enact-
ing coalition and implement a policy more to their liking- say,
y- then they risk the unified wrath of the enacting coalition.
For any policy outside the triangle connecting the ideal points
of the House, Senate, and president, there is some other policy
they unanimously prefer to it. Undoubtedly the politicians
would react to t.he bureaucratic implementation of y with new
legislation (and also presumably would find other political ap-
pointees and career bureaucrats at EPA to replace the current
bureaucratic leadership). If, however, the EPA implemented

some policy on or inside this triangle in Figure 13.3, they
might be able to get away with it. For any point inside the tri-
angle, departure from it makes either the House or the Senate
or the president worse off. But new legislation to punish an
out-of-control age ncy and its existing leadership requires the
simultaneous support of a House majority, a Senate majority,
and the president. Thus, if EPA implements some policy on or
inside the triangle, they can avoid political retribution.
Following this strategic calculus, the best EPA can do is
implement the policy x'. It is the point closest to B, the
agency's ideal, that satisfies the political requirements to
avoid political reversal. 18 The difference between x and x' is
termed bureaucratic drift by McNollgast.
Thus, we have a principal-agent relationship in which a
political principal formulates policy and creates an implemen-
tation agent to execute its details. The agent, however, has
policy preferences of its own and, unless subjected to further
controls, inevitably will implement a policy that drifts to-
wards its ideal.
A variety of controls exist that might conceivably restrict
the bureaucratic drift. Indeed, legislative scholars often point
to congressional hearings in which bureaucrats may be pub-
licly humiliated; annual appropriations decisions which may
be used to punish "out-of-control" bureaus; and the use of
watchdog agents, like the Government Accountability Office,
to monitor and scrutinize the bureau's performance. But these
all come after the fact and aren't really credible threats to the
agency (since the bureau has nipped those threats in the bud,
so to speak, by strategically implementing x' in the triangle
rather than some more extreme policy more to its liking).
18 If the bureau must observe the financial restraint of not spending more
than has been appropriated (given by the component of x on the vertical di-
mension), then x' will simply slide down the side of the triangle to the point
that is the same height as x.
The most powerful before-the-fact political weapon is the

appointment process. The adroit control of the location of B by
the president and Congress (especially the Senate), through
their joint powers of nomination and confirmation, especially
if they can arrange for appointees who more nearly share the
political consensus on policy, is a self-enforcing mechanism for
assuring realiable agent performance.
A second powerful before-the-fact weapon, one especially
emphasized by McNollgast, is procedural controls. The gen-
eral rules and regulations that direct the manner in which
federal agencies conduct their affairs are contained in the Ad-
ministrative Procedures Act. This act almost always consti-
tutes the boilerplate that is part of the enabling legislation
creating and renewing every federal agency. It is not uncom-
mon, however, for the procedures for an agency required by
the enabling legislation to be tailored to suit particular cir-
cumstances.
For example, a ,political consensus on environmental mat-
ters (policy x in Figure 13.3) might be somewhat more sympa-
thetic to businesses that have to bear the cost of retrofitting
their plants to help maintain environmental integrity than is
preferred by bureaucratic leaders in the EPA (as well as vari-
ous environmental groups). Politicians, who have to live with
the prospect of bureaucratic drift all the time, anticipate that
the bureaucrats in the EPA will want to be more aggressive
than the political consensus warrants. To reduce the prospects
of drift, the politicians can write into the enabling legislation
that the EPA must proceed on a case-by-case basis rather than
by promulgating general rules affecting a wide range of cases.
By insisting on this procedure, the politicians anticipate that
business interests are in a far better position (privileged
groups, in Mancur Olson's terms) to mobilize their supporters
to participate in the public part of agency decision making a
case at a time than are the environmental groups to mobilize
their members. Environmental groups, on the other hand,
would be better served if the agency could make decisions that

were fewer in number but of greater impact and magnitude,
since they would find it easier to mobilize for a few big pushes
than to expose their members to the incremental "torture" of
case-by-case deliberations.
To sum up, the principal-agent perspective on bureaucratic
behavior treats bureaus as agents of politicians, implementing
policies formulated by an enacting political coalition. The
problem for the politicians constituting this coalition is that
bureaucratic agents have missions, interests, and objectives of
their own that may conflict with those of the politicians. In-
deed, clever agents will take into account that to sanction a
bureau that does not do the precise bidding of an enacting
coalition, the component members of that coalition must act
in unison-this is the constitutional requirement in the sepa-
ration-of-powers political order of the United States. As long
as the agent makes sure that no alternative is preferred by
all component members of the enacting coalition to what-
ever it does, it will be spared punishment. (It is a bit like a dog
on a leash who wants to go in a slightly different direction
than its master wishes. As long as the dog makes sure the
leash is not taut, it can wander about as it pleases.) The
McNollgast approach, however, emphasizes an asymmetric
advantage for politicians (in contrast to Miller-Moe and dia-
metrically opposite to Niskanen). The enacting coalition can
anticipate bureaucratic drift and make provisions in the en-
abling legislation to reduce it, if not eliminate it altogether. Of
especial importance in this regard are the exercise of pru-
dence in the nomination and confirmation of political ap-
pointees to lead bureaus, and the stipulation of specific
administrative procedures that make it difficult for the bu-
reau to flout the will of the principals.
CASE 13. 1
CONGRESSIONAL O VERSIGHT:
P OLI CE P ATROLS, FIRE ALARMS,
AND FIRE E XTINGUISHERS
The rational choice approach to intergovernmental rela-

tions has demonstr ated the importance of principal-agent
relationships between Congress and the bureaucracy. Criti-
cal to any principal-agent relationship is monitoring- the
process whereby the principal ensures that the agent is
complying with the former's preferences rather than shirk-
ing by pursuing the agent's own preferences. In the regula-
tory arena we are quite familiar with monitoring devices.
Many readers, for example, are aware of congressional
hearings in which agencies are raked over the coals to jus-
tify their activities and their budgets.
Although these public incidents of explicit monitoring
are the most obvious, they are only part of the picture. Con-
gressional committees and subcommittees are "continuously
watchful" (in the language of the Legislative Reorganization
Act of 1946) in other ways as well. In a classic paper,
McCubbins and Schwartz* distinguish between police-patrol
and fire-alarm oversight. Police-patrol oversight, as the
name suggests, involves centralized and direct surveil-
lance--congressional committees and their staffs "cruise"
their respective jurisdictions looking for bureaucratic
malfeasors. Congress initiates inspections, supervision, and
direct instruction and oversight for its bureaucratic agents.
Fire-alarm oversight, on the other hand, is relatively decen-
·Mathew D. McCubbins and Thomas Schwartz, "Congressional Oversight

Overlooked: Police Patrols versus Fire Alarms," American Journal of Po-
litical Science 28 (1984); 165-79.
tralized, reactive, and indirect. In this instance Congress

establishes rules and procedures that enable private par-
ties (the customers or victims of the agency in question)
to monitor agency practices and bring inappropriate activi-
ties to congressional attention. The service recipients, that
is, can pull the lever on a fire alarm whenever agencies
misuse their authority. (A third form of oversight, not cov-
ered by McCubbins and Schwartz, 1night be termed fire-
extinguisher oversight. Service recipients, in this case, bring
agency malfeasance to the attention not of Congress but to
the federal courts instead. Congressional rules, in effect,
give various categories of service recipients legal standjng
to sue in court for agency abuses.)
Legislators often prefer putting fire-alarm and fire-
extinguisher oversight in place because they are less costly
and time-consuming for the legislature than police-patrol
oversight. Since legislator time and expertise is often quite
limited, having the public monitor agency activities shifts
many of these costs away from Congress and takes advan-
tage of a broader base of expertise. Fire-extinguisher over-
sight has become particularly common in recent years as
the size of the legislative agenda has enlarged and the com-
plexity of the issues on it has increased. Groups, including
the handicapped, minorities, farmers, workers, and small-
business owners, have all been given legal standing in the
jurisdiction of one or another regulatory agency. Thus, it is
fair to say that nowadays the norm is for Congress to legis-
late the requirement that agencies and courts react to com-
plaints rather than to inspect for violations itself.
Principal-agent relationships are ubiquitous in politics,
economics, and social life. Your senator, your professor, your
auto mechanic, and the pilot of your airplane are all your
agents. Which ones are monitored by police patrols? Which
submit to fire -alarm or fire-extinguisher monitoring?
A GENCY SLIPPAGE
We can draw on all three of the approaches to bureaucratic

politics presented in this chapter to point out various manifes-
tations of tension between politicians and bureaucrats. In
civics-book desc1·iptions of political institutions, it is often sup-
posed that official political decisions, as reflected in statutes
passed by Congress, signed by the president, and validated by
various courts, are the final word. The business of politicians
is the formulation of public policy, and constitutional proce-
dures specify how this is done. But policies formulated accord-
ing to even the most rigorous constitutional provisions are not
the final word. Policies are not self-implementing. Someone
must exercise the authority and expend the resources war-
ranted by a statute. This "someone" is the bureaucrat. How-
ever, if the policy actually implemented by a bureaucratic
agent departs from the policy formulated by his or her politi-
cal principals, there is agency slippage or drift. It is produced
by a combination of agent discretion, agent preferences differ-
ent fro m those of the enacting political coalition, and imper-
fect control mechanisms. The three approaches developed in
the preceding pages of this chapter highlight the different
ways this slippage manifests itself.
Budgetary Exploitation
From Niskanen we have seen that specialized bureaucratic
agents are very knowledgeable about the intricate details of
policy making in their jurisdiction, perhaps more so than even
specialists on legislative committees. Even though the latter
are themselves policy specialists, they have many other hats
to wear and roles to fill; they cannot devote the time and en-
ergy necessary to absorb all the arcane minutiae associated
with implementing a specific public policy. For bureaucratic
agents, on the other hand, these details are the core of their
very existence and a constant preoccupation. This produces an

effective informational advantage for the bureaucrat; it also
provides an opportunity. If the bureaucrat's preferences differ
from those of the enacting coalition, then he or she may use
this oportunity to pursue personal priorities. Niskanen's argu-
ment suggests that this may be accomplished by disingenu-
ously attracting t he resources and authority either to pursue
a personal agenda of policy objectives, or to siphon off re-
sources to devote to on-the-job consumption. This is made pos-
sible because, according to Niskanen, the politicians just
cannot wrap their minds entirely around the bureaucratic
production process-they cannot figure out the details of the
relationship between costs and policy outputs. This enables
the bureau to earn a "profit," displayed in Figure 13.1, which
can be put to alternative uses. If this informational advantage
is converted into opportunities for on-the-job consumption, we
have a pure instance of bureaucratic budgetary exploitation.
Bureaucratic Drift
If, on the other hand, authority and resources are used to pur-
sue other policy objectives, then .we have an instance of bu-
reaucratic drift, as described by McNollgast and depicted in
Figure 13.3. This follows from the ability of the bureaucrat to
exercise discretion in his or her pursuit of policy priorities
that may not have been warranted in statutory authority.
This policy drift, moreover, is protected from after-the-fact ret-
ribution as long as the bureaucrat is crafty enough to protect
against unanimous opposition from the components of the en-
acting coalition. In effect, if the bureaucrat can keep this key
senator or that house committee chair content, he or she may
be home free.
Bureaucracy an,d Intergovernmental Relations 435
Bureaucratic Capture
A third source of tension requires that I make some finer dis-
tinctions than heretofore. The bureaucratic drift that McNoll-
gast describes is the difference between the co1nprornise policy
arived at politically (x in Figure 13.3) and the policy imple-
mented bureaucratically (x'). In that figure, however, the care-
ful reader may have noticed that both the Senate (with ideal
point S) and the president (with ideal point P ) actually like x'
better than x. Only the House (with ideal point H) loses from
the drift in this particular example. If we retreat from the
myth that the Senate and House are unitary and possess ideal
points, instead more appropriately assuming that there are
435 ideal points scattered about H and 100 ideal points scat-
tered about S, then it is very likely that a number of represen-
tatives and senators sympathize more with the bureaucrat's
preferences than they do with either the political preference
arrived at in their legislative chamber (H and S, respectively,
in Figure 13.3) or the policy compromise arrived at among
House, Senate, a nd president (x). To push the argument one
step further, taking on board some of the factors I considered
in the previous chapter, it is likely that the committees with
jurisdiction over this bureaucracy and policy area are popu-
lated precisely with legislators who are more likely to sh ar e
the bureaucrat's preferences for larger bureaucratic output
than the preferences of their political colleagues for more
modest levels (i.e., an outlier committee). -They are in a posi-
tion to protect a drifting bureaucracy (through their gatekeep-
ing and other .agenda-control powers). Finally, behind these
legislators are the interest groups and geographic constituen-
cies whose well-being the legislators pursue; they, too, will be
pleased by the drift. Indeed, the traditional political science
literature makes frequent allusion to "cozy little tria ngles," in-
volving legislat()rs on key committees, bureaucrats, and inter-
est groups, as the dynamos behind policy implementation.
They are also referred to as "policy whirlpools" and "unholy

trinities," and the politics this arrangement sustains is called
"interest group liberalism," reflecting effective capture of the
bureaucracy by interests in the wider political world. t 9 In this
case the tension is not between politician and bureaucrat, but
rather among the politicians, bureaucrats, and interest groups
inside the policy whirlpool and those on the outside.
Coalitional Drift
At the very outset of this discussion I noted that politicians not
only want the legislative deals they strike to be faithfully im-
plemented, they want those deals to endure-they want "deals
struck to stay stuck," as the inside-the-Beltway pundits put it.
This is especiaJly problematic in American political life with
its shifting alignments and absence of permanent political
cleavages. Today's coalition transforms itself overnight. Oppo-
nents today are partners tomorrow and the reverse. This
makes dealmakers inside the legislature, and their interest-
group backers in the hustings, insecure. A victory today, even
one implemented in a favorable manner by the bureaucracy,
may be undone tomorrow. What is to be done?
To some extent legislative structure leans against undoing
the handiwork of an enacting coalition. If such a coalition
votes handsome subsidies to grain farmers, say, it is very hard
to reverse this policy without the gatekeeping and agenda-
setting resources of members on the House and Senate Agri-
culture Committees; yet their members undoubtedly partici-
pated in the initial deal and are unlikely to turn against it.
But even these structural units are unstable, old politicians
departing and new ones enlisted. The problem, then, is that
19 Theclassic in the traditional political science literature is Theodore J. Lowi,

The End of Liberalis m, 2nd edition (New York: Norton, 1979). ln economics,
the classic is George Stigler, "The Theory of Economic Regulation," Bell
Journal of Economics and Management Science 2 (1971): a 21.
Bureaucracy and Intergo vernmental Relations 437
the members of an enacting coalition whose product may not

be very durable cannot expect the kind of support from inter-
est groups and voters, in whose behalf they work, that they
might have received if they had produced a more robust re-
sult.
In short, legislatively formulated and bureaucratically
implemented output is subject to coalitional drift. 20 Today's
enacting coalition may no longer be in power tomorrow. To
prevent shifting coalitional patterns among politicians to en-
danger carefully fashioned policies, one thing the legislature
might do is insulate the bureaucracy and its implementation
activities from legislative interventions. If an enacting coali-
tion makes it difficult for its own members to intervene in im-
plementation, then it also makes it difficult for enemies of the
policy to disrupt the flow of bureau output at a later date. This
political insulation can be provided by giving implementing
agencies long lives, their political heads long terms of office
and wide-ranging administrative authority, other political ap-
pointees overlapping terms of office, and security to their
sources of revenue-like many of the "independent" regula-
tory commissions, the Federal Reserve System, and, to some
extent, the intelligence agencies. But it comes at a price. The
civil servants and political appointees of bureaus insulated
from political overseers are thereby empowered to pursue in-
dependent courses of action. Protection from coalitional drift
comes at the price of an increased potential for bureaucratic
drift. It is one of the great trade-offs in the field of intergov-
ernmental relations.
20 This idea, offered as a supplement to the McNoUgast line of argument, is

found in Murray J. Horn and Kenneth A. Shepsle, "Administrative Process
and Organizational Form as Legislative Responses to Agency Costs," Vir·
ginia Law Review 75 (1989): 499-509. It is further elabo rated in Kenneth A.
Shepsle, "Bureaucratic Drift, Coalitional Drift, and Time Consister,cy,"
Joumal of Law, Economics, and Organization 8 (1992): 111-18.
CASE 13.2
How TO TEST NISKANEN?
Most of this book has been theoretical, designed to give you

the ideas, concepts, and theories that comprise the analyti-
cal, or rational choice, approach to politics. In the cases
sprinkled throughout, I have endeavored to provide real-
world exan1ples and applications, both to display evidence
for theoretical arguments and to demonstrate how to apply
various ideas to political situations and events in everyday
life. In the present case we look at how two political econ-
omists set out to distinguish among competing theories of
intergovernmental relations. Barry Weingast and Mark
Moran* employ statistical analysis of data drawn from reg-
ulatory policy making at the Federal Trade Commission to
evaluate some of the competing claims of the theories I
have reviewed in this chapter. Their study is a useful ex-
ample of how to apply empirical analysis to rational choice
theories; it contributes to our understanding of intergovern-
mental relations as they are practiced in the real world.
In 1979, Congress, through the auspices of relevant
committees and subcom mittees, actively intervened to re-
strain the Fedleral Trade Commission (FTC) and to curb
what were seen as regulatory abuses. Policy initiatives
were criticized, FTC activities were halted, and funds were
cut off. Over the next year and a half, the FTC curtailed its
controversial activities and brought itself into alignment
with the preferences of the congressional principals who
oversaw its performance.
A Niskanen-like view of the situation is that the FTC
was a "runaway" agency that used its informational advan-
·Barry R. Weingast and Mark J. Moran, "Bureaucratic Discretion or Con-
gressional Control? Regulatory Policymaking by the Federal Trade Com-
mission," Journal of Political Economy 91 (1983): 765-800.
tage to underwrite activities · well beyond its mandate.

When Congress finally woke up to the fact that the FTC
was living too high off the bureaucratic hog, it clamped
down and brought it back into line. An alternative view con-
sistent with McNollgastian coalitional drift is that Con-
gress was in control of the FTC all along. According to this
view, a change in the composition and preferences of the
oversight committees is responsible for the intervention.
Because the FTC was not sufficiently insulated from leg-
islative intervention, and even though it may well have
been doing the bidding of a previously empowered legisla-
tive coalition, a new coalition of legislators came to power
in the late 1970s and was able to use its oversight tool kit of
hearings, procedures, and budgetary authority to move the
agency's output toward the new ideal policy they repre-
sented. In short, in the late 1970s the FTC's legislative prin-
cipal changed-through electoral turnover, retirements, and
committee reassignments. These new guys yanked on the
leash, pulling tlhe FTC into line with its preferences.
So which is it? Was the FTC a runaway Niskanian bu-
reaucracy exercising policy discretion? Or was Congress al-
ways in control, but a change in congressional preferences
required a wake-up call to the FTC from the new congres-
sional principals? To test these two views, Weingast and
Moran gathered evidence to evaluate whether preferences
of the membership of relevant oversight committees did
change in the late 1970s and whether the new committees
did indeed influence subsequent FTC behavior. The data
examined consisted of interest group ratings of legislators
by the liberal interest group, Americans for Democratic Ac-
tion (ADA) , whose scores, ranging from 0 to 100, measure a
legislator's general disposition toward liberal causes. The
scores for members of the Senate consumer affairs subcom-
mittees indicate that there was a significant conservative
drift in these bodies between 1976 and 1979 after more
than a decade of virtually no change at all. This was accom-

panied by (indeed, it reflected) a substantial turnover of
membership on these panels-liberal members had left
these panels in large numbers to be replaced by much more
conservative senators. The preferences of the oversight
groups, in short, displayed a shift fro m a proactivism in
consumer matters to an antiactivism stance during this pe-
riod. The evidence, then, is quite consistent with the Mc-
Nollgast argument. Moreover, in looking at ADA scores over
the two decades preceding this turning point, Weingast and
Moran find that the oversight panels were much more lib-
eral, according to ADA scores, than in the late 1970s, which
suggests that tine FTC was actually serving as a loyal agent
to the politicians in control at the time.
The conclusion to be drawn from this evidence is that
the FTC was hardly a "runaway" bureaucracy,, exploiting its
asymmetric informational advantage in accord with Niska-
nen's theory. Rather, it appears to have been an agent t hat
failed to notice that its principal had changed in the late
1970s. Once this principal-the more conservative legisla-
tive coalition on the oversight panel in the Senate-yanked
hard on the leash, the agent once more fell into line.
CONCLUSION
It used to be fashionable in political science to distinguish

between "politics" and "administration." Indeed, schools of
public administration and schools of government still exist
in some of our most venerable universities alongside separate
departments of politics or political science. Fortunately, politi-
cal scientists no longer hold to the notion that policy making
and policy implementation can be sealed off from one another.
In this chapter I have suggested several different ways to
think analytically about the connections between politicians
and bureaucrats. Three different, but related, connective argu-

ments have been developed. The first, associated with Niska-
nen, suggested that the permanent bureaucracy exploits its
informational advantages vis-a-vis elected politicians, leading
to bureaucratic budgets too large, bureaucrats too numerous,
and bureaucratic output too abundant. The second, associated
with Miller and Moe, took exception to Niskanen's assump-
tions of political ignorance and passivity. More informed and
proactive politicians yield a more balanced arrangement of bi-
lateral bargaining with bureaucrats, they concluded, and this
need not nurture the maladies identified by Niskanen. Finally,
McNollgast reversed Niskanen's emphasis altogether, suggest-
ing that any asymmetry that may exist favors the political
class who, as principals, create their agents, participate in ap-
pointing their leaders, sustain them with authority and
resources, monitor their activities, and ultimately pass judg-
ment over their actions.
Needless to say, there are important insights and elements
of truth in each of these approaches. This is not a horse race
in which I (or you) need to declare one of these approaches the
winner. Indeed, in combination they draw our attention to a
variety of tensions t hat exist and trade-offs th at must be made
to accommodate the sometimes conflicting priorities and pref-
erences of politicians and bureaucrats. All in all, the politics of
policy and the politics of administration are inseparable and
complicated. The theories summarized here help us to make
sense of how representative policy formulation and expert pol-
icy implementation fit togeth er.
1. The idea of term limits for elected legislators and for chairs
of legislative committees has been hotly debated for a quarter
of a century (while, of course, presidents have been term-
limited since ratification of the Twenty-second Amendment in
1951). In light of the principal-agent framework discussed in
this chapter, how might you attack the idea of term limits?
How would you defend it? Do the sa1ne arguments make sense
for bureaucrats?
2. Niskanen's model of legislative-bureaucratic relations makes

three very strong assumptions that ascribe to the bureaucracy
extraordinary power in securing an oversized budgeta: 1) The
bureau's cost schedule is private information, while the legis-
lature's demand schedule is public knowledge; 2) the bureau is
the only supplier; 3) in its bargaining with the legislature, the
bureau drives the policy-making process by making a single
take-it-or-leave-it proposal to the legislature. Which of these
assumptions do you find unrealistic or overly strong? What so-
lutions have legislatures and other principals adopted to limit
bureaucratic power in bargaining over budgets?
*3. A Niskanian bureaucracy supplies some output Q at the

following cost: TC= l.5Q 2 where costs are measured in tens of
millions of dollars. If the legislature's willingness-to-pay curve
is B = 8Q - 2Q2 , what is the maximum budget that the bureau
can secure? Be sure to consider both demand and cost con-
straints. Now the legislature institutes the following monitor-
ing scheme: With probability p the bureaucracy is audited
" See Dennis Mueller, Public Choice III (New York: Cambridge University
Press, 2003), p. 365.
(which would reveal that t he optimal level of production from

the legislature's point of view is Q0 = 1.2). If the bureaucracy
is caught producing more than Q 0 it is forced to pay a fine of
f (Q - QO) = f (Q - 1.2), which is measured in the same units as
production costs .. Write out an expression for the bureau's ex-
pected total costs when it produces less than or equal to Q 0 ,
and for when it produces more than Q0 . If p = .2, and f - 5,
how much will the bureau produce? What about if p - .5, and
f - 10? Illustrate the intuition behind these answers.
4. Members of Congress and bureaucrats both act as agents

for their principals (voters for the former, the Congress and
president for the latter). However, while representatives must
periodically face election, bureaucrats are appointed and face
no r eelection co nstraint. What is the source of preference
divergence betweeri principals and agents in each of these
cases? Does the presence of elections provide effective control
of legislators? Does the lack of elections prevent effective con -
trol of bureaucracies, or are there other mechanisms? Discuss
t he variety of before-the-fact (both institutional and r eputa-
t ional) a nd after-the-fact controls adopted by voters a nd Con-
gress to limit agent drift.
5. The following game is a styled representation of the strate-

gic interaction between a principal (who delegates some au-
thority to an agent a nd can choose whether or not to a udit
that agent's effort in any period) and an agent (who chooses
whether to "work" or "shirk"). An a udit is costly to t he princi-
pal, but he doesn't have to pay the agent if he detects shirk-
ing. The particular payoffs arise from the following values
associated with t he relevant activities. The principal earns 4 if
his agent works, but doesn't earn a nything if the agent shirks.
He pays the agent 3 to work, but if he audits and catches the
agent shirking h e doesn't have to pay the agent anything. An
audit costs one unit to conduct; thus, if the agent works and
the principal audits, the principal nets out 0. It costs the

agent 2 to do his work, so he of course prefers to shirk and not
be audited (earning his wage without doing anything!) but
fears getting audited and earning 0.
Principal:
Agent: Audit Don't Audit
I o~·- 1 I 3~·-~ I
Work 0
Shirk
Does either individual have a strategy that is optimal no mat-

ter what the other individual plays (as in the Prisoner's
Dilemma)? Are any of the four cells equilibria? What percent-
age of the time must the principal inspect to make the agent
indifferent between working and shirking? What percentage
of the time must the agent work to make the principal indif-
ferent between auditing and not auditing?
6. What motivates bureaucrats? Discuss the various theories,

and explain why answering this question is significant for un-
derstanding and controlling the bureaucracy. Does the empiri-
cal evidence validate any one theory over and above the
others?
14
Leadership
In the discussion of legislatures and bureaucracy I have made

implicit reference to their leaders-committee and subcom-
mittee chairs, party and ideological leaders, senior civil ser-
vants and political appointees. But I have not, until now,
pulled out our rational-choice magnifying glass to take a
closer look at leadership. This I do ever so briefly in the pres-
ent chapter. In doing so, however, I will resist the temptation
to write separately about presidents, governors, mayors,
prime ministers, committee chairs, cabinet officers, party
leaders, and interest group leaders. Instead, I will take this
opportunity to v.-rite about leadership more generally.
The general subject of leadership is widely studied in soci-
ology, psychology, and organizational behavior, as well as in
political science but, as far as I know, rarely thought about
from the rational-choice perspective. I can, however, mobilize
some of the bits of theory recounted in earlier chapters to see
what light they shed on leadership phenomena. First I de-
scribe leaders as agents and then think about them as agenda
setters and entrepreneurs. Since these treatments draw on
materials already covered in some detail elsewhere in this
text, I will assume that the reader has a passing familiarity
with some of the underlyi ng concepts. 1
1 I also draw on the survey of Morris P. Fiorina and Kenneth A. Shepsle, "For-
mal Theories of Leadership: Agents, Agenda Setters, and Entrepreneurs," in
Bryan D. Jones, eel., Leadership arid Politics (Lawrence: University Press of
Kansas, 1989): 17-41. The interested reader should consult this essay for an
elaborated version of the argument found in this chapter.
445
In addition to slicing up leadership in various theoretically

interesting ways, T set a second task in this chapter. I want
to explore in some detail the problem facing leaders in self-
governing groups. In order to lead, the leader, naturally
enough, needs to induce followers to follow. Bul doing this
may be costly to the leader-after all, he needs the support of
followers to retain his leadership position. This sets in motion
an interesting strategic situation.
THE L EADER AS AGENT
In ordinary language, we normally conceive of the leader as

someone in charge, the boss, the person who gives orders and
instructions- the person on whose desk, as President Harry S
Truman reminded us, the buck stops. The leader is thought to
be proactive, whereas followers are reactive. 2 While there may
be some truth to this conceptualization, it is only a partial
truth, for it does not take into account the fact that leaders
must secure the support, if not the cooperation, of their follow-
ers. This support is necessary not only to enable the leader to
accomplish various group objectives but also to retain the
leadership post itself. This is certainly true of self-governing
groups-those that officially select their leaders. But even the
authoritarian leader has to worry about followers who, under
the right circumstances, might otherwise be tempted to move
against him or her.
The leader as agent of his or her followers is no better il-
lustrated than by the experience of Henry Clay. Clay was
elected from Lexington, Kentucky, to the 10th Congress in
1810. When Congress convened in 1811, Clay, a freshman
2 Some years ago there was a New Yorker cartoon showing a formation of
geese, minus itis leader at the apex, flying off, with the "lead" goose scram-
bling to get out in front.
Leadership 447
member, was elected Speaker of the House of Representatives,

a post he did not relinquish during the fifteen year s he served
in the House. Time a nd time again he was reelected to this po-
sition, and his very first acceptance speech on November 4,
1811, the statement of the quintessential agent, tells us why:
Gentlemen. In coming to the station which you have done me the
honor to assign me-an honor for which you will be pleased to ac-
cept my thanks- I obey rather your commands than my own in-
clinations. I am sensible to the imperfections which I bring along
with me, and a consciousness of these would deter me from at-
tempting a discharge of the duties of t he chair, did I not rely con-
fidently upon your generous support.
Should the rare and delicate occasion present itself when your
s peaker shouM be called upon to check or control the wanderings
or intemperance in debate, your justice will, I hope, ascribe to his
interposition the motives only of public good and a regard to the
dignity of the house. And in all instances, be assured, gentlemen,
that I shall, with infinite pleasure, afford every facility in my
power to the despatch of public business, in the most agreeable
manner. 3
In s tating that he would "obey rather your commands than
my own inclination," and "afford every facility in my power to
the despatch of public business, in the most agreeable man-
ner," Clay was adopting the classic role of leader as follower.
He was acknowledging the fact that as a major political
leader, he had ambition to remain leader at the very least and
wholly comprehended that to do so meant to lead his followers
by serving them. He understood that the compensation of
leadership- namely, continued incumbency in the Speaker's
position-was performance based. The reward structure of
leadership ties ambition (to remain leader in the next period)
to performance (in the current period).
a James Hopkins, ed., The Papers of Henry Clay, volume I (Lexington: Uni-
versity of Kentucky Press, 1959).
This is true even for those whose incumbency as leader is

assured. The bureau chief whose position is guaranteed by
civil service practices, for instance, nevertheless depends upon
her underlings, whether it is to secure policy objectives of im-
portance to her or merely to maintain the quiet life. As long as
the leader has objectives that depend upon the contribution,
cooperation, and morale of subordinates, the latter have some
hold on her.
We saw in the previous chapter that a principal generally
depends upon his or her agent. Indeed, he or she hires an
agent in the first place because agents are often more skilled
in accomplishing particular purposes than principals them-
selves are. Leaders are a specific sort of agent. They are relied
upon by followers to coordinate their activities, to provide re-
wards and punishments for group objectives, to secure allies
and defeat opponents, and generally to "grease the skids" for
the things followers want. Leaders are normally chosen for
the specialized skills they possess-to reason, persuade, bully,
inspire, rally, intimidate, mediate, and so on. The problem for
followers, a special instance of the problem facing principals
more generally, is that the specialized skills that make a par-
ticular agent attractive as leader are also the skills by which
that agent can exploit his or her followers for private gain.
The "who will guard the guardians" question becomes "how
will we control our agent?"
As in the general case developed in the previous chapter,
leaders may be controlled by their followers through before-
the-fact and after-the-fact mechanisms. Candidates for leader-
ship positions, like agents more generally, have reputations.
Even Henry Clay, who was but a freshman legislator when
elected Speaker, was already well known in Washington as an
avid "warhawk" (supporting war against Great Britain). 4 Rep-
• He had on ly recently completed a six-month term filling in as Senator from
Kentucky, during which time he made a number of well-publicized speeches
in favor of wa r and generally displayed his charismatic capabilities.
Leadership 449
utation is, in effect, a summary of an individual's specialized

skills as leader-it reflects not only her resume of experiences
and policy predispositions but also aspects of character, pro-
bity, and talent. Reputation is undoubtedly one of the central
elements in the decision calculus of followers when they set
about selecting a leader in the first place.
Reputation not only serves as a before-the-fact perfor-
mance predictor, but also it is an asset highly valued by its
owner. I n a sense, a leader posts her reputation as a bond
when she takes on a leadership responsibility. The reputation
is put on the line, so to speak, and will either be enhanced or
tarnished on the basis of her performance. By virtue of a
leader caring about her reputation, whether to enable her to
achieve some future leadership position, to maintain her cur-
rent post, or merely to acquire "a place in history," followers
have some control over her performance.
Thus, reputation is also an after-the-fact control mecha-
nism, since it involves follower authority over the leader's fu-
ture. Certainly whenever there are others waiting in the
wings, an incumbent leader knows he or she is not indispensa-
ble. At the time of "contract renewal," he or she must be able
to answer to the satisfaction of sufficient numbers of followers
the age-old question, "What have you done for me lately?"
Thinking of leaders as agents puts an entirely different
spin on the idea of leadership found in ordinary discourse.
Leaders are dependent on followers for accolades, for coopera-
tion, and ultimately for support. Leaders may be shrouded in
authority, may act like the boss, may order and instruct, but
all of these proactive leadership maneuvers are calculated to
impress followers, thereby earning their subsequent support,
not to bully them. What this suggests, therefore, is that we
need to look beyond a leader's capabilities, which can some-
times be awesome, to his or her intentions, of which he or she
is typically a prisoner. 'lo the extent the latter depend upon
follower support, the tables are turned-the master is servant
and the servant master. We will need to keep these things in

mind as we examine next some of the capabilities of leaders.
CASE 14.1
FDR AND W ORLD WAR II
A leader is often placed in the difficult position of choosing
between what lUs or her followers want and what he or she
feels they need. This distinction was made especially clear
by the eighteenth-century English philosopher Edmund
Burke in his famous speech to the electors of Bristol. A
member of Parliament at the time running for reelection,
he made clear to his constituents that he would not go back
to Westminster as a mere delegate- someone who simply
represents and acts as an unreflective agent of constituency
opinion. Rather, he regarded himself as a trustee-someone
whose intelligence and expertise would be deployed in the
service of what he considered to be the best interests of his
constitutents. (Burke was defeated for reelection.) Henry
Clay, on the other hand, cast his role as Speaker as more
that of a delegate: "I obey rather your commands than
my own inclination." (Clay was reelected to that post for
fifteen years.) Should a leader continue to obey, or should
he follow his own inclination? The answer depends to a
large extent on how a leader interprets his mandate as an
agent.
In the years leading up to World War II, Franklin Roo-
sevelt was faced with a difficult dilemma. He followed the
Nazi march through Europe with great alarm. Public opin-
ion in the United States at the time was decidedly isolation-
ist, but Roosevelt saw a need to begin providing aid to
Great Britain a nd to prepare the country for war. He be-
lieved that delaying mobilization until Germany posed an
immediate threat to the United States would be too late.
Leadership 451
And yet, most Americans felt this was a "European prob-

lem," one in which the United States should not engage.
A delegate's course of action would have been to follow
public opinion and wait. But Roosevelt saw himself, like
Burke, as a trustee and began taking secret action early on
to prepare the nation for war. He used his skills as an
agenda setter and political entrepreneur to mobilize re-
sources and generate popular support. His ability to take
the lead- to get out ahead of public opinion-derived from
the reputation al capital he had accumulated during his
years of grappling with the economic disaster caused by the
Great Depression. Roosevelt's actions prior to World War II
exemplify the interrelatedness of agent, agenda setter, and
entrepreneur facets of leadership, as well as the importance
of reputation in facilitating leadership maneuvering room.
I examine these topics more explictly in the next several
sections below.
T HE L EADER AS A GENDA S ETTER
The alternative views of leaders that are surveyed in this

chapter are by no means mutually exclusive. Indeed, implic-
itly at least, aU of these views are compatible with the agent
perspective I just fini shed reviewing. Once a leader is con-
ceived as an agent for a coalition of followers, it is possible to
turn to the more precise sorts of things he or she does in that
capacity. In this section I do just that, envisioning the leader
as the person in charge of the group's agenda.
A group, whether civil servants in a bureau, legislators on
a committee, or members of a faculty, have a variety of things
they may wish to achieve together with only finite resources.
They can't do everything; choices must be made about where
to devote their attention and other scarce resources. Of course,
some of the activities of a group may be mandated by others.

The Social Security Administration must churn out those
checks every month. The Budget Committee must produce an
annual budget resolution. The faculty of arts and sciences
1nust determine and approve a list of degree candidates each
year prior to graduation. 5 However, it is unusual for a group to
have no discretion regarding its activities, so choices about
how to exercise that discretion must be made. And this is pre-
cisely one of the things a leader does.
Consider a department in a university, say political sci-
ence, and its chair. Many of the activities of that department
are determined by others- university rules, the dean, the
board of trustees. It must organize its curriculum for the com-
ing academic year. It must have requirements for various de-
gree programs in place. It must decide which graduate
students to admit to its doctoral program. It must see to the
administration of examinations, the grading of theses, and the
determination of prizes and honors. But there is still room on
the departmental plate for discretionary activities. Should the
department focus its attention this year on revising the grad-
uate program, or defer it to next year? Should the department
fill the vacant assistant professor post for which it has author-
ization and, if so, should it search for someone whose specialty
is rational-choice theory, congressional politics, or compara-
tive judicial policy? Should the department promote an es-
pecially promising younger colleague to tenure, or put the
promotion off for a year? Should the department begin negoti-
ating with the Department of African American Studies over a
joint undergraduate degree? The list is enormous, constrained
only by the imagination of its faculty on ways to spend time
and other resources.
6
By "must" I do not mean these things are inviolable laws of nature. like grav·
ity or the first law of thermodynamics. Rather, some things are so fundamen·
tat to a group that its very existence depends upon accomplishing them.
Failure to do so risks having the group replaced or its jurisdiction altered.
Leadership 453
Discretionary choices for a group about where to devote its

energies are highly influenced by its leader. The department
chair undoubtedly will be heavily lobbied on all the matters
listed in the previous paragraph, and some of the decisions
may not be hers alone to make. But her input will be impor-
tant, mainly because in any group the buck must stop some-
where, authority to deploy group resources must reside
somewhere, and deciding all such issues town-meeting style is
exhausting and inefficient. Most members of a departmental
faculty do not v.·ish to spend most of their waking hours on de-
partmental business- they have their own personal agendas
to prosecute. Thus, even though not all of them will be pleased
with every decision their agent makes on their behalf, the del-
egation of authority to the chair to set the department's dis-
cretionary agenda is often the efficient division of labor to
institute. The members of the department content themselves
with having had some say on who that agent would be in the
first place (a before-the-fact control mechanism), how long she
will survive in the job (an after-the-fact control mechanism),
and whether she will enjoy an uncomfortable or exalted post-
chair existence in the department (another after-the-fact con-
trol mechanism).
The chair, on the other hand, will weigh the effects of these
' various control mechanisms against her opportunity to steer
the department's agenda in the direction of her own prefer-
ences. Suppose, for instance, she cares quite a bit about the
graduate program but is not keen to promote her younger col-
league at the present time. By announcing the appointment of
a graduate review committee but not taking action on a
tenure review, she accomplishes her ends but may have to en-
dure the wrath of those with different preferences. Indeed, she
may get "rolled"-one of her seniot colleagues may, during a
facu lty meeting, argue that it is imperative that the depart-
ment move on the tenure case, a view supported by many oth-
ers in the department (even though the chair had tried to keep
that off the agenda). Since a leader cannot afford to get rolled
too frequently-if this occurs, it becomes increasingly difficult
for an agent to convince her principals that she has their in-
terests at heart- she will want to give careful consideration to
the distribution of preferences among those she leads. Like-
wise, in deciding to exercise an after-the-fact control mecha-
nism like rolling the chair, a follower must determine exactly
how much damage this will inflict on a leader he otherwise
thinks is satisfactory. 6 In short, care must be exercised to de-
tect but not to cross any number of rather fuzzy lines. In the
end the chair can extract some agenda advantages (an un-
avoidable form of agency slippage), but she must be prudent,
not piggish. This problem is normally least serious at the be-
ginning of a chair's term; at the end of a chair's term in office,
especially one who would like to retire to her own work,
agency slippage becomes more problematic (the so-called end-
game problem).
I have developed the interaction over agenda setting be-
tween a chair and her colleagues in a university department,
not only because it is interesting in its own right (at least to
the author, whose life is substantially affected by this agency
relationship), but also because it is an exemplar of the oppor-
tunities possessed by leaders of groups in general to trade off
satisfying the preferences of their followers in order to secure
some of their own objectives. Refining the group's agenda is
one of the standard activities that a group delegates to a
leader (or an agenda committee). In doing so, there is ordinar-
ily an awareness in the group that a "pound of flesh" will be
exacted by the leader-agent. Politically savvy choices, by the
group in selecting the leader and by the leader in exercising
her agenda authority, place limits on this "compensation."
Should those limits be breached, revolt soon follows.
6 He will also have to weigh the consequences of trying, but failing, to roll the
chair. The age-old adage applies: "If you go after the king, you had better get
him."
Leadership 455
Before leaving the subject of leader as agenda setter, we

should revisit .s ome of the material of Part II of this volume,
emphasizing the manipulation of agenda procedures at a more
refined level. Not only does a leader have considerable sway
over an institution's agenda "in the large," he or she also regu-
lates the fine -grained deliberation over specific items on the
agenda. This aspect of agenda setting is multifaceted.
The Speaker of the U.S. House of Representatives has a
host of opportunities to influence, directly or indirectly, how
member proposals will be treated by the full House. At the
outset, when a bill is initially "dropped in the hopper" as a leg·
islative proposal, the Speaker and his right-hand official, the
Parliamentarian, determine which committee has jurisdiction
over the proposal. As in our earlier discussion, the leader in
this case may not have complete discretion. He cannot send a
farm bill to the Merchant Marine Committee and he cannot
transmit a bill reforming the criminal code to the Armed Ser-
vices Committee. But many proposals do not match cleanly
and naturally with the established jurisdictions of House
committees, so the Speaker has some discretion. Indeed, since
t he mid-1970s, the Speaker has been given additional bill-
assignment powers, known as multiple referral, permitting her
to a ssign different parts of a bill to different committees, or to
assign the same parts sequentially or simultaneously to sev-
eral committees. We have already seen in Chapter 12 that
committees, on ce as.signed a bill, are in a position to exercise
before-the-fact gatekeeping and proposal powers and after-
the-fact negotiating and oversight powers. The Speaker is in a
position to determine which panel(s) exercise(s) these powers. 7
7
Th e Speaker is often instrumental, moreover, on exactly who serves on
which committee. Until 1910 in th e House of Representatives, th e Speaker
made all committee assignments. Even after that date, up to the present,
the Speaker's influence is felt in this absolutely crucial institutional process.
In other legisla tures, the Speaker or equivalen t ins titutional officer remains
the pivotal person in terms of committee assignments. And, of course, in
other venues, the leader is often decisive in this respect. In most academic
Once the committee or committees assigned a bill have

acted affirmatively, the whole bill or the various parts of it
acted on by multiple panels are transmitted to another arm of
the leader, the Rules Committee, which determines the spe-
cific rules under which the legislation will be considered by
the full House. Here the leader influences when debate will be
scheduled, for how long, what amendments will be in order,
and in what order they will be considered. She also rules on
all procedural points of order and points of information raised
during the debate. We saw in Part II that, when a bill is multi-
dimensional in nature, a modicum of preference diversity is
all that is required to produce majority voting cycles. This
means that the order of taking up amendments, and proce-
dural rulings on what must seem to the outsider to be minor
matters, will be decisive in determining what actually results
at the end of the process.
Although I have used the leader of a legislature to il-
lustrate points about fine-grained agenda control, I should
emphasize that a crucial resource of all leaders is their proce-
dural discretion. If every procedural issue were hardwired,
then the leader would be impotent. But rarely is this the case.
(To be so, the writers of rules would have had to anticipate
just about every possible contingency that might arise in the
history of that institution and to have bothered to write down
in the rules exactly what must be done in each circumstance.)
Thus, if a department interviews four candidates for an assis-
tant professorship, the department chair often has the oppor-
tunity to suggest the voting method that will be used to choose
among them. (Anyone opposing her suggestion will appear to
have an ax to grind.) In claiming that she needs a final rank-
departments, for example, t he chair composes committees. If, from our ear-
lier example, the department chair did not want to promote a junior col-
league to tenure, yet felt constrained from keeping the matter off the agenda
entirely, she could influence the result by having the candidate reviewed by
an artfully crafted committee t hat is likely to reflect her own doubts.
Leadership 457
ordering (so tha t if the first choice were to decline the offer,
she would know to whom to turn next), she could suggest that
each voting faculty member write down a single name-and
candidates would be ranked in terms of the total votes each
got~r that each faculty voter submit a complete rank-
ordering, from which a group rank-ordering would be derived
(see Chapter 4). Indeed, as Chapter 7 makes clear, many dif-
ferent voting systems might be employed in this circumstance,
any one of which the chair could defend by appealing to at-
tractive criteria the voting system satis fied. If she has kept
her ear to the ground, she will be able to make this "sugges·
tion" in an informed strategic fashion. Such are the agenda-
setting opportunities with which a leader is presented.
THE LEADER AS ENTREPRENEUR
The final conception of leader I consider here is that of the en-

trepreneur. Although compatible with agent and agenda-
setter conceptions, it emphasizes something slightly different.a
Entrepreneurs are leaders who, as often by self-appointment
as by selection, perform necessary services to enable a group
to accomplish some collective purpose. An entrepreneur, in
fact, may not even be a member of the group in question. In-
deed, the group in question may not even be a group in any
manifest sense. Rather it may come into being, so to speak,
because of the enterpreneur's activities. In this situation, an
entrepreneur may be seen as an agent who chooses (or creates)
a principal, rather than the more typical arrangement in
which a principal hires an agent.
8
Th e classics in the leadership literature emphasizing entrepreneurs ar e
Rich ard Wagner, "P ressure Groups and Political Entrepreneurs," Papers on
Non-Market Decision Making 1 (1966): 161-70; and Norman Frohlich, J oe
Oppenheimer, and Oran Young, Political Leadership and Collective Goods
(Princeton, N.J. : Pr inceton University Press, 1971).
I made mention in Chapter 9 of political entrepreneurs,

suggesting them as one "solution" to the problem of collective
action. I have little to add here, except to note that this con-
ception of leadership emphasizes, perhaps more than the oth-
ers we've discussed, the artistry of seizing opportunities. As
entrepreneurs, leaders shop around for unrealized possibili-
ties. These may be things the leader cares about. A young Uni-
versity of California graduate student, Mario Savio, became a
folk hero in the 1960s when he coalesced Berkeley students
into what became the Free Speech Movement. A young attor-
ney named Ralph Nader took on General Motors in the 1960s
in the first of many events that culminated in the consumer
movement. In the 1950s a black minister named Martin
Luther King Jr. led a boycott against the segregated public
transportation system in Montgomery, Alabama, in an impor-
tant early episode of what became the modern civil rights
movement. At about the same time, the separate ideological
and organizational acts of the entrepreneurs Malcolm X,
Stokely Carmichael, and Huey Newton gave rise to the Black
Power movement. In each case a political entrepreneur, com-
mitted ideologically to the cause for which he fought, seized
leadership by his very acts of leadership. Sometimes this lead-
ership involved mobilizing people who conceived of themselves
as members of a group but had failed to achieve group objec-
tives; other times it involved people who, by entrepreneurial
acts of leadership, were forged into a group. The leader reaped
not only ideological satisfaction but personal glory as well.
Implicitly, though, I have hinted that the entrepreneur
need not care very much about the issue in which he or she
plays a leadership role. Rarely will an entrepreneur be indif-
ferent. But not so rarely, he or she may "entrepreneur" an
issue for reasons other than achieving a satisfactory resolu-
tion of that issue. Sometimes the entrepreneur is a control
freak who consumes utiles from the very act of leading. Other
times, the entrepreneur derives payoffs from group members
Leadership 459
out of gratitude. Political leaders who act entrepreneurially

occasionally fall into this latter category, pushing issues that
they believe may have an electoral or campaign-finance payoff
from grateful beneficiaries. 9 Finally, it should be noted that
some entrepreneurs, especially young, politically motivated
lawyers in Washington, D.C., or one of the state capitals, play
the entrepreneurial game because they see it as a career in-
vestment, enabling them to build "can-do" reputations, net-
works of contacts, and opportunities for shoulder rubbing with
potential ment0ors.
Political entrepreneurship is not for the weak-kneed or
risk-averse. For every Mario Savio, Ralph Nader, and Martin
Luther King, there are hundreds-perhaps thousands-of
failed entrepreneurs. The market for political entrepreneur-
ship, like the market for economic entrepreneurship, rewards
creativity and insight, but many who think they have one or
both are often sadly mistaken. Sometimes it is just dumb luck
that elevates one entrepreneur to fame (or notoriety).
Agenda setting, agency, and entrepreneurship are but dif-
ferent facets of the same leadership phenomenon. What espe-
cially distinguishes them from one another is the degree to
which they reflect greater or lesser structure in the context in
which leadership is exercised. Although agenda setting has its
capacity for displays of flair, it is a relatively highly structured
activity within a tightly bounded institutional setting. Agency
activities, though still exhibiting the characteristics of a struc-
tured relationship, are a bit broader. Entrepreneurship is
broader still, emphasizing not only those actions that occur
within a fixed structure but also the actions that create or
9
The author, while interviewing congressional staff in the early 1980s, ruscov-
ered that it was not uncommon for a legislative staff director to assign
staffers to monitor committee hearings around Capitol Hill with the instruc·
tion, "Find something useful for the boss." Members of Congress, that is,
often shop around for issues that they believe they can convert to some use-
ful purpose for the mselves.
transform structure. It is, in a sense, the richest conception of

leadership but it is also the most vague. I return to the more
rigorous form of analysis in the last part of this chapter, focus-
ing on the strategic trade-offs facing a leader concerned with
maintaining and enhancing her reputation.
R EPUTATION AND L EADERSHIP
CASE 14.2
ONLY N IXON COULD Go TO CHINA
It is sometimes said in political circles that only Nixon

could have gone to China. In the early 1970s the United
States recognized the People's Republic of China (then re-
ferred to as Red China or Communist China), and Presi-
dent Richard Nixon visited the country, the first chief
executive ever to have done so. Both actions were highly
controversial at the time, given the severity of the Cold
War, the strength of anticommunist sentiment, suspicions
about Chinese motives during the end phase of the Vietnam
War, and loyalty to our ally in Taiwan.
Why was it, then, that "only Nixon" could go to China?
According to some, only a conservative with sterling anti-
communist credentials could have gotten away with it. A
liberal Democrat would have been seen as soft on commu-
nism and would not have been able to convince the public
that his or her motives were aligned with the best interests
of the nation rather than with some private ideological
agenda. Is there any merit to this argument?
The discussion of reputation and leadership presented
in this chapter can provide some help. Prior to his recogni-
tion of and trip to Communist China, Nixon had estab-
Leadership 461
lished a reputation as a "red-baiter" of the highest order.

The prospective cost to him in terms of his standing with
the right wing of the Republican party was very high, in-
deed. He coul d therefore credibly claim that he was not
prosecuting a private agenda. By bearing this cost he could
persuasively claim that he was performing as an "agent of
the nation," thereby persuading his principals (who not
only controlled his political future but also would determine
his "standing in history") that he was not shirking his offi.
cial responsibilities. A mor e liberal president, acting in the
same way, would have sent a much more ambiguous signal.
His actions would therefore have been suspect, since it
would have been difficult for the public to determine
whether he was acting in behalf of a private agenda or with
a broader view of the nation's welfare.
Similar interpretations may be offered for Lyndon J ohn-
son's civil rights policies during the 1960s and Ronald Rea-
gan's fiscal policies during the 1980s. Johnson, for many
years a Texas politician with a local reputation for using
race in unsavory ways,, became one of the political pillars of
the civil rights movement. His leadership in the U.S. Sen-
ate in the mid-1950s produced the first civil rights bill in
nearly a century. His leadership as president was essential
in passing the Civil Rights Act of 1964 and the Voting
Rights Act of 1965. Nearly twenty years later, Ronald Rea-
gan's presidency amassed the largest set of budget deficits
in the history of the Republic (surpassed recently by the
conservative George W. Bush). It is widely believed that
only a conservative president with a strong reputation for
fiscal prudence and a belief in smaller government could
have gotten away with such policies. In each of these cases,
as with Nixon, a strong reputation enables considerable
counterintuitive flexibility. As they say, "only Nixon could
have gone to China."
Every politician leader faces the problem of how most effectively

to sanction uncooperative behavior by followers. Sanctions may
be costly, but for the most successful leaders the mere threat of
such sanctions is usually sufficient. Since the strength of leader-
sh ip ca n be an important determinant of how, and how well, a
political organization fun ction s, students of political organization
need to understand the theoretical conditions for effective
leadership.
With these remarks, Randall Calvert introduces a provoca-

tive model about the trade-offs that confront the leader of a
political organization. 10 Although he focuses on legislative
leadership, the questions he worri es about are generic, apply-
ing to bureaucratic leaders, university leaders, interest group
leaders- in short, leaders of any political organization.
'l'he context is one in which a group of kindred spirits, seek-
ing to accomplish group goals, creates a leadership structure
and appoints a particular incumbent to help them solve vari-
ous collective action problems. As kindred spirits, members of
the group share much in common, but they are not perfect
copies of one another. Members of a university department,
for instance, may all want to appoint an assistant professor in,
say, American politics, but some favor a congressional special-
ist, others an expert on elections and voting behavior, still oth-
ers an authority on the presidency. Similarly, the members of
the majority party in a legis lature may share a preference for
health care legislation, but various factions among them push
for somewhat different versions (an unpleasantness that faced
President Barack Obama as he struggled with health care leg-
islation in 2009-10).
Whatever is ultimately accomplished will, on net, make
the group members better off; however, some will be more
pleased with the result than others. And those less pleased
10 "Reputation and Legislative Leadership," Public Choice 55 (1987): 81- 120.

The passage quoted in the text appears on page 81.
Leadership 463
may be tempted to defect from the group's effort, deserting it

at a crucial vote, for instance. In order to keep this prospect
from being totally debilitating, the leader must threaten, and
sometimes actually employ, sanctions against renegades. For
coordinating the group's activities, forging compromises
among the heterogeneous preferences of group members, and
holding enough of its members together to succeed in approv-
ing and implementing that compromise, the leader r eceives
"compensation." 11
Calvert assumes that in prosecuting the group's agenda of
objectives, cajoling, exhorting, and sometimes punishing group
members in th e process, the leader must maintain a minimal
level of group support. There are always leaders-in-waiting,
ready to step in to take control if the incumbent leader should
stumble. This means that sanctions by the leader against
group member s are costly, because they are administered
against members whose support is desired by the leader. The
leader must assiduously balance the benefits, in terms of
group success, against the costs, in terms of threats to her
leader ship, of employing sanctions. As Calvert puts it, "any
given punishment increases by a s mall amount the probability
that the leader w111 face a full-scale r ebellion, deposing him or
impairing his ability to lead in the future. In this sense, the
leader faces a kind of budget constraint on his ability to im-
pose sanctions."
Calvert also assumes that the precise nature of the bene-
fits to the leader (in terms both of achieving group goals and
obtaining private "compensation"), and of her costs of employ-
ing sanctions and bearing the associated ris ks, is private in-
formation. The leader knows these benefits and costs, but
followers have only the vaguest sense. By cultivating this un-
11
'l'his compensation comes in various forms, not the least of which is the op·
portunity for the leader to sq ueeze pr ivate advantage out of her agenda set·
ting powers, something discussed earlier in this chapter.
certainty, the leader is able to make followers believe that

punishment is possible more often than under conditions of
complete information. And even in those cases where punish-
ment is "too costly" (as objectively seen by a fully informed ob-
server), the leader may nevertheless engage in sanctioning
behavior in order to mislead followers into believing that her
costs are lower than they actually are. In short, uncertainty
about leadership benefits and costs allows the leader to bluff,
deceive, and dissemble- the stuff of which reputations are
made (and maintained).
In any given circumstance of leader-follower relations, the
context is bound to be rich in history and details. making it
hard to sort out the underlying factors at work. Calvert pro-
poses a very simple formal model that permits attention to be
focused on these underlying factors. In a bstracting away from
the substantive details of a specific situation, of course, much
is lost. But doing so permits the investigator to capture the ef-
fects of factors that are present, in one manifestation or an-
other, in all such circumstances. This is the advantage of a
formal model.
Display 14. l gives the formal structure of Calvert's setup.
If the follower (Mr. t) supports the leader's proposal-say by
voting for the committee chair's bill or supporting the depart-
ment chair's job candidate--the leader earns "cr edit" toward
reelection or reappointment, a (where a > 1), while follower t
receives whatever payoff he associates with achieving the
group's common objective (normalized to 0 without affecting
the argument). If, on the other hand, the follower rebels, then
the payoffs depend on what the leader does in response. If she
acquiesces in the rebellion, allowing follower t to oppose her
proposal, then she gets less than what she would have if the
follower had obeyed (again normalized to 0) and the follower
gets more (some positive level b < 1 rather than 0 if he did not
rebel). Finally, if the follower's rebellion is punished, then the
leaders bears a cost, -x1 and the follower gets b-1, a number
Leadership 465
less than 0 (since b is less th a n 1). In this latter circumstance

follower t does not know the value of -x1; he does not know
whether or not punishment in this case is costly to the leader
o:t how much (since he does not know how solid her hold on of-
fice is). Rather, follower t has beliefs, namely that -x,
is costly
(equal to -1) with probability w, and costless (equal to 0) with
probability 1-w. Although follower t has only probabilistic be-
liefs about the costliness of punishment to the leader, the
leader knows the value of -x, in advance-she is always fully
informed of the consequences of her actions.
D ISPLAY 14.1
Behavior Payoff of Lead er P ayoff of Follo we r t

Follower t obeys a 0
Follower t rebels;
Leader acquiesces 0 b
Follower t rebels;
Leader punishes -x, b-1
One more thing: followers are capable of learning. Their

belief, w, that punishment is costly ch a nges as they observe
the leader in action. Thus, leaders not only have asymmetric
information about the costliness of administering punish-
ments, they also are able to influence subsequent beliefs of fol-
lowers about th is information. If, for instance, she is able to
convince followers that punishments are mainly costless to
her, then she is likely to induce more follower loyalty to her
proposals (and thus not even have to use punishments). This
might be accomplished by administering punishments early in
her incumbency, even in those circumstances in which such
acts are really costly to her. A department chair, for example,
might deny a senior American scholar in her department an
assignment he covets to the American politics recruitment
committee, because he has announced a preference to appoint
someone with a specialty different from the one the chair

prefers. In fact, she will take a lot of heat for leaving him off
the recruitment committee-something she knows in advance,
but others in the department do not (at least not for certain).
She does it anyhow to affect the beliefs of her colleagues about
her freedom of action, conveying to them probabilistically at
least that she doesn't regard this "heat" as of much conse-
quence. Next time around, on some other issue in the depart-
ment, fewer of her colleagues will be disposed to oppose
because they believe it is more likely that she can "punish"
them costlessly.
In short, she is on her way toward establishing a reputa-
tion for not tolerating rebellious behavior. It does not mean
she will never encounter opposition, only that her colleagues
will think hard before moving against her proposals. Of
course, she too will think hard about the nature of the propos-
als she makes in the first place. Life will exist mainly and
happily in the first row of Display 14.1 in which her colleagues
enjoy the achievem ent of common objectives (if not always ex-
actly to their liking), the chair is secure in her leadership, and
punishments do not have to be meted out (looming only as a
club behind the door).
Calvert models leader reputation-building by studying the
"game" in Display 14.1 repeated exactly twice. 12 In effect, the
leader faces the possible rebellion first of follower t 1 and then
of follower t 2 • What she does in the case in which follower t 1
rebels will affect what t 2 subsequently does, inasmuch as he
will update his beliefs about how costly it is for the leader to
punish rebellion. The actual cost to the leader of punishing re-
bellion on each occasion in which it occurs (equal to -1 with
probability w and 0 with probability 1-w) is known to the
leader in advance of her choices.
12 This keeps the math manageable while at the same time allowing the ana-
lyst to see what happens to the leader over time.
Leadership 467
Calvert shows that there is actually a ''best way'' for each

of the three players to play this game. Some of the details are
obvious (but reassuring to see that his "toy model" has actually
captured them). First, if -x1 actually is 0 (costless punishment),
then the leader should always punish rebellion. Second, if
-x2 = - 1 (i.e., it is costly for the leader to punish the second fol-
lower), then she should never punish t 2 ; there would be no
purpose served in doing so (since the game ends) and thus no
point in bearing the cost of punishing.
Less obvious, and more interesting, is what the leader
should do if -x1 - -1. In providing insight on this issue, Cal-
vert's model is far more than a toy for capturing the obvious.
He shows that the leader's response to rebellion by t 1 when it
is costly to punish depends in very specific ways on the beliefs
followers hold about the leader's costs of punishment relative
to the benefits they would secure from rebelling. It takes us
too far afield to go into the details of Calvert's conclusion, ex-
cept to say that Calvert identifies very precisely when t 1
should rebel, when a leader should punish that rebellion for
certain, when a leader should "flip a coin" (or randomize in
some other way) to determine whether to punish this particu-
lar transgression, and whether, conditional on all this, t 2
should rebel (for certain or with some probability).
The important conclusion of this analysis is that leaders,
created to enable followers to achieve commonly held objec-
tives despite their temptation to defect occasionally, need to
establish a reputation for being willing to sanction uncoopera-
tive behavior. If such sanctioning were costless, leaders would
be incredibly powerful, perhaps too powerful from the vantage
point of followers. If, on the other hand, punishing followers
for transgressions were too costly, then leaders would be so
weak that they could not discourage such transgressions at
all; group efforts would be plagued by defections, reneging on
promises, and other forms of uncooperative behavior which
the leadership institution was invented to control in the first
place. But if leaders can maintain some uncertainty about the

costliness to them of punishing transgressors, then (depend-
ing on many specific features of the situation detailed by
Calvert's theoretical results) they may be in a position to build
a reputation thal causes potential rebels, at least some of the
time, to have second thoughts about rebelling. We cannot im-
prove on Calvert's concluding paragraph:
Despite the apparent bare-bones nature of the model presented
here, it serves to focus attention on a neglected aspect of political
leadership. Since a successful leader cannot be constantly moni-
toring and punishing followers, it is necessary to establish in the
followers a habit of obedience, a rule of thumb that the leader's
wishes are to be followed. Habits and rules of thumb do not occur
in a world of perfect and costless information. Thus the presence
of uncertainty, and the artful manipulation of it, are crucial for
successful leadership. Somewhat counterintuitively, this holds
true regardless of whether the leader is a ruthless dictator or a
benevolent provider of collective action.
CONCLUSIO N
Rather than writing separately about all the different kinds

of political executives we find in the real world-presidents,
premiers, CEOs, chairs, governors, mayors-I have instead
described some interpretations of the institution of political
leadership drawn from the analysis presented in earlier parts
of this book. One message, I hope, comes through clearly: An
understanding of leadership depends upon an understanding
of followership. Leaders are often invented by followers. Even
when they are not, they depend upon followers, both for the
cooperation needed to secure group objectives and for the sup-
port enabling the leader to maintain her position. This is ex-
plicitly recognized in the agent perspective on leadership
which we took to be the prevalent way of thinking about lead-
Leadership 469
ers. Accordingly, leaders must be empowered to accomplish for

their followers what the followers are unable to accomplish
without leaders, but not so empowered that they are able to
pursue a private agenda without regard for follower prefer-
ences. Leaders must be proactive but, ultimately, under the
control of their followers.
As to what leaders (and followers) actually do, I focused on
agenda setting, general entrepreneurial activity, and reputa-
tion building as elements in a leader's behavioral repertoire,
Throughout I have emphasized that leadership is a solution to
a series of problems that groups face in trying to pursue com-
mon objectives. If these problems did not exist- that is, if we
lived in a perfect world- then collective action could be pursued
with ease. But the world is not perfect, so institutions like lead-
ership must be devised as second-best solutions. And these so-
lutions, in turn, possess problems of their own-hence the need
for control mechanisms. Organizational life is a balancing act--
between leaders and followers, between controlling agents and
securing goals of principals, between exploiting the division
and specialization of labor and giving advantages to special-
ized individuals. Our attention has been focused on those
features of leadership in organizational life. For the nuance
and subtlety that accompanies the interactions between real
leaders and real followers in particular contexts, there is a
rich vein of political biography for the interested reader to
mine.
1. Benjamin Disraeli, the nineteenth-century British prime

minister, held: "I must follow the people. Am I not their
leader?" Discuss the two images of leadership implicitly com-
pared in this quotation: the leader as trustee and the leader

as delegate. Are these visions of leadership necessarily in
conflict?
2. This chapter has argued that the principal-agent frame-

work developed in Chapter 13 can be productively applied to
understanding th e delegation of authority to leaders. Discuss
the principal-agent problem in general, describe the ways in
which it applies to the problem of choosing leaders, and then
discuss a variety of before-the-fact and after-the-fact means of
controlling leaders that ameliorate the principal-agent prob-
lem. In answering this final part, you may find it useful to
contrast publicly elected officials (the president, members of
Congress) and institutional leaders (the Speaker of the House,
academic department chairs).
3. The 1\venty-second Amendment to the U.S. Constitution,

passed in 1951, limits any individual to only two terms as the
president of the United States. In contrast, members of Con-
gress do not face any limits on the number of terms in office
they can serve, although fifteen state legislatures currently
employ some form of term limits. Discuss the pros and cons of
term limits, with special reference to agency slippage and the
"endgame" problem. As applied to legislative leaders, in what
ways do terms limits serve the intended purposes of leader-
ship, and in what ways do they undermine those purposes?
Explain your reasoning carefully.
4. Part II of this book argued that aggregating individual pref-

erences into social choices can lead to incoherent, chaotic, or
manipulable decisions; Part III argued that individual inter-
ests operating together can lead to suboptimal, inefficient, and
even perverse outcomes. ln what ways does the delegation of
agenda-setting authority to leaders help to reduce these two
problems, of social choice and group cooper ation? Answer with
Leadership 471
specific examples. What new problems are created when au-

th<:>rity and discretion are vested in leaders?
*5. The following simplified treatment of electoral accounta-

bility is based on Ferejohn (1986).a Suppose that the median
voter (V) has an ideal point at zero on some single-dimen-
sional issue space, while any elected leader's (L) ideal point is
normalized to one (for example, you might think that voters
desire no corruption but politicians a lot-
. if they can get away
with it). Assume that L's utility for any outcome 0 s p s 1 is
equal to p but he also secures a payoff of T for each term in of-
fice, which is measure on the same scale, so his total payoff in
the first term is p + T. Also, payoffs in the second term are
discounted by a factor of /.. < 1 and L can be elected for only
two terms, so if he is elected for a second term his payoff is
A(p + T).
If Lis reelected for a second term, what policy will he imple-

ment? Now assume voters use a "retrospective voting strat-
egy" of the form: reelect if p s r and vote out of office
otherwise. Come up with two expressions for L's utilities, one
if he is reelectecll. and one if he is not, assuming for each case
that L sets p as high as possible consistent with the desired
electoral outcome. Then show that for voters, the optimal vot-
ing rule has r = 1 - /.. - /..Tor 0 depending on the values of/..
and T. How does voter utility in equilibrium change with /..
and T?
*6. Why is uncertainty about the leader's costs of punishment

a fundamental ingredient in Calvert's model of leadership rep-
utation building? To answer this question, first show that if x1
and x 2 are known in a two-period game, then t 1 and t 2 either
• See John Ferejobn, "Incumbent Performance and Electoral Control," Publi.c

Choice 50 (1986): 5-25.
always rebel and face no punishment, or never rebel. Then,

explain the intuition behind why uncertainty and incentives
to build a reputation can lead to punishment-even when it it
is cheaper than nonpunishment-in the first period.
7. Chapter 12 argued that legislatures face several fundamen-

tal problems in generating legislation, including majority cy-
cling, matching specific legislator interests to influence on
those issues in the legislature at large, and imperfect informa-
tion on both the policy and political impacts of legislation.
These problems could be concisely summarized as a vote ag-
gregation problem, a cooperation problem, and an information
problem. Discuss the ways in which the House and Senate
leadership, most obviously personified by the Speaker of the
House and the Senate majority leader, help to solve each of
these problems.
15
Courts and Judges
In the political science literature on rational choice approaches

to political institutions, the "c" word and the "j" word are
rarely uttered. Part of the explanation for this relative inat-
tention to courts and judges is they are enigmatic. Elected
politicians, though undoubtedly no less complex than judges,
lend themselves to simple behavioral hypotheses- they want
to be reelected; they pursue personal conceptions of "good pub-
lic policy"; they aspire to positions of influence within their re-
spective institutions. These simple behavioral hypotheses
provide leverage on understanding the operating properties of
legislative, executive, and electoral institutions. Judges, on the
other hand, resent being included in the same category as
elected politicians, preferring instead to be thought of as aloof
from the daily fray of politics. They also remain aloof from the
"trucking and bartering" of the private economy- they do not
"market" their skills and services. Indeed, their compensation
is ordinarily fixed in a manner quite independent of their per-
formance. Cong.equently, we don't quite know what to make of
judges or the legal institutions in which they operate.
Richard Posner, a federal appeals court judge as well as an
eminent former professor and currently senior lecturer in law
at the University of Chicago, writes with considerable author-
ity on this puzzle:
At the heart of economic analysis [or what in the present volume

I have called "'rational choice analysis"] of law is a mystery that is
473
also an embarrassment: how to explain judicial behavior in [ra-

tional] terms, when almost the whole thrust of the rules govern-
ing compensation and other terms and conditions of judicial
employment is to divorce judicial action from incentives-to take
away the carrots and sticks, the different benefits and costs asso-
ciated with different behaviors, that determine human action in
an economic model. ... The economic analyst has a model of how
criminals and contract parties, injurers and accident victims, par-
ents and spouses-even legislators, and executive officials such
as prosecutors- act, but falters when asked to produce a model of
how judges act. 1
As central players in important political institutions,

judges are politicians; yet their links to "constituencies" are
attenuated by the fact that, through lifetime tenure, they
don't need to have their contracts renewed by those con-
stituencies at regular intervals.2 Not only is the reelection in-
centive absent; as Posner just noted, other "carrots and sticks"
commonly found in employment relations (like performance-
based levels of compensation) are missing. What, then, moti-
vates judges? As Posner asks, "Are judges rational? Or have
the elaborate efforts made to strip them of incentives placed
their behavior beyond the reach of rational choice models?"
In addition to scrutinizing judges, in this chapter I will
focus on the courts of which judges are a part. First, I provide
an overview of exactly what courts are and what they do. Sec-
ond, I will explore the role of courts in intergovernmental rela-
tions as they stand alongside legislatures and executives in a
separation-of-powers system.
1
Richard A. Posner, "What Do Judges and Justices Maximize? (The Same
Thing Everybody Else Does)," Supreme Court Economic Review 3 (1993):
1- 41. Quotation on page 2.
2 Of course, in some s.tat es and localities judges are elected at regular inter-
vals, thus making them much more like other elected politicians. However,
judges on higher courts, in both state and federal systems, are typically ap-
pointed to lifetime posts.
Courts and Judges 475
COURTS: A B RIEF O VERVIEW
In order to undertand what animates judicial behavior, I need

to place the judge or justice3 in context by briefly considering
the role of the court system more generally. To do this I will
blithely boil down hundreds of years of jurisprudence into a
few pages. In doing so I emphasize the role of courts as dis-
pute resolvers, as coordinators, and as interpreters of rules.
Dispute Resolution
So much of the productive activity that occurs within families,
among friends and associates, even between absolute strang-
ers takes place because the participants do not have to devote
substantial resources to protecting themselves and their prop-
erty or monitoring compliance with agreements.4 For any po-
tential violation of person or property, or defection from an
agreement, a ll parties know in advance that an aggrieved
party may take an alleged violator to court. The court, in turn,
serves as a venue in which the facts of a case are established,
punishment meted out to violators, and compensation
a warded to victims. The court, therefore, is an institution that
engages in fact finding and judgment.
Many disputes are between private parties, so the court
a The term justice is reserved for members of the U.S. Supreme Court. All
other judges are called "judges," although some judges low in the pecking
order are referred to as "magistrates." (However, entirely inconsistent with
this nomenclature, one of the lowest categories of judges is "justice of the
peace.") Throughout we will si mply call all of them, with due respect, judges.
4 Naturally, some resources are devoted to protection and monitoring. How-
ever, if extraordinary resources had to be devoted, then their rising cost

would cause the frequency of the productive activities alluded to in the text
to decline, according to elementa1·y economic theory. Indeed, since the costs
of negotiating, mo n itoring, and enforcing agreements (what we political
economists call transaction costs) can be very high , they are a serious i mped-
iment to social interaction and p1·oductive activities of all sorts. Economizing
on them-by providing th e services of courts and judges, for exa mple-is one
of the great contribu tions of the modern state to social welfare.
serves principally to determine whether claims of vio-

lation can be substantiated. An employee, for example, may
sue her employer for allegedly violating the terms of a pri-
vately negotiated employment contract. Or a consumer may
sue a producer for violating the terms of a product warranty.
Or a tenant may sue a landlord for violating provisions of a
]ease. In all of these cases some issue between private parties
is in dispute. The court system provides the service of dispute
resolution.
The examples in the preceding paragraph involve ciuil dis-
putes. An entirely separate category of dispute, one in which
the courts also have a role to play, involves criminal violations.
In these cases "the public" is a party to the dispute because the
alleged violation concerns not (only) someth ing involving pri-
vate parties, but (also) a public law. This brings the public
agencies of justice into play as parties to a dispute. When an
individual embezzles funds from his partner, he not only vio-
lates a privately negotiated agreement between them (namely,
a promise of honest dealings), he violates a public law pro-
hibiting embezzlement generally. A court proceeding, in this
case, determines not only whether a violation of a private
arrangement has occurred but whether the alleged perpetra-
tor is guilty or innocent of violating a public law.
In all of these instances, the judge is responsible for man-
aging fact-finding and judgment phases of dispute resolution
(sometimes in collaboration with a jury). Thus, a large part of
the daily life of a judge involves the provision of an independ-
ent, experienced look at the facts, an assessment of whether
the dispute involves a violation of a private agreement or a
public law (or both), and finally a judgment-a determination
of which party (if either) is liable and, if so, what compensa-
tion is in order (to the private party victimized and, if judged
a criminal activity, to the larger public). Judging is a sophisti-
cated blend of reading a mystery novel, solving a crossword
puzzle, and providing wise counsel.
Coordination
Dispute resolution occurs after the fact- that is, after a dis-
pute has taken place. In a manner of speaking, it represents
a failure of the legal system, since one function of law and its
judicial institutions is to discourage such disputes in the first
instance. We may also think of courts and judges as before-
the-fact coordination mechanisms inasmuch as the anticipa-
tion of what happens once their services are called upon allows
private parties to form rational expectations and thereby co-
ordinate their actions in advance of possible disputes. A pro-
spective embezzler, estimating the odds of getting caught,
prosecuted, and subsequently punished, may think twice (or
even three or four times) about cheating his partner. Surely,
some prospective embezzlers are deterred from their crimes by
these prospects. Two acquaintances, therefore, may much
more confidently entertain the possibility of going into busi-
ness together because their partnership will flourish (or fail)
"in the shadow of the law." The sword (or is it the scales?) of
justice hangs over their collaboration.
In this sense, the court system (like the dog in Sir Arthur
Conan Doyle's story "Silver Blaze") is as important for what it
doesn't do as for what it does. 5 The system of courts and law
coordinates private behavior by providing incentives and dis-
incentives for specific actions. To the extent that these work,
there are fewer disputes to resolve and thus less after-the-fact
dispute resolution for courts and judges to engage in. What
makes the incentives and disincentives work is their power
(are the rewards and penalties big or small?), their clarity, and
the consistency with which judges administer them. Bright-
line incentives (clearly defined standards that leave little or no
room for varying interpretation), consistently employed, pro-
6 In this Sherlock Holmes adventure, the key to solving the mystery was that
a certain dog did not bark.
vide powerful motivations for private parties to resolve dis-

putes ahead of time. This sort of advanced coordination, en-
couraged by a properly functioning legal system, economizes
on the transaction costs that would diminish the frequency of,
and otherwise discourage, socially desirable activity.
Rule Interpretation
Dispute resolution and coordination tremendously affect pri-
vate behavior and the daily lives of ordinary citizens. Judges,
however, are not entirely free agents (despite the fact that
some of them are tyrants in their courtrooms). In matching
the facts of a specific case to judicial principles and statutory
guidelines, judges must engage in interpretive activity. They
must determine what particular statutes or judicial principles
mean, which of them fit the facts of a particular case, and
then, having determined all this, the disposition of the case at
hand. Does the statute of 1926 regulating the electronic trans-
mission of radio waves apply to television, cellular phones,
ship-to-shore radios, fax machines, or electronic mail? Does
the law governing the transportation of dangerous substances,
passed in 1937, apply to nuclear fuels, infected animals, or ar-
tificially created biological hazards? Often, the enacting leg-
islative body has not been crystal clear about the scope of the
legislation it passes. Indeed, a legislature acting in 1926 or
1937 hardly could have anticipated technological develop-
ments a half century later. Nevertheless, cases come up on a
regular basis, and judges must make judgment calls, so to
speak, on highly complex issues.
Interpreting thE> rules is probably the single most impor-
tant activity in which higher courts engage. This is because the
court system is hierarchical in the sense that judgments by
higher courts constrain the discretion of judges in lower courts.
If the Supreme Court rules that nuclear fuels are covered
by the 1937 law on transporting dangerous substances, then
lower courts must render subsequent judgments in a manner

consistent with this ruling. The judge in a civil or criminal trial
concernjng the shipment of nuclear isotopes from a laboratory
to a commercial user, for example, must comply in his or her
ruling with the legal interpretations passed down by the
higher courts. Also, because of the federal principle by which
the American polity is organized, federal law and interpreta-
tions thereof often trump state and local laws.
At the highest levels, courts and judges engage not only in
statutory interpretation, but constitutional interpretation as
well. Here they interpret the provisions of the U.S. Constitu-
tion or a state constitution, determining their scope and con-
tent. In determining, for example, whether the act of Congress
regulating ·th e transportation of dangerous substances from
one state to another is constitutional, the justices of the
Supreme Court might appeal to the commerce clause of the
Constitution (aUowing the federal government to regulate in-
terstate commerce) to justify the constitutionality of that act.
On the other hand, a Supreme Court majority might also rule
that a shipment of spent fuel rods from a nuclear reactor in
Kansas City to a nuclear waste facility outside of St. Louis is
not covered by t his law, since the shipment took place entirely
within the boundaries of a single state and thus did not con- ·
stitute interstate commerce.
In short, judges are continually engaged in elaborating,
embellishing, even rewriting the rules by which private and
public life are organized. In these interpretive acts they are
conscious of the fact that their rulings will not only affect the
participants in a specific case before them but will carry in-
terpretive weight in all similar cases percolating up in the
lower courts. Thus, statutory and constitutional interpretations
have a precede ntial authority over subsequent deliberation
(and, in turn, are themselves influenced by earlier interpreta-
tions).
However, all the elaboration and embellishment and "re-
--------------------------------~~~
drafting" of statutes that constitute the interpretive activity of

judges and justices are themselves subject to review. Statutory
interpretation, even that conducted by the highest court in the
land, is exposed to legislative review. If Congress is unhappy
with a specific statutory interpretation-for example, suppose
the current Congress does not like the idea of federal regula-
tion of e-mail that a federal court claimed to be permissible
under the 1926 act on electronic transmission-then it may
amend the legislation so as explicitly to reverse the court
ruling. Of course, if the court makes a constitutional ruling,
Congress cannot then abrogate that ruling through new legis-
lation. Since the famous early nineteenth-century case Mar-
bury v. Madison, the federal courts generally and the
Supreme Court in particular have come to be regarded as the
supreme interpreters of the Constitution. But Congress can
commence the process of constitutional amendment, thereby
effectively reversing judicial interpretations with which it dis-
agrees.
From this brief discussion I hope the reader now appreciates

that courts, judges, and the legal system of which they are a
part serve to resolve disputes when they arise, to discourage
disputes in the first place by providing before-the-fact coordi-
nation of expectations, and to interpret and reinterpret the
fundamental rules of the game by which the members of the
political community regulate and govern themselves. In both
their statutory and constitutional interpretive modes, courts
regulate interactions of ordinary citizens and constrain exer-
cises of authority by political officials and institutions. In
turn, they are subject to regulation by ordinary and extraordi-
nary political processes. Although the regal attire of judges
and the formality of judicial settings serve to intimidate and
to inspire awe, there is no gainsaying that courts are political
institutions and that judges are political players.
I now take up each of these last two facets separately.
First, I ask precisely how we should treat judges as political

players. Then I discuss the political role of courts in the gov-
ernment arrangement of separation of powers in the United
States.
WHO ARE THE J UDGES, AND W HAT

Do T HEY W ANT?
Judge Richard Posner provides partial answers to both of

these questions and, in the process, brings judges under the
rubric of rational-choice theory. Judges, to Posner, are mostly
"ordinary people." Just as for other ordinary people for whom
rationality is a perfectly respectable behavioral hypothesis, it
is appropriate to suppose that judges, too, are rational:
Politics, personal friendships, ideology, and pure serendipity play
too large a role in the appointment of federal judges to warrant
treating the judiciary as a collection of genius-saints miracu-
lously immune to the tug of self-interest. By treating judges and
Justices as ordinary people, my approach makes them 6t subjects
for economic analysis; for economists have no theory of genius. It
is fortunate for economic analysis, therefore, that most law is
made not by the tiny handful of great judges but by the great
mass of ordinary ones. 6
The issue for Posner, then, is how to conceptualize judging
and other judgelike behavior as rational responses to the legal
and political environment. He suggests several "types" that
capture different aspects of judicial behavior.
The Nonprofit A nalogy

The first is the manager of a nonprofit enterprise. The legal
system, Posner observes, is indeed a nonprofit enterprise. Ex-
cept for private judges (like Judge Judy or Judge Joe Brown,
6 Supreme Court Ecmwmic Reuiew 3 (1993): 1-41. Quotation is on pages 3-4.
stars of the television courtroom shows), arbitrators, and me-

diators-all of whom sell their services in the marketplace
(and thus are driven by the same profit incentives that moti-
vate other sellers of services)-judges in the official judiciary
draw fixed salaries and fringe benefits but may not pocket
fees, user charges, or other differences between receipts and
running costs. Moreover, their salaries and other benefits are
fixed independent of effort, quality of work, or any other
performance-based standard. They are salaried managers of a
nonprofit system. Posner opines that because judges cannot
claim a monetary residual like the owners of a for-profit enter-
prise, because they are not compensated in accord with the
quantity and quality of their work like other sellers of profes-
sional services, and because they have lifetime tenure, judges
should be expected on average not to work as hard as lawyers
of comparable age and talent. He qualifies this opinion by not-
ing that federal judges are nominated by political authorities
and must undergo an often thorough confirmation process, so
that it is conceivable that these filters select for industrious-
ness and other manifestations of commitment to the job.
In highlighting the nonprofit aspect of the job of judging,
Posner is able to say what judges do not maximize-namely,
profits. But nonprofit managers are not an entirely unmoti-
vated lot of ne'er-do-wells. Posner alludes to several motiva-
tional elements of the utility function of the judge as nonprofit
manager- popularity among fellow judges, law clerks, and
other court personnel; prestige in the legal and larger political
community; reputation in the academic legal world; track
record (especially the desire not to be reversed by a higher
court or the legislature). No doubt all these elements have
an effect, but Posner is skeptical that these are the central
considerations of judging (and he speaks from personal ex-
perience).
I would give slightly greater weight than Posner to am-
bition. Nonprofit m anagers are, as Posner suggests, "ordinary
people" in the sense that they do their jobs with an eye at least
partly on advancing in their world (however they conceive of
it). The program manager at the Mellon Foundation hopes to
become a vice president at the Rockefeller Foundation and, be-
yond that, perhaps the president of the Ford Foundation. 'rhe
dean of students at a small liberal arts college may aspire to
be dean of admissions at the state university and, beyond that,
perhaps dean of undergraduate education at an Ivy League
college or university. The local director of United Way may
harbor ambitions to rise through that bureaucracy to regional
and then national office. In this manner, "unter" judges hope
some day to become "ilber" judges. Federal district judges hope
for promotion to one of the Courts of Appeals, and "great men-
tioners" in Washington always have their list of appeals court
judges as candidates for the Supreme Court (The recently con-
firmed Justice Sonia Sotomayor is an apt illustration.)
Ambition for advancement, then, substitutes as a standard
for other forms of performance-based compensation. But, as
Posner notes, ambition cannot be a tremendously strong moti-
vation because the odds of advancement, even for the very
highest performing judges, are remote. It is probably more
likely that a dean of students at a high-quality midwestern
liberal arts college can become admissions dean at an Ivy
League institution than that a federal district judge will move
up the judicial hierarchy to the Supreme Court. My view,
then, is that individual judges, like other nonprofit profession-
als, rationalize the energy and commitment they invest in
their jobs by their desire for popularity, prestige, and reputa-
tion. They value these things partly for their own sake be-
cause standing among one's peers and within one's community
is part of the nature of all human beings, a perspective that
goes back at least to Aristotle. But they also value them be-
cause they conceive of these things as necessary attributes for
advancement, however remote the latter possibility might be.
In their heart of hearts, I suspect, many judges probably ex-
aggerate their own importance and, therefore, the likelihood

of lightning striking.
The Voting Analogy

A second aspect of judicial behavior that Posner highlights de-
rives from the fact that much of what a judge does may be con-
ceived of as voting. Rendering judgment is, at least in part,
the casting of a vote (for the plaintiff or defendant). As we saw
in Chapter 9, voting in mass elections is, in fact, difficult to ra-
tionalize on instrumental grounds because no single voter is
very likely to be decisive. Judges, however, are more akin to
legislators or committtee chairs in the sense that their partic-
ipation takes place in a small-group setting in which the
chances of being pivotal to an outcome are much more likely.
Indeed, some judges are committees of one (federal district
judges), others a1·e members of small panels (federal appeals
judges typically sit in panels of three), and still others are
members of moderate-sized panels (Supreme Court justices
are one of, at most, nine). Consistent ideological outliers may
rarely be pivotal in a small or moderate-sized panel, but most
others have a substantial probability of being the vote that af-
fects an outcome, at least some of the time. Since judges may
not be certain exactly when they have a chance to make or
break a majority as opposed to being a peripheral vote, they
may have an incentive to try to get their decisions "right" in
terms of their beliefs about the proper disposition of cases-
this effort may have a payoff.
Even a prospectively pivotal judge may not be moved much
by this prospect if what he or she is pivotal for is "small beer."
A vote on the merits of a case-whether in behalf of the plain-
tiff or of the defendant-has a limited impact. One of the par-
ties to the case wins and the other loses. No one else is
affected by the decision, per se. But judges "vote" not only on
who should win a given case but, through the opinions they
draft, why one side or the other should win. Judges vote not
only with ballots but also with ideas-ideas about the facts,
about judicial principles, about legal reasoning, and about
moral values. It is through the opinions they draft, rather
than the ballots they cast, that judges may influence a wider
collection of interests, since these opinions serve to constrain
lower-court judges in similar cases in the future. 7
The Spectator and Game-Player Analogies

Another type Posner explores is that of judge as spectator-
someone sitting in a theater or sports arena, or just in a com-
fortable easy chair with a mystery novel on his or her lap.
Posner does not have a passive viewer in mind here but rather
someone actively engaged in the drama taking place, em-
pathizing with one or more of the participants and puzzling
through the moral ambiguities the drama presents. At a
slightly more activist level than the spectator is the judge as
game player- as an actual participant in an unfolding drama.
As a legal case develops, according to this interpretation, the
judge, as either of these types, seeks to puzzle through the
facts on display and connect them to the morally appropriate
legal principles (as determined by the judge's own moral com-
pass). The judge derives pleasure, just like a bridge player,
simply from playing the game.
So, What Do J udges Want?

By his consideration of the different ways to think about moti-
vations of judges, and by his elimination of motivations (such
as maximizing profits and ambition for reelection) that pro-
7 Posner (1993: 19) goes on to suggest that judges, like authors and scholars,
may derive intrinsic pleasure from the act of drafting opinions, displaying
their prowess, and disseminating their views. Thus, even those judges who
have realistic beliefs about the likelihood that they will rise through the
court system on the basis of their present actions may nonetheless devote
considerable energy to those actions for the sheer pleasure they bring.
vide leverage in theorizing about entrepreneurs and politi-

cians, Posner provides ample confirmation of my claim that
judges are enigmatic actors in political life.
Posner's important contribution, I believe, is his insight
that judges are difficult to figure because any con nection be-
tween their performance and their compensation has been at-
tenuated by lifetime tenure and fixed income. The difficulty
with Posner's analysis is that (to his credit) he considered
the category of "judge" in all its splendorous variety. If, in
models of elected politicians that I have covered in previous
chapters, my net h ad been as broad, I would have had to lump
presidents and senators together with county sheriffs, city
assessors, and town dogcatchers. Undoubtedly any simple mo-
tivational hypothesis- like maximizing the probability of
reelection- would have fared much more poorly than it does
when we focus mainly on the professional politicians who
dominate our national institutions. Likewise, I believe an ap-
propriate starting point in trying to fathom the judiciary is to
focus on those judges who are "players" in the larger political
game precisely because they do have ambition, both for their
careers and their ideas about the law. Just as our models of
elected politicians are most apt for those who have reached
the political pinnacle-representatives, senator s, presidents,
and those who want to become one of these-I want now to
focus on judges of comparable stature. These judges, I believe,
may be thought of, essentially, as legislators in robes.
Legislators in Robes
In many models of legislative politics, legislators are thought
of as policy oriented. In the spatial model (as described in
Chapter 5), for example, a legislator (or, in Chapter 16, a po-
litical party), is characterized by an ideal policy and prefer-
ences that decline as one moves spatiaJly more distant from
that ideal policy. It is often unnecessary for the purposes of
the model to identify why the legislator possesses these partic-

ular preferences. But if one were to pull back the curtain, so to
speak, and speculate about why legislators hold the prefer-
ences they do, there are usually two answers.
The first is that legislative politicians actually care about
the role of government-about the policies government pur-
sues, and about the politics that get you to those policies. That
may well have been their motivation for becoming active in
politics initially. Indeed, to the extent a legislative politician
cares a bout policy because her philosophy of government is
relatively well worked out, she may be described (nonpejora-
tively) as an ideologue, or at the very least a thinker.
The second rationale for legislator policy preferences is po-
litical ambition. According to this view, a legislator may not
care much about policy per se but does care about protecting
his political flank from attack by political opponents. To do
this, he pursues policy objectives in behalf of constituents. He
is the faithful agent, serving the policy interests either of his
current constituency or of some broader constituency toward
which his ambitions point him.
We may think of judges as having policy preferences just
like legislators. To the degree that higher courts do not merely
resolve disputes between the participants in a case but, more
significantly, shape the legal context in which millions of pri-
vate citizens interact and in which thousands of public offi-
cials exercise power, these courts are critical in the formation
and implementation of public policy. Higher-court judges,
therefore, have oppor tunities to affect tbe public condition.
There is simply no way, for example, to deny the central role
played by the federal courts in transforming post-World War
II America into a considerably more integrated and equitable
society in terms of race and gender.
Now surely not all courts cases have the momentous effects
of the Dred Scott case just before the Civil War, or Brown v.
Board of Education just before the civil rights revolution. But,
within fairly broad limits, it can be claimed that big-impact

cases are often difficult to discern in advance. So, at least as a
first approximation, it may be hypothesized that judges with
policy preferences treat each case as though it might have im-
pact, either directly on national politics or indirectly through
its effects on how constitutional or statutory law is subse-
quently interpreted.
This is a radical view of judges, one with which most sit-
ting judges (or aspirants thereto in the broader legal commu-
nity) would be uncomfortable. They would be uncomfortable, I
believe, because this conception of judges makes them much
like other politicians, a view judges seem intent on discourag-
ing. This strikes me as a rather prissy attitude. Besides, it is
no badge of dishonor to be associated with wanting to have
impact on the political life of the community, whether for ideo-
logical or careerist reasons. At the end of the day, this view of
judges becomes compel1ing if it leads to interesting insights
about how politics works. I claim that the legislator-in-robes
conception of judges has precisely this potential, a claim I
elaborate in the next section.
CASE 15.l
THE B EST JUDGES M ONEY C AN BUY?
Sir Francis Bacon, one of the great philosophers, scientists,

and legal scholars in English history, was impeached after
having attained the highest judicial position in England,
Lord Chancellor. The House of Commons determined that
he had accepted twenty-eight bribes. Bacon defended his
position by claiming, "I usually accept bribes from both
sides so that tainted money can never influence my deci-
sions." The Parliament didn't buy it and sentenced him to
the Tower of London, where he remained until King J ames
pardoned him.
I present this little vignette only to add credibility to my

claim that jurists should be thought of in the same terms as
other politicians, with the same variety of motives (save
electoral ambition) that we encounter when thinking about
legislators, executives, and bureaucrats. Graft and corrup-
tion, things that do not shock or surprise us when we en-
counter them occasionally in the hurly-burly of real politics,
is undoubtedly a factor in the lives of judicial politicians as
well.
Nevertheless, legal ethics, judicial temperament, a nd
the culture in which courts and judges find themselves, on
the face of things, appear to mitigate the connection be-
tween judicial activity and self-interest. A perusal of the
Canons of Judicial Ethics of the American Bar Association
makes this easy. One canon, titled "Self-Interest," states: "A
judge should abstain from performing or taking part in any
judicial act in which his personal interests are involved."
Another, titled "Independence," states, "A judge should not
be swayed by partisan demands, public clamor or consider-
ation of person al popularity or notoriety, nor be apprehen-
sive of unjust criticism."
Do these canons of proper behavior do any good? The an-
swer is, probaibly, yes, so judicial politicians are probably
less venal and corrupt than their compatriots in the elected
branches of government. The reason is that violations of the
canons provide a basis for dismissal; judges ambitious to re-
main judges, therefore, will observe them, thereby assuring
that the judiciary, as a group, maintains a stature and pres-
tige in public opinion that other politicians do not enjoy. But
these canons probably provide far weaker incentives to in-
duce judges to park their respective ideologies outside the
courtroom door. In this sense, they share with elected politi-
cians both the motive and the opportunity to impart their
own political beliefs into their official activity.
COURTS AND I NTERGOVERNMENTAL

R ELATIONS
In conceiving of judges as legislators in robes, I am effectively

claiming that judges, like other politicians, have policy prefer-
ences they seek to implement. I needn't say much at this point
about where these policy preferences come from, only that
they exist and provide the motivation for judicial activity. (I
will have a little more to say about their origins in the con-
cluding section.)
At both the national and state level in the United States,
courts and judges are "players" in the policy game because of
the separation of powers. Essentially, this means that the leg-
islative branch formulates policy (defined constitutionally and
institutionally by a legislative process); that the executive
branch implements policy (according to well-defined adminis-
trative procedures, and subject to initial approval by the pres-
ident or the legislative override of his veto); and that the
courts, when asked, rule on the faithfulness of the legislated
and executed policy either to the s ubstance of the statute or to
the constitution itself. The courts, that is, may strike down an
administrative action either because it exceeds the authority
granted in the relevant statute (statutory rationale) or be-
cause the statute itself exceeds the authority granted the leg-
islature by the constitution (constitutional rationale). As I
noted earlier, if a decisive coalition in the legislature is un-
happy with this judicial action, then it may either recraft the
legislation (if the rationale for striking it down was statutory)
or initiate a constitutional amendment that would enable the
stricken-down policy to pass constitutional muster (if the ra-
tionale for originally striking it down was constitutional).
A number of recent scholarly works have sought to model
the judiciary in the setting outlined above. In the next few
paragraphs, I will briefly give the reader some flavor of this
analysis that employs the one-dimensional spatial model de-

scribed in Chapter 5. Some in the literature refer to this as
Marksist analysis in honor of Brian Marks, the first to attempt
to incorporate the judiciary in spatial models of policy choice.8
A number of simplifying assumptions are employed,
mainly to make the presentation less complicated; most au-
thors believe that additional complications would not change
the basic thrust of the conclusions. To begin with, it is as-
sumed that the policy options in any given setting may ade-
quately be represented by a one-dimensional interval, say
[O, 100], over which all actors-legislative, executive, and
judicial- have single-peaked preferences. Thus, the nine
Supreme Court justices have ideal points on this interval la-
beled x/, ... , x}, whereas the ideals of the 100 senators
are given by x/, ... , x8 100 , those of the 435 representatives by
xH,i . . . , xH435 • an d t h e pres1·dents
, executive
· agent by xA" (I
will ignore the complexities associated with the presidential
veto and veto overrides.) For each of these actors, preferences
are highest for their respective ideal policies and decline as
policy moves away to the left or the right from this ideal.
Those with ideals near 0 may be thought of as on the extreme
liberal end of opinion, while those with ideals near 100 are at
the extreme conservative end. Thus, to take a recent Supreme
Court case involving minority representation in municipal po-
lice and fire departments, if the [O, 100] dimension were taken
to characterize affirmative action policies, those judges, legis-
lators, and administrators favoring the setting aside of ear-
marked positions and resources for minorities (set-asides,
quotas) would have ideal points toward the zero end of the
8 For a critical review, the reader may consult Jeffrey A. Segal, "Separation-of-
Powers Games in the Positive Theory of Congres and Courts," American Po-
litical Science Reuiew 91 (1997): 28-44. The original paper on which the
preceding is based is Brian A. Marks, "A Model of Judicial Influence on Con-
gressional Policymaking," Working Papers in Political Science, P-88-7,
Hoover Institution, Stanford, Calif., 1988. Unfortunately, it has never been
published.
continuum; those favoring outreach and other expand-the-

pool-of-eligibles policies, as well as the use of race as an "ad-
justment" factor, would have ideal points somewhere in the
center of the continuum; and those who opposed affirmative
action would have ideal points clustered near the 100 end-
point.
Although each chamber of the legislature is, as we saw in
previous chapters, organizationally complex, to make things
tractable here we assume that they are simple majority-rule
institutions.9 Given single-peaked preferences, Black's Median-
Voter Theorem (see Chapter 5) tells us that we can take the
policy ideal and preferences of the median voter in the cham-
ber to represent majority preferences there. We thus may sup-
press the details of preferences of all legislators and focus
exclusively on xH and x 8 , the median-voter ideals in the House
and Senate, respectively. By the same reasoning, to know how
the court decides, given that it, too, is a majority-rule institu-
tion, we need only focus on xJ, the ideal of the median justice
on the court. Finally, we need to take account of the fact that
in the policy dimension in question, there is a status quo pol-
icy, xQ. In this simple setting, then, we can see the impact of
judicial oversight by focusing on xQ, xJ, xH, x 8, and xA. These
five "parameters" will determine the outcome of policy in any
specific application once we know the sequence in which be-
havior unfolds.
I assume at the beginning of any play of the policy game
that a status quo legislative choice, xQ, is in place-the result
of previous iterations of the game. The administrative agent,
either the current president's aide or a bureaucrat with long
9 The complexities produced by committees with gatekeeping and agenda-

setting powers, or by a chair or speaker with powers of recognition, or by
party caucuses with the power to appoint their members to committees can
be, and have been, accom modated in this literature. To ease the presenta-
tion. I have suppressed these details, but not because they cannot be incor·
porated in principle.
DISPLAY 15.1
Sequ ence of t he Policy Game
Step Player Action
0 xQ in place
1 administrative agent implements policy A

2 court strikes A or lets it stand;
if former, then it names
the new policy J
3 legislature if unhappy with policy in
place, either A or J , then
selects new policy, L
tenure, decides what policy to implement, given this legisla-

tive choice in place. 10 The court then decides whether to strike
down the agent's actions. Although, as noted above, the court
may employ either a statutory or a constitutional rationale, I
will focus only on the former. If the court declares the admin-
istrative agent's act as outside the permissible bounds pre-
scribed by the legislation, I suppose the court's majority
opinion can declare whatever policy it wishes. In effect, the
court can "legislate." Finally, if the House and Senate are
jointly dissatisfied with the outcome of this process-either
the original agent's act that the court lets stand or the new
policy enunciated by the court, then they can propose new leg-
islation. In principle, this last action sets the stage for a sub-
sequent round of the game. In this way it is possible to study
the long-term evolution of a policy. For my purposes here,
however, this is an additional layer of complexity that I do not
wish to embrace. So I assume that in step 3, if there is a new
legislative policy, it is automatically implemented and the
game ends. This sequence is given in Display 15.1.
10 See the discussion of McNollgast's model in Chapter 13.

FIGURE 15.1
0 100
Given this sequence, the distribution of parameters will

determine the rational response at each step of the process.
To see the model in action, consider the distribution given in
Figure 15. l. Existing policy, xQ, is in equilibrium, lying as it
does between the median ideal of the House and Senate. If an
administrative agent had no court to worry about, then it
would implement policy A= xH, since this is the policy closest
to its ideal that will not trigger a legislative response, despite
the bureaucratic drift. 11 But the presence of a court changes
the calculations.
Suppose the median justice's ideal, xJ, were between 0 and
xA-as might have been true of the liberal Warren court on
affirmative action issues. The court majority couldn't hope to
improve on xfl, given its preferences, so it would let the imple-
mented policy A = xfl stand. So, too, would the legislature, as
we've already seen. So, for any court to the left of the bureau-
cratic agent's ideal, the agent's action is unaffected. By the
same logic, a more conservative court, with a median justice
ideal in the interval [xA, xfl], would be similarly disposed; so
once again the bureaucratic choice would be A= xH. In each of
these cases it would appear to an unsophisticated observer
that the court is irrelevant. This, of course, would be an inap-
propriate inference, since the agent makes the choice she does
11
Recall from the analysis of the McNollgast model in Chapter 13 that any
point in the subinterval between xH and x5 (including its endpoints) is an
equilibrium in the sense that any movement away from one of these points
will be opposed by either the House or the Senate (or both). Since a legisla·
tive reaction to an agency decision requires the assent of both the House
and the Senate, these points constitute equilibria.
because she knows her preferences are sufficiently in line with

those of the court that the justices will let her action stand.
What about xJ in the interval [xH, x 5 ]? This is a circum-
stance in which, no matter what the bureaucratic choice, the
court can strike down the action and name its ideal, J = xJ, as
the new policy. Since this lies in the legislative equilibrium
set, the legislature will not react with new legislation. If the
bureaucratic agent doesn't like being reversed, then she will
anticipate the court's move in advance and simply name A - xJ
as her policy. She can do no better. Thus, we see that the pres-
ence of a court with a median ideal inside the legislative equi-
librium set is a check on bureaucratic drift. Once again,
though, if the agent does, in fact, act on her anticipation of the
court's preferences, it will appear as though the court is irrel -
evant since it will leave the agent's choice in place.
Finally, suppose xJ is to the right of the median senator-
in the interval [x8 , 100)- as may well be the case on affirma-
tive action issues for the Roberts court. Any action taken by
the bureaucratic agent will be struck down by the court; it will
declare the new policy as J = x 5 . This is the best the median
justice could hope for, since any policy closer to xJ will trigger
a legislative response-some policy in the [xw x 5 ] interval
worse from the median justice's perspective than x 5 . Of course,
a fully anticipatory response from the bureaucrat would have
her declare A = Xs• with the court allowing that policy to stand.
Even if the bureaucrat wer e, for some odd reason, to imple-
ment a policy A to the right of x5-even at the ideal policy of
the median justice, the court would still strike it down and de-
clare J • x8 . Why? If the court allowed this nominally more de-
sirable policy to stand, then the legislature would get involved
with corrective legislation, ultimately producing a policy less
desirable to the court median than x 8 .
This little example hardly does justice to the growing body
of work that seeks to integrate the courts into the intergovern-
mental policy game. 12 What it nevertheless conveys is that the

court's role may often be subtle and unobservable. 1'he equi-
librium policy of this game, as we have seen, is often sus-
tained by anticipatory behavior. The bureaucratic agent
implements a policy in the expectation that, should the court
or the legislature see it as in their respective interests, either
can act in a corrective fashion after the fact. If the agent fully
takes this expectation on board in her own calculations, then
she will make a choice that does not trigger these after-the-
fact reactions. Consequently, in many policy areas the court
will be inactive, not because it has no role to play but rather
because other players in the game have already accommo-
dated it. Where we ought to expect proactive involvement by
the court, then, are those areas in which relevant parameters
have changed but other players have failed to notice or to
react appropriately to them.
CONCLUSION
The legislator-in-robes conception of judges is, I believe, an

extremely plausible way to think about judicial actors in a
separation-of-powers regime. For different configurations of
preferences, we obtain provocative and often nonobvious im-
plications about where and whether there will be observable
judicial activity. In principle, these implications can be tested
against the experiences of the real world. This, after all, is
what social science is all about. The interested reader may
wish to play with Figure 15.1 a bit, determining how adminis-
12 A highly accessible discussion of this approach , with many citations from
other related work, may be found in John Ferejohn and Barry Weingast,
"Limitation of Statutes: Strategic Statutory Interpretation," Georgetown
Law Journal BO (1992): 565-83. Equally accessible, and especially good at
playing out the McNollgast model as it app lies to courts is McNollgast, "Pol-
itics and the Courts: A Positive Theory of Judicial Doctrine and the Rule of
Law," Southern California Law Reuiew 68 (1995): 1631-83.
trative agents, legislators, and judges will behave under alter-

native preference distributions.
Nevertheless, I cannot conclude without expressing at
least some doubts about this conception of judges. Putting
judges on a par with legislators as strategic players does not
sit well with some students of judicial politics (see the citation
to Segal in footnote 8). They view the typical judge of one of
the higher courts as a professional jurist who renders opinions
in line with a well-thought-out judicial philosophy. Legisla-
tors, presidents, and bureaucrats may strategize, but jurists
do not. They sustain this view by pointing to a jurist's body of
opinions, demonstrating that there is often a consistency to
them that would seem to be at odds with a more strategic vi-
sion . While sharing some of these doubts, I should point out
that there is no reason to believe the behavior just identified
is necessarily at odds with a strategic vision. As we saw in the
example in the previous section, when other actors, behaving
strategically, internalize the potential response of the court,
they often accommodate the court and thereby remove the
need for justices to behave strategically. It is the potential for
strategic intervention, and the belief by other politicians that
judges will behave strategically if their preferences are not ac-
commodated, th at provides the basis for the strategic vision of
judicial behavior I have developed here.
In s um, there is no gainsaying the fact that judges ar e
shadowy characters in most rational-choice theories of policy
making. The attenuation of incentives linking judicial politi-
cians to constituencies to which other politicians are linked
opens the door to a variety of alternative motivational hy-
potheses. I have reviewed several of them in this cha pter, and
developed one-legislator-in-robes hypothesis-in some detail.
Regardless of where a reader stands on this question, I hope
to have per suaded him or her that courts are essential politi-
cal institutions in intergovernmental policy making.
C ASE 15.2
L EGISLATORS IN R OBES REVISITED
In this chapter I have argued that judges are usefully

thought of as legislators in robes, motivated by policy pref-
erences and ideological goals just like other politicians. To
evaluate the legislators-in-robes hypothesis, we can turn to
the writings of judges themselves.
The hypothesis that judges make policy in addition to
interpreting laws is controversial. Robert J ackson, Supreme
Court Justice from 1941 to 1954, has written that "few ac-
cusations against the Supreme Court are made with more
heat and answered [by Court members] with less candor
than that it makes political decisions."* The lack of candor
to which J ackson alludes in the responses of justices to this
charge is, in my view, due to the presence of more politics
than the justices would care to admit to. This claim is sus-
tained by Dalin H. Oaks, Justice of the Utah Supreme
Court, who observes that "the conventional wisdom holds
that the legislat.ure makes the law, the courts interpret it,
and the executive enforces it. Like most conventional wis-
dom, this is only partly true. Judges also make law. They do
so inevitably as they interpret statutes passed by the legis-
lature, since interpretation can never be free from choices
illuminated by the creative instinct and motivated by per-
sonal preference.""t Federal Circuit Court of Appeals Judge
Howard T. Markey adds, "All people have values of some
sort, and judges are people. A per son totally devoid of values
·Robert H. Jackson, "The Supreme Court as a Political Institution," in

Alan F. Westin, ed., An Autobiography of the Supreme Court (New York:
Macmillan, 1963), p. 360.
I Dalin H. Oaks, "When Judges Legislate," in Mark W. Cannon and David
M. O'Brien, eds., Views from the Bench: 1'he Judiciary and Constitutional
Politics (Chatham, N.J.: Chatham House, 1985), p. 147.
would be a robot, an automaton." He continues, "It would

be foolish to pretend that no single Federal judge has ever,
even subconsciously, viewed his or her roles as one author-
izing a personal policy-making function. The elements of of-
fice can lead to confusion of the terms 'appointed' and
'annointed.' The temptation to 'do .good' can he strong.":j:
I would only add that, precisely because it is inevitable,
it is hardly something about which to be particularly
ashamed. Indeed, I would further claim that legislatures
often delegate this lawmaking function to courts. I noted in
Chapter 13 that the growth of governmental responsibili-
ties has led Congress to implement "fire extinguisher"
forms of monitoring and enforcement. Accordingly, citizens
and groups are given legal standing to work out legislative
detail in courts, where judges are the ultimate arbiters. In
effect, Congress knowingly gives judges substantial lati-
tude in how the laws are to be implemented. In this sense,
judges make policy and therefore are legislators in robes
not only because they are "people with values" but because
Congress has ·e ncouraged the courts to share their legisla-
tive burden in an increasingly complex policy-1naking envi-
ronment.
i Howard T. Markey, "On the Cause and Treatment of Judicial Activism,"
in Cannon and O'Brien, Views from the Bench, pp. 285-88.
1. The Chief Justice of the United States, John Roberts, drew

the following analogy- now oft-repeated by those who favor a
conservative judiciary- in the opening statement of his confir-
mation hearings in front of the Senate Judiciary Committee in
September 2005: "Judges are like umpires. Umpires don't
make the rules; they apply them. The role of an umpire and a
judge is critical ... but it is a limited role .... It's my job to
call balls and strikes, and not to pitch or bat." Explain in your
own words Roberts's vision of the role of courts. What judicial
role was he emphasizing, and what others was he implicitly
criticizing? Then, assess Justice Roberts's metaphor overall. Is
it sufficient for a Supreme Court justice to just "call balls and
strikes"?
2. The federal court system, as well as most higher state courts,

grant judges lifetime terms, and judges cannot be removed
from the bench unless they committ a serious crime. This sug-
gests that the judiciary may be the site of an extreme form of
the principal-agent problem. Explain why the principal-agent
paradigm might be a valid description of the relationship be-
tween politicians and judges, and then describe any before-
the-fact and after-the-fact means of limiting agent discretion
employed by the political principals. Then, discuss some of the
positive aspects of lifetime tenure, and explain why such a
system has been so widely adopted in the United States.
3. Sandra Day O'Connor was a moderately conservative

Supreme Court justice from 1981 to 2005 who frequently
played the role of the court's median voter by joining the
court's four other conservative justices as the deciding vote in
5-4 decisions. According to various witnesses, O'Connor ex-
pressed dismay on election night 2000 when Florida prema-
turely declared a Gore victory, and her husband is supposed to
have explained that she did not feel she could retire if a Demo-
cratic president was in power.
Assume that Supreme Court justices care only about policy,

and that their policy preferences can be represented by a one-
dimensional issue space ranging from liberal to conservative.
Also assume that any justice nominated by the president will
be approved by the U.S. Senate. If one of the current justices

were to retire, what would be the policy position of the new
court if the current president is a liberal Democrat, to the left
of the median justice? Consider three cases: retirement of the
median justice, retirement of a justice to the left of the me-
dian, and retirement of a justice to the right of the median.
Now answer the same question if the current president were a
conservative Republican, to the right of the median justice.
Does your answer help explain Justice O'Connor's desire for a

Bush election victory? How does your answer depend on
O'Connor's ideal point relative to her immediate neighbors to
the left and the right?a
4. The Judiciary Reorganization Bill of 1937 would have au-

thorized the president to appoint an extra judge to any federal
court for each sitting judge over the age of seventy. This bill
came on the heels of a string of defeats of New Deal legislation
by the Supreme Court (in everything from 5-4 votes up to
unanimous opposition). With six justices over age seventy, this
bill, if passed, would have permitted President Roosevelt to
add up to six new justices. Assuming the justices' preferences
can be represented with a unidimensional spatial model, would
the "court-packing plan" have been sufficient to guarantee
more preferable decisions in the future in all cases? Could FDR
at least have guaranteed rulings more to his liking in some
cases?
*5. Ferejohn an,d Weingast (1992)b made a novel argument

about strict interpretations of congressional statutes using a
• For further elaboration on these questions, see David Rohde and Kenneth
Shepsle, "Advising and Consenting in the 60-Vote Senate: Strategic Appoint-
ments to the Supreme Court," Journal of Politics, 69 (2008): 664-77.
b John Ferejohn and Barry Weingast, "A Positive Theory of Statutory Inter-
pretation," International Review of Law and Economics, 12 (1992): 263-79.
one-dimensional spatial model. Assume that the median legis-

lator in the two houses of Congress has ideal points at xH and
x 8 and xH < x 8 , and the Supreme Court is hearing an argu-
ment about some statute passed in a previous session of Con-
gress, which is at xQ. Suppose that the Supreme Court acts
first and can amend the statute via a ruling or leave it alone.
The Congress acts second and can replace xQ with new legisla-
tion-if both houses approve--or do nothing. Assume also that
the median Supreme Court justice's ideal point is at xQ. Would
the Supreme Court vote to alter the bill if Xu< xQ < x 8? Why or
why not? What would occur if xQ < xH or xQ > x 8 and the
Supreme Court then decides not to alter the law through a
ruling? What would then be the optimal strategy to defend the
status quo when xQ does not lie between the ideal points of the
two houses of Congress? Ferejohn and Weingast conclude that
"judges who care about statutes are well-advised to pay more
attention to political realities than to the words of statutes
themselves." Why do they reach this conclusion?
6. What motivates judges, both in the types of decisions they

make and in the amount of effort they devote to researching
and articulating their positions? Do motivations differ across
different courts- for example, between federal trial courts, ap-
pellate courts and the Supreme Court; or between elected
county judges and unelected state judges?
16
Cabinet Government and
Parliamentary Democracy
So much of the material in this volume may seem to revolve

around political life in America. Whether in describing the pri-
vate ways in which individuals organize themselves into clubs
and organizations to pursue collective purposes or the public
ways in which they constitute their official political institu-
tions, I may have conveyed the impression that I have been
doing no more than analyzing politics, American style. Prob-
lems of group choice, collective action, and institutional poli-
tics, however, are not unique to the American constitutional
order. Americans have no monopoly on voting, agenda setting,
or dividing and. specializing political labor, and I can make
this point quite forcefully by turning to the ways in which
most of the rest of the democratic world organizes its public
politics.
In this final chapter on institutions, I lead a quick tour of
institutional arrangements in parliamentary democracies.
Rather than separating power into distinct branches of gov-
ernment, each checking and balancing the others as is done in
the United States, the world of parliamentary democracy is
organized according to a principle that emphasizes the coor-
dination and concentration of power. The textbooks refer to
this as the "fusion of powers" in contrast to the American-
style "separation of powers." The centerpieces of this arrange-
503
ment are the supr emacy of parliament and the accountability

of the political executive-the cabinet, the government, or the
administration-to it.
THE PARLIAMENTARY SETTING

Electoral Arrangements
The best place to begin is with electoral arrangements. Amer -
ican legislatures, both state and federal, are elected according
to the first-past-the-post principle. Legislative candidates
campaign for office in districts, with the one r eceiving a plu-
r ality of votes there elected as the dis trict's r epr esentative. I
noted in Chapter 7 that this arrangement, which is also fol-
lowed in Great Britain and some other of its former colonies,
encourages two-party or two-candidate competition in each
electoral district (Duverger's Law). If it is the same two par-
ties in each district, then one of them will end up with a leg-
islative majority.
Although, as just noted, some parliamentary regimes like
Great Britain elect legislators in this fashion, the vast bulk of
such regimes, especially those in continental Europe (includ-
ing the democracies that emer ged in eastern Europe late in
the twentieth cen tury) elect legislators according to one of sev-
eral proportional r epresentation (PR) formulas. 1 There is
tremendous var iety in the specific PR details, but however
they are implemented, these arrangements tend to produce
parliaments in which many parties are represented. 2 Indeed,
because parliamentary seats a re distributed among many par-
ties instead of just two, it is rare for any one party to com-
mand a majority in parliament. Thus, with the exception of
1 Some parliamentary democracies, like those in Germany, J apan, and New
Zealand, have a hybrid electoral system in which some legislators are
elected from districts according to first-past-the-post, while others are deter·
mined by a PR formula.
2 See Gary Cox, Making \.btes Count (New York: Cambridge University Press,
1997).
Cabinet Government and Parliamentary Democracy 505
Britain and a few other instances in which a single party cap-

tures a parliamentary majority, it is necessary in most parlia-
mentary democ1·acies to engage in multiparty negotiations in
order to organize government.
Governmental A rrangements
In parliamentary regimes there is a division and specializa-
tion of labor. Tihe House of Commons does not run British
politics. Nor does the Dutch 'Iiveede Kamer, the German
Bundestag, the Japanese Diet, or the Norwegian Starting run
the politics of the Netherlands, Germany, Japan, or Norway,
respectively. Rat her, each parliament "elects" a government to
serve as the executive arm of the regime. I thus need to de-
scribe exactly what a "government" is on the one hand, and
how a country's parliament "elects" one on the other.
The politica] executive in a parliamentary democracy is,
with a few exceptions that I won't stop to consider here, cho-
sen by parliament. This executive is called the government; it
is also known as the cabinet or council of ministers. It is a col-
lection of senior politicians each of whom is the head of a de-
partment or ministry of state. In nearly every parliamentary
regime there is a finance minister, foreign minister, interior
minister, defense minister, justice minister, education minis-
ter, environment minister, and so on.
Thus, there is a division and specialization of labor at two
different levels. The first distinguishes between legislature
and executive. The legislature selects the executive in the first
place; keeps it in place or replaces it with a different one; and
considers various pieces of legislation, the most important of
which is the annual budget. 3 I will discuss these features mo-
3 It should be noted that legislatures in parliamentary regimes are not the
hyperactive lawma king engines we encounter in the American setting. Typi-
cally, parliaments vote a few broad delegations of statutory authority which
are then implemeroted by the government. In the United States, on the other
hand, Congress and the various state legislatures legislate with much
greater frequency and in a much more fine-grained fashion.
mentarily. The second level of division and specialization of

labor is in the government itself. In a manner quite parallel to
the arrangements involving committees in the U.S. Congress,
each ministry of state has jurisdiction over specified dimen-
sions of public policy, called a ministerial portfolio. In this do-
main the minister and his or her senior civil servants have
considerable discretion to interpret statutory authority and
implement public policy.
Another way to think about this arrangement is as a chain
of principal-agent relationships. Parliament as principal dele-
gates executive authority to a collective agent, the cabinet.
The cabinet as principal, in turn, delegates discretionary au-
thority in various policy jurisdictions to its agents, namely
particular cabinet ministers. To control its agents, and coordi-
nate their activities, the cabinet employs various before-the-
fact and after-the-fact mechanisms; usually one member of
the cabinet plays the role of policeman and maestro-the
prime minister. Likewise, to control its agent, parliament, too,
exercises before-the-fact and after-the fact authority. We refer
to this authority as that of "making and breaking govern-
ments," since parliament votes a cabinet into office in the first
place and, if it is unhappy with cabinet performance after the
fact, may vote it out of office and replace it with an alternative
cabinet. 4
The Government Formation Process

A typical sequence of events in one cycle of parliamentary
democracy begins with a triggering event like an election.
When the new parliament convenes there are often a number
of parties represented, no one of which commands a majority
of votes on its own. Each party has an endowment consisting of
4
These details are developed in Michael Laver and Kenneth A. Shepsle, Mak-
ing and Breaking Governments: Cabinets and Legislatures in Parliamentary
Democracies (New York: Cambridge University Press, 1996).
a certain number of seats under its control and a set of policy

priorities and positions on which it has just conducted its elec-
toral campaign.5 Also at this time there will be a government
in place-the cabinet that was running the country just before
the election. The first order of business facing the new parlia-
ment is whether to retain the status quo cabinet or replace it
with a new cabinet. If the political fortunes of the various par-
ties have changed as a result of the election, it is highly likely
that a new cabinet will be selected, one better reflecting the
balance of political forces in the new parliament. But since no
one party is a majority in its own right, the new government
will have to reflect the preferences of a coalition of political
parties. Government formation requires coalition building.
Anyone watching this process, at least th at part of it that
is not conducted behind closed doors, will know when it has
reached its conclusion. This is when a parliamentary majority
allows a new set of senior politicians to take over the reins of
each of the ministries of state. This transition is normally ac-
companied by much fanfare, including a press conference in
which the new ministers "meet the public" and announce their
policy goals. In some parliamentary regimes, there must be
a formal vote of investiture in which the new government is
formally approved by a parliamentary majority. In others,
the head of state (the king or queen or president) simply an-
nounces the new government. In all cases the new govern-
ment survives at the pleasure of parliament. At any time,
& In effect, each party may be treated as a unitary actor. Unlike legislative
parties in the United States, which contain within them a wide range of
opinion and whose members are relatively free to pursue their own private
objectives, parliamentary parties are far more homogeneous in terms of pol-
icy preferences, and their leaders have powerful mechanisms by which to
control the rank and file. Parliamentary party leaders, of course, must keep
their followers happy, as I argued more generally in Chapter 14, but they
may proceed in a manner that is best for their party as a whole. In effect,
then, we may think of the parliamentary party as acting as if it had well·
defined preferences, as manifested in the beliefs and choices made by its
leaders.
parliament may entertain a motion of no confidence in the

government, usually moved by an opposition party politician;
if this passes, then the government must resign and the gov-
ernment formation process starts up anew. 6 Alternatively, the
government itself may introduce a motion of confidence, or at-
tach such a motion to an important bill before the parliament.
If that motion or bill should be defeated, then the government
is obliged to resign. 7
So, the government formation process ends with the crea-
tion of a new government-a new set of managers of the
various departments of state. But what of its beginning? Or -
dinarily, when either a new parliament convenes, an old gov-
ernment is defeated in a confidence procedure, or the old
government resigns for whatever reason, the head of state
appoints a leading politician, called a formateur, to try to as-
semble a new government. In some countries either the con-
stitution explicitly, or constitutional conventions implicitly,
restrict the discretion of the head of state. He or she must, for
example, allow the prime minister of the previous government
first crack at forming a new government. Or he or she must
allow the leader of the largest party in the new parliament
first shot. In other cases, the head of state is free to choose
whoever he or she wishes as formateur. (In some circum-
stances, the head of state appoints an informateur-a re-
spected elder statesman with no personal ax to grind-to
consult widely in the political community and then advise him
or her on whom to appoint formateur.) The formateur negoti-
6 Indeed, a vulnerable government normally can anticipate whether a parlia·

mentary majority no longer has confidence in it; as a result the government
will resign in advance of the motion-of-no-confidence procedure so that the
latter is not carried out to its ignominious conclusion.
7 Why would a government do this? The chief reason is to raise the stakes of
defeating the bill to which the motion is attached. It is a power play, a show-
down move, in which the government puts parliament on notice that if it
should defeat the government's bill, the government will come crashing
down around its ears.
ates with other parties in assembling a distribution of minis-

terial portfolios to political parties- a government-that a
parliamentary majority is prepared to support. The parties re-
ceiving a seat at the cabinet table (by being assigned a minis-
terial portfolio) are said to be the government parties. Those
not included in government are the opposition parties (though
some of these m ay lend the government support in parlia-
ment). I examin e this process shortly, as well as the motiva-
tions that drive it.
Once in place, the new government governs. Each minister
attends to day-to-day administration with senior civil ser-
vants in his or her ministry and, together with other minis-
ters, meets regularly in cabinet to formulate and coordinate
overall governmental policy. This will entail not only the im-
plementation of existing policy but also the formulation of new
policy. And the latter means that the government will need to
have firm control of parliamentary politics, setting parlia-
ment's agenda and making sure it toes the line. Thus, there is
a political tension between government and parliament, with
the former in the d!1ver's seat because of its control over day-
to-day management of political affairs; but since the latter
"makes" the government in the first place, it also can "break"
the government . In short, the government governs subject to
keeping the confidence of parliament.
A MENAGERIE OF GOVERNMENTS
In the U.S. Con,gress, the House and Senate are organized by

the party capturing a majority of seats in that chamber. It is
always the case in the modern era for either the Democrats or
Republicans to win a chamber majority (although it wasn't
always this way, especially early in American constitutional
experience). In contrast, the process of government formation
in parliamentary democracies briefly described in the last few
pages produces governments of quite a variety of sizes and

shapes (Table 16.1).
The number of parties represented at the cabinet table can
be one or many-in which case the executive is a single party
or a coalition, respectively. The number of seats in parliament
controlled by the government parties can constitute either a
majority or a minority. Thus, there are four different types of
government. Their relative frequencies in Europe since 1945
are given parenthetically in Table 16. l.
Unified governments, which arise in slightly more than one
of every eight governments in the period of study, are formed
by parties that win outright parliamentary majorities. This is
a relatively rare event, which occurs frequently in Great
Britain but only occasionally anywhere else in Europe. In this
case the majority party votes itself all the cabinet portfolios.
The only serious threat to these "juggernauts" is intraparty
factionalism. One faction of the majority party, if sufficiently
numerous, can bring the government down by deserting it on
a key vote. 8
Much more common, occurring about half the time in the
postwar era, is a cabinet composed of several parties that
jointly control a parliamentary majority. These multiparty ma-
jority governments need to control their rank and file in order
to continue in office, but the ordinary tools of party discipline
are typically sufficient for this task. A ''backbencher" who
threatens to stray off the reservation may find that his or her
parliamentary career prospects suddenly become bleak, not to
speak of the fact that the party organization will refuse to
back him or her at the next election.
8
In Great Britain in 1994, the Conservative government of Prime Minister
John Major was a unified government, but only tenuously so. It had a bare
majority in the House of Commons and thus was potentially vulnerable to
factional blackmail from within its partisan ranks. In parliamentary votes
on expanding the role of the European Union in the spring of 1994, so-called
Euroskeptics inside the Conservative Party threatened to desert the govern-
ment (which favored further European integration). Major called their bluff
and won ... that time.
Cabinet Government and Parliarnentary Democracy 511
TABL E 16.1
P ARLIAMENTARY GOVERNMENT FORMS*
Governmen t P a rties in P arliament

Majority Minority
Single Party Unified Single-party
(.134) minority (.238)
Executive
Coalition Multi party Multiparty
majority (.500) minority (. 128)
•Cell entries give the proportion of each type as determined by Strom for
a sample of Western parliamentary democracies, 1945-1982.
SOURCE: Kaare Strom, Minority Government and Majority Rule (New York:
Cambridge University Press, 1990), p. 61.
A more serious threat to these governments is that during

their incumbency, one or more of the governing parties will be-
come dissatisfied and resign, essentially uniting with the op-
position to bring down the government.9 Short of pulling out of
a sitting government in order to defeat it, a governing party
may insist on a reallocation of portfolios. If, for example, re-
cent poll results suggested that one of the minor government
parties was rising in popularity at the expense of one of the
major government parties, the former could insist on a cabinet
reshuffle in which it received more portfolios, or more impor-
tant portfolios.
A single-party minority government sounds pretty strange
to American ears (although I suggest shortly that the reader
has more familiarity with it than he or she realizes). It is a
government in which one party receives all cabinet portfolios,
9 If one of the governing parties pulls out, then, as we noted earlier, a n actual
vote of no confidence need not occur. If t he parties remaining in the govern-
ment believe they no longer have the s upport of parliament, the government
may simply resign, triggering a new round of government formation. Alter-
natively, the prime minister may petition the head of state to dissolve parlia-
ment and call for new elections.
but that party has less than a majority of parliamentary

seats. It governs at the sufferance of an opposition that, in
combination, controls a majority of parliamentary seats. How
can this be? It is able to retain office, as I shortly demonstrate,
because the opposition, even though a majority, cannot agree
on an alternative that it prefers. Imagine a large social demo-
cratic party-slightly left of center~ontrolling, say, 40 per-
cent of parliamentary seats, that faces a radical left-wing
party, an extremist right-wing party, and a religious party,
controlling roughly 20 percent of the seats each. It is quite
likely that anything preferred to a social democratic minority
government by the left-wingers is strongly opposed by the
right-wingers and religious partisans, whereas anything pre-
ferred to the social democrats by the two more conservative
parties is opposed by the left-wingers. The opposition majority
simply cannot get its act together, allowing the social democ-
rats to survive in office even though they comprise but a mi-
nority of the whole parliament. This is not at all an unusual
scenario and, as the evidence reported in Table 16.1 suggests,
something like it occurs nearly one-quarter of the time in
Western parliamentary democracies.
To complete the picture, about one in every eight govern-
ments in the postwar Western experience is a coalition gov-
ernment whose partners control less than a parliamentary
majority. Typically they will be center-left or center-right
coalitions that, like the single-party minority governments de-
scribed in the previous paragraph, split the opposition.
Before turning to an analysis of government formation, it
is worth mentioning that Americans often get depressed over
a phenomenon conventionally referred to as divided govern-
ment. This is a condition in which one party, say the Republi-
can Party, controls the White House, but the Democrats
control one or both of the legislative chambers. After the 2006
elections, Republican George W. Bush occupied the White
House, but the Democrats had captured both houses of Con-
gress. It is alleg,e d by inside-the-Beltway pundits that this is a

formula for gridlock as partisan majorities in the different
branches of gov,e rnment prevent anything from happening. 10
Such pundits often cast an alluring look across the Atlantic,
touting the advantages of parliamentary government. Yet, the
data of Table 16.1 suggest that parliamentary regimes are also
highly prone to divided government. Multiparty majority coali-
tion governments are those that control a parliamentary ma-
jority, but divided government occurs within the executive.
Single-party minority governments have a unified executive
but, lacking a m ajority in parliament, are divided in exactly
the same way American governments often are-an executive
whose party lacks a legislative majority. (Thus, despite sound-
ing pretty strange to American ears, single-party minority gov-
ernments are structurally identical to the divided governments
with which Americans have had frequent experience.) Minority
coalition governments are divided both within the executive
and between executive and legislature. Indeed, European par-
liamentary experience suggests that governments are divided
about seven-eighths of the time, a proportion roughly equal to
the postwar American experience. P arliamentary government,
in short, is no cure for divided government.
ANALYZING GOVERNMENT FORMATION
I now want to analyze how governments of various types actu-

ally are formed. Government formation in parliamentary
democracies is a subject that has been studied analytically for
10
This does not ofte n comport well with the actual facts (though this never
stopped a pundit worth his or her salt). President Reagan enjoyed .enor·
mous s uccess in h is first term despite the fact that the Republicans did not
control a majority in the House of Representatives. Alternatively, both
Jimmy Carter and Bill Clinton enjoyed (hardly the right word!) unified gov·
ernment (in 1977- 81 and 1993-95, respectively), but each would undoubt-
edly have given his eye teeth for a record of accomplishments match ing that
of Reagan's first term.
more than forty years. Scholars have produced an immense

literature and I shall not be able to explore every nook and
cranny. 11 Instead, I develop an example in some detail. The ex-
ample typifies many of the possibilities that are encountered
in the real world, but it is hardly typical, since the real world
contains an enormous amount of variation (as suggested in
Table 16.1). My purpose, however, is not so much to present
the reader with a full-blown model of government formation
as it is to show how rational politicians think through all the
many permutations of the government formation game in
order to pursue their objectives effectively.
German Example: Basic Setup

In Germany throughout the last quarter century, four parties
won seats in the German parliament, the Bundestag: the
Christian Democrats (CD), the Social Democrats (SD), the
Free Democrats (FD), and the Greens (G). The CDs are a clas-
sic center-conservative party-procapitalist on economic is-
sues, proactive in foreign policy, and traditional on social
issues. The SDs are a classic social democratic party-left.
leaning on economic matters, pacifist in the foreign realm, and
progressive on social issues. The FDs are a liberal party-
free-market oriented on economic issues, moderate in foreign
policy, and progressive on social issues. The Greens, finally,
are more left-leaning than the Social Democrats on economic
and foreign affairs and more progressive than the Free De-
mocrats on social issues.
Although the seat totals for each party varied from year to
year, the strategic structure in Germany did not. Specifically,
rarely did a single party win an outright majority of seats.
Second, the CDs were sufficiently numerous that they could
11 The standard bibliographic reference surveying this field is Michael Laver
and Norman Schofield, Multiparty Government (New York: Oxford Univer-
sity Press, 1980).
form a majority with any of the other three parties. Third, a

majority excluding the CDs r equired that all of the remaining
parties coalesce. Thus, the set of possible majority coalitions
in the Bundestag during this period included
· CD-SD
• CD-FD
• CD-G
• SD- FD-G
as well as any of the above with an additional party added.
A government (or cabinet), recall, is a distribution of minis·
terial portfolios among the parties (more accurately, among
senior party politicians). Suppose there were onl y two key cab-
inet ministries in the government-Finance (F) and Foreign
Affairs (FA) . Display 16.1 lists all the possible German gov-
ernments involving these ministries. There are four ways to
assign the Finance portfolio to a party and, for each of these,
four ways to assign the Foreign Affairs portfolio, giving a total
of sixteen possiible governments. 12 In the case at hand, then,
there are four possible majority coalitions in the Bundestag
(listed above) and sixteen possible governments (given in Dis-
play 16.1).
12 If, as is often the case in the real world, there are more than the four par·
ties in the example (say, p in number), and more than two portfolios (say, q
in number), then there are pq possible ways to divvy the portfolios up
among parties, and thus pq possible governments. Even for moderate values
of p and q, this number can grow rather large. With ten parties (as was ap-
proximately t he case at various times in countries like Belgium or Italy)
and fifteen government ministries, for instance, there are ten million billion
different conceivable governments. We would have to harvest an entire for-
est to print a list like the one in Display 16.1!
D ISPLAY 16.1
POSSIBLE P ORTFOLIO ALLOCATIONS IN THE B UNDESTAG
P a r ty holding
Fin ance p ortfolio (F) For e ig n Affa irs po r tfolio (FA)
1. CD CD
2. CD SD
3. CD FD
4. CD G
5. SD CD
6. SD SD
7. SD FD
8. SD G
9. FD CD
10. FD SD
11. FD FD
12. FD G
13. G CD
14. G SD
15. G FD
16. G G
German Example, Continued: Party Preferences

and Majority Preferences
We now need to consider how each party (and ultimately how
each parliamentary majority) assesses each of the sixteen gov-
ernments in Display 16.1. There is no right answer to this
matter, since a party's assessment of a particular government
will depend on what the party cares about. In the literature
on government formation, two motivational hypotheses have
been seriously entertained.
DISPLAY 16.2
OFFICE-SEEKING P ARTY P REFERENCES OVER
ALTERNATIVE G E RMAN G OVERNMENTS*
CD SD FD G
{l} {6} {11} {16}
{2, 3, 4} {5, 7, 8} {9, 10, 12} {13, 14, 15}
{5, 9, 13} {2, 10, 14} {3, 7, 15} {4, 8, 12}
{6, 7, 8, 10, 11, {1, 3, 4, 9, 11, {l, 2, 4, 5, 6, {1, 2, 3, 5, 6,
12, 14, 15, 16} 12, 13, 15, 16} 8, 13, 14, 16} 7, 9, 10, 11}
* The numbers in each column refer to the goverments listed in Display
16.1. A party prefers higher-listed to lower-listed governments and is in-
different among bracketed governments.
-
OFFICE-SEEKING. The first, the office-seeking motivation, as-
sumes that parties care only about getting into office. Thus,
the CDs, in looking over the possibilities in Display 16.1, will
most like government 1, giving them both portfolios. If they
car e more about financial matters than foreign affairs (and I
will assume this, for the sake of argument, for all parties),
they will next prefer governments 2, 3, and 4 (among which
they are indifferent), giving them the Finance portfolio and
some other party the Foreign Affairs portfolio. Next they will
prefer governments 5, 9, and 13 in which they get the Foreign
Affairs portfolio and some other party gets Finance; again,
they are indifferent among these three governments. Finally,
they least prefer (and are indifferent among) governments 6,
7, 8, 10, 11, 12, 14, 15, and 16, in which they are excluded
from government altogether. Display 16.2 gives the office-
seeking preferences of each of the four parties.
Since we know which combinations of parties constitute
Bundestag major ities (the bulleted list above), we can com-
pute majority p1·eferences over the sixteen possible govern-
ments from these individual preferences. For example, the

CD- FD majority prefers government 3 (CDs get Finance and
FDs get Foreign Affairs) to government 6 (SDs get both portfo-
lios), since 3 ranks ahead of 6 in both party preference order-
ings. It is a bit tedious, and not particularly enlightening in
this particular example, to examine all sixteen governments
to see how majorities rank them. It is tedious because there
are 120 such comparisons to be checked (though a computer
with even a nwdest amount of power could do this in the blink
of an eye); it is unenlightening because, as Display 16.2 re-
veals, there is a substantial amount of indifference in party
preferences, the latter because the only thing parties consider
i::> whether they are in government or not. For these reasons I
go no further in an analysis based on office-seeking, though
the reader now can see how it could be carried out. (This is
one of those proverbial "homework exercises" that the enter-
prising reader may wish to pursue.)
POLICY-SEEKING. The second motivational hypothesis, that of

policy-seeking parties, is richer in its possibilities, and ar-
guably a more realistic conjecture about party behavior as
well, so I explore it somewhat more deeply. Needless to say,
ambitious politicians love the prospect of advancing to high of-
fice. The ego boost of power, not to speak of the creature com-
forts of a commodious office, an army of defere ntial assistants,
the services of a leather-upholstered limousine, and frequent
travel to exciting world destinations, all contribute to the
allure of office. There is no gainsaying the potency of the
office-seeking motivation to explain the wishes of individual
politicians. But parliamentary politicians are not entirely free
agents. They are members of parties which have been elected
to the parliament to acco1nplish things. Party leaders must
check the acquisitive instincts and other private motives of
their senior party rnembers to ensure that these partisan pur-
poses are not jeopardized.
In particular, I conjecture that a more compelling motiva-

tional consideration derives from the fact that parties serve
external constituencies, delivering policies desired by those
constituencies in exchange for electoral support. External con-
stituencies don't really care whether a politician from the
party is a cabinet minister or not. They do care about the poli-
cies produced by the government in power. Always attentive to
electoral considerations and ever mindful of the fact that elec-
toral constituencies are the principals for which the party is
an agent, party leaders seek to install governments that are
disposed toward policies preferred by these constituencies.
This may mean pushing to get fellow party members ap-
pointed to the cabinet, but it may not. Thus, despite the indi-
vidual politician's lust for office, the party system, electoral
competition, and the principal-agent relationship between
parliamentary parties and external constituencies check and
constrain this motivation. 13
To see what is involved in a policy-seeking analysis of gov-
ernment formation, we need to know what policies a party
wants to pursue on the one hand, and what policies a party
believes other_parties are pursuing on the other. To facilitate
this, I employ the by now familiar spatial model initially de-
veloped in Chapter 5. Based on our crude characterization of
the German parties earlier in this section of the present chap-
ter, we portray their policy preferences in a simplified, two-
dimensional spatial map (Figure 16.1).
The horizontal dimension captures economic policy, with a
left-wing radical positioned to the left and a right-wing conser-
vative to the right. The vertical dimension describes foreign
policy, with a more pacifist orientation at the bottom and a
more activist int-ernational posture at the top. Thus, given the
ia I also note that it doesn't matter for our purposes whether a party is "re-
ally" committed to the policies it pursues (ideological parties), or instead
whether it does so only to secure electoral benefits from satisfied con-
stituents (instrumental parties).
FIGURE 16.1
FOREIGN
POLICY
ECONOMIC POLICY
preferences of their electoral supporters, the CDs are the most

activist party in the Bundestag on foreign policy issues and
one of the more conservative on economic issues. The black dot
next to their name locates the CD "ideal" policy on both dimen-
sions. In a similar manner we locate the other three parties.
I assume that the horizontal dimension falls into the juris-
diction of the Finance Ministry and the vertical dimension is
the responsibility of the Foreign Affairs Ministry. I also
assume-and this is a strong assumption-that all parties in
the Bundestag know that whichever party secures each of the
ministries will push its ideal policy in that jurisdiction. 14
14 It is a strong assumption because it stipulates not only that a minister has
a great deal of authority in his or her jurisdiction (something that any ob·
server would accept), but also that other players on the parliamentary
scene have virtually no influence over policy once the minister has been
I can now state precisely what is meant by policy seeking.

A policy-seeking party will, if given the authority to do so, im-
plement its ideal policy in the jurisdictions in which that au-
thority applies. A party that obtains both portfolios, therefore,
will implement its ideal point. A party that obtains one portfo-
lio will implement the component of its ideal point that ap-
plies to the jurisdiction of that portfolio. 1'hus, if the CDs form
a single-party minority government, they will implement the
policies associated with their ideal point . If, on the other
hand, the CDs and FDs form a multiparty majority govern-
ment, with the CDs getting the Finance portfolio and the FDs
the Foreign Affairs portfolio, then this government will imple-
ment the policies associated with the point identified by the
intersection of the horizontal line through the FD ideal and
the vertical line through the CD ideal. (It is the point from
which the shaded petals originate-these will be described
momentarily.) That is, the CD-FD government will implement
the CD's ideal economic policy and the FD's ideal foreign
policy.
German Example, Concluded: Equilibrium in

Government Formation
In Figure 16.1 the four horizontal and the four vertical lines
through each of the party ideal points define the foreign policy
and the economic policy, respectively, that each of the four
parties would implement if it secured the relevant cabinet
portfolio. Thus, each of the sixteen possible governments (the
ones listed in Display 16.1) is associated with one of the six-
designated. The only recourne for a parliamentary majority is to t:.opple the

government. As long as a minister does not cross that line, he or she is oth-
erwise unconstrained. Tu say that this assumption is strong is effectively to
say that it is very unrealistic. Models, as I hope I made clear way back in
Chapter l, are always um·ealistic; the important issue is whether they are
"close enough for government work" to permit some insight. This the reader
must judge after thinking about the argument in its entirety.
teen points in Figure 16. l defined by the intersection of one

horizontal and one vertical line. For example, the point at the
intersection of the vertical line through the CD ideal and the
horizontal line through the FD ideal locates the policies of
the CD-FD government in which the CDs get the Finance
portfolio and the FDs the Foreign Affairs portfolio. I refer to
these sixteen points so defined as the lattice of possible gov-
ernments.
The ideal policies not only define what policies a party
would implement if it were given the power to do so, but they
also allow us to characterize party preferences over other poli-
cies. As in Chapter 5, I make the simplifying assumption that
parties prefer policies closer to their ideal to those farther
away from their ideal. With this assumption we can analyze
the viability of alternative governments, each of which is
judged by the various Bundestag parties in terms of the poli-
cies they are postulated to pursue.
Consider the CD-FD government just discussed. (Through-
out this discussion, the first-named party obtains the Finance
portfolio and the second-named the Foreign Affairs portfolio.)
To determine how the CDs assess this government, draw a cir-
cle through the CD-FD point centered on the CD ideal (as
shown in Figure 16.1). The CDs are indifferent among all the
points on this circle and prefer any point inside it (because
each point on the circle is exactly as far away from the CD
ideal as is the CD-FD point, and all points inside the circle
are closer to the CD ideal than the CD- FD point). The circle is
an indifference curve, and the area inside it is a preferred-to
set in the language of Chapter 5. Notice that the only govern-
ments the CDs prefer to the CD-FD government are the
single-party minority government, in which the CDs get both
portfolios enabling them to implement their ideal point, and
the FD-CD government in which they swap portfolios.
The three other circles in Figure 16.1 reflect the prefer-
ences of the remaining parties regarding the CD-FD govern-
Cabinet Government and Parlia1nentary Democracy 523
ment. Th see how parliamentary majorities feel about this gov-

ernment, we look at relevant intersections of these circles. Re-
call that there are four majority coalitions (the bulleted list
given earlier). The shaded petal pointing northeast is the set
of policies preferred by the CD-FD Bundestag majority to the
policy that a CD- FD government would implement. While
there are policies that this majority prefers, there is no gov-
ernrnent (alternative allocation of portfolios as reflected in the
various lattice points) that it prefers. There are two shaded
petals pointing northwest (one is only slightly bigger than the
other), reflecting the policies that the CD-SD and CD- G Bun-
destag majorities prefer to the policies that would be imple-
mented by the CD-FD government. Again, there are policies
each of these majorities would rather see implemented, but no
government (no lattice point) that would do it. Finally, only
barely visible to the naked eye is a sliver of a shaded petal
pointing almost due south- the policies preferred by the
SD- FD- G B undestag majority. Again, this majority prefers no
government to CD-FD .
In short, no other government is preferred by any of the bul-
leted parliamentary majorities to the CD-FD government. In
the language of Chapter 5, this government has an empty win-
set relative to th e fifteen alternatives the Bundestag might
consider. If this government were installed and set about im-
plementing the CD ideal on economic policy and the FD ideal
on foreign policy, then it would constitute an equilibrium of
the government formation process. Any motion of no confi-
dence would fail, since a Bundestag majority would not pass
such a motion unless it ha d an alternative government in
mind to replace the incumbent. 15
It is actually rather remarkable that such an equilibrium
exists, since typically in multidimensional spatial analysis
16 In fact, this is actuaDy t he government that dominated German politics

throughout the last quarter century (with occasional exceptions).
there is no majority-rule equilibrium (see Chapter 5). The rea-

son there is one in this example is because we are considering
only the sixteen lattice points, not the entire two-dimensional
space of points. I have restricted focus to the points on the lat-
tice because the government formation process chooses gov-
ernments, not policies. If the Bundestag chose policies directly,
then there would be no need for a government at all, since
Germany would be run directly by its parliament. The fact is,
however, that in Germany and every other parliamentary
democracy, its parliament does not run the country; it chooses
a government which does. So our focus on governments (the
lattice of points) rather than policies (the entire space of
points) is appropriate.
Let me conclude this brief equilibrium analysis by noting a
couple of additional findings. First, although we did find an
equilibrium in the German example-an empty-winset point
from among those on the lattice-it does not always work out
so nicely. Examples can be produced in which no lattice point
is an equilibrium. In this case, majority preferences cycle over
all the lattice points. However, systematic empirical work on
the real world suggests that equilibrium is very likely. 16
Second, an empty-winset lattice point may not be the only
equilibrium. By this we do not mean that there can be more
than one empty-winset point, for this is virtually impossible. 17
Rather, I am suggesting that there is a different way to think
about equilibrium. We display this for our German example in
Figure 16.2. There the analysis focuses on a CD single-party
16 This evidence is found in Laver and Shepsle, Making and Breaking Govern-
ments, chapters 6-9.
17 Informal proof- Suppose x and y are two lattice points. Logically, either x is
preferred by some majority toy, or y is preferred by some majority to x (both

cannot be true). Of course, a third possibility is that x and y tie in a vote be-
tween the two. lf either of the first two scenarios holds, then the losing can-
didate does not have an empty winset, so that at most one of them does.
Only in the third scenario is it possible for there to be more than one empty-
winset lattice point.
FIGURE 16.2
FOREIGN
POLICY
1 - - - - - - + --
SD
G
ECONOMIC POLICY
minority government in which the CDs receive both portfolios

and implement their idea) policy. Clearly this is the best re-
sult for the CDs-they would certainly participate in no Bun-
destag majority trying to replace this government. The only
majority which excludes the CDs is the SD-FD-G majority.
Indifference curves through the CD ideal point for each of
these three parties are shown in the figure, and their common
intersection is sh aded. This shaded area describes the winset,
W(CD,CD), the set of policies that all three parties prefer to
the one implemented by the CD government. Notice that there
are three lattice points in this winset (indicated by x). Thus, a
Bundestag majority prefers any government giving the CDs
the Finance portfolio and any other party the Foreign Affairs
portfolio to an all-CD government: W(CD,CD) - {(CD,FD),

(CD,SD), (CD,G)}.
I claim, nonetheless, that t,he all-CD government is an
equilibrium. But I need another assumption, one I believe the
reader will not find offensive. And this is that no party may be
forced into government against its will. Put differently, each
party has a veto over every government in which it is a partic-
ipant. It can block such governments from forming by simply
refusing to join. With this assumption in hand, notice t,hat the
contents of W(CD, CD) given in the previous paragraph con-
tains three governments, all of which include the CDs. Thus,
by our last assumption, the CDs can block those governments
from replacing (CD,CD). So we conclude: If an all-CD govern-
ment were in power, then no motion of no confidence could suc-
ceed, because the only prospective replacement governments
preferred by a Bundestag majority would be vetoed by the
CDs, and the only replacement governments that could not be
vetoed by the CDs are not preferred by a Bundestag majority.
Okay, the reader might concede, an all-CD government is
indeed an equilibrium. But how could such a government have
been selected in the first place? That is, why would a Bundes-
tag majority have ever handed over the keys to all ministries
to the CDs, especially if they appreciated the conclusion we
just drew? This is a very good question, but I believe there is a
good answer. Im agine that once upon a time the CDs were a
majority party all by themselves. Naturally enough, they
formed an all-CD government, since they had the votes in the
Bundestag to do that. Now suppose there is a new election,
and the CDs lose their majority. The new parliament convenes
to form a government, but the CDs are the incumbent sitting
government and they stay in power until they are replaced by
the new parliament. In this situation, as Figure 16.2 graphi-
cally displays, the CDs can remain in power as a minority gov-
ernment because the new parliamentary majority is unable to
replace it with something they prefer that is veto-proof.
Cabinet Governrnent and Parliamentary Democracy 527
I thus have derived two different kinds of government

equilibrium. One, displayed in Figure 16.1, is an allocation of
portfolios to which no other allocation is preferred, an empty-
winset government. The second, displayed in Figure 16.2, is
an allocation of portfolios to which other allocations are pre-
ferred, but these are vulnerable to veto.
CONCLUSION
This chapter has served as something of an antidote to the

possible misimpression that rational behavior in institutional
contexts is a distinctly American practice. In fact, institutional
arrangements elsewhere, though they surely differ in funda-
mental ways, nonetheless serve some of the same purposes
that I portrayed American institutions as serving in the other
chapters of this part of the volume. Thus, political labor is
divided and specialized; authority is delegated; and before-
the-fact and after-the-fact control mechanisms are utilized.
Rational actors in these political settings exercise sophistica-
tion and foresight in determining what actions will best serve
their interests.
In the case of parliamentary behavior, in which the govern-
ment (a specialized agent) is selected by a parliamentary ma-
jority (the principal), partisan political leaders calculate about
government formation in terms of objectives they wish the se-
lected government to serve. If office-seeking animates their
behavior, then they will maneuver, whether in voting to bring
a government down or to form one in the first place, to include
their partisan colleagues at the cabinet table. If policy-seeking
drives behavior, then the calculation will be in terms of which
policies they prefer to see implemented. I have done no more
than develop a specific example in detail. But I hope to have
displayed how one might analyze politics, whatever the insti-
tutional context.
1. Like any complex institution, parliamentary democracies

feature a division and specialization of labor, between the par-
lia ment and the government it elects on the one hand, a nd be-
tween the various cabinet portfolios or ministries within the
government on the other. Discuss these arrangements in gen-
eral terms. Which institutions are tasked with which activi-
ties? What purposes does this administrative setup serve? Is
there any delegation involved, and if so, how do the principals
keep their agents in line?
2. Mueller (2003)A provides an example of coalition formation

in a one-dimensional setting to show that behavioral hypothe-
ses strongly affect predictions of how coalition governments
will form. Suppose some country with PR ends up with the fol-
lowing distribution of 100 parliamentary seats among five
parties, ordered from left to right: A(l5) < B(28) < C(ll) <
D(33) < E(l4) . First, find all minimum winning coalitions
(MWC)- that is, coalitions with greater than fifty votes but
for which removaJ of any one party will make them a minority.
Then, find the smallest minimum winning coalition. Then,
find the MWC with the fewest members, which might be a
plausible prediction if bargaining becomes more complex as
the number of pa rties increases. Then, find all MWCs for
which the parties are adjacent in the political space. Is one of
these hypotheses more believable than the others? Does it de-
pend on whether you think politicians are policy-seeking or
office-seeking?
•See Dennis Mueller, Public Choice III (New York: Cambridge University
Press, 2003), pp. 280-81.
3. Draw two sets of axes ranging from 0 to 10, with one dimen-
sion representing Finance a nd the other Defense, and suppose
that there are three parties (A, B, and C) with ideal points
(1, 2.5), (3.5, 1) and (9, 7). Any coalition of two parties (with
one controlling each portfolio exclusively) is sufficient for a
majority government, but no one party has an outright major-
ity. What is the set of possible minority and majority govern-
ments? Is there a stable majority government coalition (i.e.,
with an empty winset)? Is there a minority government that is
a stable equilbrium (if we assume that no party can be forced
into a government against its will)? Illustrate all of your an-
swers graphically and briefly explain your logic, taking special
care to explain how minority governments come about.
4. In proportional representation systems it is almost always

true that no single party wins an outright majority of votes.
Using the logic of Downsian electoral competition, explain
why this might be so.
5. After reading this chapter, what is your reasoned opinion

a bout the merits of proportional representation within parlia-
mentary democracies, relative to the U.S. system of single-
member districts and presidential government? In your
discussion, consider at least representation; divided govern-
ment and gridlock; the government formation and breakdown
process; and effective governa nce.
17
Final Lessons
To summar ize this volume in all but the most superficial fash-
ion is more than the reader is likely willing to tolerate. So I do
so superficially. Part I provided the basic building blocks for
political analysis- individuals, their preferences, and their
beliefs. I emphasized there the importance of analysis and,
with the rational choice paradigm, I nailed my colors to the
mast by laying out an analytical approach.
Part II focused on group choice by voting. I demonstrated
that groups, unlike individuals, could not always be treated as
unitary because the aggregated preferences of group members
did not always add up. Neither the method of majority rule
nor, from Arrow's Theorem, any other "reasonable" method of
preference aggregation can guarantee rational group prefer-
ences. This fact, in turn, means that individuals controlling
motion making and other agenda mechanisms can strategi-
cally manipulate things to their advantage. Combine this with
the fact that there are many ways in which to stage an elec-
tion or structure a vote, and the bottom line is that the result
of group deliberation is, in many respects, artifactual. It is de-
pendent on the individual preferences and beliefs of group
members, to be sure. But it also depends upon who makes
motions, in what order motions are made, what the voting
rule is, and many other features besides. One might well want
the outcome of group choice to depend on individual prefer-
ences and beliefs. But it is considerably less clear that it is de-
530
Final Lessons 531
sirable for group choice to depend upon the many arbitrary

features of the processes by which preferences and beliefs are
aggregated.
Part III examined groups as ongoing entities, many of
whose decisions and actions are taken by methods other than
voting. I took the notion of "cooperation" as fundamental and
built it up from two-person bargaining (over such things as
whether or not to drain a marsh) to multiperson coordination
and collective action problems. We saw that many of the diffi-
culties facing modern societies are, in fact, manifestations of
the inability of large groups to coordinate individual behavior
in mutually profitable ways. Consequently, in equilibrium,
life is not all sweetness and light: public goods are underpro-
duced, while public ''bads" are overproduced; there are too
many negative externalities and too few positive externalities;
common resources are mismanaged, overutilized, and poorly
superintended. But this is only to observe that human soci-
eties are not utopias and that there is always room for
improvement.
Part IV examined some of the ways in which human soci-
eties engage in such improvement. They invent institutions,
some of which allow for more efficient performance (taking ad-
vantage of a division and specialization of labor) and others
of which enhance coordination, standardization, and the "ra-
tionalization" of expectations. Some of these institutions are
involved in the making of selections (legislatures and par-
liaments), others in the implementation of selected options
(bureaucracies), while still others orchestrate (leadership in-
stitutions) and validate (courts and judges).
The methodology implicit in our various forays into "ana-
lyzing politics" is portable. You can take it with you, so to
speak, to analyze novel political situations as they present
themselves to you. So many of our life experiences take place
in group settings, ranging from two-person emotional part-
nerships, to small-scale attachments (for example, rooming
groups or families), to larger associations (like religious

congregations or amateur sports leagues). Politics, as I en-
deavored to argue in the very first chapter, attaches to all
such group interactions. To analyze these situations, you need
to appreciate what it is the individuals want. what they be-
lieve, what actions are feasible for then1, and what are the
rules of interaction-the mechanisms by which individual ac-
tions (chosen for r ational reasons) are transformed into out-
comes. It is also often useful to know a bit of history- about
what transpired to produce the circumstance you are now in-
terested in analyzing; this history is often instructive as we
seek to figure out what people want, what they believe, and
what they think others want and believe.
Another lesson I hope the reader takes from this book is
that the methodology of analysis I have described in these
pages applies to the "small p" politics of private life as well as
it does to the ''big p" politics of public life. Politics attaches to
all group phenomena. lt doesn't just take place in forums of
official deliberation. You don't need robes or uniforms or for-
mal proceedings to have politics. The politics of the kitchen
table is as interesting and complex as the politics of the cabi-
net table.
Finally, analyzing social life with the methodology of ra-
tional choice spans political science, sociology, economics, law,
and philosophy and probably other subjects as well. Within
political science, the rational choice approach is not limited to
the study of American politics, though I drew many examples
and illustrations from that arena. As Chapter 16 endeavored
to illustrate, rational choice approaches to the study of parlia-
mentary politics are well developed. So, too, are other aspects
of comparative politics. International relations, whether in
the study of national security, trade, or other forms of strate-
gic interaction, is something that has been subjected to ra-
tional choice analysis for many decades now. Even political
philosophy lends itself to the rational-choice approach. It is
Final Lessons 533
hard to imagine a complete examination of what is just, or

good, or fair wit hout repairing to a consideration of what is ra-
tional for those involved.
If you've taken some of these lessons away from a reading
of this volume, then the enterprise of writing it has been worth
it to me (and the enterprise of reading it has been worth it to
you). If you haven't, then return to Chapter 1 and try it again!
In either case, I encourage you to dip into the ever-growing
literature on the subject, some of the more important contri-
butions to which I have footnoted along the way. I believe the
approach surveyed in these pages enriches the study of poli-
tics, but with this volume you have only scratched the surface.
Index
AARP, 303 Black's Single-Pea kedness Condi-

abortion issue, 183n tion, 82-83, 90, 92- 96
Academy of Motion Picture Arts and fairness ver sus consistency, 73, 82,
Sciences, 212-13 89
acid rain, 331 Gibbard-Satterthwaite Theorem
Administrative Procedures Act, 429 and, 172
after-the-fact authority, 385-86, independence of irrelevant alter·
388 natives,69-70,80-81
agency slippage (drift), 433-37 on legislative intent, 75-76
agenda power, 125, 131, 388 majority rule and, 76-81
agenda procedure, 47 May's Theorem and, 76-81
agenda setters McKelvey's Chaos Theorem and,
defined, 47 154
group choice and, 185-88 nondictatorsbip condition, 70, 80-
leaders as, 445, 451-57 81
agents overview, 71-74
leaders as, 446-.51 restrictions on conditions, 81-84
principal-agent relationship, 420- unanimity condition, 69, 80- 81
32 universal domain condition, 69, 80,
air pollution, 322-23, 330-34 82, 84, 109
Alger, Horatio, 232 value restriction condition, 84
altruism, Prisoners' Dilemma and, The Art of Political Manipulation
253- 59 (Riker), 173
amendment control rule, 392-93 average value voting strategy, 187
American Automobile Association, avoid-the-worst voting strategy, 187
317 Axelrod, Robert, 244
Americans for Democratic Action
(ADA), 439-40 backward induct ion, 163- 67
American Truckers Association, 317 Bacon, Francis, 488
Anderson, John, 122 bargainin g, bilateral, 417- 20
Andreoni, James, 257 Barry, Brian, 50
anonymity condition (May's Coro!· before-the-fact power
lary), 77-78, 80 agenda power, 125, 131, 388
antimajoritarian restriction (Arrow's appointment process, 429
Theorem), 64 behavioral economics, 13n
Apple Computers, 363n beliefs/belief systems
appointment process, 429 defined, 16- 17, 30-31,289
approval voting, 194- 95, 198- 99 in democratic societies, 299
Aristotle, 305 multiperson cooperation and, 289-
Armies of the Night (Mailer), 263 91
Arrow, Kenn eth, 68, l l 7n, 213n in rational choice model, 16-17,
Arrow's Theorem 30-33
antimajoritar ian restriction, 64 Bentley, Arthur, 265- 66
assumptions, 68- 72, 76 bilateral bargaining, 417- 20
background, 67-68 bilateral monopoly, 419
535
536 Index
Ilirney, .James, 222 2:~ campaign contributions, 387- 88

Bismarck, Otto von, 1fi4 cap-and-trade program, 322, 833
Rlack, Duncan, 82, 154 - 55 Carmichael, Stokely, 458
Black Power movement, 458 Carnahan, Jean, 22
Black's Median-Voter Theorem Carnahan, Mel, 21-22
centripetal tendency, 116-17 Carter, Billy, 122n, 5l:ln
legislators and, 389, 492 CBI (Confederation of British Indus-
overview,96-99. 105, 123 try), 283
Plott's Theorem and, 108 central tendency. 208
Black's Single-Peakedness Theorem, certainty, rational behavior and, 31-
82- 83, 90, 92- 96 32
bliss point, 92 Civil Rights Act (1964), 461
I31unL, Matt, 21-22 civil rights politics
Board of Education, Brown u., 487 collective action, 265, 284
Borda count, 193, 195, 198-200 Dred Scott case, 487
Bradley, Bill, 66- 67 Johnson and, 461
Brams, Steven, 194 legislative intent and, 75
breast cancer amendment, 162 Civil War, tax reform and, 9, 64- 67
British politics, 117 class action suits, 250
Brown u. Board of Ediu:ati-0n, 487 Clay, Henry
Buchan, Nancy, 258 as agent, 446- 48, 450
budgeting heresthetics and, 182, 183n
exploitation, 433 34 Polk and, 222-23
intergovernmental relations and, strategic voting and, 182
408-17 Clean Air Act, 332
zero-based, 132- 33, 152 Clinton. Bill
bureaucracy Perot and, 178-79
agency slippage, 433- 38 presidential vetoes and, 168
bilateral bargaining, 417-20 radio licenses and, 324
budget-maximizing, 408-17 spatial elections, 121
Congressional oversight, 431- 32 unified government and, 513n
principals and agents, 420-30 Clinton, Hillary, 23, 36- 37, 122
testing Niskanen's model, 438-40 closed-rule committee system, 125,
bureaucratic drift, 428. 430, 434- 35 128-34, 152. 393
Burke, Edmund, 450- 51 closed shops, 287n
Bush, George H. W., 122, 178- 79 cloture, defined, 360
Bush, George W. coalitional drift, 436-37
budget deficits, 461 coalitions
divided government and, 512 enacting, 425, 430
Gore and, 122 governm ent formation process,
presidential vetoes, 168-69 507, 512
by-product theory (Olson), 283-86, Coaso, Ronald, 310
291-92 coherence
13yrd, Robert, 331 32, -~76 Arrow's Theorem and. 67- 68
majority rule and, 59- 60, 72
cabinet government trade-off with power, 72- 73
cabinet reshuffle, 511 Cold War, 239-41
defined, 505 collective action
German example, 515 by-product theory, 283- 86, 291-92
overview, 505- 6 conflict of interest, 278- 80
cabinet reshuffle, 511 experiential behavior, 289-92
Calvert, Randall, 120, 462- 68, 471 group basis of politics, 264-68
Index 537
historical perspective, 262-64 divided government and, 512

instrumental behavior, 289- 90 division of labor in, 360
large-scale, 283-86 interest group influence and, 396-
The Logic of Collective Action 97
(Olson), 280- 91 jurisdictional authority, 383-87,
marsh-draining game, 280-81 506
as multiperson cooperation, 268- legislative intent, 75
75 organization of, 509
nonuniqueness of goals, 276- 78 pork-barrel politics, 239-41
political entrepreneurs, 286-89, Powell Amendment, 174-76
291-92 presidential vetoes and, 167-69
problems with, 64n principals and agents, 420-21
public supply and, 315 seniority rule in, 399
theories of voting and, 293- 301 specialization of labor and, 363
collective preference ordering, 201 tax reforms, 66
college admissions, need-blind, 157- consistency condition (Arrow's Theo-
58 rem), 73, 82, 89
Colonial Penn insurance, 303 consumers in rational choice model,
committee systems 18- 19,30
closed-rule, 125, '1.28-34, 152, 393 convergence
defined, 124 electoral competitions and, 116,
jurisdictional authority and, 383- 121-22
87 spatial model and, 144
in legislatures, 381 cooperation
monitoring and control, 389- 95 among legislators, 369-73
multiple referral, 455 collective action and, 268-75
open-rule, 125-26, 134-38, 393 costly enforcement, 250-51
power of, 387, 456 Dictator Game, 256
seniority rule, 399 enforcer incentives, 251-52
standing, 363n, 365 Hobbes on, 233- 34
subcommittees, 398-401 imperfect enforcement, 251
common lands problem, 335-42 inducing via external enforcement,
comparability property, 24, 27 247-52
comparative statics, 127-28 inducing vi.a internalized values,
compliance, legislatures and, 377-79 245- 47
compromise, strategic thinking and, Mafioso Dilemma, 246-50
136- 38 marsh-draining game, 235-39,
Condorcet procedure, 194-95, 198- 241-44, 269, 280
200 multiperson, see multiperson coop-
Condorcet's paradox, 53-61, 67 eration
Condorcet wi nner, 57, 63n paradox of, 238-41
Confederation of British Industry rugged individualism and,
(CBI), 283 232
conflict of interest, 278- 80 Stag Hunt game, 260
Congress, see also House of Repre- tit-for-tat strategy, 244
sentatives; Senate two-person, 235-41
committee systems, 382- 83, 391- cooperation dividend, 241-42
92 coordination
congressional pay raise dilemmas, collective action and, 276- 80
159- 63 courts and, 477-78
constitutional amendments and, leadership and, 397-401
480 legislatures and, 397-401
538 Index
courts division of labor

coordination, 477-78 institutions and, 357-61
dispute resolu tion, 4 75-76 jurisdictions and, 365, 380- 83
intergovernmental :relations and, in parliamentary regimes, 505-6
490-96 Dole, Robert J., 162
overview, 475-81 dominant strategy, 219
rule interpretation, 479- 81 double oral auctions, 326
Cox, Gary, 203-4, 208-9, 224 Downs, Anthony, 111 -23, 154-55
Croson, Rachel, 258 Downs's electoral competition model
Cuisinart Corporation, 250n centripetal tendency, 116- 17
cumulative voting (C\.}, 207 overview, 111-14, 119- 23
Cuomo, Mario, 12 1 three-candidate version, 118-19
CV (cumulative voting), 207 two-candidate version, 114-18
cyclical majorities Doyle, Arthur Conan, 477
Condorcet's paradox, 53- 61, 67 Dred Scott case, 487
"divide the dollars" game, 61- 64 drift
legislatures and, 373-74 agency slippage, 433-37
majority rule and, 109 bureaucratic, 428, 430, 434-35
probability of, 58-59 Droop quota, 211- 12
underlying problems, 373-74 Ducks Unlimited, 267
Duverger's Law
Danforth, John, 75-76 defined, 6, 216
Darman, Richard, 67 two-candidate competition, 216,
Davis, Leonard, 303 224-25, 504
decision-making models
closed-rule committee system, 125, earmarking, 241n
128-34, 152 Easton, David, 11
open-rule committee system, 125- economics
26, 134-38 classes of actors, 18-19, 30
overview, 7-8 comparative statics, 127- 28
pure majority rul e, 125-34 controversy in, 20
universal domain co ndi tion and, production function, 270
80 supply-side, 66
delegating authority transaction costs, 358
in institutions, 366-67 An Economic Theory of Democracy
in legislatures, 383-87 (Downs), 112
democratic peace hypothesis, 6 electoral competition model (Downs),
Denzau, Arthur, 176 111-23
deregulation, 323 electoral systems, 191, 202, 504- 5
Dictator Game, 256 enacting coalitions, 425, 430
discharge petitions, 384 endowments, defined, 506- 7
dispersed tendency, 208 entrepreneurs
dispute resolution, 475-76 collective action and, 286- 89,
Disraeli, Benjamin, 469 291-92
distributive politics, 60, 63 leaders as, 288- 89, 445, 457- 60
district-based systems, 6, 216, 224- Environmental Protection Agency,
25 332-33,426-29
divided government, 5 I 2, 513n Equal Rights Amendment, 137-38
"divide the dollars" ga me Euclidean preferences, 145
cyclical majorities and, 61- 64 even-if form (Arrow's Theorem), 71-
geometry of majority rule, 99-102 74
tax politics and, 64 excludability of goods, 307, 309
Index 539
Expected Utility Principle, 33, 297 Prisoner's Dilemma, 236n, 244,

experiential behavior 246- 47, 253- 59
basis for voting, 299 strategic interdependence, 236
defined,289- 91 Garrison, Greg, 213
evidence regarding, 291- 92 gatekeeping authority
extensive form game, 163n agenda power, 125, 131, 388
externalities (public goods) legislatures and, 384
air pollution, 330-34 PAC influences and, 396
common lands problem, 339- 42 Geertz, Clifford, 4
defined, 325 General Motors, 458
Harvard University experiment, German parliament, 514-27
325- 27 Gibbard, Allan, 172
individual transferable quotas, Gibbard-Satterthwaite Theorem,
339-40 172- 74
regulation and, 3 32 global warming, 322
tax-or-subsidy so.lution, 328- 30 Golden Rule, 244
Gore, Al, 122
Facebook .com, 351 Governing the Commons (Ostrom),
fairness condition (Arrow's Theorem), 338
73, 82, 89 Government Accountability Office,
FCC (Federal Communications Com· 428
mission), 323-24 government formation process
Feddersen, Timothy, 218- 21 analyzing, 513-27
Federal Communications Commis· equilibriu m in, 521-27
sion (FCC), 323- 24 office-seeking, 517-18
Federal Trade Commission, 417n, overview, 506-13
438-40 policy-seeking, 518-21
Fehr, Ernst, 343- 48 government parties, 509
Ferejohn, John, 471, 501 Great Depression
Fiorina, Morris, 145, 148 Roosevelt and, 451
fire-alarm oversight, 431-32 tax reform and, 9, 64- 67
fire-extinguisher oversight, 432 group choice models, see also collec-
first-past-the-post (FPP), 193n, 204- tive action; cooperation; public
7, 216-17 goods; strategic behavior
first-preference majority rule, 45- agenda setting and, 185- 88
46 analysis exercises, 41-49
Fishburn, Peter, 194 Arrow's Theorem on, 68-71
focal point , 272 defined, 8
formateur, 508 electoral systems, 202-22
Forsythe, Robert, 256 majority rule and, 53-89, 145- 48
FPP (first-past-the-post), 193n, 204- manipulation and, 9
7, 216-l7 voting methods, 192- 202
free-rider problem, 344-48 Guarnaschelli, Serene, 219
free speech movement, 284, 458 gun control issue, 183n
Gachter, Simon, 343- 48 Hamilton, Jay, 362n

game theory Hansen, John Mark, 292
dominant strategy, 219 Hardin, Garrett, 335
finite number of r epetitions, Hardin, Russell, 50
243n Hare system (STV), 210- 13
jury verdicts and, 218 Harvard University, 325- 27
political science and, 163n heresthetic, 180- 83
540 Index
highway system, interstate, 316-17 rational choice model, see rational

Hindmoor, Andrew, 15 choice model
Hobbes, Thomas, 233- 34, 250, 252 rugged individualism, 232
Holden, Bob, 21 uncertainty in, 8-9
Holmes, Oliver Wendell, 76 individual transferable quotas
Holmes, Sherlock, 477n (ITQs), 339- 40
Hotelling, Harold, 111-12 informateur, 506
House Internal Security Committee Information and Legislative Organi-
(HISC), 390n zation (Krehbiel), 394
House of Representatives institutions, see also specific types of
agenda setting and, 455 institutions
bureaucratic drift and, 435 closed-rule committee system, 125,
committee systems, 363n, 382-83, 128-34, 152
391-92, 398, 400, 455-56 delegating authority, 366-67,
division of labor in, 358, 360- 61 383- 87
intergovernmental relations, 493- division of labor, 357- 61
94 jurisdictions, 364-66
jurisdictional authority, 383- 87 monitoring, 366- 67, 388-95
organization of, 509 multiperson cooperation and,
Powell Amendment, 174-76 279
principal-agent relationship and, open-rule committee system, 125-
426-28 26, 134-38
structure and procedure, 379 politics and, 356
suspension-of-the-rules mode, pure majority rule, 125-34
360-61 regular procedure and, 357-61
tax reform, 65 scientific knowledge and, 318-20
House Un-American Activities Com- specialization of labor, 361 - 65,
mittee (UUAC), 390n 380-83
Hume, David instrumental behavior, 289-90, 371-
marsh-draining problem, 235-39, 72
241-44, 269, 280, 337 instrumental rationality, 17
on system of Jaws, 3:56- 57 interest group influence, 396- 97
Humphrey, Hubert, 12.:2n, 177-78 intergovernmental relations
agency slippage and, 433-37
ideal point, 92, 114 bilateral bargaining and, 417-20
ideology, see beliefs/belief systems budget-maximizing bureaucrats
IEC (independence of entry of and,408-17
clones), 223 Congressional oversight, 431-32
impossible result (Arrow's Theorem), courts and, 490-96
71-74 defined, 408
incoherence, see coherence motivational considerations, 408-
income tax, 65 10
incumbency effect, 6 Niskanen's model, 410-15, 438-40
independence of entry of clones principals and agents, 420-30
(!EC), 223 variations on Niskanen's model,
independence of irrelevant alterna- 415-17
tives (Arrow's Theorem), 69-70, intersection symbol, 94, 94n
80-81 interstate highway system, 316-17
indifference curves, 101-4 intransitivity, 50, 57
individual choice models investiture, vote of, 507
Arrow's Theorem on, 68-71 investors in rational choice model,
defined, 7 18-19
Index 541
ITQs (individual transferable quo· principal-agent relationship and,

tas), 339-40 448
reputation and, 448- 49, 460-68
Jackson, Andrew, 182- 83 legislation, see also specific acts
Jackson, Jesse, 121 acts of Congress, 420- 21
Jackson, Robert, 498 air pollution, 332
Jefferson, Thomas, 182-83 amendment control rule, 392- 93
Jobs, Steve, 363n judges and, 486-88
Johnson, Lyndon, 376, 461 legislative intent, 74- 76
jointness of supply, 307-9 Powell Amendment, 174-76
judges presidential vetoes and, 167- 69
corruption and, 4 88-89 principal-agent relationship and,
game-player analogy, 485 426-28
as legislators, 486- 88, 498-99 tax reform, 9, 64- 67
motivations of, 485-86 legislative intent, 75
nonprofit analog-.1, 481-84 legislative oversight, 386, 431- 32
Posner on, 473- 74, 481 Legislative Reor ganization Act
spectator analogy, 485 (1946), 386, 431
voting analogy, 484- 85 legislatures
judicial system, see co urts; judges budget-maximizing bureaucrats
Judiciary Reorganization Bill (1937), and, 408- 17
501 cabinet government and, 505
jurisdictions campaign contributions, 387-88
delegating a uthority, 366-67, closed-rule committee system, 125,
383-87 128-34, 152, 393
division of labor and, 365, 380- committee systems, 363n, 365, 381
83 compliance and, 377-79
institutions and, 364-66 Congressional oversight, 432
ministerial portfolios and, 506 cooperation among legislators,
monitoring activi ties, 366-67, 369- 73
388- 95 coordination and, 397-401
jury verdicts, voting rules and, 217- defined, 202n
22 delegating authority, 383-87
justices, Supreme Court, 475n division of labor in, 360-61
electoral arrangements, 504-5
Kansas-Nebraska Act (1854). 183 information and, 376- 77
Keating, Charles H., Jr., 387, 397 interest group influence, 396-97
Kennedy, John F., 3n jurisdictions, 364-66, 380-95
killer amendments, 172-73 leadership and, 397-401, 462
King, Martin Luther, Jr., 458 majority cycles a nd, 373-74
Krehbiel, Keith, 394, 404 matching influence and interest,
374- 76
land ta.x, 65 monitoring and control, 388-95
leadership open-rule committee system, 125-
as agenda setters, 445, 451-57 26, 134-38,393
as agents, 446- 51 pay raise dilemmas, 159- 63
coordination and, 397-401 Powell Amendment, 174-76
as entrepreneurs. 288-89, 445, presidential vetoes and, 167-69
457-60 public services and, 320-21
legislatures and, 397-401, 462 pure majority rule, 125-34
political entrepreneurs and, 288- s patial models of, 123- 41
89 specialization of labor, 380- 83
542 Index
legislatures (cont.) Marksist analysis, 491

structure and proced ure, 379-95 Marsh, David, 283
term limits, 470 marsh-draining game
underlying problcmF, 373-79 collective action and, 280- 81
Leviathan, 234, 244, 250 common lands problem and, 337
L<>vine, Michael E.. 185-88 examples in politics, 239
lighthouse services, 310-11 multiperson cooperation and,
limited vote (LV) method, 206-7 269
The Logic of Collective Action two-person cooperation and, 235-
(Olson), 280- 91 38
Lombardi, Vince, l l:l two-person cooperation with re-
lotteries, defined, 32- H3 peat play, 241- 44
LV Oimited vote) method, 206- 7 mathematical notation, 69n, 94n
maximization paradigm in rational
Madison, James, 356 choice model, 29-30
Madison, Marbury v., 480 :May, Kenneth. 78
Mafioso Dilemma, 246-50 May's Corollary, 77-81
Magnuson, Warren, 181, 183n May's Theorem, 77-81
Mailer, Norman, 263 McCain, John, 122, 318
Major, John, 117, 5t0n McCarty, Nolan, 304
majority cycles, sec cyclical majori- McCaskill, Claire, 21 2:!, 25, 30-31
ties McCubbins, Mathew, 425, 431 - 32
majority rule McKelvey, Richard, 219
Arrow's Theorem, 76-84 McKelvey's Chaos Theorem
coherence and, 72 Arrow's Theorem and, 154
collective clarity and. 85-86 defined, 109-10
com plicated geometry of. 99-110 legislative gridlock and, 404
Condorcet's paradoJC and, 67 majority-rule regime and. 138-41.
cyclical majorities, 53-67 147-48
first-preference, 42. 45-46 McMillan, John, 400n
group choice and, 5:1 89, 145-48 McNollgast
jury verdicts and, 219 20 on bureaucratic drift, 428, 430,
majority cycles and, 109 434-35
pure, 125- 34 on equilibrium, 494n
by round-robin tournament, 42- name origin, 425
43, 46 on procedural controls, 429
simple geometry of, 91 99 testing Niskanen, 439- 40
spatial election app lications, 111- Meirowitz, Adam, 304
23 methodological individualism, 17
spatial models of legislatures, military-industri al complex, 317-18
123-41 Miller, Gary, 418- 22, 430
Malcolm X, 458 Miller, John, 257
Malkevitch, Joseph, J 92n Minas, Sayer, 254
manipulation ministerial portfolios, 506, 515-16
group choice models and, 9 misrepresentation
heresthetics, 180- 83 agenda setting and group choice,
strategic behavior, 170-74 185-88
strategic voting and, 170, 176-79 sophisticated behavior and, 184
of tax policy, 9 MMR (method of majority rule), 76-
Marbury v. lvfadison, 480 81
Markey, Howard T., 498- 99 mode ls, 7, see also group choice mod-
Marks, Brian, 491 els; rational choice model
Index 543
Moe, Terry, 418-22, 430 on bureaucrats, 408-9, 420, 422,

Mondale, Walter, 122 430
monitoring activities critics of, 418-20
Co ngressional oversight, 431-32 Niskanen's model, 410- 15, 438- 40
in institutions, 366-67 variations on Niskanen's model,
in legislatures, 388-95 415- 17
in principal-agent relationships, Niskanen's model
423-24 intergovernmental relations and,
monopolies 410- 15
bilateral, 419 testi ng, 438-40
lighthouse services, 311 variations on, 415- 17
monopoly proposal power, 125 Nixon, Richard, 122n, 177- 78, 460-
monotonicity co ndition (May's Corol- 61
lary), 77- 78, 80 Noll, Roger, 425
moral hazard problem, 367 nondictatorship condition (Arrow's
Moran, Mark, 438- 40 Theorem)
motion of confidence, 508 defined, 70
motion of no confide nce, 508 Gibbard-Satterthwaite Theorem
mot ivation in rational choice model, and, 172
18- 20 group decision-making and, 80
Moynihan, Daniel Pa trick, 161- 62 restricting, 81
Mueller, Dennis, 528 NSF (National Science l<~oundation),
multiparty majority governments, 320
510-11, 513 nuclear disarmament, 239-41
multiperson cooperation
belief systems and , 289-91 Oaks, Dalin H., 498
collective action as, 268-75 Obama, Barack
coordination and, 276-80 cap-and-trade program, 322
monopoly rights and, 3lln Clinton and, 36, 122
multiple referral, 455 national defense and, 318
Murkowski, Frank Ii., 162 O'Connor, Sandra Day, 500
Olson, Mancur
Nader, Ra lph, 458 by-product theory, 283- 86, 289,
national defense, disarmament and, 291
239- 41, 318 The Logic of Collective Action, 268,
National Riffe Association, 183n 280- 91
National Science Foundation (NSF), political entrepreneurs, 286-89
320 privileged groups, 292, 318, 429
NATO, 282-83 selective incentives, 359
need-blind college ad missions, 157- O'Neill, Tum, 212- 13
58 OPEC (Organization of Petroleum
negative externalities, 325, 328 Exporting Countries), 311-14
network effects, 351 open-rule committee system, 125- 26,
neutrality condition (May's Corol- 134-38, 393
lary), 77- 78, 80 opposition parties, 509
New Deal politics, 75 ordering principles
Newton, Huey, 458 defined,25
New Yorker, 446n rational choice model an d, 29- 30
New York Stock Exchange, 351 Ordeshook, Peter, 293, 298- 99, 304
New York 1imes, 212 organizational slack, 416n
Niskanen, William Organization of Petroleum Exporting
budgetary exploitation, 433-34 Countries (OPEC), 311- 14
544 Index
organization theory, 367 political entrepreneurs, 286- 89,

Oscar nominations, 212 13 291-92
Ostrom, Elinor, 338, 340- 41 political manipulation, see manipu-
lation
PACs (political action committees), political parties
396-97 Duverger's Law and, 6, 216
Palfrey, Thomas, 219, 304 plurality voting systems and, 173
paradox of cooperation, 238-41 presidential vetoes and. 169
parliamentary democracy proportional representation. 214-
analyzing government formation, 15
513-27 political science
electoral arrangements, 504-5 game theory and, 163n
forms of, 511 historical perspective, 3-6
governmental arrangements, 505- models and theories, 7- 10
6 politics
government formation process, Bismarck on, 154
506-9 defined, 11
organization of, 503 distributive, 60, 63
partial abstention, 207 group basis of, 264- 68
pay raise dilemmas, 159- 63 institutionalization of, 356
peak associations, 265 public goods and, 309- 14
Pepper, Claude, 287- 88 Politics (Aristotle), 305
Perot, Ross, 122, 178 79 Polk, James K., 222- 23
Pesendorfer, Wolfgang, 218- 21 popular sovereignty, 191
piece-rate method of compensation, pork-barrel politics, 239 41
363 positive externalities, 325, 328
Plott, Charles, 107- 9. 145, 148, 185- Posner, Richard, 473- 74, 481- 86
88 Possum Hollow Rod and Gun Club,
Plott's Theorem, 107- 9, 146- 48 265, 268
"plumping" (election phrase), Powell, Adam Clayton, 175-76
206n-207n Powell Amendment, 174- 76
pluralistic political system, 265-68, power
280 after-the-fact authority, 385-86
plurality runoff, 193 95. 197, 199 Arrow's Theorem on trade-offs,
plurality voting systems 72-73, 88- 89
considerations, 197- 202 gatekeeping authority, 125, 131,
Cox on, 203- 4 384
cumulative voting, 207 jurisdictional authority, 366-67,
defined, 173 383-87
equilibrium in, 207- 10 monopoly proposal power, 125
first-past-the-post, 193n, 204-7, multiple referral, 455
216-17 proposal power, 384
limited vote method, 206-7 preference
proportional representation and, Arrow's Theorem on, 68- 69
203-4,215-17 cyclical majority, 53- 67
single nontransferable vote. 205- Euclidean, 145
6, 209 majority vote, 42- 43
single transferable vote, 210-13 misrepresenting, 184
types of, 193-97 in rational choice model, 14, 20-29
police-patrol oversight, 431 single-peakedness condition and,
political action committees (PACs), 150
396- 97 strategic thinking and, 163-67
Index 545
preference aggregat ion, 73, 77-78, plurality voting systems and, 203-
195 4, 215- l 7
preference diversity, 41-42, 372 proposal power, 384
preference heterogeneity, 371-74, Proposition 174 (CaUfornia), 138
390 Protestant work ethic, 156
preference ordering public goods
Arrow's Theorem on, 69-70 common lands problem, 335- 42
Condorcet's paradox and, 55-56 dcfined,307
defined,25 defining terms, 306-9
society and, 60 excludability, 307, 309
preference revelations, 43 externalities, 325-34
presidential vetoes, 167-69 jointness of supply, 307- 9
PricewaterbouseCoop ers, 213 policy-relevant information and,
prime minister, 506 377
principal-agent relati:onship politics and, 309-14
Congressional oversight, 431-32 property rights and, 323-24
denned, 367n public supply, 314- 23
enacting coalitions. 425 punishing free-riders, 344-48
intergovernmental r elations and, radio spectrum as, 323-24
420- 30 scientific knowledge as, 318-20
leadership and, 448 public supply
parliamentary regimes and, 506 cap-and-trade program, 322-23
Prisoners' Dilemma collective action and, 315
about, 236n deregulation and, 323
altruism and trust in, 253-59 interstate highway system, 316-
common lands problem and, 337 17
Mafioso Dilemma and, 246-47 national defense and, 317- 18
marsh-draining game a nd, 244 politica l intervention and, 314-15
private goods privatization and, 322
defined, 306-7 railio spectrum and, 323- 24
excludability, 307, 309 scientific knowledge, 318-21
jointness of supply, 307-9 time horizons, 321
privatization, 322 Pirate Radio (film), 324
privileged groups pure majority rule, 125-34
agenda power and, 131
Olson on, 292, 318, 429 radio spectrum, 323- 24
privileged motion, 146 rational choice model
problem of order, 234 actors in, 18-19, 30
procedural controls, 429 beliefs in, 16-17, 30- 33
producers in rational choice model, comparability property, 24, 27
18-19 defined, 13
production function, 270 environmental uncertainty and,
Profiles in Courage (Kennedy), Sn 30-33
property rights external environment in, 16
common lands problem, 335-42 maximization paradigm, 29-30
controlling externalities, 332 motivation in, 18-20
public goods and, 323- 24 preference in, 14, 20- 29
proportional representation self-interest in, 15
Cox on, 203- 4 transitivity property, 25- 26, 28,
Duverger's Law and, 6, 216 29n
electoral arrangements, 504 rational foresight, strategic behavior
examples, 214- 15 and, 159-70
546 Index
rationality assumption (Arrow's The- cloture in, 360

orem), 69-72, 76 committee systems, 363n, 383,
Reagan, Ronald 391-92
budget deficits, 461 division of labor in, 360
convergence and, 122 interest group influence and, 397
divided government and, 513n intergovernmental relations, 493-
fiscal policies, 461 94
beresthetical maneuvers and, jurisdictional authority, 383-87
183n Magnuson maneuver. 181
tax reform and, 9, 67 matching influence and interest,
reciprocal cooperation, see two- 376
person cooper ation organization of, 509
reputation and leadership, 448- 49, principal-agent relationship and,
460-68 426- 28
Revenue Acts (1932/1938), 66 structure and procedure, 379
r eversion point, 124n testing Niskanen, 439-40
Riker, William H. tiebreaking rule, 98n
The Art of Political Manipulation, seniori ty rule, 399
173 Sen's Value-Restriction Theorem, 84
heresthetic and, 180-82 sequential runoff, 193, 195, 198- 99
May's Theorem and, 80 set-theor etic symbol, 69ri
Powell Amendment and, 176 shades-of-gray experiment, 28
voting analysis, 293, 298-300 Shepsle, Kenneth , 176
rivalrous goods, 307n simple plurality voting, 193-95, 197
Road and Track, 253n sincere preference revelation, 43
Roberts, John Glover, Jr., 495, 499- sincere voting strategy, 187
500 Singer, Isaac Bashev is, 252
Roe u. Wade, 83 single-elimination tournament, 4 7rt
Roosevelt, .Franklin D., 450-51 si ngle nontransferable vote (SNTV),
Rosenthal, Howard, 304 205-6, 209
round-robin tournaments single-party minority government,
group incoherence and, 67 511- 12
majority rule by, 42-43, 46 single-peakedness condition
rugged individualism, 232 defined,82-83,92- 96
rule interpretation by courts, 479-81 geometric representation to, 90
preferences violating, 150
Safire, William, 207n single transferable vote (STV), 210-
saline-solution experiment, 28 13
Salisbury, Robert H ., 292 s lavery issue, 182-83, 222- 23
Satterthwaite, Mark, 172 Smith, Adam, 328
Savio, Mario, 458 SNTV (single nontransferable vote),
Schelling, Thomas, 272 205-6, 209
Schlozman, Kay, 267 Social Choice and Individual Values
school voucher system, 138 (Arrow), 68
Schwartz, Thoma3, 431-32 social choice theory, 73, 85, 223
scientific knowledge, 318-21 Sophie's Choice (Styron), 27
selective incentives, 359 sophisticated outcome, 167
self-interest in rational choice model, sophisticated preference revelation,
15 43
Sen, Amartya, 84 sophisticated voting
Senate agenda maneuvering and, 170,
bureaucratic drift and, 435 174-76
Index 547
defined, 167n, 176 compromise and, 136-38

jury verdicts and, 218-19 pay raise dilemmas and, 159- 63
misrepresentation ar.d, 184 sophisticated outcome and, 167
Powell Amendment, 174-76 strategic voting
strategic voting comparison, l67n, congressional pay ra ise case, 161-
179 63
Sotomayor, Sonia, 483 defined, 167n
spatial models examples, 173-74
closed-rule committee system, 125, issue manipulation and, 170, 176-
128- 34, 152 79
complicated geometry of majority sophisticated voting comparison,
ruJe, 99- 110 167n, 179
electoral competition, 111-23 STV (single transferable vote), 210-
oflegislatures, 123-41 13
open-rule committee system, 125- Styron, William, 27
26, 134-38 subcommittees, 398- 401
opportunities afforded with, 142- subsidies, enernalities and, 328-30
44 sunset provisions, 124n, 132-33
pure majority ruJe, 125-34 supply-side economics, 66
simple geometry of majority rule, Supreme Court
91- 99 compromise decisions, 83
specialization of labor intergovernmental relations and,
institutions and, 361-65 491
legislatures and, 380-83 justices and, 475n
in parliamentary regimes, 505-6 as legislators, 498
Stag Hunt game, 260 rule interpretation, 478-80
standing committee system, 363n, specialization of labor and, 362- 63
365 suspension-of-the-rules, 360- 61
Stavins, Robert, 322
"stopping rules," 125n take-it-or-leave-it proposals, 125
strategic behavior Talent, J im, 21- 22, 31
agenda setting and group choice, truces
185-88 Civil War and, 64- 67
defined, 174 externalities and, 328-30
Gibbard-Satterthwaite Theorem, manipulation of, 9
172-74 Tax Reform Act (1986), 66
heresthetic, 180-83 term limits, 4 70
legislatures and pay raise dilem- testing Niskanen's model, 438- 40
mas, 159-63 Thatcher, Margaret, 117
~pulation, 170-74, 180-83 theory, 9
misrepresentation, 184-88 thick description, 4
need-blind college admissions, third-party enforcement, 248, 252
157-58 tie-in sales, 311
Powell Amendment, 174-76 Tierney, John, 267
presidential vetoes, 167- 69 time horizons, 321
rational foresight, 159-70 tit-for-tat strategy, 244
strategic voting and, 176-79 tobacco industry, 183n
rategic interdependence, 236 Tocqueville, Alexis de, 266
~ategic manipulation, see manipu· transaction costs, 358
lation transitivity property in rational
.ategic preference revelation, 43 choice model, 25-26, 28, 29n
ategic thinking Truman, Harry S, 446
548 Index
trust, Prisoners' DHemma and, 253- not wasting votes, 177-79

59 popular sovereignty, 191
Twenty-second amendment, 470 voting methods
two-candidate competition collective action and, 293-301
Downs's electoral competition collective preference ordering,
model, 114-18 201
Duverger's Law, 216, 224-25, 504 considerations, 197-202
two-person cooperation defined, 191
overview, 235-41 jury verdicts and, 217-22
with repeat p1ay, 241-45 rational theories of voting, 300
types of, 192-97
unanimity rule Voting Rights Act (1965), 399, 461
Arrow's Theorem on, 69, 80-81
defined, 45 Wade, Roe u., 83
jury verdicts and, 218, 220-21 wage system. 363
uncertainty. 31. 33n Wagner, Robert (book reviewer),
umfied g011ernments, 510, 513n 286-87,289,291-92
unions, organizing, 287n Wagner, Robert (Senator), 287-88
United Nations Food and Agricul- Wagner Act (1935), 287
tural Association, 339 Walker, Jack L., Jr.. 292
universal domain condition (Arrow's Wallace, George, 122, 177-78
Theorem) Warren, Earl, 494
defined, 69 Warsaw Pact, 282-83
Gibbard-Satterthwaite Theorem Washington, George, 335
and, 172 Washington Post, 162
group decision-making and, 80 weak preference relation, 24n-25n
restricting, 82, 84, 109 wealth tax, 65
U.S. Postal Service, 315n Weingast, Barry. 425, 438-40, 501
utile, defined, 236n willingness-to-pay curve (Niskanen's
utility function, 92 model), 410-11, 417, 422
utility number. 32 workers in rational choice model,
wor1~ w1:r
8
value restriction condition (Sen's n, 450-51 /
Theorem), 84 Wozniak, Steve, 363n
Verba, Sidney, 150
vetoes, presidential, 167-69 X, Malcolm, 458
vote of investiture, 507
voters, see also sophisticated voting; Yes, Minister (tv series), 4 l8n
strategic voting YouTube.com, 348
collective action and, 293-301
ideal point for, 114 zero-based budgeting, 132-33, 152
PRAISE FOR
Analyzing Politics
"Analyzing Politics is the single best source for
understanding how powerfully rational choice
theory has influenced the study of politics. It does
not dumb down the important core concepts, yet
it remains entertaining, vivid, and accessible. My
students not only read it-they love it!"
-James Fowler, University of California, San Diego
"This is a superb introduction to the analysis of

political action in rational terms, uniting economics
and political science. It is one of the rare books that
double as a student-friendly classroom guide and a
treatise from which experts can learn."
- Judge Richard A. Posner
"Undergraduate education in political science is too

often removed from the research frontier-failing
to inspire students with the thrill of scientific
discovery. Shepsle's Analyzing Politics bucks this
trend. The experiments in this new edition were
developed by Shepsle over more than a decade of
innovative classroom teaching and are a terrific
addition."
- Ethan Bueno de Mesquita, University of Chicago
"This book is a lucid introduction to the rational-

choice approach to the study of politics. Shepsle
presents a wealth of conceptually sophisticated
material in clear and simple terms."
-Torsten Persson, Stockholm University
"Analyzing Politics has become the introduction to

rational choice approaches to politics. Clear and
lively, this book provides a fresh and comprehensive
approach to the field."
-Barry Weingast, Stanford University
flf°f )

SHEPSLE, Keneth - Analyzing Politics - Rationality, Behavior and Institutions PDF

Uploaded by

Copyright:

Available Formats

SHEPSLE, Keneth - Analyzing Politics - Rationality, Behavior and Institutions PDF

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

SHEPSLE, Keneth - Analyzing Politics - Rationality, Behavior and Institutions PDF

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What are some of the main concepts discussed in the book?

What are some of the main concepts discussed in the book?

Additional Praise for Analyzing Politics

(continued from back cover)

"I regularly use this book in my classes in development and

"With its clear presentation of concepts and many examples,

"Wonderfully written with powerful and well-chosen exam-

"This concise but thorough introduction to rational choice is

Series Editor, Kenneth A. Shepsle

Analyzing Politics is the cornerstone of The New Institutional-

Analyzing Congress by Charles Stewart III

Copyright IC> 2010, 1997 by W. W. Norton & Company, Inc.

All rights reserved

Manufacturing by Courier Companies, inc.

Library of Congress Cataloging-in-Publication Data

1 It Isn't Rocket Science, but ... 3

3 Getting Started with Group Choice Analysis 41

11 Institutions: General Remarks 355

The first edition of Analyzing Politics appeared more than a

tried to acknowledge them in footnotes at the relevant places.

This book pretends to be modest, but it really isn't. Although

Consider this illustration. We are all familiar with reproaches

Rarely, however, are they more than exercises in storytelling

census of some particular population is just a bunch of numbers until it is

contempt of the majority for opinions of the minority party).

public-affairs commentators, and politics junkies. There is

M ODELS AND THEORIES

uation. We thus expand upon our ele mentary model of individ-

risky career choices of politicians. On other occasions, I play a

based on the conduct of intergovernmental relations between

6 An answer key is provided to instructors at wwnorton.com/nrl.

to real-world situations. Your job is to engage the material,

7 David Easton, The Political System (New York: Knopf, 1953).

viduals interacting, maneuvering, dissembling, strategizing,

I n a nalyzing politics I shall take what b as come to be known

"behavioral economics." For a fine review of this revisionst interpretation,

Individual wants, which I refer to as preferences, can be in-

order to proceed. For us, preferences are one of the givens of a

3 Methodologically, this is very similar to the approach of economists, who

But preferences, tastes, and values are not all there is to

nect instruments to outcomes. Acting in accord both with one's

tional relations aggregate all the way up to the nation-state as

To motivate a rational model of political behavior, let's begin

an eye on the overall long-term financial return. This is laid

behavior with a simple set of assumptions? The modern the-

THE SIMPLE LOGIC OF

Our first building block is the notion of preference. We must

variety of contexts, our building blocks must be developed in

Claire McCaskill graduated from law school in 19.78 and

mained on the ballot and he actually won the election! A

We can be reasonably confident that McCaskill preferred x

McCaskill ended up running for (and winning, as it hap-

In the remainder of this section I abstract from the specific

ences." Similarly, "xliy" means Mr. i is indifferent between

like x and y, either x R,y or y R, x or both. In words, a person has complete

Property 2: Transitivity. The strict preference relation is

bility). And the comparison they permit are internally consis-