Alliance Formation and Conflict Initiation: The Missing Link

Author(s): Anessa L. Kimball

Source: Journal of Peace Research, Vol. 43, No. 4, Special Issue on Alliances (Jul., 2006), pp.
371-389
Published by: Sage Publications, Ltd.
Stable URL: http://www.jstor.org/stable/27640349
Accessed: 15/05/2010 07:45

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. 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/action/showPublisher?publisherCode=sageltd.

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.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Peace
Research.

http://www.jstor.org
 ? 2006 Journal ofPeace Research,
vol. 43, no. 4, 2006, pp. 371-389
Sage Publications (London, Thousand Oaks, CA
andNew Delhi) http://jpr.sagepub.com
DOI 10.1177/0022343306064816
]k4^???\??^

Alliance Formation and Conflict Initiation:

The Missing Link*

ANESSA L. KIMBALL
de science L'Universit? Laval
D?partement politique,

Existing research on the connection between alliance formation and conflict initiation has explicitly
focused on the direct effect of alliances on conflict by including some measure of alliance behavior as
an independent variable in models of conflict behavior. Existing research misspecifies the relationship
between alliances and conflict, because alliance formation and conflict initiation are shaped by many
of the same factors (in particular, regime type and capabilities), and alliance formation decisions are
to conflict initiation decisions. Thus, alliance formation and conflict initiation should be
endogenous
modeled in a system of equations where a set of variables shapes alliance formation and conflict directly,
and indirectly affects conflict through the decision to ally. The author estimates a two-equation
probit
model that accounts for the endogenous nature of alliance formation decisions and, thus, for the indi
rect effects of variables like regime and power on conflict. Results suggest that the effect of regime on
alliance behavior differs across time periods. Finally, the model provides evidence that the total effects
of variables like power and regime on conflict are, in fact, mediated by how those variables influence
the decision to ally.

Introduction and, in fact, most scholars that the

agree
is The literature treats
relationship complex.
What is the nature of the relationship
alliances as a cause of conflict as if
potential
between dyadic alliance formation and
alliances arise from a divorced
process largely
conflict initiation? Answers to that question from the that leads to conflict. In
process
in research about the alliance
existing it is almost the case that the
fact, certainly
conflict relationship are complicated. One
decision to ally and the decision to fight arise
reason for the the
possible variety among from of the same factors. When
many causal
findings is misspecification of the relation
two states decide to the very that
ally, things
ship between alliance formation and conflict,
made them also the chances
ally shape they

* fight, but probably diminish those chances

Special thanks go to Dave Clark, Brett Ashley Leeds, for the very reason that committed to
they
Will Moore, Sarah Croco, Jacques Hymans, Bill Reed, and
an alliance. I an and
the anonymous reviewers for their helpful comments on develop argument
earlier versions of this manuscript. This project was pre model linking alliances and conflict in this
sented at the 2003 Annual Meeting of the Midwest Politi
fashion based on three claims: (1) alliance
cal Science Association, the 2004 Merriam Junior Master
Class in Formal Theory, and theWorld Politics Workshop formation decisions and conflict initiation
at Binghamton University. The data used here and a web decisions are not of one
independent
appendix are available at http://www.prio.no/jpr/datasets
and bingweb.binghamton.edu/-akimball. Correspondence:
another, (2) alliance formation and conflict
Anessa.Kimball@pol.ulaval.ca. initiation are caused some of the same
by

371
372 journal of Peace Research volume 43 I number 4 /july 2006

variables, and (3) those variables exert direct expected utility model deduced that allies
effects on the chances of conflict and indirect ought to be
more likely to engage in bilateral
effects through the decision to ally. Those conflict, because third parties will not inter
claims considered
together suggest
a differ vene and any movement in foreign policy by
ent theoretical of the relation allies from one another should increase
specification away

ship between alliance formation and conflict the initiator's expected utility for conflict.

initiation than has been explored previously Bueno de Mesquita's (1981) empirical
in the literature and, moreover, suggest that evidence supports that claim, but Ray (1990)
specification can also be linked to a specific provides persuasive evidence that the

econometric
technique permitting
an em
relationship found by Bueno de Mesquita is
pirical
test of my theoretical expectations.
sensitive to the exact definition of conflict
Existing
research on the alliance?conflict utilized and the cases included. After revising
nexus tends to focus on the direct effect of the data, Ray (1990: 80-82) finds some
alliances on conflict by the inclusion of some evidence of a
positive correlation between

of measure of alliance behavior into alliance ties and conflict proneness between
empiri
cal models of conflict behavior. Some of that dyads. However, he is careful to
point
out that

research that alliances increase the correlation does not causation (i.e. the
argues equal
chances of conflict, and some
argues that relationship might be spurious). Despite the
alliances decrease those chances. Further, attractive nature of the 'friends-as-foes' claim,
there are a number of studies whose findings which others noted earlier based on observed
or
implications
are indeterminate. After evidence (Holsti, Hopmann & Sullivan,
exploring
the existing research, I will return 1973; Sabrosky, 1980), Ray (1990) and
to the claims that I have made and argue that others (Vasquez, 2000) cast doubt upon it,
for the direct effects based on the fact that an initiator's
accounting exclusively positive
of alliances on conflict mischaracterizes their for war is a necessary
expected utility only
since alliance ties and conflict condition for conflict to occur, but not a
relationship,
arise from many of the same causes. I argue sufficient one. However, other scholars using
that the effects of variables like power and different sets of cases have uncovered a

are because between alliance and

regime complex they shape positive relationship
alliance formation and, through
that process, participation in conflict (e.g. Levy, 1981,
exert indirect effects on conflict. Thus, a 1983, for great-power dyads; Kim, 1989, for
proper econometric of the alliance coalitions). scholars examin
specification Finally,
is necessary to account for the the characteristics of alliances themselves
relationship ing
total effects of power and on conflict have also found a
regime positive relationship
behavior. between some alliance characteristics
specific
and an increased risk of conflict (Senese &

The Direct Effect of Alliance on Vasquez, 2004; Gibier, 1996, 2000; Siverson
& King, 1980). Therefore, the findings of
Conflict Is Positive
several research do to a
programs point
Some of the earliest empirical explorations positive relationship.
into the between alliances and
relationship
conflict uncovered a correlation
positive on
The Direct Effect of Alliance
between them (Singer & Small, 1966, 1968).
Conflict Is Negative
This supports the notion that states
finding
might form
alliances in anticipation of Among the first pieces of empirical research
conflict. Further, Bueno deMesquita's (1981) to assert a
negative relationship between
 Anessa L. Kimball Missing Link 373

alliance ties and conflict was Bremers 'Dan

ship in its recognition that both processes
gerous Dyads' (1992). Bremer presents might be caused by changes in capabilities
bivariate results a relation and that states can in either
showing positive engage process
between various of alliances and as a response to those (Waltz, 1979).
ship types changes
the chance of conflict between Likewise, scholars toward as
increasing point regime
states. However, his multivariate
findings
influential both in shaping choices about
indicate that the positive relationship conflict (Russett, 1993; Maoz & Russett,
between alliance ties and conflict reverses 1993) and choice about alliances (Lai &
direction after
accounting
for
contiguity Reiter, 2000; Simon & Gartzke, 1996;
between states. Since then, scholars have Siverson & Emmons, 1991). If these and
regularly controlled for the effect of alliance other factors contribute to alliance and
ties in dyadic models of conflict (recent conflict decisions, then
existing research mis

examples include Reed, 2000; Russett & specifies the relationship between alliances
Oneal, 2001; and Clark & Regan, 2003), and conflict and fails to capture the total

and results consistently suggest

a
negative effects of the factors that shape both
relationship. Additionally, intuitions about outcomes. The consequences of this type of
the of alliances based on can be substantial. Econo
consequences misspecification
formal theory (Leeds, 1999; Ritter, 2003) metrically, this misspecification is a form of
suggest that states form alliances either to omitted variable bias, which violates the
to each other or to overcome of a zero correlation between the
provide signals assumption
the commitment Game-theoretic variables and the error terms.
problem. independent
alliance formation models generally posit Subsequently, the estimated error term

that states engage in alliance formation to absorbs the effect of the excluded variable
information to other actors, and that (here the factors that the alliance for
convey shape
increase in available information about mation in parameter esti
process), resulting
adversaries to decrease their mates that are both biased and inconsistent
potential ought
risk of engaging in conflict.1 These findings (Long, 1997). Moreover, existing analyses
are consistent with the following general fail to examine the conflict and alliance
? are
claim if alliances formed to formation processes and, therefore,
prevent jointly,
conflict, then the direct effect of alliance on they fail to correctly capture the relationship
conflict to be between the
ought negative. processes. Unnecessarily
the alliance formation and conflict
assuming
to be not con
processes independent only
aMore Nuanced
Specifying strains models in undesirable but also
ways,
Relationship Between Alliance and
the theoretical to
ignores opportunity
Conflict Willingness
consider how the decisions to ally and to
One possible explanation for the divergent fight
are related. In
particular, allying
and

expectations and findings in the alliance fighting are both influenced by regime and
literature is that scholars have the Insofar as and
neglected capabilities. regime capabili
fact that the processes that alliances ties determine alliance formation, their
produce
and conflict are not different. effects on conflict are either exacerbated or
altogether
Realism points
to a more nuanced relation ameliorated by the decision to ally. By
the sources of conflict and alliance
treating
1 This is consistent with work by Reed that
(2003) formation most research examines
separately,
suggests that the information provided by bilateral trade sources on
the direct effects of those
ties between states leads dyads with more substantial trade only
ties to have a decreased risk of fighting one another. conflict.
374 journal of Peace Research volume 43 / number 41 july 2006

The argument I propose is unique insofar several ways, such as

examining the relation
as it includes elements from the between and alliance for
bargaining ship major powers
literature and contemporary research on the mation (Levy, 1981, 1983), how changes in
sent alliance across alliances conflict (Kim,
uncertainty-reducing signals by power shape
formation. A lack of information about 1989), and how changes in power might lead
relative capabilities
is a crucial factor in pre to alliance termination (Leeds, 2003; Leeds
dicting conflict behavior (Reiter, 2003), but & Savun, 2004; Morrow, 1991). Not only
the formation of an alliance provides crucial do capability levels convey information
information to the alliance about about the value that a state adds to an
partners
each other's
capabilities and preferences alliance, but they also help states to develop
(Bearce, Flanagan & Floros, 2006; Ritter, expectations of the utility of the alliance in
2003). In fact, the formation of an alliance the event that it is challenged.2 Relative
ought
to mediate
uncertainty between the capability levels inform potential members
alliance both about the benefits that accrue from alliance
partners surrounding power
and Because both power and formation. We consider relative capa
preferences. might
preferences, and the uncertainty
surrounding
bilities as being related to the trade-offs that
them, are known to be related to conflictual states make to alliances. As one
might join
behavior (Vasquez, 2000; Russett & Oneal, state becomes
increasingly preponderant
2001), the formation of an alliance provides within the alliance, the other state may be
revised information about alliance partners unwisely relying upon its partner for its
that was otherwise unavailable, and its
military capacity, suggesting it is incurring
absence could have increased conflict domestic risks as it trades
greater security

propensity between the partners. Thus, the

personal/internal security for allied security.
of information that states seek and Thus, as one state becomes more
type preponder
receive when form alliances is the same ant relative to the other, alliances should be
they
information that
shapes conflict behavior less likely to form, since states should prefer
between states, and, therefore, the decisions from
attaining security gains alignment
about both of those processes are linked. rather than developing security dependencies
This link between the conflict and alliance through alignment. Moreover, since states

processes not
only suggests
a more
complex with equal levels of capacity have greater
relationship than initially posited in the uncertainty regarding
unobserved military
literature, but also supports the
exploration capacity (Bueno de Mesquita, Morrow &
of different econometric
techniques that can Zorick, 1997), the formation of an alliance
account for the
complexity of the between them is a method through which
alliance-conflict In the each can obtain reliable information about
relationship. coming
sections, I discuss the claims in the extant the other's actual military capabilities.
literature about the sources of alliance for
HI: As relative increase (toward
mation and the sources of conflict. I then capabilities
the chances that the
those in an preponderance),
join explanations empirical an
dyad forms alliance decreases.
model designed to capture the total effects of
those causal variables. While the literature is in relative agree
ment about how alliance
capabilities shape
formation, there is little consensus on the
Claims About Alliance Formation
2 Studies on alliance
reliability are consistent with that
The relationship between military capacity claim (Leeds, Long & Mitchell, 2000; Leeds et al., 2002;
and alliance behavior has been explored in Sabrosky, 1980).
 Anessa L. Kimball Missing Link 375

role of regime in the decision to ally.3 Sig provide anecdotal evidence that states may

naling arguments developed in the past ally with partners with whom they share
decade suggest that pairs of states with some rivals.Moreover, Kim (1989) finds that
jointly similar regimes have a different alliance blocs often give
rise to rival alliance
to understand and the blocs, which increases the chances of conflict
capacity interpret

signals sent to each other (Fearon, 1994; occurring.

Schultz, 2001; Werner, 2000) because of
H4: As the number of rivals that a dyad
how democratic institutions shape beliefs shares increases, the
States that dyads willingness
about credibility. share regime to increase.
some ally will
type have convergent inter
similarity
ests ex ante as
they
are
subject
to the same I have discussed the existing claims within
domestic constraints that shape their the literature and suggested some hypotheti
to in either alliance forma cal about the influence of
capacity engage expectations
tion or conflict behavior.4 Thus, I posit that regime type and relative capabilities on
states with similar (i.e. autocratic alliance formation. Moreover, since alliance
regimes

dyads or democratic dyads) are more likely formation

shapes
conflict behavior, the total
to form alliances than are dissimilar
pairs of effect regime type and capabilities on
of
states. conflict behavior is composed of their direct
effects in the conflict equation plus their
H2: Joint democracy increases willingness
indirect effects from the alliance
to form equation,
alliances.
which can only be known if the processes are
H3: Joint autocracy increases to examined Thus, research has
willingness jointly. previous
form alliances. failed to account for the total effects of
regime and relative capabilities on conflict
Finally, I consider the influence of
behavior because of the misspecified
common interests on the of states
willingness between alliances and conflict.
to One measure of common interests is relationship
ally. a
The section presents series of
the number of rivals that a shares. following
dyad
total-effect propositions about the effect of
Rivals consistent threats to
represent security,
relative capabilities and regime type on
and to the extent that states have common
conflict behavior, given that those factors
rivals, have common interests with
they a states
also the chances that of
to issues and should be more shape pair
respect security an alliance.
form
to one The
willing ally with another. preva
lent alliance blocs that operated throughout
the Cold War (NATO and theWarsaw Pact) Conflict: Expectations of Total
3 Siverson & Emmons Effects
(1991) provide support for the idea
that democracies ally in numbers higher than expected by
The conflict literature iswell developed and
chance. Simon & Gartzke (1996) analyze the same data
and conclude that there is little correlation between alliance unified a set of factors
reasonably regarding
dyads and regime type. However, Lemke (2001) finds fault held to shape the chances of states fighting
with their argument. Lai & Reiter (2000) find evidence of
one another. This section will briefly develop
the flocking phenomenon only after 1945, while the
opposites-attract phenomenon existed mainly before 1945. and
expectations regarding regime, power,
Readers should keep in mind that the proportion of
status. The bulk of this section,
democracies in the state system has steadily increased over major-power
time, which may be a partial explanation for the divisive however, will focus on
developing expec
ness in the literature. tations about how those will influ
factors
4 For
examples of domestic constraints (on cooperation), ence
see Putnam (1988) and Milner the likelihood of conflict, given that
(1997); and (on conflict),
see Bueno de Mesquita et al. (1999, 2003). also the chances that states
they shape ally.
376 journal of Peace Research volume 431 number 41july 2006

The democratic peace literature (Russett, Accounting for the Distance Between
1993; Maoz & Russett, 1993) suggests that States

pairs of democracies behave more

peacefully Opportunities for states that are willing to
in than do other of states. There interact or are
pairs pairs cooperatively conflictually
is also some limited evidence of an autocratic shaped by the interaction possibilities for
peace as well (Raknerud & Hegre, 1997; states (Most & Starr, 1989). Proximity
Beer & 20025), between states influences the number of
Peceny, Sanchez-Terry, sug
that of autocracies interactions are to have; it may
gesting dyads composed they likely
are less conflict as well. Thus, I expect also to consolidate common interests or
prone help
that the direct effect of joint democracy (or exacerbate disputes.6 Gibier (1996, 2000)
joint autocracy) will be to decrease the points
out that a number of alliances actually
chances that a in conflict. I involve territorial settlement and that terri
dyad engages

expect the direct effect of dyads with at least torial settlement causes
unusually pacifistic
one to increase the chances that alliances. Thus, does supply a
major power contiguity
a dyad engages in conflict (Bremer, 1992, constant
opportunity
for alliance formation

2000), because major powers have both as well as for conflict. On average,
states that

interests and the are close to each other are more to

global capacity, increasing likely ally
chances that their interests collide with those because of constant so
opportunity, contigu
of other states.
Finally, within the dyad, each ity should increase a dyad's willingness to
state's assessment of its need to
fight
and its ally, while increasing distance between them
chances of winning a conflict are shaped by should decrease their willingness to ally.7
relative I argue that Since constant inter
capabilities. increasing contiguity provides
between states will increase their action for states, I expect it to
parity opportunities

willingness to fight (the chances of conflict), increase a

dyad's willingness
to engage in

while one state conflict as well. Since the direct effects of

increasing preponderance by
within the dyad reduces the likelihood of some of the factors of interest in both
conflict between a
pair
of states. Thus, the alliance behavior and conflict behavior have
direct effect of relative capabilities (moving been discussed, Iwill discuss the total effects
from to is to decrease of some of those same factors of interest on
parity preponderance)
the chances of a dyad engaging in conflict conflict behavior, because the total effects

(Clark & Regan, 2003). Most existing need not be the same as the direct effects.

research can those direct effects;

capture only
however, the model of conflict behavior that Total Effect of Relative Capabilities on
I present allows me to uncover the total
Conflict
effects of some variables of interest. Before The dyadic distribution of power is influen
specifying my expectations regarding
the tial in the formation of alliances and in the
total effects, I will discuss the of onset of conflict. States at are more
importance parity
for the influence of distance to form alliances than are states with
accounting likely
between states as an factor that and are states at
important divergent capabilities, parity
interstate behavior. 6
shapes Geographical proximity plays an important role in inter
state wars (Bremer, 1992, 2000; Vasquez, 1993), yet
Bremer (1992) found that the relationship between
alliances and war disappears when contiguity is added, and,
5 Reiter & Stam
(2003) re-analyze their models using therefore, proximity plays a key role in state behavior.
7
directed dyad data, and their results suggest that some of However, major powers may not be subject to distance
the findings of Peceny, Beer & Sanchez-Terry (2002) concerns because they are more likely to ally regardless of
change after accounting for which state initiated the distance, since they can project their capabilities and, in
conflict. turn, their interests (Levy, 1983).
 Anessa L. Kimball Missing Link 377

more likely to fight one another. That states Total Effect of Regime Type on Conflict
at parity aremore likely to fight is the finding Just as the effects of power on conflict might
most common in the literature on power and be mediated by the decision to ally, so might
conflict. However, when two states with the effects of regime similarity. The total effect
similar capabilities ally, the chances they will of joint democracy on conflict is actually
fight should decline, given the very fact that positive when those democracies are allied.
allied. Thus, alliance formation On its face, this result seems counter-intuitive
they may
well mediate the effect of on conflict. but makes sense the
parity actually considering rarity
Put another way, while parity might
makes of conflict and the particular rarity of conflict
states more likely to fight, their fighting is among democracies. Democratic
pairs
are

conditional on whether or not

they form an more likely to ally and less likely to fight, yet
alliance. in an alliance makes when do have contact
Investing they ally, they greater
that new ally less attractive and thus over shared interests than otherwise and thus
fighting
reduces the chances of conflict. have more to Stein
opportunity disagree.
That states at parity are less likely to fight (1990) suggests that pairs of states need both
because are allies a and conflictual interests, so
they might provide partial complementary

explanation for why the empirical findings that states have

something
to formalize their
on the distribution of power and conflict are cooperation around. Democratic conflict may
somewhat mixed. Because the effect of power be rare, as the democratic
extraordinarily

parity on conflict ismediated in dyads that peace literature demonstrates, but this finding
ally, models that test only the direct effect of suggests that the few times democracies

capabilities
on conflict, without
accounting actually do fight might well be because of their
for these indirect effects of power on alliance close as alliance
quarters partners. Democracy
formation, might
misestimate the chances of
might keep them at peace, but allying might
conflict. The total effect of power parity on make them quarrel.
conflict on whether or not states
depends Total Effect of Joint Democracy on Conflict:
ally. If states do not ally, then traditional
models conflict on an alliance
Joint democracy will have a positive total
regressing effect on conflict.
variable and a power variable probably
estimate the effect of power correctly. On the The total effect of joint autocracy on conflict
other hand, for states that do those to be as well. Autocratic
ally, ought positive pairs
models overestimate the effect of on of states are more to with one
parity likely ally
conflict, when they fail to account for the another and less likely to fight each other.
indirect effect of parity on conflict through However, similar to their democratic
the formation of an alliance. The total effect once autocratic states are
counterparts,
of on conflict, alliance forma allied, also have more to
parity given they opportunities
tion, should be negative. Contrast this to the
disagree with each other. Despite the
traditional expectation inmuch of the litera negative direct effect of alliance formation on

ture that parity will have a

positive effect on conflict, the total effect of joint autocracy is
the chances of conflict. Thus, I expect as with to conflict behavior,
positive respect
follows: because joint autocracy positively shapes
alliance formation and its effect on conflict

Total Effect of Capabilities: Parity will have a behavior8 is magnified by its positive
total effect on the of
negative probability 8 There as to the direction of the direct
is uncertainty
conflict. effect of joint autocracy on conflict, as dissension exists
within the literature.
378 journal of'Peace Research volume 431 number 41 july 2006

influence on alliance formation. Thus, variable from the same

arising explanatory
alliance formation mediates the relationship variables that influence conflict. Therefore, it
between regime similarity and conflict makes sense to estimate the
equations pre
behavior.
dicting conflict and alliance formation

Total Effect on simultaneously.

of foint Autocracy Conflict: Both of the outcome variables for these
Joint autocracy will have a total
positive are
processes responses, a
effect on conflict. binary suggesting
pair of seemingly unrelated probit equations
The expected total effect of regime and with correlated errors in a recursive system
relative capabilities on conflict behavior is is appropriate. Greene (2003: 715) specifies
different from the expected direct effect of the bivariate probit model9 below:
and relative on conflict
regime power
behavior for two reasons. First, alliance for =
Prob[^1 l,j2=l|X1>X2]
= u>
mation is
shaped by regime and relative capa <P2(X'l?l + yy2,X'2?2,P)
bilities as well, and, thus, the total
marginal
effect must account for the influence of this Thesecond dependent variable (y2)
indirect effect. Second, the direct effects of appears on the side of the first
right-hand
alliance formation on conflict behavior is equation, the model recursive. Since
making
negative. Thus, empirical models that this model is a full-information maximum

exclusively account for the direct effects of likelihood it allows for all
specification,
and relative on conflict combinations of the dependent vari
regime capabilities possible
behavior are not only misspecifying the ables (conflict [yx = 1], alliance [y2 = l];10
alliance-conflict
relationship, but, more conflict [yi = 1], no alliance [y2 = 0]; no
importantly, could lead to incorrect infer conflict [yi = 0], alliance [y2 = 1]; no conflict
= =
ences about the effects of those crucial vari
[yi 0], no alliance [y2 0]) as opposed to
ables on conflict. In the following section, I the censored probit (see Reed, 2000) or
suggest how this revised alliance?conflict other Heckman-style (selection) models
can be examined an where inclusion in the y2 is
relationship using equation depen
empirical model. dent upon yl being observed.11 These errors
are assumed to be distributed bivariate
normal (&2) and rho (p) measures; 'the cor
Integrating Expectations into an
relation between the disturbances of the
Empirical Model the omitted factors'
equations, (Greene,
Thus far, I have made the claim that alliance 2003: 717).
formation affects conflict but in an unusual The recursive bivariate probit model is
way; the factors that affect conflict also
shape consistent with the theoretical story that I
alliance formation and, thus, exert direct and have It allows the simultaneous
explored.
indirect effects on the chances of conflict. estimation of both initia
processes (conflict
This claim implies that alliance formation is tion and alliance formation) from
arising
endogenous to the variables generally held to
influence the likelihood so 9 This
of conflict, model is implemented in Stata 8.0 using the

decisions about conflict initiation 'biprobit' command.

originate 10 States do
engage in both behaviors in the same time
from within the of alliance forma
process period. In 1936, Ecuador and Peru signed a treaty demar
tion. Thus, alliance formation is an cating their border and, inDecember of the same year, they
explana a conflict over that border (Simmons, 1999).
began
tory variable shaping a dyad's willingness 11
Thus, while this is a simultaneous equation model, it is
to fight and, in addition, is a dependent not a selection model.
 Anessa L. Kimball Missing Link 379

some of the same

independent variables, non-directed politically relevant dyads by
regime type and relative capabilities in par year from 1816 until 2000.14 The dependent
ticular. Moreover, it allows for the inclusion variables are coded a 1 when
dichotomously;
of the influence of alliance formation on a an alliance is formed15 or conflict is initi
dyad's probability of conflict. The empirical ated16 based on the Correlates of War
model where y\ is a binary indicator of (COW) Alliance dataset (v.3.0) (Gibier &
conflict initiation and y2 is a binary indica Sarkees, 2004) and theMilitarized Interstate
tor of alliance formation is12 dataset (v.3.0), Most of
Dispute respectively.
the variables were
independent generated
P (conflict initiation) = a ? ?x relative using EUGene (v.3.04) (Bennett & Stam,
- - 2000). Variables generated from EUGene
capabilities ?2 joint democracy ?3 joint
+ + include capabilities, distance,
autocracy ?^ contiguity /J5major powers democracy,
? + ?? status, alliance indi
y2 alliance formation autocracy, major-power
cators, and conflict initiation, while the
data are from data
P (alliance formation) = a - ?x relative shared-rivalry computed

+ ?3 joint provided by Stinnett & Diehl (2001).

capabilities + ?2 joint democracy Relative capabilities are based on COW
? of distance + /?5
autocracy ?^ log CINC scores (Singer, Bremer & Stuckey,
+ /3? shared rivals + E?
contiguity
1972) and represent the ratio of the stronger
state in the dyad to the combined capabili
It is to notice that alliance forma
important ties of the two; the variable ranges from .5
tion in both as an inde
appears equations,
(parity) to 1 (preponderance). Distance is
pendent variable in the first and a dependent calculated using the distance between
variable in the second. This econometric
national If two states share a border,
capitals.
structure captures the nature of are
endogenous they coded 0; otherwise the inter-capital
the alliance variable, avoids simultaneity is calculated. uses
distance This work the
and will allow me to examine the
problems, natural of distance in an attempt
logarithm
direct, indirect, and total effects of power to normalize the measure, and if the distance
and regime on conflict behavior (Greene, between states is zero, then the variable is
2003: 716).13 This empirical structure is a coded 0. Major-power dyads are coded 1 if
very close of the theoretical
representation either state in the is a power
dyad major
claim that capabilities and regime shape to the COW
during the given year according
conflict behavior both directly and indirectly, list of major powers. Finally, shared-rivalry
through the choice to ally.

14 In order to maximize the number of observations, I set

Data and Methods the contiguity level to 5 so that dyads included in the
sample may have 151?400 miles of land or ocean separ
The data used in this came from a
analysis ating them, resulting in 94,163 possible politically relevant
of sources. The unit of is dyad years.
variety analysis 15Alliance formation is scored 1 when a new alliance is
formed or a different alliance is formed from the previous
alliance. The data do not differentiate between renewed
12 I also account for or new alliances of the same type as the previous alliance.
temporal dependence among the
observations through the use of time splines (Beck, Katz & Moreover, this variable accounts for any type of new
Tucker, 1998). alliance ? here, I do not differentiate between alliance
13 Bivariate
probit models, like this one, have been types, though that is an interesting project for the future.
16 Conflict
employed previously in the US politics literature; see initiation is scored 1when both states in the
specifically Zorn (2002) but also Martin & Wolbrecht dyad engage in a militarized interstate dispute where the
(2000). hostility levels for both states are greater than 3.
380 journal of PEACE RESEARCH volume 43 I number 4/July 2006

data were from data etc.) exert direct

computed rivalry (power, regime, simple,

provided by Stinnett & Diehl (2001), and effects on the chances of conflict. The

this variable is the sum of the number of implicit assumption this model (and those
shared rivals for a dyad in a given year. This like it in the literature) makes is that alliance
variable from 0 to 9, with a mean decisions are to conflict and to the
ranges exogenous
to .03 and a standard deviation of .26. other causes of conflict. The two
equal anticipated
Democracy and autocracy come from the
equation models reported next explicitly
Polity III data merged in EUGene (Jaggers relax that assumption.
& Gurr, 1995). Democracy is scored on a The second column of Table I reports esti

scale from 0 to 10, with 10 being the highest mates for a

simultaneously estimated, seem

score for The value is unrelated, bivariate model, to

democracy. autocracy ingly probit
scored on a scale of 0 to 10; it is a measure enable me to evaluate the extent to which

of the extent that the political institutions are alliance formation and conflict onset are

closed in a state.17 related, though I do not posit a direct

between the two behaviors
relationship
(since alliance formation is omitted from the
Results
conflict The estimates in the
equation).
Table I reports a single probit model of conflict equation
are consistent with my
conflict initiation including alliance forma expectations and with the
single-equation
probit model of conflict;
tion as an variable. This model
independent joint democracy
is, on the whole, representative of
existing and joint autocracy decrease the chances that

empirical models18 in the conflict literature, states

fight,
while major-power dyads have an

and the estimated

relationships
are consist increased willingness to fight. The signifi
ent with those in that cance of p that the omitted factors
commonly reported suggests
literature. between these two are different
processes
The effect of alliance formation on from zero, some evidence that this model is
conflict is negative and insignificant in the underspecified.
To
remedy
at least part of

model, and the direc the I estimate a model

single-equation probit underspecification,
tion of the coefficient is consistent with that including alliance formation in the conflict
of the literature allies are In so, I relax the exo
portion reporting equation. doing fully
less likely to fight. As is the case in most geneity assumption and allow alliance for

empirical work on conflict, the effect of mation to have a direct effect on conflict and
alliances is treated in this model as if it arises for the variables that influence both alliance
exogenously and as if the other variables formation and conflict to exert more

complex direct and indirect effects.

Bivariate Probit Model 2 reports parame
17 In
general, any country with a democracy score greater ter estimates for a model which accounts for
than 6 is considered a democracy, both institutionally and
with regard to personal liberties and participation. Simi the recursive or nature of the
endogenous
larly, any country with an autocracy score greater than 6 is between alliance formation and
considered a full autocracy. relationship
18This in its conflict decisions. These results
model differs from Bremers (1992) model willingness
exclusion of economic advancement and militarization as that alliance formation decisions
suggest
well as its inclusion of cases from 1966 to 2000. This
decrease to in
model varies from Clark & Regans (2003) conflict willing dyadic willingness engage
ness model in its exclusion of dyadic trade. Finally, this conflict when for the
accounting endoge
model differs from the models that Russett & Oneal nous nature of those decisions. This finding
(2001) report in its exclusion of trade dependence and
international as well as the that states that form alliances with
organization membership suggests
number of cases before 1886. one another are indeed less likely to fight
 Anessa L. Kimball Missing Link 381

Table I. Probit and Bivariate Probit Estimates

Bivariate 1 Bivariate 2
Single probit probit probit

Conflict willingness equation

Relative capabilities -0.845* -0.867* -0.885*
(.094) (.094) (.093)
Contiguity 0.353* 0.351* 0.391*
(.030) (.030) (.030)
Joint democracy -0.517* -0.514* -0.514*
(.066) (.066) (.065)
Joint autocracy -0.071* -0.071* -0.039

(.037) (.037) (.036)

Major-power dyad 0.501* 0.460* 0.533*
(.051) (.050) (.051)
Alliance formation -0.010 -0.548*
(.056) (.091)
Constant -0.240* -0.179* -0.112*

(.095) (.095) (.095)

Alliance formation equation
Relative capabilities -0.910 -0.916*
(.073) (.073)
Log of distance -0.532* -0.518*
(.018) (.019)
Contiguity -3.608* -3.508*
(.133) (.135)
Joint democracy -0.101* -0.098*
(.039) (.039)
Joint autocracy 0.124* 0.120*

(.025) (.025)
Sum of shared rivals 0.335* 0.302*
(.025) (.027)
Constant 4.228* 4.136*
(.148) (.150)
N 92,079 92,079 92,079
Chi-square (df) 3,250.45* (10) 9,372.99* (19) 9,786.82* (20)
Log-likelihood -5,814.26 -14,972.13 -14,937.11
P 0.210* (.029) 0.442* (.045)
Base probability 0.00397 NA 0.00081

*/><.0l.
Robust standard errors are reported in parentheses.
All models estimated with Beck, Katz & Tucker (1998) controls for temporal dependence.

each other, once we account for (1) the direct distribution results in a score of 70,19 exceed
effect that alliance formation has on conflict ing the value needed for significance (for 1 df)
(2) the simultaneous nature of at the .01 level and that the second
propensity, suggesting
the decisions, and (3) the link between the bivariate probit model is the better model.
two processes. A log-likelihood test of the
19 LLR tests = -2 - = 70.048.
bivariate probit models the chi-square ((-14972.135) (-14937.111))
using
382 journal of'Peace Research volume 431 number 41 july 2006

But the direct effects are only part of the are consistent with my expectations and the

story and are the part that has been democratic peace research program, the
explored
in the literature for some time. The indirect findings for joint autocracy further contrib
effects of the variables in the alliance forma ute to the research on a autocratic
potential
tion not the sources as well.
equation only represent peace21
of alliance formation, but also influence the I also report base probabilities for the
chances of conflict through their effects on models from Table II which include the
alliance formation. The estimates in the effect of The base
regime similarity. proba
alliance formation equation
indicate that bility of both outcomes (alliance formation
autocrats are more
likely
to
ally and that and conflict initiation) being observed in a
democrats are less likely to ally. This effect of given year when a dyad is composed of
joint democracy is inconsistent with my democratic states is about 2 in 10,000 before

expectations but is consistent with Simon & 1945, and it decreases to about 1 in amillion
Gartzke's (1996) claim that there is little cor after 1945. The same relationship is observed
relation between type and alliance for autocratic where chances
regime jointly dyads,

dyads. Table II, however, a more

go from 5 in 10,000 pre-1945 to 3 in a
provides
nuanced picture of the
relationship
between million post-1945. The base probability, for
regime type and alliance formation behavior democratic dyads, of a
pair of states
forming
across time. an alliance and not engaging in conflict is
II reports bivariate probit estimates
Table about 7 in 100 before 1945 and actually
comparing the pre-1945 time period to the decreases to about 4 in 100 after 1945,
years between 1945 and 2000.20 Interest whereas for autocratic dyads the chances of

ingly,
a
comparison of the results for joint alliance formation and peace before 1945 are
democracy suggests that democracies were about 4%, and they increase after 1945 to
less likely to ally before 1945 but more likely about 10%. Those probabilities suggest that
to do so after 1945, despite the aggregate states are far more likely to ally and be at
effect reported inTable I (Bivariate Probit 2) peace in any given year than ally and fight
being negative. Moreover, the results for each other. The risk of allying and fighting
joint autocracy are similar, as
joint autocracy was far higher for jointly similar dyads before
increased willingness to ally before 1945 and I945 than in the modern era. Moreover,

decreased willingness to ally after 1945. jointly

autocratic dyads, for the most part,
Taken
together,
these results indicate that appear to be twice as likely (or more) to
joint regime type significantly influences engage in those interstate behaviors relative

alliance formation but does so differently to jointly democratic dyads. Finally, the
across time These results of the coefficient on alliance for
periods. provide magnitude
support for Lai & Reiters (2000) claim that mation is larger in the post-1945 time
democracies do flock after 1945 but, period, suggesting that the pacifying effects
contrary to their results, these results also of alliance formation within the dyad are
suggest that autocracies flocked before 1945. different, upon the time domain.
depending
Moreover, across the bivariate models As is the case in most non-linear models,
probit
reported in both Tables I and II, joint regime these coefficients allow us to evaluate only
decreases a chances for direction and statistical rather
type uniformly dyad's significance,
conflict. While the joint democracy results
21 That the effect of joint autocracy on conflict behavior
loses significance across time periods provides some insight
20 I thank an anonymous for suggesting as to why the effect of joint autocracy exhibits instability
reviewer dis
aggregating the time domain. in other models of conflict behavior.
 Anessa L. Kimball Missing Link 383

Table IL Bivariate Probit Estimate Comparisons, 1816-1944 and 1945-2000

1816-1944 1945-2000

Conflict willingness equation

Relative capabilities -0.695* -0.839*
(.146) (.116)
Contiguity 0.224* 0.622*

(.042) (.041)
Joint democracy -0.665* -0.233*
(.117) (.073)
Joint autocracy -0.128* -0.022

(.059) (.043)
Major-power dyad 0.442* 0.451*
(.062) (.099)
Alliance formation -0.220* -1.750*
(.083) (.127)
Constant -0.090* -0.423*
(.144) (.120)
Alliance formation equation
Relative capabilities -1.200* -0.487*
(.110) (.115)
Log of distance -0.519* -0.313*
(.025) (.031)
Contiguity -3.306* -2.102*

(.175) (.246)
Joint democracy -0.327* 0.210*
(.066) (.041)
Joint autocracy 0.164* -0.078*
(.037) (.046)
Sum of shared rivals 0.274* 0.109*
(.034) (.078)
Constant 4.851* 1.252*
(.203) (.259)
N 36,570 55,509
Chi-square (df) 5,978.65* (20) 2,756.07* (20)
Log-likelihood -7,018.4512 -6,384.5347
P 0.290* (.050) 0.787* (.037)

Base probability =
P(l,l) (Jt. dem. 1) 0.00016 0.0000010
Base probability aut. = 0.00051 0.0000029
P(l,l) (Jt. 1)
Base probability =
P(0,l)t (Jt. dem. 1) 0.067 0.041
Base probability aut. = 0.036 0.106
P(0,l)t (Jt. 1)

*/><.0L
*where
P(0,1) is the base probability of forming an alliance and not fighting.
Robust standard errors are reported in parentheses.
All models estimated with Beck, Katz & Tucker (1998) controls for temporal dependence.

than the total effects these variables exert on conflict given the effects of variables in both
the chances of interstate conflict. The total equations. I computed in Table III total
effects essentially comprise the chances of
marginal effects (Christofides, Stengos &
384 journal of Peace Research volume 43 I number 41july 2006

Swindinsky, 1997) for the two variables of For dichotomous variables (like the joint
principal interest: capabilities and joint regime variables), the total effect accounts for

regime. the marginal effect when the variable equals

Table III reports marginal effects of 1, minus the
marginal
effect when the
in and on the variable 0. The variables
changes capabilities regime equals joint regime
likelihood of dyadic conflict. Total marginal and the major-power variable all have
effects for the bivariate probit model account positive marginal effects,23 ranging from .4
for the direct effect of a variable of interest, to .9. Thus, when a dyad is jointly demo
as well as the indirect effect of that variable cratic, its chances of conflict are increased by
through the endogenous variable (here, about 58% relative to the base probability of
alliance formation), using the joint probabil conflict for a mixed regime dyad, which is
equal to chances of about 8 in 10,000. The
ities on the chances of conflict. For a con

marginal effect of being a major-power dyad

tinuous variable such as relative capabilities
that appears in both the total is to increase the chances of conflict about
equations, by
marginal effect is the sum of the direct and 44% relative to the base probability of
the indirect effects and will 'account for the conflict. The total marginal effects reported
direct effect of a in that variable on for are consistent with
change regime similarity my
the probability that yY [conflict] equals one, expectations and point
to the critical conse

and . . . the indirect effect of the in of for only the direct

change quences accounting
this variable on
the probability that y2 effect of regime
on conflict.

[alliance formation] equals 1 in the equation For the

endogenous variable, alliance for

which, in turn, affects the probability that j/j mation, the total
marginal
effect of the for

equals one' (Greene, 1998: 298). Thus, the mation of an alliance between a
pair of
states

total marginal effect of changes in relative is to decrease their chances for conflict by
capabilities22 on the probability of dyadic about .4%. This provides support for the
conflict initiation is -0.002, that, claim that the effect of alliances on conflict
suggesting
as relative move toward to be Moreover, this
capabilities prepon ought negative. finding
derance, the effect on the chances for the claim that alliances
marginal provides support
of a dyad engaging in conflict are reduced by do cause pacifistic behavior between the
about .2%. However, the total and, thus, an answer to
marginal signatories provides
effect of an increase in relative on the research in this article.
capabilities question posited

reducing the chances for conflict for a dyad The connection between dyadic alliance
at is much about .5%, and that formation and conflict initiation is
parity larger, negative.

negative effect diminishes as one state I also compute marginal effects

Finally,
becomes increasingly preponderant in the for joint democracy and joint autocracy
dyad. This lends support to the work of based on the disaggregated models reported
Bueno de Morrow & Zorick
Mesquita,
who that the effect of 23 Some readers
(1997), suggest may be wondering why the marginal
increased is much more to con effect of joint democracy on the likelihood of conflict ini
capabilities tiation is positive even though the coefficients for joint
conflict behavior when states are
sequential democracy in both equations are negative. This is because
at These total the marginal effect when democracy equals 1 is the sum of
parity. reported marginal
the alliance formation (X'2?2)s minus the product of p
effects are consistent with the total effect
multiplied by the sum of the conflict initiation (X'^?^s.
for relative
expectation capabilities previ The product of p multiplied by the sum of the conflict ini
outlined. tiation (X\?{) is positive and larger than the negatively
ously
signed sum of the alliance formation (X2?2), resulting in
a positive
marginal effect on the chances of conflict initia
22 Relative are set at their mean, which is .89.
capabilities tion when joint democracy equals 1.
 Anessa L. Kimball Missing Link 385

Table III. Effects of Selected Variables on the Likelihood of Dyadic Conflict Initiation
Marginal Using
Bivariate Probit Model 2 (from Table I)

Percentage
Direct effect Plus Indirect effect Total effect change

Continuous variables
Relative capabilities (mean) -0.0024 0.0000 -0.0024 -0.2450
Power parity (.5) -0.0043 -0.0009 -0.0051 -0.5140
Power preponderance (.99) -0.0010 -0.0002 -0.0012 -0.1180

Endogenous variable
= 1 Minus = 0
whenj/2 whenjy2
Alliance formation 0.0008 0.0046 -0.0038 -0.3788

Dichotomous variables
when x = 1 Minus when x = 0

Joint democracy 0.5855 0.0008 0.5847 58.467

Joint autocracy 0.5885 0.0008 0.5877 58.765
Major-power dyad 0.4407 0.0008 0.4399 43.986

(FromTable II)
Joint democracy before 1945 0.2145 0.0001 0.2144 21.443

Joint democracy after 1945 0.9873 0.000001 0.9873 98.728

Joint autocracy before 1945 0.3249 0.0010 0.3239 32.390
Joint autocracy after 1945 0.9625 0.000003 0.9625 96.245

Base probability of both alliance formation and conflict initiation (P(l,l) =.000808

inTable II.While the marginal effects for all negative direct effects for some dichotomous
of the variables are the mag variables of interest, the total effect
regime positive, marginal
nitude of the effects is much larger in the can still increase the risk of conflict occurring
recent era. The marginal effect of jointly between states, because
marginal effects are

similar regime type appears to increase the computed using all of the estimated
base probability of alliance and conflict by parameters in the model, so that the sign
on

about 20-30% in the pre-1945 era, while the the sum of the (X'?)s may not retain the
effect for the post-1945 era is about three direction of the estimated coefficient.
times as While these are With reference to the
large.24 percentages hypothetical expec
in are evaluated tations, all of the variables as
large magnitude, they nearly perform
relative to increasing the base probability of expected, aside from the notable exception
both alliance formation and conflict for both for joint democracy. I discussed the potential
periods (reported at the bottom of Table II), explanations
for the unexpected result for

which is vanishingly small from a likelihood joint democracy by disaggregating the time
of single digits in the hundreds-of-thousands domain and showing
a
change
in the direc

(to the millions in the post-1945 period). tion of the variable across time. Thus, overall,
These marginal effects suggest that, despite the results reported
here are consistent with

my expectations for both the hypotheses and

24 The total effect for joint democracy (and joint autoc total effects. the best evidence that
Perhaps
racy) reported above in this table falls between the range of
research the relation
the total effects reported from the disaggregated time previous misspecifies
period. ship between alliance formation and conflict
386 journal of Peace Research volume 43 / number 41july 2006

initiation is that the indeterminate relation formation in the single probit model was

ship in the single probit model is corrected intedeterminate. However, linking

the

(to negative) when the relationship ismore decision processes and

positing
a direct effect

appropriately specified by accounting for all between them results in a

negative
relation
of the combinations of possible outcomes, ship between alliance formation and dyadic
via the seemingly unrelated bivariate probit willingness to engage in conflict. Thus, the
model that estimates both processes simul
missing link in existing research on the
taneously. Additionally, evidence of the relationship between alliance formation and

consequences of that misspecifica conflict initiation is the interconnection

potential
tion when one the total between these as
appear compares decisionmaking processes,

marginal
effects for the variables of interest
they are both shaped by regime and power,
(reported in Table III) to the direct effect as well as a
properly specified
econometric

expectations specified in the single probit technique

to capture the theoretical argument.
model of conflict. Finally, these results also Moreover, these results also provide
more

lend to claim that 'there are evidence to that the influence of

support Levy's suggest

clearly a variety of plausible theoretical regime type

on alliance formation may be
between alliances and war (or peace)' upon the years analyzed.
linkages dependent
(Levy, 1981: 584), and the results suggest that Further, the total marginal effects for
the connection between alliance formation
regime and power are shaped by all of the
and conflict initiation does depend upon the estimated parameters
as well as the correla

specification of the relationship and method tion between the error terms of each

of empirical analysis,
as well as the time
equation. The result of methodologically
under these claims is positive and
period analysis. incorporating
substantial total effects on the
marginal
probability of dyadic conflict initiation for
Conclusions
several variables such as
important regime
This project sought to determine whether similarity. Thus, the connection between
alliances cause
pacifistic behavior among the alliance formation and conflict initiation is
and, as such, it represents a first sensitive to the of the relation
signatories, specification
take at
analyzing
the influence of alliance ship. And the marginal effects of power and
formation on other interstate behaviors25 are reactive to both the sum of
by regime type

examining dyadic alliance and conflict the individual equation {X'?)s and the
behavior. The results reported here provide direction of p, suggesting that the interactive
evidence that the effects of alliance on nature of the is not
processes transparent.
conflict behavior may be misestimated or Moreover, it also suggests that alliance

subject to bias, if both the alliance formation behavior cannot be

accurately controlled' for

process and its effect on conflict initiation are in single probit models of conflict behavior
omitted from our
predictive models of by accounting for only the direct effects of
conflict initiation.26 The effect of alliance alliances, regime type, and power, since the
25 Gowa total effects of are in the
(1994) investigates the direct effect of alliance marginal regime
participation on bilateral trade behavior between signa direction to the direct effects.
opposite
tories. However, she does not account for the simultane
ous nature of the processes. Furthermore, I acknowledge the possibility
26 A web to this article (available at http:// of a direct between conflict
appendix relationship
www.prio.no/jpr/datasets and http://bingweb.binghamt.on. initiation and alliance formation in the same
edu/-akimball) replicates table A5.1 from Russett & Oneal
(2001 ) and finds support for the claim that these processes period, though I do not explore the direct
ought to be examined together. effect of conflict initiation on alliance
 Anessa L. Kimball Missing Link 387

formation here, because my claim is about Perception, and Escalation, American Political

the direct relationship between alliance for Science Review 91(1): 15-27.

mation and conflict initiation and the simul Bueno de Mesquita, Bruce; James D. Morrow,

taneous nature of those decision processes. It Randolph M. Siverson & Alastair Smith,
1999. An Institutional Explanation of the
is a direction for future research. Addition
Democratic Peace', American Political Science
ally, it is possible that the type of alliance Review 93(4): 791-807.
signed by states shapes conflict behavior Bueno de Bruce; Alastair Smith,
Mesquita,
(Bremer, 1992), and, while this project does M. Siverson & James D. Morrow,
Randolph
not differentiate between alliance type (e.g. The Logic of Political
2003. Survival. Cam
defense pact, offense pact, non-aggression MA: MIT Press.
bridge,
etc.), that is a that can be Louis Thansis &c Robert
pact, project Christofides, N.; Stengos
pursued in the future. Finally, the results of Swindinsky, 1997. 'On the Calculation of
these that alliance formation Effects in the Bivariate Probit
analyses suggest Marginal
increases behavior between Model', Economics Letters 54(2): 203-208.
pacifistic signa
claims of Bueno de Clark, David H. & Patrick M. Regan, 2003.
tories, the
refuting
to A Statistical Tech
Mesquita (1981), while supporting the 'Opportunities Fight:
nique for Modeling Unobservable Phenom
arguments posed by formal theory suggesting
ena', Journal of Conflict Resolution 47(1):
that the costs incurred though alliance for
94-115.
mation to deter states from
ought engaging Fearon, D., 1994. 'Domestic Political
James
in a costly conflict with one another.
Audiences and the Escalation of International

Disputes', American Political Science Review

References
90(3): 715-735.
Gibier, Douglas M., 1996. Alliances That Never

Bearce, David H.; Kristen M. & Balance: The Territorial Settlement Treaty',
Flanagan
Katharine M. Floros, 2006. 'Alliances, Conflict Management and Peace Science 16(2):
Internal Information, and Military Conflict 75-97.
Member States', International Gibier, M., 2000. Alliances: Why Some
Among Douglas
60(3): in press. Cause War and Why Others Cause Peace', in
Organization
Beck, Nathaniel; Jonathan N. Katz & Richard John A. Vasquez, ed., What Do We Know

Tucker, 1998. Time in About War? New York: Rowman & Littlefield
'Taking Seriously
Binary Time-Series?Cross-Section Analysis', (145-165).
American Journal of Political Science 42(4): Gibier, Douglas M. & Meredith Reid Sarkees,
1260-1288. 2004. 'Measuring Alliances: The Correlates of

Bennett, D. Scott & Allan Stam, 2000. EUGene: War Formal Interstate Alliance Dataset,
A Conceptual Manual', International Inter 1816-2000', Journal of Peace Research 41(2):
actions 26(2): 179-204. 211-222.

Bremer, Stuart, 1992. Dyads: Con Gowa, Joanne, 199]4. Allies, Adversaries, and Inter
'Dangerous
ditions the Likelihood of Interstate national Trade. Princeton, NJ: Princeton Uni
Affecting
War, 1816?1965', Journal Resolu versity Press.
of Conflict
tion 36(2): 309-341. Greene, William H., 1998. 'Gender Economics

Bremer, Stuart, 2000. 'Who Whom, Courses in Liberal ArtsColleges: Further

Fights
A. of Economic
When, Where, and Why?', in John Results', Journal Education 29(4):
ed., What Do We Know About War? 291-300.
Vasquez,
New York: Rowman & Littlefield (23-37). Greene, William H., 2003. Econometric Analysis,
Bueno de Mesquita, Bruce, 1981. The War 5th edn. Saddle River, NJ: Prentice
Trap. Upper
New Haven, CT: Yale University Press. Hall.
Bueno de Mesquita, Bruce; James D. Morrow Holsti, Ole; Terrence Hopmann & John D.
& Ethan R. Zorick, 1997. Sullivan, 1973. and Disintegration in
'Capabilities, Unity
388 journal of'Peace Research volume 43 /number 41july 2006

International Alliances: Studies. Levy, Jack S., 1983. War in the Modern Great
Comparative
New York: Wiley. Power 1495-1975. KY:
System, Lexington,
Keith & Ted R. Gurr, 1995. University Press of Kentucky.
Jaggers, 'Tracking
Democracy's Third Wave with the Polity III Long, J. Scott, 1997. Regression Models for Cate
Data', Journal Peace Research 32(4): and Limited Dependent Variables.
of gorical
469-482. Thousand Oaks, CA: Sage.

Kim, 1989. 'Power, Alliance, and Maoz, Zeev & Bruce Russett, 1993. 'Normative
Woosang,
Major Wars, 1816-1975', Journal of Conflict and Structural Causes of Peace', American
Resolution 33(2): 255-273. Political Science Review 87(3): 624-638.
Lai, Brian & Dan Reiter, 2000. 'Democracy, Martin, Andrew D. & Christina Wolbrecht,
Political Similarity, and International 2000. 'Partisanship and Pre-Floor Behavior:

Alliances', Journal of Conflict Resolution 44(2): The Equal Rights and School Prayer Amend
203-227. ments', Political Research Quarterly 53(4):
Leeds, Brett Ashley, 1999. 'Domestic Political 711-730.
Institutions, Credible Commitments, and Milner, Helen, 1997. Interests, Institutions, and
International American Journal Domestic Politics and Inter
Cooperation', Information:
of Political Science 43(4): 979-1002. national Relations. Princeton, NJ: Princeton

Leeds, Brett Ashley, 2003. Alliance Reliability in University Press.

Times of War: State Decisions to Morrow, James D., 1991. Alliances and Asym
Explaining
Violate Treaties', International metry: An Alternative to the
Organization Capability Aggre
57(4): 801-827. gation Model of Alliances', American Journal
Leeds, Brett Ashley & Burcu Savun, 2004. 'The of ^PoliticalScience 35(4): 904-933.
to Ter A. &
Way We Were:
Explaining Decisions Most, Benjamin Harvey Starr, 1989.
minate Alliances', paper at the and International Politics.
presented Inquiry, Logic,
Annual Meeting of the Midwest Political Columbia, SC: University of South Carolina
Science Association, Chicago, IL, 15-18 April Press.
://archive, allacademic. Mark; Caroline C. Beer & Shannon
(http com/publication/ Peceny,
Sanchez-Terry, 2002. 'Dictatorial Peace?',
getfile.php?file=docs/mpsa_proceeding/2004
American Political Science Review 96(1):
04-13/25107/mpsa_proceeding_25107.pdf&
PHPSESSID=a48f456ad4cccbf29e32baa21ef 15-26.

66106). Putnam, Robert D., 1988. 'Diplomacy and

Leeds, Brett Ashley; Andrew G. & Sara Domestic Politics: The of Two-Level
Long Logic
Mitchell, 2000. Games', International 42
McLaughlin 'Reevaluating Organization
Alliance Reliability: Specific Threats, Specific (Summer): 427-460.
Promises', Journal Resolution 44(5): Raknerud, Arvid & H?vard 1997. 'The
of Conflict Hegre,
686-699. Hazard of War', Journal of Peace Research

Leeds, Brett Ashley; Jeffrey M. Ritter, Sara 34(4): 385-404.

McLaughlin Mitchell & Andrew G. Long, Ray, James Lee, 1990. 'Friends as Foes: Inter
2002. Alliance Treaty and Pro national Conflict and Wars Between Formal
Obligations
visions, 1815?1944', International Interactions Allies', in Charles Gochman &cAlan Sabrosky,
28(3): 237-260. eds, Prisoners of War? Nation-States in the

Lemke, 2001. the Similarity Modern Era. Lexington, MA: Lexington

Douglas, 'Measuring
of States', paper at the annual (73-92).
presented
meeting of the American Political Science Reed,William, 2000. A Unified StatisticalModel
Association, San Francisco, CA, 30 August of Conflict Onset and Escalation', American
2 September. Journal ofPolitical Science 44(1): 84-93.
Levy, Jack S., 1981. Alliance Formation and War Reed, William, 2003. 'Information & Economic
Behavior: An Analysis of the Great Powers, Interdependence', Journal Resolu
of Conflict
1495-1975', Journal of Conflict Resolution tion 47 {I): 65-71.
24(4): 581-613. Reiter, Dan, 2003. 'Exploring the Bargaining
 Anessa L. Kimball Missing Link 389

Model of War,' on Politics 1(1): 1915-1945', in J. David ed., Quanti

Perspectives Singer,
27-43. tative International Politics: and
Insights
Reiter, Dan & Allan C. Stam, 2003. 'Identifying Evidence. New York: Free Press (247-288).
the Culprit: Democracy, and Singer, J. David; Stuart Bremer & John Stuckey,
Dictatorship,
Dispute Initiation', American Political Science 1972. 'Capability, Uncertainty, and Major
Review 97 (May): 333-337. Power War, 1820-1965', in Bruce Russett,
Ritter, Jeffrey, 2003. A Signaling Theory Model ed., Peace, War, and Numbers. New York: Sage
to Alliance Polities', paper presented (19-48).
Approach
at the Annual of the International Siverson, M. & Juliann Emmons,
Meeting Randolph
Studies Association, Portland, OR, 25 1991. 'Birds of a Feather: Democratic Politi

February-1 March. cal Systems and Alliance Choices in the Twen

Russett, Bruce, 1993. Grasping the Democratic tieth Century', Journal of Conflict Resolution
Peace: Principles for
a Post-Cold War World. 35(2): 285-306.
Princeton, NJ: Princeton University Press. Siverson, Randolph M. & Joel King, 1980.
Russett, Bruce & John Oneal, 2001. Attributes of National Alliance
Triangulat Membership
and and War American
ing Peace: Democracy, Interdependence, Participation, Journal of
International New York: Political Science 24 (February): 1-15.
Organizations.
Norton. Stein, Arthur A, 1990. Why Nations Cooperate?
Sabrosky, Alan Ned, 1980. 'Interstate Alliances: Ithaca, NY: Cornell University Press.
Their Reliability and the Expansion of War', Stinnett, & Paul F. Diehl, 2001. 'The
Douglas
to
in J. David Singer, ed., The Correlates of War Path(s) Rivalry: Behavioral and Structural
II: Testing Some Realpolitik Models. New York: Explanations of Rivalry Development',
Free Press (161-198). Journal ofPolitics 63(3): 717-740.
Schultz, Kenneth A., 2001. Democracy and Vasquez, John A., 1993. The War Puzzle. Cam
Coercive MA: Cam bridge, MA: University Press.
Diplomacy. Cambridge, Cambridge
bridge University Press. Vasquez, John A., ed., 2000. What Do We Know
Senese, Paul & John A. Vasquez, 2004. 'Alliances, About War? New York: Rowman & Littlefield.
Territorial Disputes, and the Probability of Waltz, Kenneth N., 1979.
Theory of International
War: for Interactions', in Paul E Diehl, Politics. New York: McGraw Hill.
Testing
ed., The Scourge War: New Extensions on an Werner, Suzanne, 2000. 'The Effects of Political
of
Old Problem. Ann Arbor, MI: of on the Onset of Militarized
University Similarity
Press (189-221). Disputes, 1816-1985', Political Research
Michigan
Simmons, Beth A., 1999. 'Territorial Disputes Quarterly 53(2): 343-374.
and Their Resolution: The Case of Ecuador Zorn, Christopher, 2002. 'US Government Liti
and Peaceworks in the Federal
Peru', Papers 27. Washington, gation Strategies Appellate
DC: United States Institute of Peace Courts', Political Research Quarterly 55(1):

(http://www.usip.org/pubs/peaceworks/pwks 145-166.

27.html).
Simon, Michael W. & Erik Gartzke, 1996.
'Political
System Similarity and the Choice of ANESSA L. KIMBALL, b. 1979, PhD in
Allies: Do Democracies Flock Together or Do Political Science (Binghamton University,
Attract?', Journal Resolu SUNY, May 2006); Assistant Professor of
Opposites of Conflict
tion 40(4): 617-635. Political Science, L'Universit? Laval (2006? ).
& Melvin 'National Current research interests: domestic
Singer, J. David Small, 1966. politics
Alliance Commitments and War Involvement, and alliance behavior; the distributional

1815-1945', Peace Research (Inter dilemma and foreign policy; non-traditional

Society
the of cooperative
national) Papers 5: 109-140. security strategies; design
& Melvin 1968. Alliance and bargaining.
Singer, J. David Small, institutions; negotiation
and the Onset of War,
Aggregation