Intervention in Conflicts from a Network Perspective

Abstract

Whether neutral or on the side of a combatant, third-party states’ intervention in ongoing interstate conflicts is a triadic phenomenon which involves ties between a joining state and the two originators of the dispute. Existing studies on this topic have failed to fully capture the triadic nature of intervention, preferring instead to focus either on the joiner’s motivations or on the distinct dyadic relationships between joiners and the two separate combatants. Building on classic structural theories of triadic balance and on prior work by Maoz et al. (2007), in this article we address the triadic aspect of both mediation and “joining behavior”. The nature of the triadic relations among disputants and third parties influences not just the likelihood of intervention, but also the type of intervention. When triadic relations are unbalanced, third parties are more likely to intervene as intermediaries. On the contrary, when triadic relations are balanced, third parties are more likely to intervene in a partisan manner. We explore our main hypotheses by constructing a triadic data set that combines Corbetta and Dixon’s (2005) data on partisan third-party interventions and Frazier and Dixon’s (2006) data on neutral (intermediary) interventions in militarized interstate disputes with a friendship–hostility scale extracted from international events data (IDEA and COPDAB).

Keywords

intervention networks structural balance third parties triads

Introduction

In the 1960s, Rosenau (1969) complained that the term intervention was loosely used to refer to a variety of distinct behaviors ranging from peacekeeping to meddling with a state’s domestic affairs that were only marginally connected to one another. Since then scholars have made great strides in addressing Rosenau’s concerns. Research on third parties has developed into a variety of subfields separated from one another by considerations about: the type of third parties (e.g. states versus IGOs); third parties’ activities (e.g. mediation versus peacekeeping); third parties’ intentions (e.g. bringing peace versus promoting one side’s victory); the timing of the intervention (e.g. peace-making versus post-conflict peace-building); and the settings of the intervention (e.g. interstate versus intrastate conflicts).

All of these subfields share the notion that the phenomenon of intervention, regardless of its type, is largely shaped by the triadic relationship between an intervening third party and the two sides involved in a conflict.¹ Yet, the vast majority of studies on mediation or partisan joining behavior fails to describe the full extent of this triadic relationship, preferring to focus either on the motivations of the intervening third party alone, or on its separate dyadic relations with the two disputants. The potential issues associated with this practice are non-trivial (see Corbetta, 2010). The traditional theoretical and empirical models adopted in the discipline do not offer easy solutions to the problem of capturing a set of relations that simultaneously change along three dimensions and often move in opposite directions.

In this article we offer a non-traditional answer to this problem by resorting to theory and methods developed in social network analysis. We conceive of friendly and unfriendly relations as creating a network among international social actors—states, in our case—which is itself constituted by triadic “building blocks”. Structural balance theory provides a set of guidelines for developing expectations about triadic relationships. These guidelines lead us to hypothesize that neutral, mediatory interventions will more likely occur in association with unbalanced triadic relations—roughly, situations in which a third party has similar relations with the two conflict originators. Conversely, partisan interventions will be more prevalent when triadic relations are balanced—instances in which a third party has opposite relations with the two conflict originators.² We test these expectations by focusing on states’ intervention in ongoing Militarized Interstate Disputes (MIDs) (Jones et al., 1996; Ghosn et al., 2004) for the post-WWII period. We derive data on partisan interventions from Corbetta and Dixon (2005) and data on neutral, mediatory interventions from Frazier and Dixon (2006). We assess triadic relational ties using a friendship–hostility scale built from international events data (IDEA and COPDAB).

Neutral and Partisan Interventions

In the study of interstate conflict the term “intervention” has been and is used interchangeably to indicate at least two different types of third-party actions. On the one hand, intervention may refer to instances in which third-party actors become involved in a neutral fashion with the goal of delivering a peaceful resolution to the conflict. Mediation is the technique most often associated with neutral intervention, but other techniques such as facilitation, conciliation, and arbitration are sometimes considered. On the other hand, intervention may refer to instances in which third parties become involved in conflicts in a partisan fashion, by taking sides with one of the disputants, with the goal to steer the conflict to an outcome that is most favorable to the side they support. Partisan interventions are often referred to as “joining behavior” and are most commonly associated with conflicts in which a third party directly fights alongside its favored disputant. Other, less extreme partisan intervention techniques exist, however, and they range from expressions of diplomatic support to the provision of military support short of direct fighting—for example, military supplies and assistance. A common thread running through works on both mediation and joining behavior is that they are based on specific assumptions—sometimes stated, sometimes unstated—about the existence, direction, and strength of ties between third parties and the disputants.³

Research on both types of intervention is extensive, and an entirely separate article would be necessary to provide an adequate summary of both literatures. For the purpose of this article, it is important to notice that research on neutral intervention in interstate conflicts has largely—although not exclusively—revolved around three interrelated questions: the identity and motivations of mediators; the timing of the intervention; and the effectiveness of mediation. Works that have primarily dealt with the first big question have been concerned with comparisons between different states, and between states and other non-state actors—IGOs in particular—as potential mediators in interstate wars, disputes, and crises (see, for example, Young, 1967; Butterworth and Scranton, 1976; Touval and Zartman, 1985; Bercovitch, 1996; Dixon, 1996; Bercovitch and Schneider, 2000; Leng and Regan, 2003; Greig, 2005; Terris and Maoz, 2005; Beardsley et al., 2006; Frazier and Dixon, 2006; Greig and Diehl, 2006). With regard to states as mediators, most of these works stress the interveners’ underlying pragmatic interest in the peaceful resolution of the conflict and the existence of particular ties with both of the disputants. Similarities in regime ties—especially between democratic mediators and democratic combatants—and the importance of the mediators’ perceived neutrality has been stressed in this literature.

The mediators’ neutrality is of particular importance because it implies equidistance in relations between a third party and the dispute’s originators. Recent works, however, have questioned the notion that neutral mediators are more likely to be effective. Specifically, the insight from this perspective is that biased mediators who lean more heavily toward one of the disputants than the other are more credible when they relay information between the originators (see, for example, Kydd, 2003, 2006; Smith and Stam, 2003; Rauchhaus, 2006; Savun, 2008). Some scholars have pushed the “biased third parties argument” well past the informational role of mediators, proposing that powerful, biased third-party states are more effective peacemakers because they can impose settlements on the least favorite disputants (see Favretto, 2009). Although they are insightful, it can be argued that these works purposely blur the lines between mediation, deterrence, and coercion, essentially returning to the categorical confusion among types of intervention against which Rosenau (1969) argued. Most importantly, many of the works within this line of research remain theoretically and empirically ambiguous about the sources of the biased third parties’ preferences.

Research on partisan joining behavior has been equally concerned with third-party states’ motivations, and has focused on a set of assumptions about a joiner’s relationship with the disputants (see Yamamoto and Bremer, 1980; Bremer, 1982; Cusack and Eberwin, 1982; Bremer, 1992). This body of research revolves around two interrelated questions: Which states actually join? Which side of a dispute do they join? Because of the nature of these questions, assumptions about the relationship between joiners and dispute originators are somewhat more explicit than in the case of neutral mediation. In the context of joining behavior, the basic premises are that third-party states (1) have a strategic interest in the victory of one side in the conflict, and (2) will lean more toward one disputant than the other. Thus, both questions may be reinterpreted as asking which states have a biased interest toward one party in the dispute, and how strong that bias must be for a third party to actually jump into the fray.

Scholars interested in the expansion of conflicts in the geographical space have identified alliances and proximity as some of the main sources of third-party bias (see, for example, Most et al., 1989; Siverson and Starr, 1991). Studies from a strategic perspective, instead, have argued that joining behavior is highly dependent on the disputants’ expectations about the likelihood that one side’s allies—or the other side’s enemies—will actually intervene (see, for example, Kim, 1991; Smith, 1996; Werner, 2000; Fearon, 2002; Quackenbush, 2006; Yuen, 2009; Grant, 2010). The general conclusion is that partisan joining occurs when resolute third parties face resolute challengers in a classic extended deterrence scenario. Other scholars, however, have actually explored what factors shape third parties’ preferences so that some joiners are sufficiently biased to be resolute, while others are not. In a seminal article on the topic, Altfeld and Bueno de Mesquita (1979) suggested that partisan joiners derive utility from both aiding their favorite side in a conflict and steering the conflict toward their preferred outcome. The presence of uneven alliance ties between a third-party state and the disputants was key to the joiners’ utility functions.⁴ Successive works have shown how, in addition to alliances, asymmetries in the distribution of capabilities, (dis)similarities in regime type, asymmetry in trade relations, and economic interests also contributed to identifying which third-party states actually intervene and which sides they align themselves with (see, for instance, Werner and Lemke, 1997; Enterline, 1999; Reiter and Stam, 2002; Aydin, 2008; Corbetta, 2010).

Thus, research on partisan joining tends to model intervention as a function of a third party’s preferences toward the disputants, which are shaped by its relations with the two combatants. However, most works are limited in that they treat such relations as if they were independent of each other. Some scholars have lamented the inability of theoretical and empirical models to include all three relations among a joiner and two disputants (Quackenbush, 2006; Yuen, 2009; Corbetta, 2010). But even when researchers treat the ties between a third party and the two joiners as formally interdependent, their expectations often translate into empirical tests that assume independence between the models’ moving parts. It is when formal and traditional approaches start buckling under the weight of having to model simultaneous relationships between three actors that network analysis becomes useful, both theoretically and empirically. The ability to model and test the existence of relations among multiple actors is the essence of network analysis. As seen in the next section, network theorists have proposed, long ago, very simple models of triadic relations among social actors. In their simplicity, such models generate clear-cut, effective expectations about whether we should observe neutral or partisan state intervention in ongoing conflicts.

Balance Theory and Intervention

Social network analysis is often erroneously conceived as a set of methodological techniques for the descriptive and inferential analysis of social networks. In fact, in addition to providing a variety of statistical tools, social network analysis also comprises a vast array of theoretical approaches that explain and predict the existence of ties among a network’s constituent units and/or the structure of a network as a whole. Such theories have been developed and applied in various fields ranging from, of course, sociology (Wasserman and Faust, 1994) to communication (Monge and Contractor, 2003) to economics (Jackson, 2008), and they are becoming increasingly popular in political science and international relations. The advantages and limitations of network analysis as a theory of international relations have been explored elsewhere (for example, Hafner-Burton et al., 2009; Kahler, 2009) and are too extensive to review here. Rather, we are interested in exploiting the fact that network theorists have previously dealt with questions concerning triadic social relations, and we bring their knowledge to bear on the issue of intervention in ongoing conflicts.

Social networks are most often conceived as emerging from a series of dyadic relations, which are often treated as networks’ basic relational components. However, networks can be structured around more complex components and, while there is no theoretical limit to the complexity of such relational components,⁵ triadic relations are one step above dyads in order of compositional complexity. Knoke and Yang (2008: 13–14) refer to triadic relations as the third level of network analysis, and psychologists and other social scientists have found triadic structures useful for the analysis of “sentiment” ties—for example, liking, friendship, antagonism. Psychological and sociological theories about triadic relations are usually attributed to Heider (1946, 1958), Harary (1959), Cartwright and Harary (1956), and Harary et al. (1965). Heider’s (1946, 1958) “balance theory” is based on the notion that individuals seek consistency and balance in their interpersonal relations. At its most simplistic, balance theory can be summarized into four basic propositions about the relationships between one ego and two alters:

1. my friend’s friend is my friend;

2. my friend’s enemy is my enemy;

3. my enemy’s friend is my enemy;

4. my enemy’s enemy is my friend.

More formally, triadic relations can be graphically represented as in Figure 1, where the triangle’s vertices represent social actors and the triangle’s edges represent social relations. Relationships of friendship are noted with a positive sign, while relationships of antagonism are noted with a negative sign. The absence of a signed line implies neither like nor dislike (Wasserman and Faust, 1994: 222).

Figure 1.

Heider’s Basic Model

In Heider’s original formulation, all triadic relations having the same sign—either all positive or all negative—were considered balanced (Scott, 2000: 12). Cartwright and Harary (1956) further revised the theorem by specifying that triadic balance can be achieved if the algebraic multiplication among these relationships has a positive sign. That is, when the overall sign of a triadic cycle is positive, the triadic relationship is considered balanced (Wasserman and Faust, 1994: 224–225). More generally, Cartwright and Harary refer to a graph in which ties among actors are either positive or negative as a signed graph, or s-graph, and state: “an s-graph is balanced if and only if all paths joining the same pair of points have the same sign” (Cartwright and Harary, 1956: 286). The multiplication of all possible combinations in a triadic cycle produces eight possible social states.

Since this theorem is general, rather than limited to triads, Cartwright and Harary (1956) are considered responsible for the extension of Heider’s balance theory from the psychological, individual level of analysis to large-group dynamics (Scott, 2000). This extension is complemented by a second theorem, which they name the “structure theorem” and consider as the second necessary and sufficient condition of structural balance:

An s-graph is balanced if and only if its points can be separated into two mutually exclusive subsets such that each positive line joins two points of the same subset and each negative line joins points from different subsets. (Cartwright and Harary, 1956: 286)

Harary provided the mathematical proof of each theorem in a series of papers in the Michigan Mathematical Journal.⁶

Although we should always be careful when translating individual or group-level theories to states, the implications of balance theory for the study of conflict and cooperation at all levels are quite obvious (Scott, 2000). Structural balance theory has been used, for example, in anthropological studies of alliances in times of war (Wasserman and Faust, 1994: 221) and in the study of cooperation among politicians and community elites (Knoke, 1994). A few conflict scholars have also tapped the potential of this approach and have applied it in practice. Both Lee et al. (1994) and Crescenzi (2007), for example, use variants of Heider’s theorem and Cartwright and Harary’s elaboration of it to explain how states learn from the behavior of other states and how they develop a reputation for being hostile or friendly. Maoz et al. (2007), in particular, note that both realism and liberalism develop expectations about states’ behavior on the basis of indirect relations among states that reflect Heider’s structural balance approach. Using Militarized Interstate Disputes and Correlates of War data to develop matrices expressing indirect enmity or friendship among states, Maoz et al. (2007) successfully manage to explain the causes of relational unbalances and use the propositions of balance theory—enemy of my enemy, enemy of friend, and friend of my enemy—to predict the occurrence of alliances, interstate disputes, and war.

The work of Maoz and his associates on indirect triadic relationships among states has far-reaching implications for research on conflict intervention. As noted earlier, research on both neutral and partisan intervention in ongoing interstate conflict is predicated on stated or—more often—unstated expectations about third-parties’ behavior as a result of their preferences for the two disputants. But while most research on intervention fails to consider triadic relations simultaneously, balance theory speaks directly to the issue of structural relations among three actors. Let i and j be the combatants involved in an interstate dispute, and let k be a potential third-party intervener. Since i and j are fighting, their relationship is naturally one of enmity. Three triadic structures are of particular importance: those marked by a negative sign between i and j in Table 1. In graph form these relations can be visualized as in Figure 2.

Table 1.

Possible Combinations of Triadic Relations

k–i	k–j	i–j	Relation characteristic
+	+	+	Balanced
+	+	−	Unbalanced
+	−	+	Unbalanced
+	−	−	Balanced
−	+	+	Unbalanced
−	+	−	Balanced
−	−	+	Balanced
−	−	−	Unbalanced

Source: Khanafiah and Situngkir (n.d.).

Figure 2.

Balanced and Imbalanced Triadic Relations

Structural balance theory assumes that social actors (states) prefer harmony to tension and expects that they will strive to achieve or preserve balanced relations because they lead to harmony. In panel A of Figure 2, third party k experiences an unbalance while having friendly relations with both i and j. In these circumstances, third party k can restore balance either by antagonizing one of the disputants—turning their relationship from positive to negative—or by changing the relationship between the disputant themselves, turning it from “–” into “+” (Scott, 2000).⁷ Because it is commonly assumed in the intervention literature that taking active sides with one party is more costly than other conflict management techniques, the latter option is obviously preferable to the former. Thus, in the presence of structural unbalance with positive relationships between a third party and both combatants, we expect neutral, mediatory interventions to be more likely than non-neutral interventions or non-interventions.

On the contrary, when third party k has a friendly, positively signed relationship with one disputant but not with the other, the situation is one of balance, as shown in panels B and C of Figure 2. The “folk logic” prevailing in this case is the “enemy of my friend is an enemy”. Under these circumstances, structural balance theory predicts that k will have an incentive to preserve the structural status quo. The only way in which this can be achieved is by ensuring that the positive relationship with its preferred alter is not changed. This increases k’s incentive to take an active side in the ongoing conflict with its preferred disputant. As a result, in the presence of structural balance with a positive relationship between a third party and j and a negative relationship between a third party and i, we expect non-neutral, partisan interventions to be more likely than neutral interventions or non-interventions. We further expect third party k to intervene on the side of the disputant with whom it shares a positive tie. If k shares a positive tie with i, it will then be more likely to join on i’s side. It will join on j’s side if it shares a positive tie with j. Thus, structural balance theory generalizes and formalizes the string of disparate expectations—noted above—developed in the joining behavior literature about the side which third parties are expected to support when they enter a conflict.

Two sets of triadic relations pose a particular challenge. The first set of relations concerns situations in which third party k has negative relations with both i and j. Given the three negative signs in the triadic cycle, formally the situation is one of unbalance. With such triadic configurations we would expect third-party involvements in this context to be neutral, intermediary interventions. However, the i← (–) →k← (–) →j cycle is inconsistent with structural balance theory, as per the structural theorem outlined earlier, and it produces no specific prediction either concerning the behavior of social actors in general or concerning third-party intervention in particular.⁸ The unbalance a third party experiences when two enemies are fighting is not of the same nature as the unbalance experienced when two friends are fighting. Therefore, we expect the likelihood of any kind of intervention in this triadic configuration to be considerably smaller than that of other situations of unbalance.

The second set of anomalous triadic relations occurs when a third party has no relations with either or both disputants. The problem here is one of transitivity. In general, a triad is transitive if when i→j and j→k, then i→k (where → indicates the presence of a tie). Triads where more than one of these conditions are not satisfied are intransitive. Cartwright and Harary (1956) and Harary et al. (1965), however, call these triads “disjointed” and assert that they do not create particular problems. Where there is no relationship between k and i, or between k and j, a triad can be considered “vacuously balanced” (Cartwright and Harary, 1956), and balance can be assessed simply by multiplying the two present ties (Harary et al., 1965). If the resulting sign is positive, the triad is vacuously balanced. If the sign is negative, it will be vacuously unbalanced. Triads such as these—which Wasserman and Faust (1994: 243–246) call “vacuously transitive”—generate expectations similar to those outlined above. If, for example, k has no relationship with j but its tie with i is positive, the triad is weakly unbalanced, and k will be more likely to intervene in a neutral fashion. If, on the other hand, k has no relationship with i but its tie with j is negative, the triad is weakly balanced, and k will be more likely to intervene in partisan fashion on the side of i.⁹ Where k has ties to neither i nor j, third party k will experience no structural pressure toward any kind of intervention.¹⁰ These hypotheses are summarized graphically in Figure 3. In the following section, we outline how ties between states have been measured and we proceed to testing these hypotheses about neutral and partisan intervention.

Figure 3.

Structural Balance Hypotheses About Intervention

Social Network Analysis and the Study of Intervention

Because we are interested in exploring the predictions of structural balance theory on both mediatory and partisan intervention in conflicts, we need to identify comparable data on the two types of interventions, recorded with similar procedures and for the same set of interstate conflicts. Frazier and Dixon’s (2006) data set on neutral interventions in militarized interstate disputes (MIDs) for the 1946–2001 period and Corbetta and Dixon’s (2005) data on partisan interventions in MIDs for the same time period provide the two dependent variables needed for this study. In addition to mediation, Frazier and Dixon’s (2006) data record information on a range of intervention techniques—from verbal appeals to actual mediations—in which third parties do not explicitly take sides with either MID originators and are, at least on the surface, interested in the peaceful resolution of the dispute. Corbetta and Dixon’s (2005) data, instead, record interventions in ongoing MIDs in which third-party states explicitly take sides with one of the dispute originators, and their actions range from diplomatic expressions of support or opposition to fighting on the side of the favorite combatant.¹¹ The dichotomous occurrence of a third-party state’s mediatory intervention regardless of intervention technique, as recorded by Frazier and Dixon (2006), is the dependent variable for the structural unbalance hypothesis. The dichotomous occurrence of a partisan intervention—again, regardless of technique—as recorded by Corbetta and Dixon (2005), is the dependent variable of interest for the structural balance hypothesis. Because partisan interventions are expected to occur on the side with which k has a positive relationship, we further dichotomize this dependent variable into intervention in favor of i or in favor of j.¹² Based on the two data sets, there are 360 instances of neutral interventions and 1426 instances of partisan interventions (525 on the side of i and 901 on the side of j) in MIDs between 1946 and 2001.

In both data sets the dispute-triad constitutes the basic unit of analysis, and the triadic structure easily lends itself to the study of the effects of triadic balance. However, because both data sets only record realized interventions, it was necessary to supplement them by including potential intervening states and disputes in which no intervention occurs, in order to avoid obvious selection biases. The two data sets were first merged with the a list of post-1946 dispute-dyads from Maoz’s Dyadic MIDs data set (see Maoz, 2005), and then augmented with a list of all potential third-party states in a given dispute. The unit of analysis in the resulting data set is list of states members of the international system for a given year to produce all possible third party k-disputant i-disputant j triads for each dispute.¹³ The resulting data set contains 274,624 triads, 98,664 of which are politically relevant.¹⁴

Structural balance theory leads to the expectation that intervention in interstate disputes is the expression of underlying networks of cooperative and antagonistic relations, the basic constituent unit of which is triads. In order to identify networks of cooperative and antagonistic relations we employ data from both the Conflict and Peace Data Bank (COPDAB) (Azar, 1980) and Virtual Research Associates’ Integrated Data for Events Analysis (VRA IDEA) (Bond et al., 2003). COPDAB data cover the 1948–1978 period and are derived from about 70 different sources. IDEA data, instead, cover the 1990–2004 period and are based on machine-coded Reuters news feeds.¹⁵ In both data sets, the events take the initiator–action–target form, and each action is weighted according to a scale measuring relative levels of conflict and cooperation. Cooperative interactions receive a positive score, while negative interactions receive a negative score.¹⁶ Because states engage in a variety of simultaneous interactions with one another, we aggregate event data scores at the yearly level.¹⁷ Event data also have a high level of reciprocity, and reciprocated relations generate an addition set of expectations that extend beyond the scope of this analysis (see Cartwright and Harary, 1956). Although it may be interesting to assess the impact of the relative strength of triadic ties, structural balance theory—at least in Cartwright and Harary’s original formulation—focuses on the sign of relational ties rather than on their relative strength. To preserve consistency with the theoretical framework, we remove directionality in the data by applying the weakest link principle (Dixon, 1993, 1994), thus considering the lowest level of interactions between two states as the baseline for their relationship of friendliness/enmity for a given year.¹⁸

In summary, for each year covered by the COPDAB and IDEA data sets we produce either a positively signed (friendly) or negatively signed (unfriendly) score for each third party–disputant relationship in a triad. All relations between disputants i and j are by definition negatively signed. These positive and negative relations are used to estimate dichotomous measures of triadic balance or unbalance. When COPDAB or IDEA scores between a potential joiner and the disputants are both positive, we are in the presence of an unbalanced relation. When instead COPDAB or IDEA scores between a potential joiner and the disputants have opposite signs—one is negative and the other is positive—we record a balanced triad. When no interaction is recorded, we assume indifference between two states and assign a score of zero.¹⁹ Triads where either the k↔i or the k↔j relationship is zero are vacuously balanced or unbalanced, as indicated earlier, depending on whether the non-missing relationship is positive or negative. Triads in which both k↔i and k↔j are zero are intransitive and, therefore, belong to the reference group in the empirical tests that follow.

Because structural balance or unbalance in a triad may have nothing to do with the occurrence of neutral mediation or partisan joining, we control for the three factors most commonly found in the conflict literature to affect intervention: capabilities, alliances, and similarity in regime type. We strive to operationalize these control variables so that they are triadic measures rather than attributes of dyads or of the component units. We use the Alliance Treaty Obligations and Provisions data set (Leeds et al., 2000, 2002) to control for alliance ties. Specifically, we create two separate dichotomous triadic measures of structural balance for alliance ties. If k has alliance ties with both i and j, the triad is unbalanced with regard to alliances. This may affect the likelihood that k will intervene in a mediatory fashion. For the test of the structural balance hypothesis about partisan interventions, a triad is coded as balanced with regard to alliances if a third party has an alliance with i but not with j, and vice versa. We address the notion that similarity in regime type matters in interventions by creating two structural balance variables using the well-known Polity Democracy–Autocracy score (Marshall and Jaggers, 2002). We code a positive tie in a triad if two states are mature democracies (greater than or equal to 6 on the Polity DEM-AUT score) or mature autocracies (less than or equal to −6 on the Polity DEM-AUT score). Thus, for the hypothesis about mediatory interventions, a triad is coded as unbalanced—a dichotomous measure—according to regime type if both the k↔i and k↔j dyads are mature democracies, or if both are mature autocracies. For the hypothesis about partisan interventions, a triad is coded as balanced—a dichotomous measure—according to whether k↔i are mature democracies (or autocracies) but k↔j are not, or vice versa. In order to derive a triadic measure of capabilities, we employ the Correlates of War’s Cumulative Index of National Capabilities (CINC) scores (Singer et al., 1972; Singer, 1988). Research on intervention suggests that intervention is more likely to occur when a third party has sufficient capabilities, relative to the disputants, to alter the balance of power between them or to enforce the terms of a mediated agreement. This is usually estimated with a triadic capability ratio obtained by dividing a potential joiner CINC score by the total amount of capabilities in the triad. We follow the same operational procedure.

Unlike most studies of social networks, in this article we are not explicitly interested in modeling the dynamics that give rise to a specific network structure. Rather, following structural balance theory, we are interested in how network structure affects social actors’ behavior—that is, a third party’s decision to mediate, intervene in, or abstain from an ongoing conflict. In short, while in many social network analyses the configuration of a network is the dependent variable of interest, in this study the social network structure is the main independent variable. In addition to the application techniques commonly used for the inferential analysis of networks, such as Exponential Random Graph-family models, our context is constrained by other practical limitations. For the inferential portion of our analysis we depart from network methodology and resort to the more familiar logit framework. Specifically, we employ multinomial logit analysis. We recode the dichotomous dependent variables described above into a single dependent variable with three mutually exclusive categories: non-intervention, mediation, and partisan intervention. The results of the analysis are presented in the following section.

Results

Before moving to the inferential analysis, we provide a simple visual representation of the network structures inherent in the data. We present a single network that is illustrative of the intervention dynamic we are exploring. Figure 4 shows the network of friendliness/enmity ties created by COPDAB data for the 1962 Sino-Indian war, a conflict that attracted a sizable number of both neutral and partisan interventions. Countries are represented by circles or diamonds and are labeled with Correlates of War country abbreviations. The circles representing the two disputants—India (IND) and China (CHN)—have been blown up in size compared with third-party states. During the Sino-Indian war of 1962 all third parties that joined in partisan fashion sided with India—the United States of America (USA) Canada (CAN), the United Kingdom (UKG), France (FRN), Malaysia (MAL), Australia (AUL), and New Zealand (NEW)—and they are shown in the graph by diamond-shaped nodes. Countries that intervened as neutral intermediaries—the Soviet Union (RUS), Sweden (SWD), Guinea (GUI), Tanzania (TAZ), Egypt (EGY), and Syria (SYR)—are represented, instead, by smaller circles. The numeric values associated with each tie in the graph represent the minimum conflict-cooperation score between two countries. Visual inspection of the graph suggests that in most cases, the expectations of structural balance theory are supported. Four of the six neutral third parties (represented by circles) have unbalanced or vacuously unbalanced relations with the two disputants, having positive ties with both India and China. The exceptions are Syria, which should intervene in partisan fashion, and Guinea, which should not intervene at all. Similarly, six of the seven cases of partisan intervention occur in conjunction with balanced or vacuously balanced triadic relations—these “biased” third parties have a positive relation with India but a negative or null relation with China. The only wrong prediction occurs in the case of Australia, which has friendly ties with both disputants and should intervene in intermediary fashion.

Figure 4.

Structural Balance Relations and Intervention in the 1962 Sino-Indian War

While possibly telling of the dynamic inherent in structural balance, this is but one dispute in the entire post-WWII period. The relationship between structural (un)balance and specific types of intervention may be spurious because triadic balance/unbalance may itself be the result of other factors that produce friendly or unfriendly relations among states. We further test the relationship between triadic balance/unbalance and intervention using multinomial logit analysis, which allows us to control for triadic alliance balance/unbalance, triadic regime balance/unbalance, and triadic capability ratio. The results of two sets of multinomial logit regressions—one for the years covered by the COPDAB data and one for the years covered by the IDEA data—are reported in Tables 2 and 3. The dependent variable’s categories are non-intervention, mediation, and partisan intervention. In all models, we employed robust standard errors and clustered on triads. Because in the case of participatory interventions the “direction” of the intervention matters as much as its occurrence, we further test whether balanced triadic relations in favor of i lead to partisan intervention on the side of i and, conversely, whether balanced triadic relations favoring j cause a joiner to actually support j. This third set of results is shown in Table 4.

Table 2.

Structural Unbalance and Neutral Interventions, 1948–1978 and 1990–2001

	1948–1978		1990–2001
	COPDAB		IDEA
	All Triads	Pol. Relevant	All Triads	Pol. Relevant
Mediation
Unbalanced triad	1.331***	0.996***	2.728***	1.921***
	(0.253)	(0.272)	(0.485)	(0.528)
Alliance ties	1.094**	0.648	0.555	0.159
	(0.409)	(0.427)	(0.543)	(0.481)
Regime similarity	−0.0571	−0.115	−0.916	−0.748
	(0.359)	(0.394)	(0.512)	(0.570)
Capability ratio	2.362***	2.787***	3.926***	4.990***
	(0.540)	(0.447)	(0.729)	(0.944)
Constant	−7.159***	−6.275***	−9.251***	−9.122***
	(0.450)	(0.425)	(0.460)	(0.779)
Intervention
Unbalanced triad	−0.203	−0.713***	1.602***	0.848**
	(0.220)	(0.212)	(0.342)	(0.278)
Alliance ties	1.085***	0.533	−0.188	−1.272
	(0.266)	(0.287)	(0.782)	(0.740)
Regime similarity	−0.318	−0.427	−3.234***	−3.000**
	(0.306)	(0.308)	(0.902)	(0.988)
Capability ratio	2.464***	3.170***	1.792*	2.100***
	(0.458)	(0.398)	(0.843)	(0.627)
Constant	−5.716***	−4.918***	−6.917***	−5.900***
	(0.408)	(0.407)	(0.667)	(0.615)
N	107815	41901	98114	33947
pseudo R2	0.072	0.140	0.092	0.121
AIC	12474.5	8387.7	2987.2	2080.8

Standard errors in parentheses. *p < 05; **p < 01; ***p < 001.

Table 3.

Structural Balance and Partisan (Non-Neutral) Interventions, 1948–1978 and 1990–2001

	1948–1978		1990–2001
	COPDAB		IDEA
	All Triads	Pol. Relevant	All Triads	Pol. Relevant
Mediation
Balanced triad	1.052	0.648	0.910	0.723
	(0.577)	(0.616)	(0.518)	(0.505)
Alliance ties	1.303***	0.597*	0.842*	0.257
	(0.312)	(0.282)	(0.428)	(0.444)
Regime similarity	0.654	0.353	0.251	0.00616
	(0.410)	(0.452)	(0.642)	(0.727)
Capability ratio	2.751***	3.033***	4.121***	4.974***
	(0.493)	(0.422)	(0.670)	(0.769)
Constant	−7.626***	−6.557***	−9.245***	−8.896***
	(0.371)	(0.377)	(0.449)	(0.560)
Intervention
Balanced triad	2.083***	1.575***	0.657**	0.443
	(0.200)	(0.168)	(0.231)	(0.236)
Alliance ties	0.771***	0.0494	2.358***	1.763***
	(0.206)	(0.190)	(0.223)	(0.219)
Regime similarity	0.596	0.484	1.080***	1.087***
	(0.399)	(0.391)	(0.287)	(0.226)
Capability ratio	2.627***	3.085***	2.107**	2.141***
	(0.411)	(0.362)	(0.720)	(0.587)
Constant	−6.374***	−5.415***	−8.260***	−7.285***
	(0.312)	(0.311)	(0.509)	(0.521)
N	107815	41901	98114	33947
pseudo R2	0.126	0.165	0.135	0.152
AIC	11748.9	8142.4	2846.6	2007.0

Standard errors in parentheses. *p < 05; **p < 01; ***p < 001.

Table 4.

Structural Balance and the Direction of Partisan Interventions, 1948–1978 and 1990–2001 (Politically Relevant Triads)

	1948–1978		1990–2001
	COPDAB		IDEA
	On side i	On side j	On side i	On side j
Mediation
Balance towards i	−0.490		0.0787
	(0.612)		(0.922)
Balance towards j		1.258		1.266*
		(0.708)		(0.634)
Alliance ties	0.661*	0.599*	0.263	−0.147
	(0.261)	(0.273)	(0.435)	(0.291)
Regime similarity	0.366	0.273	0.0141	0.138
	(0.427)	(0.439)	(0.716)	(0.719)
Capability ratio	3.017***	2.923***	4.979***	4.737***
	(0.383)	(0.392)	(0.756)	(0.685)
Constant	−6.499***	−6.504***	−8.842***	−8.699***
	(0.400)	(0.345)	(0.550)	(0.494)
Intervention
Balance towards i	2.172***		0.707
	(0.300)		(0.516)
Balance towards j		2.448***		1.290**
		(0.216)		(0.466)
Alliance ties	0.447	−0.156	2.699***	1.471***
	(0.286)	(0.221)	(0.657)	(0.244)
Regime similarity	0.604*	0.493	1.900***	0.745**
	(0.293)	(0.478)	(0.391)	(0.253)
Capability ratio	3.486***	2.913***	2.160***	1.847**
	(0.389)	(0.338)	(0.496)	(0.662)
Constant	−6.994***	−5.848***	−9.732***	−7.239***
	(0.293)	(0.344)	(0.689)	(0.561)
N	40711	39418	33125	32629
pseudo R2	0.172	0.161	0.209	0.126
AIC	4475.4	5975.0	1018.2	1559.2

Standard errors in parentheses. *p < 05; **p < 01; ***p < 001.

The hypotheses derived from structural balance theory are largely supported, with a few exceptions. Overall, the effects of triadic (un)balance on neutral mediation and partisan intervention are more clearly detectable for the Cold War era, covered by the COPDAB data, than for the post-Cold War period, covered by the IDEA data. Table 2 indicates that unbalanced triadic relations are positively associated with neutral, mediatory interventions. These results are more clearly discernible for the 1948–1978 period when triadic unbalance, as expected, is also negatively or insignificantly related to biased interventions. However, for the post-Cold War period, when we used IDEA data to generate the structural unbalance variable, the triadic unbalance variable remains statistically significant but fails to distinguish between neutral and non-neutral intervention. The number of mediatory and partisan interventions for the 1990–2001 years is relatively low compared to the Cold War era, and simple cross-tabulation (not shown here) reveals the presence of an almost equal number of mediation attempts and partisan interventions in correspondence with triadic unbalanced relations.

Table 3 shows that after accounting for the set of covariates commonly associated with third parties’ motivations for intervening in ongoing conflicts, triadic balance remains a significant predictor of partisan interventions for both the 1948–1978 period and for the 1990–2001 period. As expected, structural balance only predicts biased interventions, as its effect on the likelihood of neutral mediation is not statistically significant. Triadic balanced relations fail to correctly predict partisan intervention in the post-Cold War era only when politically relevant triads are taken into account. A possible explanation lies in the fact that the post-Cold War years no longer see proxy wars involving the major powers as third parties.

As shown in Table 4, the expectation that third parties will try to preserve structural balance by intervening in biased fashion is supported for the most part when the direction of structural balance is considered. When structural balance points to a positive relationship between third party k and disputant i, a biased intervention on the side of it is more likely to occur than mediation or intervention on the side of j. Conversely, when a positive relationship exists between k and j, intervention is more likely to be on the side of j. Even here, however, the patterns of biased third-party behavior in the post-Cold War period, covered by the IDEA data, are not as clear-cut as the patters detected in the 1948–1978 period, covered by the COPDAB data. The results in Table 4 are based on politically relevant triads, and in this case structural balance fails to predict significantly the direction of partisan intervention for the 1990–2001 period. Results not shown here indicate that when the analysis is conducted on all triads, triadic balance correctly predicts the direction of intervention for both side i and side j even in the post-Cold War era.²⁰

It is worth pointing out that these results are considerably “watered down” by the fact that the reference category (no intervention) is massively larger than the mediation and partisan intervention categories, even when only politically relevant triads are taken into account. Although the results are not reported here, we conducted the same analysis by changing the reference category for the multinomial logit model. When partisan intervention is the reference category for unbalanced covariates, or when mediation is the reference category for balanced covariates, the multinomial logit models detect more clearly the different effects of structural balance and structural unbalance, respectively.

The results described up to this point are estimated only with triads that are fully (un)balanced. We cannot fully corroborate the expectations of structural balance theory in the case of vacuously balanced and vacuously unbalanced triads. In Table 5, the models test the effect of vacuously unbalanced and vacuously balanced ties on mediation and intervention for the 1948–1978 and 1990–2001 time periods. The results from Table 5 indicate that vacuous balance correctly predicts partisan intervention during the 1948–1978 period; however, vacuous unbalanced relations fail to distinguish between mediation and partisan intervention regardless of the time period we consider. At best, what can be said is that triadic relations must be completely transitive for the model to clearly predict which type of intervention, if any, will be observed. The prediction derived from structural balance theory that a signed relation will weigh more than no tie or a neutral tie in vacuous triads may not entirely apply in international relations. However, the fact that vacuously balanced triadic ties correctly predict biased interventions lead us to suspect that these results are driven by the nature of the COPDAB and IDEA data, where missing data and neutral conflict/cooperation scores are likely to mask the existence of potentially hostile relations.²¹

Table 5.

Vacuous (Un)balance, Mediation, and Intervention, 1948–1978 and 1990–2001

	1948–1978		1990–2001
	COPDAB		IDEA
Mediation
Vacuously unbalanced triad	0.709***		0.904***
	(0.207)		(0.245)
Vacuously balanced triad		−0.299		0.00528
		(0.474)		(0.919)
Controls not shown
Intervention
Vacuously unbalanced triad	0.757***		1.525***
	(0.129)		(0.336)
Vacuously balanced triad		0.563**		0.432
		(0.215)		(0.551)
Controls not shown
N	107815	107815	98114	98114
pseudo R2	0.076	0.090	0.090	0.135
AIC	12412.0	12227.6	2995.8	2848.8

Standard errors in parentheses. *p < 05; **p < 01; ***p < 001.

The aforementioned exceptions should not divert attention from the overall picture provided by these results. In all cases, the multinomial logit equations are statistically significant and fit the data fairly well. Preserving balanced triadic relations or restoring balance when it is lost appears to be important to third-party states. The logic outlined by Maoz et al. (2007) with regard to conflict initiation is confirmed when conflict expansion and conflict management are considered. When a conflict involves a friend and an enemy, third-party states are more likely to intervene in a partisan fashion. However, when a conflict involves two friends, triadic balance is more likely to be restored through neutral mediation.

With regard to the control variables, capability ratio is an unsurprising predictor of both neutral and partisan intervention. Strong third parties are likely to have broader foreign policy interests and, therefore, to be more likely to intervene in the first place. Strong third parties also have the wherewithal to coerce or enforce mediated agreements when they intervene neutrally, or to change the military outcome of an ongoing conflict when they intervene in a biased fashion. The impact of alliance ties is somewhat less clear than usually argued in the alliance literature. When a third party is allied with both disputants, the alliance triad is structurally unbalanced and the third party is not significantly more likely to intervene in either neutral or partisan fashion. Third parties are more likely to intervene in partisan fashion when they are allied with only one of the combatants, although this tendency is not clearly distinguishable from the tendency to attempt mediation. It is possible that allies with a friend involved in a dispute will attempt mediation before having to intervene on the partner’s side. Somewhat surprisingly, triadic relations of regime similarity fail to predict both neutral and partisan interventions. This finding clashes with arguments about the tendency to accept mediation from states with similar traits and with arguments about the importance of regime similarity in the decision to intervene in biased fashion (Werner and Lemke, 1997; Corbetta, 2010). Likely, conflating similarity between democracies with similarity between autocracies does a disservice to arguments about the importance of democracy in providing assistance in conflict and in offering and accepting mediation. Offering and accepting neutral or partisan interventions may not be part of the “toolkit” of authoritarian leaders, and considerably different results may emerge if only similarity between democratic states is taken into account.

Nonetheless, one cannot rule out that the effects of variables commonly associated with mediation and intervention in dyadic studies disappear when we move to a different level of analysis. The vast majority of previous studies have managed to account for only dyadic relations for most variables of interest in the phenomenon of intervention. The strength of structural balance theory, instead, is that it allows us to develop hypotheses that account for the relationships of three states simultaneously: a joiner and two disputants. Network analysis and its theory of structural balance move us past the limitations common to the issue of intervention, making empirical comparison with previous explanations difficult at best. In most cases, network analysis presupposes that dyadic relations constitute the foundational component of a network. However, when there are strong theoretical bases for assuming that structures more complex than a dyad are the basic unit of a network, comparison with non-structural theories of international peace and conflict becomes more challenging.

Discussion and Conclusions

Research on third parties and intervention in conflict has grown considerably in recent years. This growth reflects two simultaneous efforts. One is the effort to better understand conflict dynamics beyond the important consideration that war and peace are the result of the interactions between two countries. The inception and termination of crises, disputes, and wars are more fully understood when we consider expectations about the behavior of third-party states and third-party states’ own stakes in ongoing conflicts. The second effort concerns international relations scholars’ attempts to extend the formal and empirical examination of international political processes beyond the interstate level of analysis.²² Most theoretical and empirical efforts to explore third parties’ conflict management or the expansion of interstate conflict, however, have remained anchored to the premise that relationships between third parties and conflict originators can be decomposed into a series of dyadic relations that can be explored with familiar theoretical and empirical approaches. The implication is that higher order dependencies are either not relevant to the intervention process or negligible. This premise is rarely made explicit, yet it creates limitations in the study of both mediation broadly defined and joining behavior.

We have argued that social network analysis tools can be useful in the study of neutral and non-neutral interventions because they explicitly incorporate triadic relations both at the theoretical and empirical levels. Structural balance theory is founded on the notion that triadic relations are more than the simple sum of a set of three dyadic relations. The theory expects social actors to seek consistency in their triadic relations and, therefore, it easily lends itself to the development of hypotheses about neutral and partisan intervention. Triadic (un)balance is a structural attribute which we have attempted to capture with event data (COPDAB and IDEA) that may reflect underlying friendly or antagonistic relations within sets of three states. While providing some support for previous findings at the dyadic level concerning the importance of well-known factors (for example, capabilities) in the study of interventions, the present results highlight how prior research may have overlooked a key component of the phenomenon of intervention: namely, the triadic dimension. As expressed in relations of friendship and enmity, triadic relations of structural balance are neither the simple summation of a set of dyadic links nor the spurious outcome of a known set of antecedent causes (e.g. regime-type similarities or alliance ties). Because state-of-the-art research on both mediatory (e.g. Favretto, 2009) and partisan (e.g. Yuen, 2009) intervention seems to emphasize strongly the importance of “cut-points” in a third party’s preferences toward both disputants in ongoing conflicts, scholars of neutral and non-neutral conflict management techniques ought to pay growing attention to the triadic level of analysis.

We believe, of course, that social network analysis provides a promising avenue of research on the topic of third parties’ conflict management. However, this approach still has important weaknesses, some of which are reflected in the study. At the theoretical level, we purposely avoid the debate, which has been handled elsewhere (e.g. Hafner-Burton et al., 2009), of whether network theories developed in other social sciences can be applied wholesale to international relations. Empirically, considerable work remains to be done. As seen in this article, one of the major weaknesses of social network approaches is that they still struggle with handling valued and signed relationships. Even when they have access to continuous relational data, social network analysts are sometimes forced to transform them into discrete-level (dichotomous) data because of the difficulty of handling valued and signed matrices. The COPDAB and IDEA event data used in this article provide us with continuous measures of friendship and antagonism, but we have been forced to treat such measures essentially as dichotomous, focusing only on their direction rather than on their magnitude. Thus, the amount of information we have been able to bring to bear to address the problem of triadic relationships is limited compared with the wealth of data available. This problem is bound to often frustrate researchers in a data-rich subfield such as international relations.

We think, nonetheless, that this is an important line of research to pursue, especially with regard to questions concerning third parties’ conflict management. As just stated above, recent research about intervention stresses the importance of third parties’ bias—and not just informational bias—toward the disputants both in cases of neutral and non-neutral intervention. What we may call the relative social distance between a mediator or joiner and the conflict originators is said to affect the occurrence, timing, and effectiveness of the intervention. Existing studies incorporate such directional “bias thresholds” but fail to quantify them explicitly. This leaves several fascinating questions without answers. For example, what is the actual level of friendship between k and i and between k and j at which third party k stops mediating and starts coercing in neutral interventions? Further, how much does third party k have to lean toward i, rather than j, before triadic relations switch from unbalance to balance and the intervention shifts from neutral to partisan? As we have shown in this article, the data to measure such thresholds exist, and social network analysis provides interesting ways for doing so. These are questions that require an answer at the triadic level, and we plan to direct our future efforts to address them.

Footnotes

Acknowledgements

The authors would like to thank Zeev Maoz, Glenn Palmer, Xun Cao, Yon Lupu, Mike Ward, Charles Boehmer, and the anonymous reviewers for their helpful comments, and Bill Dixon and Derrick Frazier for sharing their data. All errors in the manuscript are our own.

1

Many interstate and intrastate conflicts involve more than two sides and more than two originators. However, the vast majority of international conflicts begin as confrontations between two sides. For the purpose of this article we focus on interstate conflicts that start as dyadic affairs. We break conflicts involving more than two originators into their dyadic sub-components.

2

We provide a more precise definition of balanced and unbalanced triadic relations in the following sections.

3

The term “intervention” has been used in the same way in the study of intrastate conflicts. As originally noted, intervention is additionally used to refer to instances in which an external state intervenes in the domestic politics of another state, often in the context of a domestic or international crisis. We limit the scope of this article to interventions in interstate conflicts and employ the term accordingly.

4

See also Bueno de Mesquita (1989) and Joyce (2008) for an alternative agent-based approach, based on , on belligerent intervention and the expansion of war.

5

At the upper limit, one can treat the entire network as the basic unit of analysis.

6

7

Structural balance theory assumes that the sign of the relationships between k and i and j matters more than the relative strength of those relationships. If, for example, k has a positive relationship with both i and j (while i and j are fighting), the relationship remains unbalanced even if k’s friendship with i is stronger than its friendship with j.

8

According to the structural balance theorem, networks must be divisible into two subgroups for researchers to be able to determine whether they are balanced or unbalanced. Actors in an all-negative triad must be placed in three different subgroups. This is compatible with the theory of clusterability, but not with structural balance. On this issue, see , especially chapter 6.

9

Intuitively, third party k’s preferences are for positive relationships with a disputant over no relationships over negative relationships. In cases of vacuously unbalanced triads—for example, (i← (Ø) →j)—fighting on the side of j, rather than mediating, may generate future negative relationships with i. Triadic balance is more easily restored by either creating a positive relationship with i or by turning the conflictual relationship between i and j into a positively signed one, or both. Instead, in cases of vacuously balanced triads—for example, (i← (Ø) →j)—balance can be more easily preserved by creating a positive relationship with i by fighting on its side than by turning the negative relationship with j into a positive one. This is consistent with both the “the enemy of my enemy is my friend” folk logic and with findings that the motivation to oppose an enemy weighs more than the incentive to help a friend in partisan joining (see Corbetta, 2010).

10

While structural balance theory deals with structural pressure, we cannot rule out that in cases of no relationship with either disputant third party k may experience larger societal pressures to intervene neutrally because it is perceived as being unbiased.

11

When the third-party actions reach the militarized threshold, data largely overlap with the MID data set, which also records militarized third-party interventions.

12

These correspond to side A and side B, respectively, in data.

13

In both data sets disputes with multiple originators are broken into a series of originating dyads. Also, in both data sets some of the recorded intervention techniques—for example, neutral requests for a cease-fire or partisan diplomatic expressions of support—can be carried out even by small, distant states. Placing artificial restrictions on the type of third-party states to be included may introduce bias in the results. Thus, with the exception of micro-states, we first place no selection restriction on the third-party states to be included. We later repeat the analysis by including only politically relevant third parties as a robustness check.

14

Political relevance was determined by whether (1) a third party is a major power state, (2) the dispute is a major-power dispute, or (3) a third party is geographically contiguous to either disputant.

15

The unavailability of event data for the 1979–1989 period leads to a loss of 303 cases of partisan interventions and 49 cases of mediation as recorded in the Corbetta and Dixon (2005) and data sets, respectively.

16

See Goldstein (1992) and for a discussion of the conflict-cooperation scales in the two data sets and their comparability.

17

See for a discussion of the pros and cons of event data aggregation.

18

This simplification makes the inherent assumption that for a dyad to be coded as cooperative, cooperation must be reciprocated. However, the presence of unilateral conflict is sufficient to consider the dyad hostile.

19

For a far more detailed treatment of the limitations and biases of all event data in general, and the IDEA and COPDAB data sets in particular, see King and Lowe (2003) and .

20

The 1990–2000 decade is characterized by many disputes involving Russia and the former Soviet republics or countries in the former Soviet bloc. Few of these disputes draw any partisan intervention. Similarly, many disputes during this decade involve Iraq, the United States, and Great Britain. By coding rules Iraq is almost always coded as being on side j (side B) of a dispute, and this may affect the results to some extent.

21

Even in the case of vacuously (un)balanced triads, however, the displayed results are greatly affected by the fact that the no-intervention reference category is so large. When the reference category is changed, structural (un)balance clearly distinguishes between mediation and intervention.

22

This is not to say that international relations researchers have neglected other levels of analysis in the empirical study of international politics. Efforts to explore systematically structural international processes abound. Yet, in terms of sheer numbers, they tend to be fewer than the more prevalent studies that emphasize dyads and their relations. See for a discussion of the implications of scholars’ focus on dyads in the study of international conflict.

Renato Corbetta is an Associate Professor of Political Science at the University of Alabama-Birmingham. His research focuses on interventions in conflict, on networks in international relations, and on major powers and status in international politics.

Keith A. Grant received his PhD in Political Science from the University of Arizona in 2010, and is currently a Senior Political Scientist with Guardian Six Consulting. Keith’s primary research interests lie at the intersection of international security and global governance, where he focuses on how latent features of the international system influence the behavior and security of states.

References

Altfeld

Michael F.

de Mesquita

Bruce Bueno

. 1979. Choosing side in war. International Studies Quarterly 23(1): 87–112.

Aydin

Aysegul

. 2008. Choosing sides: Economic interdependence and interstate disputes. Journal of Politics 70(4): 1098–1108.

Azar

Edward E.

1980. The Conflict and Peace Data Bank (COPDAB) project. Journal of Conflict Resolution 24(1): 143–152.

Beardsley

Kyle

Quinn

David

Biswas

Bidisha

Wilkenfeld

Jonathan

. 2006. Mediation style and crisis outcome. Journal of Conflict Resolution 50(1): 58–86.

Bercovitch

Jacob J.

1996. Resolving International Conflicts: The Theory and Practice of Mediation. Boulder, CO: Lynne Rienner.

Bercovitch

Jacob J.

Schneider

Gerald

. 2000. Who mediates? The political economy of international conflict management. Journal of Peace Research 37(2): 145–165.

Bond

Doug

Bond

Joe

Churl

Jenkins

J. Craig

Taylor

Charles Lewis

. 2003. Integrated Data for Events Analysis: An event typology for automated events data development. Journal of Peace Research 40(6): 733–745.

Bremer

Stuart A.

1982. The contagiousness of coercion: The spread of serious international disputes, 1900–1976. International Interactions 9(1): 29–55.

Bremer

Stuart A.

1992. Dangerous dyads: Conditions affecting the likelihood of interstate war, 1816–1965. Journal of Conflict Resolution 36(2): 309–341.

10.

Bueno de Mesquita

Bruce

. 1989. The contribution of expected utility theory to the study of international conflict. In Handbook of War Studies, ed. Midlarsky

Manus

, pp. 143–169. Boston: Allen and Unwin.

11.

Butterworth

Robert L.

Scranton

Margaret E.

1976. Managing Interstate Conflict, 1945–1974. Pittsburgh, PA: University Center for International Studies.

12.

Cartwright

Dorwin

Harary

Frank

. 1956. Structural balance: A generalization of Heider’s theory. Psychological Review 63(5): 277–293.

13.

Corbetta

Renato

. 2010. Determinants of third parties’ intervention and alignment choices in ongoing conflicts, 1946–2001. Foreign Policy Analysis 6(1): 61–85.

14.

Corbetta

Renato

Dixon

William J.

2005. Danger beyond dyads: Third-party participants in interstate disputes. Conflict Management and Peace Science 22(1): 39–61.

15.

Crescenzi

Mark J. C.

2007. Reputation and interstate conflict. American Journal of Political Science 51(2): 382–396.

16.

Croco

Sarah E.

Teo

Tze Qwang

. 2005. Assessing the dyadic approach to interstate conflict processes: A.k.a. ‘dangerous’ dyad-years. Conflict Management and Peace Science 22(1): 5–18.

17.

Cusack

T. R.

Eberwin

W. D.

1982. Prelude to war: Incidence, escalation, and intervention in international disputes, 1900–1976. International Interactions 9(1): 9–28.

18.

Dixon

William J.

1993. Democracy and the management of international conflict. Journal of Conflict Resolution 37(1): 42–68.

19.

Dixon

William J.

1994. Democracy and the peaceful settlement of international conflict. American Political Science Review 88(1): 14–32.

20.

Dixon

William J.

1996. Third party techniques for preventing conflict escalation and promoting peaceful settlement. International Organization 50(6): 653–682.

21.

Enterline

Andrew J.

1999. Some other time: The timing of third party intervention in dynamic interstate conflicts, 1816–1992. Paper presented at the 33rd Annual Meeting of the Peace Science Society (International). Ann Arbor, MI, 8–10 October.

22.

Favretto

Katja

. 2009. Should peacemakers take sides? Major power mediation, coercion, and bias. American Political Science Review 53(2): 248–263.

23.

Fearon

James D.

2002. Selection effects and deterrence. International Interactions 28(1): 5–29.

24.

Frazier

Derrick V.

Dixon

William J.

2006. Third-party intermediaries and negotiated settlements, 1946–2000. International Interactions 32(4): 385–408.

25.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart A.

2004. The MID3 data set, 1993–2001: Procedures, coding rules, and description. Conflict Management and Peace Science 21(2): 133–154.

26.

Goldstein

Joshua S.

1992. A conflict and cooperation scale for WEIS events. Journal of Conflict Resolution 36(2): 369–385.

27.

Grant

Keith A.

2010. The indirect dimensions of conflict and cooperation. PhD dissertation. Tucson, AZ: University of Arizona.

28.

Greig

Michael J.

2005. Stepping into the fray: When do mediators mediate? American Journal of Political Science 49(2): 249–266.

29.

Greig

Michael J.

Diehl

Paul F.

2006. Softening up: Making international conflict more amenable to diplomacy. International Interactions 32(4): 355–384.

30.

Hafner-Burton

Emilie M.

Kahler

Miles

Montgomery

Alexander H.

2009. Network analysis for international relations. International Organization 63(3): 559–592.

31.

Harary

Frank

. 1959. On the measurement of structural balance. Behavioral Science 2(4): 316–323.

32.

Harary

Frank

Norman

Robert Z.

Cartwright

Dorwin

. 1965. Structural Models: An Introduction to the Theory of Directed Graphs. New York: Wiley.

33.

Heider

Fritz

. 1946. Attitudes and cognitive organization. Journal of Psychology 21: 107–112.

34.

Heider

Fritz

. 1958. The Psychology of Interpersonal Relations. New York: John Wiley.

35.

Hummon

Norman P.

Doreian

Patrick

. 2003. Some dynamic of social balance processes: Bringing Heider back into balance theory. Social Network 25(1): 17–49.

36.

Jackson

Matthew O.

2008. Social and Economic Networks. Princeton, NJ: Princeton University Press.

37.

Jones

Daniel M.

Bremer

Stuart A.

Singer

J.D.

1996. Militarized interstate disputes, 1816–1992: Rationale, coding rules, and empirical patterns. Conflict Management and Peace Science 15(2): 163–213.

38.

Joyce

Kyle A.

2008. Third parties and the dynamic of conflict expansion. PhD dissertation. University Park, PA: Pennsylvania State University.

39.

Kahler

Miles

. 2009. Networked Politics: Agency, Power, and Governance. Ithaca, NY: Cornell University Press.

40.

Khanafiah

Deni

Situngkir

Hokky

. N.d. Social balance theory: Revisiting Heider’s balance theory for many agents. ( http://www.au.af.mil/au/awc/awcgate.htm )

41.

Kim

Chae-Han

. 1991. Third party participation in wars. Journal of Conflict Resolution 35(4): 659–677.

42.

King

Gary

Lowe

Will

. 2003. An automated information extraction tool for international conflict data with performance as good as human coders: A rare events evaluation design. International Organization 57(3): 617–642.

43.

Knoke

David

. 1994. Political Networks: A Structural Perspective. 2nd edn. Cambridge: Cambridge University Press.

44.

Knoke

David

Yang

Song

. 2008. Social Network Analysis. 2nd edn. Thousand Oaks, CA: Sage.

45.

Kydd

Andrew

. 2003. Which side are you on? Bias, credibility, and mediation. American Journal of Political Science 47(4): 597–611.

46.

Kydd

Andrew

. 2006. When can mediators build trust? American Political Science Review 100(3): 449–462.

47.

Lee

Sung Chull

Muncaster

Robert G.

Zinnes

Dina A.

1994. The friend of my enemy is my enemy: Modeling triadic international relationships. Synthese 100: 333–358.

48.

Leeds

Ashley B.

Mitchell

Sara McLaughlin

Long

Andrew G.

2000. Reevaluating alliance reliability. Journal of Conflict Resolution 44(5): 686–699.

49.

Leeds

Ashley B.

Ritter

Jeffrey M.

Mitchell

Sara McLaughlin

Long

Andrew G.

2002. Alliance treaty obligations and provisions, 1815–1944. International Interactions 28(2): 237–260.

50.

Leng

Russell J.

Regan

Patrick M.

2003. Social and political cultural effects on the outcome of mediation in militarized interstate disputes. International Studies Quarterly 47(3): 431–452.

51.

Maoz

Zeev

. 2005. Dyadic MID dataset version 2.0. ( http://psfaculty.ucdavis.edu/zmaoz/dyadmid.html )

52.

Maoz

Zeev

Terris

Leslie G.

Kuperman

Ranan D.

Talmud

Ilan

. 2007. What is the enemy of my enemy? Causes and consequences of imbalanced international relations. Journal of Politics 69(1): 100–115.

53.

Marshall

Monty G.

Jaggers

Keith

. 2002. Polity IV project: Political regime characteristics and transitions, 1800–2002. Typescript. University of Maryland.

54.

Monge

Peter R.

Contractor

Noshir

. 2003. Theories of Communication Networks. New York: Oxford University Press.

55.

Most

Benjamin A.

Starr

Harvey

Siverson

Randolph

. 1989. The logic and study of the diffusion of international conflict. In Handbook of War Studies, ed. Midlarsky

Manus

, pp. 111–139. Boston: Allen and Unwin.

56.

Quackenbush

Stephen L.

2006. Identifying opportunity for conflict: Politically active dyads. Conflict Management and Peace Science 23(1): 37–51.

57.

Rauchhaus

Robert W.

2006. Asymmetric information, mediation, and conflict management. World Politics 58(2): 207–241.

58.

Reiter

Dan

Stam

Alan C.

2002. Democracies at War. Princeton, NJ: Princeton University Press.

59.

Rosenau

James N.

1969. Intervention as a scientific concept. Journal of Conflict Resolution 13(2): 149–171.

60.

Savun

Burcu

. 2008. Information, bias, and mediation success. International Studies Quarterly 52(1): 25–47.

61.

Scott

John

. 2000. Social Network Analysis: A Handbook. 2nd edn. Thousand Oaks, CA: Sage.

62.

Shellman

Stephen M.

2003. Time series intervals and statistical inference: The effects of temporal aggregation on event data analysis. Political Analysis 12(1): 97–104.

63.

Singer

David J.

1988. Reconstructing the Correlates of War dataset on Material Capabilities of States, 1816–1985. International Interaction 14(2): 115–132.

64.

Singer

David J.

Bremer

Stuart

Stuckey

John

. 1972. Capability distribution, uncertainty, and major power war, 1820–1965. In Peace, War, and Numbers, ed. Russett

Bruce

, pp. 19–48. Beverly Hills: Sage.

65.

Siverson

Randolph

Starr

Harvey

. 1991. The Diffusion of War: A Study in Opportunity and Willingness. Ann Arbor, MI: University of Michigan Press.

66.

Smith

Alastair

. 1996. To intervene or not to intervene: A biased decision. Journal of Conflict Resolution 40(1): 16–40.

67.

Smith

Alastair

Stam

Allan

. 2003. Mediation and peacekeeping in a random walk model of civil and interstate war. International Studies Review 5(4): 115–135.

68.

Terris

Lesley G.

Maoz

Zeev

. 2005. Rational mediation: A theory and a test. Journal of Peace Research 42(5): 563–583.

69.

Touval

Saadia

Zartman

I. William

. 1985. International Mediation in Theory and Practice. Boulder, CO: Westview.

70.

Wasserman

Stanley

Faust

Katherine

. 1994. Social Network Analysis: Methods and Applications. Cambridge: Cambridge University Press.

71.

Werner

Suzanne

. 2000. Deterring intervention: The stakes of war and third party involvement. American Journal of Political Science 44(4): 720–732.

72.

Werner

Suzanne

Lemke

Douglas

. 1997. Opposites do not attract: The impact of domestic institutions, power, and prior commitments on alignment choices. International Studies Quarterly 41(3): 529–546.

73.

Yamamoto

Yoshinobu

Bremer

Stuart A.

1980. Wider wars and restless nights: Major power intervention in ongoing wars. In The Correlates of War II: Testing Some Realpolitik Models, ed. Singer

David J.

, pp. 85–119. New York: Free Press.

74.

Young

Oran

. 1967. The Intermediaries: Third Parties in International Crises. Princeton, NJ: Princeton University Press.

75.

Yuen

Amy

. 2009. Target concessions in the shadow of intervention. Journal of Conflict Resolution 53(3): 745–773.