Introduction: Research and Methodology
on arguments supported by anecdotes, the new data sets
allowed international relations scholars to systematically test,
revise, and retest their theories, a process that has ultimately
contributed to a great deal of accumulated knowledge about the
effects of alliance formation. The next section describes many of
the most influential of these studies.
Systematic Studies of Alliances
One of the first empirical tests of the alliance-war relationship
was Singer and Small’s (1966) examination of the behavior of
major states. They found that states ranking high on alliance
activity also rank high on war engagement, initiation, and bat-
tle deaths. This study provided the first systematic evidence
that alliances were associated with war, which contradicted the
traditional, realist balance of power literature. Following this
work, Wallace (1973) and Bueno de Mesquita (1975, 1978)
examined the effects polarization has had on war onset. Wal-
lace found that very high or very low levels of polarity in the
system produced war. Refining the indicators of polarity,
Bueno de Mesquita showed that the “tightness” of an alliance
system is not necessarily related to war onset. He found, how-
ever, that increases in the tightness of alliances are correlated
with war. Discrete alliances, on the other hand, are not associ-
ated with war.^2
Levy (1981) reformulated the questions about alliance for-
mation and war and with a limited extension of the Correlates
of War data found that, with the exception of the nineteenth
century, most alliances are followed by war. More significantly, a
relationship between power and alliance formation was found:
all so-called great power alliances in the sixteenth, seventeenth,
and twentieth centuries were followed by war within five years
(Levy 1981, 597–598, tab. 7).
Alliances between major states may also be linked to
increased probabilities of war in a second and less obvious way,
according to Vasquez (1993, 163–164). Schroeder (1976) has
discussed how major, “predatory” states may sign non-aggres-
sion pacts in order to remove potential adversaries before
attacking smaller states. Moul (1988, 34–35) shows that there is
an increased likelihood for unequal states to go to war in the
presence of a non-aggression pact with another state. These
pacts of restraint, or pacta de contrahendo,limit the expansion
of war but promote hostilities between unequals.
The studies by Moul (1988) and Schroeder (1976) are
important indicators that alliances can, in some instances, be
used as methods of limiting the expansion of war. In most cases,
however, alliances have been shown to be an important link in
the spread of war after war begins. For example, Siverson and
King (1979, 1980) have shown that participation in a multilat-
eral war is often determined by alliance ties. The presence of
alliances contributes to expanded wars as smaller states bring
their major state allies into their wars. A similar analysis links
alliances to the types of wars fought. Vasquez (1997) has shown
that wars among noncontiguous, rival states almost always hap-
pen in the presence of an alliance. The original territorial issue
that begins a conflict will drag noncontiguous allies into an
expanded war; the implication is that, had the alliance not
existed, the war was much more likely to remain bilateral.
Alliances may also have some unanticipated effects regarding
which side states choose to fight on in a coming war. A small lit-
erature exists on the probability of friends becoming foes.
Expected utility theory notes that it is sometimes rational for
allied states to end up in war with each other instead of against
a common foe (Bueno de Mesquita 1981). Most arguments
start between friends or acquaintances, and it follows that
alliances may increase the probability of war between states. Ray
(1990), however, has noted that this probability is quite small
and is dependent on both the definition of war and the defini-
tion of alliance (in both cases, dyadic or multilateral). Bremer
(1992) has seemingly ended this controversy by controlling for
the proximity of the allies. Noting that allies are more likely to
fight each other than unallied states, Bremer also finds that the
addition of contiguity into a multivariate analysis eliminates the
relationship between alliance ties and war. Although the logic of
opportunity is consistent with Bueno de Mesquita’s argument,
Bremer’s findings suggest that allies are becoming foes not
because of their alliances but because they are neighbors. Allies
that are not neighbors are not likely to go to war with each
other. Bremer (1992) did find, however, that allies that were
highly militarized had an increased chance of fighting each
other. This suggests that Bremer may have statistically isolated
the cases in which alliances promoted war between signatories.
Despite these results, the dangers of alliance formation
should be noted. Alliances generally increase rather than
decrease the chances of war against nonallied states, and even
Bremer’s research has shown that alliances can increase the
chances of war even against allies in certain situations. Both
these findings run counter to realist prescriptions advocating
the quickly formed alliance, and both underscore the dangerous
potential that alliances can have for naive policymakers.
The findings presented above, which show that alliances are
often associated with war, have been tempered, however, by
their inability to explain a large number of anomalous, peaceful
alliances. Nineteenth-century alliances, for example, were much
more likely to be followed by peace than by war. Maoz (2000)lii- Bueno de Mesquita has discounted these findings as a potential level
of analysis fallacy, that is, one or two states could be driving the sys-
temic polarization results even though the remaining states in the
system are moving away from polarization. Nevertheless, the polar-
ization studies seem to provide evidence that intense polarization is a
necessary systemic condition for large-magnitude wars (Vasquez
1993, 248–258). Although polarization is not necessarily associated
with the war proneness of single alliances, alliances formed during pe-
riods of intense polarization do have an increased chance of being fol-
lowed by large-magnitude wars. This interpretation is consistent with
the findings of Siverson and King (1979) and Sabrosky (1985) on war
expansion as well as Organski and Kugler’s analysis of world wars
(1980, 49–56).