The Complexity of Interdependence

Author(s): John A. Kroll

Source: International Studies Quarterly, Vol. 37, No. 3 (Sep., 1993), pp. 321-347
Published by: Wiley on behalf of The International Studies Association
Stable URL: https://www.jstor.org/stable/2600811
Accessed: 01-06-2019 00:44 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 [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

The International Studies Association, Wiley are collaborating with JSTOR to digitize,
preserve and extend access to International Studies Quarterly

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
International Studies Quarterly (1993) 37, 321-348

The Complexity of Interdependence

JOHN A. KROLL

Dartmouth College

"Interdependence" has long been a cardinal element in theoretical

analyses of international politics, but its utility has been hampered by
disagreements over both the meaning of the term and its links to changes
in state behavior. This article provides a general framework for resolving
those debates by using a game-theoretic analysis developed by Kelley and
Thibaut. Their work demonstrates that past calls for choosing between the
concepts of vulnerability and sensitivity are misplaced. The first measures
dependence, the second interdependence. Only by using both-and by
combining them with a third measure, independence-can we fully model
the interactions of states in international politics. With such a complete
model we can then begin to understand why some seemingly baneful
issues in international politics (such as Mutual Assured Destruction) can
still lead to cooperative outcomes, while some apparently benign issues
(such as free trade) often lead to conflict. Finally, the article offers some
preliminary propositions on the connection between interdependence
and cooperation, tying those propositions to the measures of interdepen-
dence developed by earlier analysts.

Recent political events have brought a new urgency to attempts at understanding

interdependence in international politics. Simple answers such as "interdependence
creates irreversible political unions" and "interdependence is politically mean-
ingless" are belied by the divergent outcomes of these events. On the one hand, the
seemingly indestructible political and economic structure of the Soviet Union has
devolved into a loose, and perhaps unstable, Commonwealth of Independent States.
On the other hand, the European Community, in spite of some setbacks, seems to
have acquired a momentum that is inexorably moving its members (and many states
on its periphery) toward stronger economic and political integration. In addition,
increasing international policy coordination at multilateral institutions such as the
United Nations, the International Monetary Fund, the General Agreement on
Tariffs and Trade (GATT), and the Group of Seven suggests that all states are
finding fewer areas in which unilateral action seems the most efficacious.
Past research concerning interdependence in international politics has been
both wide-ranging and contentious. Broad theoretical works have tried to demon-

Author's note: An earlier version of this paper was presen-ted at the joint an-nual conlvention of the British
Interniationlal Studies Association- an-d the Internlationlal Studies Association1, Lon-don, 28 March-1 April 1989 and as
CISA Working Paper No. 72, Cen-ter for Initernlationlal an-d Strategic Affairs, UCLA. The author would like to express his
appreciation- to Anne Florin-i, Richard Rosecrance, Arthur Stein1, Denlis Sullivan-, Oran Young, Jacqueline Dufault, and
the reviewers an-d editors of ISQfor their commen-ts on this article.

? 1993 Initerilatiolnal Studies Association.

Published by Blackwell Publishers, 238 Maini Street, Cambridge, MA 02142, USA, an-d 108 Cowley Road, Oxford OX4
1JF, UK.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 322 The Complexity of Interdependence

strate the value (or the inadequacy) of the concept for understanding state behavior
(Young, 1968-69; Waltz, 1970; Morse, 1970, 1972; Clark and Welch, 1972; Cooper,
1972; Little and McKinlay, 1978; Goldman and Sj6stedt, 1979; Baldwin, 1980; Jones
and Willetts, 1984; Dell, 1987; Keohane and Nye, 1989; Rosenau and Tromp, 1989;
Stein, forthcoming), while detailed empirical works have debated the existence or
nonexistence of trends that may be occurring in the levels of interdependence
(Deutsch and Eckstein, 1961; Rosecrance and Stein, 1973; Katzenstein, 1975;
Rosecrance et al., 1977). More recently, several analysts have explored the
connections between trade and political cooperation (Polachek, 1980; Gasiorowski
and Polachek, 1982; Domke, 1988). This past literature has illuminated three major
questions for research in this area: How do we define interdependence-as
sensitivity or as vulnerability? Is interdependence a new concept in international
politics-or has it always been there as a core element of the field? And how does
this concept of interdependence fit into common views on power in international
politics-is it a form of power or is it an alternative to power?
Without answers to these fundamental questions, research on interdependence
may be of little use in understanding the political changes now taking place in the
international system. This paper addresses these questions and seeks to provide
answers through four discussions: first, a review of the literature mentioned above,
outlining the origin of these points of contention; second, presentation of a new
framework for understanding interdependence and the issues it raises for research
in international politics; third, a preliminary application of this framework to issues
of international security and trade; and, finally, a brief reflection on future research
on interdependence and the value of this newly outlined approach for that work.
In brief, the conclusions of the paper are as follows: A debate that sets measures
of vulnerability against measures of sensitivity is done in vain; the first measure is
linked to the concept of dependence and the second is linked to the concept of
interdependence. Because, as we shall see, these two concepts are theoretically
distinct, analysis of state interactions must include measures of both. In addition,
economic interdependence may be a relatively new force in international politics,
if by interdependence we mean the need to coordinate policies to reach common
gains. Finally, the link between interdependence and increased cooperation in
international relations is not automatic. Such a link depends on the nature of that
interdependence and its relationship to the other elements of state interaction:
independence and dependence.

The Issues Surrounding "Interdependence"

Defining Interdependence

Perhaps the major disagreement that lies within the literature on interdependence
is also the most basic: how to define the concept. Is interdependence best seen as a
function of vulnerability or of sensitivity? Several analysts (Hirschman, 1945; Waltz,
1970; Baldwin, 1980) have argued that interdependence should be seen as a
question of a state's vulnerability, the cost it would face in breaking a relationsh
Interdependence is a case of mutual vulnerability, where two actors find themselves
in a relationship that would create large costs for both of them should it break
down. If such costs are, however, unequally distributed, then one can talk of a

IKeohanle anid Nye's work is an exception- to this apparen-t dichotomy. They talk of sensitivity an-d vulnlerability as
beiing two "dimen-sion-'s" of interaction, ones that exist in both depen-den-ce and in-terdependen-ce (1989:11-12).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 323

dependent relationship. Waltz argues that sensitivity measures of interdependence

may be "economically more interesting" but are also "politically less important"
(1970:206, 210). Baldwin concurs with Waltz's main conclusions, arguing that
interdependence should be seen as a case of vulnerability for three reasons: first, it
matches historical usage; second, sensitivity interdependence (defined below) can
vary inversely with vulnerability, thus distorting common usage; and, third,
sensitivity can be called just that, whereas interdependence should be reserved for
vulnerability issues (Baldwin, 1980:490-491) .2
Other analysts have offered an alternative view of interdependence, one that
stresses the increased sensitivity states feel when their relations with one another
increase (Cooper, 1972; Tollison and Willett, 1973; Deutsch, 1978; Whitman,
1979). Under such a definition, two states become more interdependent when
events that take place within one state have an impact upon events taking place in
another state. If vulnerability focuses on the costs of breaking relations, sensitivity
can be said to focus on the costs of maintaining it (Keohane and Nye, 1989:12). A
specific argument for the economic origins of such sensitivity in trade relations
appears in the work of Tollison and Willett (1973:268). They argue, "An increase
in elasticities (i.e., in interdependence) will, ceteris paribus, increase the degree of
simultaneity of price movements (i.e., increase the degree of integration)." Such
price equalization is a function of the amount of trade between two economies.3
Rosecrance et al. combine these two elements to argue that high interdependence
between economies exists where one finds both high horizontal interdependence
(measured through trade flows) and vertical interdependence (measured through
correlated factor prices). They go on to argue that such measures of sensitivity
"reveal interdependence much more directly than notions of 'vulnerability' and
'invulnerability' " (1977:427-428).

Power versus Interdependence

A second area of contention revolves around the question of whether

interdependence should be seen as a form of power relations or as an alternative to
power relations. The choice has important implications. Theorists who see
interdependence as an issue of vulnerability also see it as a form of power. For
example, Baldwin (1980:495, 505), after concluding that interdependence is based
on vulnerability, and therefore is a case of mutual dependence, goes on to examine
dependence as a form of power. His conclusion seems to be that to be dependent is
to lack power and to be independent is to have power. In addition, Hirschman
(1945:v) has shown how states can translate the economic power that grows out of
dependence into "political pressure and leverage." Independence in these analyses
becomes the opposite of dependence; to increase one, you must decrease the other.
One cannot both increase and decrease power at the same time.
Other theorists who focus on the sensitivity side of interdependence seem to be
arguing for a conception that places it as an alternative to power. Perhaps the best
example of such an approach is John Burton's work (1972). Burton rejects the
"billiard-ball" model of international politics, with its image of states crashing
against one another, their final trajectory to be determined by the force behind

2Baldwin's article provides a rich review and anlalysis of both past an-d present uses of the term interdependence. In
addition- to Hirschman-, Baldwini cites Sir Norman An-gell, Ramsay Muir, an-d Francis Delaisi as some of the authors
who define the con-cept of in-terdepen-dence as onle based on relations that are costly to break (1980:481-484).
3The linlk between trade and factor price equalization is predicted by the Heckscher-Ohlin Theorem. See Caves
ai-dJones (1981:118).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
324 The Complexity of Interdependence

each actor's movement. Instead, he argues for a "cobweb" model of international

politics, where the seemingly Lilliputian restraints of day-to-day enterprise are soon
found to have so bound the traditional organs of the state Gulliver that the idea of
state power becomes obsolescent. Interdependence is not a form of vulnerability to
other states, it is a new form of relationship with them, one in which neither state
can act without some form of coordination with the other.

Something Old or Something New?

The above observation leads us to the final issue concerning interdependence

covered in this essay: Is interdependence something new in international politics or
has it always been there? Those theorists who use their research to answer the
question "Has interdependence increased over time?" are, in effect, suggesting that
interdependence is some new force that may or may not be appearing in
international politics. However, if interdependence simply means that states are
able to influence one another, it is not new; war, hardly an event limited to modern
times, places states in a situation in which they are highly interdependent
(Rosecrance et al., 1977:426).
If interdependence is to be seen as a modern occurrence in international
politics, it must be distinguished from such ubiquitous historical phenomena. And,
if interdependence, or a form of interdependence, is being held out as an explana-
tion for the appearance of a new type of behavior by states, then its definition must
include some means of distinguishing itself from the relations states have always
faced. And, finally, if interdependence is to mean something more than simply
interaction among states, some theoretical foundation is needed for linking
patterns of such interaction to patterns of state behavior.

Components of Interdependence

The extensive use of game theory in analyses of international politics demonstrates

that interdependence has always had a role in international politics, for a key
strength of game theory is its ability to reveal interdependent relationships. To be
interdependent, states must face a set of payoffs that depend, to some degree, on the
behavior of the other actor. Without such a strategic interaction, there can be no
interdependence.4 If there is some variation in payoffs across cells of the matrix,
interdependence may be present; if there is no variation, interdependence is absent.
However, the typical game theory analysis of actors' behavior holds much more
information than the mere absence or presence of interdependence. Harold
Kelley and John Thibaut (Thibaut and Kelley, 1959; Kelley and Thibaut, 1978), in
a study of interpersonal relations, offer an original method for analyzing such
matrices, a method that brings forth the different forms of interaction that actors
experience-independence, dependence, and interdependence-tied together in
a common structure. (The summary below is only a small part of the numerous
insights these authors provide concerning interdependence.) Kelley and Thibaut
argue that these three components constitute distinct elements in the definition of
interaction between actors. (See Figure 1.) In the remainder of this paper I use the
term interaction to refer to this common structure, the composite relationship
between actors. I use the terms independence, dependence, and interdependence to
describe elements of that relationship. (The justification for these distinctions and
the definitions of other terms in the chart are offered below.)

4For a broader analysis of strategic behavior versus parametric behavior see Elster (1 979:chap. 1).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 325

INTERACTION

INDEPENDENCE DEPENDENCE INTERDEPENDENCE

Concordance

Correspondent Non-Correspondent

FIG. 1. The Elements of Interaction

Kelley and Thibaut argue that a strategic relationship can be broken down into
three constituent parts. First, actors can possess some independent control over the
payoff they receive, even in a strategic relationship; this they call Reflexive Control.
It is the degree to which an actor can influence its payoff by its own choice of
strategy, regardless of the strategy chosen by its opponent. Second, an actor's final
payoff can be dependent upon the actions of its opponent; this Kelley and Thibaut
call Fate Control. With such Fate Control an actor sees that, no matter which
strategy it chooses, part of its final payoff is determined solely by its opponent's
choice of strategy. Finally, there can be an element of the total payoff that the actor
can achieve only through interdependent, or joint, actions; this Kelley and Thibaut
call Behavior Control. The stronger such Behavior Control is, the more an actor
must coordinate its strategy with that of its partner to maximize its payoffs.
For example, imagine two people who decide that one of their rooms is looking
rather rundown and needs to be repainted. Jerry is very good at painting windows
and moldings; he is meticulous in his work, but he is also very slow. Elaine is very
good at painting the walls; she works much faster, which allows her to paint large
areas without leaving streaks, but she is apt to stray when painting trim. In this
relationship Jerry possesses some independence (Reflexive Control); he can
increase his satisfaction with the appearance of the room if he uses his skills to
produce nicely painted woodwork. However, Jerry is also dependent upon Elaine
(Fate Control) because he knows that the walls will look their best only if she
chooses to paint them. Finally, Jerry is interdependent with Elaine (Behavior
Control) because no matter who paints what, he will be satisfied with the room only
if they coordinate their actions; one must paint the walls while the other paints the
trim. If they both paint the same component, the room will remain unfinished.
This decomposition of a strategic interaction can be illustrated more formally by
applying it to a game matrix taken from Kelley and Thibaut (1978:36-38). Figure 2
clearly demonstrates the most fundamental requirement for a strategic
relationship. Neither actor's payoff can be predicted without knowledge of the
strategy followed by the other actor. In this matrix Actor B can choose between two
strategies, b, and b2, while Actor A can choose between a, and a2. Each cell of the
matrix contains the payoffs the two actors receive from their combined choices of
strategy. Actor B receives the payoff in the lower left corner of each cell, while
Actor A receives the payoff in the upper right corner. In addition, we can go
beyond that recognition of a strategic interaction by using Kelley and Thibaut's
method. Let us focus first on Actor B.
Figure 3 starts with the same matrix, but shows only the payoffs for Actor B. We
then can dissect that matrix. The first element of the relationship between Actors B

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
326 The Complexity of Interdependence

A
al a2

15 20

B bl 20 8
-2 6
b2 15 -2
FIG. 2. Sample Game Matrix

Actor A

al a2

Actor B |8+||7.5
ACOBbi 20 F |75
IjIII
14. 1+I2 II

b2 i0 114.5 0 1 0.5 -2
B'S TOTAL GAME B's RC B's FC B's BC

WEIGHTS: RCB = + 7.5 FCB = + 14.5 BCB = -2.5

FIG. 3. Analysis of Actor B's Components of Interaction

and A is the degree to which B can control its payoffs irrespective of what A does.
The control B possesses lies in its ability to choose between strategy b, and b2. The
first step in determining the extent of this control is to calculate the average payoff
B receives from strategy bl, using the payoffs from each of the cells linked to this
strategy, by using the following values: Bb, = (Bbiai + Bbia2)/2. In this equation, Bb1
is the average payoff Actor B gets from strategy bi, while Bbiai is the payoff Actor B
receives in cell bial and Bbia2 is the payoff Actor B receives in cell bla2 . Applying this
formula to Figure 3 we have (20 + 8) /2 = 14. Next, the average payoff B receives
from strategy b2 is calculated from payoffs in each of the cells associated with it: B b2
= (Bb2ai + Bb2a2)/2. In Figure 3 we have (15 + (-2))/2 = 6.5. Finally, we determine
the relative value of each strategy by taking the absolute value of the difference
between these two average payoffs: RCB = IBb1- Bb2I. In this case we have 114 - 6.51 =
7.5. On average, B receives 7.5 units more from strategy b, than from strategy b2, no
matter which strategy A chooses, giving B a Reflexive Control (RC) of 7.5. This value is
placed in the cells related to bl, the strategy tied to this greater average payoff.5
B's Fate Control (FC) is the degree to which B's payoff is controlled by Actor A,
irrespective of what B does (i.e., the higher this number, the more control A has

5Kelley and Thibaut are not arguing that Actor B will ever receive a payoff of 7.5 units because of this Reflexive
Control. Siince such a payoff does not appear in anly of the matrix cells, it canlnlot. However, this measure of Reflexive
Control (the average payoff B will receive from each strategy) does tell us the degree to which B can have an impact on
the final payoff it will receive from a particular interaction- with Actor A, irrespective of anly actions taken- by that actor.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 327

over B's payoff). A's power to influence that payoff is based on its ability to choose
between strategy a, and a2. The first step in determining this power is to calcul
the average payoff B receives when Actor A chooses strategy a,, using the following
values: Bai = (Balb, + Balb2)/2. In Figure 3 we have (20 + 15)/2 = 17.5. On averag
B will receive a payoff of 17.5 if A chooses strategy al. Next, the average payoff B
receives if A chooses strategy a2 is calculated in the same fashion: Ba2 = (Ba2b, +
Ba2b2)/2. In Figure 3 we have (8 + (-2))/2 = 3. On average, B will receive a payoff
of 3 if A chooses strategy a2. Finally, we determine the relative value of each
strategy by taking the_absolute value of the difference between their two ave
payoffs: FCB = IBai - Ba21. In this case we have 117.5 - 31 = 14.5. Thus, on average,
14.5 units of B's payoff depends on the strategy chosen by A, independent of the
strategy chosen by B. To that degree, then, one can say this FCB shows the degree
to which B prefers A to choose strategy a1 or strategy a2. Or, to put it another way,
it is the degree to which B's final payoff depends on A's choice of strategy.
Finally, we have the degree to which B's final payoff is a function of joint action
taken by both Actor B and Actor A. This is B's Behavior Control (BC), the degree
to which B's payoff depends upon B coordinating its strategy choice with that
taken by Actor A. Or, another way to phrase it is to say that B's BC reflects the
degree to which B, by changing its strategy, can influence A's incentive to choose a
particular strategy. The easiest calculation of this value comes from determining
the payoff needed to make RCB + FCB + BCB equal the original payoff in each cell
of the total game. For example, cell bla1 in the Total Game matrix has a payoff of
20. Because RCB and FCB for this cell already add to 22 (i.e., 7.5 + 14.5), the value
of BCB in this cell must be -2 in order to bring the total back down to the required
20. The complete set of values for BCB appears in the final matrix: alb1 and a2b2
have a value of -2 and b1a2 and b2a1 have a value of 0.5.
The total effects of these elements on B's payoffs are summarized in the weights
for each type of control, the magnitude of difference between payoffs created by
each element. The total effect of RCB is +7.5 (7.5 - 0); B prefers strategy b1 because
its average payoff is 7.5 units higher than strategy b2, irrespective of A's choice of
strategy. The total effect of FCB is +14.5 (14.5 - 0); B prefers A to adopt strategy a1
because its average payoff for B is 14.5 units higher than strategy a2, irrespective of
B's choice of strategy. The total effect of BCB is calculated by measuring the
difference between payoffs along the two diagonals. In this case that is (-2) - 0.5 =
-2.5. B's average payoff is 2.5 units higher if it adopts a strategy different from A's
choice (selecting b1 if A chooses a2 and b2 if A chooses a1).6 This analysis can be
repeated for Actor A, giving us the complete picture of the interaction between the
two actors (Figure 4).7
With Kelley and Thibaut's analysis we can now describe fully the structure of
interaction that exists between Actors A and B. First, these two actors get very little
from joint action with each other (BCA and BCB), and what little benefit there is
leads each actor to prefer a different set of outcomes. (In the MBC matrix, B
prefers a1b2 or a2b1 to gain its best payoff, whereas A prefers alb1 or a2b2
best, or least-worst, payoff.) But they are still highly dependent. Each actor has
more control over the other actor's payoff (Fate Control) than it does over its own
payoff (Reflexive Control). A's Reflexive Control over its own payoffs-RCA-is less
than A's Fate Control over B's payoffs-FCB, and RCB is less than FCA. In addition,
these relationships are pulling the two actors in different directions. RCB and FCA

61n order to preserve additional information from the matrix, Kelley and Thibaut give each weight value a sign of
+ or -. The + symbol denotes that the highest payoff for an actor occurs in the upper left cell, while a - sign is given if
the highest payoffs occur in other cells. This sign allows one to compare actors: Do both receive their best payoffs in
the same cell or not?
7Both actors' various types of control can be combined to form three types of control within each matrix: Bilateral
Reflexive Control (BRC), Mutual Fate Control (MFC), and Mutual Behavior Control (MBC).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
328 The Complexity of Interdependence

Actor A

a] a2

15 20 o 6.5 15.5 15.5 -o.s -2

Actor B bl 20 8 7.5 7.5 + 14.5 0 + -2 0.5

b2 15 -2o 6.5 14.5 ? o ? 0.5 -2 5

TOTAL GAME BRC MFC MBC

RCB = +7.5 FCB = + 14.5 BCB = -2.5

WEIGHTS:

RCA = -6.5 FCA =+15.5 BCA = + 1.5

FIG. 4. Summary of Both Actors' Components of Interaction

both have the same sign. Thus, B's own interests lead it to pick the strategy that
benefits A. On the other hand, RCA and FCB have different signs, pulling A to take
actions that do not give B its best payoff. Actor B can get Actor A to follow the
strategy B prefers (a,) only by threatening to follow a strategy that is not in B's best
interest (b2).
These methods have some important implications if we wish to turn them around
and use Kelley and Thibaut's terms-Reflexive Control, Fate Control, and Behavior
Control-to construct (rather than dissect) matrices of strategic interaction. As
Kelley and Thibaut note (1978:50), the mathematics behind their decomposition of
game matrices resembles an analysis of variance for each of the two actors' payoffs.
Their approach extracts the three forms of variance that exist in a two-by-two
matrix: variance across rows, variance across columns, and variance across diagonals.
These three measures can fluctuate independently of one another; or, to put it
more formally, they are orthogonal.8 Suppose that we do not know the payoffs in
the Total Game matrix faced by two actors, but do have information about their
respective elements of Reflexive Control, Fate Control, and Behavior Control.
Kelley and Thibaut's analysis tells us that if we have information on each one of
these elements, for each actor, we have a complete description of the strategic
forces at work in their relationship (i.e., we have information on all the elements of
variance in their relationship). However, since these three elements are orthogonal,
we cannot use information about one element to infer information about the
others. (For example, knowing an actor's Fate Control does not also tell us how
much Behavior Control it possesses.) We can use Kelley and Thibaut's three
elements as building blocks for describing an actor's strategic interaction with
others, but that description will be complete only if we have measures that supply
information for each one of these elements.
This framework is not just a new way of defining interdependence. Kelley and
Thibaut also provide the means for establishing a theoretical connection between
the level and type of interactions states face and the influence of these connections
on an actor's behavior. This can be done by an examination of the relationships

8For example, take Figure 3. There we see that Actor B faces the followiing values: RCB = 7.5, FCB = 14.5, and B
= -2.5. We could add to B's Reflexive Control and make its RCB = 17.5 by addin-g 10 poin-ts to the payoffs of cells bla
and bha2 in the RC matrix. Actor B's preferen-ce for strategy bh over strategy b9 has nlow in-creased from 7.5 to 17.5.

The Total Game matrix, because it reflects the sum of the three elements, must also change. The payoff in cell bla,
changes from 20 to 30, and the payoff in cell bja2 chan-ges from 8 to 18. However, B's Fate Con-trol and Behavior
Con-trol remain the same: FCB will still be 14.5, anid BCB will still be -2.5.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 329

that are formed between the elements of Reflexive Control, Fate Control, and
Behavior Control, both within the structure of one actor's payoffs and between the
structures of both actors' payoffs.
Kelley and Thibaut categorize those additional relationships using two terms:
correspondence and concordance. Correspondence refers to the degree to which the
actors' interests coincide within the Behavior Control matrix. Following the rule
discussed earlier, each actor's BC weight (the difference between its highest BC
payoff and its lowest BC payoff) is given a sign, positive if the highest payoff lies in
the alb, and a2b2 cells, and negative if it lies in the a1b2 and a2b, cells. If both
actors' weights of Behavior Control are positive in sign, or both are negative in
sign-signifying that they can both benefit at the same time-Kelley and Thibaut
rate the actors' Behavior Control as correspondent. If we assume that each actor has
the same choice in strategies, then a positive correspondence signifies that they
benefit from following identical strategies (e.g., both drive along the left side of the
road); whereas a negative correspondence signifies that they benefit from adopting
complementary strategies (e.g., one drives the car while the other acts as navigator).
When they cannot benefit at the same time (i.e., one has a negative BC and the
other a positive BC), then they are said to have noncorrespondent BC.9 For
example, suppose the two strategies available to each actor represent possible arms
control treaties on naval deployments. Actor A is a major naval power; Actor B is a
minor naval power. Actor A wants to retain its military advantage and is adamant in
opposing any treaty that B may offer. Actor A wants to use its opposition to the
treaty to demonstrate that all naval arms control is inherently unworkable and can
do so only by actively disagreeing with whatever proposal B makes. Actor B, on the
other hand, sees arms control as a way to compensate for its naval weakness;
therefore it is willing to support any treaty A supports because it simply wants to set
a precedent for controlling naval forces. The matrix in Figure 4 contains an
occurrence of such noncorrespondent BC; the two actors cannot both acquire
their highest payoff (+0.5 for B, -0.5 for A) at the same time.
Concordance focuses on the relationships among the matrices of Reflexive Control,
Fate Control, and Behavior Control. Here, we focus not on the relationship between
actors, but on the relationship between the payoffs received by any one actor. Does
each of these elements of interaction-Reflexive Control, Fate Control, and
Behavior Control-provide the actor with the highest payoff in the same cell? In
other words, do these payoffs reinforce one another, leading an actor to one
strategy that maximizes its payoff in each matrix? Or, do they counter one another,
leading to cross-cutting pressures over the choice of a strategy? In Figure 4, Actor
B's Reflexive Control (RCB = +7.5) is concordant with B's Fate Control over A (FCA
= +15.5),10 whereas A's Reflexive Control (RCA = -6.5) is discordant with A's Fate
Control over B (FCB = +14.5). Actor B can hurt Actor A only by suffering a loss itself
(moving to strategy b2 moves A from an average payoff of 15.5 to one of 0, but at the
same time B's average payoff drops from 7.5 to 0). Actor A can threaten B without
loss (moving from strategy a, to a2 causes B's average payoff to drop from 14.5 to 0,
but A's average payoff increases from 0 to 6.5). This paragraph demonstrates that
Reflexive Control, which by its nature is about independence rather than
interdependence, is an important element in understanding the ability of Fate
Control and Behavior Control (or, as I argue below, dependence or
interdependence) to influence an actor's policy choices.

9This conicept of correspon-den-ce applies only to the Behavior Control matrix. By their very structure, the
Reflexive Control an-d Fate Control matrices will have a corresponden-ce of zero, neither negative nor positive (Kelley
an-d Thibaut, 1978:35).
IOFCA is the degree of Actor A's payoff controlled by the action-s of B.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
330 The Complexity of Interdependence

Back to the Debates

A review of the above discussion demonstrates the compatibility that lies between
the literature on international interdependence and Kelley and Thibaut's work on
strategic interactions. The first seeks to show how certain concepts-vulnerability
and sensitivity-can be used to define the overall strategic interaction of two or
more states. The second takes the opposite approach and shows how such
composite structures can be broken down into their constituent factors: Fate
Control, Behavior Control, and Reflexive Control. This section demonstrates how
the concepts from the literature on interdependence match the factors defined by
Kelley and Thibaut. It then goes on to argue that by recognizing these connections
(vulnerability with Fate Control and sensitivity with Behavior Control) we can
enhance our use and understanding of the terms dependence and interdependence.
Let us first look at the connections between Fate Control and vulnerability. Fate
Control measures the degree to which an actor's payoff is determined solely by the
actions of its partners. In other words, it is the degree to which an actor's final
payoff is vulnerable to the choice of strategy taken by those partners. These phrases
ring consistent with much of the literature on interdependence as vulnerability.
For example, Baldwin describes vulnerability interdependence as a case in which
"one thing ... must rely upon something else for fulfillment of a need" (1980:475).
Hirschman argues that the gains a state receives from international trade make it
dependent upon the state that "bestows" those gains; they are "that part of a
country's well-being which it is in the power of its trading partners to take away"
(1945:vii, 19). We see that both Fate Control and vulnerability seek to gauge the
same force: the degree to which one actor can control the benefits its partner
receives from the relationship.
A comparable case can be made for a similarity between the concepts of
Behavior Control and sensitivity. Behavior Control measures the degree to which
two actors can reach certain payoffs only by coordinating their strategies. An
actor's final payoff is not determined solely by its own choice of strategy (as under
Reflexive Control) or solely by its opponent's choice of strategy (as under Fate
Control), but, instead, by the interaction of those choices. This same interactive
effect lies at the heart of many discussions of sensitivity interdependence. For
example, in an article often cited as the archetype of this literature, Richard
Cooper writes as follows:

As nations become increasingly interdependent, ... countries desiring to pursue

tax or regulatory policies that deviate widely from those policies in other countries
will find themselves stimulating large inflows or outflows of funds, firms, or
persons; these induced movements will in turn weaken the intended effects of the
policies, or make them more costly. (1972:166)

Sensitivity measures the degree to which states must coordinate changes in their
policies (i.e., not "deviate widely") to achieve the sought-after benefits of those
new policies. This same emphasis on linkage appears in Rosecrance et al.
(1977:427), when the authors describe interdependence as a situation in which a
change in one state effects a comparable change in another state, causing the
states to move up and down the "ladder of international position" together.11

IlDuriing the collapse of the Europeain Exchan-ge Rate Mechanism in September 1992 much was made of the costs
borne by Great Britain- aind Italy as they were forced to devalue their currencies in the face of high interest rates in
Germany. Less notice was given- to how this failure to coordinate fiscal and monetary policies has also hurt Germanly.
The same inflow of foreign currencies that caused the Deutschmark to appreciate also raised the monley supply in
Germanly by some 5 percen-t (Hutton, 1992:6), creating the same inflationary pressure that the Bundesbank's policy
was originally designled to forestall. In the long run-, Germany may suffer as much as any country in Europe because
of this failure to coordinate fiscal and monetary policies.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 331

Both sensitivity and Behavior Control seek to measure the same influence: the
degree to which an actor can reach certain benefits only by orchestrating its
choice of strategy with that of its partner.
This discussion suggests the following conclusions about analyzing the strategic
interaction of states in the international system: (1) We should not equate mutual
dependence with interdependence. The first term is tied to issues of vulnerability
and denotes a situation in which two actors each possess some unilateral control
over the payoff received by the other; the second term is tied to issues of sensitivity
and denotes a situation in which an actor shares control of its final payoff with
another actor. If both actors face this second situation, then one has a case of
mutual interdependence (i.e., mutual Behavior Control). (2) Under this rubric, the
terms positive interdependence and negative interdependence are precise. The first refers
to situations where both actors, by following identical or complementary strategies,
can benefit at the same time. The second refers to situations where an actor can
reach its best payoff only by denying the other actor in the relationship its best
payoff. (3) Kelley and Thibaut's framework demonstrates that both dependence
and interdependence must be included in analyses of a strategic interaction.
Neither one, alone, gives us a complete picture of the forces at work in such
situations. Moreover, that complete picture requires the addition of a third element,
independence-the degree to which an actor can still control its own payoffs even
when locked into a strategic relationship. And (4), by linking Reflexive, Fate, and
Behavior Control with independence, dependence, and interdependence, we see
that the latter must also be orthogonal. In other words, if we are to use them to fully
define a strategic relationship, we must have an independent measure of each one.
There is no sense in which one measure, be it vulnerability or sensitivity, can take
preeminence over all others.
Identifying these elements and their connections allows us to understand why
interdependence has been so slippery a term in the study of international politics.
Such strategic interaction is not simply a question of losing control of one's payoffs
from a relationship, or of seeing another actor gain control over one's payoffs, or of
requiring increased coordination between actors. Nor can one assume that the
identification of change in any one element implies a change in the others. Instead,
studying strategic interplay requires a study of all three elements: interdependence,
dependence, and independence. Each plays a role in determining the changes in
behavior that the totality of such interactions can bring about.
We see that the debate over choosing between vulnerability and sensitivity as the
single measure of interdependence is misplaced. Each element can exist in a
relationship at the same time and each element can vary without influencing the
other. In addition, the international politics literature is flawed in its assumption
(normally an implicit one) that a rise in a state's vulnerability and sensitivity will
decrease its ability to act unilaterally. The above analysis shows that neither
interdependence nor mutual dependence is the inverse of independence. Such
independence can exist in its own right. Therefore, demonstrating that inter-
dependence or dependence has increased does not mean, automatically, that
independence has gone down. Indeed, it may increase along with dependence or
interdependence, producing little net effect on a state's overall situation. A full
understanding of any set of international interactions can take place only if we
have measured each element of those interactions and examined the degree of
concordance between those elements.
The same consonance appears in Kelley and Thibaut's final term. As we've seen,
correspondent Behavior Control is the equivalent of positive interdependence,
whereas noncorrespondent BC is the equivalent of negative interdependence. If
we adopt Baldwin's criterion of "historical usage" in selecting a definition for inter-
dependence (see above), this affinity between Behavior Control and the common

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
332 The Complexity of Interdependence

use of the terms positive interdependence and negative interdependence reinforces the
argument for distinguishing between the mutual dependence and vulnerability of
Fate Control (where such terms do not apply) and the interdependence and
sensitivity of Behavior Control (where they do apply). (Thus, based on his own
terms, Baldwin's rejection of sensitivity as the measure of interdependence is
premature.)

Summarwy

This section has been an analysis of the techniques developed by Kelley and
Thibaut for analyzing the strategic qualities of a game-theory matrix and the
parallels between those techniques and the extant literature on theories of inter-
national interdependence. We have seen that to assess the influence of
interactions on a relationship between actors we need three separate elements of
information:

* First, we have to establish the absence or presence of the three components

of interaction-Reflexive Control, Fate Control, and Behavior Control-
recognizing that each one exists independently of the others.12 These three
components coincide with the concepts of independence, dependence, and
interdependence discussed in the international politics literature.
* Second, we have to establish the degree of correspondence within the
Behavior Control matrix. Is the interdependence between the two actors a
positive or a negative relationship?
* Finally, what is the relationship between the components of interaction? Are
they concordant, collectively steering an actor toward one paramount
strategy, or are they discordant, imposing contradictory forces upon an
actor's strategic choices? Concordant changes are much more likely to
produce noticeable modifications in an actor's behavior.

Typical Matrices of Interaction

To illustrate their influence, in this section I use the three elements of interaction
as tools to analyze two common game matrices: a zero-sum game and a pure
coordination game. Together they outline the extremes of a strategic relationship:
pure conflict versus pure cooperation.13 Using Kelley and Thibaut's framework we
shall see that coordination is not synonymous with cooperation (because the zero-

12Kelley aind Thibaut also arialyze the relative weights of these components; for example, is there more FC thai
RC? This paper limits itself to a discussion- of the presence and absen-ce of these components for the sake of brevity.
13By "cooperation" I mean outcomes in which a chan-ge in strategy by either actor would not in-crease both actors'
payoffs (e.g., the Cooperate, Cooperate cell in a Prisoner's Dilemma). Under this definition all other outcomes are
cases of con-flict-either they are Pareto-deficien-t (e.g., the Defect, Defect cell in the Prisoner's Dilemma) or one
actor is exploiting the other (e.g., the Cooperate, Defect cells in the Prisonler's Dilemma an-d Chickenl matrices).
Rapoport, Guyer, and Gordon- state that such joint maximization results from a "cooperative orien-tation." They also
go on to call it the result that comes from a Pareto-optimal strategy (1976: 117, 18-19). (There is some debate in the
field about such usage of the term (Hamburger, 1979:86). The question- seems to be whether all or on-ly some of the
outcomes that are not Pareto-deficient should be called Pareto-optimal.) With this definition one canl talk about
cooperation- through exchange (Fate Con-trol), cooperation- through coordination- (correspon-den-t Behavior
Con-trol), and cooperation- through con-gruity (Reflexive Con-trol). Keohanle (1984:51) argues that this last element
should be distinguished from other forms of cooperationi and relegated to its owIn category-harmonly-because both
actors can advance their interests without anly form of en-gagement. As Kelley and Thibaut have showni, however, such
apparent harmony must be included in discussionis of cooperation; if that harmony is discordanit with the other
elemen-ts of a strategic in-teraction it canl actually underminie attempts at cooperationi.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 333

sum game requires just such coordination) and that the game matrix most often
used as an example of cooperation (the coordination game) is based on interde-
pendence and sensitivity rather than on dependence and vulnerability.

Zero-Sum Games

A zero-sum game matrix is the quintessential model of pure conflict between actors:
each actor gains only at the expense of the other actor. (See Figure 5.) In each cell
the cost of gain by one actor is borne by the other actor, leaving the total payoff for
both actors at zero. Zero-sum games can be used successfully to model conflicts over
fixed-sum resources: land, ideology, wealth (a la mercantilism), and so on.

Actor A

al a2

5 -5 0 0 0 0 5 ~~~~~~~~~~~-5
ActorB bJL-5 J5 =| 0 + ? | 5 5

b2 5 -5 1 1 ? I O -s

TOTAL GAME BRC MFC MBC

RCB = O FCB = O BCB = - 10

WEIGHTS:

RCA = O FCA = O BCA = + 10

FIG. 5. Zero-Sum Matrix

Because each actor's payoff depends on the strategy chosen by the other actor,
one can easily say that they are in an interdependent relationship. Using Kelley
and Thibaut's method of analysis, the nature of this relationship becomes clearer.
In such a perfectly symmetric zero-sum matrix neither actor possesses either
Reflexive Control (RC)-independent control over their own payoffs-or Fate
Control (FC)-control over their opponent's payoffs.14 But they are very
interdependent in terms of their Behavior Control (BC)-the degree to which
they must coordinate their strategies to control the final payoff. However, their BC
is noncorrespondent; both actors cannot reach their best payoffs at the same time.
Actor A can reach its maximum payoff only if B does not, and Actor B can reach its
maximum payoff only if A does not. This decomposition demonstrates that both
actors are indeed locked into a strong interdependent relationship; neither actor is
able to reach its maximum payoff without coordinating its strategy with that of its
opponent. It also demonstrates the absence of a power relationship, through the
possession of either RC or FC by the actors.

14This does not have to be the case. One can construct a symmetric zero-sum game in which there is no BC but
there is RC aiid FC. The goal here is to show that zero-sum games can engender BC (i.e., interdependence), even
though they are fun-damen-tally conflicting and that this noncorrespondent BC is the converse of the positive
in-terdependence contained in a coordin-ation- matrix. Zero-sum games do iiot necessarily engender nonlcorresponldent
BC. It is also possible to construct a zero-sum matrix that genlerates zero RC, FC, and BC; simply put the values (-5, 5)
in each cell. Since there is no variation across cells there is nio indepen-dence, no dependence, and no interde-
penden-ce in the matrix. This wide variety of outcomes demon-strates why it would be un-wise to simply equate zero-
sum games with niegative in-terdependen-ce.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
334 The Complexity of Interdependence

Actor A

al a2

Actor B b 2 8 3 3=+ + v

TOTAL GAME BRC MFC MBC

RCB = 3 FCB = 0 BCB = - 10

WEIGHTS:
RCA = 0 FCA = -3 BCA = + 10

FIG. 6. Zero-Sum Matrix with Reflexive and Fate Control

One can add RC and FC to a zero-sum matrix by disrupting the symmetry

between actors of the payoffs in the matrix. (See Figure 6.) Doing so reveals an
added element of linkage which exists in such zero-sum games. Giving Actor B
some Reflexive Control (by increasing the average payoff of strategy b1 from 0
to 3) without changing the payoff from Behavior Control (which remains at 10)
automatically gives B some Fate Control over Actor A (-3). The ability of RC and
FC to vary independently, discussed earlier, is removed by making the payoffs of
both actors within each cell linked (together they must always add up to zero). As
noted in the discussion of Figure 3 above, such opposing signs for each payoff
demonstrate the conflict of interest between A and B: the policy that benefits B
most (b1) is the least beneficial to A.

Pure Coordination Games

At the other end of the cooperation/conflict spectrum is a pure coordination game.

(See Figure 7.) In such a game both actors benefit at the same time, but only if they
coordinate their strategies. In addition, in its purest form, neither actor has a prefer-
ence over which strategy is chosen. The premier example is deciding which side of
the road traffic should stay on. No one cares whether they drive on the right or left,

Actor A

al a2

Actor B bi 5 0 ? 1 ? ? ? ? | ? ?

b2 0 I s 1 W W o I I o I o I I o
TOTAL GAME BRC MFC MBC

RCB = 0 FCB = 0 BCB = + 5

WEIGHTS:

RCA = 0 FCA = 0 BCA = + 5

FIG. 7. Coordination Matrix

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 335

but they are quite concerned with coordinating their choices with the other actors.
A decomposition of such a matrix may seem a little trivial since both the
Reflexive Control and Fate Control matrices are empty, and the Behavior Control
matrix is a duplicate of the overall matrix. However, the absence of Fate Control
and Reflexive Control, as in the case of a zero-sum matrix, demonstrates that the
interdependence of the two actors, in the symmetrical form of the game, comes
solely from the need to coordinate their policies. In the case of a coordination
game, this need can lead to mutual benefit, as demonstrated by the matching signs
of the BC figures.

Summary

This section offered two additional elements to our understanding of interdepen-

dence. Interdependence, as a concept, is not limited to either positive or negative
relationships between actors. Thus, if we are to talk about the influence of
interdependence upon state behavior, one must specify which type of relationship
exists. Are states linked in such fashion as to allow them to both reach their best
payoffs at the same time (positive interdependence)? Or are they in a relationship
in which only one actor at a time can reach those maximum benefits (negative
interdependence)?15 As we saw above, that specification also has important rami-
fications for the manner in which additional elements of interaction such as RC
and FC affect the payoffs actors face (at least in the case of constant-sum games).
We have also seen that at least one form of interaction, Behavior Control, can
exist without either state possessing any type of power over the actor. In both of the
above examples, zero-sum games and coordination games, the payoff each actor
receives is a function of the strategy followed by the other actor. But in neither case
can one say that one actor, or even both actors, has, or have, power over the other
(in the sense that they cannot influence the other actor without some rebound
effect upon themselves). To conclude:

* Interdependence is not limited to positive relationships. It can be negative

or positive. Such characterizations have important implications for the effect
such interdependence will have both on the behavior of actors and on the
manner in which other elements of interaction are introduced into the
relationship.
* Analytically, one can isolate an element of interaction, Behavior Control,
that is based on coordination, not on exchange. Thus, to talk about the
strategic interaction of actors, we must be prepared to talk about both
interdependence (MBC), based on coordination, and mutual dependence
(MFC), based on exchange.

15Usilng this conception, one should resist the temptation to declare all wars as cases of negative interdependence.
If capitulation will benefit a state (because it stops the destruction of that state), then the two warring states are in a
situation of mutual dependence. They can cooperate by exchange: for example, Actor A will stop bombing Actor B if
it gives A a slice of territory. Actor A ben-efits from the acquisition of the new territory; Actor B benefits, one assumes,
because the value of peace exceeds the value of the lost territory. War involves negative in-terdependen-ce only when
one state wants peace at any price and the other wants war above all else. A possible historical example is the
ultimatum presented by Austria-Hungary to Serbia in 1914. Serbia, desperate to avoid war, was willin-g to agree to
almost any demand placed upon it. Austria-Hun-gary, equally desperate to humiliate Serbia, was determinled to reject
any agreement puLt forth by it (Remak, 1967:111-115). Each actor's strategy was directly tied to the other's, but they
could n-ot both succeed at the same time. This is the essen-ce of n-egative interdependen-ce.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 336 The Complexity of Interdependence

Interdependence and MAD

The above would be a very interesting essay on the concepts of independence,

dependence, and interdependence even if we were to stop here, but it would be of
little use to students of international politics unless we went on to demonstrate that
the concepts can also be applied to the issues, theories, and measures of interde-
pendence found in international politics. Figure 8 demonstrates the first appli-
cation of this technique, an analysis of a security issue: Mutual Assured Destruction
(MAD). (The goal here is to model the essential concept of deterrence based on
MAD, not any particular nuclear balance.) Such an analysis raises two important
points for theories of international politics. First, dependence is not a concept that
can be limited to economic issues; it also occurs in security issues. As such,
dependence is not a term that can be monopolized by political economists and
used to argue that increased economic ties have created some fundamentally new
reality in state relationships. Such changes must be sought elsewhere. Second,
mutual dependence, although closely tied to the exercise of power over others, can
lead to cooperation, under appropriate conditions.
MAD exists when two states possess a sufficient number of nuclear weapons to
inflict unacceptable damage on an opponent's civilization. The balance of such a
condition rests on each side possessing weapons that can survive a first strike by an
opponent and still inflict unacceptable damage in retaliation. If one accepts the
argument of MAD in its simplest form, a state receives no advantage from attacking
its opponent with nuclear weapons. Given the invulnerability of an opponent's
weapons, state A, the attacking state, receives no military benefit from launchin
first strike-Actor B still possesses the ability to retaliate and destroy A's civilization.
Nor is there any political or economic gain for A in its destruction of B-Actor A has
no more land, no more wealth, and so on than it had before the attack.16 Actor A's
attack merely inflicts costs upon opponent B by destroying its economy and
population. (This lack of gain is indicated in Figure 8; the payoffs for Actor A
remain the same in cell BPEACE, APEACE and in cell BPEACE, AWAR-)
In addition, retaliation by the attacked opponent follows a set pattern. First, it
imposes similar destruction on the attacker, even though the attacker has struck
first. (E.g., Actor A suffers the same as Actor B in the BWAR, AWAR cell even thou
A may have attacked first. This ability to retaliate with assured destruction even
after an attack is the essence of minimum deterrence.) Also, the retaliating state

16This set of payoffs comes from statements such as Robert McNamara's: "Nuclear warheads are not military
weapons in the traditional sense and therefore serve no military purpose other than to deter one's opponent from
their use" (1987:16). It also comes from statements such as Schelling's: "A weapon that can hurt only people, and
cannot possibly damage the other side's striking force, is profoundly defensive; it provides its possessor no incentive
to strike first" (1960:233). RobertJervis (1978a) argues that the deterrence theories of Schelling (and of other
theorists such as Brodie) can be modeled as a game of Chicken. However, it is hard to un-derstand two shifts in
payoffs that the Chicken matrix requires. First, in that matrix a unilateral attack nlot R S.F.

only hurts one's opponent, it also benefits the attacker; this later shift runs contrary Retr
to the Schelling quote and implies that nuclear weapons canl be used for gain. Stan
Second, in Chicken, the attacked opponent suffers by retaliating against its attacker. 4,
However, if the opponent has already seen its civilization destroyed by attack, it is CHICKEN MATRIX
hard to imagin-e how retaliation- makes this situation- worse. The confusion- arises from
Jervis's use of the expression "stand firm" to con-flate two differen-t strategies: the threat of war and war. If onle's
opponent yields, theii stand firm seems to mean- that on-e has followed a strategy of threatening war-thus, the
outcome 4,2: you gain a political advantage whereas your opponen-t has accepted a political loss without the
destruction- of war. If your opponen-t also stan-ds firm, theni, suLddenly, that same strategy becomiies the equivaleint of
war: you attack, they attack, both receive their worst payoffs, an-d on-e arrives at the outcome 1,1. This coniflation of a
strategy of threaten-ing niuclear war an-d on-e of fighting nuLclear war obscures the dilemma of ex ante anld ex post
interests described below, whereas the matrix and an-alysis given- here highlight that dilemma.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 337

Actor A

Peace War

Actor B Peace 0 -6 ? ? ? 6 ? | 6 6
6 W 0 | 0 0 6 6
War 0 -6 I 6 6
TOTAL GAME BRC MFC MBC

RCB = O FCB = 6 BCB = O

WEIGHTS:

RCA = 0 FCA = 6 BCA = O

FIG. 8. Mutual Assured Destruction

has nothing to lose by countering the initial attack-its civilization has already
been destroyed. (E.g., the payoffs for Actor B are the same in cell BPEACE, A
and BWAR, AWAR.)
Figure 8 also dissects this initial matrix, breaking it down into the constituents of
Reflexive Control, Fate Control, and Behavior Control identified by Kelley and
Thibaut. The inability of states to use nuclear weapons for traditional methods of
gain (e.g., seizing land or economic assets without destroying them) appears
sharply in the absence of Reflexive Control: nuclear weapons do not increase a
state's independence. Because states can use nuclear weapons to hurt (by
attacking) or help (by not attacking) other states they possess Fate Control over
those states. Finally, the absence of Behavior Control demonstrates the conundrum
of MAD: there is no reason to retaliate for an attack by your opponent. Because
states receive no benefit from coordinating their policies with those of other states
(i.e., there is no BC) they do not profit from retaliating against their attacker; they
simply impose pain upon it. Ex ante policies of MAD may be useful to deter an
opponent, but ex post policies of MAD offer no benefits to the retaliating state. The
relationship between states under MAD is one of mutual dependence (MFC) based
on the exchange of restraint-I will not attack you if you do not attack me. This is
different from the issues of coordination that would occur if there were elements
of interdependence in the BC matrix.

Trade and Interdependence

The purpose of this section is to demonstrate how Kelley and Thibaut's process can
analyze the game matrices that model another aspect of world politics, one that is
most often linked to research on international interdependence and depen-
dence-world trade. Below is an argument for establishing the basic matrices of
international trade and, then, the breakdown of those matrices.

The Structure of Trade: Large States versus Small States

The most important bit of information in describing a state's position in the realm
of international trade is its economic size. Neoclassical trade theory, in discussing
the role of tariffs in international commerce, distinguishes between those policies
available to economically large states and those available to economically small

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
338 The Complexity of Interdependence

states.17 Economically large states are, by definition, states whose presence in the
international market is so strong that they are able to influence international price
lines through the amount of goods they buy or don't buy: they are price makers.
Economically small states, because they are such small consumers, cannot
influence world price lines: they are price takers. Because of the difference in their
ability to influence international prices, large states and small states receive very
different payoffs in choosing between strategies of free trade and closure.
Small states never benefit from tariffs; such taxes simply increase the cost of
imported goods, reducing consumer purchasing power and leading manufacturers
into areas of inefficient production. Their national income always suffers. Small
states have no influence over the international market; therefore, their best policy is
always free trade.
Large states, because they have the ability to influence international prices, can
indeed benefit from tariffs if the tariffs are used appropriately. If a large state
imposes an optimum tariff, it can increase its national income (Scitovsky, 1942;
Johnson, 1953-54). Tariffs decrease demand at home by making imported goods
more expensive. But large states, by their economic size, can make this decrease in
demand so strong that world prices begin to fall (in response to decreasing
demand versus constant supply). An optimum tariff is one that causes world prices
to drop sufficiently so that the losses incurred from reduced trade are exceeded by
the gains from paying less for the imported goods.18 Under this optimum tariff
national income has increased. These different payoffs for large states and small
states lead to very different strategic interactions. The matrices generated by
comparing two large states trading with one another and a small state trading with
a large state illustrate this contrast.
Using the above argument for optimum tariffs, one quickly sees that two large
states facing one another over the issue of free trade versus closure are in a
Prisoner's Dilemma game (Panchamukhi, 1961; Riezman, 1978; Conybeare, 1984;
Kroll, 1992). (See Figure 9.)19 If we take mutual free trade (3, 3) as our starting
point, we can assess the relative advantages and disadvantages of changing strategy.

17The following discussion assumes that neoclassical trade theory is correct in arguing that the one element of
trade a state can effectively control is prices and that those prices, as measured by that state's terms of trade,
determine the payoffs it faces when dealing with the international market. There are alternative arguments. Some
analysts contend that other national interests (e.g., supporting infant industries, ensuring full employment,
protecting strategic industries) drive state policy. Neoclassical theory rejects these arguments on the grounds that
there are alternative policies (most often, direct subsidies to selected industries) that perform the stated goal more
efficiently; for these goals tariffs are always a second-best policy choice (Caves andJones, 1981:218-222). Hegemonic
stability theory argues that free trade occurs because hegemon-s both benefit from open trading systems and have the
power to administer such systems. As Conybeare (1984:61-62) has shown, however, the economic power of
hegemons makes protectionism, not free trade, their dominant strategy. A third alternative focuses on the role of
domestic interest groups in the setting of commercial policy. International trade and shifting economic environ-
ments can hurt certain domestic groups leading them to prefer tariffs over free trade. Each of these works notes,
however, that, although they have a theory that explains why certain groups demand tariffs, they cannot explain why
a state supplies those tariffs (Cassing, McKeown, and Ochs, 1986:845-846; Wallerstein, 1987:731; Rogowski, 1989:20).
In other words, these works provide theories of interest group formationi, but they do n-ot necessarily provide theories
of state behavior.
If we were at the stage of fully developed theory, this assumption- that neoclassical trade theory accurately
represents a state's payoffs from trade would be tested, as all assuLmption-s are, by the success of the empirical
propositions derived from it (Waltz, 1979:10). At the momen-t, it is a plausible and orthodox starting point for
examining some of the issues facing those who seek a theory that links economic interdepenence to state behavior.
18Home prices are still high, limiting consumer demand, but the state sees only the world prices. The differenlce
between the two is the level of protection, which is pocketed by the state through the collection- of tariff taxes.
19I have changed the normal set of payoffs slightly by making un-ilateral defection give the defecting actor five
points rather than the more tradition-al four points. This adds an elemen-t of Behavior Control to the matrix, as we
shall see. Behavior Con-trol does n-ot have to occur in a Prison-er's Dilemma, but if it does, it always shows the two
actors' BC to be correspon-den-t in their in-terests.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 339

Large State: A
Free
Trade Tarifs

Free 3 5 O 1.5 2.5 2.5 0.5 1

Large Trade 3 1 I= 0 + 2.5 + 0.5 1
State: B 1 2 O 1.5 K 0 1 0.5
Ta riffs 5 2 11.5 11.5 1 2.5 O 1 0.5

TOTAL GAME BRC MFC MBC

RCB = - 1.5 FCB = 2.5 BCB = - 0.5

WEIGHTS:
RCA = - 1.5 FCA = 2.5 BCA = -0.5

FIG. 9. Large State/Large State Trade

Each actor faces a temptation to defect from free trade since by doing so it
increases its payoffs (from 3 to 5); this increased payoff comes to the defecting
actor through the improvement of its terms of trade vis-a-vis its opponent. Terms of
trade is a relative concept; therefore, if Actor B improves its payoffs by putting up a
tariff, Actor A suffers a loss (moving from 5 to 1). Actor A then has an incentive to
retaliate, by adopting its own tariffs, restoring the original terms of trade (moving
from 1 to 2). This new equilibrium occurs at a lower level of trade, leaving both
actors worse off than under free trade (compare 3, 3 to 2, 2).
Small states, in their dealings with large states, face a very different set of payoffs.
(See Figure 10.) Two major, and interrelated, changes have taken place. First,
Actor B, the small state, cannot benefit from the imposition of a tariff on imports.
Small states, by definition, cannot improve their terms of trade because they can
only be price takers in the international market. A tariff put in place by a small
state reduces the total payoff the state receives from trade (from 5 to 3).20 Even if
the small state's terms of trade have been adversely affected by the tariffs of a large
state, retaliation by the small actor will only worsen the outcome (from 2 to 1).
Again, small states can no more influence their terms of trade in retaliation than
they could in unilateral defection, therefore the tariff only worsens their position.
Free trade is always the dominant strategy for small states.

Large State: A
FreeTarif
Trade Tarifs

Free 3 5 O 2 O O 3 3
Small Trade 5 2 1.5 1.5 + 2.5 o + 1 0.5
State: B 3 W 2 O O 3 3
Tar iffs3 1 o o 2.5 O 0.5 I

TOTAL GAME BRC MFC MBC

RCB = 1.5 FCB = 2.5 BCB = 0.5

WEIGHTS:
RCA = -2 FCA = O BCA =

FIG. 10. Small State/Large State Trade

2500ue might argue over the placement of 2B and 3B inl the matrix, suggesting that the two be reversed. For
sake of this argument such a switch would not affect the conclusions this analysis draws. For example, switching the
payoffs 2B and 3B would still leave Free Trade as the dominant strategy for a small state.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
340 The Complexity of Interdependence

Second, not only are small states unable to improve their lot through tariffs, they
also cannot affect the payoffs received by trading partners who are large states.21 As
noted above, such adverse effects come from the fact that terms of trade is a
relative concept: If mine improve, yours must suffer. If a small state cannot
improve its terms of trade, it cannot hurt its trading partner's terms of trade. Thus,
as Figure 10 illustrates, the choice of strategy by the small state, Actor B, has no
influence on the large state's (Actor A) payoffs (3 vs. 3 or 5 vs. 5). The only action
that can influence the large state's payoffs is its own decision to impose an
optimum tariff, which can increase its payoff from 3 to 5.

The Elements of Trade Interdependence and Dependence

Having constructed these two composite matrices which states face in their trading
relations, we can now use Kelley and Thibaut's method to analyze the forms of
interaction embedded in them. Figure 9 also illustrates the breakdown of the large
state/large state trading relationships. One can see from this figure that trade
among large states entails a form of linkage between member actors. To be
specific, trade entails mutual dependence, as shown in the Fate Control the actors
experience. Each trading actor feels the degree to which its payoffs are determined
by the choice of strategy invoked by its trading partner. These two actors can be
said to be mutually dependent upon one another: 2.5 units of influence over their
final payoffs will, in this example, be determined by their partner's action,
irrespective of the strategy choice they make.
Because mutual dependence (FC rather than BC) dominates the actors' rela-
tionship, cooperation has to be based on exchange (rather than coordination).
However, such coordination is undermined by the discordance between each
actor's FC and RC. The two actors' Fate Control pushes them toward some form of
exchange that would make Free Trade/Free Trade the outcome, but their
Reflexive Control pushes them toward an outcome in cell Tariffs/Tariffs. Two
large states facing the issue of free trade versus protection do indeed face a
dilemma in their interactions. They can receive the better payoffs of free trade only
if they can impose some structure that allows them to resist the temptation of
following that which is under their control, their dominant strategy, and that
ensures that their trading partner also installs such a structure.22
One also sees that the Behavior Control (BC) is extremely low in this breakdown
of the Prisoner's Dilemma. This dearth of BC means that a change in strategy by
one actor has little effect on the other actor's choice of strategy (i.e., it does not
make it more or less desirable for the other actor to change strategies). There is
little incentive for the two actors to coordinate their choices. As noted above, it is
this need to coordinate activities that Kelley and Thibaut label interdependence
(as opposed to the mutual dependence based on exchange in MFC).

2IConybeare (1984) argues otherwise, stating that retaliation by a small state again-st a large state hurts both actors.
Small states are always price takers in the internationlal market, so tariffs imposed in retaliation have no more effect
than tariffs imposed in unilateral defection.
22Thus, most of the recent literature on cooperation in Prisoner's Dilemma has focused oni techniiquies to alter
this domination of independence over mutual dependen-ce, on the grounds that in the lonig run both actors will be
better off cooperating. For example, Michael Taylor's argument is that actors need to reduce the temptationl to
defect to make continued cooperation the preferred strategy (1987). Jervis's solution to the Security Dilemma
(1978b) can be explained in a slightly different way. Offensive security can cause coniflict because it creates
discordance between RC anid FC (you are more secure, but you are also more able to hurt your opponenit); defenisive
security causes less conflict because it does not create discordanice betwveeni RC anid FC (you are more secure, but are
no more able to hurt your opponent).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 341

We also see that such mutual dependence does not preclude the continued
existence of some independence in the control each actor has over the payoff it
receives. Both states still possess some Reflexive Control (1.5 units in this example);
that is, they can influence the payoffs they receive irrespective of what the other actor
does. In this case, a Prisoner's Dilemma matrix, that Reflexive Control reflects their
possession of dominant strategies: each actor is better off when it imposes tariffs.
In addition, close examination of such Prisoner's Dilemma matrices reveals that
the addition of significant elements of correspondent Behavior Control is rather
difficult. For example, increasing the payoffs of mutual free trade relative to the
other payoffs could inject a substantial amount of BC into the matrix, increasing
the payoff both states could receive from coordinating their policies by both
adopting low tariffs. But trade theory states that unilateral defection will always pay
better than cooperation, at least to some degree. Therefore, if there were some way
to increase the benefits of mutual free trade, the temptation for unilateral
defection (and thereby the RC each actor possesses) must also increase.
However, this leaves some discussions of interdependence in a bit of a quandary,
for such change in Behavior Control is the essence of Tollison and Willett's
argument about the connection between interdependence and factor price
equalization. They argue that as economies become integrated, and the prices in
those markets are tied together, decision makers lose some of their economic
autonomy (1973:268). Independent fiscal policies are quickly thwarted by
international forces flowing into a domestic market made porous by the free
movement of trade, labor, and capital. (See the example of German interest rates
and inflation rates in footnote 11.) Such price sensitivity increases the pressure on
states to coordinate policies with one another to control the effects of their
economic policies. It is this coordination that is reflected in the Behavior Control
matrix of Kelley and Thibaut's method for analyzing interdependence.
Putting these arguments together, we arrive at the following conclusions about
the economic interaction that arises from trade between economically large states.
First, economic trade theory demonstrates that two trading partners are, in part,
mutually dependent upon one another for the gains they can receive from such
trade. Second, conflict arises in such trade relationships because states still retain
partial independence in such transactions, an independence that draws them away
from cooperation. And third, arguments about the importance of price sensitivity
in economic interdependence are in effect saying that trade, over time,23 leads to
an increased need to coordinate policy-an outcome that is structurally different
from an exchange process of mutual dependence.
The relationship between economically large states and small states is quite
different. A decomposition of this matrix, however, still leads to some insights.
(The breakdown of this matrix is given in Figure 10.) As noted in the discussion of
the original matrix for small state/large state trade, if one concentrates solely on
economic theory, economically small states possess only one optimum policy
toward trade; they should always keep their markets open no matter what policy
their trading partners follow. Free trade always increases the national income when
compared to tariffs. However, if we look at the politics of the strategic interaction
that takes place when large states and small states trade, we see that there are some
major drawbacks to the relationship for small states.

23R. J. Barry Joiies (1984) argues that one must iniclude the concept of time in discussing vulnierability
initerdepen-denice. I am arguin-g that one muLst include time in order to un-derstan-d the con-cept of sensitivity
iinterdepeindeince.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
342 The Complexity of Interdependence

Examining Figure 10 we see that small states do indeed possess a dominant

strategy of free trade. However, its large-state trading partner still possesses the nor-
mal dominant strategy for large states: tariffs. As a result, strong asymmetries in the
bargaining positions arise. A large proportion of the small state's payoff is under the
control of the large state (reflected in the small state's Fate Control: FCB = 2.5),
whereas the large state's payoff is free from influence by the small state (FCA = 0). In
addition, whereas the small state retains an incentive to coordinate its policies rela-
tive to the large state (BCB = .5), the large state faces no such restraint (BCA = 0).
Most important, this absence of FC and BC experienced by the large state can be
reversed only by changing the basic nature of the relationship between the large
state and the small state. The absence of FC and BC is a function of the small state's
inability to influence the payoffs received by the large state. As long as this remains
the case, the large state's payoffs will be solely a function of its own, independent
choice of strategy (i.e., pure RC, with no countervailing FC as in the case of large
state/large state interactions). Therefore, whereas economic theory correctly points
small states toward a dominant strategy of free trade, political analysis demonstrates
that large states, following a dominant strategy of tariff imposition, will exploit those
small states.
Comparing the decompositions of these two matrices (large state/large state vs.
large state/small state) demonstrates that the dependency felt by small states is not
just a component of interdependency; it is another form of political relationship.
Small states face a relationship with large states that is qualitatively different from
the ones large states face when trading with one another. This relationship cannot
be modeled by simply putting together the two separate sets of payoff from each
state. Large state/small state interactions generate a unique set of payoffs for the
actors involved.

The Study of International Interdependence

Offering a new method for analyzing an old issue requires some justification. A
reader should be given some compensation for the time invested in exploring that
new framework. This paper has argued that Kelley and Thibaut's approach to
analyzing strategic interactions passes muster because, in addition to providing a
mathematically elegant way of analyzing interdependence, it provides insight into
some of the questions that have vexed the interdependence literature in the field
of international politics: defining interdependence and outlining the theoretical
links between interdependence and political behavior.
We have seen that strategic interaction is much more complex than many
writers have argued. Within it are elements of independence (Reflexive Control),
dependence (Fate Control), and interdependence (Behavior Control). Each ele-
ment exists in its own right, measuring a unique element of variance in the total
relationship. The key definitional question for future research into such inter-
actions is not exclusion of concepts, but rather the inclusion of all three concepts.
In addition, by including all three concepts one also recognizes that power
remains an element of interaction, but that power by itself cannot fully describe all
relations within that interaction. Dependence (FC) is not the only element of
power in interdependence. There is also the power of independence (RC), and
the relationship between these two factors (the degree of concordance between FC
and RC) plays an important role in influencing the behavior of actors. There is also
interdependence (BC), an element that focuses on the benefits of coordination
rather than exchange. Such an element can play an important role either in
cooperation (e.g., the correspondence of a coordination game) or in conflict (e.g.,
the noncorrespondence of a zero-sum game).

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 343

Finally, we have seen that the concept of interdependence, by itself, is not neces-
sarily new to international politics. Historical examples of cooperative interde-
pendence and conflictive interdependence are easy to generate. Interdepen-
dence-as the term is used in the literature, referring to all strategic
interactions24-is not limited to economics. Mutual Assured Destruction is a prime
example of mutual dependence in the realm of security, one based on the
threatened use of force. If economic interdependence is to be a concept that
describes some new element of explanation for international politics, it will require
a better understanding of how economic forces affect the political choices
embedded in the three elements of interaction. The implication seems to be that
processes such as trade are dramatically increasing the amount of cooperative BC in
the matrix, but this is an empirical question that must be answered through
empirical research.
Figure 11 lays out these conclusions in diagrammatic form. Different forms of
interaction can lead to different forms of cooperation and conflict. Structures
defined by pure Reflexive Control or pure correspondent Behavior Control should
lead to cooperation automatically. In the first case there is no need for conflict
because each state can maximize its interests independently of the behavior of
other states; in the second case the mutuality of interests requires the simplest of
coordination before cooperation becomes the stable outcome. At the other
extreme, cases of pure noncorrespondent Behavior Control must lead to conflict;
both actors cannot maximize their benefits at the same time. In between is a range
of cases that may lead to cooperation under certain conditions. Two states
possessing mutual Fate Control may be able to deter each other from conflict,
assuming they are able to install infallible institutions of retaliation such as MAD.25
Other cases, such as the Prisoner's Dilemma of trade between large states, are
more complex. Here, two states face mutual Fate Control, which could lead to a
deterrent form of cooperation, but the two states face an additional problem: each
actor possesses a discordant element of Reflexive Control-FC can lead to free
trade, but RC leads them to closure. Cooperation, in such cases, can come about
only if states deter both other states and themselves from defection by instituting a
set of policies that fulfill both requirements.

F COOPERATION
AUTOMATIC

\4COOPERATION|
IIMPOSSIBLE |

FIG. II. Interdependence and International Cooperation

24As noted earlier, throughout this central section of the paper I have used the term interdependence to refer to one
element of the strategic relationships that can take place between actors. Indeed, the major theme of this paper has
been that the normal use of the term interdependence in the literature has been misleading because it mistakenly
conflates the elements of independence, dependence, and interdependence. However, since the remainder of this
paper is an analysis of that literature, I have reverted to this more common usage. To do otherwise would involve an
awkward rewording of the extant literature. Therefore, in the following discussion the term interdependence will again
refer to the broad relationship between states. Reflexive Con-trol, Fate Control, and Behavior Control will refer to the
components of that relationship.
251 am using the term institution here in the sense offered by Oran Young (1989:32). He defines institutions as a
set of social practices and conitrasts them with organizations, which are material enitities.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
344 The Complexity of Interdependence

This need for balance in self-restraint and other-restraint has appeared in

analyses both of the Prisoner's Dilemma and of recent trading systems. Taylor
(1976) and Axelrod (1984) argue that cooperation in the Prisoner's Dilemma
comes about only when policies of both contingent cooperation (i.e., threatened
retaliation against others) and low temptation (self-restraint) are in place.
Keohane's analysis of the GATT (1986) argues that free trade is possible only when
states follow policies of specific reciprocity (tit-for-tat retaliation) and diffuse
reciprocity (a general commitment to free trade). Such policies, he argues, create
the set of norms needed to maintain cooperation. Kroll (1992)-focusing on
interests rather than norms-argues that free trade systems collapse when states
abandon policies of contingent discrimination (a manipulation of other states'
payoff by bilateral tariff reductions and retaliatory tariffs) or systemic commitment
(a self-binding that comes from long-term treaties, avoidance of trading blocs, and
most-favored-nation clauses). Only when both of these policies are in place can
states create a set of national trade interests that form the type of cooperative
equilibrium described by Taylor and Axelrod.
The role of interdependence measures is to help define which of these structures
is present or evolving in international affairs. The relationship between a state and
the international market seems correlated to the concept of Reflexive Control.
Small states benefit from free trade, irrespective of the policies of other states in the
market; large states benefit from imposing an optimum tariff, irrespective of the
policies of other states. Each state possesses a dominant strategy. Vulnerability
measures tap the degree of Fate Control in the system: states that are highly
dependent on certain geographic markets can see their terms of trade manipulated
by states that control those markets. And finally, sensitivity measures may give us a
sense of the degree to which Behavior Control is emerging as states find their local
economies more and more integrated into the international market.
Unfortunately, when judged against this argument, most empirical works on
interdependence in the field of international relations are incomplete. Many studies
focus solely on developing a measure, or set of measures, that can reflect the level of
interdependence between or among states. Such works normally conclude by
stating that interdependence has gone up, gone down, or cycled through history
(Deutsch and Eckstein, 1961; Katzenstein, 1975). Others have taken this a step
further, assessing these levels of interdependence by comparing them to inter-
regional relations in what most scholars would accept as a highly interdependent
market, the United States. This approach ends with a conclusion on the degree to
which international politics has begun to resemble a highly integrated market
(Rosecrance et al., 1977). Each of these works leaves its central premise-increased
interdependence (or increased "positive interdependence") leads to increased
political cooperation-implicit and untested. Only a few works have gone on to
make that assumption explicit and to test it by looking at the empirical links
between economic flows and the degree of international cooperation and conflict
(Polachek, 1980; Gasiorowski and Polachek, 1982; Domke, 1988).
This paper takes issue with these works, on two grounds. First, whereas the
juxtaposition of competing definitions of interdependence against the historical
record may be valuable in assessing the degree to which interdependence is truly a
variable of international politics, our ultimate goal should be to link those measures
to a dependent variable of political behavior. If we limit ourselves to identifying the
patterns of interdependence, we will not be able to assess the value of such a
variable for explaining political behavior. Interdependence must be used to form
testable hypotheses by linking it to variables of cooperation and conflict. However,
the works cited above which do test this link between economics and behavior are
also incomplete.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 345

As noted above, there is no direct link that ties increased transactions between
states (e.g., trade, mail, phone calls) to changes in cooperation. Some transactions
may lead to conflict: for example, the creation of a dependent trading relationship
that allows one state to use its economic power to force another state to act against
its own interests. However, if that power is matched by the other state, the
relationship may remain relatively peaceful, with mutual benefits reached by
deterrence. Thus, there is no reason to expect a clear correlation between the
amount of trade between two states and the degree of cooperation exhibited in
their relationship. The issue of structure, and the manner in which states respond
to that structure, must be interposed.
For example, using the above analysis we can make a few tentative predictions
about the connection between changes in international trading relationships and
changes in international political relationships. Sharp rises in an actor's
independence (perhaps through the creation of a trading bloc) combined with
constant or declining levels of dependence could overwhelm the stabilizing effects
of such institutions as the GATT, leading to a collapse of free trade. The benefits of
protectionism would overwhelm whatever costs the members of GATT could
impose. On the other hand, a sharp decline in economic independence and
dependence (created by a transformation that left the world populated solely by
economically small states) would leave actors facing the kind of positive
interdependence that would most likely lead to spontaneous cooperation. To use
Keohane's lexicon (1984), these states would be moving toward harmony.
International trade institutions such as the GATT would become less necessary for
maintaining trade cooperation since the discordance in state trading relationships
would have declined.
Providing a foundation for such theoretically grounded empirical work has been
the goal of this paper. Kelley and Thibaut's work provides two major advances in
research in interdependence in international politics. First, it offers a framework
for incorporating the many definitions and measures of interdependence that exist
in the literature-thus providing a coherent set of independent variables to
examine. Second, it provides a theoretical explanation of those variables and their
relations to one another which, when combined with substantive issues such as
international trade, may provide the foundation for developing sound dependent
variables of predicted changes in state behavior.

References

AXELROD, R. (1984) The Evolution of Cooperation. New York: Basic Books.

BALDWIN, D. (1980) Interdependence and Power: A Conceptual Analysis. International Organization 34
(Autumn): 471-506.
BURTON,J. (1972) World Society. Cambridge: Cambridge University Press.
CASSING, J., T. McKEowN, AND J. OCHS (1986) The Political Economy of the Tariff Cycle. American
Political Science Review 80 (September): 843-862.
CAVES, R., AND R.JONES (1981) World Trade and Payments: An Introduction. Boston: Little, Brown.
CLARK, C., AND S. WELCH (1972) Western European Trade as a Measure of Integration: Untangling the
Interpretations. Journal of Conflict Resolution 16 (September) :363-382.
CONYBEARE, J. (1984) Public Goods, Prisoners' Dilemmas and the International Political Economy.
International Studies Quarterly 28 (March) :5-22.
COOPER, R. (1972) Economic Interdependence and Foreign Policy in the Seventies. World Politics 24
(January) :158-181.
DELL, E. (1987) The Politics of Interdependence. New York: St. Martin's Press.
DEUTSCH, K. (1978) Thle Analysis of International Relations, 2d ed. Englewood Cliffs, NJ: Prentice-Hall.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 346 The Complexity of Interdependence

DEUTSCH, K., AND A. ECKSTEIN (1961) National Industrialization and the Declining Share of the
International Economic Sector, 1890-1959. World Politics 13 (January):267-299.
DOMKE, W. (1988) War and the Changing Global System. New Haven: Yale University Press.
ELSTER, J. (1979) Ulysses and the Siren: Studies in Rationality and Irrationality. Cambridge: Cambridge
University Press.
GAsIOROWSKI, M., AND S. W. POLACHEK (1982) Conflict and Interdependence: East-West Trade and
Linkages in the Era of Detente. Journal of Conflict Resolution 26:709-729.
GOLDMAN, K., AND G. SJOSTEDT (EDS.) (1979) Power, Capabilities, Inzterdepenzdence: Problems in the Study of
International Influence. Newbury Park, CA: Sage.
HAMBURGER, H. (1979) Games as Models of Social Phenomena. San Francisco: W. H. Freeman.
HIRSCHMAN, A. 0. (1945) National Porver and the Structure of Foreign Trade. Berkeley: University of
California Press.
HUTTON, W. (1992) Exchange Rate Realignment Heralds Return to Mission Control. Guardian Weekly,
September 20, 6.
JERVIS, R. (1978a) Deterrence Theory Revisited. CISA Working Paper No. 14. UCLA: Center for
International and Strategic Affairs.
JERVIS, R. (1978b) Cooperation under the Security Dilemma. World Politics 30 (January) :167-214.
JOHNSON, H. (1953-54) Optimum Tariffs and Retaliation. Review of Economic Studies
21 (1953-1954):142-153.
JONES, R.J. BARRY (1984) "The Definition and Identification of Interdependence." In Interdependence on
Trial: Studies in the Theory and Reality of Contemporary Interdependence, edited by R. J. BarryJones a
P. Willetts, pp. 17-63. New York: St. Martin's Press.
JONES, R. J. BARRY, AND P. WILLETTS (EDS.) (1984) Intterdependence ont Trial: Studies in the 7Theory and Reality
of Contemporary Interdependence. New York: St. Martin's Press.
KATZENSTEIN, P. (1975) International Interdependence: Some Long-term Trends and Recent Changes.
International Organization 29 (Autumn): 1021-1034.
KELLEY, H. H., AND J. W. THIBAUT (1978) Interpersonal Relations: A Theory of Interdependence. New
Wiley.
KEOHANE, R. 0. (1984) After Hegemony: Cooperation and Discord in the World Political Economy. Princeton,
NJ: Princeton University Press.
KEOHANE, R. 0. (1986) Reciprocity in International Relations. International Organization
40 (Winter):1-27.
KEOHANE, R. O., AND J. S. NYE, JR. (1989) Power and Interdependence, 2d ed. Glenville, IL: Scott,
Foresman.
KROLL, J. (1992) Avenues of Commerce: Free Trade and Protectionism in International Politics. Unpublished
manuscript.
LITTLE, R., AND R. D. MCKINLAY (1978) Linkage-responsiveness and the Nation State: An Alternative
Conceptualization of Interdependence. British Journal of International Studies 4:209-225.
McNAMARA, R. (1987) Blundering into Disaster: Surviving the First Century of the Nuclear Age. New York:
Pantheon.
MORSE, E. (1970) The Transformation of Foreign Policies. World Politics 22 (April):371-392.
MORSE, E. (1972) "Transnational Economic Processes." In Transnational Relations and World Politics,
edited by R. Keohane andJ. Nye,Jr., pp. 23-47. Cambridge: Harvard University Press.
PANCHAMUKHI, V. R. (1961) A Theory of Optimum Tariff Policy. Indian Economic Journal
9(October) :178-198.
POLACHEK, S. W. (1980) Conflict and Trade. Journal of Conflict Resolution 24 (March): 55-78.
RAPOPORT, A., M. GUYER, AND D. GORDON (1976) The 2 X 2 Game. Ann Arbor: University of Michigan
Press.
REMAK,J. (1967) The Origins of World War I: 1871-1914. New York: Holt, Rinehart, and Winston.
RIEZMAN, R. (1978) Tariff Retaliation from a Strategic Viewpoint. Southern Economic Journal 48
(January) :583-593.
ROGOWSKI, R. (1989) Commerce and Coalition: How Trade Affects Domestic Political Alignments. Princeton,
NJ: Princeton University Press.
ROSECRANCE, R., AND A. STEIN (1973) Interdependence: Myth or Reality? World Politics
26(October):1-27.
ROSECRANCE, R., A. ALEXANDROFF, W. KOEHLER, J. KROLL, S. LAQUEUR, AND J. STOCKER (1977) Whither
Interdependence? International Organization 31 (Summer) :425-471.
ROSENAU, J., AND H. TROMP (EDS.) (1989) Interdependence and Conflict in World Politics. Brookfield, VT:
Avebury.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms
 JOHN A. KROLL 347

SCHELLING, T. (1960) The Strategy

SCITOVSKY, T. (1942) A Reconsideration of the Theory of Tariffs. Review of Economic Studies
9 (Summer):89-110.
STEIN, A. (forthcoming) "Governments, Economic Interdependence, and International Cooperation."
In Behavior, Society, and Nuclear War, vol. 3, edited by P. E. Tetlock, J. L. Husbands, R. Jervis, P. C.
Stern, and C. Tilly. New York: Oxford University Press.
TAYLOR, M. (1976) Anarchy and Cooperation. London: Wiley.
TAYLOR, M. (1978) The Possibility of Cooperation. Cambridge: Cambridge University Press.
THIBAUT,J. W., AND H. H. KELLEY (1959) The Social Psychology of Groups. New York: Wiley.
TOLLISON, R., AND T. WILLETT (1973) International Integration and the Interdependence of Economic
Variables. International Organization 27 (Spring) :255-271.
WALLERSTEIN, M. (1987) Unemployment, Collective Bargaining, and the Demand for Protection.
American Journal of Political Science 31 (November) :729-752.
WALTZ, K. (1970) "The Myth of Interdependence." In The International Corporation, edited by C.
Kindleberger, pp. 205-223. Cambridge: Harvard University Press.
WALTZ, K. (1979) Theory of International Politics. Reading, MA: Addison-Wesley.
WHITMAN, M. v. N. (1979) Reflections of Interdependence: Issues for Economic Theory and U.S. Policy.
Pittsburgh: University of Pittsburgh Press.
YOUNG, 0. (1968-69) Interdependencies in World Politics. InternationalJournal 24 (Autumn) :726-750.
YOUNG, 0. (1989) International Cooperation: Building Regimes for Natural Resources and the Environment.
Ithaca, NY. Cornell University Press.

This content downloaded from 170.239.168.67 on Sat, 01 Jun 2019 00:44:47 UTC
All use subject to https://about.jstor.org/terms