Failing to succeed? The cumulative impact of international mediation revisited 1

Abstract

Previous studies argue that the cumulation of mediations generally decreases uncertainty and ultimately helps to solve a conflict peacefully. Yet the empirical evidence for this is inconclusive, suggesting that existent theoretical accounts might be incomplete. For addressing this shortcoming, the paper develops an argument for a curvilinear relationship between the cumulation of mediations and mediation effectiveness. Based upon this rationale, it is less likely that initial mediation attempts succeed, but—after a turning point has been reached—subsequent mediations are more likely to induce an effective resolution of a dispute. The author also argues that this effect should be stronger if the same actors mediate over the course of a conflict. The empirical implications of this theoretical revision are tested in a quantitative framework employing data for 1816–2001. The results provide strong support for the theory.

Keywords

Cumulative impact international mediation mediation effectiveness

Introduction

A recent UN report states that as an “effective tool to prevent, contain, and resolve disputes, mediation is important at all stages of possible conflict cycles” (UN 2011). In other words, international mediation, that is, “a mode of negotiation in which a third party helps the parties find a solution which they cannot find by themselves” (Touval and Zartman, 1985) has proven to be a crucial and effective instrument for the nonviolent resolution of several inter- and intrastate disputes. In other cases and at different stages of a conflict, however, mediations have failed and, hence, it seems they do not always or necessarily contribute to solving disputes at all. The reason for this might be, first and foremost, that domestic and international conflicts tend to be very complex, sometimes involving high stakes. The belligerents, and even the (potential) mediator(s), face uncertainty about many factors pertaining to the outbreak of fighting, which might hamper effective conflict resolution in turn (see, e.g. Beardsley 2008, 2011: 26; Fearon, 1995).

The existent literature argues here that mediation generally decreases uncertainty and that the cumulation of mediations, that is, a number of consecutive efforts of mediation over the course of a dispute, facilitates learning,² which can ultimately help to solve a conflict peacefully (e.g. Beardsley, 2008; Bercovitch and Gartner, 2006; Greig, 2005; Greig and Diehl, 2006; Kydd, 2003, 2006; Regan and Aydin, 2006; Regan and Stam, 2000; Savun, 2008). For example, Bercovitch and Houston (2000: 183) stress that “the feedback from previous events includes information, experience, learning, and understanding gained by the mediator and the parties”. Put differently, a present mediation attempt in a conflict might build upon experiences and learning processes stemming from preceding third-party interventions, making it potentially more effective than the earlier mediations (see also Heldt, 2009: 137f). Eventually, we would therefore expect to see a positive and linear relationship between the cumulation of mediations and mediation success. It might be surprising, however, that the empirical findings drawn from the existing literature are inconclusive, thus suggesting instead that previous theoretical accounts might be incomplete. Ultimately, we therefore do not know if and how the cumulation of mediations positively influences the outcome of mediation attempts.

In order to shed more light on this ambiguity, the following paper seeks to make a twofold contribution. First, I develop an argument for a curvilinear relationship between the cumulation of mediations and mediation effectiveness. While I concur with extant arguments that mediation generally decreases uncertainty and facilitates learning, I argue that initial mediation attempts are unlikely to reveal a sufficient amount of information because issue complexity is most complex and belligerents’ ideal points are farthest away at this stage of a conflict (Beardsley, 2008: 725f). Hence, it is actually more likely that a conflict will see initial settlement attempts failing—yet only until a turning point has been reached at and after which the belligerents are more receptive to the information provided by a cumulation of mediations. Mediations that occur after this tipping point are then more likely to lead to an effective resolution of the conflict. This theoretical revision complements Heldt (2009: 133), who examines the sequencing of any kind of peaceful third-party intervention: “paradoxically then, initial mediation efforts may (or perhaps even should) fail to ultimately generate an effective resolution of a conflict”. Moreover, this work could contribute to the ripeness theory (e.g. Zartman, 1989, 2001). According to this framework, mediators may enter a dispute before it is ripe or “softened-up” (Greig and Diehl, 2006) for effective third-party mediation. While initial interventions are thus less likely to succeed, these initial mediation efforts could ripen the conflict for subsequent resolution despite their ineffectiveness and, after a sufficient point of ripeness, that is, a turning point, has been reached, we observe an increasing probability of mediation effectiveness.

Second, I also contend that this U-shaped relationship between the cumulation of mediations and mediation effectiveness should be stronger, that is, reach the turning point faster and induce more effective outcomes, if the same actors mediate over the course of a dispute. This is because their own experience from past mediation attempts weighs more heavily on policy-makers than the efforts and experience from others (Jervis, 1976; Levy, 1994; Pickering, 2002; Reiter, 1996). That being said, the cumulation of mediations should nonetheless affect mediation outcomes when different mediators intervene, since information can disseminate over various actors and channels. These “indirect links” can thus compensate for instances where the same mediators might not be able (or willing) to intervene again (Böhmelt, 2009; Dorussen and Ward, 2008), although the observed outcome is likely to be less pronounced than in the case of an unvarying set of interveners.

Against this background, I thus seek to develop and empirically test a causal argument on the way in which mediation efforts can build upon one another to produce more effective results over time. Particularly owing to this emphasis on the curvilinear impact of a cumulation of mediations, my research presents a counterintuitive argument that explains why mediation can be seen as unsuccessful when being studied in an isolated fashion, but that it proves to be effective when examined within the overall process of conflict resolution. This rejects the naive treatment of different mediations as independent from each other. Melin (2011: 698) emphasizes accordingly that conflict resolution attempts “are not independent, with subsequent efforts at least commencing with more information than was available at the previous one” (see also Greig and Regan, 2008: 765). Therefore, my research adds to the existent work that highlights potential inconsistency problems of mediation (e.g. Beardsley, 2008, 2011).

The remainder of the article proceeds as follows. First, I review the current literature on mediation that points to the cumulation of mediations and the importance of uncertainty and learning processes therein. This allows me to demonstrate that previous theoretical accounts seem incomplete and that their corresponding empirical results might be misleading. I then develop a theoretical revision along the two pillars outlined above, that is, a framework that combines rationales about asymmetric information with psychological theories of learning. Afterwards, I describe the research design and test the empirical implications of my theory by analyzing quantitative data from the Issue Correlates of War (ICOW) project (Hensel and Mitchell, 2007; see also Hensel, 2001; Hensel et al., 2008). The final section concludes and discusses the implications of my research.

The cumulation of mediations, uncertainty, and learning: literature review

Disputes, especially at their early stages, are complex, involving many uncertain parameters, and belligerents simply lack all information that may be necessary for fully evaluating their situation (see, e.g. Blainey, 1988; Fearon, 1995). This uncertainty also pertains to a (potential) third-party intervener: “once a mediator gets involved in a negotiation, she typically spends considerable effort learning about the dispute and the parties involved” (Kydd, 2003: 600; see also Powell, 2004). However, mediation might address this and decrease uncertainty as it can ensure the efficient transfer of information, which belligerents would not be able to convey without a third party (Beardsley, 2008, 2011: 102f; Kydd, 2003, 2006; Morgan, 1995; Pillar, 1983; Powell, 1999; Rauchhaus, 2006; Regan and Aydin, 2006; Regan and Stam, 2000; Savun, 2008; Smith and Stam, 2003; Svensson, 2007, 2009). In more detail, mediators can pool information about the preferences, positions or interests of the belligerents, and then disseminate it to the actors; they can also keep the lines of communication between the antagonists open, provide alternative views on issues, or identify sets of mutually acceptable agreements (Beardsley, 2011: 36ff).³ As Regan and Aydin (2006: 740) state:

outside interventions can provide a more objective view of conditions and possible outcomes of the conflict, which help adversaries to update their beliefs about the likely outcomes. Absent an outside intervention, the information held by the warring parties is at best asymmetrical, and neither side has a unilateral incentive to honestly convey its military capabilities, expectations of victory or defeat, or the value of a settlement that it would accept for the fear of exploitation by its adversary.

Regan and Aydin (2006: 740) conclude that “the key to successfully solving disputes, therefore, is to reduce the asymmetry of information about capabilities and incentives”. This decreases the risk of continued fighting and, simultaneously, increases the chances that the parties will agree on the peaceful resolution of a conflict (see Beardsley, 2008).

However, many mediation attempts turn out to be ineffective and they are not able to end a dispute peacefully. For instance, the ICOW data (Hensel and Mitchell, 2007; see also Hensel, 2001; Hensel et al., 2008) identify 934 different mediation attempts between 1816 and 2001, while only 219 (23.5%) of those attempts were fully effective to the extent that they actually ended an issue claim. Therefore, one mediation attempt on its own may be insufficient for uncovering all (or a necessary amount of) information. The previous literature (e.g. Bercovitch and Gartner, 2006; Greig and Diehl, 2006; Heldt, 2009; Regan and Aydin, 2006; Regan and Stam, 2000) then actually points out that only a series of succeeding mediations may develop a cumulative effect that positively influences the overall conflict resolution outcome. More specifically, the basic argument common for this work is that every mediation “sets a precedent on which the next attempt can build” (Heldt, 2009: 137f). Mediations thus “convey information and initiate learning processes that will, in turn, enable the antagonists to identify areas of agreement” (Heldt, 2009: 138; see also Greig, 2005; Greig and Regan, 2008: 765; Leng, 1998; Melin, 2011: 698; Regan and Stam, 2000). Regan and Stam (2000: 244) argue for that reason that the cumulation of mediations “increases the knowledge that the parties hold regarding their own and their adversaries’ positions, potential compromises, and the costs of continued conflict”. Hence, it is proposed that “as the number of third-party mediations accumulates, so will also the capital of peace” (Heldt, 2009: 139).

The existent literature, however, shows inconclusive evidence for this proclaimed positive linear impact of a cumulation of mediations on mediation effectiveness (see also Heldt, 2009: 135). For example, Bercovitch and Gartner (2006: 340f) argue indeed for a positive impact of a series of mediations on the outcome of present mediation interventions owing to learning and reciprocity effects. Although the authors find that previous mediations induce the same or even more complete outcomes, Bercovitch and Gartner (2006) only study previous mediations from the same actor. Heldt (2009) examines minor armed conflict dyads in 1993–2004 and concentrates on conflict resolution in terms of negotiated ends in a context of emerging intrastate conflicts. He reports a positive relationship between the total number of mediation attempts and the likelihood of negotiated resolutions. That being said, Heldt (2009) also obtains a strongly negative impact for the cumulative impact of good office interventions.

Similarly, Regan and Stam (2000) focus on interstate militarized disputes and seek to predict dispute duration. These scholars claim that “a cumulative effect is at hand to the extent that the same mediator carries out mediation” (Heldt, 2009: 138), while their theory eventually refers to a linear specification: the greater the previous mediation efforts over the course of a dispute, the higher the effectiveness of a mediation attempt under study. In fact, Regan and Stam (2000) find a negative relationship between the number of mediation attempts by the same actor and dispute duration. Greig and Diehl (2006: 362), moreover, state that, owing to previous mediations, belligerents “may gain greater insights into the other’s position, and perhaps even begin to empathize with their enemy, as they become more exposed to one another and encounter information about each other that conflicts with their existing adversarial images” (see also Princen, 1992; Stein, 1996). The cumulative effect of mediations, therefore, establishes trust and norms of interaction between the mediator and the parties, which facilitates agreement (Bercovitch and Gartner, 2006; Greig and Diehl, 2006: 362; Kelman, 1996; Lederach, 1997; Rubin, 1992). Greig and Diehl (2006: 373) obtain evidence for a positive relationship between the number of previous mediations and the likelihood of mediation onset; the impact of the former on mediation effectiveness remains unclear in this work, though, since it is not directly evaluated. However, Greig and Diehl (2006: 362) themselves raise doubts that previous mediation attempts can positively influence the outcome of a series of third-party interventions.

Böhmelt (2011) includes a variable indicating whether it is a particular mediating group’s first, second, third, etc. attempt in a specific dispute. This approach accounts for learning and reciprocity over the course of mediations and his mediation cumulation variable is significant at the 1% level with a negative linear impact. Thus, subsequent mediation attempts of the same mediator(s) may be actually less effective. Greig (2005; see also Heldt, 2009: 131) also obtains evidence for a negative linear relationship between the cumulation of mediations and third-party intervention effectiveness—despite his original argument that mediation is more likely to be successful as disputants develop “a rapport with the mediator and gain experience mediating their conflict” (see also Bercovitch, 2002; Zubek et al., 1992).

To sum up, the literature highlighted the role of the cumulation of mediations and the importance of uncertainty and learning processes before, but the results are inconclusive.⁴ Studies hardly find systematic support for the argument that the more previous mediations before an intervention under study, the more effective this latter effort will be. This empirical ambiguity points to two conclusions. First, although the existent work recognizes learning effects, reciprocity and thus a cumulative effect of mediations in conflicts, most scholars argue for a linear trend. As I will contend in the following, however, the claim for a linear trend seems implausible. Second, as a result of the first point, incorrect theoretical claims induce misleading empirical specifications. In fact, one could even think that those linearly specified empirical models actually capture the duration of a dispute—which, however, does not necessarily relate to a cumulative effect of mediations. Ultimately, we obtain spurious correlations that might not be able to tell us if and how mediations can exert a cumulative impact on resolution outcomes. Based upon my theoretical revisions in the next section, I thus extend existent empirical models accordingly as well.

To the best of my knowledge, Bercovitch et al. (1991) provide one exception in this context. These scholars do not find a linear relationship between the cumulation of mediations and mediation effectiveness, but instead a curvilinear trend. More specifically, Bercovitch et al. (1991: 13) emphasize that:

[o]ur data indicate a slight increase in the probability of successful mediation after one or two previous attempts (32%). After this point, however, the probability of success begins a long decline. A mediator entering a conflict after three or four previous attempts at mediation will have no better chance of mediating successfully than one who is the first to attempt mediation (the probability of success in both instances is 23%). If seven or more attempts at mediation have preceded a mediator’s intervention, the probability of success is just 13%. Even the most persistently mediated conflicts (i.e., with 10 or more mediation attempts) only achieve average success (22%).

These findings are summarized in Figure 1. Against this background, the utility of my study can be extended in a twofold sense. First, although Bercovitch et al. (1991) find empirical evidence for a curvilinear relationship that somewhat mirrors my own findings below, these scholars do not provide a theoretical argument for why this pattern is actually observed. In fact, Bercovitch et al. (1991) originally expected a linear relationship between the cumulation of mediations and effectiveness. Second, I also extend their work both theoretically and empirically by distinguishing attempts conducted by the same mediator(s) from interventions involving new third parties. Bercovitch et al. (1991) do not make this distinction.

Figure 1.

The cumulative impact of mediations on mediation effectiveness in Bercovitch et al. (1991). Graph displays predicted probabilities for effective mediation according to Bercovitch et al. (1991: 13).

The cumulative impact of international mediation—a theoretical revision

My theoretical framework building upon rationales about uncertainty and learning approaches departs from and adds to the existent work in the following ways. First, although I concur with the previous literature that mediation can in principle decrease uncertainty about the conflict and the reasons for its onset by shedding light on the “adversaries’ positions, potential compromises, and the costs of continued conflict” (Regan and Stam, 2000: 244; see also Beardsley, 2008: 725f), a single mediation or a few early, that is, initial intervention efforts are unlikely to substantially reduce uncertainty and to create a sufficient amount of trust. At the early stages of a conflict, the level of uncertainty is considerably high, the conflict as such may seem severely complex, the actors involved simply do not know much about their counterpart, and the belligerents’ ideal points are farthest away (Beardsley, 2008: 725f, 2011: 108; Blainey, 1988; Fearon, 1995; Kleiboer, 1996; Powell, 1999). Thus, I argue that early mediations are actually more likely to fail.⁵ If subscribing to this claim, the likelihood of failure might even increase over the course of initial mediations, since “repeated diplomatic failure may sour the parties on future initiatives, convincing them that such actions are a waste of time as their differences are irreconcilable” (Greig and Diehl, 2006: 362; see also Regan and Stam, 2000: 243). The early UN Security Council mediations in the conflict between India and Pakistan over the Kashmir region clearly provide an example here: essentially, all of these third-party interventions failed. In other words, I expect the rate of effectiveness to decrease with early, that is, initial mediation attempts in a dispute—despite the fact that a single mediation might be able to provide information to some extent.

This view is in line with existing learning theories, which argue that decision-makers rely on cognitive structures to order and prioritize incoming information and to make inferences about the environment (Bennett, 1999; Tetlock and McGuire, 1985). Since such structures shape what leaders see in the world, new incoming and perhaps contradictory information is often discounted (Pickering, 2002: 318) and the belligerents stick to their views that originally led to the outbreak of a conflict, ultimately causing initial mediations to be more likely to fail. However, I claim that this trend toward less effective outcomes is likely to change at some point over the cumulation of mediations. When disconfirming evidence becomes overwhelming, that is, when more and more mediations fail, “policymakers begin to reconsider their cognitive structures and the policies that such structures helped to forge” (Pickering, 2002: 318; see also Levy, 1994: 304f). As Tetlock (1991: 27f) notes, “policymakers reappraise basic premises only after repeated failures”. Therefore, while initial mediation efforts are likely to fail in generating an effective resolution of a conflict, the belligerents should become more receptive to the uncertainty-decreasing and trust-building impact of mediation at that point, when the failure of earlier mediations has indeed become overwhelming. As of this point, actors are more willing to learn from these previous efforts and to change their patterns of behavior. A sufficient amount of information has then been uncovered, the antagonist are open to the provided information by mediations and willing to reconsider their positions, and I expect mediations that occur after this turning point to be increasingly more likely to lead to an effective resolution of a conflict. Consistent with this argument, Greig and Diehl (2006: 362) state that “in this conception, it may actually be rational for a third party to offer mediation or for the parties to negotiate, knowing that such effort will fail; the expectation is that such failure is but a prelude, and indeed a prerequisite, to success in subsequent diplomatic efforts” (see also Heldt, 2009: 133).

Note that this argumentation somewhat mirrors the logic of the ripeness theory (Zartman, 1989, 2001; see also Greig and Diehl, 2006; Greig and Regan, 2008). In the words of Bercovitch (2004), “a ripe moment describes a phase in the life cycle of the conflict where the parties feel exhausted and hurt, or where they may not wish to countenance any further losses and are prepared to commit to a settlement, or at least believe one to be possible”. In other words, mediation tends to work best before conflicts become “very heated”, although they have to become heated enough for the parties to feel a need to resolve them (e.g. Greig and Diehl, 2006; Greig and Regan, 2008; Regan and Stam, 2000). Thus, there is a very small space in which the conflict is ripe for third-party intervention. If mediation occurs before this space, it is unlikely to succeed. In light of this, mediators do not necessarily enter a dispute when it is ripe for resolution, but rather get involved before an effective resolution can be achieved. Since a conflict is not yet ripe for effective third-party mediation under these circumstances I outlined above, earlier attempts are more likely to fail. However, despite their failure, past mediations ripen the conflict for subsequent resolution and, after a sufficient point of ripeness, that is, when the turning point has been reached, we observe an increasing probability of mediation effectiveness.⁶ Based upon this, I depart from the postulated linear relationship between the cumulation of mediations and mediation effectiveness by arguing for a curvilinear relationship.⁷

Hypothesis 1. The relationship between a series of mediations and mediation effectiveness is characterized by a U-shaped relationship.

Second, I also add to the existing literature by arguing that the postulated relationship from the first hypothesis should be particularly strong when a cumulation of mediations is performed by the same mediator(s). This claim is based upon how information about interests, intentions, and so on, from previous efforts is transmitted to present attempts. Recall that any mediation attempt “sets a precedent on which the next attempt can build” (Heldt, 2009: 137f). However, a state’s own experience from previous attempts is generally likely to weigh more heavily on policy-makers than the experiences of others (Huth and Russett, 1984; Jervis, 1976; Leng, 1983; Levite et al., 1992; Levy, 1994; Pickering, 2002; Reiter, 1996). Conflict resolution crucially depends on the “credible and truthful transmission of information” (Dorussen and Ward, 2008: 193; see also Greig and Diehl, 2006: 363ff). If the same mediators intervene, they can rely on a direct transmission of information, that is, a dense and short connection of information flows, and a mediator does not have to gather “indirect information” from intermediary sources, that is, other mediators (Böhmelt, 2009; Dorussen and Ward, 2008). Dense and short connections to information pools convey information with lower costs, more precisely and faster (Dorussen and Ward, 2008: 197). This ultimately leads to even more decreased uncertainty and even more trust, as well as more strongly established mutually accepted norms than in those cases where mediators must rely on other sources (see Dorussen and Ward, 2008: 194f; Hafner-Burton and Montgomery, 2006). According to this mechanism, it is generally plausible that stronger and more efficient information flows, that is, if a mediator can rely on her own past experience, lead to a higher level of mediation effectiveness.⁸

In turn, I argue that the experience of other mediators’ interventions is less vivid, more difficult to recall, and hence less relevant (Pickering, 2002). Information about interests, intentions and understandings is disseminated here via other mediators that intervened previously and is not dependent on one’s own experience alone. These longer chains in a network of mediators (Böhmelt, 2009; Dorussen and Ward, 2008) might then be able to transmit information and experience in order to affect learning processes, but these indirect sources of information are less likely to reveal private information, and they are less efficient and effective than direct experiences. That being said, indirect sources of information, that is, previous mediations conducted by other actors, can be (imperfect) substitutes for “direct sources” when the same mediators might not be able (or willing) to intervene after they did before (see Böhmelt, 2009; Dorussen and Ward, 2008: 195). Therefore, mediation efforts of others are still likely to lead to the outcome as postulated in the first hypothesis, but this should be less strongly pronounced than in cases where mediators can draw upon their own experience.⁹

Hypothesis 2. The expected effect postulated in H1 should be stronger (weaker) if the same mediators (other mediators) intervene over the course of a dispute.

Research design

Data, dependent variable and methodology

For testing the empirical implications of the theoretical framework, I employ data from the ICOW project (Hensel and Mitchell, 2007; see also Hensel, 2001; Hensel et al., 2008). These data cover territorial claims in the Western Hemisphere and Western Europe, river claims in the previous two regions plus the Middle East, and maritime claims in the Western Hemisphere and Europe. Claims are identified according to explicit evidence of contention involving official representatives of two or more nation states over the issue type in question. The advantages of the ICOW data are twofold.¹⁰ First, we can compare cases across various issues of contention. Second, the data’s cases are not limited to situations that become militarized at some point, which addresses the problem of selection bias to some extent. The temporal domain of my study is from 1816 to 2001, while the data are essentially cross-sectional. However, some temporal dependence may persist as I argue that interventions are unlikely to be independent of previous conflict management efforts over a claim between a specific dyad (see also Greig and Regan, 2008; Melin, 2011). To address this problem, I cluster the standard errors on each claim per dyad and consider a duration variable that is specified below.

I rely on Touval and Zartman (1985) for capturing the concept of mediation. Hence, I do not consider cases that experienced military interventions or bilateral conflict management, but attempts to settle each ICOW claim through peaceful third-party interventions.¹¹ The unit of analysis comprises a contentious issue claim with a third-party mediation attempt.

The dependent variable measures the effectiveness of third-party settlement attempts along five categories. First, a third-party mediation may be completely ineffective and there is no agreement among the belligerents. Second, the antagonists may be able to reach an agreement with the help of third parties, but at least one of the disputants does not ratify the treaty. Third, even if there is an agreement that addresses the issues that led to the dispute in the first place, at least one of the belligerents does not comply with it. Fourth, although all of the warring parties comply with the agreement that was reached, this still may not end the claim. Finally, if all conflict states comply with an agreement, the claim is ended, and the effectiveness of a third-party intervention is considered to be at its maximum. Since the dependent variable follows an ordinal scale, I use ordered logit regression models to test my hypotheses. Table 1 gives an overview of effectiveness.

Table 1.

The effectiveness of international mediation—descriptive statistics, 1816–2001

	Frequency	Percentage
No agreement	1071	53.42
Agreement	145	7.23
Ratification	74	3.69
Compliance	496	24.74
End of claim	219	10.92
Total	2005	100.00

Core explanatory variables: the cumulation of mediations

For the operationalization of the key explanatory variables, I rely on the following two approaches. First, consistent with the previous literature, the cumulation of mediations is measured by a variable counting the number of previous mediation efforts in a dispute (previous mediation attempts).¹² Out of 2005 observations in my data, about 48% have seen at least one previous mediation attempt (N = 962). In turn, the mean value for that subset that only considers issue claims with at least one previous mediation is 4.13 preceding third-party interventions, with a minimum of 1 (N = 339/962) and a maximum of 23 (N = 4/962). Not surprisingly, the frequency of observations decreases with a larger number of previous mediation attempts. For example, already more than 75% of those observations with at least one previous mediation have seen five or fewer earlier attempts (N = 758/962). On the other hand, about 9% of those observations with at least one preceding third-party intervention have more than 10 previous mediation attempts (N = 87/962). In order to model the hypothesized curvilinear relationship between the cumulation of mediations and effectiveness, I also include the square term of this variable in my models.

Further, the second hypothesis argues that the cumulative impact of mediations should be stronger if the same actors mediate over the course of a dispute. Thus, I also use a variable that counts how many previous mediations were conducted by that mediator of the intervention under study (previous mediation attempts (same actor)). While this operationalization seems straightforward in the case of unilateral mediations, it becomes more difficult if more than one mediator is present, that is, if a coalition of actors intervened and multiparty mediation occurred (see Böhmelt, 2011, 2012). I decided to employ a relatively conservative approach by considering only the exact same set of actor(s). Put differently, if a third-party intervention under study is a multiparty mediation, my count item here only considers those previous mediations that were conducted by the exact same composition of actors. Hence, one deviation from that composition is sufficient for it not being taken into account by this variable. In descriptive terms, out of 2005 observations in my data, fewer than 10% have seen at least one previous mediation attempt with the same set of actors (N = 194). The mean value for that subset only considering earlier mediations with the same set of interveners is 2.26, with a minimum of 1 (N = 89/194) and a maximum of 9 (N = 1/194). Moreover, out of these 194 issue claims with at least one previous settlement attempt, fewer than 5% have seen more than five previous mediation attempts that comprised the same actors. Similar to the specification for the first hypothesis, I include a square term of this variable as well.

Control variables

I also control for other influences of mediation effectiveness, although they are not explicitly addressed by my theory (see Kleiboer, 1996). First, any mediation attempt before an intervention under study must have failed at least to some extent. Otherwise, an issue claim would have been fully resolved and the dispute over this claim completely ended. However, the degree to which these mediations have failed differs. For example, while one intervention might have reached at least some level of agreement among the belligerents (effectiveness_t_–2 = 1), a succeeding mediation could have failed completely according to the effectiveness measure above (effectiveness_t_–1 = 0). In order to control for these different “degrees of effectiveness”, I consider an effectiveness ratio of previous attempts, that is, the count of all previous mediations that were partially effective divided by the count of all earlier mediations a conflict has seen so far.

Second, in order to show that my operationalizations for the cumulation of mediations do not simply pick up the duration or level of ripeness of an issue claim, I incorporate the following controls that were introduced in the “hurting-stalemate” literature (e.g. Bercovitch, 2004; Greig and Diehl, 2006; Greig and Regan, 2008: 761). Hence, despite the similarities between my argument and the ripeness theory (Zartman, 1989, 2001), these two mechanisms are essentially different and the proclaimed effect of a curvilinear impact of the cumulation of mediations should hold even when controlling for those variables in the hurting-stalemate literature. To this end, the intensity or importance of an issue claim is captured by salience, which measures the characteristics of an issue claim. Following Hensel and Mitchell (2007: 5), the sample’s variable range is [0; 12], with higher values indicating greater salience. Duration, that is, a variable that measures the time elapsed from the start of an issue claim until a mediation attempt under study occurred, addresses any persisting temporal dependencies and the argument that mediation attempts should be less likely to succeed in long-standing conflicts (Beardsley et al., 2006; Bercovitch and Gartner, 2006; Bercovitch et al., 1991; Kressel and Pruit, 1989). It also captures the claim that information problems decrease with conflict duration, as disputants are likely to learn over the course of a crisis independent from any third-party intervention (e.g. Blainey, 1988; Fearon, 1995, 2004: 290). Although the previous two items should measure the intensity of a dispute fairly accurately, the salience and duration items might not fully capture influences like fatalities over the course of a dispute. Therefore, I also include a dummy variable indicating whether a mediation attempt is conducted during an ongoing militarized interstate dispute (MID) and binary variables for whether a dispute broke out owing to river or territorial claims, since the type of issue claim might also influence the degree of intensity of conflicts (Hensel, 2001; see Hensel et al., 2008). Maritime claims, which are included in the ICOW data as well, constitute the baseline category.

Third, I include the natural log of the ratio of the stronger belligerent country’s capability to that of the weaker antagonist (Singer et al., 1972). The rationale for this item stems from the fact that mediation is more likely to succeed when neither side has a clear military advantage (e.g. Bennett and Stam, 1996: 242; Mason and Fett, 1996: 550). On the contrary, power discrepancies increase the risk that the more powerful party has few incentives to agree to a peaceful settlement as it stands a better chance of winning a conflict militarily.

Fourth, I also consider the fighting parties’ regime type. The data are taken from Polity IV’s polity2 item (Marshall and Jaggers, 2002), relying on a weakest-link specification, where the least democratic country in a belligerent dyad influences the likelihood of mediation effectiveness. Democracies have a common structural background and share institutional procedures, which increases the chances that the antagonists agree on a peaceful settlement (Allee and Huth, 2006; Dixon, 1993, 1994; Greig, 2005; see also Böhmelt, 2011).¹³

Finally, the style of a mediation attempt may influence the prospects for peaceful settlements. Dixon (1996), for example, finds that more enforcing strategies seem to induce more effective outcomes (see also Beardsley et al., 2006). The ICOW data include dichotomous variables on functional techniques (which attempt to address the use of the claimed territory but do not address sovereignty questions), procedural attempts (which address future efforts to settle a claim but do not address sovereignty directly) and substantive attempts (which address sovereignty over part or all of a claimed territory). I include a dichotomous variable for procedural and substantive settlements, respectively, while functional attempts are used as a baseline category.¹⁴

Empirical analysis

The results of my analysis are summarized in Table 2, where I estimate four models. First, model 1 focuses on the cumulative impact of mediation per se while other explanatory variables or the item’s focus on whether the exact same actor(s) mediated before are not considered. Second, model 2 substitutes the original mediation cumulation variables for those variables that take into account the composition of previous mediations. Finally, models 3 and 4 follow the same procedure as models 1 and 2, respectively, but here I include the control covariates as well. Therefore, models 1 and 3 test my first hypothesis, while models 2 and 4 examine the validity of the second hypothesis. Although the signs and standard errors of nonlinear models’ coefficients can be interpreted directly, they cannot be considered as slopes or elasticities. Hence, to unveil the implied magnitude of my substantial findings, Table 3 reports the average change in the predicted probability across categories of effectiveness as an explanatory variable changes from its minimum to its maximum (while holding all other variables at their means).

Table 2.

The cumulative impact of international mediation, 1816–2001

	Model 1	Model 2	Model 3	Model 4
Previous mediation attempts	−0.143		−0.132
	(0.036)***		(0.040)***
Previous mediation attempts²	0.008		0.006
	(0.002)***		(0.002)***
Previous mediation attempts (same actor)		−0.211		−0.785
		(0.114)*		(0.173)***
Previous mediation attempts (same actor)²		0.045		0.156
		(0.022)**		(0.037)***
Effectiveness ratio of previous attempts			0.404	0.624
			(0.273)	(0.347)*
Territorial claim			−0.041	−0.043
			(0.149)	(0.152)
River claim			−0.131	−0.154
			(0.232)	(0.236)
MID			−0.333	−0.340
			(0.185)*	(0.185)**
Duration			−0.000	−0.000
			(0.000)	(0.000)
Procedural settlement			−0.281	−0.323
			(0.140)**	(0.140)**
Substantive settlement			−0.575	−0.573
			(0.159)***	(0.150)***
Salience			−0.048	−0.057
			(0.026)*	(0.026)**
Capability ratio			−0.097	−0.092
			(0.035)***	(0.034)***
Democracy			−0.008	−0.009
			(0.010)	(0.009)
Observations	2005	2005	1657	1657
Log pseudo likelihood	−2463.10	−2472.48	−2177.35	−2174.99

Robust standard errors clustered on dyad claim in parentheses.

Significant at 10%; ** significant at 5%; *** significant at 1% (two-tailed).

Table 3.

Predicted probabilities of mediation effectiveness, 1816–2001

	Model 1	Model 2	Model 3	Model 4
Previous mediation attempts	−19.8%		−21.7%
Previous mediation attempts²	31.2%		26.2%
Previous mediation attempts (same actor)		−14.3%		−23.4%
Previous mediation attempts (same actor)²		28.3%		34.5%
Effectiveness ratio of previous attempts			4.0%	6.2%
Territorial claim			−0.4%	−0.4%
River claim			−1.3%	−1.5%
MID			−3.3%	−4.0%
Duration			−2.5%	−3.5%
Procedural settlement			−2.8%	−3.2%
Substantive settlement			−5.7%	−5.7%
Salience			−5.7%	−6.7%
Capability ratio			−9.2%	−8.7%
Democracy			−1.6%	−1.8%

Change of predicted probabilities of effectiveness is shown for each independent variable when moving from the minimum to the maximum; all other variables held at their mean values; significant probability estimates in bold.

Before discussing the main results, I briefly discuss the findings of the control covariates. Owing to space limitations, however, I primarily focus on those variables that are statistically significant. First, MID has the expected negative sign and is significant at the 10% level (at least). If an issue conflict comprises a continuing MID, the effectiveness of a mediation effort is substantially lower. Arguably, this seems to stem from the higher intensity of such a dispute. Furthermore, my findings for the different styles of mediations reveal surprising insights. The variables procedural and substantive settlements are negatively signed, meaning that these strategies perform significantly worse than less enforcing, functional techniques. Although this goes against earlier studies (e.g. Beardsley et al., 2006; Dixon, 1996), my inferences could be influenced by the use of the data approach I chose, as indicated in footnote 10. Third, when moving from the minimum toward the maximum of salience, the predicted probability of reaching the most effective settlement decreases on average by about 7% across each category of effectiveness. This essentially mirrors the finding for MID. Fourth, my estimations support the claim on belligerents’ power discrepancies and that they decrease the chances of an effective peaceful settlement. Substantially, when moving from the minimum toward the maximum of capability ratio, the predicted probability of reaching the most effective outcome decreases on average by about 9% across each category of effectiveness. Finally, duration, that is, the time elapsed from the start of an issue claim until a mediation attempt under study occurred, merits special attention owing to my argument on information, learning and the importance of time therein. This variable, however, is statistically insignificant in either model estimation, lending support to the claim above that my operationalization of the cumulation of mediations does not pick up the duration of a dispute as such.

Turning to the variables of main interest, I find strong support for my hypotheses. First, the results confirm the first hypothesis arguing for a nonlinear relationship between the cumulation of mediations and mediation effectiveness, thereby emphasizing that the previous work might have overlooked the theoretical revision I suggest in this paper. Figures 2 and 3 illustrate this relationship graphically. While the x-axis pertains to the number of previous mediation attempts (either from the same actor(s) or not), the predicted probabilities of seeing either value 3 or 4 on effectiveness are shown at the y-axis. According to these graphs, effective mediations have a baseline probability of 42–50% where no previous mediations occurred. If an intervention can rely upon earlier mediations, however, the probability of effectiveness first decreases and then, after a turning point has been reached, substantially increases again. Table 2 shows that the turning point associated with a cumulation of mediations lies at around 10 previous mediations when the composition of earlier attempts is not considered (about 2.5 for those cases where the exact same mediator(s) intervened before). Therefore, on one hand, there is evidence that belligerents and mediators alike do learn from previous interventions. On the other hand, however, a single mediation or a few previous mediations may not be sufficient enough, and it also seems—paradoxically then—that earlier mediations are indeed more likely to fail in order to ultimately achieve a peaceful resolution of a conflict (Heldt, 2009: 133). The curvilinear relationship as argued for in my first hypothesis and depicted in Figures 2 and 3 is in its form similar to Bercovitch et al. (1991: 13) and robust over a wide variety of model specifications. Adding or suppressing controls from the models does not alter this result.

Figure 2.

The cumulative impact of mediations on mediation effectiveness, 1816–2001. Graphs display predicted probabilities for effective mediation attempts based on model 1 (black solid line) and model 2 (gray solid line). Dashed lines pertain to 90% confidence intervals. All variables apart from the variable at the x-axis and its square term are held at their means.

Figure 3.

The cumulative impact of mediations on mediation effectiveness, 1816–2001. Graphs display predicted probabilities for effective mediation attempts based on model 3 (black solid line) and model 4 (gray solid line). Dashed lines pertain to 90% confidence intervals. All variables apart from the variable at the x-axis and its square term are held at their means.

I also obtain strong support for my second hypothesis. Although mediators of a current intervention seem to draw lessons from earlier efforts even if these specific actors did not mediate before, mediators and the belligerents apparently learn more effectively from and are more receptive to earlier attempts in which the same third parties intervened. This observation is based on two findings. First, the graphs based on models 2 and 4 reach the turning point more quickly than those accumulations of mediations that do not take into account the composition of previous attempts. More specifically, the turning point at which negative margins of effectiveness turn into positive ones is at 2.3 and 2.5 in models 2 and 4, respectively. That being said, the rate of effective mediations does not increase until nine (model 1) or 11 (model 3) previous interventions when focusing on the cumulation of mediations that have not necessarily been conducted by the same third parties. Second, a cumulation of mediations with the same intervener(s) needs fewer efforts to reach an effective settlement of an issue claim. While these mediation sequences take fewer than 10 settlement attempts, those mediation sequences that do not control for the identity of previous mediators do need more than 20 third-party interventions in order to reach an equally high level of effectiveness.

However, although the cumulation of mediations of the same actors is more effective than those that comprise different mediators, it is also evident that my first hypothesis holds regardless of the actor compositions of preceding interventions. In other words, information is being transferred as well, yet indirectly. Precisely these indirect sources of experience can work as (imperfect) substitutes within the context of mediation when the same mediators might not be able (or willing) to intervene after they did before (see Böhmelt, 2009; Dorussen and Ward, 2008: 195). In sum, learning through the mediation efforts of others does occur and it does matter, but it is less strongly pronounced than in cases where mediators can draw upon their own experience.

Robustness

In order to ensure the robustness of my findings, I changed a variety of model specifications and re-ran the estimates. First, Clarke (2005) shows that the inclusion of control variables may actually increase the bias instead of decreasing it. However, Tables 2 and 3 and Figures 2 and 3 demonstrate that making amendments in this regard does not affect the substance of my findings.

Second, the dependent variable’s “ratification” category has only 74 observations, meaning that this value could be an outlier category, which overestimates my results. In order to address this, I recoded effectiveness by grouping the “agreement” and “ratification” categories together. Similarly, given my rather small sample of events, I also considered all models using a less stratified dependent variable by employing a dichotomous item for effectiveness with the value of 0 matching categories 0–2 and the value of 1 matching categories 3 and 4 of the original variable. Both changes did not alter my core findings.

Third and as indicated above, I considered all kinds of voluntary third-party interventions, which includes both limited forms of mediation and heavy mediations such as arbitration and adjunction (Rauchhaus, 2006: 224). In order to rule out the possibility that my results are driven by the latter category, I re-estimated all models but only considered unbinding third-party mediations. The findings virtually stay the same though.

Fourth, before we can examine the factors driving mediation effectiveness, mediation has to occur. There are many underlying factors that are likely to influence both the decision to mediate and mediation effectiveness (e.g. Beber, 2012; Böhmelt, 2010; Gartner and Bercovitch, 2006). However, the decision to mediate is itself a strategic consideration and thus we might face problems of selection bias. Although I tried to mitigate the consequences of selection bias with several strategies in the research design, these problems may persist for some covariates. In order to deal with this issue in more depth, I ran alternative estimations using a Heckman selection model (Heckman, 1979). For these calculations, the specifications are as follows. First, I employed the entire ICOW dataset, that is, I did not drop cases without third-party mediation and, hence, used as my unit of analysis any contentious issue whether or not it attracted settlement efforts. Second, the onset of mediation signified the dependent variable in the selection equation and I used effectiveness as the dependent item for the outcome stage. Finally, the explanatory variables for the first stage comprised the control variables from above, which also represent usual factors that explain mediation onset, that is, the intensity of a claim dispute, belligerents’ regime type and their capability scores as well as variables for modeling time dependencies.¹⁵ I did not change the model specifications from my core estimations above in order to model the second stage. The results from these calculations, nevertheless, showed that it is unlikely that unmeasured influences in the first stage associated with unmeasured influences in the outcome equation bias my parameter estimates. The parameter ρ, which indicates whether the error terms in the selection and the outcome equation are correlated, is statistically insignificant. Furthermore, my core results do not change substantially in the selection model, and I still obtain evidence for a U-shaped relationship between the cumulation of mediations and effectiveness.

Finally, the operationalization for those variables that consider whether the exact same (set of) actor(s) mediated over the cumulation of mediations follows a relatively strict and conservative coding decision. For unreported model estimations, I somewhat relaxed this by also counting those previous mediation attempts in which at least one of the actor(s) of a mediation effort under study intervened. This change, nevertheless, did not affect my results either.

Conclusion

Does the cumulation of mediations positively influence peaceful settlement attempts in conflicts? What is the underlying mechanism for this? The existent literature dealt with these questions extensively, but its results remained ambiguous. In order to shed more light on this, I revised and extended previous theoretical accounts in two ways. First, initial mediations are more likely to fail because the level of uncertainty and complexity may be highest under those circumstances and the antagonists’ ideal points are farthest away at the beginning of a conflict (Beardsley, 2008: 725f, 2011: 108; Blainey, 1988; Fearon, 1995; Powell, 1999). Based upon this, I developed an argument for a U-shaped relationship between the cumulation of mediations and mediation effectiveness. Second, I then adopted previous arguments for the claim that experiences and learning develop more strongly if the same actors mediate over the course of a dispute—without denying that indirect information transmissions may work as a substitute for direct effects. My empirical findings strongly supported both stated hypotheses.

This research might have several important implications for both academia and policy-making. First, I believe this work helped us improving our understanding of how international mediation can be effective and, more generally, of how single mediation attempts add to the cumulation of third-party interventions. Therefore, with regard to policy advice, decision-makers belonging to either side in a conflict, that is, any belligerent or mediator, should not be discouraged if initial mediation efforts fail. On the contrary, these interventions essentially only pave the way toward an effective and ultimate settlement of a dispute and are by no means indicators for the chances of eventual success (see also Heldt, 2009).

Second, the length of this path toward ultimate peace substantially depends on who mediates. Effective outcomes can be achieved more quickly if the same mediators establish a cumulation of mediations. Hence, the belligerents and mediators alike should have a strong interest in the actor composition staying the same over the course of a conflict—given sincere motives. In other words, the same set of mediators should generally be preferred over different interveners, although different mediators still seem to be more effective than cases without any kind of third-party intervention.

From a scholarly perspective, third, my study not only tried to close a gap in the existent research on the cumulation of mediations, but also emphasized again that mediation attempts cannot be studied independently from each other (see also Greig and Regan, 2008; Melin, 2011). From a narrow and arguably naive point of view, if a single third-party intervention fails, it is ineffective. However, broadening the perspective and examining the effectiveness of this single attempt after integrating it to the whole cumulation of mediations over the course a conflict may lead to the conclusion that the picture of intervention success looks inherently differently. Therefore, my research contributes to the existent work highlighting the inconsistencies of mediation that may persist (e.g. Beardsley, 2008, 2011).

Finally, in particular, Heldt (2009) points to the different types of third-party interventions and even different strategies or tools over the course of a dispute. For example, there are functional techniques, procedural attempts and substantive attempts; there are good offices, arbitration, adjunction or peacekeeping. Heldt (2009: 133) concludes accordingly that, given the different characteristics of these, “it appears inconceivable that the choice of tool is inconsequential for the outcome” of a peace process. Hence, although I controlled for the different styles and techniques of mediation attempts, it seems worth making the effort to further disaggregate the cumulation of mediations against this background.

Footnotes

Acknowledgements

I am grateful for a research grant provided by the Folke Bernadotte Academy and I thank Lena Kiesewetter for her excellent research assistance. Three anonymous reviewers and the editor of CMPS, Glenn Palmer, also provided valuable input that helped to improve the paper. The data and syntax code for replicating the article’s results can be obtained from the author upon request.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

1.

A previous version of this paper was presented at the Annual Convention of the International Studies Association (San Diego, CA, USA, 1–4 April 2012) and the Jan Tinbergen European Peace Science Conference (Berlin, Germany, 25–27 June 2012).

Notes

References

Allee

Huth

(2006) Legitimizing dispute settlement: International legal rulings as domestic political cover. American Political Science Review 100(2): 219–234.

Beardsley

(2008) Agreement without peace? International mediation and time inconsistency problems. American Journal of Political Science 52(4): 723–740.

Beardsley

(2011) The Mediation Dilemma. Ithaca, NY: Cornell University Press.

Beardsley

Quinn

Biswas

Wilkenfeld

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

Beber

(2012) International mediation, selection effects, and the question of bias. Conflict Management and Peace Science 49(4): 397–424.

Beck

Katz

Tucker

(1998) Taking time seriously: Time-series cross-section analysis with a binary dependent variable. American Journal of Political Science 42(4): 1260–1288.

Bennett

(1999) Condemned to Repetition? The Rise, Fall, and Reprise of Soviet–Russian Military Interventionism, 1973–1996. Cambridge, MA: MIT Press.

Bennett

Stam

(1996) The duration of interstate wars, 1816–1985. American Political Science Review 90(2): 1260–1288.

Bercovitch

(2011) Theory and Practice of International Mediation: Selected Essays. New York: Routledge.

10.

Bercovitch

(2002) Introduction: Putting mediation in context. In: Bercovitch

(ed.) Studies in International Mediation. New York: Palgrave Macmillan, pp. 3–24.

11.

Bercovitch

(2004) International mediation and intractable conflict. In: Burgess

Burgess

(eds) Beyond Intractability. University of Colorado, BO: Conflict Information Consortium.

12.

Bercovitch

Gartner

(2006) Is there method in the madness of mediation? Some lessons for mediators from quantitative studies of mediation. International Interactions 32(4): 329–354.

13.

Bercovitch

Houston

(2000) Why do they do it like this? An analysis of the factors influencing mediation behavior in international conflicts. Journal of Conflict Resolution 44(2): 170–202.

14.

Bercovitch

Anagnoson

Wille

(1991) Some conceptual issues and empirical trends in the study of successful mediation in international relations. Journal of Peace Research 28(1): 7–17.

15.

Blainey

(1988) The Causes of War. New York: The Free Press.

16.

Böhmelt

(2009) International mediation and social networks: The importance of indirect ties. International Interactions 35(3): 298–319.

17.

Böhmelt

(2010) The effectiveness of tracks of diplomacy strategies in third-party interventions. Journal of Peace Research 47(2): 167–178.

18.

Böhmelt

(2011) Disaggregating mediations: The impact of multiparty mediation. British Journal of Political Science 41(4): 859–881.

19.

Böhmelt

(2012) Why many cooks if they can spoil the broth? The determinants of multiparty mediation. Journal of Peace Research 49(5): 701–715.

20.

Clarke

(2005) The phantom menace: Omitted variable bias in econometric research. Conflict Management and Peace Science 22(4): 341–352.

21.

Dixon

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

22.

Dixon

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

23.

Dixon

(1996) Third-party techniques for preventing conflict escalation and promoting peaceful settlement. International Organization 50(4): 653–681.

24.

Dorussen

Ward

(2008) Intergovernmental organizations and the Kantian peace—A network perspective. Journal of Conflict Resolution 52(2): 189–212.

25.

Ellis

Mitchell

Prins

(2010) How democracies keep peace: Contextual factors that influence conflict management strategies. Foreign Policy Analysis 6(4): 373–398.

26.

Fearon

(1995) Rationalist explanations for war. International Organization 49(3): 379–414.

27.

Fearon

(2004) Why do some civil wars last so much longer than others? Journal of Peace Research 41(3): 275—301.

28.

Gartner

Bercovitch

(2006) Overcoming obstacles to peace: The contribution of mediation to short lived conflict settlements. International Studies Quarterly 50(4): 819–840.

29.

Gent

Shannon

(2011) Decision control and the pursuit of binding conflict management: Choosing the ties that bind. Journal of Conflict Resolution 55(5): 710–734.

30.

Greig

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

31.

Greig

Diehl

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

32.

Greig

Regan

(2008) When do they say yes? An analysis of the willingness to offer and accept mediation in civil wars. International Studies Quarterly 52(4): 759–781.

33.

Hafner-Burton

Montgomery

(2006) Power positions—International organizations, social networks, and conflict. Journal of Conflict Resolution 50(1): 3–27.

34.

Heckman

(1979) Sample selection bias as a specification error. Econometrica 47(1): 153–161.

35.

Heldt

(2009) Sequencing of Peacemaking in Emerging Conflicts. In: Aggestam

Björkdahl

(eds) War and Peace in Transition. Changing Roles of External Actors. Lund: Nordic Academic Press, pp. 128–146.

36.

Hensel

(2001) Contentious issues and world politics: The management of territorial claims in the Americas, 1816–1992. International Studies Quarterly 45(1): 81–109.

37.

Hensel

Mitchell

(2007) Issues Correlates of War Project: User Manual for ICOW Data, Version 1.1. Available at: http://www.icow.org

38.

Hensel

Mitchell

Sowers

Thyne

(2008) Bones of contention: Comparing territorial, maritime, and river issues. Journal of Conflict Resolution 52(1): 117–143.

39.

Huth

Russett

(1984) What makes deterrence work? Cases from 1900 to 1980. World Politics 36(4): 496–526.

40.

Jervis

(1976) Perception and Misperception in International Politics. Princeton, NJ: Princeton University Press.

41.

Kelman

(1996) The interactive problem-solving approach. In: Crocker

Hampson

Aall

(eds) Managing Global Chaos: Sources of and Responses to International Conflict, Washington, DC: United States Institute of Peace Press, pp. 501–519.

42.

Kleiboer

(1996) Understanding success and failure of international mediation. Journal of Conflict Resolution 40(2): 360–389.

43.

Kressel

Pruitt

(1989) Conclusion: A research perspective on the mediation of social conflict. In: Kressel

Pruitt

(eds) Mediation Research. San Francisco, CA: Jossey-Bass, pp. 241–262.

44.

Kydd

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

45.

Kydd

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

46.

Lederach

(1997) Building Peace: Sustainable Reconciliation in Divided Societies. Washington, DC: United States Institute of Peace Press.

47.

Leng

(1983) When will they ever learn? Coercive bargaining in recurrent crises. Journal of Conflict Resolution 27(3): 379–419.

48.

Leng

(1998) Reciprocity in recurring crises. International Negotiation 3: 197–226.

49.

Levite

Jentleson

Berman

(eds) (1992) Foreign Military Intervention: The Dynamics of Protracted Conflict. New York: Columbia University Press.

50.

Levy

(1994) Learning and foreign policy: Sweeping a conceptual minefield. International Organization 48(2): 279–312.

51.

Marshall

Jaggers

(2002) POLITY IV Project: Political Regime Characteristics and Transitions, 1800–2002. Dataset User’s Manual. College Park, MD: University of Maryland.

52.

Mason

Fett

(1996) How civil wars end: A rational choice approach. Journal of Conflict Resolution 40(4): 546–568.

53.

Melin

(2011) The impact of state relationships on if, when and how conflict management occurs. International Studies Quarterly 55(3): 691–715.

54.

Morgan

(1995) Untying the Knot of War: A Bargaining Theory of International Crises. Ann Arbor, MI: University of Michigan Press.

55.

Pickering

(2002) War-weariness and cumulative effects: Victors, vanquished, and subsequent interstate intervention. Journal of Peace Research 39(3): 313–337.

56.

Pillar

(1983) Negotiating Peace: War Termination as a Bargaining Process. Princeton, NJ: Princeton University Press.

57.

Powell

(1999) In the Shadow of Power: States and Strategies in International Politics. Princeton, NJ: Princeton University Press.

58.

Powell (2004) Bargaining and learning while fighting. American Journal of Political Science 48(2): 344–361.

59.

Princen

(1992) Intermediaries in International Conflict. Princeton, NJ: Princeton University Press.

60.

Pruitt

Carnevale

(1993) Negotiation in Social Conflict. Pacific Grove, CA: Brooks/Cole.

61.

Quinn

Wilkenfeld

Smarick

Asal

(2006) Power play: Mediation in symmetric and asymmetric international crises. International Interactions 32(40): 442–470.

62.

Rauchhaus

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

63.

Regan

Aydin

(2006) Diplomacy and other forms of intervention in civil wars. Journal of Conflict Resolution 50(5): 736–756.

64.

Regan

Stam

(2000) In the nick of time: Conflict management, mediation timing, and the duration of interstate disputes. International Studies Quarterly 44(2): 239–260.

65.

Reiter

(1996) Crucible of Beliefs: Learning, Alliances, and World Wars. Ithaca, NY: Cornell University Press.

66.

Rubin

(1992) Conclusion: International mediation in context. In: Bercovitch

Rubin

(eds) Mediation in International Relations: Multiple Approaches to Conflict Management. New York: St Martin’s Press, pp. 249–272.

67.

Savun

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

68.

Singer

Bremer

Stuckey

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

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

69.

Smith

Stam

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

70.

Stein

JGr

(1996) Image identity and conflict resolution. In: Crocker

Hampson

Aall

(eds) Managing Global Chaos: Sources of and Responses to International Conflict. Washington, DC: United States Institute of Peace, pp. 93–101.

71.

Svensson

(2007) Bias, bargaining, and peace brokers: How rebels commit to peace. Journal of Peace Research 44(2): 177–194.

72.

Svensson

(2009) Who brings which peace: Biased versus neutral mediation and institutional peace arrangements in civil wars. Journal of Conflict Resolution 53(3): 446–469.

73.

Tetlock

(1991) Learning in US and Soviet foreign policy: In search of an elusive concept. In: Breslauer

Tetlock

(eds) Learning in US and Soviet Foreign Policy. Boulder, CO: Westview Press, pp. 20–61.

74.

Tetlock

McGuire

(1985) Cognitive perspectives on foreign policy. In: Long

(ed.) Political Behavior Annual. Boulder, CO: Westview Press, pp. 255–273.

75.

Touval

Zartman

(1985) Introduction: Mediation in theory. In: Touval

Zartman

(eds) International Mediation in Theory and Practice. Boulder, CO: Westview Press, pp. 7–17.

76.

UN (2011) Summary Report—United Nations Mediation: Experiences and Reflections from the Field. New York: United Nations.

77.

Wilkenfeld

Young

Asal

Quinn

(2003) Mediating international crises: Cross-national and experimental perspectives. Journal of Conflict Resolution 47(3): 279–301.

78.

Zartman

(1989) Ripe for Resolution: Conflict and Intervention in Africa. Oxford: Oxford University Press.

79.

Zartman

(2001) The timing of peace initiatives: Hurting stalemates and ripe moments. The Global Review of Ethnopolitics 1(1): 8–18.

80.

Zubek

Pruitt

Peirce

McGillicuddy

Syna

(1992) Disputant and mediator behavior affecting short term success in mediation. Journal of Conflict Resolution 36(3): 546–572.