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 the 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
232 Analyzing Politics

A WORLD WITH NO COOPERATION:

WHAT WOULD IT BE LIKE?

In nineteenth- and early twentieth-century America, mu.ch

was made of the virtue of "rugged individualism." People Wete
thought to be virtuous if they were self-reliant-if they deve1_
oped the requisite coping skills and other forms of "hum<:ln
capital" to enable them to survive and prosper in a world fl.l.ll
of opportunities, to be sure, but full of pitfalls and hazards a.s
well. This ideology assumed mythic proportions, personified in
the famous Horatio Alger stories about a young lad who su.c.
ceeded in a cruel world by dint of individual effort alld
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 alld
self-reliant, but families. Responsibility for, and cooperation
within, families (sometimes of a rather extended sort) we:i-e
values held dear in an earlier era, putting to shame the puj-,
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
t
Cooperation 233

l 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, but 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 through 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-

! •
234 Analyzing Politics

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:

Two-PERSON COOPERATION

Individual behavior typically involves bearing some costs in

order to secure 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 Part 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
 236 Analyzing Politics

to as strategic interdependence. We can analyze the situa tiori

ystematically 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 1. 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?

1
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.
3
Some readers may recognize this payoff matrix. In game theory it is known
as the Prisoners' Dilemma. Two petty criminals are arrested for a burglary.
If 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 O utiles). But if one squeals and the other keeps quiet,
then 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

DISPLAY 8.1
HUME'S MARSH-DRAINING GAME*

Farmer B's Choice

Drain marsh Do not drain marsh
(Cooperate) (Do not cooperate)
Farmer A's Choice
Drain marsh 1, 1 -1, 2
(Cooperate)
Do not drain marsh 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 O 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,
238 Analyzing Politics

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 flourish so each
farmer gets a O 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.

"See David Hme, A Treatise of Human Nature, ed. L.A. Selby-Bigge and P. H.
Nidditch (Oxford: Clarendon Press, 1975 [17371). I thank Dr. Mark Yellin for
bringing this to my attention.
'
Cooperation 239

CASE 8.1
THE PARADOX OF COOPERATION:
NUCLEAR 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 the 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 countrie in-
curred substantial economic and psychological costs main-
taining these arsenals. Since neither country was able to
generate a clear advantage in its nuclear threat, and since
use of these weapons was suicidal, 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 0; 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.
240 Analyzing Politics

SOVIET UNION
Maintain Disarm
UNITED 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 cooperation. The term "pork-
barrel politics" refers 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-
billion 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 has not been all or nothing. It has pro-
ceeded, in a sense, in baby steps: ''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. Thu ,
everyone knows that despite everyone being better off in a
pork-free world, everyone will continue to push for projects
for his or her own 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, because 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 earmarhing,
see Kenneth A. Shepsle, 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-PERSON COOPERATION WITH
REPEAT PLAY

In Hume's example, 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 relationships appropriately to cap-
242 Analyzing Politics

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 noncooperative 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,
seems 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 end. 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
5
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
all 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.
244 Analyzing Pou».
vzcs

elicited cooperation from the other, thereby encouraging e~ch

to try it again. In short, a little positive reinforcement ~a ,
well set them on a "cooperation path" for quite some time. J:.._"1s
_)
the "shadow of the future"-the prospect of cooperation d:i_vi-
dends not just now but stretching out over the longer ha~-
that make cooperative moves look very attractive.
In his famous work on this subject, Robert Axelrod6 c<:\lls
the behavioral strategy of "being nice" the first time and th.en
on each succeeding occasion, doing what the other guy did th~
time before the tit-for-tat strategy. The first time, cooper~te.
The next time, cooperate if your colleague cooperated the last
time. But don't cooperate if he didn't the last time, and dl:)n'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 e<'l.ch
farmer playing his tit-for-tat strategy in the repeat plaJ, 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 l:'eli-
gious principle (like the Golden Rule), although their behaVior
seems to exhibit it. Nor is there a sword-wielding Leviatl\an
making the two cooperate. Instead, each of two ruggedly ill.di-
vidualistic, rational egoists has, by virtue of being embedrjed
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 l'l.ote
that there is an "evil twin" to the norm of reciprocal coop~ra-
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 far rner
will "punish" the other for failing to cooperate the time before.

6 Robert Alexrod, The Euolution 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 MECHANISMS
INDUCING 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

7 This situation is nicely analyzed in Avinash Dixit and Barry Nalebuff, Thinh-
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.
246 Analyzing Politics

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 reread 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.
I 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 O 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, -13
(Cooperate)
Squeal -13, -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 B, which
stands for "bad," is a very large number'

squeal), since this would change his payoff from O 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 -13 utiles.
But the game theoretics, in my view, do not give sufficient
psychological weight to that big negative payoff of -j3. I sus-
pect that neither of the soldiers would take a chance on being
the only one to squeal (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
 248 Analyzing Politics

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. In economic contexts it is normally assumed 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).
Imagine 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, I 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 fly-by-night opera-

9
In a sense, though, there may not be a clear distinction between these two
mechanisms all the time. In the presence of external enforcement, individ-
uals will often take this fact on board 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. Coercing 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.'?
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
third-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 her guarantee, the buyer will require a
break on the price. This difference between the price a buyer would pay
under a fully enforceable guarantee and one that is not enforceable is, in
effect, the insurance premium the buyer requires.
 250 Analyzing Politics

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 them 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,
imperfect enforcement, and the incentives of the enforcer.

COSTLY ENFORCEMENT. When your food processor breaks down,

and the merchant from whom you purchased it tells you that
it is not his responsibility but rather that of the manufacturer,
how much bother are you prepared to undertake to redress
your grievance? You might assume a low-cost burden, like
writing a letter of complaint to the manufacturer. But you
most likely would not hire a lawyer to take the manufacturer
to small claims court. The fact of the matter is that enforce-
ment is not costless, and "small" departures from cooperative
agreements (so that the guarantor is forced to honor her guar-
antee only if the guaranteed party bears a substantial cost)
are likely to go unpunished, at least most of the time. Some-
times institutions arise to handle these problems. An enter-
prising lawyer might enter a "class action" suit on behalf of all
those victimized. Each victim will have suffered a small cost,
perhaps, but if there is a large enough number of victims, the
total amount at stake could be rather large-large enough to
interest a lawyer whose take is a proportion of the settle-
mentl-! Nevertheless, before invoking Hobbes's Leviathan ~s a
11
In fact, precisely this happened in the 1980s in regard to a food processor
manufactured by the Cuisinart Corporation. A class action suit was lodged
against the company and, in a settlement, the company compensated every
purchaser of its food processor. The author was a beneficiary! My recollec-
tion is that each beneficiary was sent a catalog of Cuisinart products from
which a choice might be made. My heavy frying pan is still in use nearly
three decades later.
Cooperation 251

solution to the problem of capturing cooperation dividends,

one ought to take into account the probable costs of imple-
menting enforcement. They may not be worth the candle.

IMPERFECT ENFORCEMENT. Even if the costs are small enough to

permit enforcement to operate, it may not do so perfectly. It
may not be that much bother, for example, to take a local mer-
chant or landlord to small claims court for cheating you. But
there is no guarantee that the judge will come to the "correct"
judgment. Enforcement is always bound to be imperfect be-
cause human judgments are fallible on the one hand, and be-
cause the landlord may be the judge's brother-in-law on the
other. The point is that, despite what first-year law students
are told about the majesty of the law, litigation is no less sub-
ject to human error and corruption than anything else that
human beings undertake. Judges, police officers, parole offi-
cers, official investigators, and other "players" in the justice
system are often called upon to make judgment under less-
than-ideal circumstances. Mistakes will be made.

ENFORCER lNCENTIVES. As long as the normal execution of en-

forcement is not too expensive, and as long as mistakes are
neither too frequent nor too egregious, then these possibilities
will not much diminish the value of third-party enforcement.
Put more constructively, in many circumstances neither costli-
ness nor imperfections will be so out of line as to undercut the
value of outside enforcement. The incentives-of-enforcers prob-
lem, however, is not so easy to duck. Mind you, most econo-
mists do precisely this: They assume, by fiat, that enforcement
involves an honest discovery procedure in which fault is deter-
mined and compensation ordered. This, in turn, provides in-
centives for people not to defect from cooperative agreements
in the first place. But it does raise the issue of what institu-
tional features provide the incentives for enforcement agents
to do this. Why shouldn't the town traffic officer, for example,
252 Analyzing Politics

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-
lem for third-party enforcers is resolved, one should not em-
brace this solution.

12
The incentives-of-enforcers problem is demonstrated vividly by the case of
New York City's parking meters. According to Road and Track (June 1994.
p. 15), half of all of the city's parking-meter change collectors were charged
with stealing over $1 million from their collections.
Cooperation 253

PRELIMINARY 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 third-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 the topics of n-person cooperation and collec-
tive action.

EXPERIMENTAL CORNER

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
254 Analyzing Politics

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." 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 cooperating, 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
256 Analyzing Politics

that player is a heartless competitor or that she simply felt

insecure about her partner's trustworthiness? We could
never answer such 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 have 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
alll). Forsythe et al." 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 et al., "Fairness in Simple Bargaining Experiments,"

Games and Economic Behavior 6 (1994): 347-69.
Cooperation 257

(giving nothing) and generous types (giving lots) exist

across societies."
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 whatever money he gives
to the other participant will be multiplied by a factor p. For
Andreoni and 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, and 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 Henrich et al., in a superb 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 that only ·
among their American college student participants was allocating O per-
cent the most common outcome; among the Orma, a tribe in Kenya, the
most common offer was 50 percent. See Joseph Henrich et 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 GARP: An Exper-
imental Test of Consistency of Preferences for Altruism," Econometrica
70 (2002): 737-53.
258 Analyzing Politics

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 and his partner. Another 11 percent seemed to
be maximizing overall social welfare, and the final 24 per-
cent 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 that altruism is present,
but not omnipresent, in social interactions.
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 (like 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, they also 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 help us under-

stand the Prisoners' 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 reach 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 laboratory 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!).

PROBLEMS AND DISCUSSION QUESTIONS

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
emissions.
260 Analyzing Politics

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:"

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 3, 1 0, 0
Movies I 0,0 1, 3

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, and religion. Discuss how each of
these social institutions has evolved features to facilitate coop-
eration, with reference to external enforcement, internal en-
forcement, and repeated interaction.

I
I,