Content uploaded by Emilie M. Hafner-Burton
Author content
All content in this area was uploaded by Emilie M. Hafner-Burton on Aug 06, 2014
Content may be subject to copyright.
War, Trade, and Distrust: Why Trade
Agreements Don’t Always Keep the Peace*
EMILIE M. HAFNER-BURTON
University of California, San Diego
ALEXANDER H. MONTGOMERY
Reed College
There is growing evidence that preferential trade agreements (PTAs) provide strong
institutional incentives to prevent international conflict among member states, often
creating the conditions of trust that can help prevent militarized aggression. We provide
an approach to the study of how international institutions influence conflict behavior
that considers how PTAs exclude as well as include members and create asymmetrical
relationships among members that could exacerbate conflict. PTAs do more than create
expectations of economic gains and reduce opportunism; they also create hierarchical
relations between states, which can encourage conflict under different conditions due to
distrust. We theorize these conditions for militarized international disputes, develop
appropriate measures using social network analysis, and test our expectations on new
PTA data during the period 1950 to 2000.
KEYWORDS: centrality; militarized disputes; preferential trade agreements; social
network analysis
Does trade interdependence promote peace among states? A long intellectual tradi-
tion suggests as much, as liberals have argued that the formation of global trade ties
and institutions creates incentives to settle disputes before they come to violence
(Angell, 1913; Doyle, 1997; Mitrany, 1966; Nye, 1971; Russett and Oneal, 2001).
Trade with foreign nations fosters economic interdependence among governments,
generates expectations about wealth gains, now and into the future, and may even
create a sense of community among nations (Deutsch, 1957; Gartzke, 1998; Haas,
1960; Polachek, 1980; Russett et al., 1998). International institutions like preferential
trade arrangements (PTAs) or the World Trade Organization (WTO) that organize
trade ties reinforce these peace-making processes. They provide a formal institu-
tional mechanism through which states lengthen the shadow of future trade
*
The authors share equal responsibility for this article and their names appear in alphabeti-
cal order. Please direct correspondence to both authors. The authors thank Zeev Maoz,
Charles Boehmer, and two anonymous reviewers for valuable comments on earlier drafts
of this article, as well as Jon Pevehouse and Edward Mansfield for sharing their data.
257
Conflict Management and Peace Science
ÓThe Author(s). 2012. Reprints and permissions:
http://www.sagepub.co.uk/journalsPermissions.nav
[DOI: 10.1177/0738894212443342]
Vol 29(3): 257–278
relations, credibly commit to economic interdependence, and cooperate to over-
come coordination problems. In the process, trade institutions increase trust through
creating strong expectations of future gains, enhancing political relations and preser-
ving peace among trade partners (Fernande
´z and Portes, 1998; Mansfield and
Pevehouse, 2000; Mansfield et al., 1999; Schiff and Winters, 1998).
Various scholars have contested this liberal argument (Baldwin, 1980; Barbieri,
1996; Buzan, 1984; Gilpin, 1981; Hirschman, 1945), yet most of the evidence avail-
able today suggests that the liberals might be right: economic interdependence can
promote collective security among trade partners, and trade institutions are central
to the peacekeeping process. Studies illustrate that trade partners are systemati-
cally less likely to resort to military conflict than states that do not trade (Gartzke
et al., 2001; Maoz, 2009; Oneal et al., 1996, 2003). Others show that when trade
becomes institutionalized through PTAs, members of the same arrangement are
even less likely to go to war than are other states and that PTAs are more effective
in preventing conflict as trade between members grows (Dembinski et al., 2004;
Mansfield and Pevehouse, 2000, 2003; Mansfield et al., 1999).
However, we have reason to be skeptical that mutual membership in trade institu-
tions always or primarily dampens conflict. While it is possible that institutional inter-
actions reveal or alter states’ interests to create trust among members, which may in
turn inhibit conflict, we suggest that this trust-building property only holds under
conditions of symmetrical relationships between states in institutions. Large asymme-
tries in these relationships, by contrast, can actually undermine trust between states
and thereby lead to additional militarized conflict in the international system. More
specifically, we argue that the distribution of ties in the network formed through
PTAs can lead to large asymmetries in dependence and influence, thus undermining
trust and promoting conflict.
Hafner-Burton and Montgomery have shown that neither international govern-
mental organizations (IGOs) nor PTAs simply provide a benign forum for state
interaction (Hafner-Burton and Montgomery, 2006, 2008, 2009). They also create
social networks that define hierarchies of social order in which states hold relative
positions of power over each other. Like military power (Waltz, 1979), states’
social power in this network of institutional memberships creates expectations for
behavior and gives states the ability to coerce, bribe, reward, or punish others,
defining the conditions under which acts of military aggression or cooperation are
rational strategies of action. These positions are emergent properties of the over-
all pattern of institutional memberships that define the social network among
states, and can just as easily provide incentives for conflict as for cooperation.
In this article, we draw upon the insights of social network analysis
1
to investi-
gate how states in the international system form relationships through their
1
For a general primer on social network analysis, see Scott, 2000; for a technical primer,
see Wasserman and Faust, 1994; on social network analysis (SNA) in international rela-
tions, see Hafner-Burton et al., 2009. For some prominent examples of SNA in IR, see
Dorussen and Ward, 2008; Elkins, 2009; Elkins et al., 2006; Hafner-Burton and
Montgomery, 2006, 2008; Ingram et al., 2005; Manger et al., 2008; Maoz, 2009, 2010; Maoz
et al., 2005, 2007; Ward and Hoff, 2007; Ward et al., 2007.
Conflict Management and Peace Science 29(3)
258
membership in international trade institutions, like PTAs and the WTO, and how
these relationships in turn create an international structure that can shape violent
conflict. In the realist view, mutual memberships in international institutions—in
social network analysis terms, ties—are epiphenomenal: they have no impact on
states’ propensity for conflict, which is only characterized by relative disparities in
military capabilities or trade gains (Mearsheimer, 1994). In the liberal view, ties
bind states positively: trade institutions produce symmetrical relationships of
amity that solve problems of relative gains that can lead to war through multiple
mechanisms. Neither assumption, however, accurately describes how trade institu-
tions create incentives for conflict. Institutional ties between actors shape conflict,
but they do not necessarily have positive effects.
Rather, ties place states into particular positions of power within a social net-
work. Social network analysis allows for conceptualization and measurement of
the emergent properties of networks such as institutions. As a conceptual frame-
work, network analysis moves from the individualist perspective, in which institu-
tional membership is considered to affect state behavior simply through
membership in individual institutions, to a perspective that emphasizes the emer-
gent properties of networks of institutions, whether from the perspective of the
direct network of states or from both direct and indirect ties as well. Rather than
measuring the effects of institutions through totaling up the number that a country
joins or the characteristics of individual institutions, network analysis measures
the collective effects of direct and indirect ties through institutions on state
behavior.
Here, we focus on two particular social positions—states’ direct sensitivity
dependence (hereafter dependence) on each other and their relative access central-
ity due to both direct and indirect ties in the social network formed by PTAs.
2
Conceptually, for any pair of states, the relative dependence of one state on
another increases with the number of joint memberships and decreases with the
number of non-joint memberships. States that have a large number of ties to other
states have high access centrality; this quality increases further for states that have
ties to other well-connected states. We define and describe both terms in detail in
the pages to come. We hypothesize that asymmetries in dependence or access cen-
trality increase conflict since they undermine trust between states.
In the following pages, we position our argument within the broader context of
debates over the effects of international organizations and economic interdepen-
dence on conflict through trust and develop our core propositions. We introduce
new data on state membership in trade institutions from 1950 to 2000 and derive
our principal indicators using social network tools. We use these tools to empiri-
cally identify and describe states’ structural positions within the social network
created by ties, and systematically test our argument using cross-national data on
interstate military disputes. In the process, we take great care to replicate previous
work where possible and to build on existing scholarship.
2
On different forms of dependence, see Maoz, 2009, 2010. On access centrality, see
Hafner-Burton et al., 2009.
Hafner-Burton & Montgomery: War, Trade, and Distrust
259
How Institutions Can Lead to Trust
Since the time of Immanuel Kant, scholars of international politics have proposed
that peace is possible through the forging of an international social structure based
on liberal principles of rights, obligations, and law (Kant, [1795] 1991). In the mod-
ern world, international institutions and economic interdependence create this
social structure through the establishment of authority, providing a foundation for
the building of trust among nations in an anarchic world and the diffusion of the
value of cooperation (Angell, 1913; Zimmern, 1936). Europe has been an impor-
tant example. Looking to Europe, integration theorists have argued that economic
integration between nations can, over time, create a communal understanding and
interest in collective security, as learned forms of cooperation in markets spill over
into areas of higher security politics (Haas, 1971; Mitrany, 1966, 1976; Nye, 1971).
PTAs, like the European Community, can merge the interests of age-old rivals by
establishing a community of states and peoples long divided (Wallace, 1994).
While integration theorists looked toward the experience of Europe, Karl
Deutsch proposed the more general concept of a security community, where eco-
nomic integration among liberal states could create a sense of community based on
trust and reciprocity (Deutsch, 1957). A pluralistic security community forms
among members that hold similar core values grounded in common institutions
and mutual responsiveness, creating a shared identity, loyalty, and sense of ‘we-
ness’ among states (Adler and Crawford, 2002). Building from Deutsch, Adler and
Barnett argued many years later that this sense of community among states is not
simply a European phenomenon, but can and does exist internationally (Adler and
Barnett, 1998). States within this community have the ability to cultivate peaceful
methods of interacting, although they do not always succeed or choose to do so.
The underlying theme of both theoretical traditions has been adopted by scho-
lars seeking to provide evidence that international institutions in general and trade
institutions in particular assuage conflict by generating trust and mutual identifica-
tion and proffering norms of peace-making among states with common institutional
ties. In their influential work on the democratic or ‘‘Kantian’’ peace, Russett and
Oneal argue that shared memberships in intergovernmental organizations (IGOs)
increase cooperation among nations and reduce their propensity to go to war
(Oneal and Russett, 1999; Russett and Oneal, 2001). Among the various means
through which IGOs can have a pacifying effect is the generation of narratives of
mutual identification through the construction of mutual identity and the legitima-
tion of norms developed within the organizations that facilitate cooperation and
create common interests. Bearce extends this line of argument to commercial insti-
tutions and suggests that the key to generating trust is improved contact between
potential rivals (Bearce, 2003). Trade institutions, accordingly, can promote peace
by bringing government decision makers together in repeated interactions that can
generate bonds of trust
3
among potential adversaries and reduce problems of
3
Schiff and Winters (1998) similarly assume that trade between neighboring states creates
trust and reduces the likelihood of conflict and that international organizations can serve a
trust-building function. See also Schiff and Winters (2002).
Conflict Management and Peace Science 29(3)
260
misperception. Mansfield and Pevehouse correspondingly suggest that PTAs can
produce focal points around which states shape their expectations for acceptable
behavior, preventing breakdowns in cooperation that might otherwise lead to vio-
lence (Garrett and Weingast, 1993; Mansfield and Pevehouse, 2000).
Why Trade Institutions Don’t Always Keep the Peace
Many before us have been skeptical of the claim that interdependence promotes
peace among states. It is well understood that international institutions can have
adverse effects on conflicts among member states, mismanaging crisis situations
and worsening conflict intensity (Gallarotti, 1991), or producing rivalry among
states due to their relative social positions (Hafner-Burton and Montgomery,
2006). We are nevertheless among the first to directly tackle the principal claims
supporting the liberal thesis that trade institutions dampen conflict, and to pro-
pose an explanation for why conflict often characterizes outcomes.
This is important because we observe significant instances of violent conflict
4
between PTA members: the 1990s alone included border clashes between
Armenia and Azerbaijan, members of the Commonwealth of Independent
States (CIS); the outbreak of war in the Great Lakes, with foreign involvement
in the Democratic Republic of Congo from Angola, Namibia, Rwanda, Uganda,
and Zimbabwe, all members of the Common Market for Eastern and Southern
Africa (COMESA); the Iraqi invasion of Kuwait and violent border clashes
between Egypt and Sudan, all members of the Council of Arab Economic Unity
(CAEU); and fighting between India and Pakistan, members of the South
Asian Association for Regional Cooperation (SAARC). North and South
Korea frequently are involved in violent incidents; both are members of the
Global System of Trade Preferences Among Developing Countries (GSTP). A
majority of these disputants are also members of the WTO. Powers contends
that in Africa, 16% of all militarized international disputes registered by the
Correlates of War data from 1950 to 1992 occurred between PTA members
(Powers, 2003, 2004). These examples show clearly that members of the same
trade institution can and do conflict, that conflict often breaks out into violence,
and that commerce is frequently not enough to keep the peace. They stand in
sharp contrast to the liberal expectation that trade institutions dampen conflict
throughanincreaseintrust.
Trade institutions do increase repeated contact between members; however,
contact does not necessarily build trust or a sense of community. The lessons of
European integration theory suggest that building community through upgrading
the common interest between PTA members requires a minimum level of homo-
geneity: a pluralist social structure, a high level of economic and industrial devel-
opment, and ideological similarity (Haas, 1960). Security communities are also
most likely to develop through economic relations among Western nations, as
even the most institutionalized forms of integration in the developing world
4
As per the usual convention, we define ‘‘violent’’ MIDs to be any MID in which at least
one death occurred.
Hafner-Burton & Montgomery: War, Trade, and Distrust
261
cannot be said to create the mutual identification at the core of the concept
(Bearce, 2003). Although evidence suggests that economic integration has led to
the formation of a collective identity and trust among member states of the
European Union over time, it is well understood that ‘‘democratic features of lib-
eral democracies enable the community in the first place’’ (Russett and Oneal,
2001: 166). The liberal argument that trade institutions dampen conflict by build-
ing trust among leaders to overcome commitment problems consequently chiefly
applies to the Western world of advanced democratic nations. Yet the overwhelm-
ing majority of trade institutions manage trade between partners that include at
least one developing or nondemocratic state, and there is no evidence to show
that these institutions build trust over asymmetrical distribution of gains.
Boehmer, Gartzke, and Nordstrom cogently argue that states that belong to many
different international institutions may have a greater number of international
interests to competitively defend and a greater array of opportunities to enact
aggressive behavior in defense of those perceived interests (Boehmer et al., 2002).
We extend this argument one step further; trade institutions create and shape
states’ interests, affecting not only the number of potential issues for dispute, but
also establishing conditions that can lead to distrust. Institutions do this by placing
states in social positions of power within international relations, which shape
expectations for behavior by defining which issues are legitimate for contestation
via military means and enable states to coerce, bribe, reward, or punish each
other. We address this possibility in the next section.
How Trade Institutions Undermine Trust and Fuel Conflict
Although we do not share the liberal view that trade institutions necessarily or
regularly promote peace among members, we are equally skeptical of critics’ argu-
ments maintaining that they have no relationship to war or a negative relationship.
Rather, we argue that trade institutions are neither inherently likely nor unlikely
to keep the peace among member states. States form social networks through
membership in these institutions. Mutual memberships create ties between states
and, although the strength of these ties increases with additional joint member-
ships, they do not necessarily create positive or negative bonds between states.
These ties define states’ relative positions in social hierarchies in the international
political economy. Like the balance of military power, these positions are state
characteristics that are measured (and have their effects) relative to other states,
shaping the conditions under which conflict or cooperation become rational strate-
gies of action.
We are concerned with two particular types of relative social positions: depen-
dence and access centrality. Take two states, A and B; state A’s dependence on
state B is proportional to the number of trade institutions both belong to and is
inversely proportional to the total number of trade institutions state A belongs to.
In other words, A’s dependence on B increases if A and B join additional trade
Conflict Management and Peace Science 29(3)
262
institutions together, but A’s dependence on B decreases if A joins additional
trade institutions without B. Like differences in military capabilities, different lev-
els of dependence create structural motives for peace among trading partners in
some circumstances, and conflict under others (Hafner-Burton and Montgomery,
2006). Here, we focus on one particular type of dependence: direct dyadic sensitiv-
ity dependence due to the PTA network. It is direct because the measure relates
primarily to direct links to each of the two actors in question; dyadic because it
looks at the strength of the link between the two actors; and sensitivity because it
measures how sensitive each actor is to the other due to the link rather than the
potential opportunity cost of breaking it. Although multiple different kinds of
dependence (direct and indirect; monadic, dyadic, and systemic; sensitivity and
vulnerability; single and multi-dimensional) are relevant to relations between
states (Maoz, 2009, 2010), we are focusing on this particular type and measuring it
in this way in this article because it provides for a direct contrast between the
mechanism of trust through interdependence, typically also measured as direct
dyadic sensitivity interdependence (i.e. the strength of a tie between two states,
the number of PTAs that both belong to), and the mechanism of distrust through
dependence, which can be measured by dividing the liberal measure by the total
number of PTAs that each state belongs to.
Specifically, we expect that distrust will characterize conditions of asymme-
trical dependence. We consequently expect the liberal prediction to be most
robust when states hold positions of dependence vis-a
`-vis each other that are
relatively equal; under these conditions, mutual memberships in trade institu-
tions are most likely to produce trust and, therefore, to create conditions favor-
able to conflict avoidance. However, we expect that states will be more likely to
conflict with members of their trade institutions when one state is much more
dependent on the other; under these conditions, disparity in dependence is more
likely to lead to distrust, and therefore, provide incentives for the distrustful
state to initiate a conflict.
Hypothesis 1: Dyads characterized by greater relative disparity in dependence will be
more likely to engage in militarized disputes than dyads characterized by relative equal-
ity in dependence.
While the (direct dyadic sensitivity) dependence of two states on each other can
be derived from their direct membership ties alone, access centrality is an emer-
gent property of the ties between all states in the entire system. There are three
(major) classes of centrality measures—access, brokerage, and efficiency—
which measure the extent to which an actor is generally well-connected to oth-
ers, can act as a broker between actors, and can reach other actors quickly
(Hafner-Burton et al., 2009). Here, we use the access centrality measure since
this family of measures has been shown to be important in the relationship
between international institutions and conflict in general and PTAs in particular
(Hafner-Burton and Montgomery, 2006, 2008, 2009). Conceptually, a state’s
Hafner-Burton & Montgomery: War, Trade, and Distrust
263
access centrality in a network is proportional to the ties received from other
states. A state with high access centrality—that is, one with a large number of
strong ties to other important states—is particularly socially powerful and
important in the network of trade institutions. Further, we believe that a large
number of ties to other important states will increase that state’s capabilities fur-
ther, and so we use a measure of access centrality (eigenvector centrality) that
reflects this. With large disparities in centrality, we expect that (similar to our
predictions on dependence) trust will be more likely to form among states that
have equal access centrality, while distrust and therefore conflict will character-
ize relationships between pairs of states with large disparities in access central-
ity.
5
Disparities in social power thus operate opposite to disparities in material
power, where (according to realist arguments) equality leads to increased con-
flict and a preponderance of power suppresses it.
Hypothesis 2: Dyads characterized by greater relative disparity in access centrality will
be more likely to engage in militarized disputes than dyads characterized by relative
equality in access centrality.
Large disparities in dependence and access centrality create asymmetries in states’
abilities to socially coerce, bribe, reward, or sanction each other; more influential
and access central states simply have more social capital to expend. These states
will distrust marginal countries that are not well integrated into their networks
and will be more likely to escalate conflict rather than resolve disputes peacefully.
Conversely, less influential states have little ability to respond to social coercion
and so may choose to respond in a material way through military or economic
action. Disparities thus produce incentives for forum-switching for such states.
Highly dependent or peripheral states are excluded from communities and fear
domination within economic forums, and thus rather than attempting to compete
in the socioeconomic sphere where disparities are large, countries have incentives
to enter militarized conflicts instead. Expectations and abilities combine to under-
mine trust across these dyads.
Research Design
We test our two hypotheses using pooled cross-national time-series data on state
dyad-years. Our attention is focused on all dyads
6
from the period 1950 to 2000,
although we replicate our results on politically relevant dyads in order to facilitate
5
Dependence is inherently a directed dyadic quantity that defines a relative hierarchy
between two states, whereas access centrality is a monadic quantity that defines an abso-
lute hierarchy among all states. We reduce both quantities to undirected dyadic quantities,
since we argue that it is the relative position of two states with respect to each other that
affects conflict.
6
We choose to use all dyads instead of politically relevant dyads since the effects captured
by taking only a subset of states (power projection capabilities, distance between dyads)
are already included in our model.
Conflict Management and Peace Science 29(3)
264
comparison across studies that consider only the latter sample.
7
In order to ensure
comparison with studies supporting the liberal thesis that international institutions
dampen conflict, we base our analysis on Mansfield and Pevehouse, who consider
the effects of PTAs on conflict (Mansfield and Pevehouse, 2000); we also draw
insights from Oneal and Russett’s analysis of the effects of IGOs on conflict.
8
We incorporate their data and update them in several ways. In order to exam-
ine the post-Cold War period, we rely on an alternative trade institution data
set, since Mansfield and Pevehouse’s data only consider PTA membership from
1950 to 1985. We consider membership in a greater number of PTAs and
observe membership from 1950 to 2000.
9
In our sample, we include membership
in the General Agreement on Trade and Tariffs (GATT) and the WTO because
these trade institutions are just as likely as PTAs to offer a basis for mutual
trust. We base our GDP and trade data on Gleditsch’s expanded set (Gleditsch,
2002), our measures of regime type from Polity IV (Marshall and Jaggers,
2003), and generate the remainder of our data using EUGene (Bennett and
Stam, 2000, 2005). We employ Beck, Katz and Tucker’s splines to correct for
temporal dependence in the dependent variable (Beck et al., 1998; Tucker,
1999), and calculate our social network variables using the sna package in R
(Butts, 2007; R Development Core Team, 2007). Regression analysis was done
in STATA (Stata Corporation, 2005).
Trade institutions exhibit a great deal of institutional variation. Ideally, any
study comparing the effects of relative institutional ties on conflict with the effects
of absolute ties would also include information about varying institutional quali-
ties, such as dispute settlement mechanisms or security aims. Unfortunately, these
data are simply not available for most trade institutions. Following both Mansfield
and Pevehouse (2000) and Oneal and Russett (1999), we consequently adopt the
simplifying assumption that trade institutions can be analyzed as if they supply
homogenous institutional qualities across agreements. We thus assume that social
network properties that emerge through a given institution (such as NAFTA) cre-
ate equivalent types of social ties to those created by another institution (such as
ASEAN).
The liberal perspective suggests that increased information through organiza-
tional mechanisms as well as social interaction itself and a feeling of sameness
and mutual identification decreases conflict. Consequently, both components of
the liberal argument suggest that the conflict-decreasing properties of trade
7
In order to facilitate replication of our findings, we have made our data, including our
PTA and state samples, available at the following locations: http://irps.ucsd.edu/ehafner/
and http://ahm.name/.
8
Oneal and Russett’s (1999) study gives full details of their model specification and their
results.
9
Data were collected by Emilie Hafner-Burton using sources from the WTO, McCall
Smith (2000), and Schott (2003). We would like to thank Ed Mansfield, Jon Pevehouse,
Bruce Russett, and John Oneal for generously sharing their data.
Hafner-Burton & Montgomery: War, Trade, and Distrust
265
institutions should be proportional to membership, not simply dichotomous;
more arrangements mean more information, and similarly more common iden-
tification. In order to test our social network hypotheses, we thus create a count
measure of mutual membership in institutions. This is important because
Mansfield and Pevehouse rely on a binary measure of mutual PTA ties between
dyads (Mansfield and Pevehouse, 2000). This measure allowed them to test
whether two states with mutual membership in any PTA are less likely to dis-
pute than two states without membership. Our count measure of mutual mem-
bership better fits both the liberal hypotheses and our social network
hypotheses.
We start by deriving a general measure of mutual membership in trade institu-
tions. We incorporate all trade institutions in the sample, including non-
preferential institutions such as GATT, PTAs composed of other PTAs such as
the EU-Gulf Cooperation Council, and non-reciprocal arrangements such as the
Cotonou arrangement and the numerous EU arrangements with individual states
outside of the EU. We treat all memberships as symmetrical and equal for the
purposes of calculating liberal and social network variables, since co-membership
in any of these institutions is a mutual affiliation that could transmit information
or increase affinity.
For each year, we observe the nstates and ktrade institutions that exist for that
year, forming an nby kaffiliation matrix A.
10
Each entry is either 1 (if a state is a
member of an institution) or 0 (if not). We then convert the affiliation matrix A
into a sociomatrix Sby multiplying the matrix by its transpose (S=A
0A). Each
off-diagonal entry s
ij
is equal to the number of trade institution that states iand j
have in common (our variable PTASAME
ij
), while the diagonal s
ii
is equal to the
total number of trade institutions country ibelongs to. The diagonal elements are
the total number of PTAs a given country belongs to; the off-diagonal elements
indicate the number of PTAs two countries share.
We then define the dependence of state ion state jas the number of shared
memberships of iand jdivided by the total number of trade institution member-
ships of state i, producing a dependence matrix S
ˇ.
DEPENDENCEij =
sij =sij=sii
An actor with high access centrality is the recipient of many ties. If we believed
that only direct ties mattered, we would use a simple form of access centrality,
degree centrality, computing by summing over an actor’s incoming ties. However,
according to our hypothesis, it is not simply the incoming ties that matter, but also
the importance of the actors sending those ties. Consequently, we define the access
centrality of a state as the sum of the state’s ties to the other actors in the system,
10
An affiliation matrix is a social network term for a special case of a two-mode matrix. A
two-mode matrix has two distinct types of entities; an affiliation matrix is a two-mode
matrix with only one set of actors. See Wasserman and Faust (1994: 29–30).
Conflict Management and Peace Science 29(3)
266
Table 1. Estimates of the Effects of PTAs on MIDs, 1950 to 2000: Replications and Core Models
(1) Replications of M&P (2000) (2) Base Model (3) Network Model
Variable
1950–1985 1950–2000 1950–2000 1950–2000
pr dyads pr dyads all dyads all dyads all dyads
PTA
ij
3.98E-02 (1.33E-01) 4.76E-02 (8.91E-02) 0.35 (0.09)***
PTA
ij
x TRADE
ij
22.14E-04 (1.13E-04)
+
23.50E-06 (8.44E-06) 9.17E-06 (1.23E-05)
PTA
ij
x GDP
L
3.17E-06 (1.32E-06)*8.52E-08 (3.37E-07) 22.41E-07 (4.85E-07)
PTA
ij
x GDP
H
27.64E-07 (3.62E-07)*4.65E-10 (7.76E-08) 26.56E-08 (8.05E-08)
PTASAME
ij
22.82E-02 (1.18E-02)*22.99E-02 (1.30E-02)*
DEPENDENCE
D
0.22 (0.10)*
ACCESS CENTRALITY
D
2.46 (0.83)**
DEM
L
24.57E-02 (1.04E-02)*** 26.33E-02 (8.56E-03)*** 27.58E-02 (8.62E-03)*** 27.93E-02 (9.60E-03)*** 27.65E-02 (9.52E-03)***
DEM
H
3.23E-02 (9.81E-03)** 3.70E-02 (8.30E-03)*** 4.36E-02 (7.49E-03)*** 4.53E-02 (7.82E-03)*** 4.00E-02 (7.38E-03)***
GROWTH
L
1.77E-03 (4.66E-03) 22.81E-03 (3.84E-03) 22.10E-03 (3.84E-03) 27.70E-03 (3.57E-03)*27.71E-03 (3.54E-03)*
TRADE
ij
22.77E-05 (1.68E-05)
+
25.73E-06 (1.52E-05) 22.64E-05 (1.79E-05) 21.75E-05 (7.12E-06)*21.65E-05 (6.79E-06)*
GDP
L
1.05E-06 (2.28E-07)*** 6.50E-07 (1.95E-07)** 1.26E-06 (3.24E-07)*** 6.34E-07 (3.05E-07)*5.86E-07 (2.90E-07)*
GDP
H
2.61E-07 (5.54E-08)*** 1.32E-07 (3.03E-08)*** 4.45E-07 (2.32E-08)*** 3.34E-07 (4.16E-08)*** 3.28E-07 (4.37E-08)***
CAPRATIO
ij
22.91E-03 (6.61E-04)*** 23.25E-03 (7.08E-04)*** 21.81E-03 (5.50E-04)** 22.91E-03 (9.20E-04)** 22.91E-03 (9.15E-04)**
ALLIES
ij
20.15 (0.13) 24.76E-02 (1.07E-01) 0.22 (0.11)
+
3.79E-02 (1.26E-01) 3.57E-02 (1.27E-01)
HEGEMONY 1.08E-02 (2.12E00) 1.81 (1.90) 4.90 (1.97)*1.23 (1.91) 20.40 (1.89)
CONTIG
ij
1.32 (0.16)*** 1.27 (0.15)*** 2.94 (0.13)*** 24.98 (0.51)*** 25.07 (0.49)***
MAJPOWER
ij
1.17 (0.15)*** 1.19 (0.16)***
LOGDIST
ij
21.04 (0.07)*** 21.05 (0.07)***
N 25209 42320 407209 407209 407209
All models are estimated using logit and include a natural spline function with three knots, omitted from the table due to space considerations. Numbers in parentheses
are panel corrected standard errors. Pr dyads are politically relevant dyads.
+
p\.10; *p\.05; ** p\.01; *** p\.001.
Hafner-Burton & Montgomery: War, Trade, and Distrust
267
Table 2. Robustness Checks of the Estimates of the Effects of PTAs on MIDs, 1950 to 2000
(3) (4) (5) (6) (7)
Variable
Networks Minimal Model PR dyads DISPUTE Trade Dep
PTASAME
ij
22.99E-02 (1.30E-02)*26.72E-02 (1.29E-02)*** 26.16E-02 (1.97E-02)** 26.47E-03 (1.14E-02) 24.57E-02 (1.24E-02)***
DEPENDENCE
D
0.22 (0.10)*0.37 (0.16)*3.90E-02 (8.92E-02) 0.22 (0.09)*0.30 (0.09)**
ACCESS CENTRALITY
D
2.46 (0.83)** 3.49 (0.90)*** 2.24 (0.78)** 2.10 (0.71)** 2.87 (0.91)**
DEM
L
27.65E-02 (9.52E-03)*** 26.12E-02 (8.79E-03)*** 28.90E-02 (9.34E-03)*** 27.98E-02 (1.01E-02)***
DEM
H
4.00E-02 (7.38E-03)*** 3.26E-02 (7.75E-03)*** 3.79E-02 (7.44E-03)*** 4.72E-02 (7.58E-03)***
GROWTH
L
27.71E-03 (3.54E-03)*24.43E-03 (3.63E-03) 25.59E-03 (3.20E-03)
+
23.51E-03 (3.61E-03)
TRADE
ij
21.65E-05 (6.79E-06)*25.80E-06 (4.40E-06) 21.60E-05 (5.53E-06)**
TRADEDEP
L
26.71 (8.60)
GDP
L
5.86E-07 (2.90E-07)*5.60E-07 (1.55E-07)*** 4.25E-07 (2.46E-07)
+
GDP
H
3.28E-07 (4.37E-08)*** 2.00E-07 (3.22E-08)*** 2.88E-07 (4.45E-08)***
CAPRATIO
ij
22.91E-03 (9.15E-04)** 22.93E-03 (6.94E-04)*** 23.28E-03 (8.42E-04)*** 22.76E-03 (9.92E-04)**
ALLIES
ij
3.57E-02 (1.27E-01) 27.20E-02 (1.10E-01) 5.56E-02 (1.17E-01) 0.12 (0.13)
HEGEMONY 20.40 (1.89) 0.42 (1.91) 2.73 (1.78) 25.73 (1.81)**
CONTIG
ij
25.07 (0.49)*** 24.15 (0.52)*** 23.29 (0.64)*** 24.24 (0.52)*** 24.55 (0.62)***
MAJPOWER
ij
1.19 (0.16)*** 1.69 (0.15)*** 20.28 (0.15)
+
1.31 (0.16)*** 1.82 (0.16)***
LOGDIST
ij
21.05 (0.07)*** 20.96 (0.07)*** 20.57 (0.08)*** 20.90 (0.07)*** 20.98 (0.09)***
N 407209 534407 42320 407209 407209
All models are estimated using logit and include a natural spline function with three knots, omitted from the table due to space considerations. Numbers in parentheses
are panel corrected standard errors. PR dyads are politically relevant dyads.
+
p\.10; *p\.05; ** p\.01; *** p\.001.
Conflict Management and Peace Science 29(3)
268
weighted by the access centrality of the actors sending ties to that state. In practice,
we compute the eigenvector centrality
11
of the dependence matrix S
ˇ.
12
We then convert these measures into undirected dyadic form by computing
the difference for each quantity: DEPENDENCE
D
= |DEPENDENCE
ij
–
DEPENDENCE
ji
|and ACCESS CENTRALITY
D
= |ACCESS CENTRALITY
i
–
ACCESS CENTRALITY
j
|.
Statistical Analyses
We begin by replicating the core findings presented in Mansfield and Pevehouse
(2000) using our new trade institution data. In order to do so, we estimate the fol-
lowing model:
MIDij =b0+b1PTAij +b2(TRADEij3PTAij)+b3(GDPL3PTAij)+b4(GDPH3PTAij )
+b5DEML+b6DEMH+b7GROWTHL+b8TRADEij
+b9GDPL+b10GDPH+b11CAPRATIOij +b12 ALLIESij +b13HEGEMONY
+b14CONTIGij +eij
ð1Þ
A militarized international dispute (MID
ij
) occurs when a state threatens or enacts
military force against another state. The observed value of the dependent variable
is binary, equaling 1 if a dyad ij begins a MID in a given year t, and 0 if no MID is
observed. This variable is often termed MID onset, in contrast to DISPUTE
ij
,
which equals 1 for each year of a MID rather than just the first year. We include
MID joiners as experiencing a MID onset in the year that they join.
To replicate, we employ independent variables consistent with Mansfield and
Pevehouse and effectively lagged by one year. The core of their analysis rests on
the results of PTA membership on conflict. The authors measure whether a pair
11
Bonacich power centrality (a generalization of eigenvector centrality, see Bonacich,
1987) is a frequently-used measure in the access centrality family of measures (Hafner-
Burton et al., 2009) that takes into account through the parameter bhow much access
comes from being connected to other well-connected nodes (b.0) or to more isolated
nodes (b\0). The latter happens in exchange networks (Cook et al., 1983). Since our net-
work is not an exchange network and our theoretical expectation is that social influence
comes from being connected to others who are well-connected in institutional networks,
we choose b.0. There is little guidance in the literature as to how to choose b,andso
consequently we default to the simpler eigenvector measure, which occurs when the para-
meter bapproaches the reciprocal of the largest eigenvalue of the sociomatrix.
12
Using the directed dependence matrix S
ˇinstead of the symmetrical raw sociomatrix S
can be problematic in cases where a network contains unreciprocated ties. In these cases,
alternate, eigenvector-like measures should be used. However, in the dependence matrix,
every outgoing tie has an incoming tie as well. Consequently, eigenvector approaches are
equally valid (Bonacich and Lloyd, 2001).
Hafner-Burton & Montgomery: War, Trade, and Distrust
269
of states ij share membership in any PTA during a given year t, drawing upon their
sample of institutions from 1950 to 1985. They call this binary variable PTA
ij
and
expect that dyads sharing any mutual memberships will be less likely to conflict.
We replicate this variable in our new and updated sample including more institu-
tions and covering the period 1950 to 2000. Since Mansfield and Pevehouse’s sam-
ple appeared to exclude general trade institutions that were not agreements (e.g.
GATT), for the purposes of initial replication in Table 1 we exclude these institu-
tions and compute PTA
ij
;
13
using PTA
ij
, and we re-create several interaction terms
with GDP and trade variables, described below. However, for the reasons stated
above, we use a continuous measure that includes these institutions for our exten-
sions of the model in Table 2, PTASAME
ij
.
Four additional variables are used to discriminate between the effects of PTAs
and the effects of other liberal forces influencing conflict. Mansfield and
Pevehouse test a ‘weak link’ hypothesis about the dependence of democracy.
DEM
L
measures the political character of the less democratic state in a dyad,
which they expect to be the stronger determinant of conflict behavior. Because a
MID can result from the actions of a single state, they argue that MID likelihood
mainly depends on the level of political constraint experienced by the weak link—
the less constrained state in each dyad (or the less democratic state). The variable
ranges from –10 for a state characterized by extremely autocratic political institu-
tions, to 10 for a state characterized by extremely democratic political institutions.
Following Mansfield and Pevehouse, our second model also includes the regime
type of the more democratic state in the dyad, DEM
H
. This weak link theory
extends to economic growth as well. Since states with lagging growth may have
incentives to launch wars to divert the public’s attention from poor economic con-
ditions, both studies include a measure of economic growth. GROWTH
L
mea-
sures the percentage change in GDP per capita during the previous three years
for the state in each dyad that has experienced the smallest change. TRADE
ij
measures the sum of i’s exports to and imports from jin year t.
The authors consider additional variables to control for realist expectations
about the causes of conflict. GDP
H
and GDP
L
measure the real Gross Domestic
Product of the state with the highest and lowest national income in 1996 US dol-
lars, respectively. CAPRATIO
ij
is the ratio of the stronger state’s military
capability—measured by averaging its share of world population, urban popula-
tion, military expenditures, military personnel, iron and steel production, and
energy consumption—to that of the weaker dyad member. This may increase con-
flict (if the stronger state is tempted to take over the weaker one) or decrease it
(if the stronger state deters the weaker state from attacking). ALLIES
ij
equals 1 if
the dyad members were linked by formal mutual defense treaties, neutrality pacts,
or entente, and equals 0 otherwise. This variable is important to control for the
common wisdom that allies are generally less likely to fight each other than
13
For the subset of overlapping dyads between our samples, our PTA
ij
is highly correlated
with their variable (~0.84).
Conflict Management and Peace Science 29(3)
270
non-allied states because they share a common security interest. The authors also
include the variable HEGEMONY in order to control for the strength of the most
powerful state relative to other states in the international system. This variable is
computed by the percentage of total global GDP generated by the state with the
largest GDP in the previous year (the United States for every year in our sample).
Finally, CONTIG
ij
controls for the potential that MIDs result when at least one
member of a dyad can reach the other member with effective military force. The
variable equals 0 if two states are not directly or indirectly contiguous and 1 if
they share a territorial boundary.
14
In the first column of Table 1, we report estimates of our replication of
Mansfield and Pevehouse, Model (1), using our measure, PTA
ij
. In order to be as
consistent as possible with their analysis, in this column we limit our time span
from 1950 through 1985
15
and our sample to politically relevant dyads. In the sec-
ond column, we expand the replication through 2000 and in the third column we
replicate this second model on an expanded sample of states that include all
dyads.
When using the same variable construction as Mansfield and Pevehouse (binary
PTA
ij
, excluding organizations composed of PTAs and general trade institutions),
the same timespan (1950–1985), and the same dyads (politically relevant), we find
very similar results. The coefficient on PTA
ij
does switch signs from their results,
but as both are small and highly insignificant, this is a consistent finding. Our coef-
ficients on the three interaction terms are in the same direction as their results,
although at lower levels of significance. The remaining variables also have similar
results, except for HEGEMONY, which is insignificant in our tests. However,
when we expand our timespan through 2000, we find that none of the PTA-related
variables are significant; moreover, when we move to all dyads, we find that PTA
ij
is highly significant and positive, indicating that shared membership in a PTA
overall actually leads systematically to increased conflict.
Using these results, we build a base model, Model (2), from which we compute
all subsequent models. For the reasons stated previously, we believe that a count
measure that includes all trade institutions is more compatible with both versions
of the liberal hypothesis (as well as a better measure for construction of social net-
work variables); the anomalous finding that PTAs seem to increase conflict in
some models adds empirical support for our argument that a binary measure is
inadequate. Consequently, we include our continuous measure PTASAME
ij
that
includes all trade institutions in all remaining regressions instead of PTA
ij
.
Because the interaction terms between membership in trade institutions and trade
and GDP are insignificant when we expand the time horizon of the study either
with all dyads or politically relevant ones, we exclude them from Model (2). We
also introduce two additional control variables used by Oneal and Russett in their
studies of the effects of IGOs on MIDs: LOGDIST
ij
controls for the natural
14
See Mansfield and Pevehouse (2000) for complete descriptions of their variable codings.
15
Due to limits on the temporal span of the GDP variables (1950–2000), where we include
this variable our first year is 1951.
Hafner-Burton & Montgomery: War, Trade, and Distrust
271
logarithm of mileage between the two capitals (or major ports for the super-
powers) of dyad partners, while MAJPOWER
ij
controls for the effects of great
powers.
16
The variable takes on a value of 0 if a dyad is made up of minor powers
and 1 if it contains at least one great power. Both are important to control for rea-
list explanations of war.
MIDij =b0+b1PTASAMEij
+b2DEML+b3DEMH+b4GROWTHL+b5TRADEij
+b6GDPL+b7GDPH+b8CAPRATIOij +b9ALLIESij +b10HEGEMONY
+b11CONTIGij +b12MAJPOWERij +b13 LOGDISTij +eij
ð2Þ
Using Model (2) as our base, we add DEPENDENCE
D
and ACCESS
CENTRALITY
D
in Model (3) to test our social network hypotheses about depen-
dence and access centrality. We report estimates of the parameters from Model
(3) in the final column of Table 1. Our results show that membership in trade
institutions can both increase and decrease conflict among members. Consistent
with the liberal expectation, dyads linked by more ties to trade institutions are less
likely to go to war. However, in contrast with the liberal expectation, we find two
circumstances under which memberships in trade institutions can significantly
increase the likelihood of conflict among members. When dyads are characterized
by relative disparities in dependence or access centrality within the network of
trade ties that characterize the structure of the international political economy,
states are more likely to fight. When dyads are characterized by relative equality
in dependence or access centrality, states are less likely to engage in militarized dis-
putes, controlling for the number of overall memberships.
We also test the empirical import of our findings. Using CLARIFY (Tomz
et al., 2003), we set all variables at their median levels, then increase each variable
of interest separately to its 95% level. Increasing PTASAME
ij
in this way
decreases MID onset by 7%. By contrast, increasing DEPENDENCE
D
and
ACCESS CENTRALITY
D
to the same level increases MID onset by 17% and
10% respectively. To think of this a different way, suppose that two states A and B
at the median in all other variables belong to just one joint PTA in total, so
PTASAME
ij
is 1 and DEPENDENCE
D
is 0. Now suppose that state A joins a
PTA with state C as well. DEPENDENCE
D
then becomes 0.5, while PTASAME
ij
remains at 1. The chance of militarized dispute onset increases by 10% as a result.
If state A and state B joined a second PTA together, by contrast, PTASAME
ij
would become 2, while DEPENDENCE
D
would stay at 0, decreasing militarized
dispute onset by 3.5%. While these are not enormous, in terms of the influence
that states’ individual choices can have, they are quite important.
Our findings, moreover, are robust. We have taken a number of steps to pro-
vide results that are as consistent with as many different sample and variable spec-
ifications as possible. Although we cannot report all of those steps here, we do
16
Oneal and Russett have a slightly different set of countries as major powers; we use the
COW2 dataset to determine major power status.
Conflict Management and Peace Science 29(3)
272
address some of the more important issues. Table 2 offers estimates across four
new models. In Model (4), we test a minimalist model by eliminating all realist
and liberal variables while retaining geographic variables, social variables, and
PTASAME
ij
. In Model (5), we test whether our results are consistent when we
truncate our sample to politically relevant dyads. In Model (6), we consider an
alternative specification of the dependent variable that takes on a value of 1
throughout the entire dispute rather than just the onset. Finally, Model (7) offers
different specifications of the relationship between trade, GDP, and conflict:
instead of three separate components, we include Oneal and Russett’s measure of
trade dependence TRADEDEP
L
, equal to the lower ratio of bilateral trade to
GDP in a given dyad. We also substituted in Zeev Maoz’s dependence balance
measure (Maoz, 2009, 2010), which has a 0.97 correlation with our own, in all
models; the results were substantively unchanged.
Our results are quite stable across models, with some variation. In the minimal
model, all three remaining variables (PTASAME
ij
,DEPENDENCE
D
, and
ACCESS CENTRALITY
D
) have a greater effect at the same or better significance
levels. When we restrict our sample to only politically relevant dyads, the coeffi-
cient on DEPENDENCE
D
becomes insignificant. This is an interesting finding,
since it could indicate that the effects of PTA dependence (but not access central-
ity) are limited to countries that fall outside of traditional definitions of ‘‘politically
relevant’’ states. We are conducting further research to investigate the causes of
this particular finding, although we strongly prefer measuring political relevance
through the inclusion of variables rather than through the exclusion of cases.
When using DISPUTE as the dependent variable instead of dispute onset (MID),
PTASAME
ij
becomes insignificant, while the coefficent on DEPENDENCE
D
increases significantly and the coefficient on ACCESS CENTRALITY
D
decreases
slightly. When including the effects of trade dependence, we find that while trade
dependence itself is insignificant, the parameter estimates for ACCESS
CENTRALITY
D
and PTASAME
ij
increase, while DEPENDENCE
D
decreases
slightly.
Conclusion
PTAs are not inherently trust-building or rivalry-inducing institutions and mem-
bership does not automatically reveal previously hidden information about mili-
tary capacity, build trust, or ensure commitment to peace. Instead, there is
significant variation in the effects of trade institution membership across states in
the international system. While there are a group of states that fulfill the condi-
tions for the liberal peace, for many states at the fringes, membership can actually
increase conflict by defining hierarchies of access centrality and dependence that
cause conflict with those at the bottom of those hierarchies.
Our findings therefore strongly confirm the liberal belief that trade institutions
can keep the peace under some circumstances—when they create ties among states
with relatively equal social positions within the international political economy,
facilitating trust and reciprocity. Yet trade institutions do not always keep the
Hafner-Burton & Montgomery: War, Trade, and Distrust
273
peace, and the liberal mechanisms are not universally applicable. We show that
trade institutions can also create relative disparities in social status among mem-
bers and that these disparities increase the likelihood of militarized conflict. Far
from being vehicles of peace, trade institutions generate inequalities of social
power that can lead to distrust and, ultimately, to conflict among nations.
Social network analysis gives us a way to conceptualize and tools to measure
the effects of international institutions beyond simple membership effects. In par-
ticular, it allows us to think about asymmetries created by the uneven distribution
of ties across the international system. Here, we have identified two important
asymmetries in the international trade institution network that lead to increased
militarized conflict: PTA access centrality and direct dyadic PTA sensitivity
dependence. Without the use of network analysis, such results would be difficult
to conceive of and impossible to measure. These two measures do not exhaust the
possible effects of the network of PTAs on conflict. Other centrality measures
related to the brokerage and efficiency families could have significant effects on
conflict through different mechanisms, while the network also creates potentially
important dependencies at the monadic, dyadic, and systemic levels through direct
and indirect effects and through both sensitivity and vulnerability. Measuring
these effects will be important for the future agenda of network analysis in inter-
national relations.
References
Adler, E., and M. N. Barnett. 1998. Security Communities. Cambridge Studies in Interna-
tional Relations. Cambridge: Cambridge University Press.
Adler, E., and B. Crawford. 2002. Constructing a Mediterranean regime: A cultural
approach. Paper presented at The Convergence of Civilizations? Constructing a Medi-
terranean Regime conference. 6–9 June. Arrabiba Monastery, Fundacao Oriente, Lis-
boa, Portugal.
Angell, N. 1913. The Great Illusion: A Study of the Relation of Military Power to National
Advantage. 4th rev. and enl. edn. New York: Putnam.
Baldwin, R. E. 1980. Economic Development and Growth. 2nd edn. Huntington, NY: R. E.
Krieger.
Barbieri, K. 1996. Economic interdependence: A path to peace or a source of interstate con-
flict? Journal of Peace Research 33(1): 29–49.
Bearce, D. H. 2003. Grasping the commercial institutional peace. International Studies
Quarterly 47(3): 347–370.
Beck, N., J. N. Katz and R. Tucker. 1998. Taking time seriously: Time-series–cross-section
analysis with a binary dependent variable. American Journal of Political Science 42(4):
1260–1288.
Bennett, D. S., and A. C. Stam. 2000. EUGene: A conceptual manual. International Interac-
tions 26: 179–204.
Bennett, D. S., and A. C. Stam. 2005. EUGene v. 3.1. (http://eugenesoftware.org)
Boehmer, C., E. Gartzke and T. Nordstrom. 2002. Do intergovernmental organizations
promote peace? Manuscript. (http://www.columbia.edu/~eg589/pdf/IGOsandpeace01130
2b.pdf)
Bonacich, P. 1987. Power and centrality: A family of measures. American Journal of Sociol-
ogy 92(5): 1170–1182.
Conflict Management and Peace Science 29(3)
274
Bonacich, P., and P. Lloyd. 2001. Eigenvector-like measures of centrality for asymmetric
relations. Social Networks 23(3): 191–201.
Butts, C. T. 2007. sna: Tools for Social Network Analysis. R package v. 1.5. (http://erzuli.s-
s.uci.edu/R.stuff)
Buzan, B. 1984. Economic structure and international security: The limits of the liberal case.
International Organization 38(4): 597–624.
Cook, K. S., R. M. Emerson, M. R. Gillmore and T. Yamagishi. 1983. The distribution of
power in exchange networks: Theory and experimental results. American Journal of
Sociology 89(2): 275–305.
Dembinski, M., A. Hasenclever and W. Wagner. 2004. Towards an executive peace? The
ambivalent effects of inter-democratic institutions on democracy, peace, and war. Inter-
national Politics 41(4): 543–564.
Deutsch, K. W. 1957. Political Community and the North Atlantic Area: International Orga-
nization in the Light of Historical Experience. Princeton, NJ: Princeton University Press.
Dorussen, H., and H. Ward. 2008. Intergovernmental organizations and the Kantian peace:
A network perspective. Journal of Conflict Resolution 52(2): 189–212.
Doyle, M. W. 1997. Ways of War and Peace: Realism, Liberalism, and Socialism. New York:
Norton.
Elkins, Z. 2009. Constitutional networks. In Networked Politics: Agency, Power, and Gov-
ernance, ed. M. Kahler, pp. 43–66. Ithaca, NY: Cornell University Press.
Elkins, Z., A. T. Guzman and B. A. Simmons. 2006. Competing for capital: The diffusion of
bilateral investment treaties, 1960–2000. International Organization 60(4): 811–846.
Fernande
´z, R., and J. Portes. 1998. Returns to regionalism: An analysis of nontraditional
gains from regional trade agreements. World Bank Economic Review 12(2): 197–220.
Gallarotti, G. M. 1991. The limits of international organization: Systematic failure in the
management of international relations. International Organization 45(2): 183–220.
Garrett, G., and B. R. Weingast. 1993. Ideas, interests, and institutions: Constructing the
EC’s internal market. In Ideas and Foreign Policy: Beliefs, Institutions, and Political
Change, eds J. Goldstein and R. O. Keohane, pp. 173–206. Ithaca, NY: Cornell Univer-
sity Press.
Gartzke, E. 1998. Kant we all just get along? Opportunity, willingness, and the origins of
the democratic peace. American Journal of Political Science 42(1): 1–27.
Gartzke, E., Q. Li and C. Boehmer. 2001. Investing in the peace: Economic interdepen-
dence and international conflict. International Organization 55(2): 391–438.
Gilpin, R. 1981. War and Change in World Politics. Cambridge and New York: Cambridge
University Press.
Gleditsch, K. S. 2002. Expanded trade and GDP data. Journal of Conflict Resolution 46(5):
712–724.
Haas, E. 1960. International integration: The European and the universal process. Interna-
tional Organization 4: 607–646.
Haas, E. B. 1971. The study of regional integration: Reflections on the joy and anguish of
pretheorizing. In Regional Integration: Theory and Research, eds L. N. Lindberg and S.
A. Scheingold, pp. 3–42. Cambridge, MA: Harvard University Press.
Hafner-Burton, E. M., and A. H. Montgomery. 2006. Power positions: International organi-
zations, social networks, and conflict. Journal of Conflict Resolution 50(1): 3–27.
Hafner-Burton, E. M., and A. H. Montgomery. 2008. Power or plenty: How do interna-
tional trade institutions affect economic sanctions? Journal of Conflict Resolution 52(2):
213–242.
Hafner-Burton & Montgomery: War, Trade, and Distrust
275
Hafner-Burton, E. M., and A. H. Montgomery. 2009. Globalization and the social power
politics of international economic networks. In Networked Politics: Agency, Power, and
Governance, ed. M. Kahler, pp. 23–42. Ithaca, NY: Cornell University Press.
Hafner-Burton, E. M., M. Kahler and A. H. Montgomery. 2009. Network analysis for inter-
national relations. International Organization 63(3): 559–592.
Hirschman, A. O. 1945. National Power and the Structure of Foreign Trade. Berkeley, CA:
University of California Press.
Ingram, P., J. Robinson and M. L. Busch. 2005. The intergovernmental network of world
trade: IGO connectedness, governance and embeddedness. American Journal of Sociol-
ogy 111(3): 824–858.
Kant, I. [1795] 1991. Toward perpetual peace: A philosophical sketch. In Kant’s Political
Writings, ed. X. Reiss, pp. 93–130. Cambridge: Cambridge University Press.
Manger, M. S., M. A. Pickup and T. A. B. Snijders. 2008. When country interdependence is
more than a nuisance: The longitudinal network approach. Paper presented at the 104th
Annual Meeting of the American Political Science Association. 28–31 August. Boston,
MA.
Mansfield, E. D., and J. C. Pevehouse. 2000. Trade blocs, trade flows, and international con-
flict. International Organization 54(4): 775–808.
Mansfield, E. D., and J. C. Pevehouse. 2003. Institutions, interdependence, and interna-
tional conflict. In Globalization and Armed Conflict, eds G. Schneider, K. Barbieri and
N. P. Gleditsch, pp. 233–250. London: Routledge.
Mansfield, E. D., J. C. Pevehouse and D. H. Bearce. 1999. Preferential trading arrange-
ments and military disputes. Security Studies 9(1–2): 92–118.
Maoz, Z. 2009. The effects of strategic and economic interdependence on international con-
flict across levels of analysis. American Journal of Political Science 53(1): 223–240.
Maoz, Z. 2010. Networks of Nations: The Formation, Evolution, and Effect of International
Networks, 1816–2001. New York: Cambridge University Press.
Maoz, Z., L. G. Terris, R. D. Kuperman and I. Talmud. 2005. International relations: A
network approach. In New Directions for International Relations: Confronting the
Method-of-Analysis Problem, eds A. Mintz and B. M. Russett, pp. 35–64. Lanham, MD:
Lexington.
Maoz, Z., L. G. Terris, R. D. Kuperman and I. Talmud. 2007. What is the enemy of my
enemy? Causes and consequences of imbalanced international relations, 1816–2001.
Journal of Politics 69(1): 100–115.
Marshall, M. G., and K. Jaggers. 2003. Polity IV Project: Political Regime Characteristics
and Transitions, 1800–2002. Integrated Network for Societal Conflict Research, 30 Sep-
tember. (http://www.cidcm.umd.edu/inscr/polity/)
McCall Smith, J. 2000. The politics of dispute settlement design: Explaining legalism in
regional trade pacts. International Organization 54(1): 137–180.
Mearsheimer, J. J. 1994. The false promise of international institutions. International Secu-
rity 19(3): 5–49.
Mitrany, D. 1966. A Working Peace System. Pittsburgh, PA: Quadrangle.
Mitrany, D. 1976. The Functional Theory of Politics. London: London School of Economics
and Political Science.
Nye, J. S. 1971. Comparing common markets: A revised neo-functionalist model. In
Regional Integration: Theory and Research, eds L. N. Lindberg and S. A. Scheingold, pp.
192–231. Cambridge, MA: Harvard University Press.
Oneal, J. R., and B. M. Russett. 1999. The Kantian peace: The pacific benefits of democ-
racy, interdependence, and international organizations, 1885–1992. World Politics 52(1):
1–37.
Conflict Management and Peace Science 29(3)
276
Oneal, J. R., F. H. Oneal, Z. Maoz and B. Russett. 1996. The liberal peace: Interdepen-
dence, democracy, and international conflict, 1950–85. Journal of Peace Research 33(1):
11–28.
Oneal, J. R., B. Russett and M. L. Berbaum. 2003. Causes of peace: Democracy, interde-
pendence, and international organizations, 1885–1992. International Studies Quarterly
47: 371–393.
Polachek, S. W. 1980. Conflict and trade. Journal of Conflict Resolution 24(1): 55–78.
Powers, K. L. 2003. Regional trade agreements as security institutions: Managing interna-
tional conflict through integration in security & natural resources. Paper presented at
Edinburgh Joint Sessions of Workshops, Workshop No. 9: Geography, Conflict and
Cooperation.
Powers, K. L. 2004. Regional trade agreements as military alliances. International Interac-
tions 30(4): 373–395.
R Development Core Team. 2007. R: A Language and Environment for Statistical Comput-
ing v. 2.7.0. (http://www.R-project.org)
Russett, B. M., and J. R. Oneal. 2001. Triangulating Peace: Democracy, Interdependence,
and International Organizations. New York: Norton.
Russett, B., J. Oneal and D. Davis. 1998. The third leg of the Kantian tripod for peace:
International organizations and militarized disputes, 1950–85. International Organization
52(3): 441–467.
Schiff, M., and A. L. Winters. 1998. Regional integration as diplomacy. World Bank Eco-
nomic Review 12(2): 271–295.
Schiff, M., and A. L. Winters. 2002. Regional cooperation, and the role of international
organizations and regional integration. World Bank Policy Research Working Paper
2872, July.
Schott, J. J. 2003. Free Trade Agreements: US Strategies and Priorities. Washington, DC:
Institute for International Economics.
Scott, J. 2000. Social Network Analysis: A Handbook. 2nd edn. London: Sage.
Stata Corporation. 2005. Stata Statistical Software v. 8.2. (http://www.stata.com)
Tomz, M., J. Wittenberg and G. King. 2003. Clarify: Software for Interpreting and Presenting
Statistical Results v. 2.1. (http://www.stanford.edu/~tomz/software/clarify.pdf)
Tucker, R. 1999. BTSCS: A Binary Time-Series–Cross-Section Data Analysis Utility v. 4.0.4.
(http://www.vanderbilt.edu/~rtucker/programs/btscs/)
Wallace, W. 1994. Regional Integration: The Western European Experience. Washington,
DC: Brookings.
Waltz, K. N. 1979. Theory of International Politics. New York: McGraw-Hill.
Ward, M. D., and P. D. Hoff. 2007. Persistent patterns of international commerce. Journal
of Peace Research 44(2): 157–175.
Ward, M. D., R. M. Siverson and X. Cao. 2007. Disputes, democracies, and dependencies:
A reexamination of the Kantian peace. American Journal of Political Science 51(3): 583–
601.
Wasserman, S., and K. Faust. 1994. Social Network Analysis: Methods and Applications.
Cambridge: Cambridge University Press.
Zimmern, A. E. 1936. The League of Nations and the Rule of Law, 1918–1935. London:
Macmillan.
EMILIE HAFNER-BURTON is an associate professor at the School of International Relations and
Pacific Studies and director of the School’s new Laboratory on International Law and Regulation.
She has served as professor of politics and public policy at Princeton University, where she held
Hafner-Burton & Montgomery: War, Trade, and Distrust
277
joint appointments in the Department of Politics and the Woodrow Wilson School for
International and Public Affairs. She has also served as research scholar at Stanford Law School
and fellow of Stanford’s Center for International Security and Cooperation (CISAC). Previously,
she was postdoctoral prize research fellow at Nuffield College at Oxford University, and was a
recipient of MacArthur fellowships at Stanford’s CISAC and an affiliate at the Center for
Democracy, Development and the Rule of Law at Stanford University.
ALEXANDER H. MONTGOMERY is an associate professor at Reed College. He has a BA in
Physics from the University of Chicago, an MA in Energy and Resources from the University of
California, Berkeley, and an MA in Sociology and a PhD in Political Science from Stanford
University. While at Stanford, his work was supported by a National Science Foundation Graduate
Fellowship. He has been a joint International Security Program/Managing the Atom Project
Research Fellow at the Belfer Center for Science and International Affairs, Kennedy School of
Government, Harvard University, and a post-doctoral fellow at the Center for International
Security and Cooperation, Stanford University. During 2012–3, he will be a Council on Foreign
Relations International Affairs Fellow in Nuclear Security.
Conflict Management and Peace Science 29(3)
278
