Review: Alliances, Balance, and Stability

Reviewed Work(s):
The Origins of Alliances. by Stephen M. Walt
The Balance of Power: Stability in International Systems. by Emerson M. S. Niou; Peter C.
Ordeshook; Gregory F. Rose
Glenn H. Snyder

International Organization, Vol. 45, No. 1. (Winter, 1991), pp. 121-142.

Stable URL:
http://links.jstor.org/sici?sici=0020-8183%28199124%2945%3A1%3C121%3AABAS%3E2.0.CO%3B2-H

International Organization is currently published by The MIT Press.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/mitpress.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.

http://www.jstor.org
Thu Aug 23 11:48:58 2007
Alliances, balance, and stability
Glenn H. Snyder

Stephen M. Walt. The Origins of Alliances. Ithaca, N.Y.: Cornell

University Press, 1987.
Emerson M. S. Niou, Peter C. Ordeshook, and Gregory F. Rose.
The Balance of Power: Stability in International Systems. New York:
Cambridge University Press, 1989.

Alliances and alignments are surely among the most central phenomena in
international politics. Yet we have no theory about them that remotely ap-
proaches the richness of our theories about war, crisis, deterrence, and other
manifestations of conflict. What might explain this? Perhaps it is their ubiq-
uity, given their informal as well as their formal manifestations, and con-
sequently the difficulty of isolating them as objects of study. George Liska,
in the opening sentences of his Nations in Alliance, which after nearly three
decades remains the only comprehensive theoretical treatment, put it this
way: "It is impossible to speak of international relations without referring
to alliances; the two often merge in all but name. For the same reason, it
has always been difficult to say much that is peculiar to alliances on the
plane of general analysis."' Thus, would-be alliance theorists, realizing that
they do not really want to write a general theory of international relations,
retreat to something manageable-another analysis of NATO, perhaps.
However, we do have various partial theories, each focusing on a partic-
ular aspect of alliances or approaching them from a distinctive perspective.
For example, the theory of collective goods was in vogue during the 1970s

1. George Liska, Nations in Alliance: The Limits oflnterdependence (Baltimore, Md.: Johns
Hopkins University Press, 1962), p. 3. For a comprehensive inventory and discussion of the-
oretical propositions in a more "behavioral" vein, see Ole R. Holsti, P. Terrence Hopmann,
and John D. Sullivan, Unity and Disintegration in International Alliances: Comparative Studies
(New York: Wiley, 1973).

International Organization 45, 1, Winter 1991

0 1991 by the World Peace Foundation and the Massachusetts Institute of Technology
122 International Organization

and early 1980s and became quite well developed as an explanation for why
the United States carried a larger share of the arms burden in NATO than
its European allies did. But this theory was rarely extended to other alliances,
particularly pre-bipolar alliances, or to other dimensions of alliance p ~ l i c y . ~
There have been occasional attempts to apply sociological coalition theory,
most interestingly to the U.S.-Soviet-China triangle.3The theory of "struc-
tural balance," which posits a tendency toward a positive product in triadic
relationships (three positives, or two negatives and a positive), has proven
quite fruitful in the analysis of pre-1914 alliance d i p l ~ m a c yN-person
.~ game
theory, apparently the most promising formal approach, has had little ap-
plication until r e ~ e n t l y There
.~ have been some interesting recent formal-
izations from a broadly "rational choice" per~pective.~ But these various
efforts do not cumulate; they remain partial and idiosyncratic.
In this article, I will try to suggest some issues and themes that a com-
prehensive theory of alliances ought to address, while commenting on two

2. The seminal article is Mancur Olson and Richard Zeckhauser's "An Economic Theory of
Alliances," Review of Economics and Statistics 48 (August 1966), pp. 266-79. For typical
examples of the elaborations and critiques that followed, see James Murdock and Todd Sandler,
"A Theoretical and Empirical Analysis of NATO," Journal of Conjict Resolution 26 (June
1982), pp. 237-65; and Bruce M. Russett and Harvey Starr, "Alliances and the Price of Pri-
macy," in Bruce M. Russett, ed., What Price Vigilance? The Burdens of National Defense
(New Haven, Conn.: Yale University Press, 1970), pp. 91-127. For applications of the theory
to pre-1939 systems, see Wallace Thies, "Alliances and Collective Goods: A Reappraisal,"
Journal of Conjict Resolution 31 (June 19871, pp. 298-332; and Barry Posen, The Sources of
Military Doctrine: France, Britain and Germany Between the Wars (Ithaca, N.Y.: Cornell
University Press, 1984).
3. See Dina Zinnes, "Coalition Theories and the Balance of Power," in S. Groennings, E. W.
Kelley, and M. Leiserson, eds., The Study of Coalition Behavior (New York: Holt, Rinehart
& Winston, 1970), pp. 351-69; and James C . Hsiung, "Sino-U.S.-Soviet Relations in a Triadic-
Game Perspective," in James C. Hsiung, ed., Beyond China's Independent Foreign Policy
(New York: Praeger, 1985), pp. 107-32.
4. For the original theory, see Fritz Heider, The Psychology of Interpersonal Relations (New
York: Wiley, 1958). See also Howard F. Taylor, Balance in Small Groups (New York: Van
Nostrand Reinhold, 1970). For applications to international relations, see Robert Jervis, "Sys-
tems Theories and Diplomatic History," in Paul Gordon Lauren, ed., Diplomacy (New York:
Free Press, 1979), pp. 212-45; Brian Healey and Arthur Stein, "The Balance of Power in
International History: Theory and Reality," Journal of Conjict Resolution 17 (March 1973),
pp. 33-61; and H. Brooke McDonald and Richard Rosecrance, "Alliance and Structural Balance
in the International System: A Reinterpretation," Journal of Conjict Resolution 29 (March
1985), pp. 57-83.
5. The seminal work applying n-person game theory to political science is William H . Riker's
The Theory of Political Coalitions (New Haven, Conn.: Yale University Press, 1962), but its
emphasis is on legislative rather than international coalitions. The pioneer in international
applications is R. Harrison Wagner, whose works include "The Theory of Games and the
Problem of International Cooperation," American Political Science Review 77 (June 1983), pp.
330-46; and "The Theory of Games and the Balance of Power," World Politics 38 (July 1986),
pp. 546-76.
6. For recent works in this genre, see Michael F . Altfeld, "The Decision to Ally: A Theory
and Test," Western Political Quarterly 37 (December 1984), pp. 523-44; James D. Morrow,
"On the Theoretical Bases of a Measure of National Risk Attitudes," International Studies
Quarterly 31 (December 1987), pp. 423-38; and Michael D. McGinnis, "A Rational Model of
Regional Rivalry," International Studies Quarterly 34 (March 1990), pp. 111-37.
 Alliances 123

recent heroic efforts, one in the traditional mode and the other in the formal
mode.
The first task in any theoretical investigation is to define the object of
analysis. What exactly is an "alliance"? Arnold Wolfers' definition is at-
tractive for its simplicity: an alliance is "a promise of mutual military as-
sistance between two or more sovereign states."' An alliance is a "promise,"
an explicit mutual declaration of future intent. This sets it apart from inten-
tions or expectations of assistance that may arise from sources other than
a promise (more about this presently); it also distinguishes alliance from the
physical act of assisting others, whether or not in fulfillment of a prior
promise. Alliances involve military collaboration. This immediately distin-
guishes them from all nonmilitary associations, ranging from economic car-
tels, such as the Organization of Petroleum Exporting Countries, to political
groups, such as the British Commonwealth. Further, they entail military
cooperation against particular other states, which differentiates them from
universal collective security organizations, such as the League of Nations
and the United Nations. Finally, the restriction to "sovereign states" rules
out government ties with nongovernmental entities, such as revolutionary
groups. The definition leaves open the issue of what form the "assistance"
will take; thus, it embraces not only offensive and defensive alliances but
also agreements not to use force, as in neutrality and nonaggression agree-
ments. A definition of this sort should be followed by a typology that sets
out in greater detail the possible variations on these basic forms and on other
characteristics such as the contingencies that will activate the promise.*
However, our theory must not be limited to formal alliances, thus defined.
What we really want to understand is the broader phenomenon of "align-
ment" of which explicit alliance is merely a subset. Alignment refers to the
expectations held by policymakers concerning the question "Who will de-
fend whom?" or, more broadly, "Who will support whom and who will
resist whom to what extent and in what contingencies?" Expectations of
support or opposition stem from a great variety of sources, most of which
can be crudely grouped into three categories: (1) strength inequalities, (2)
conflicts and common interests among states, and (3) past interaction, in-
cluding the negotiation of formal alliances.
Degrees of strength inequality form the basis for the familiar distinction
between bipolar and multipolar system structures. Each of these ideal types
logically generates different sorts of alignment expectations. In a bipolar
system, characterized by extreme inequality, the superpowers have clear
strategic interests in resisting each other's expansion. Consequently, the

7. Arnold Wolfers, "Alliances," in David L. Sills, ed., International Encyclopedia of the

Social Sciences (New York: Macmillan, 1968), p. 268.
8. The best typology currently available is Russett's. See Bruce Russett, "An Empirical
Typology of International Military Alliances," Midwest Journal of Political Science 15 (May
1971), pp. 262-89.
124 International Organization

lesser states can expect to be defended by one if attacked by the other-

unless, of course, they have already been conceded to the "sphere of influ-
ence" of their attacker. States in a multipolar system, by contrast, may lack
confidence of being supported if attacked because the strategic interests of
other states are ambiguous. The others will be uncertain whether an attack-
er's expansion is to their detriment or to their advantage if they have not
yet clearly identified the attacker as an enemy or as a potential ally. Even
if the others view the aggressive state as a threat to themselves, each may
be tempted to stand aside in the hope that someone else will provide the
collective good of resisting it.9 Thus, a multipolar system structure in which
capabilities are distributed evenly does not by itself imply any alignments.
However, moderate differences in capability may generate some alignment
expectations in a multipolar system. For example, two weaker states will
expect mutual support against a stronger state that threatens them both,
while two strong states with a weaker state between them will tend to be
rivals and protective of the weaker state against each other. Geographical
features, by their effects on offensive or defensive capabilities, may also
induce alignments.
Because of the relative weakness of structurally induced expectations,
alignments in a multipolar system will be largely a function of specific,
nonstructural conflicts and affinities-for example, territorial disputes or
ideological or ethnic attractions and repulsions. A state will expect some
degree of support from states with which it shares interests and will expect
opposition from others with which its interests are in conflict. These ex-
pectations will be weaker if the lines of conflict and commonality are different
on different issues; they will be stronger if these lines are mutually rein-
forcing. In turn, these expectations will be either reinforced or undermined
by the expectations that are generated by strength inequalities. The conflict
and commonality factor will combine with the strength inequality factor to
produce a pattern of alignment-an overall pattern of expectations regarding
support or opposition from others-that exists apart from any interaction
or behavior.
But alignments are also affected, of course, by the overt interaction of
states. Dispute settlements, supportive or opposing communications, eco-
nomic transactions, crises, wars, and other kinds of interaction, including
alliance agreements, will change expectations about who will support whom
in what contingencies and with what likelihood. The negotiation of a formal
alliance is simply an interaction episode that has such change as its main
purpose; for other interactions, it is usually secondary or incidental. Alli-
ances generally strengthen preexisting alignments by introducing elements
of precision, legal and moral obligation, and reciprocity. Thus, they usually
have greater effects on alignment than other interactions do; however, as in
the occasional case of a merely "paper" alliance, they may nbt. Some
9. See Kenneth N. Waltz, Theory oflnternational Politics (Reading, Mass.: Addison-Wesley,
1979), p. 165.
 Alliances 125

nonalliance interactions, such as the Anglo-French entente before World

War I, may have a greater effect on expectations than some formal alliances
do. Some alliances may run counter to the prior pattern of tacit alignment
and thereby transform that pattern.
A theory of alliances must give almost equal billing to relations of enmity.
To quote Liska again: "Alliances are against, and only derivately for, some-
one, or s ~ m e t h i n g . " ' ~
The broader concept of alignment should be under-
stood to include "alignment against" as well as "alignment with." Here we
begin to appreciate the problems of distinguishing allied relations from in-
ternational relations in general. Adversary relations provide the raison d'etre
for alliances and alignments; so much is obvious. Beyond this, conciliation
of an adversary can be an alternative to alliance against it; a state must
weigh whether one or the other yields the greater surplus of security gain
over cost. A state's cooperative moves toward an adversary may trigger an
ally's defection if the moves are interpreted as a prelude to the state's own
defection; but short of this, such moves may increase bargaining leverage
over the ally. Too much support of an allied state may embolden it too much
in confrontations with its adversary, but an attempt to restrain it may weaken
deterrence of the opponent. The greater the threat posed by an adversary,
the greater is the cohesion of the alliance. And the more the allies pull
together, the more threatening they appear to their opponents, and the more
the latter will likewise cohere. Alliance and adversary relations are a seam-
less web; any change in one will impinge on the other. Policy directed
primarily toward one of these realms must take account of its side effects
in the other.

Traditional theory

Stephen Walt's The Origins of Alliances does not completely fill the alliance
theory void, but it makes an impressive start. As the title suggests, his study
is principally about alliance formation rather than the politics of alliances
after they form. Walt asks two central questions: Why do states ally? And
how do states choose their friends? He then advances five hypotheses:
(1) states ally against states that threaten them-that is, they "balance";
(2) states ally with states that threaten them-that is, they "bandwagon";
(3) states choose allies of similar ideology; (4) foreign aid attracts allies;
and (5) political penetration facilitates alliance. In testing the hypotheses
against the history of alliances in the Middle East between 1955 and 1979,
Walt finds strong support for the balancing hypothesis, little for bandwa-
goning, modest support for ideological attraction, and little for aid and pen-
etration as causes of alliance.
Walt proposes to replace the traditional theory of "balance of power"
with a more comprehensive "balance of threat." Balance-of-power theory,
10. Liska, Nations in Alliance, p. 12.
126 International Organization

he argues, is too one-dimensional and thus distorting in its claim that states
ally in order to balance off the power of others. Actually, "states ally to
balance against threats rather than against power alone."" The level of threat
facing a state is a function not only of the distribution of power but also of
the geographic proximity, offensive capabilities, and, in particular, the per-
ceived intentions of others.
I wondered, on first reading, just how new this reformulation was. Was
it a "major revision," as the dust jacket announces, or was it only a "re-
finement," as the text describes it? The first three factors simply appear to
be components of "power" or "capability"; had we not recognized them
all along, at least implicitly? As for the fourth factor, intentions, had we not
always understood that balancing coalitions formed only (or usually) against
states that were apparently aggressive as well as powerful? A quick check
of several titles in the balance-of-power literature produced mixed results. l 2
Some authors describe the balancing process (when it is conceived as a
process rather than a static equilibrium) as something that is triggered by
the existence of "imperialist" or "revisionist" states. But others do imply
that the mere possession of great power will be enough to stimulate coun-
tervailing alliances. Hardly ever are powerful capabilities and expansionist
intentions explicitly separated as independent sources of threat. Nor, typi-
cally, are geographic factors introduced as either positive or negative mod-
ifiers of capabilities. Thus, in disaggregating these variables and then rein-
tegrating them in the concept of "threat," Walt has made a useful theoretical
clarification, one that brings balance-of-power theory more into line both
with common sense and with the historical record.
Walt has not quite buttoned up his revision, however. If what is balanced
is "threat," rather than "power," of what does the balancing consist? At
several points, he says that the more "dangerous" the threat (and he clearly
has in mind the "intentions" component of the threat), the more power will
be mobilized against it. Thus, the coalitions that formed against Germany
in World Wars I and I1 were far larger than necessary for "balance of power,"
and a large number of powerful states have coordinated their policies against
Libya, despite that state's modest capabilities. Threatening intentions are
therefore balanced by mobilizing extra power.I3 But this seems to ignore the
intentions factor on the defensive side. If threats are comprised of two broad
components-capabilities and intentions-might not the defensive reaction

11. Walt, The Origins of Alliances, p. 5.

12. The books checked were as follows: Hans J. Morgenthau, Politics Among Nations (New
York: Knopf, 1973); Inis L. Claude, Jr., Power and International Relations (New York: Random
House, 1962); Edward V. Gulick, Europe's Classical Balance of Power (New York: Norton,
1967); and Paul Seabury, ed., The Balance of Power (San Francisco: Chandler, 1995). Authors
who assert that a balance does form or should form merely against power sometihes warn that
a powerful state may develop aggressive intentions.
13. Walt, The Origins of Alliances, pp. 22 and 25.
 Alliances 127

to threats be similarly disaggregated? That is, would it not be reasonable to

postulate'both a balance of "capabilities" and a balance of "intentions" or
"resolve"? Then an adversary's aggressive inclinations might be balanced
by the defenders' firm resolve to resist. Or a defensive alliance that was
capable of blocking its adversary might not have achieved overall balance
if the adversary doubted the will of the allies to stand by each other in crisis
and war.I4 Thus, Britain and France failed to balance Germany at the time
of the Munich crisis even though their combined capabilities were equal or
superior to Germany's capabilities.
In Walt's theory, threatened states face a choice between balancing (al-
lying against) or bandwagoning (allying with) the threatening power. Bal-
ancing is the more frequent response because to bandwagon is to place undue
trust in a powerful state's benevolence. However, states occasionally join
an aggressor either to share in the spoils or to divert the aggressor's attack
away from themselves. They are most likely to do so if they are weak, if
strong allies are not available for balancing purposes, or if the threatening
state is believed to be appeasable. In the closing phase of a war, some states
may bandwagon with the apparent victor.
That states balance more than they bandwagon hardly comes as a surprise
to students of international politics, as Walt recognizes. He points out,
however, that policymakers often believe the opposite. German leaders,
before World War I, believed that "strength attracts" and thought that they
could coerce Britain into neutrality or even alliance with Germany. Soviet
leaders apparently believed that they could intimidate Norway and Denmark
into not joining NATO. U.S. policymakers have often believed that allies
would defect to the Communist side if they were to lose confidence in U.S.
strength and reliability. This mistaken belief, Walt asserts, has led to foreign
policy excesses-too much intervention, too much military spending, and
so forth. If the United States did less, its allies would do more for themselves,
since they prefer balancing to bandwagoning. One may agree (as I do) with
Walt's critique of U.S. policy and yet wonder whether the excesses should
be attributed to a belief that states in general are prone to bandwagon. Fears
that allies might defect (perhaps to the opponent) if insufficiently supported
have always afflicted policymakers in an anarchic international system; such
fears are not incompatible with a belief that balancing is the norm.I5
14. Walt approaches this formulation when he says that "an imbalance of threat occurs when
the most threatening state or coalition is significantly more dangerous than the second most
threatening coalition" (ibid., p. 265). This implies that threats are balanced when both sides
are equally "dangerous." But it seems odd to refer to a defender as "dangerous," and this
choice of terms fails to distinguish the resolve component from the capabilities component of
each side's threat.
15.Walt has clarified this in a subsequent article. The issue, he says, is the "relative propensity
for states to balance or bandwagon." If bandwagoning is the dominant tendency, alliance leaders
must "work very hard" to prevent defections. If balancing is preferred, less effort is required.
See Stephen M. Walt, "Testing Theories of Alliance Formation: The Case of Southwest Asia,"
International Organization 42 (Spring 1988), p. 283.
128 International Organization

The balancing-bandwagoning dichotomy is, in some ways, an unfortunate

vehicle for the analysis of alliance formation. First, states may "balance"
by means other than alliance-for example, by armament or military action.
Walt's analysis leaves the misleading impression that the balance-of-power
theory is entirely about alliance making. A fully developed theory of alliances
would have to deal with choices between alliances and other means to se-
curity, and for that purpose, the balance-of-power framework would be
appropriate, but Walt does not attempt that. Second, and more seriously,
the dichotomy obscures the full range of choices within the alliance realm
and inhibits a more discriminating analysis. States do have alignment options
other than allying with or against a threatening power. One is neutrality,
whether formal or informal. Another is to improve relations with third par-
ties, but short of alliance. Still another is to conciliate or compromise with
the threatening state.
When the threat is short of Hitlerian dimensions, reducing it by conciliation
is a legitimate and rational alternative to allying with or against it. It is the
preferable tactic when the threat is an illusory one of the kind that arises
out of a "security dilemma." The choice between conciliation and balancing
involves optimizing among security, autonomy, and intrinsic values: con-
ciliation buys security at some cost to intrinsic values; balancing buys it at
the cost of autonomy.16 Many combinations of balancing and conciliation
are conceivable, as diplomatic history makes abundantly clear. For example,
after making a colonial entente with Britain in 1907 that had anti-German
balancing overtones, Russia reassured the Germans by settling certain con-
flicts with them in 1910. The Anglo-Russian entente, like the previous
Anglo-French entente in 1904, illustrates how conciliation of one adversary
may amount to balancing against another. All such nuances are lost in a
theory that limits choices to balancing and bandwagoning."
Nor does this dichotomy tell us much about how states choose their allies
and shape their alliances when their motive is strictly balancing. At least in
a multipolar system, there may be several plausible partners. Logically, a
state will choose the one that provides the greatest surplus of security benefit
over cost. But the calculus by which it might compare alternative allies in
these terms cannot be generated by Walt's theory. Nor does the theory
specify the determinants of the kind of alliance the state will choose to form:
bilateral versus multilateral, unilateral guarantee versus reciprocal pledge,
alliance versus entente, clear versus ambiguous commitment, and so forth.
Nor can anything be deduced from the theory concerning how prospective

16. For a discussion of these trade-offs, see Morrow, "On the Theoretical Bases of a Measure
of National Risk Attitudes."
17. Walt has subsequently relaxed the dichotomy. Balancing and bandwagoning are ideal
types which actual behavior only approximates: "balancing against a potential threat does not
require unremitting hostility to it." States should "maximize security by aligning dith one side
while maintaining cordial relations with the other." See Walt, "Testing Theories of Alliance
Formation," p. 315.
 Alliances 129

partners divide up the alliance's benefits and what contributes to "bargaining

power" in this process.
The concept of bandwagoning is at once the most original and the most
elusive part of the theory advanced by Walt.I8 He differentiates two kinds:
offensive and defensive. Offensive bandwagoning is alignment with a dom-
inant state in order to share in the spoils of victory. Defensive bandwagoning
is a "form of appeasement"; a state aligns with an ascendant or aggressive
state in order to avoid being attacked. A variant of the latter is realignment
with the opponent out of disillusionment with a defensive alliance. In his
opening theoretical statement, Walt tries to restrict defensive bandwagoning
to appeasement via "alliance" or "alignment," but the elasticity of the latter
term leaves uncertain just what kinds of appeasement qualify. In his historical
examples and in later writings, Walt appears to expand the definition to
cover appeasing behavior not involving alliance, such as abstaining from an
alliance against the threatening state or offering the state large and unre-
ciprocated concessions under pressure and with a sense of fear and vulner-
ability. Negotiation of detente is not to be considered bandwagoning, nor is
maintaining "reasonably good relations" with a threatening state while ba-
sically balancing against it.I9 Apparently, Walt would not include appease-
ment at the expense of a third state; or, at any rate, he does not cite as an
example of bandwagoning the British and French appeasement of Germany
at the expense of Czechoslovakia during the 1930~.~O
Part of the problem is that the balancing-bandwagoning dualism cuts across
two other forms of dualism-offensive versus defensive alliances and re-
sistance to versus accommodation of an opponent-but it does not match
either one. Balancing alliances are clearly defensive, and some bandwagon-
ing alliances are offensive; an analysis from this perspective would place
bandwagoning in the context of offensive alliances in general. Walt does not
attempt this; apart from the bandwagoning type, offensive alliances are ab-
sent from his scheme. (It could hardly be otherwise once it is posited that
all alliances are responses to "threat.") Walt's analysis is closer to the
second broad dichotomy wherein balancing alliances would figure as a form
of resistance and defensive bandwagoning alliances as a form of accom-
modation. But there are, of course, many ways to accommodate an adversary
state besides alliance with it. Thus, Walt's focus on alliances blocks a full
analysis of the question of when states will resist (balance) their opponents
18. The concept was briefly discussed by Kenneth Waltz, Walt's teacher, in Theory of
International Politics, p. 126. Waltz credits Stephen Van Evera for the label.
19. Walt, "Testing Theories of Alliance Formation," p. 315.
20. Obviously, these states were not balancing either. Walt cites the Nazi-Soviet Pact and
the alignments of several East European states with Germany in the 1930s as examples of
bandwagoning. The diplomatic history of the 1930s does not support Walt's thesis that balancing
predominates over bandwagoning, although it was hardly a typical decade. See Robert G.
Kaufman, "Balancing and Bandwagoning Reconsidered: Alignment Decisions and Non-Deci-
sions Against Nazi Germany Between the Two World Wars," unpublished manuscript, n.d.,
Colgate University, Hamilton, N.Y.
130 International Organization

and when they will conciliate them. Those instances of conciliation, espe-
cially of major "appeasement," that have motivations similar to those of
the bandwagoning alliance ought logically to be labeled bandwagoning. But
this excludes the lesser forms and degrees of conciliation that are not sim-
ilarly motivated and that may play a role in the formation of alliances gen-
erally-for example, as preludes to alliance, bargaining ploys, means of
keeping options open, and hedges against excessive dependence on an ally.
Given Walt's "neorealist" orientation, it is odd that the variable of system
structure plays little role in his analysis. He apparently believes that his
generalizations are equally applicable to multipolar and bipolar systems,
although he seems to have the post-1945 bipolar system chiefly in mind.21
Greater sensitivity to structural differences would have enriched his theory.
For example, bandwagoning is logically more likely in a multipolar system
than in a bipolar one. Balancing is inhibited in a multipolar system because
of collective goods anxieties ("if I try to balance against this aggressor, will
others do the same?") and hopes ("somebody else will do the job"). Bal-
ancing may also be hindered by ambiguity about which state poses the
greatest threat. Bandwagoning is encouraged by the thought that there are
other targets toward which an aggressor's energies may be diverted and
there are other potential allies that a state may call upon for help in the event
that its aggressor-ally turns against it. In a bipolar system, bandwagoning is
less likely not only because of the virtual certainty that the superpower
protector will continue to balance off the superpower threat but also because
the threat itself is so unambiguous. Small states are less tempted to band-
wagon with the threat, since they need not fear being deserted if they balance
and since bandwagoning so clearly risks subservience.
Walt could have made use of such reasoning to explain why U.S. poli-
cymakers are excessively concerned about the reliability of U.S. allies: they
simply fail to appreciate the structural constraints on bandwagoning in a
bipolar system. In the new multipolar world toward which we are presently
drifting, such concerns may become more appropriate.
Nevertheless, Walt deserves credit for having initiated theorizing about a
phenomenon hitherto almost ignored. Further development would seek to
understand bandwagoning as a systemic process and to determine how it
relates to other such processes. Bandwagoning apparently is the alliance
equivalent of the domino theory in adversary relations.22 As Robert Jervis
has noted, both are theories of positive feedback or instability; a departure

21. Walt does recognize that the Middle East is a "multipolar, regional subsystem embedded
within a bipolar global system" and that this structural condition gives certain distinctive
characteristics to alliance formation in the area-notably a preference for superpower support
over alignment with regional actors. See Walt, The Origins of Alliances, p. 266.
22. Compare Deborah Welch Larson, "Bandwagon Images in American Foreign Policy:
Myth or Reality?" and Robert Jervis, "Domino Beliefs and Strategic Behavior," both in Robert
Jervis and Jack Snyder, eds., Dominoes and Bandwagons: Strategic Beliefs and Superpower
Competition in the Eurasian Rimland (New York: Oxford University Press, forthcoming).
 Alliances 131

from equilibrium begets further departures. The opposite of or the antidote

to both is balancing, a process of negative feedback that maintains or restores
equilibrium by reassuring allies and deterring opponents.23Either the positive
or the negative process may be dominant at any given time. Thus, early in
the career of an aggressive state, when it is still believed to be appeasable,
its bandwagon may gain passengers and dominoes may fall; but later, when
the aggressor appears bent on domination of the whole system, the balancing
tendency will prevail.
One may cavil about the global representativeness of alliances in the
Middle East and yet be enormously impressed by Walt's middle chapters
wherein the hypotheses are tested. They are a fine piece of theoretical-
historical analysis, richly informative as well as theoretically enlightening.
All of the hypotheses turn out pretty much as the author predicts in his
earlier theoretical commentary. For example, balancing during 1955-79 was
much more common than bandwagoning (only 12 percent of the alignments
were of the bandwagoning type, and they were fragile). And, as predicted,
states balanced against threats, not solely against power. In the Middle East,
however, states have been concerned almost entirely with regional threats.
They remain indifferent to the global balance because they can do little to
affect it, because the superpowers, being relatively distant, are not perceived
as serious threats, and because if one superpower were to develop aggressive
inclinations, it would be deterred by the other. When the Middle East states
ally with a superpower, they do so for protection against a local threat, while
the superpowers themselves are interested principally in balancing off each
other, either by opposing the other's regional clients or by enticing them
into realignment.
Ideology has only modest effects on the choice of allies, Walt finds. In
the Middle East, the effects have been minimal for local alliances but some-
what stronger for alliances with the superpowers. The difference is explained
by the general tendency of security considerations to dominate ideological
ones. Because locals feel most insecure vis-a-vis other locals, they are con-
strained to choose regional allies according to security interests rather than
ideological affinities when the two conflict. Since they do not feel threatened
by either superpower (partly because the superpowers deter each other),
they are free to align with the one that is most congenial ideologically.
Walt further challenges the conventional wisdom when he considers for-
eign aid and political penetration as instruments of alliance. The conventional
wisdom says that aid buys alignment or at least influence. Not so, says Walt:
an aid relationship develops as a result of previous alignment; it does not
cause alignment. Since small states choose their patrons on grounds other
than the prospect of aid, the giving of aid rarely yields significant leverage
to the patron. As evidence, Walt cites the Soviet and U.S. failure to garner
132 International Organization

influence with aid to Arab states. He notes that the United States has been
somewhat more.successfu1 with Israel.

Formal theory

Until recently, one of the most conspicuous gaps in alliance theorizing has
been the lack of formalization and, in particular, the absence of any sustained
attempt to apply n-person game theory. The explanation, I think, lies not
so much in the mathematical complexity of the theory, as is often supposed,
as in the difficulty of operationalizing the theory in international relations
terms. At first glance, this theory, with its components of coalition formation,
winning and losing coalitions, defections, bargaining, indeterminacy of part-
ners, and so forth, seems ready-made as a model for alliances, but the
appearance is deceptive. There is a considerable gap between the realities
of international life and the entities, assumptions, and workings of the theory;
a prerequisite to fruitful application is to close or at least narrow this gap.
Among the key concepts in n-person theory are those of "weights," "win-
ning," "coalition," and "payoff." "Weights" translate fairly straightfor-
wardly as "military power and potential." The other notions are more prob-
lematic. In the theory, a coalition wins when its aggregate weight is greater
than that of its complement-that is, greater than the aggregate weight of
all players outside the coalition. But international alliances begin to produce
value when they are far short of controlling over half of the power resources
in the system-presumably when they outweigh their most likely opponents,
and sometimes even short of that.
In the theory, "winning" occurs at some discrete point in time, when the
winners receive their payoffs, which are generally equivalent to the losses
of the losers-that is, the game is zero-sum. But in international relations,
there is no such clear decision point, nor is the game zero-sum. If payoffs
are conceived as security, they are experienced in a continuous stream over
time, a stream that expands and contracts with changes in the environment
or in relations within the alliance. And security gained by an alliance is not
necessarily equivalent to security lost by its opponents.
Even the notion of "coalition" does not have a clear-cut empirical ana-
logue. In cooperative game theory, whether or not a coalition exists is un-
ambiguous: either an agreement has been made or it has not; and if it has
been made, it will be fulfilled. In international life, however, the existence
of a coalition does not depend on formal agreement. Mutual expectations
of military support may follow from a variety of behaviors short of formal
alliance or merely from a coincidence of interests. And it is always uncertain
whether these expectations will be fulfilled, however they are generated.
Thus, the existence of a coalition is a probabilistic matter, or, in other words,
the value of an alliance must be discounted by a "loyalty coefficient."
 Alliances 133

A difficult concept to operationalize is that of "payoff." A peculiar char-

acteristic ~f international relations, absent from more highly organized con-
texts, is the interdependence of weights and payoffs. That is, not only do
the comparative weights of competing coalitions determine which one gains
and which one loses, but the gains and losses themselves consist largely of
changes in relative weights-that is, in relative power or capability. In short,
a considerable part of the payoff in each play of the alliance game is an
improved or reduced capacity to make gains in future plays.24 This feature
distinguishes international coalitions from legislative and other coalitions in
hierarchical societies in which a win or a loss does not generally change the
relative power of the players but pays off in intrinsic values. International
payoffs include power (strategic) values as well as intrinsic (ultimate) values,
so the calculus is complex. The most important payoff value is security,
which is a blend of strategic and intrinsic values. If security is defined as a
probability of continued enjoyment of values presently held, the "proba-
bility" is largely a function of power relations, while the "values presently
held" are intrinsic. Alliances and alignments may produce intrinsic values
other than security, such as prestige, domestic stability, and economic ben-
efits. Offensive alliances, of course, pay off, in addition, in anticipated or
actual gains by coercion or war.
Finally (although this certainly does not exhaust the problems that might
be listed), there is the problem of constructing a "characteristic functionH-
a list of aggregate values for all conceivable alliances-which is necessary
to make any n-person model "run." This is difficult empirically because the
values of alliance for individual states will always be to some degree incom-
mensurate and thus nonadditive. The obvious way to deal with this is to
devise some surrogate for alliance value which is commensurate and additive
across alliance members, but any such surrogate is bound to be quite un-
realistic.
Advancing fearlessly into this thicket, Emerson Niou, Peter Ordeshook,
and Gregory Rose have given us the first book-length attempt to apply
n-person game theory to international relations. It is more rigorous and more
broadly systemic than Walt's book, and alliances per se are somewhat less
central; philosophically, however, both books are squarely in the "realist"
tradition. While Walt offers a refinement of realist balance-of-power thinking,
Niou and his colleagues claim to have formalized it.
However, for these latter authors, "balance" does not mean equilibrium
of power or a process of approaching or returning to equilibrium. It means
"stability," of which there are two kinds: system stability and resource
stability. The central type is system stability, defined as a distribution of
such that no "essential" state or coalition of states can be
eliminated. Resource stability is a condition in which no state or coalition
24. Niou, Ordeshook, and Rose, The Balance of Power, p. 25.
25. The word "resources" is tacitly defined as "military power potential."
134 International Organization

has the incentive or means to alter the current distribution of resources.

Both kinds of stability exist simultaneously in only one case: when one state
controls exactly half of the resources in the system. In this case, the state
would not be permitted to expand further, since that would enable it to
eliminate all others; moreover, the others would not seek to expand at each
other's expense, since any conflict between them would tie up resources
necessary for collective resistance to the near-preponderant state. When no
state controls as much as half of the total resources, the system is resource-
unstable but may be system-stable: redistributions can occur, but the in-
dependence of every essential actor may still be guaranteed. System stability
is equivalent to the traditional notion of the preservation of the independence
of the great powers. However, resource stability is not quite equivalent to
the conventional idea of stability against war, since resource instability is
only a necessary, not a sufficient, condition for war.
The core of the theory is the part pertaining to system stability. It rests
on several assumptions, five of which are crucial: ( I ) resources are infinitely
divisible and freely transferable; (2) states try to maximize their resources,
consistent with survival; (3) states prefer to secure resources by negotiation
rather than by war (even though war is assumed to be costless); (4) all states'
resources grow at the same rate; and (5) states have complete information,
notably about one another's resource holdings. It follows, the authors assert,
that states will voluntarily cede resources to a stronger challenger or attacker
if necessary to avoid elimination and that, in turn, the challenger will accept
the resource transfers in preference to gaining the same amount by war.26
Thus, if one state is threatened by another state or by an aggressive coalition,
it can save itself by devising a "viable counter," of which there are two
types.
In the first type of counter, the victim simply transfers enough resources
to the threatening state or to the largest member of the threatening coalition
to give that state exactly half of the power resources in the system, thereby
making it "near-preponderant." This "freezes" the system: all the smaller
states will defend each other against challenge by the near-preponderant
one, and the latter is deterred. Thus, curiously, the principal means to pre-
vent a challenging state from achieving preponderance is to deliberately make
it just short of preponderant. Stability does not depend on the actual oc-
currence of the transfers but, rather, on the common knowledge that all
states have sufficient resources to "buy stability" in this way if necessary.
In the second type of counter, the victim threatens to form a coalition
stronger than the challenging state or coalition and perhaps including one

26. It is not clear why peaceful transfers are preferred over war gains, since wars are assumed
to be costless. Zero war costs are assumed in order to simplify the model; the assumption does
not affect the theoretical results. See Niou, Ordeshook, and Rose, The Balance of Power, p.
52.
 Alliances 135

or more members of the latter, with the goal of counterchallenging the re-
maining members. The turncoats will, of course, have to be offered more
than they expected in their first coalition to induce them to switch sides.
But then the remaining members of the first coalition can similarly induce
members of the second to switch, the second coalition can respond in kind,
and so on. No state can be offered more than half of the resources in the
system, since that would confer preponderance. The cycle ends when mem-
bers of a losing coalition, in order to avoid elimination, offer to transfer
enough resources to the largest of the winners to make it near-preponderant.
The offer will be accepted (again, "freezing" the system) because the ad-
vantaged state would rather make its gains peacefully than via war. Thus,
the two types of viable counter are not really independent; ultimately, system
stability depends on the feasibility of large voluntary resource transfers.27
Stated in game theory terms, system stability obtains when all actors are
"essential," meaning that each is an indispensable member of some "min-
imum winning coalition." In that case, each member either is able to mobilize
a viable countercoalition when threatened or is able, by itself or jointly with
others, to transfer sufficient resources to the threatener to give the latter
half of the total resources in the system.
The notion that a threatened state, in order to preserve its sovereignty,
would voluntarily transfer enough resources to its threatener to place the
latter on the brink of preponderance is the linchpin of the theory; therefore,
it must pass particularly severe muster. The issue is not so much whether
the transfer is physically feasible but whether a state would ever find it
desirable or necessary to make it. Niou, Ordeshook, and Rose point out that
such transfers are made all the time, but their examples-the granting of
spheres of influence, the payment of indemnities and reparations, the grant-
ing of trade concessions, and the transfer of territory as the consequence of
wars-fall well short of the scale implied by their theory, which is the
voluntary transfer of anything short of the entire territory of the homeland.
Of course, states sometimes do surrender large amounts of territory in the
face of overwhelming power, but generally they do so with the intention of
avoiding futile war costs, not with the intention of giving the aggressor half
of the total resources in the system and thereby obtaining the support of
others for the defense of some remnant of their territory. Moreover, a single
victim state, even if it were "essential," might not have enough territory to

27. One is tempted to suggest a parallel between the two types of viable counter and Walt's
choice between balancing and bandwagoning. The second type of counter, the countercoalition,
is crudely analogous to balancing. The ceding of resources has the flavor of the defensive form
of bandwagoning or simply of appeasement. However, there seem to be no examples in history
of an aggressor being appeased to the extent of giving it half of the power in the system. And
the motive for appeasement, in its conventional sense, is to moderate an aggressor's intentions,
not to increase its capabilities.
136 International Organization

grant near-preponderance to its aggressor, in which case it would have to

count on further transfers from friends not themselves under direct threat.28
The authors say that they need the assumption of large resource transfers
in order to portray a "frictionless systemM-that is, to get down to the
bedrock of first principles before introducing the complications of reality.
Certainly this is desirable in any theoretical enterprise, and theorists can be
forgiven for unrealistic assumptions if they produce useful results. After all,
we excuse Newton for assuming that objects unaffected by gravity would
fly off into space in a straight line, since it did help him devise the laws of
gravity. But is this assumption of large resource transfers similarly justified?
Is it necessary to the authors' theoretical argument?
Apparently, the authors thought it necessary for two related reasons. First,
some mechanism was needed for potentially realizing the ultimate condition
for stability-the possession of half of the system's resources by a single
state. The potential, not the actuality, was necessary and sufficient to guar-
antee the survival of essential actors. Second, the assumption was needed
to resolve a logical dilemma in applying a particular solution concept, the
"bargaining set." This concept, applied to a group of three maximizing
players, predicts a set of three feasible winning two-party coalitions with
the third party's payoff reduced to zero. Since a zero payoff means elimi-
nation in the international game, this is hardly a model of "stability." More-
over, if a winning coalition proposed to divide the system's resources fifty-
fifty, the third party would have no viable counter, since any successful bid
would grant preponderance to one of the others. Hence, the device of peace-
ful ceding of resources was introduced to permit the third party to save itself
(to "buy stability," as the authors put it) by buying off a member of the
threatening coalition at a price short of suicide. The logical fit between
bargaining-set theory and a desired theoretical outcome-stability-was im-
proved, but at considerable cost to empirical realism.
Theoretical assumptions must pass not only a usefulness test but also a
plausibility one: there must be no alternative assumption that is more plau-
sible and similarly useful. I believe there is such an alternative in this case.
Assume simply that states will defend a victim of attack if the success of
the attack would give the attacker a preponderance of resources. Or, in the
case of an aggressive coalition, assume that the allies of the largest aggressor
will defect and join the victim if it becomes apparent that otherwise the
largest state would gain preponderance. In either case, the attack would be
deterred by the anticipation of this outcome. Why would a victim give up a
large amount of its resources in order to persuade others to come to its
defense if it is strategically rational for others to do that anyway? The victim

28. For example, France's capitulation in 1940 did not give Germany near-preponderance,
nor would have the capitulation of all of Western Europe; concessions by the Uhited States,
the Soviet Union, or both would have been required to preserve remnants of territory and
sovereignty for the Western European countries.
 Alliances 137

could always hope that it would preserve more of its territory in allowing
others to defend it than in ceding whatever was required to preserve its
sovereignty. These others might prefer to restrict the aggressor to something
less than half of the total resources of the system; they could not assume,
as these authors do, that all states will grow at the same rate, and they might
prefer to avoid the eternal vigilance and constraint on their own policy that
would be entailed in permitting one state to be near-preponderant. This
scenario could be formally stated in a manner consistent with bargaining-
set logic, although it would require the addition of certain quite plausible
assumptions: that alliance bargains are not necessarily kept, that bargaining
continues during war, and that members of an attacking coalition are unequal
in strength or do not conquer resources at the same rate. Thus, realism
would be improved at some cost to parsimony-a desirable exchange in this
case.
Niou, Ordeshook, and Rose provide an empirical test of their model, using
great power alliance formation between 1871 and 1914 to "discern histori-
cally the forces it describes and the extent to which those forces appear to
operate despite the frictions characterizing reality."29 Their central hypoth-
esis is that states will form coalitions which are only large enough to win
and which maximize the ratio of the alliance's gain to its resources. An index
of resources for the six states involved is constructed from standard com-
ponents: production of coal, iron ore, pig iron, and steel; military-age male
population; mobilization efficiency; and force projection. The gain to any
winning coalition is operationalized as the amount of resources that a losing
coalition would have to transfer to the largest member of the winning co-
alition to give that member exactly half of the resources in the system. That
amount, then, is divided among the members of the alliance according to
their proportional resource contributions. This operationalization is, of course,
true to the initial statement of the theory and is therefore subject to the
anomalies mentioned earlier; it strains credibility to think of it as a plausible
measure of the value of alliances.
The "process" part of the theory-its description of alliance bargaining
and the preservation of stability-hardly shows up in the historical assess-
ment. The notion of "ceding," which is central to the model, is virtually
absent from the history. The only concrete reference is to the possibility of
Germany ceding Alsace-Lorraine to France in the 1880s to block a
Franco-Russian alliance, a concession which, the authors comment, "al-
most certainly would have led to the demise of the nascent German state."30
Although threats and counterthreats of alliance occasionally appear in the
history, they are not translated formally into bargaining-set dynamics.
Regarding outcomes, the results of the empirical assessment are hard to
appraise because there is no summary statement of correct and incorrect
29. Niou, Ordeshook, and Rose, The Balance of Power, p. 223.
30. Ibid., p. 262.
138 International Organization

predictions. Some of the alliances that actually formed were consistent with
theoretical expectations; others were not. In the latter cases, the authors
resort to a variety of selective ad hoc explanations, most of which appeal
to variables clearly present in reality but omitted from the model. Thus,
Germany's decision to join a "losing" alliance with Austria in 1879 rather
than a winning one with Britain is explained by Britain's lack of interest in
alliance with Germany at the time and by Anglo-German economic com-
petition. Germany's failure to choose the higher-paying Russian alliance over
alliance with Austria is explained by Germany's "mistrust" of Russia and
by its desire to keep Britain, Russia's enemy, unaligned. Certain other well-
known factors, such as Bismarck's belief that he could count on Britain in
any case because of its common interests with Austria and that he could
control Austria but not Russia, are not mentioned.
Most of these improvisations make reference to interests of or conflicts
between certain parties-obvious determinants of alliance behavior which
are practically ignored in the model. Interests appear in the operationalized
model only peripherally and crudely, as when certain alliance possibilities
such as Germany-France or Austria-Russia are labeled "infeasible" owing
to severe conflicts. Common interests and less-than-severe conflicts of in-
terest simply do not appear in the model; alliances are valued according to
their "winning" potential, regardless of commonalities or conflicts among
the allies or between them and their opponents. Thus, in being faithful to
the "power side" of realist theory, the authors' model slights the equally
central "interest side" and winds up being a caricature of realism rather
than a model of it.
Conceiving alliance values in terms of resources does capture the dis-
tinctive element of international politics mentioned earlier: that gains and
losses are counted more in terms of strategic values or relative power than
in terms of intrinsic values. It also provides a precise measure of value that
is commensurate and transferable across states. And, of course, strategic
values are at the center of the realist paradigm that the authors are attempting
to model. These are important strengths of their method. But these advan-
tages are gained at a considerable cost in verisimilitude. First, calculating
alliance payoffs strictly as potential resource gains at others' expense im-
plicitly casts alliances as offensive. Defensive alliances presumably are val-
ued chiefly for the "security" they provide for values presently held, not
for their capacity to acquire more, although a superiority in resources would
be conducive to security. The authors admit that their criterion of value
"may be irrelevant" to defensive alliance^,^' but they then proceed to test
their theory against a historical period in which most alliances were defensive
or primarily so. Consequently, the values they attribute to alliances between
1871 and 1914 can hardly be taken seriously.
Computing the security value of defensive alliances would require a fairly

31. Ibid., p. 234

 Alliances 139

complex calculus in which resource comparisons are only one component.

Security increments from alliance consist chiefly of a decrease in the prob-
ability of attack (deterrence), an increase in the probability of winning if
attacked (defense), and a reduction in the list of potential attackers (preclu-
sion). The amount of security gained by a state in each of these dimensions
is partly a function of the state's military resources compared with those of
its adversary. It is also a function of the degree of conflict with its adversary
(which bears on the probability of being attacked), the resources contributed
by its ally, and the reliability or "loyalty" of its ally. Crudely speaking, the
greater the shortfall between the state's resources and those of its adversary
and the greater the conflict of interest between the state and its adversary
(the greater the "threat," as Walt would put it), the more the state will gain
from alliance, assuming that its ally's capabilities do make up the shortfall.
From these gains must be subtracted the cost of the state's commitment to
its ally, which is a function (again, crudely) of the ally's resource deficit and
degree of conflict with its own adversary, the state's general loss of diplo-
matic freedom of action, and the state's cost of settling any conflicts it might
have with its ally. Still other factors could be added to either the benefit or
cost side-for example, the possible effects of the alliance in triggering future
events, such as the formation of a counteralliance, that would affect the
members' security. The resultant would be a "net security value," to which
would have to be added or subtracted various intrinsic values other than
security to get an aggregate value for each alliance member.
Now, admittedly, this fuller specification of alliance value would create
severe operational problems, which the method of Niou, Ordeshook, and
Rose avoids. Even if a measure of alliance payoffs in such terms could be
devised for individual states, there is the further problem of making payoffs
across alliance members commensurate so that aggregate alliance values
could be computed and transferability of values could be assumed. It is not
clear whether these problems can be solved. If they cannot be solved,
n-person game theory probably must be confined to a heuristic role. But
that is at least better than employing it empirically with an unrealistic in-
dicator of payoffs.
The authors' claim to have formalized realism is moderately justified. It
is not a complete formalization, as demonstrated by their ad hoc handling
of the concept of interests. However, most of their theoretical conclusions
(if not their route in reaching them) are congruent with traditional realist
thinking. No state is allowed to become preponderant; no essential state can
be eliminated; the goal of each state is to maximize power consistent with
survival; and the threat or anticipation of coalitions, rather than their actual
formation, is what maintains stability-all of these ideas resonate with tra-
ditional realism. On the other hand, curiously little (given the book's title)
is said about "equilibrium" or tendencies toward it. Indeed, the authors
assert that an equal or uniform distribution of power is largely irrelevant to
resource and system stability. Their emphasis on resource (power) max-
140 International Organization

imization, rather than security, and their conclusion that bipolar systems
are less stable than multipolar systems make their approach more an example
of old-style realism than of "neorealism."
Comparison of the two books under review is difficult because of their
different methodologies and purposes, but some interesting contrasts do
emerge. One obvious difference is that for Niou, Ordeshook, and Rose,
alliances are offensive and driven by greed; for Walt, they are chiefly de-
fensive and are driven by fear. The coauthors seem, again, to have been too
much influenced by the logic and dynamics of their model and not enough
by the world they are modeling. Walt's stance is the more realistic one. Walt
takes full account of conflicts and commonalities of interest, notably in
adding the factor of "intentions" to the "threat" against which states bal-
ance. Niou and his colleagues consider only "capabilities" ("resources")
in their model and introduce interests only verbally and partially. Both books,
to their credit, pay particular attention to geographical factors, but they
reach dismayingly different conclusions about them. For Walt, geographical
proximity is a source of threat. For Niou, Ordeshook, and Rose, proximity
is stabilizing, while distance is destabilizing: since resources have to be
discounted for distance, fewer resources have to be transferred to a nearby
aggressor than to a distant one to "buy stability." In their interesting chapter
on geography, the three have formalized the notion of "balancer" and the
special positions of "central" and "peripheral" powers. They have also
formalized the idea of preventive war. Thus, for these authors, alliances are
only one of several phenomena associated with the balance of power, while,
for Walt, alliances are the focus of analysis, and the balance of power is a
subsidiary theme. Niou and his colleagues are more systemic than Walt;
they approach alliances as a process central to systemic stability, whereas
Walt sees them principally as an instrument of statecraft.
So different in style, these books make different sorts of contributions.
Walt's work demonstrates that there is still room for theoretical progress in
the traditional realist mode. The contribution of Niou, Ordeshook, and Rose
is less in theoretical substance than in method.32It is the first comprehensive
formalization of realist balance-of-power theory, a valuable ground-breaking
effort despite its substantive flaws.

Toward a theory of alliances

Where do these two books leave the theory of alliances and the balance of
power? What remains to be done? Neither book is or claims to be a general
theory of alliances. Much is omitted from both that would have to be included

32. On the whole, the authors' mathematics are accessible to the novice.
 Alliances 141

in such a theory-for example, the alliance bargaining process, the sources

of intra-alliance influence, the benefits and costs of alliance, a typology of
alliances and alignments, the interplay between alliance and adversary re-
lations, the durability of alliances and the politics of maintaining them, the
causes of their dissolution, and the effects of domestic politics.
Further progress toward a comprehensive theory of alliances must come
to terms with Liska's warning cited earlier, the warning that it is hard to
isolate alliances and alignments, broadly conceived, from international re-
lations in general. No doubt he meant, in 1962, international security rela-
tions, but the thought is still forbidding. I believe the problem can be handled,
however, by a parsimonious ordering of the contexts of security cooperation.
There are two types of context. The "horizontal" one encompasses the
whole domain of security policy choice, ranging from armaments to alliances
to military action, and cuts across two salient fields of interaction, alliance
relations and adversary relations. In this interactive or behavioral realm, the
central research task is to understand the links between alliances and other
security "games." The "vertical" context, which is one of system structure,
interests, and domestic politics, generates the preferences and constraints
that enter into interaction choices.
A good deal can be accomplished deductively within each of these con-
texts. In the vertical context, for example, we can peel off several layers of
reality by assumption for the purpose of exposing "first principles" a la
Newton or Niou, and we can then replace them one by one to assess sec-
ondary and tertiary variables. Thus, we might look first at the incentives for
alliance in an anarchic but conflictless and power-egalitarian world, next at
the effects of power inequality, and then at the effects of each state's common
and conflicting interests with the others in the system. The exercise would
reveal the nature of "general" interests generated by structure alone, the
nature of "particular" interests that are in conflict or held in common be-
tween specific states, and the manner in which these different kinds of
interest affect alignments.33
In the horizontal context, we need to know a great deal more about the
relationships of power and interdependence between allies and how these
relations interplay with adversary relations. Each state faces an optimizing
problem that cuts across both sets of relations. Thus, in its relations with
an ally, it may seek an optimum trade-off between the risk of abandonment

33. 1 attempt such an exercise in "Alliance Theory: A Neorealist First Cut," Journal of
International Affairs 44 (Spring-Summer 1990), pp. 103-25. I have drawn on this article for
several other ideas in the present review essay; see pp. 103-5 and 108-9. The progressive
relaxation of assumptions is also part of the method of Niou and his colleagues. For instance,
they relax their initial assumption of equal rates of growth to develop their model of preventive
war.
142 International Organization

and the risk of e n t r a ~ m e n t but

, ~ ~the location of that optimum will depend
on the degree of threat it faces from its adversary. The greater the threat
(as Walt would define that term), the greater the risk of entrapment the state
will accept as the price of increased support from the ally; the smaller the
threat, the higher the risk of abandonment it will tolerate. In its relations
with an adversary, the state will try to strike an optimum balance between
security and other (nonsecurity) values. If, for example, the state wishes to
cede some nonsecurity values to its adversary in order to reduce conflict
and thus enhance security, it may be constrained in this by the fear that its
ally will interpret the concession as imminent defection, thereby causing the
ally to defect. Or if the state considers increasing security by tightening
alliance bonds, it may worry about becoming too dependent on the ally and
too vulnerable to entrapment. Clarification of all such trade-offs, which can
become quite complex, especially when armament and action choices are
thrown into the equation, can be handled best by formalized deduction. Here
the work of Niou and his collaborators blazes a useful trail.

34. For a discussion of these risks, see Glenn H. Snyder, "The Security Dilemma in Alliance
Politics," World Politics 36 (July 1984), pp. 461-96. Apparently, the concepts of entrapment
and abandonment were initially proposed by Michael Mandelbaum in The Nuclear Revolution:
International Politics Before and After Hiroshima (New York: Cambridge University Press,
1981), pp. 151-52.