American Political Science Association The American Political Science Review

Absolute and Relative Gains in International Relations Theory
Author(s): Robert Powell

Source: The American Political Science Review, Vol. 85, No. 4 (Dec., 1991), pp. 1303-1320
Published by: American Political Science Association
Stable URL: http://www.jstor.org/stable/1963947
Accessed: 26-02-2017 18:53 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Political Science Association is collaborating with JSTOR to digitize, preserve and extend
access to The American Political Science Review
This content downloaded from 195.49.216.244 on Sun, 26 Feb 2017 18:53:46 UTC
All use subject to http://about.jstor.org/terms
ABSOLUTE AND RELATIVE
GAINS IN INTERNATIONAL
RELATIONS THEORY
ROBERT POWELL
University of California
Berkeley
he problem of absolute and relative gains divides neoliberal institutio

and structural realism. The former assumes states focus primarily on their absol
and emphasizes the prospects for cooperation. The latter supposes states are large
cerned with relative gains and emphasizes the prospects for conflict. Existing work in
international relations theory generally traces the differences between these two theories
to different assumptions about states' preferences. Using a simple game-theoretic model,
this essay offers a reformulation of the problem of absolute and relative gains that links
changes in the states' behavior, the feasibility of cooperation, and especially the states'
concern for relative versus absolute gains explicitly to changes in the constraints facing
the states. Many of the differences between neoliberal institutionalism and structural
realism appear as special cases of the model.
lrhe problem of absolute and rela- about states' preferences lead to different
tive gains divides two of the most influ- expectations about the prospects for inter-
ential approaches to international rela- national conflict and cooperation. The
tions theory. Neoliberal institutionalism more states care about relative gains, the
assumes that states focus primarily on more a gain for one state will tend to be
their individual absolute gains and are seen as a loss by another and the more dif-
indifferent to the gains of others. Whetherficult, it seems, cooperation will be.2
cooperation results in a relative gain or However, tracing different expectations
loss is not very important to a state in about the likelihood of conflict and coop-
neoliberal institutionalism so long as it eration to different assumptions about
brings an absolute gain. In terms of pref- states' preferences poses an important
erences, this focus on absolute gains is theoretical difficulty for international
usually taken to mean that a state's utility relations theory.3 Jervis (1988), Keohane
is solely a function of its absolute gain. In (1984), Lipson (1984), Stein (1983, 1984),
contrast, neorealism, or structural realism and the contributors to the volume edited
assumes that states are largely concernedby Oye (1986) have recently called for
with relative rather than absolute gains. trying to bring at least some aspects of the
In the anarchy of international politics, study of international political economy,
"relative gain is more important than which is usually taken to be the province
absolute gain" (Waltz 1959, 198). A state's of neoliberal institutionalism, and the
utility in structural realism is at least part- study of security affairs, which is usually
ly a function of some relative measure like taken to be the province of structural
power. 1 These differing assumptions realism, within a single analytic frame-
AMERICAN POLITICAL SCIENCE REVIEW

VOLUME 85 NO. 4 DECEMBER 1991
American Political Science Review Vol. 85
work. But if neoliberal institutionalism the type or nature of states' preferences,

and structural realism really do make fun- would be impossible.
damentally different assumptions about I offer a reformulation of the problem
states' preferences, then efforts to unify of absolute and relative gains and take a
these approaches with a third-image ex- step toward bringing the study of interna-
planation cannot succeed. tional conflict and cooperation within a
To use Waltz's analogy (1979, 89-91) single analytic framework.4 I show that
between political structures and economic many important aspects of neoliberal
market structures, it will be impossible toinstitutionalism and structural realism
explain the differences between neoliberalmay be seen as special cases of a very sim-
institutionalism and structural realism ple model of the international system in
over, say, the prospects for international which changes in the states' behavior, the
cooperation in the same way that econo- feasibility of cooperation, and especially
mists explain the differences between out- the states' concern for relative versus
comes in a perfectly competitive market absolute gains are explicitly linked not to
and a monopoly. In that explanation, different assumptions about the states'
economists assume that the goals or preferences but to changes in the con-
preferences of the unit, in this case the straints facing the states. The model thus
firm, are the same in both a perfectly shifts the focus of analysis away from
competitive market and in a monopoly: a preferences to constraints.5
firm seeks to maximize its profits. What In the model, states are assumed to be
varies in moving from one market struc- trying to maximize their absolute gains.
ture to the other are the constraints under That is, a state's utility depends solely on
which a firm attempts to maximize its the absolute level of economic welfare it
profits. Thus, changes in the market con- attains. This is in keeping with neoliberal
straints and not in the units' preferences institutionalism. But the states are trying
account for the variation in a firm's to maximize their economic welfare
behavior in the two different market within the constraints imposed by an
structures. If, however, neoliberal institu- anarchic international system in which
tionalism and structural realism are the use of force, in keeping with structural
actually based on fundamentally different realism, may be at issue. When the cost of
assumptions about states' preferences, using force is sufficiently low that the use
then moving from a neoliberal institu- of force actually is at issue, cooperative
tional to a structural realist setting wouldoutcomes that offer unequal absolute
see a change in the attributes of states, gains cannot be supported as part of an
that is, their preferences. In this sense, equilibrium even though the states' pref-
neoliberal institutionalism and structural erences are defined only over their abso-
realism would be about essentially dif- lute level of economic welfare. This in-
ferent types of units, one primarily con- ability to cooperate is in accord with the
cerned with absolute gains, the other with expectations of strucural realism, though
relative gains. Consequently, the locus of the assumption that states are maximizing
the differences between neoliberal institu- their absolute gains is not in keeping with
tional and structural realist explanations its usual formulations. If the use of force
of international behavior would be in the is not at issue because fighting is too cost-
first or second images. A third-image uni- ly, then the results are more in accord
fication of these approaches, which would with neoliberal institutionalism. For
explain changes in states' behavior in example, cooperative outcomes that
terms of changes in the constraints facing could not be sustained when the use of
the states and not in terms of changes inforce was at issue now become feasible.
1304
International Relations Theory
Thus, many aspects of neoliberal institu- anarchy and cooperation. Structural real-
tionalism and structural realism appear as ism generally associates anarchy with a
special cases of the model. concern for relative gains and a lack of
The model developed below is extreme- cooperation (Waltz 1979, 105). Neoliberal
ly simple. It aims primarily at conceptual institutionalism, however, argues that
clarification, not empirical application. anarchy in the sense of a "lack of a com-
To this end, it has been deliberately mon government" (Axelrod and Keohane
designed to show, in the simplest possible1986, 226; Keohane 1984, 7) does not
well-defined game-theoretic model, thatimply a lack of cooperation (Keohane
many of the differences between neo- 1984, 65-84). The present analysis helps
liberal institutionalism and structural to elucidate the sources of this difference.
realism can be traced explicitly to the con- The ability or inability to enforce rules of
straints facing states, rather than to their behavior is relevant only if the physical
preferences. The extreme simplicity of the environment defined by the system's con-
model clarifies the relation between constraints is such that one of the possible
straints and the problem of absolute and behaviors is to use one's relative gain to
relative gains that although present in one's advantage and to the disadvantage
more complicated models, would be more of others. If there are no such opportuni-
difficult to discern there. The develop- ties, then the inability to commit oneself
ment of more complicated and less con- to a promise not to use a relative gain to
trived models awaits future work.6 But one's advantage is moot. Thus, the con-
despite the model's simplicity, its analysiscern for relative gains arises from both
makes three new points that are relevantanarchy and the constraints that define
to international relations theory. Most the range of possible behavior. Although
narrowly, it suggests that cooperation insuch a system is clearly implicit in struc-
some circumstances may be even more tural realism, neither neoliberal institu-
difficult to achieve than has been pre- tionalism nor structural realism fully
viously appreciated. Some agreements appreciates the significance of the sys-
that offer equal absolute gains-and tem's constraints in the origins of the con-
therefore no relative gain-cannot be sus- cern for relative gains.
tained in equilibrium. The reason is thatOne consequence of the failure of both
cheating on the agreement would bring structural realism and neoliberal institu-
large relative gains. tionalism to appreciate the role of the
The second and third points address system's constraints in the problem of
two other broader issues in international absolute and relative gains is the use
relations theory. The model offers a sim- of repeated games in both of these ap-
ple formal example showing that Waltz's proaches to model the international sys-
notion of political structure is unable tem to implicit in structural realism (e.g.,
account for important changes in the feas- Jervis 1978, 171; Keohane 1984; Oye
ibility of international cooperation. The1986). As will be shown, relative gains in
political structure as Waltz defines it repeated games, including the repeated
(1979, 79-101) remains constant through- prisoner's dilemma, cannot be used to
out the analysis; but the feasibility of one's advantage. Thus, even if states are
cooperation varies. Thus, variations in unable to make binding agreements in
what Waltz takes to be the structure of the these models, relative gains in repeated
political system cannot explain the varia- games do not matter. This renders any
tion in the feasibility of cooperation in the analysis of structural realism's under-
model. standing of the relation between anarchy,
The second issue is the relation between relative gains, and cooperation based on
1305
these models problematic. to punish uncooperative behavior today

Finally, a more complete understanding with future sanctions. If the shadow of the
of the relation between the system's con- future looms sufficiently large, then the
straints and the origins of the concern for future costs to uncooperative behavior
relative gains also shows there is nothing will outweigh the immediate gains; and,
theoretically special about the possible weighing costs against benefits, even ego-
use of force. If the nature of military tech-istic states will cooperate. This logic is, in
nology is such that one state can turn a turn, formalized in the neoliberal institu-
relative gain to its advantage and the dis- tional analysis with a repeated prisoner's
advantage of others, then these con- dilemma in which mutual cooperation can
straints will induce a concern for relativebe sustained as an equilibrium outcome
gains and this may impede cooperation with a strategy of punishing defection
absent any superior authority to ensure should it occur.8
that these gains not be used in this way. First, I will present a simple neoliberal
But if, for example, the nature of an institutional model of the problem of con-
oligopolistic market is such that a firm flict and cooperation. The model has been
can use a relative gain in market share to explicitly designed to capture the essence
increase its long-run profits at the expense of the neoliberal institutional analysis in
of other firms, then this system will also the simplest possible formal setting and to
induce a concern for relative gains that be as similar to a repeated prisoner's
may make cooperation difficult. The con- dilemma as possible. This similarity helps
cern for relative gains may characterize to clarify the relation between the formal-
many domains, and a more refined under- ization used here and that employed in
standing of the origins of this concern other work on neoliberal institutionalism.
helps to identify them.7 Second, I will demonstrate that the model
I will examine a very simple neoliberal actually does capture the essence of the
institutional model of the problem of neoliberal analysis problem of coop-
cooperation. The examination shows that eration. Finally, repeated games are
the repeated games offer poor models of shown to be poor models of the problem
the problem of conflict and cooperation in of absolute and relative gains and
structural realism. I will then modify the cooperation in structural realism.
model so that the constraints of the modi- The simple neoliberal institutional
fied model create opportunities for a state model is a three-by-three game that is
to turn relative gains to its advantage, played twice. The states, in keeping with
which in turn, induces a concern for rela- neoliberal institutionalism, will try to
tive gains. I then use this model to study maximize the absolute sum of their first-
how changes in these constraints affect and second-period payoffs. The second
the feasibility of cooperation. play of the game casts a shadow of the
future onto the first play of the game. But
because the game is only played twice,
A Simple Neoliberal there is no shadow of the future to affect
Institutional Model the prospects of cooperation in the second
period. Thus, the problem of cooperation
The essence of the neoliberal institu- reduces to seeing if the shadow of the
tional analysis of the problem of coopera- future makes cooperation possible in the
tion is that the shadow of the future may first period. In this way, the two-period
lead the egoistic states hypothesized in repeated game provides a formal setting
structural realism to cooperate. Repeated for studying the problem of conflict and
interaction gives each actor the ability cooperation that is simpler than the in-
1306
finitely repeated prisoner's dilemma.9 Figure 1. A Neoliberal

The three-by-three game, which will be Institutional Model
played twice, will be called 'and is de-
rived from the prisoner's dilemma by add-
ing a third strategy to the two-by-two
S2
prisoner's dilemma. There are two states, F T C
Si and S2, in this game; and each has three
strategies: F, T, and C. Figure 1 shows the
payoffs to the various possible outcomes. F 3,3 1,4 -1,0
The four cells at the upper left form a sim-
ple prisoner's dilemma in which each state
strictly prefers playing T to F regardless of
whether the other state is playing T or F. S1 T 4,1 2,2 -1/2,0
But if both states play T, they are worse
off than if both had played F. The out-
come (T, T) is Pareto-inferior to (F, F). If
one state plays C, then that state's payoffs
are zero regardless of what the other state C 0,-1 0,-1/2 0,0
does-while the other state obtains 0,
-1/2, or -1 if it plays C, T, or F,
respectively.
It is important to emphasize that the
game composed of two plays of 6is an
extreme theoretical simplification. 6has consequently, the option of closing the
been explicitly designed so that it captures economy is equivalent to imposing an
the essence of an infinitely repeated embargo.10 The game 4' then, represents
prisoner's dilemma and the neoliberal the case in which a state gains if it is the
institutional analysis of the problem of only state to impose a tariff but both
cooperation in the simpler formal setting states lose if they both impose tariffs; that
of a two-period repeated game. At this is, the outcome in which both states
level of abstraction, serves as a substan- impose tariffs (T, T) leaves both states
tive model of the international system or worse off than the free trade outcome,
of the international economy only in the (F, F). If both states close their economies
very general ways that the repeated and effectively embargo each other by
prisoner's dilemma does in neoliberal playing C, there will be no trade. The
institutional or structural realist analysespayoffs in this case are less than the pay-
of the problem of cooperation. Neverthe- offs obtained if both states had imposed
less, 6' may be given a more concrete an optimal tariff.1 If only one of the
interpretation that is roughly based on an states closes its economy, there will still
optimal tariff model, and describing this be no trade regardless of what the other
interpretation may help to make the state does; and the state that closed its
analysis clearer. Building on Johnson's economy receives the same payoff regard-
(1953) seminal work on optimal tariffs, less of what the other state does. But the
the strategies F, T, C may denote free and state that had expected to trade, either
open trading policies, imposition of the freely or with tariffs, suffers a loss due
optimal tariff, and closure of the economy to a misallocation of its productive
by trade prohibition, respectively. By resources. Accordingly, each state prefers
assumption, there are only two states; to close its economy if the other does. For
1307
the sake of clarification, the neoliberal in- T in the second period if (F, F) is the first-
stitutional model composed of playing 4' period outcome and C if not and, similar-
may be loosely interpreted as a very rudi- ly, for S2 to play Fin the first period and T
mentary model of strategic trade. 12 in the second period if (F, F) is the first-
The formal aspects of the neoliberal period outcome and C if not. In essence,
institutional critique of structural real- the state that remained faithful to free
ism's analysis of the problem of coopera- trade during the first period by playing F
tion are based on a repeated prisoner's is punishing the defector in the second
dilemma in which the shadow of the period by imposing an embargo. Clearly,
future is large enough to induce coopera- (F, F) is the first-period outcome if both
tion. In order to relate the present formal- states follow these strategies. These strate-
ization to that employed in the neoliberal gies also constitute an equilibrium if
institutional critique, the payoffs of the neither state can improve its payoff by
model developed here have been con- deviating from its strategy, given that the
structed so that the shadow of the future other player is following its strategy. If
does induce cooperation in the model. both states follow these strategies, each
More formally, the neoliberal institu- receives three in the first period and two
tional model has been explicitly designed in the second for a total payoff of five. If,
so that (F, F) is a first-period equilibrium however, a state deviates in the first
outcome. 13 period by playing T, it will do better in
To show that (F, F) can be a first-period that period by realizing a payoff of four.
equilibrium outcome, the equilibria of the But given the other state's strategy of
neoliberal institutional model will be imposing an embargo in the second period
determined. In equilibrium, the outcome by playing C if (F, F) was not the first-
on the last play of a repeated game mustperiod outcome, the most that a defector
be an equilibrium of the stage game that can attain is a second-period payoff of
is being repeated.14 In particular, the zero. This yields a total payoff of four;
second-period outcome of the game com- that is, the future cost of defection, which
posed of two plays of the stage game 4'is to obtain zero rather than two in the
must be an equilibrium of 4' 4g in turn, second period, outweighs the immediate
has two pure-strategy equilibria.15 In the gain to defecting, which is four rather
first equilibrium, each state imposes the than three in the first period. No player
optimal tariff and receives a payoff of 2. has any incentive to deviate from the
In the second equilibrium, each state em- strategies just described, so these strate-
bargos the other by closing its economy to gies constitute an equilibrium whose first-
trade. This yields the payoffs (0,0 ). Thus, period outcome is (F, F).17 Thus, threats
in the game in which 6is played twice, to punish can be used to sustain coopera-
the only second-period outcomes that can tion in this model. Similar strategies will
be part of an equilibrium are (T, T) and also support (F, T) and (T F) as equilib-
(C, C).16 rium outcomes. 18
Playing 4'twice does, however, affect As in the neoliberal institutional cri-
the first-period outcomes that cantique of structural realism, anarchy does
be sus-
tained in equilibrium. Indeed, the free not imply a lack of cooperation. The
trade outcome (F, F) as well as the out- equilibria of this model are consistent
comes in which one state imposes a tariff, with the lack of a common government in
(T. F) and (F, T), can occur in equilib- the sense that they are subgame perfect. 19
rium. For example, the equilibrium strate- This means that carrying out the threat-
gies that lead to (F, X in the first period ened punishment of imposing an embargo
are for S1 to play F in the first period and in equilibrium is in the threatener's own
1308
self-interest. Neither state can improve its payoffs, the state suffering a relative loss,
payoff by backing down and not follow- while certainly preferring a higher payoff,
ing through on its threat, given the other cannot improve its payoff by altering its
state's strategy.20 Because implementing strategy. In this sense, this state is uncon-
the threat to punish deviation is in the cerned by its relative loss.
threatening state's own interest, the equi- Although the repeated prisoner's dilem-
libria supporting (F, F), (T, f, or (F, T) do ma provides the formal foundations for
not implicitly require that some external the neoliberal institutional critique of
authority exists to enforce threats or structural realism's pessimistic assessment
promises that are in a state's interest to of the prospects for cooperation in anar-
make but not necessarily to implement chic systems, repeated games, including
should the time come to do so. In this the repeated prisoner's dilemma and the
sense, these equilibria are in keeping with simple neoliberal institutional model
the notion of international anarchy as the developed here, formalize structural real-
absence of a central authority. Coopera- ism's understanding of the international
tion, even in anarchy, is possible.2" system and, especially the role of force in
In sum, the repeated prisoner's dilemma that system very badly. This, in turn, sig-
has been widely used to provide the for- nificantly weakens any analysis of the
mal underpinnings of the neoliberal insti- problem of relative gains and the feasibili-
tutional analysis of the problem of inter- ty of cooperation based on this type of
national conflict and cooperation. The formalization. Repeated games are poor
repeated game composed of twice playing models for two reasons.
4'formalizes much of the essence of the The first is more readily apparent but
neoliberal institutional critique of struc- less directly related to the problem of
tural realism in a simpler setting. The absolute and relative gains and the effects
shadow of the future makes cooperation of the concern for relative gains on the
possible even in an anarchic system in problem of cooperation. Structural real-
which each state judges "its grievances ism, as well as realism,22 focuses on a sys-
and ambitions according to the dictates tem in which states have the option of
of its own reason" (Waltz 1959, 157) using force if that seems to them to be in
because there is no authority that is their best interest and in which the use
superior to the states. In particular, the of force may transform the system. For
cooperative outcome in which both states Aron, international relations "present one
maintain open economies and receive original feature which distinguishes them
equal absolute gains can be supported in from all other social relations: they take
equilibrium. This equilibrium offers eachplace within the shadow of war" (Aron
state a payoff of five in the repeated 1966, 6). For Waltz, force in international
game. The absolute gains are equal, so relations "serves, not only as the ultima
there are no relative losses. There are, ratio, but indeed as the first and constant
however, also equilibrium outcomes in one. . . . The constant possibility that
which the states' absolute gains are un- force will be used limits manipulations,
equal and, therefore, one of the states suf- moderates demands, and serves as an
fers a relative loss. For example, in the incentive for the settlement of disputes"
equilibrium in which (F, T) is the outcome (1979, 113). And for Gilpin, hegemonic
in the first period, S1 receives four in the war is a principle means of systemic
first period and two in the second for a change (1981, 42-49).
total of six. S2 receives one in the first Given the central role that the possible
period and two in the second for a total of use of force plays in structural realism, it
three. Because these are equilibrium would seem that any model of the inter-
1309
national system envisioned in structural The second reason that repeated games
realism would have to satisfy two require- formalize structural realism's conception
ments: (1) the option of using force should of the international system poorly is more
be represented in the model; and (2) the subtle but also more directly related to the
model should allow for the possibility problem of absolute and relative gains
that the use of force might, whether inten- and the feasibility of cooperation. The
tionally or not, change the system. For a concern for the relative gains in structural
repeated-game model, the first require- realism arises because a state's relative
ment means that one of the options in the loss to another state may be turned
stage game that is being repeated should against it as that other state pursues its
correspond to the option of using force. own ends: "If an expected gain is to be
If, for example, the system is being for- divided, say, in the ratio of two to one,
malized with a repeated prisoner's dilem- one state may use its disproportionate
ma, then defecting should be taken to rep- gain to implement a policy intended to
resent the option of attacking. Repeated damage or destroy the other" (Waltz
games can satisfy this requirement. 1979, 105). If, therefore, a model is to
But repeated games cannot satisfy the provide a good formalization of structural
second requirement, and this renders realism's conceptualization of the prob-
them poor models of the international lem of relative and absolute gains, there
system implicit in structural realism. If the should be some way for a state in that
international system is modeled as a model to use a relative gain to its advan-
repeated game, then the state of the tage and to the detriment of the other
system (i.e., the actors, the options avail- state. Unless such an option exists in the
able to the actors in each period, and the model, there is no reason for a state to be
payoffs to the various possible combina- troubled by a relative loss, and the model
tions of actions) remains constant. Every is not capturing what structural realism
period looks like every other period in a sees as the essence of the problem of rela-
repeated game. Nothing the actors do can tive gains and cooperation. But no such
change the state of the system. Fighting a option exists in repeated games; conse-
hegemonic war in one period does not quently, they assume away what struc-
affect the constraints facing the actors in tural realism takes to be the essence of the
the next period. If, for example, both problem of relative gains and cooperation.
states defect in any round of a repeated That there is no way for a state's rela-
prisoner's dilemma and this is interpreted tive loss to be turned against it in a
as fighting a hegemonic war, then a hege- repeated game is easiest to see by con-
monic war can never change the system; sidering how actors sustain cooperation in
for immediately after the war, the same repeated games and then showing that a
states simply play the prisoner's dilemma state's ability to induce the other state to
again and the game continues on as if cooperate is unaffected by any relative
nothing had happened. Put another way, losses. To sustain cooperation in any two-
using a repeated game to model the inter- actor game, each actor must be able to
national system is to say that the system make the long-run costs of defection for
in 1939 was formally equivalent to the the other actor outweigh the immediate
system that emerged in 1945. Because gains. The future costs that each actor can
force can never change the system if this impose on the other determine whether or
system is formalized as a repeated game, not the actors can sustain cooperation.
formalizing the international system in- But in a repeated game, an actor's ability
herent in structural realism in this way is to make the future costs of defection
quite problematic. higher than the immediate gain does not
1310
depend on how well that actor has done in A Structural Model of the
the past relative to the potential defector. Problem of Absolute
If, for example, a sanction must be and Relative Gains
imposed for five periods in a repeated
game to make the cost of defection exceed I develop a very simple and highly styl-
the gain, then it will always take five ized formalization of the problem of abso-
periods to do this regardless of how well lute and relative gains. States will now
the actors have done in relation to each have the explicit option of using force if
other before the sanction is imposed. that seems to them to be in their best
Because the ability to impose future costs interest. More importantly, the model has
does not depend on how well the states been explicitly designed so that relative
have done relative to each other, achiev- gains and losses matter because they
ing a relative gain or suffering a relative affect a state's ability to use force success-
loss does not affect a state's ability to fully to further its interests. As will be
induce cooperation with threats of future seen, these constraints actually do induce
punishment. A state that has suffered a a concern for relative gains. For this
series of relative losses can threaten to reason, this model offers a better formali-
impose just as much future punishment on zation of the structural realism's under-
the other state as it would have been able standing of the problem of absolute and
to threaten had it not suffered these rela- relative gains.
tive losses. Relative losses do not impede Because the constraints -in the model
formally induce a concern for relative
a state's ability to enforce cooperation in a
repeated game. There is, therefore, no gains, the model may be used to examine
reason to be concerned with relative gains how changes in these constraints affect
or losses. Thus, the formalism of a the states' relative concern for absolute
repeated game omits what structural real- and relative gains and the feasibility of
ism sees as the very essence of the prob- cooperation. This examination shows that
many important aspects of neoliberal
lem of absolute and relative gains. For this
reason, repeated games formalize the institutionalism and structural realism
international system implicit in structural emerge as special cases of the model. If the
realism very badly.23 use of force is at issue because the cost of
A more satisfying analysis of the prob- fighting is sufficiently low, cooperation
lem of relative gains and the feasibility ofcollapses in the model. This is in keeping
cooperation requires a model in which the with the expectations of structural real-
option of using force is represented ism. But if the use of force is no longer at
explicitly and in which a state's relative issue, cooperation again becomes feasible.
loss may be turned against it. Using the This is more in accord with neoliberal
neoliberal institutional model analyzed institutionalism. The model thus offers a
here as a point of departure, I next synthesis of the apparently conflicting
develop a model that satisfies these twoperspectives of neoliberal institutionalism
requirements. In that model, each state's and structural realism on the problem for
ability to use force successfully to achieverelative gains and the feasibility of coop-
its ends will depend on how well it has eration by explaining variations in the rel-
previously done relative to the other ative concern of relative gains and in the
state. corresponding feasibility of cooperation
in terms of changes in the constraints fac-
ing the states and not in terms of the
attributes of the states.
The model also addresses two other
1311
Figure 2. The Option of Fighting underlying this analysis is that the

stronger a state is economically, the more
likely it is to prevail in war. Thus, relative
gains matter because they affect how the
states are likely to fare in the event of war
A A
and thus affect the states' future expected
payoff. It is important to emphasize, how-
ever, that relative gains are significant not
NA NA
because a state's utility is a function of
S1 S2 them-the states are still trying to max-
imize their absolute gains-but because
the constraints imposed by the underlying
technology of war makes it possible for a
issues in international state to use relations
its relative gains to itstheory
advan-
First, the structure tage andof the
to the internation
disadvantage of the other
system as Waltz defines state.24 it will be sho
to remain constant Tothroughout formalize a stylized technology
the of ana
sis. Yet the feasibility of cooperation warfare in which relative gains and losses
varies. This means that Waltz's notion of affect a state's ability to prevail in the
structure is unable to account for these event of war, the payoffs to attacking will
changes in the feasibility of cooperation. be taken to depend on what happened on
Second, the model clarifies the relation the first play of 4' The states are assumed
between anarchy and cooperation by to be roughly equal before the game
helping to explain why anarchy does not begins, so that the states will fight to a
imply a lack of cooperation in some sys- stalemate if there is war and if the differ-
tems but does impede cooperation in ence between the first-period payoffs is
others as neoliberal institutionalism has too small to give one state a significant
shown. military advantage. More specifically, if
The specter of war is introduced into one state attacks and the difference be-
the model by assuming that each state tween the first-period payoffs is less than
must decide whether or not to attack 3, then both states will fight to a draw,
between the two plays of 6As illustrated paying a fighting cost of 3.5, and then
in Figure 2, after the first play of & S1 play 6'for a second time. These payoffs
chooses between attacking, A, or not are shown in Figure 3, where 6'- t3.51 is
attacking, NA. If S1 attacks, there is war. the payoff matrix obtained by subtracting
If S1 does not attack, then S2 must decide 3.5 from each payoff in 6If the difference
between attacking or not. If S2 attacks, between the first-period payoffs is at least
there will also be war. If S2 does not 3, then the state with the higher payoff
attack, then there is no war and the game will be assumed to be sufficiently strong
ends with the second play of & economically that it will prevail if there is
To complete the specification of the war. If a state prevails, then its payoff to
modified game, the payoffs to attacking the entire game is what it received on the
must be defined. This specification first play of 4, which must have been 4
depends, in turn, on making a fundamen- (for otherwise the difference between its
tal assumption about the nature or tech-payoff and the other state's payoff would
nology of warfare; and it is at this point have been less than 3) less the cost of
that relative gains formally enter the anal-fighting, 3.5, plus the payoff to victory,
ysis. The very simple, highly stylized which will be taken to be 6. In sum, the
assumption about the nature of warfare total payoff to prevailing is 6.5. Defeat
1312
Figure 3. The Payoffs Figure 4. The Payoffs if

If Fighting Brings Stalemate S1 Will Prevail
? - (3.5) - (3.5) (6.5, -2.5) (6.5, 2.5)
A A A A
? / NA NA *
E : NA NA ?
S. S2 S S2
means a loss more of difficult

3.5than has been
due previously to
ing plus 0, appreciated. Because will
which (F, F) offers both
be t
off to defeat. Thus, the total payoff if a states equal absolute gains and conse-
state is defeated is the first-period payoff, quently no relative gains, it might seem
which must have been 1 if the difference that it would be possible to sustain this
between the states' first-period payoffs outcome in equilibrium. This, however,
was at least 3, less the cost of fighting, is not the case. Although this outcome
3.5, for a net of -2.5. These payoffs are offers no relative gain, deviating from it
depicted in Figure 4. does; and this means that it cannot be an
The prospect of war renders coopera- equilibrium outcome. If one state plays F
tion in the first play of impossible. and the other cheats with T, then the lat-
Clearly, the outcome (T, F) cannot be sus- ter will secure a large relative gain and
tained as a first-period equilibrium out- then attack and prevail. Each state, there-
come. If Si imposes a tariff and S2 main- fore, has an incentive to deviate from F if
tains an open economy, then Si is sure to the other player is playing F, so (F, F) can-
attack and defeat S2 because this maxi- not be an equilibrium outcome. Anarchy
mizes Si's payoff. The game would endand the possible use of force may even
with payoffs (6.5, -2.5). S2, however, prevent the realization of agreements that
can do better than this by playing Tin the provide equal absolute gains if deviating
first period. This will give S2 a first periodfrom the agreement would yield large
payoff of 2; and, because war would relative gains.
mean stalemate, neither state will attack. These results are in keeping with the
S2 will therefore end the game with at expectations of structural realism.
least two. Thus, S2 has an incentive to Although each state in the model is trying
deviate from F in the first period in orderto maximize its absolute level of economic
to deprive SI of its relative gain; and this welfare, it is doing so within an anarchical
means that (T, F) cannot be part of an structure whose underlying technology of
equilibrium. A similar argument shows warfare means that the relative gains of
that (F, T) cannot be an equilibrium another can be turned against it so as to
outcome. reduce its future level of economic wel-
The effects of the specter of war on the fare. These constraints force the states to
outcome in which both states maintain be concerned about relative gains in the
open economies (F, F) illustrate a point sense that outcomes that offer unequal
that suggests cooperation may even be absolute gains or the prospect of unequal
1313
absolute gains should one state cheat on C in the second period because of Si's
an agreement cannot be sustained in equi- deviation from F. In this case, S1's payoff
librium. The only first-period equilibrium will be 4 + 0 = 4. Si, therefore, has no
outcomes in the model are those in which incentive to deviate from the strategy
both states impose tariffs (T, 7) or close specified; nor, by a symmetric argument,
their economies (C, C). has S2. These strategies thus constitute an
But if the constraints facing the states equilibrium that leads to the cooperative
change, the conclusions are more in ac- first-period equilibrium outcomes.
cord with the expectations of neoliberal In brief, cooperation collapses in the
institutionalism. In particular, coopera- model when the use of force is at issue.
tion reemerges once the use of force is not This is in keeping with structural realism.
at issue. To formalize this, suppose that But if the cost of war is sufficiently high
the cost of fighting rises from 3.5 to 7. that the use of force is no longer at issue,
Then the payoff to victory is 4-7 + 6 = then cooperation again becomes possible;
-1, to defeat, 1 -7 + 0 = -6. In these and this is in accord with neoliberal insti-
circumstances, a state prefers not to at- tutionalism. Thus, the simple model
tack even if it has secured a relative gain developed here integrates much of struc-
on the first play of 4 Even if sure of pre- tural realism and neoliberal institutional-
vailing, no state will resort to force; and ism within a common analytic frame-
cooperation again becomes feasible in the work, in that many aspects of these two
model. To sustain (F, F), for example, the approaches appear as special cases of the
state's equilibrium strategies would be to present model.
play F in the first play of 6 not attack, The structural model illuminates two
and then play T in the second period if other issues in international relations
both states played Fin the first period. If a theory. First, it formally shows that
state deviated from F in the first period or Waltz's notion of a political structure
if a state attacked and the war ended in a (1979, 79-101) cannot explain the varia-
stalemate, then the states would sanction tion in the feasibility of cooperation con-
each other during the second period by sidered here. The political structure, as
playing C. To see that these are, in fact, Waltz defines it, remains constant and
equilibrium strategies that do support unchanged throughout the analysis. Con-
(F, F) in the first period, note that if bothsequently, structural changes as Waltz
states follow them, then each receives 3 in conceives of them cannot account for the
the first period, does not attack, and then variations in the likelihood of coopera-
obtains 2 in the second period for a totaltion. Explaining the range of cooperative
of 5. If S1 deviates from F by playing T, itbehavior examined here requires a more
will receive 4 in the first round and then detailed examination of the system's con-
must decide whether or not to go to war. straints than Waltz's definition of struc-
If it does, it will prevail, but its total pay-ture permits.
off will be the first-period payoff of 4, less To see that the three characteristics that
the cost of war, 7, plus the payoff to vic- define a political structure for Waltz, i.e.,
tory, 6. The overall payoff of deviating in the distribution of capabilities, the func-
this way is 3. This, however, is less than tional differentiation or nondifferentia-
Si would have obtained had it followed tion of the units, and the ordering prin-
the original strategy. Si, therefore, has no ciple, remain constant, note that the game
incentive to deviate by playing T and then is completely symmetric. There are only
attacking. Si also has no incentive to devi- two states, and the only difference be-
ate by playing T in the first period and tween them is that one is labeled Si and
then not attacking, for S2 will be playing the other is labeled S2. Thus, as the cost of
1314
fighting varies, there is no change in the government to ensure that states do not
distribution of capabilities or in the func- exploit these opportunities. In contrast
tional differentiation of the states. The with the structural model, anarchy does
ordering principle also remains that of not imply a lack of cooperation in the
anarchy, in that the games are always neoliberal institutional model, because
analyzed in terms of subgame perfect relative gains do not matter in that model.
equilibria. The political structure, there- As I have shown, each state's ability to
fore, remains constant, although the feasi- deter defection with the threat of future
bility of cooperation varies. punishment does not depend on how well
The second issue in international rela- that state has done relative to the other in
tions theory is the relation between anar- a repeated game. The constraints in the
chy and cooperation. Cooperation col- neoliberal institutional model do not offer
lapses when the use of force is at issue. the states a way to exploit a relative gain.
But anarchy in the sense of a "lack of Consequently, there is no need for a com-
common government" to enforce agree- mon government to ensure that the states
ments is not in and of itself the cause of do not exploit opportunities to turn a
this collapse. No such authority exists in relative gain to their advantage, because
the neoliberal institutional model in there are no such opportunities in the
which 6'was simply repeated twice or in model.
the structural model. Yet, cooperation In sum, two factors combine to induce
was an equilibrium outcome in the former a concern for relative gains and make
and in the latter when the use of force was cooperation difficult. The first factor,
sufficiently costly. Anarchy, as neoliberal which heretofore has only been implicit in
institutionalism argues, does not logically structural realism's description of the
imply a lack of cooperation (Keohane international system, is that the con-
1984, 65-84). straints defining the system create oppor-
What accounts for the lack of coopera- tunities for one state to turn relative gains
tion is not anarchy for this is present in to its advantage and to the disadvantage
all of the models. Rather, the explanation of other states. The second factor, on
lies in the different sets of constraints that which both structural realist and neo-
define what the states can do in anarchy. liberal institutional analyses of the prob-
In the structural model when the use of lem of relative gains and cooperation
force is at issue, as it is in the international have focused, is anarchy.25 If opportuni-
system envisioned in structural realism, ties to exploit relative gains exist, then the
relative gains matter. In these circum- absence of a common government to en-
stances, the constraints in the structural sure that the states do not exploit these
model formally create an opportunity for opportunities may impede cooperation.
each state to exploit its relative gain to its By emphasizing that structural realism
own advantage and to the detriment of implicitly assumes that the nature of mili-
the other state. Such constraints will then tary conflict means that the potential use
induce a concern for relative gains unless of force creates a set of constraints in
there is a central authority that can ensure which states can exploit their relative
that no state will exploit the opportunity gains, I may have obscured a more gen-
afforded by these constraints. Coopera- eral point. The general problem confront-
tion, therefore, collapses for two reasons. ing a state in this system is one of con-
The constraints create opportunities for a strained optimization in which the units
state to exploit its relative gains to its are trying to maximize their absolute level
advantage and to the disadvantage of the of economic welfare subject to a set of
other state; and there is no common constraints in which a unit's current rela-
1315
tive gain may be translated into a future insights? Any firm judgment about this
absolute gain for that unit and a future must await further work. As a first step
absolute loss for the other units. If this toward this judgment, I have begun to
more general description also applies to describe some of the insights that follow
other systems, then the present analysis from examining a simple model in which
may also help to explain why cooperation states are assumed to be attempting to
may be difficult to achieve in those sys- maximize their absolute level of economic
tems even though the units in those sys- welfare in an anarchic international sys-
tems are attempting to maximize their tem in which an absolute gain but relative
absolute gains and the possible use of loss today can become an absolute loss
military force is not a relevant part of thetomorrow.
problem. If, for example, long-run profit This simple model takes a step toward
maximization within an oligopolistic mar-bringing neoliberal institutionalism and
ket structure depends on relative gains instructural realism within a single analytic
terms of, say, market share, then this framework by showing that many aspects
analysis would suggest that it may be very of these two approaches can be seen as
difficult to achieve international coopera- special cases of the model. In keeping with
tion and collaboration even if national the expectations of structural realism,
security concerns are not at issue.26 states are concerned about relative gains
Describing the problem faced by states when the possible use of force is at issue.
in the international system in the more Cooperative outcomes that offer unequal
abstract terms of the optimization prob- absolute gains cannot be an equilibrium in
lem confronting them may suggest pre- this system. Indeed, even some agree-
viously unappreciated parallels between ments that offer equal absolute gains, and
this and other problems where the possi- therefore, no relative gain cannot be sus-
ble use of force is not the source of the tained in equilibrium, because one state
concern for relative gains. can achieve a relative gain by defecting
from the agreement. The prospects for
cooperation are, however, sensitive to the
Conclusions costs of fighting. If the use of force is no
longer at issue, then a state's relative loss
The problem of absolute and relative will not be turned against that state. Rela-
gains divides neoliberal institutionalism tive gains no longer matter, and coopera-
and structural realism. The former tion now becomes feasible. This is in
assumes that states are primarily con- keeping with the expectations of neo-
cerned with absolute gains while the latter liberal institutionalism.
assumes that states are primarily con- The model also furthers the analysis of
cerned with relative gains. What to two other issues in international relations
theory. First, it provides a simple formal
assume about states' preferences is a theo-
retical question, not an empirical one. example showing that Waltz's characteri-
States as rational unitary actors do not zation of a system's political structure in
exist. They are a theoretical construct. terms of its ordering principle, the func-
Thus, the question of whether states max- tional differentiation of its units, and the
imize absolute gains or are concerned distribution of capabilities across the units
about relative gains is empirically mean- cannot account for the variations in the
ingless. The real question is, Which feasibility of cooperation examined here.
assumption about state preferences is Changes in the cost of war do not affect
more useful? Which in the context of a the system's structure in Waltz's formula-
broader theory gives rise to better tion but do affect the potential for coop-
1316
eration in the model. Second, the model based on the same set of assumptions about states'
helps to clarify the relation between anar- preferences that structural realism employs (1984,
chy and cooperation. Anarchy, as neo- 67). However, Gowa (1986) in general and Grieco
(1988a) in particular argue that the repeated
liberal institutionalism has shown, does prisoner's dilemma on which Keohane bases much of
not imply a lack of cooperation. Rather, his analysis (1984, 65-84) is not consistent with
two factors combine to make cooperation structural realism.
difficult. The first is anarchy, defined as a 4. For other attempts to integrate the analysis of
conflict and cooperation, see Lipson 1984, Oye
lack of a common government that can 1986, and Stein 1984.
enforce commitments. The second factor 5. Explaining states' concern for relative gains in
(crucial to structural realism but only terms of constraints rather than preferences is inci-
implicit in it) is that the system be charac- dentally more in keeping with the original notions of
terized by a set of constraints that present structural realism. Indeed, what makes structural
realism structural and distinguishes structural real-
the states with opportunities in which ism from realism is that the concern for relative
they can use relative gains to their advan- gains is induced by the system's structural con-
tage and to the disadvantage of others. straints (Waltz 1959, 34-36).
Absent such opportunities, relative gains 6. See Powell (1991) for an effort in this
direction.
cannot be exploited, and there is no need
7. For Waltz, the concern for relative gains also
for a central authority to ensure that extends beyond the international system to other
these nonexistent opportunities are not self-help systems (1979, 91). But it is difficult to
exploited. define self-help precisely.
8. For examples using the repeated prisoner's
dilemma to formalize the problem of cooperation see
Axelrod 1984, Keohane 1984, and Oye 1986.
Notes 9. As is well known, if the prisoner's dilemma is
repeated a known finite number of times and there is
I thank David Collier, Michael Desch, James complete information, there is no cooperation in
Fearon, Jeff Frieden, Avery Goldstein, Joanne equilibrium. There must be infinitely many repeti-
Gowa, Joseph Grieco, Keisuke Iida, Ethan Kapstein, tions for there to be cooperation in equilibrium. If,
Robert Keohane, James Morrow, Ronald Mitchell, therefore, one is to study the problem of cooperation
Kamal Shehadi, R. Harrison Wagner, and Celeste formally in the context of a repeated prisoner's
Wallander for helpful comments and criticisms. dilemma, one must focus on an infinitely repeated
1. For discussions of neoliberal institutionalism's prisoner's dilemma. The present formulation permits
emphasis on absolute gains and realism's or struc- the formal study of cooperation in the simpler set-
tural realism's emphasis on relative gains, see Gilpin ting of a two-period repeated game.
1981, 18-25; Gowa 1986, 172-79; Grieco 1988a, 10. One of the weaknesses of this interpretation is
1988b, 1990; Hoffmann 1973, 6-12; Jervis 1988, that the states in 6are choosing their policies simul-
334-36; Keohane 1984; Keohane and Nye 1977, taneously and must endure significant losses before
23-37; Keohane and Nye 1987, 728-33; Krasner they can change their policies in the second period.
1983, 356; Lipson 1984, 12-18; Mearsheimer 1990, A better substantive model would not entail simul-
11, 44-45; Morgenthau 1967; Rosecrance 1981, taneous decisions or would allow the states to
705-7; Rosecrance 1986, 44-63; Snidal 1981, 3-6; change their policies before suffering significant
Stein 1983, 134; Stein 1984, 383-86; Viner 1948, losses.
1-29; Waltz 1959, 34-36, 196-98, 219-21; Waltz 11. Johnson (1953) shows that if both states
1979, 102-28; and Wolfers 1962, 67-115. impose optimal tariffs, then, depending on the elas-
2. Discussions of absolute and relative gains and ticities of demand, one or both states will suffer
the problem of cooperation include Gowa 1986; compared to the free trade outcome. Optimal tariffs
Grieco 1988a, 1988b, 1990; Lipson 1984; and Snidal cannot make both states better off. (In this analysis,
1991.
optimal tariffs are assumed to leave both states
3. For specific examples characterizing the con- worse off relative to the free trade outcomes. Conse-
cern about relative gains in terms of preferences, see quently, the payoffs to (F, F) are larger than those to
Gilpin 1981, 18-21; Grieco 1988a, 1988b, 1990; Jer- (T. 7) in K) He also establishes that there will be a
vis 1988, 335; Keohane 1984, 66; Lipson 1984, 15; positive volume of trade even if both states have
Mearsheimer 1990, 11, 44-45; Snidal 1991; and Stein imposed an optimal tariff. This implies that the
1983, 134; Stein 1984, 382. Keohane also claims that states obtain higher payoffs even with mutual
many aspects of neoliberal institutionalism can be tariffs than they would if there were no trade. Thus,
1317
payoffs to (T, 7) are higher than the payoffs to an mated in the limit if the game is repeated sufficiently
outcome in which at least one of the states plays C. often and the actors do not discount the future too
12. For an introduction to, and survey of, some much (Benoit and Krishna 1985; Fudenberg and
of the recent work on strategic trade, see Cohen Maskin 1986).
1990, Krugman 1986, Richardson 1990, Stegemann The existence of multiple equilibria suggests that
1989, and Stem 1987. one purpose of institutions and regimes may be to
13. Recall that the problem of cooperation re- focus expectations on a particular equilibrium. This
duces to what can happen in the first period because is certainly in keeping with the notion that regimes
'is only played twice, which implies that there is no are "sets of implicit or explicit principles, norms,
shadow of the future in the second period. rules, and decision-making procedures around
14. If the equilibrium of a repeated game did not which actors' expectations converge in a given area
require the states to play an equilibrium of the stage of international relations" (Krasner 1983, 2).
game in the last period, then one of the states wouldAlthough suggestive, this is formally rather ad hoc.
have an incentive to deviate from the purported Ideally, one would like to formalize the institution
equilibrium strategy in the last period of the game. within the model with the result that the model has a
This implies that the purported equilibrium strate- unique equilibrium. For an important effort along
gies cannot actually be equilibrium strategies these lines see Morrow 1990.
because no actor can ever have any positive incen- 19. See Rasmusen 1989, 83-89 for a discussion of
tive to deviate from an equilibrium strategy. subgame perfection.
15. I will focus exclusively on pure-strategy equi- 20. To see that the equilibrium supporting the
libria and disregard mixed strategies. (F, F) is subgame perfect, let Si's strategy be to play F
16. The combination of strategies (T, 7) is an in the first period and T in the second period if
equilibrium of because no state has any incentive (F, F) is the first-period outcome and C if not. Simi-
to deviate from its strategy, given the other state's larly, S2's strategy is to play Fin the first priod and T
strategy. If SI plays T, then playing T offers S2 itsin the second period if (F, F) is the first-period out-
highest payoff. Conversely, if SI playsfl, then alsocome and ' if not. As shown above, neither state
offers S2 its highest payoff. Similarly, (C, C) is an has any incentive to deviate from its strategy, given
equilibrium even though it is Pareto-dominated by the other state's strategy, so this combination of
(T, 7) because no state has any incentive to deviatestrategies constitutes a Nash equilibrium in which
from its strategy, given the other state's strategy. If(F, F) is the first-period outcome. To establish sub-
SI plays C, then playing C offers S2 its highest pay-game perfection, it must also be shown that neither
off; conversely, if SI plays C, then C also offers S2 its state has any incentive to deviate from carrying out
highest payoff. the threatened embargo, given the other state's stra-
17. The reason that cooperation can arise in equi- tegy and that (F, F) was not the first-period outcome.
librium in the game composed of a finite number ofSuppose that SI played T, rather than F, in the first
repetitions of 4but not in a finite number of repeti-period. Given Si's strategy of playing C in the sec-
tions of the prisoner's dilemma has to do with a veryond period if, as is assumed, (F, F) was not the first-
peculiar characteristic of the prisoner's dilemma, period outcome, then S2 has no incentive to deviate
which is that the minmax payoff equals the equilib-from its strategy of embargoing SI; that is, S2 cannot
rium payoff. This is true of very few other games; improve its payoff by playing something other than
and when the minmax payoff does not equal the C, given Si's strategy of playing C. Similarly, given
equilibrium payoff, cooperative outcomes, as well S2's strategy of playing C in the second period if
as any other individually rational outcome, can be(F, F) was not the first-period outcome, then SI has
supported in equilibrium in a finitely repeated gameno incentive to change its strategy by playing some-
if the number of repetitions is sufficiently large andthing other than C, given that S2 is playing C.
the actors do not discount the future too much. Neither state has any incentive not to carry out its
Indeed, the technical reason for adding the strategy threat, given the other's strategy; and the equilib-
C to the prisoner's dilemma formed by the strategies rium is subgame perfect.
Fand Tis to make the minmax payoff unequal to the 21. Although the equilibria of the neoliberal insti-
equilibrium payoff. This makes it possible to sustain tutional model are subgame perfect, the desire to
cooperation in the first period and thus to capture conduct this analysis in the simplest formal setting
the essence of the infinitely repeated prisoner's possible creates some difficulties. These equilibria
dilemma in a simpler setting. For further discussion are not renegotiation-proof (Farrell and Maskin
of the equilibria of repeated games, see Benoit and 1989). Surmounting this would require moving to a
Krishna 1985, Fudenberg and Maskin 1986, and more complicated formal setting that would tend to
Kreps et al. 1982. obscure the basic points the model is intended to
18. The problem of multiple equilibria frequently illustrate.
plagues repeated games. Indeed, almost any out- 22. See n. 5.
come can be supported in equilibrium. More formal- 23. Gowa (1986) and Grieco (1988a, 1988b, 1990)
ly, any individually rational payoffs can be approxi- have also criticized the use of the repeated prisoner's
1318
dilemma to model structural realism's conception of 12:3-42.

the problem of cooperation. Their criticisms, how- Jervis, Robert. 1978. "Cooperation under the Secur-
ever, focus on the states' preferences and the ity Dilemma." World Politics 30:167-214.
assumption that states are trying to maximize their Jervis, Robert. 1988. "Realism, Game Theory, and
absolute gain. Cooperation." World Politics 40:317-49.
24. The phrase "technology of warfare" is used Johnson, Harry. 1953. "Optimal Tariffs and Retalia-
here because the nature of military technology in tion." Review of Economic Studies 21:142-53.
this formalization constrains the states in the inter- Keohane, Robert. 1984. After Hegemony. Prince-
national system in much the same way that the tech- ton: Princeton University Press.
nology of production constrains firms in an oligopo- Keohane, Robert, and Joseph Nye. 1977. Power and
listic market. Interdependence. Boston: Little, Brown.
25. For examples of this focus, see Oye 1986. Keohane, Robert, and Joseph Nye. 1987. "Power
26. Indeed, Grieco (1990) studies the problem of and Interdependence Revisited." International
relative gains in the context of negotiations between Organization 41:723-53.
the European Community and the United States Krasner, Stephen. 1983. "Regimes and the Limits of
over nontariff barriers, a context in which it is dif- Realism." In International Regimes, ed. Stephen
ficult to attribute any concern for relative gains to Krasner. Ithaca: Cornell University Press.
the effects that a relative loss may have on the prob- Kreps, David, Paul Milgrom, John Roberts, and
ability of survival. This analysis helps explain why Robert Wilson. 1982. "Rational Cooperation in a
this concern may nevertheless still be present at least Finitely Repeated Prisoner's Dilemma." Journal
at the level of firms and why they may try to induce of Economic Theory 27:245-52.
the government to reflect these concerns. Krugman, Paul, ed. 1986. Strategic Trade Policy
and the New International Economics. Cam-
bridge: Massachusetts Institute of Technology
References Press.
Lipson, Charles. 1984. "International Cooperation
Aron, Raymond. 1966. Peace and War. New York: in Economic and Security Affairs." World Poli-
Doubleday. tics 37:1-23.
Axelrod, Robert. 1984. The Evolution of Coopera- Mearsheimer, John. 1990. "Back to the Future."
tion. New York: Basic Books. International Security 15:5-56.
Axelrod, Robert, and Robert Keohane. 1986. Morgenthau, Hans. 1967. Politics among Nations,
"Achieving Cooperation under Anarchy." In 4th ed. New York: Knopf, 1967.
Cooperation under Anarchy, ed. Kenneth Oye. Morrow, James. 1990. "Modelling International
Princeton: Princeton University Press. Regimes." Presented at the annual meeting of the
Benoit, Jean-Pierre, and Vijay Krishna. 1985. American Political Science Association, San
"Finitely Repeated Games." Econometrica 53: Francisco.
905-22. Oye, Kenneth, ed. 1986. Cooperation under Anar-
Cohen, Benjamin. 1990. "The Political Economy of chy. Princeton: Princeton University Press.
International Trade." International Organization Powell, Robert. 1991. "In the Pursuit of Power and
44:261-81. Plenty." University of California, Berkeley.
Farrell, Joseph, and Eric Maskin. 1989. "Renegotia- Typescript.
tion in Repeated Games." Games and Economic Rasmusen, Eric. 1989. Games and Information.
Behavior 1:327-60. Oxford: Basil Blackwell.
Fudenberg, Drew, and Eric Maskin. 1986. "The Richardson, J. David. 1990. "The Political Economy
Folk Theorem in Repeated Games with Discount- of Strategic Trade." International Organization
ing or with Incomplete Information." Econo- 44:107-35.
metrica 54:533-54. Rosecrance, Richard. 1981. "International Theory
Gilpin, Robert. 1981. War and Change in World Revisited." International Organization 35:691-
Politics. New York: Cambridge University Press. 713.
Gowa, Joanne. 1986. "Anarchy, Egoism, and Third Rosecrance, Richard. 1986. The Rise of the Trading
Images." International Organization 40:172-79. State. New York: Basic.
Grieco, Joseph. 1988a. "Anarchy and the Limits of Snidal, Duncan. 1991. "Relative Gains and the Pat-
Cooperation." International Organization 42: tern of International Cooperation." American
485-507. Political Science Review 85:701-26.
Grieco, Joseph. 1988b. "Realist Theory and the Stegemann, Klaus. 1989. "Policy Rivalry among
Problem of International Cooperation." Journal Industrial States." International Organization
of Politics 50:600-24. 43:73-100.
Grieco, Joseph. 1990. Cooperation among Nations. Stein, Arthur. 1983. "Coordination and Collabora-
Ithaca: Cornell University Press. tion." In International Regimes, ed. Stephen D.
Hoffmann, Stanley. 1973. "Choices." Foreign Policy Krasner. Ithaca: Cornell University Press.
1319
Stein, Arthur. 1984. "The Hegemon's Dilemma." Waltz, Kenneth. 1959. Man, the State, and War.
International Organization 38:355-86. New York: Columbia University Press.
Stem, Robert, ed. 1987. United States Trade Policies Waltz, Kenneth. 1979. Theory of International
in a Changing World Economy. Cambridge: Politics. Reading, MA: Addison-Wesley.
Massachusetts Institute of Technology Press. Wolfers, Arnold. 1962. Discord and Collaboration.
Viner, Jacob. 1948. "Power Versus Plenty As Baltimore: Johns Hopkins University Press.
Objectives of Foreign Policy in the Seventeenth
and Eighteenth Centuries." World Politics 1:
1-29.
Robert Powell is Associate Professor of Political Science, University of California,

Berkeley, CA 94720.
1320

American Political Science Association The American Political Science Review

Uploaded by

Copyright:

Available Formats

You might also like

American Political Science Association The American Political Science Review

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

American Political Science Association The American Political Science Review

Uploaded by

Copyright:

Available Formats

Absolute and Relative Gains in International Relations Theory

Author(s): Robert Powell

he problem of absolute and relative gains divides neoliberal institutio

AMERICAN POLITICAL SCIENCE REVIEW

work. But if neoliberal institutionalism the type or nature of states' preferences,

these models problematic. to punish uncooperative behavior today

finitely repeated prisoner's dilemma.9 Figure 1. A Neoliberal

Figure 2. The Option of Fighting underlying this analysis is that the

Figure 3. The Payoffs Figure 4. The Payoffs if

? - (3.5) - (3.5) (6.5, -2.5) (6.5, 2.5)

means a loss more of difficult

dilemma to model structural realism's conception of 12:3-42.

Robert Powell is Associate Professor of Political Science, University of California,

You might also like