Geographic proximity and third-party joiners in militarized interstate disputes

Abstract

A rich literature addresses how a state’s capabilities, its desire to aid or exploit a warring neighbor, and its alliance commitments determine whether or not the state joins an ongoing conflict. However, an important geopolitical consideration – proximity to the location of the ongoing conflict – has yet to be examined. The authors argue that states are more likely to join conflicts that occur close to their territories than conflicts that are located at a greater distance, and that accounts that do not pay attention to this distance are incomplete. Proximity to the location of an ongoing conflict affects the opportunity for a state to join (by decreasing costs), while also affecting the state’s willingness to join (by increasing the potential threat to the state’s security). A series of statistical models provide evidence for the authors’ claims: a state’s opportunity to join and its willingness to aid or exploit a neighbor in conflict, or to fulfill its alliance commitments, are each conditioned by its proximity to the location of the conflict. This conditioning effect of dispute location is important because it helps account for cases that appear to contradict the expectations of existing arguments regarding capabilities, contiguity, and alliances – such as when weak, non-contiguous, and non-allied states join ongoing conflicts and strong, contiguous, and allied states do not join.

Keywords

conflict location geography international conflict third-party intervention

Introduction

Does a potential joiner’s distance to the location of an ongoing conflict influence whether or not it will join the conflict? The decision as to whether or not to join an ongoing conflict – potentially placing troops in the line of fire and sinking scarce resources to aid or exploit foreign powers – ranks as one of the most significant foreign policy decisions faced by a government.¹ Existing approaches identify being the neighbor and/or ally of a state at war as conditions that increase the likelihood of a third party joining that war (Siverson & Starr, 1991). Yet, relatively few conflicts expand beyond the initial belligerents and of those that do, only a small number of neighbors and allies join. Moreover, a significant number of joiners are, in fact, non-contiguous and non-allied.² Focusing on contiguity and alliances alone, we therefore lack the ability to predict which contiguous/allied states are more (or less) likely to join, or to explain any additional cases of joining – specifically, why non-contiguous and non-allied third parties join ongoing disputes.

We argue that by focusing on the geographical location at which the initial belligerents come into conflict, it is possible to discriminate between different probabilities of joining, even among contiguous and allied third parties. Third parties are more likely to join ongoing conflicts that are proximate to the state, and they are more likely to join conflicts involving neighbors and/or allies when the conflicts involving these neighbors and allies occur closer to their own territory. Our conflict joining story incorporates the role of contiguity and alliances, but argues that the effect of these factors are conditioned on the distance between the potential joiners and the conflict itself. A dispute originated by the Soviet Union helps to illustrate this point.

The three-month dispute beginning in July 1961 between the Soviet Union and the USA, known as the Berlin Crisis, resulted in the construction of the Berlin Wall beginning in August 1961. This dispute represented the climax of disagreements between the Soviet Union and the other (Western) occupying states over the status of Berlin and increasing frustration regarding the considerable flow of refugees from the eastern (Soviet controlled) sectors to the western sectors of the city. Subsequent to this dispute’s bilateral origins in July 1961, a number of states joined the dispute in August (East Germany on the side of the Soviet Union and West Germany, Great Britain, and France on the side of the USA) and in October (both Poland and Czechoslovakia amassed troops at their borders).

According to conventional arguments we have little by which to determine whether, for instance, Afghanistan, Japan, Iran, or Poland (each as contiguous neighbors of the Soviet Union) would have been most likely to join the ongoing dispute. Since the dispute itself was located in Berlin, it seems implausible to anticipate the involvement of Afghanistan (4,800 km), Japan (9,000 km), or Iran (3,565 km). However, Poland’s joining (only 533 km from the location of the dispute) – both in terms of its opportunity to deploy force and its willingness to strengthen its buffer against potential spillover – seems quite logical.³

Until recently, a dearth of data has meant that it has not been possible to study the influence that location has on state behaviors across a variety of interstate conflict processes. The Militarized Interstate Dispute Location dataset fills this gap, detailing the precise geographic location of MIDs from 1816 to 2001 (Braithwaite, 2010). We use these data to test three related claims: the influence of state capabilities, alliances, and contiguity in determining joining behaviors are each conditioned by the proximity of a state to the location of the ongoing conflict, with the influence of each factor diminishing with increasing distance between the state and the conflict location. A series of statistical models return considerable evidence confirming our hypotheses.

The article proceeds as follows. First, we detail the predominant accounts of conflict joining in the conflict processes literature, focusing on arguments pertaining to capability, contiguity, and alliances. Throughout this discussion, we note the key omission of geographic conflict location, and highlight the logic driving our expectations regarding the influence of a state’s proximity to a conflict on its joining decisions. We then outline the research design used to test our hypotheses and discuss the findings, as well as the implications of these findings for future research.

Who joins? And why?

In their study of the causes of war, Singer & Small (1972: 35) state that ‘the geographic loci [of wars] are, from our point of view, of limited interest’. Their preference to focus on competing power parity and preponderance arguments, as well as assessing the pacifying qualities of democratic regimes, provided a cue for most subsequent research on international conflict (cf. Bennett & Stam, 2004).⁴ The tendency to ignore the location at which a conflict occurs is even found in most research on the geography of conflict, which has typically conceptualized association between phenomena (e.g. conflicts) as consisting primarily of a correlation between the entities (e.g. states) that participate in or experience the phenomena (cf. Anselin & O’Loughlin, 1990; Siverson & Starr, 1991). For example, research on conflict diffusion focuses on factors that provide states with the opportunity (e.g. contiguity) and/or willingness (e.g. alliances) to join a conflict (Siverson & Starr, 1991). Our theoretical innovation is to bring geography into the conflict joining story – arguing the effects of these factors are conditioned on the distance between the potential joiner and the ongoing dispute.

We consider Siverson & Starr (1991) as our point of departure. Their models of war diffusion identify being contiguous to, or allied with, an initial belligerent in an ongoing conflict as treatments that increase a state’s likelihood of being infected by the conflict. Their opportunity and willingness framework has played a central role in the development of theories and empirical tests of joining behaviors (cf. Corbetta, 2010; Werner & Lemke, 1997). We join them (and Starr, 2005) in building on the work of the Sprouts (1969). Specifically, we view interaction opportunities as the set of factors that define and constrain the range of actions available to states. We argue the most notable determinant of a state’s interaction opportunities is its ability to project power beyond its borders.

We also join them in viewing willingness as reflecting the actors and events existing or occurring in close proximity as carrying greater salience than those more distant. Simply stated, more proximate issues are of greater relevance because they pose a graver threat to the state and/or offer greater potential benefits through interdependence. Thus, whereas Siverson & Starr (1991) conceived of contiguity as demonstrative of opportunity (versus alliances, which are a mark of willingness), others suggest that contiguity also serves as an indicator of willingness (cf. Clark & Regan, 2003; Gartzke, 2012). We build on this literature, branding both alliances and contiguity as a mark of willingness.

Distance to conflict and the opportunity and willingness to join

We augment the opportunity and willingness framework by focusing on the geographical context of the decision to join. We focus on the effect that distance between a potential joiner and the conflict location has on a state’s ability to project its capabilities to influence the conflict, and thus its opportunity to join. Additionally, proximity to a conflict also affects the likelihood a state will be affected by spillover from it, and is therefore also likely to influence the state’s willingness to join. This proximity effect influences the decisions of non-contiguous and non-allied states, as well as contiguous and allied states. This leads to hypotheses based on both independent and conditional effects stemming from the central role of geographic distance.

The capability a state is able to use in the pursuit of its political goals has a fundamental role in competing theoretical accounts of conflict processes. The realist tradition is particularly concerned with capability because it is central to the balance of power in the international system (Waltz, 1979). States are concerned with the acquisition of power by other states because it threatens their own security and survival. Military conflict provides states with an opportunity to gain additional power which, in turn, could lead to a change in the distribution of power in the international system. As a result, a state is likely to consider the threat posed to its external security by an ongoing conflict when deciding whether or not to join (Haldi, 2003). Extant research demonstrates more capable states are more likely to join ongoing conflicts (Altfeld & Bueno de Mesquita, 1979; Huth, 1988).⁵

However, this research assumes that power is projected equally across all distances and to all locations globally. As Bueno de Mesquita (1981: 103) argues, ‘a nation’s power must decline monotonically with distance’. Distance, he shows, exerts a debilitating effect on a state’s capacity to engage in conflict. Additionally, he notes the rate of decline is surely greater for weaker states and the rate of decline decreases as a consequence of major technological advances.

It is therefore necessary to adjust our expectations regarding capability to reflect the deleterious effect of distance. But how should we do so? Bueno de Mesquita (1981) draws inspiration from Boulding (1962) in contributing a theoretically appealing and operationally feasible adjustment for the miles per day over which a state could plausibly transport the personnel and materials required for the execution of a military operation. He suggested daily distances of 250 miles (1816–1918), 375 miles (1919–45), and 500 miles (1946 onwards).

In an effort to test Boulding’s (1962) claim, we argue states constantly make discriminatory decisions when allocating their scarce resources. Given the cost of using military force, the decision to employ force is not taken lightly. However, when force projection requirements are lower – the closer a conflict is to a given state – the costs of joining are less prohibitive.

In addition to having the opportunity to join an ongoing conflict, a state must be willing to do so. Building upon the Galtonian principle that all actions are related to one another but more proximate actions are more related than distant actions (Galton, 1889), we assume states typically pay greater attention to events occurring locally than they do to events occurring at a greater distance (Gleditsch, 2002). Issues and events close to home have the greatest bearing upon the establishment of foreign policy priorities. The relevance of an ongoing dispute to other states might develop in a couple of ways: (a) the state may stand to profit more from joining an ongoing conflict that is proximate, as it may be able to affect the local distribution of capabilities in its own favor; (b) the state may be directly at threat from spillover from the ongoing conflict (e.g. refugees, direct fighting) unless it acts to bring an alternative outcome. Accordingly, in addition to arguing proximity to a conflict affects a state’s ability to project power to affect the outcome of that conflict, we also argue proximity to the location of an ongoing conflict ensures the conflict represents a more significant potential threat to a state’s national security. Both opportunity and willingness lead to the same expectation:

Hypothesis 1: The likelihood that a state will join an ongoing conflict decreases with increasing distance to the conflict location.

In addition to the independent effect of distance to conflict location, we also derive two additional predictions, stemming from the empirical observation that not all neighbors and allies of states involved in disputes choose to join those disputes, and that not all non-neighbors and non-allies choose not to join. The opportunity and willingness framework offers little to discriminate between joiners and non-joiners when conditions of contiguity and/or alliance are not satisfied. We argue that distance to the conflict location provides such a means of discrimination. The further away a potential joiner is from a conflict (above and beyond being neighbors and allies, or non-neighbors and non-allies), the less likely such states are to join an ongoing conflict.

While power may be the most widely discussed correlate of conflict processes, the most robust empirical finding in the conflict literature tends to report contiguity matters most. Following the demonstration by Richardson (1960) that states with multiple neighbors are more likely to be infected by conflict than those with fewer or no neighbors, a significant quantitative literature has shown that contiguity has the foremost influence on the likelihood of conflict. This finding is consistent for major powers (Wallensteen, 1981), rivalries (Vasquez, 1983), across intensities of conflict (Diehl, 1985), and across time periods (Holsti, 1991). In terms of conflict joining decisions, Siverson & Starr (1991) and Starr & Thomas (2002) identify contiguity with an initial belligerent in a conflict as a ‘treatment’ that, in part, proxies the opportunity for a state to join the conflict.

Moving beyond merely characterizing contiguity as an indicator of opportunity, Vasquez (1995) argues static patterns of contiguity matter only insofar as they represent extant claims to territories. This territorial conflict logic – at the core of both the ‘Steps to war’ (Senese & Vasquez, 2008) and ‘Territorial peace’ (Gibler, 2012) – thus moves contiguity from being a mere indicator of opportunity to one that reflects the interest a state holds towards a given issue, and its willingness to pursue that interest.

Hypothesis 2: The likelihood that a state will join a conflict involving a neighboring state is conditioned by distance, with the greatest likelihood in cases in which the location of the dispute is proximate to the territory of the potential joiner.

A second crucial determinant of joining according to the existing literature is the presence of an alliance between a potential joiner and an initial belligerent. Alliances and, most commonly, defense pacts play a crucial role in conflict processes according to both neoliberal and neorealist accounts. In both traditions, constellations of alliances are viewed as a map of intersecting state interests. In the case of joining behaviors, existing results suggest that a state is more likely to join an ongoing conflict if it is allied to one of the initial belligerents (Leeds, 2003; Siverson & Starr, 1991; Smith, 1996).⁶

In an effort to embrace the geography of alliance relationships, Gartzke & Gleditsch (2004) examine the commitments of democratic allies. They control for proximity by coding whether or not a potential joiner and an allied initial belligerent in a conflict are contiguous as well as by examining the distance between the capital cities of a potential joiner and an allied initial belligerent. Their results, which show joining is more likely when a third party is proximate to an ally, are robust to these alternative specifications. However, no research has yet examined the influence of alliances on joining behaviors in the specific context of the distance to the conflict itself.

Just as we argued being contiguous to an initial belligerent of a conflict may enhance a state’s willingness to join the conflict if that conflict is local, we also suggest that while states may wish to commit consistently to all alliance relationships, they are limited by scarce resources, and thus will commit first and foremost to those disputes that potentially have the greatest bearing on their own security.

Hypothesis 3: The likelihood that a state will join a conflict involving an allied state is conditioned by distance, with the greatest likelihood in cases in which the location of the dispute is proximate to the territory of the potential joiner.

Proximity model of joining

Increasing distance from a conflict location reflects loss of strength (opportunity) and loss of interest (willingness) for a potential joiner. To directly examine the extent to which distance diminishes both opportunity and willingness, we develop two models. In the first model, we discuss a new modification of a state’s ability to project power. Traditional models of conflict processes (including those of joining behaviors) typically include a variable for material capabilities. Bueno de Mesquita (1981) offers a variant on this approach, which discounts capabilities by distance from the state to an initial belligerent. We modify this approach by discounting capabilities by the distance from the state to the actual conflict location. It is important to note we do not view these two measures as being substitutes for one another. We make this modification not to present a competitive test of their value but rather to better specify the costly influence of the distance across which the state will actually have to project its power. In other words, while we conceive of the distance between an initial belligerent and potential joiner as an indicator of the shared interest of those two states, we strongly believe the distance between the potential joiner and the conflict location is more directly relevant as an indicator of both the actual power projection cost to the potential joiner and the plausible impact the conflict could have on their security environment.

In the second model, we directly examine the claim that the effects of both being contiguous to and allied with an initial belligerent on the likelihood of joining are conditioned by distance. We introduce interaction terms between each of these variables and the distance between the state and the conflict location. The expectations of Hypotheses 1 and 2 imply we should observe different rates of joining for contiguous and non-contiguous and allied and non-allied potential joiners as distance increases. Before detailing the statistical tests of our hypotheses, it is informative to offer descriptive statistics of these divergent rates of joining.

The differing mean distances to ongoing conflicts show that among both contiguous and non-contiguous potential joiners, those that actually join ongoing conflicts are more proximate to the ongoing conflict than are non-joiners: 1,490 km versus 2,924 km in the case of contiguous states and 5,263 km versus 7,865 km in the case of non-contiguous states.⁷ Moreover, contiguous joiners have the shortest mean distance (1,490 km) to the ongoing conflict, with non-contiguous non-joiners at the greatest mean distance (7,865 km) from the ongoing conflict. These summary statistics clearly corroborate Hypothesis 2: states that join an ongoing conflict are (on average) closer to the conflict location than those that do not join for both contiguous and non-contiguous states.

The mean distances to ongoing conflicts for allied and non-allied joiners and potential joiners broadly conform to Hypothesis 3: states that join an ongoing conflict are (on average) closer to the location of a conflict than those that do not join for both allied and non-allied states. For allied states these distances are 2,906 km versus 5,608 km for joiners and non-joiners, respectively, while for non-allied states the distances are 3,583 km versus 7,677 km. The shortest mean distance to ongoing conflicts is for allied joiners (2,906 km) and the longest mean distance is for non-allied non-joiners (7,677 km).

Research design

In order to test our hypotheses, we use the Militarized Interstate Dispute (MID) dataset (Ghosn, Palmer & Bremer, 2004) to identify both our sample of disputes and all cases of third parties joining (if any) in each dispute.⁸ Our sample only includes MIDs between 1816 and 2001 with a dispute hostility level of 4 or 5 (i.e. ‘use of force’ or ‘interstate war’) and for which hostilities lasted for more than one day. We use these MIDs because they indicate military force was used by at least one state. These are the most serious disputes and we expect third-party decisions to be most consistently observable during these disputes.

Our dataset includes all third parties that could potentially join an MID (Corbetta, 2010; Joyce, Ghosn & Bayer, forthcoming; Melin & Koch, 2010). These potential joiners are all states in the international system (except the initial belligerents) during the MID according to the Correlates of War (COW) State System Membership List (Correlates of War Project, 2008). While some research includes only third parties that actually joined a MID in their sample (Siverson & Starr, 1991; Werner & Lemke, 1997), this is problematic because important information about third parties that could have joined a MID but chose not to is discarded. Since a third party is continually updating its decision to join or remain neutral as a dispute persists, a decision not to join at one point in time does not preclude joining in the future. For each year in each MID we include all states that could potentially join a MID.^9,10 Our sample includes 1,231 MIDs, 211,730 potential joiner years (observations), and 514 cases of joining.¹¹

Our dependent variable measures whether or not a potential joiner joined an ongoing MID (regardless of which side it joined) after the MID’s start date. We create a binary variable that equals 1 if a potential joiner joined an ongoing MID and 0 otherwise. We use a logit model and cluster each initial belligerent–potential joiner dyad within each MID. Thus, we assume there is not independence across initial belligerent–potential joiner dyads within the same MID, which is important when there are multiple initial belligerents in a MID.

Independent variables

Capability

We use the COW Composite Index of National Capability (CINC) (Singer, Bremer & Stuckey, 1972) to measure the capability of the initial belligerents and each potential joiner. We construct three capability variables: (1) no adjustment to a potential joiner’s capability using the loss-of-strength gradient (LSG), (2) adjustment to a potential joiner’s capability using the LSG based on the distance between the capital city of a potential joiner and the initial belligerents,¹² and (3) adjustment to a potential joiner’s capability using the LSG based on the distance between the capital city of a potential joiner and the MID location.¹³

When calculating the distance between the capital city of a potential joiner and the initial belligerents, we follow Bueno de Mesquita (1981: 108) and discount a potential joiner’s capability to the location where it is stronger. We account for the fact that some MIDs have multiple initial belligerents by finding the shortest distance between a potential joiner and all of the initial belligerents.

We followed Bueno de Mesquita (1981: 104), whose indicator ‘depends on the number of days it takes to transport a major military operation’. Bueno de Mesquita defines the transportation range as 402 km per day (1816–1918), 604 km per day (1919–45), and 805 km per day (1946 onwards).¹⁴ We then adjusted a potential joiner’s CINC score using the formula from Bueno de Mesquita (1981: 105): $A d j u s t e d C I N C = {C I N C}^{{l o g}_{10} [(d i s t a n c e / k i l o m e t e r s p e r d a y) + (10 - e)]}$ , where distance is either the distance between the capital city of a potential joiner and the initial belligerents or the distance between the capital city of a potential joiner and the MID location. For the variable that does not account for the LSG we simply use the CINC score of a potential joiner.

The capability contribution a potential joiner can offer to the initiator is calculated as the capabilities of the initiator and potential joiner relative to the combined capabilities of the initiator, target, and potential joiner minus the capabilities of the initiator relative to the initial belligerents: $\frac{C_{I} + C_{P J}}{C_{I} + C_{T} + C_{P J}} - \frac{C_{I}}{C_{I} + C_{T}}$ , where $C_{I}$ , $C_{T}$ , and $C_{P J}$ represent the capabilities of the initiator, target, and potential joiner, respectively. The term on the left represents the probability the initiator wins if the potential joiner joins the initiator. The term on the right represents the probability the initiator wins if the potential joiner does not join (i.e. remains neutral).¹⁵ The capability contribution a potential joiner can offer to the target is calculated similarly except the term on the left represents the probability the target wins if the potential joiner joins the target and the term on the right represents the probability the target wins if the potential joiner does not join: $\frac{C_{T} + C_{P J}}{C_{I} + C_{T} + C_{P J}} - \frac{C_{T}}{C_{I} + C_{T}}$ . Finally, we measure the maximum capability contribution a potential joiner could offer to either initial belligerent: $m a x (\frac{C_{I} + C_{P J}}{C_{I} + C_{T} + C_{P J}} - \frac{C_{I}}{C_{I} + C_{T}}, \frac{C_{T} + C_{P J}}{C_{I} + C_{T} + C_{P J}} - \frac{C_{T}}{C_{I} + C_{T}})$ . If a potential joiner’s capability contribution to the initiator (target) is greater than its capability contribution to the target (initiator) then we use that value to represent the capability contribution a potential joiner can make to either initial belligerent.¹⁶ In our statistical models we use the maximum capability contribution a potential joiner could make to either initial belligerent.¹⁷

Contiguity

Contiguity is measured by determining whether the potential joiner shares a land border or is separated by 241 km of water or less from an initial belligerent as determined by the COW Direct Contiguity dataset (Stinnett et al., 2002). We create a binary variable that equals 1 if the potential joiner is contiguous to an initial belligerent and 0 otherwise.

Alliance

We measure whether a potential joiner had an alliance obligation with an initial belligerent using the Alliance Treaty Obligations and Provisions (ATOP) dataset (Leeds et al., 2002). The ATOP dataset provides the precise obligations undertaken by each member of an alliance. We examine three types of alliance obligations: defense, offense, and neutrality. Defense and offense pacts obligate a member of an alliance to provide military assistance in the event of military conflict involving another member of the alliance and neutrality pacts involve an explicit promise not to join a conflict against an ally and not to facilitate an adversary’s success in the event of a conflict with the ally.

Since two states in the same alliance do not necessarily undertake symmetric obligations, we coded the obligation of the potential joiner to the initial belligerent of an MID and not vice-versa. In order for a potential joiner and an initial belligerent to be coded as having an alliance obligation, at least one alliance had to be in effect for the full duration of an MID. We create a binary variable that equals 1 if a potential joiner had a defensive, offensive, or neutrality obligation with either initial belligerent and 0 otherwise.

Regime type

We measure regime similarity between a potential joiner and the initial belligerents using the Polity IV dataset (Marshall & Jaggers, 2010). We compute regime similarity by calculating the difference between a state’s democracy and autocracy scores, which range from $- 10$ to $+ 10$ , to which we add 10, so scores range from 0 to 20. States with scores between 16 and 20 (inclusive) are coded as democracies and states with scores between 0 and 4 (inclusive) are coded as autocracies. We create two binary variables, one that equals 1 if a potential joiner and either initial belligerent are jointly democratic and 0 otherwise, and one that equals 1 if a potential joiner and either initial belligerent are jointly autocratic and 0 otherwise.

Distance to MID location

We use the MIDLOC dataset to measure the great circle distance from the capital city of each potential joiner to each MID onset location (Braithwaite, 2010). Each record details the approximate location of the first hostilities of a MID; this onset location is then used to proxy the overall MID location. Thus, a single location is used to represent each MID.¹⁸ Following common convention, latitude and longitude point coordinates are recorded for each observation. Latitude is measured as the angular distance, in decimalized degrees (DD), of a point north or south of the equator. Longitude is measured as the angular distance, in DD, of a point east or west of the prime (Greenwich) meridian. Our distance variable measures the great circle distance (in km) between the longitude and latitude coordinates of a potential joiner’s capital city and the longitude and latitude coordinates of the MID location using the World Geodetic Survey (WGS84) parameters.¹⁹

Results: Location, location, location

The first set of models compare the effect of different LSG adjustments on the likelihood of third-party joining. The results are shown in Table I. The first column presents the results for the standard approach, which does not discount a third party’s capability by distance. The second column presents the results when a third party’s capability is adjusted for distance to the initial belligerents. The third column presents the results when a third party’s capability is adjusted for distance to the MID location.

These results clearly demonstrate a third party is more likely to join an ongoing conflict as its ability to influence the outcome of that conflict increases. The results in Column 1, which do not take distance into account, are consistent with findings in the existing literature. However, the coefficient associated with the adjusted capabilities measures is larger when we adjust using distance to the initial belligerents, and largest when we adjust using distance to the conflict location itself.

Figure 1 shows the effect of a third party’s capability on the predicted probability of joining an ongoing MID for four configurations of the contiguity and alliance variables: (1) Not Contiguous, Not Allied, (2) Contiguous, Not Allied, (3) Not Contiguous, Allied, (4) Contiguous, Allied. These configurations allow us to not only examine how the predicted probability of joining changes as a third party’s capability varies for each configuration but also for what range of a third party’s capability the predicted probability is statistically significant. In this figure, the y-axis represents the predicted probability of joining and the x-axis represents the range of third-party capability values. When calculating these

Table I.

Logit models of third-party joining using different loss-of-strength gradient specifications, 1816–2001

	No LSG adjustment	LSG adjustment using distance to initial belligerents	LSG adjustment using distance to MID location
Capability	2.09 $^{* *}$	2.41 $^{* *}$	2.52 $^{* *}$
	(0.18)	(0.18)	(0.18)
Contiguity	2.23 $^{* *}$	2.14 $^{* *}$	2.14 $^{* *}$
	(0.11)	(0.10)	(0.10)
Alliance	1.18 $^{* *}$	1.15 $^{* *}$	1.15 $^{* *}$
	(0.11)	(0.11)	(0.11)
Democracy	0.12	0.10	0.11
	(0.12)	(0.12)	(0.12)
Autocracy	–0.04	–0.05	–0.06
	(0.10)	(0.10)	(0.10)
Intercept	–7.27 $^{* *}$	–7.22 $^{* *}$	–7.20 $^{* *}$
	(0.08)	(0.08)	(0.08)
N	211,730	211,730	211,730
$- 2 \times$ Log-likelihood	6282.73	6248.18	6240.67
Likelihood ratio $χ_{(5, 5, 5)}^{2}$	1562.88 $^{* *}$	1697.35 $^{* *}$	1683.75 $^{* *}$

Significance levels (two-tailed): * 5%; ** 1%. Coefficients with robust standard errors in parentheses.

predicted probabilities we set both the democracy and autocracy variables equal to 0 since both are not statistically significant in any of the models. For each configuration we set the contiguity and/or alliance variable equal to 0 or 1 depending on the configuration.²⁰

This figure displays several results. First, for all four configurations the predicted probability of joining increases as a third party’s capability increases regardless of whether or not the third party’s capability is adjusted for distance.²¹ Second, the predicted probability of joining is generally highest for third parties both contiguous

Figure 1.

Predicted probability of a third party joining an ongoing MID as a third party’s capability increases

to and allied with an initial belligerent, followed by those contiguous to but not allied with an initial belligerent, then by those allied with but not contiguous to an initial belligerent and, finally, by those neither contiguous to nor allied with an initial belligerent. Third, when a third party’s capability is low then adjusting a third party’s capability to account for distance does not make much difference in the predicted probability of joining across all four configurations. However, as a third party’s capability increases, the differences between the three specifications is more evident. In general, at the higher end of the capability range, the predicted probability of a third party joining is underestimated when a third party’s capability is not discounted for distance or is discounted using its distance to the initial belligerents. The differences are most prevalent for the configuration where a third party is most likely to join (Contiguous, Allied).

While the results are similar regardless of which distance measure is used, we do not consider our distance to MID location measure to be a substitute for the commonly used measure of distance between a potential joiner and an initial belligerent.²² The distance between a potential joiner and an initial belligerent is typically included in model specifications to account for the importance of similarity, affinity, or rivalry between states in determining conflict joining. In contrast, our measure of distance to MID location directly captures the theoretical expectation that proximate MIDs negatively affect states’ security environments. Furthermore, theoretically speaking, adjusting a potential joiner’s capability using the LSG based on distance to MID location is the most appropriate way to do so in studies of conflict joining because it accurately captures the costly nature of distance on the projection of power.

The second set of models examine the conditional effects of distance on third-party joining. The results are shown in Table II. The results without the conditional effects are shown in Column 1 (these are the same as Column 3 of Table I). The results when we add distance to MID location are shown in Column 2. Here we can clearly see that as a potential joiner’s distance from the conflict location increases, potential joiners are statistically significantly less likely to join an ongoing conflict. This effect is above and beyond that of the distance adjusted capability variable, which suggests, in line with Hypothesis 1, that the distance to the conflict location affects a state’s willingness to join an ongoing conflict, as well as its opportunity to do so.

Our arguments in Hypotheses 1 and 2 suggest that distance not only has a direct bearing on a third party’s willingness to join an ongoing conflict but also a conditional effect via contiguity and alliances. The third column of Table II presents the results of these conditional effects, captured by two interaction terms: (1) contiguity and distance to MID location, and (2) alliance and distance to MID location. We also include the capability variable based on the distance to MID location LSG adjustment. It is only possible to directly interpret the constitutive term of these interaction terms for those cases in which the other term is equal to zero. In line with previous research, when distance to the conflict location is equal to zero then states contiguous to or allied with an initial belligerent are more likely to join an ongoing conflict. In order to examine the conditional effects, we calculated the predicted probability of a third party joining an ongoing MID for the four configurations of the contiguity and alliance variables as a potential joiner’s distance to a MID location increases. These configurations allow us to examine not only how the predicted probability of joining changes as distance to MID location increases for each configuration but also for what range of distance to MID location the predicted probability remains statistically significant.

Figure 2 shows the predicted probability of a third party joining an ongoing MID as its distance to a MID location increases for each configuration of the contiguity and alliance variables. These predicted probabilities are based on Model 3 in Table II. In this figure, the y-axis represents the predicted probability of joining and the x-axis represents the range of distance to MID location. The dashed line corresponds to the predicted probability and the solid lines correspond to the 95% confidence interval. For all four configurations of the contiguity and alliance variables the predicted probability is statistically significant, that is, the confidence interval does not include zero across the entire range of distance to MID location. When

Table II.

Logit models of third-party joining using different distance to MID location specifications, 1816–2001

	Model 1	Model 2	Model 3
Capability	2.52 $^{* *}$	2.25 $^{* *}$	2.27 $^{* *}$
	(0.18)	(0.19)	(0.19)
Contiguity	2.14 $^{* *}$	1.55 $^{* *}$	1.99 $^{* *}$
	(0.10)	(0.11)	(0.16)
Contiguity x Distance			–0.0002 $^{* *}$
			(0.00006)
Alliance	1.15 $^{* *}$	1.05 $^{* *}$	0.63 $^{* *}$
	(0.11)	(0.10)	(0.14)
Alliance x Distance			0.0002 $^{* *}$
			(0.00003)
Distance		–0.0002 $^{* *}$	–0.0002 $^{* *}$
		(0.00002)	(0.00002)
Democracy	0.11	0.15	0.09
	(0.12)	(0.12)	(0.12)
Autocracy	–0.06	–0.11	–0.07
	(0.10)	(0.10)	(0.10)
Intercept	–7.20 $^{* *}$	–6.10 $^{* *}$	–6.12 $^{* *}$
	(0.08)	(0.13)	(0.15)
N	211,730	211,730	211,730
$- 2 \times$ Log-likelihood	6240.67	6114.13	6077.91
Likelihood ratio $χ_{(5, 6, 8)}^{2}$	1683.75 $^{* *}$	1506.55 $^{* *}$	1430.69 $^{* *}$

Significance levels (two-tailed): * 5%; ** 1%. Coefficients with robust standard errors in parentheses.

calculating these predicted probabilities we set the capability variable equal to its mean value and both the democracy and autocracy variables equal to 0 since both variables are not statistically significant. For each configuration we set the contiguity and/or alliance variable equal to 0 or 1 depending on the configuration. When either of these variables is set to 1 we vary its interaction with distance to MID location from the minimum to the maximum of distance to MID location.

In this figure, our results show that the likelihood of a third party joining an ongoing MID, regardless of whether it is contiguous to and/or allied with an initial belligerent, is conditional on its distance to the MID location. The results of existing research are located at 0 on the x-axis. A third party not contiguous to or allied with an initial belligerent has the smallest predicted probability of joining. A third party contiguous to but not allied with an initial belligerent has a much higher probability of joining followed by those not contiguous to but allied with an initial belligerent. Finally, a third party contiguous to and allied with an initial belligerent has the highest probability of joining. These results are consistent with findings in the extant literature.

Our results are more nuanced than those in the existing literature. Across all four configurations of the contiguity and alliance variables the predicted probability of a

Figure 2.

Predicted probability of a third party joining an ongoing MID as a third party’s distance to a MID location increases

third party joining an ongoing MID decreases as its distance from the MID location increases.²³ These results show that two of the variables most strongly associated with third-party joining, contiguity and alliance, are affected by the distance a third party is from the conflict location. In particular, the strongest effects of distance are when a third party is contiguous to one of the initial belligerents, with the strongest effect for a third party both contiguous to and allied with an initial belligerent. The probability of joining for a third party contiguous to and allied with an initial belligerent steadily declines as the distance the third party is from the conflict location increases. When a third party is not contiguous to but allied with an initial belligerent the predicted probability of joining decreases as the distance the third party is from the MID location increases but not markedly so. Additionally, the predicted probability of joining is higher for those third parties neither contiguous to nor allied with an initial belligerent if those third parties are close to the MID location. Moreover, since our distance adjusted capability variable is held constant across the four configurations, these results are not simply due to a third party’s opportunity to join a conflict but also reflect a third party’s willingness to do so. In sum, the results presented in Table II and the graphical evidence in Figure 2 provide strong support for Hypotheses 1 and 2.

Conclusion

One of the most important foreign policy decisions faced by states is whether or not to join ongoing military conflicts. In this article we argued proximity to the location of an ongoing conflict influences a third party’s opportunity to join by decreasing costs and willingness to join by increasing the potential threat to the state’s security. We found strong empirical evidence for this argument. We began by noting the surprising reality that existing research on conflict joining has neglected to model the impact of conflict location on joining decisions. Our claim, that joining decisions are heavily conditioned by distance to the conflict location, garners considerable empirical support, both in terms of a state’s opportunity and willingness to join. States are much more likely – whether contiguous or not, allied or not – to join proximate conflicts than distant ones. We began with a discussion of the Berlin Crisis in 1961. The empirical results strongly support the notion that Poland joined this MID because of its close proximity to the MID location and not just because it was a neighbor of the Soviet Union; after all, Japan, Afghanistan, and Iran (among many other neighbors) were never likely to join given their vast distance from the MID.

Our theoretical discussion and empirical results suggest it is not sufficient to examine the proximity of a third party to an initial belligerent either through contiguity or using the distance between the third party and an initial belligerent. Analysts should discount capabilities by a measure of the state’s distance from the conflict location and also condition a third party’s contiguity and alliance relationships with an initial belligerent by its distance from the conflict location.

Future research on third-party joining in ongoing conflicts should consider how other variables (e.g. economic relationships) that influence whether or not a third party joins are conditioned by distance to the conflict. Additionally, future research should prioritize a more critical engagement with the dynamic impact of distance on conflict joining. Herein, we have operationalized distance from a single MID onset location. The MIDLOC dataset includes data on incident-level locations for the 1993–2001 time period, which provides the potential to analyze the robustness of our findings throughout the full duration of a conflict (i.e. as hostilities move closer to [or further from] a potential joiner’s territory). Additionally, our distance to MID location measure provides temporal variation across disputes that contiguity and the distance between states (usually) do not. An interesting extension of our analysis would be to examine the effect of distance to MID location on conflict joining in the context of repeated disputes between the same states. We have also treated MIDs as uniform conflicts with invariant issues at stake by focusing on all MIDs that involved a sufficient level of severity. It is conceivable that MIDs with distinct issues at stake may attract states at different distances with varying likelihood. Furthermore, research could also focus on how the costs associated with projecting power over distance affect the likelihood and nature of state contributions to peacekeeping forces and to allied actions, such as those conducted under the auspices of the ‘Coalition of the Willing’. Finally, future research could explore the differential impacts of distance on alternative forms of intervention – military, economic, and diplomatic. Beyond these immediate possible extensions, there are likely a great many studies of conflict processes that would benefit from greater sensitivity to the location of conflict. We hope the present article encourages further research of such topics.

Footnotes

Replication data

The data and commands for the empirical analysis as well as an appendix can be found at .

Acknowledgments

We thank Amber Boydstun, Rebecca Glazier, Heather McKibben, Randy Siverson, two anonymous referees, and the editor for helpful comments.

Notes

References

Altfeld

Michael F

Mesquita

Bruce Bueno de

(1979) Choosing sides in wars. International Studies Quarterly 23(1): 87–112.

Anselin

Luc

O’Loughlin

John

(1990) Spatial econometric analysis of international conflict. In: Chatterji

Manas

Kuenne

(eds) Dynamics and Conflict in Regional Structural Change, Festschrift in Honor of Walter Isard. New York: New York University Press, 325–345.

Bennett

D Scott

Stam

Allan C

(2000) EUGene: A conceptual manual. International Interactions 26(2): 179–204.

Bennett

D Scott

Stam

Allan C

(2004) The Behavioral Origins of War. Ann Arbor, MI: University of Michigan Press.

Boulding

Kenneth E

(1962) Conflict and Defense: A General Theory. New York: Harper & Row.

Braithwaite

Alex

(2010) MIDLOC: Introducing the Militarized Interstate Dispute Location dataset. Journal of Peace Research 47(1): 91–98.

Bueno de Mesquita

Bruce

(1981) The War Trap. New Haven, CT: Yale University Press.

Clark

David H

Regan

Patrick M

(2003) Opportunities to fight: A statistical technique for modeling unobservable phenomena. Journal of Conflict Resolution 47(1): 94–115.

Corbetta

Renato

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

10.

Correlates of War Project (2008) State system membership list, v2008.1 (http://correlatesofwar.org).

11.

Diehl

Paul F

(1985) Contiguity and military escalation in major power rivalries, 1816–1980. Journal of Politics 47(4): 1203–1211.

12.

Galton

Francis

(1889) Comment on E. B. Tylor ‘On a method of investigating the development of institutions: Applied to laws of marriage and descent’. Journal of the Royal Anthropological Institute 18: 270.

13.

Gartzke

Erik

(2012) Where nations fight: Capabilities, interests and contiguity in interstate disputes. Unpublished manuscript.

14.

Gartzke

Erik

Gleditsch

Kristian Skrede

(2004) Why democracies may actually be less reliable allies. American Journal of Political Science 48(4): 775–795.

15.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart

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

16.

Gibler

Douglas M

(2012) The Territorial Peace: Borders, State Development, and International Conflict. Cambridge: Cambridge University Press.

17.

Gleditsch

Kristian Skrede

(2002) All International Politics Is Local: The Diffusion of Conflict, Integration, and Democratization. Ann Arbor, MI: University of Michigan Press.

18.

Gleditsch

Kristian Skrede

Ward

Michael D

(2001) Measuring space: A minimum distance database and applications to international studies. Journal of Peace Research 38(6): 739–758.

19.

Haldi

Stacy

(2003) Why Wars Widen: A Theory of Predation and Balancing. Portland, OR: Frank Cass.

20.

Holsti

Kalevi

(1991) Peace and War: Armed Conflicts and International Order, 1648–1989. Cambridge: Cambridge University Press.

21.

Huth

Paul K

(1988) Extended Deterrence and the Prevention of War. New Haven, CT: Yale University Press.

22.

Joyce

Kyle A

Ghosn

Faten

Bayer

Reşat

(forthcoming) When and whom to join: The expansion of ongoing violent interstate conflicts. British Journal of Political Science.

23.

Leeds

Brett Ashley

(2003) Alliance reliability in times of war: Explaining state decisions to violate treaties. International Organization 57(4): 801–827.

24.

Leeds

Brett Ashley

Ritter

Jeffrey M

Mitchell

Sara McLaughlin

Long

Andrew G

(2002) Alliance treaty obligations and provisions, 1815–1944. International Interactions 28(3): 237–260.

25.

Lemke

Douglas

(2002) Regions of War and Peace. Cambridge: Cambridge University Press.

26.

Marshall

Monty G

Jaggers

Keith

(2010) POLITY IV Project: Political Regime Characteristics and Transitions, 1800–2010. Fairfax, VA: Center for Systemic Peace, George Mason University.

27.

Melin

Molly M

Koch

Michael T

(2010) Jumping into the fray: Alliances, power, institutions, and the timing of conflict expansion. International Interactions 36(1): 1–27.

28.

Moul

William Brian

(1988) Balances of power and the escalation to war of serious disputes among the European great powers, 1815–1939: Some evidence. American Journal of Political Science 32(2): 241–275.

29.

Reiter

Dan

Stam

Allan C

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

30.

Richardson

Lewis F

(1960) Statistics of Deadly Quarrels. Chicago, IL: Quadrangle.

31.

Senese

Paul D

Vasquez

John A

(2008) The Steps to War: An Empirical Study. Princeton, NJ: Princeton University Press.

32.

Singer

J David

Small

Melvin

(1972) The Wages of War, 1816–1965: A Statistical Handbook. New York: John Wiley & Sons.

33.

Singer

J David

Bremer

Stuart

Stuckey

John

(1972) Capability distribution, uncertainty, and major power war, 1820–1965. In: Russett

Bruce M

(ed.) Peace, War, and Numbers. Beverly Hills, CA: Sage, 19–48.

34.

Sinnott

(1984) Virtues of the haversine. Sky and Telescope 68(2): 159.

35.

Siverson

Randolph M

Starr

Harvey

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

36.

Smith

Alastair

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

37.

Sprout

Harold

Sprout

Margaret

(1969) Environmental factors in the study of international politics. In: Rosenau

James N

(ed.) International Politics and Foreign Policy. New York: Free Press, 60–72.

38.

Starr

Harvey

(2005) Territory, proximity, and spatiality: The geography of international conflict. International Studies Review 7(3): 387–406.

39.

Starr

Harvey

Dale Thomas

(2002) The nature of contiguous borders: Ease of interaction, salience, and the analysis of crisis. International Interactions 28(2): 213–235.

40.

Stinnett

Douglas

Tir

Jaroslav

Schafer

Philip

Diehl

Paul F

Gochman

Charles

(2002) The Correlates of War Project contiguity data, version 3. Conflict Management and Peace Science 19(2): 58–66.

41.

Vasquez

John

(1983) The tangibility of issues and global conflict: A test of Rosenau’s issue area typology. Journal of Peace Research 20(2): 179–192.

42.

Vasquez

John

(1995) Why do neighbors fight? Proximity, interaction, and territoriality. Journal of Peace Research 32(3): 277–293.

43.

Wallensteen

Peter

(1981) Incompatibility, confrontation, and war: Four models and three historical systems, 1816–1976. Journal of Peace Research 18(1): 57–90.

44.

Walt

Stephen M

(1987) The Origins of Alliances. Ithaca, NY: Cornell University Press.

45.

Waltz

Kenneth

(1979) Theory of International Politics. Reading, MA: Addison-Wesley.

46.

Weidmann

Nils B

Kuse

Doreen

Gleditsch

Kristian Skrede

(2010) The geography of the international system: The CShapes dataset. International Interactions 36(1): 86–106.

47.

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.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB