Shared identity and bargaining outcomes

The literature on conflict resolution and bargaining draws heavily on the rationalist bargaining framework in which parties in conflict proffer settlement offers that may be accepted, resolving the conflict at least temporarily, or rejected, leading either to war or to continued opportunities to make offers. Offers are derived in equilibrium from the preferences of the parties, the outside options and/or cost of war, and, usually, the timing of the offers made. Each of these may be common knowledge among the parties or may be uncertain, although typically the nature of the uncertainty (i.e. the distributions from which the relevant parameters are drawn) is common knowledge even when the parties’ preferences are not.

As should be apparent from this exceedingly brief summary of the rationalist bargaining model (see Powell Reference Powell2002 for a review of this literature), there is nothing explicitly prohibiting the incorporation of identity. Yet, until recently, identity played little role in most rationalist accounts, including bargaining. Partially as a consequence of this, the literature on bargaining and conflict resolution has, for the most part, treated identity fairly superficially, as we noted in the introduction. To rectify this, we develop a theory of the use of identity in conflict resolution in two parts. In this section, we describe the effect of static identities on bargaining outcomes; in the fourth section, we make actors’ identities endogenous. Splitting the development of the theory in this manner allows us to make clear first the incentives for identity manipulation, and second the manner by which these incentives might be obtained by a skilled mediator. The current section also serves as a self-contained theory of the role of identity in bargaining, though this is not the main focus of the paper.

Our theory draws on several lines of research, all of which utilize a fluid conceptualization of identity. A broad and growing literature across political science, one stemming indirectly from the psychological foundations of social identity theory (SIT), assumes that individuals may choose to manifest the aspect of their identity most beneficial to them. That is to say, not only can the salience of identity vary by context, as in SIT, but salience itself becomes a choice variable. A subset of this scholarship, deriving insights from game theory and behavioral game theory, formalizes the incentives leading to and benefits gained from identity choice. Like Dickson and Scheve (Reference Dickson and Scheve2006), Penn (Reference Penn2008), Shayo (Reference Shayo2010) and others, we detail the effect of identity in a rationalist framework, in our case focusing on the role of shared identity in altering bargaining outcomes.Footnote 4

We begin by defining what we mean formally by identity, choosing a conceptualization that relies on the extensive literature on identity salience (for reviews, see Monroe, Hankin, and Van Vechten Reference Monroe, Hankin and Van Vechten2000; Kurzban Reference Kurzban2003; Michalski and Helmig Reference Michalski and Helmig2008; Stryker Reference Stryker2008). Our theory shares with this literature the assumptions that specific identities can be made salient at certain times and under certain circumstances, and that such contextually activated identities can have predictable consequences for behavior. Specifically, we assume that there are numerous identity aspects or dimensions (hereafter called ‘IDs’) along which actors may identify themselves and that actors may choose which IDs are most salient to them at any given time, though constrained by their own self-concepts. For example, an individual might view it as beneficial to emphasize linguistic similarities in one instance and religious similarities in another (see Posner Reference Posner2005). The idea that actors can vary the salience of multiple aspects of their identity is not novel (in fact, it is a central tenet of identity theory and SIT; see Stets and Burke (Reference Stets and Burke2000) for a review of the key concepts in both literatures, as well as linkages between them). However, we depart from (and contribute to) this literature by allowing for more fluid specifications of identity while addressing the complications that arise from deliberately chosen identity manifestations.

Formally, we assume that each actor's identity is defined by a vector of weights between zero and one assigned to each element of the set of all possible IDs.Footnote 5 IDs with non-zero weights are associated with aspects of an actor's identity; IDs with weights of zero, which will be the overwhelming majority if the set of all possible IDs is large, are those completely absent from one's self-concept. In this framework, which is similar to that of Shayo (2010),Footnote 6 we do not categorize types of identity (e.g. ethnic or linguistic), unlike many models of identity choice (see, e.g. Posner Reference Posner2005). Instead, each potential aspect of identity (such as Spanish speaker or practitioner of Sunni Islam) is its own dimension. The weights on the aspects need not add to one, and may in fact be completely uncorrelated with each other. For example, assume that there are only four possible aspects of identity in the world: Sunni, Shi'ite, Iraqi, and merchant. In this simple example, each person's identity looks like this: (a, b, c, d), where a is the weight on the Sunni ID, b is the weight on the Shi'ite ID, c is the weight on the Iraqi ID, and d is the weight on the merchant ID. The first two, a and b, are likely to be strongly and negatively correlated regardless of the political context at that moment. The third, c, may or may not be positively correlated with one of a or b, depending on the distribution of each religion in the country. The fourth might be weakly or not at all correlated with either of the previous three.

Our formalization allows individuals to have any values of a, b, c, and d between 0 and 1. As previously stated, 0 and 1 define the extremes: the identity vector of a person who is solely a Sunni would be (1, 0, 0, 0), while the vector of a person who is wholly a Shi'ite would be (0, 1, 0, 0). Note that an actor can be extreme in multiple dimensions at once. Someone with the identity vector (1, 0, 1, 1) would identify maximally as a Sunni Iraqi merchant. These extremes typify common characterizations of SIT: individuals automatically (often unconsciously) divide themselves and others into ‘in-groups’ (comprising people with like identities) and ‘out-groups’ (whose members share identities that differ from, or even oppose, those of the in-group), and tend to favor the group with which they identify (Kramer and Brewer Reference Kramer and Brewer1984; De Cremer and Van Vugt Reference De Cremer and Van Vugt1999; Huddy Reference Huddy2004). While accurate in many circumstances, such characterizations betray an oversimplified view of identity expression, for even seemingly static racial identities may be affected by context (for historical discussions of the malleability of racial designations, see Smedley in Stokes et al. Reference Smedley2001; Nobles Reference Nobles2000, chap. 1 and 2). In our example, an immigrant to the United States from a formerly mixed part of Baghdad might have an identity vector that looks like (0.7, 0.2, 0.8, 0.5) if she, though Sunni, on occasion attended Shi'ite services with friends while growing up, was strongly nationalistic, and, while not a merchant herself, was the child of one.

Explicit in this example is the notion of identity change – although she remains the same person, our example Iraqi found her identity changing upon exposure to alternative religious practices. How does this change occur? We draw on the extensive literature on identity salience in postulating that identity changes endogenously via repeated social interactions. One might think of this in sociological terms – the more time one spends within a group, the more one identifies with that group (see, e.g. Aronson Reference Aronson2008) – or the process can be an affective one, as described by Smith (Reference Smith1999), in the sense that an actor's emotional well-being is enhanced by group membership. Alternatively, one could think of this as the natural outcome of repeated rational interactions: if there is a benefit to being a member of a certain group (in terms of improved transactions with members of that group), then frequent transactions with a specific group will induce a rational actor to strengthen her in-group affiliation. We discuss this further in the fourth section . For our purposes now, it suffices that the core of the theory is that interacting with others who manifest similar aspects of identity leads to an increase in the weights that one assigns to those identity aspects. On the basis of this logic, we would expect that if we visited our Iraqi immigrant again several years later, she would place a higher weight on the Iraqi ID if she had located in an enclave strongly devoted to the maintenance of Iraqi culture than if she had instead chosen to reside in a more ethnically mixed area. We would also expect this trend to increase the more time this living arrangement held.

Because of the importance of social interaction in the formation of identity, individuals cannot freely select a desired identity. A politician might suggest that he would be an effective advocate for a particular racial or ethnic minority, but he will have difficulty claiming his in-group status if his history does not support his alleged affiliation. Continuing our example, our Iraqi immigrant would have an easier time evoking a feeling of commonality with other Sunnis than a wholly secular politician who happened to be born into a Sunni family, despite both nominally having the same religious affiliation. If the politician were sufficiently clumsy in his appeals, he might even suffer a penalty for making the attempt, as those to whom he was professing similarity might come to view him not only as inauthentic, but also as a dissembler.

This leads naturally to the role that identity plays in influencing outcomes. Our discussion of contextually determined identity, helpful in capturing real-world nuance, presents problems here. While an in-group/out-group duality allows for straightforward payoff comparisons in most situations – payoffs to outcomes are higher when individuals interact with members of their own group than when they interact with those who are outside of their group – now the translation between identity and outcomes is less clearly defined. To overcome this challenge, we focus on the deliberate manifestation of an ID. By manifesting an ID, we merely mean expressing a particular aspect or dimension of one's identity at a given time and in a given context (see Tajfel (Reference Tajfel1982a) for a similar working definition). In the example of the previous paragraph, the politician was attempting to manifest the Sunni aspect of his identity.

Manifestation of an ID within this theory is a conscious choice, made after considering the situation. Consistent with rationalist theories of identity choice, we assume that identity manifestations can entail behavioral prescriptions and have social consequences that affect one's utility (Posner Reference Posner2005; Dickson and Scheve Reference Dickson and Scheve2006; Penn Reference Penn2008; Shayo Reference Shayo2010).Footnote 7 Successfully manifesting the same identity aspects as those of the others with whom one interacts can improve the outcome of one's interactions, while unsuccessfully doing so can make the outcome worse. What ‘improve’ and ‘worse’ mean depend on the context; we discuss their import in bargaining shortly. First, we must flesh out manifestation further. We assume that the attempt of manifestation is costly in time, effort, and the potential resources spent in attempting to ‘pass’ as that ID. Further, we assume that the probability of manifesting an identity aspect successfully is increasing in the weight on the ID, while the cost of manifestation is decreasing in the weight. Put differently, it is easier to convince others that you are of a certain group – to ‘pass’ as a member of that group – the more a part of that group you truly are. At one extreme, an actor with a weight of 1 on an ID would pay little to no cost to manifest it, and suffer almost no chance of failure; the ID is fundamental to the actor's self-concept, and this is clear to all with whom the actor comes into contact. At the other extreme, an actor with a weight of 0 on an ID would pay a large cost to manifest it, and would most likely fail because the ID is so far outside the actor's experience as to almost preclude successful passing. All IDs to which an actor has never been exposed fall into this latter extreme.

As an example, assume that our Sunni politician was trying to reach a compromise deal among a group of people who all strongly identify as Sunni. Since manifesting a Sunni identity aspect would improve each interaction between the politician and members of the Sunni group, doing so might also improve the chances of coming to a successful deal. Let us assume that the politician's weight on his Sunni ID is 0.3 before meeting with the group, and that the probability of a successful manifestation is exactly equal to the weight on that ID.Footnote 8 Then there is a 30% chance of manifesting successfully, which would net the politician the benefit from doing so, but a 70% chance of unsuccessfully manifesting the Sunni ID, which would garner the politician a penalty. In deciding whether or not to make the attempt at manifestation, the politician must determine whether the expected payoff from this lottery over outcomes exceeds the cost of manifestation. If it does, then the politician should make the attempt, even though it might fail. If it does not, he should rely upon other means to broker this deal.Footnote 9

The role of the ID weight is important in this example. Were the politician to have a weight of 1 on the Sunni ID, he could manifest it with relative ease if he so decided. After all, it would be near costless to do so, and the attempts would almost always be successful. In contrast, a weight of 0 would leave the politician without manifestation as a valid option, since it would most likely fail, and the attempt would be costly as well. The 0.3 value we chose for the example, between these two extremes of 0 (no weight on a given ID) and 1 (highest ID weight), introduces uncertainty. The attempt might succeed, but it might fail. We expect this to be the norm, and thus believe that manifesting an aspect of one's identity is a risky action in nearly all cases. Formally, even the simplest scenario in which manifesting an ID is an option produces an expected utility calculation: an actor chooses to manifest an ID if P(success) × Benefit + (1 – P(success)) × Penalty > Cost.

What about for the more complex scenario of bilateral bargaining? The standard rationalist (bilateral) bargaining model provides a useful guide. Leaving out the shadow of the future, there are three common relevant factors that affect the manner in which actors bargain under uncertainty. The first is their preferences over outcomes, as formally represented by their (expected) utility functions. Should an actor's preferences change to align more closely with the other actor's preferences, the actor's offers would change and a resolution would become more likely. The second factor is the actors’ beliefs about the opposing actors’ preferences, typically represented formally by having one actor uncertain of the type of the other actor. Should an actor's beliefs over the distribution of types of the second actor shift toward types that have preferences more closely aligned with the first actor, the first actor would be more likely to make an offer designed to produce a resolution to the conflict. This is important for two reasons. One, should the second actor's preferences actually be altered, in line with the first factor, this shift in beliefs would make an actual resolution more likely as well. Two, experimental evidence suggests that actors may form overly critical assessments of the other actors in a bargaining scenario (Dickson Reference Dickson2009); altering actors’ beliefs would directly counteract this effect. Finally, along these lines, the third factor is the quality of information revealed during the negotiation. Should the quality of information revealed during the bargaining process improve, information revelation is more likely to reduce uncertainty, diminishing the likelihood of (continued) conflict (Fearon Reference Fearon1995).

We postulate that the identities that the actors manifest during bargaining alter all three of the aforementioned factors (preferences over outcomes, beliefs about opponents’ preferences, and information quality), and we detail one specific manifestation outcome in the next section: the manner in which a signal of shared identity – that is, both actors successfully manifest a common ID – aids in conflict resolution, as suggested in the previous paragraph. Before doing so, however, it is important to unpack who the actors in our theory are meant to represent.

We have deliberately used the vague term ‘actors’ in this section as the theory presented thus far does not specify the nature of those engaging in the bargaining process. Although it is perhaps most natural to think of identity on the individual level, the logic in this section goes through equally well in the case of states. The large literature on state identity (e.g. Campbell Reference Campbell1992; Hopf Reference Hopf2002; Mattern Reference Mattern2005; Rieker Reference Rieker2006; Rousseau Reference Rousseau2006; Abdelal et al. Reference Abdelal, Hererra, Johnston and McDermott2009; Risse Reference Risse2010; see also Mitzen Reference Mitzen2006 for a similar aggregation argument) provides one way forward. Just as one may aggregate individual-level identity upward, so may we aggregate individual-level identity weights to the state level to produce an identity vector for the state.Footnote 10 Some care must be taken in this process, of course; if a particular ethnic group is prohibited from affecting state policy or state leadership going forward, presumably the identity weights it contributes to state identity have far less of an effect on the bargaining process than those associated with the dominant ethnic group. Further, one must assume that states behave as unitary actors during the bargaining process, although this is common in formal models of bargaining.

We must be more circumspect, however, in our definition of ‘actor’ when making identity weights endogenous in the fourth section. As the primary mechanism for altering identity weights we propose is social, it is far less clear how to aggregate up to the state level. While we do briefly suggest how this may be done at the end of that section, we choose instead to focus on individual-level actors in the fourth and fifth sections. In addition to eliding the practical problems with aggregation, we believe that this is a more appropriate assumption in modeling mediation. Fundamentally, mediation is an elite-driven process in which one or more third-party mediators interact repeatedly on a personal level with appointed representatives of the bargaining parties. Treating the actors as individuals better captures this dynamic.Footnote 11

Shared identity and conflict resolution

We now apply our theory of identity in bargaining to the specific case of shared identity in conflict resolution. International negotiations are not simple, and they often involve bargaining over several dimensions at once. Uncertainty is prevalent in perceptions of the preferences and the capabilities of both sides, external pressure to resolve a conflict may be lacking, and trust may be in short supply between disputants. Strong identity cleavages, of the type likely to arise due to in-group/out-group dynamics during conflict or the strategic behavior of actors looking for stronger bargaining positions, can exacerbate these problems. However, the converse of this is also true: shared identity aspects can mitigate the same problems, improving the likelihood of successful resolution. In this section, we detail several causal pathways by which this may happen. We keep the discussion general, tying it to mediation and the associated literature in the fifth section.

As discussed in the previous section, the more salient an identity aspect is, the more an actor is able to manifest it successfully. If it were also true that an actor's successful manifestation of an ID was associated with superior outcomes when this ID was salient for the other actor, then rational actors, like those posited in the typical bargaining model, would frequently choose to manifest salient, shared IDs. After all, doing so would, with a high probability, garner both actors better outcomes.Footnote 12 Thus, in this case, negotiations in the presence of a shared ID that is salient to both sides would be accompanied by frequent manifestations of this ID, assuming both actors were aware of this commonality. This in turn would lead to an increased chance of a successful conflict resolution, as we discussed briefly at the end of the previous section. We go into further detail on this point below. To achieve this beneficial outcome, however, we first must justify the missing assumption, and specify what lies behind the association of shared ID manifestation with superior outcomes.

We propose three fundamental, closely related causal mechanisms inspired by the social psychological literature and directly connected to the precise mechanisms outlined in the last section with respect to the standard bargaining model; these mechanisms in turn imply others more closely tied to the political science literature on negotiations. The first mechanism is an increase in trust between the negotiating parties. Manifestations of shared identity can signal cultural sameness, which leads not only to an association of the other party with oneself, but also to an increased belief that the manifesting party shares one's own positive characteristics, such as integrity or character (see Otten and Moskowitz Reference Otten and Moskowitz2000; Otten and Wentura Reference Otten and Wentura2001). Those that share such positive characteristics are believed to be more likely to enter into negotiations in good faith and less likely to renege when it comes time to implement an agreement.Footnote 13 The second is an increase in an actor's belief of commonality. Expressions of common identity suggest common purpose and shared intent, and the notion of ‘common fate’ is widely documented in studies of racial and ethnic group politics (Dawson Reference Dawson1994, Reference Dawson2001; Stokes-Brown Reference Stokes-Brown2003, Reference Stokes-Brown2006; Sanchez Reference Sanchez2006, Reference Sanchez2008; Junn Reference Junn2008; for a review, see McClain et al. Reference McClain, Johnson Carew, Walton and Watts2009). Finally, the third is an improvement in communication between the actors. Research on SIT (particularly, Turner Reference Turner1985) mentions information quality as a mechanism for group identification: people tend to have more information about members of their in-group than they do about folks in the out-group, and this explains in-group bias. Manifestation by nature is performed in the common cultural idiom – or even the same language – of the ID, an idiom well understood by actors with high salience on that identity aspect (see Estes Reference Estes1994; Pavlenko and Blackledge Reference Pavlenko and Blackledge2004). It stands to reason that communications accompanying manifestation are thus more likely to be understood, particularly because shared identity facilitates the transmittal of information across actors (see Wilder Reference Wilder1981; Park and Rothbart Reference Park and Rothbart1982; van Bavel, Parker and Cunningham Reference van Bavel, Packer and Cunningham2008).

The combination of improved trust, commonality, and communication has numerous secondary effects depending on the context. Although it is not necessary, it is helpful to view the role of shared identity in conflict negotiations through the lens of bargaining theory. In this context, improvements in trust, commonality, and communication affect three distinct aspects of actors’ bargaining as discussed in the previous section: their preferences over outcomes, their beliefs about the opposing actors’ preferences, and the quality of information revealed during the negotiation.

Although this point is general, it aids in understanding to explore it in the context of a specific empirical example. Consider the century-old dispute between Argentina and Chile over the Beagle Channel (a strait separating the islands of the Tierra del Fuego). Previous attempts (in 1908, 1966, and 1977) by the British Crown to resolve this conflict were ineffective; in the case of 1978, England's involvement actually worsened the conflict between these South American countries (see Laudy Reference Laudy2000; Corbacho Reference Corbacho2008). In response to these failed settlement attempts, particularly the 180-day negotiations of the previous year, Pope John Paul II sent a personal message to the (then) president of Buenos Aires (Raúl Alfonsín) and General of Santiago (Augusto Pinochet) urging a peaceful solution. Both countries allowed the Roman Pontiff to mediate the dispute through the good offices of Cardinal Antonio Samoré, his special emissary. The resulting 1982 resolution (known as the Treaty of Peace and Friendship) granted the islands to Chile while giving many of the maritime rights to Argentina (see Parish Reference Parish2006 for details). Why might this mediation have succeeded where previous attempts at settlement failed?

We start by considering the actors’ own preferences.Footnote 14 As noted in the previous section, a negotiation becomes more likely to yield a mutually beneficial settlement the closer are the actors’ preferences. Thus, any factor that shifts the preferences of the actors toward each other can help enable conflict resolution. Certainly, such shifts can occur for reasons unrelated to identity. For example, the Vatican may have succeeded in part because of the duration of the conflict: both Chile and Argentina may have simply wanted the dispute over the Beagle Channel to end sufficiently greatly so as to overcome opposing interests. In other words, past failures may have nurtured Argentina's and Chile's desire for resolution for its own sake.

Increases in both trust and perceptions of commonality arising from manifestations of a shared ID can also affect an actor's preferences, in one of two ways. First, recent research on ‘group-based’ self-regulation shows that a more positive view of others can decrease one's desire to have members of the out-group do less well (see Abrams Reference Abrams1994; Otten, Mummendey and Blanz Reference Otten, Mummendey and Blanz1996; Sassenberg, Kessler and Mummendey Reference Sassenberg, Kessler and Mummendey2003). If demonization of the ‘other’ results in an increased utility benefit for doing well at the expense of one's opponent, then increased trust reduces the size of this utility benefit, so that it has less effect on the equilibrium bargain, widening the bargaining zone. Second, the increase in trust may be felt as well by other aligned actors, who may then increase external pressure to come to a resolution. If one derives utility from the support of other aligned actors, then this too will increase the probability of a successful resolution to the conflict.

While the exact terms of the Beagle Channel resolution are not publicly available, it seems clear that Pope John Paul II appealed to the Christian sensibilities of the claimants, using an identity-based approach. Catholicism is the national religion of both nations, and this shared religious identity, if made salient in the context of the negotiations, could have played an important role in resolving the Beagle Channel conflict (see Laudy Reference Laudy2000, 293; for a more general discussion of the usefulness of faith-based mediators in conflict resolution, see Bercovitch and Kadayifci-Orellana Reference Bercovitch and Kadayifci-Orellana2009), in part by enhancing trust. This approach highlights commonality and spurs external pressure. In this case, the Pope's envoy, Samoré, mobilized domestic support for the peace plan, which in turn pressured Argentinean and Chilean military leaders to compromise (Garrett Reference Garrett1985; Parish Reference Parish2006; Corbacho Reference Corbacho2008). Furthermore, the emissary's involvement allowed the representatives of these nations to ‘save face,’ for they found it easier to make concessions proposed by a neutral arbitrator than to signal weakness by giving in to demands made by the other side bilaterally (see Laudy Reference Laudy2000, 293; Bercovitch and Kadayifci-Orellana Reference Bercovitch and Kadayifci-Orellana2009). External pressure, like increased trust, can change each party's utility from a settlement. Note that disputants may deliberately seek mediation to spur external pressure, as there is evidence that leaders may choose mediation when seeking political cover for concessions (Beardsley Reference Beardsley2010).

Next, consider an actor's perceptions of the other party's preferences. A shift, for example, in Chile's perceptions toward viewing Argentina's preferences as being more in line with its own can be an accurate representation of true preferences, or a misperception brought about due to bias. Both cases make a resolution more likely, but different causal mechanisms underlie them. Misperception due to bias arises from the same sources as a shift in one's own preferences: improved trust and commonality. As stated above, common identity signals a similarity in positive characteristics. Because individuals expect members of their in-group to share their beliefs and values (Robbins and Krueger Reference Robbins and Krueger2005), it is natural – even if not necessarily rational – to extend this similarity to other dimensions. This can reduce the effect of the negative perception of the opposing side in bargaining (Dickson Reference Dickson2009), which makes resolution less likely. In contrast, a more accurate representation of true preferences arises from communication and commonality. Improved communication raises the signal-to-noise ratio of discussion, and commonality can indicate true similarity in preference and falsely perceived similarity. Combined, they enable an actor to discern more accurately the preferences of the other disputant, getting away from the biases introduced by viewing each other through a false self/other dichotomy and by the position-taking of the disputants. Among other benefits, this decreases the belief that one's opponent will renege on the agreement during implementation, helping to reduce the credible commitment problem. It is in this spirit that the Pope's emissary advises, as the following passage from the United Nations Report on the Beagle Conflict suggests, that the nations’ ‘ancient’ linguistic and denominational identities should trump Chile and Argentina's secular (arguably more superficial) grievances:

The Mediator even goes as far as to suggest that these countries will ‘forget’ their disagreements if they submit themselves to the Will of God (United Nations 1977, 246). This was an ‘identity play’ that Queen Elizabeth II (who was a protestant) would have been less able to execute, and, given the feuding nations’ overwhelmingly Catholic populaces, one can credit the Pope's success in mediating this conflict in part with his ability to shift the dialogue of the negotiations away from territorial disagreements while reminding the claimants of their common religious affiliation. Moreover, the above passage suggests that the Pope and his emissary made clever use of the tactic of ‘altercasting’ (see Pratkanis Reference Pratkanis2000): the mediation process added to the salience of the disputants’ religious identity, placing the representatives of Chile and Argentina into the roles of ‘good Catholics’, and, consequently, persuading them to behave in a manner consistent with these imposed roles (in essence, attempting to compel the disputants to settle their territorial differences peacefully).

Finally, consider the quality of information unrelated to preferences. This category includes the relative power of both disputants, the probability each party might renege upon the agreement and the chance of third-party enforcement of the original agreement, and so on. Improved communication decreases uncertainty in all these aspects of the negotiation, decreasing the chances of war (Fearon Reference Fearon1995). As noted above, shared identity facilitates improved communication in several ways. Should the shared ID be linguistic or be correlated strongly with a linguistic ID, so that it too is substantially shared, then communication absent translation would typically be more effective at conveying information. A common cultural idiom may also make misunderstandings less likely, as cultural references would be better understood. With respect to the Beagle Channel mediation, this common cultural framework existed within the rituals and requirements of Catholicism.

Thus, there are numerous pathways by which a shared identity aspect can increase the likelihood of a successful conflict negotiation. As we briefly noted, however, conflict tends to exacerbate existing identity cleavages or even engender new ones. Using the above example, before the conflict the salience of nationalist identities Argentinean and Chilean may have been low, with many individuals having non-zero salience in both simultaneously, but by the time negotiations occurred most actors’ saliences in nationhood might have been strong and dichotomous. In this case, identity would provide little benefit, and the reverse of some of the processes detailed here, spurred by manifestations of non-shared identities, would increase the belief that the parties had substantially different preferences, decrease their individual benefits for coming to a resolution, and lower the signal-to-noise ratio of the negotiations. As we explain in the fifth section, it is unlikely that parties in bilateral negotiations in this context would hit upon identity plays focusing on shared identity aspects themselves, even if compromise in bargaining were their primary goal. This is one way in which a skilled mediator can be useful. In the following two sections, we describe how identity weights are endogenous to the process of interpersonal interaction, and we discuss the potential usefulness of third-party mediation in the context of this broadened theory.

A theory of endogenous identity formation

Thus far, we have treated the weights on actors’ identity aspects, or IDs, as fixed. As discussed briefly in the second section, however, implicit in our definition of an actor's identity vector is the idea of identity change: at a basic level, these weights must come from an actor's social context. As this context varies over the course of an individual's life – and for the time being, we focus on individuals, narrowing our definition of actor – so too should the weights on a person's IDs. Social context dictates the array of individuals with whom one engages in repeated social interactions; these social interactions alter identity (McGuire et al. Reference McGuire, McGuire, Child and Fujioka1978; Weigert, Teitge and Teitge Reference Weigert, Teitge and Teitge1986; Grotevant Reference Grotevant1992; Caputi Reference Caputi1997; Elbedour, Bastien and Center Reference Elbedour, Bastien and Center1997; Côte and Levine Reference Côte and Levine2002). Our theory rests on a simple assumption, supported by several causal pathways: successful manifestations of an ID experienced during a potentially gainful interaction with another person can increase the weight one places on that ID. Gainful interactions here are any in which manifesting a shared ID has a benefit; we identified several benefits in the previous two sections for bargaining interactions intended to resolve conflict.

It is natural to assume that increasing the weight on an ID in response to the possibility of garnering a benefit is a rational action, and certainly this interpretation is consistent with our theory. In other words, purely rational responses to potential gains would induce actors to increase the weights on the IDs that are most likely to produce these gains, as we have assumed. After all, the static model of which this is an extension assumes that the chance of a successful manifestation is increasing in the weight, while the cost of manifestation is decreasing in it. Therefore, one is more likely to be able to signal a shared ID via its manifestation in response to another's manifestation the higher one's weight is on that ID. One need not assume rational responses, though, for our key assumption to hold. Simply assuming that one desires to be part of the in-group, as signaled by others’ manifestations of the in-group ID, is sufficient motivation for one to seek to increase one's weight on that ID, consistent with our theory. And there are many reasons why an individual might want to be part of the in-group, some of which may not be rational, or even rationalizable.Footnote 15

There are also many ways in which an individual may increase the weight on an ID. For example, each successful manifestation by another person might lead one to increase one's weight on that ID by a set amount every time, or by an amount dependent on the present value of the weight. The latter option might better model a scenario in which, once one's weight had moved close to 1 or 0, little further movement would occur. Practically, we expect that even those rationally attempting to adjust their ID weights, as opposed to merely responding unconsciously to the present context, would be limited in this regard by the malleability of their self-concepts. As an illustration, we return to the politician from our example in the second section. Extensive discussions with fellow Sunnis might result in unconscious adoption of behavioral prescriptions associated with Sunni Islam, increasing the importance of the Sunni ID in his self-concept, raising its weight in his identity vector. Or he could recognize the need for an increased weight on his Sunni ID and consciously mimic these same behavioral prescriptions, but fail to do so compellingly due to his inability to internalize the identity shift. His ID weight would shift only slowly, and he would commit frequent faux pas. The politician might, though, truly be able to ‘feel their pain,’ so to speak, internalize the identity shift, perhaps with abundant research and coaching designed to aid in this endeavor, and shift his ID weight rapidly. Whether this process is unconscious or conscious, fast or slow, however, our argument holds so long as the direction of the movement in ID weight is correct.

While we have presented the interaction within which individuals might choose to manifest an aspect of their identities as something separate, the interaction itself is really just one more piece of historical context of the sort that gave rise to that person's identity in the first place. Thus, the final component of our theory of endogenous identity formation specifies that manifesting an identity aspect can also alter one's identity weights. The politician might fail to convince other Sunnis of his sincerity, but the act of manifesting might increase his weight on his Sunni religion. Passing successfully requires convincing others of the validity of your claim, which often necessitates study of and practice in the behavioral prescriptions associated with the ID. The expenditure in time and potentially resources this necessitates may make it easier for you to manifest that identity in the future, increasing the weight on that ID.

Central to the theory's functioning is the idea of feedback. High-weight IDs are less costly to manifest and more likely to succeed – this leads to more frequent manifestation, and more successful manifestation. More frequent and successful manifestations imply yet higher weights on that ID, and induce higher weights in others on that ID, leading to even more frequent and successful manifestations. This sort of positive feedback process is often associated with path-dependent dynamics (Page Reference Page2006): the greater the attention to the behavioral prescriptions associated with a particular ID, the more beneficial is additional attention in the future.Footnote 16 One way of looking at the role of the mediator in an identity play, as discussed in the next section, is to initiate and encourage this cycle of positive feedback, allowing the disputants to take advantage of the benefits of shared identity described in the previous section.

Before moving on, however, it is once more helpful to examine the nature of the actor in our theory. We have discussed identity formation and change as a consequence of repeated interactions, and have therefore focused on the level of the individual, as individuals are the ones participating in these interactions. Simple aggregation to the state level of such interactions strikes us as particularly unwarranted, since not all individuals within the state will have experienced the same interactions. Because of this, we view the application of our theory to the case of mediation, described in the next section, as an elite-level theory: the representatives of the disputants and the mediator engage in repeated interactions in the context of coming to an agreement, which is then implemented in some fashion by the disputing states.Footnote 17 This does not mean, however, that the aggregate identities of states must remain static as those of their representatives change. Interactions with opinion leaders and perhaps the media as well can also lead to identity change by the same process described here. Representatives can, via repeated interactions, influence the identities of state leadership; further, access to, and sometimes control of, the media and opinion leaders provide state leadership with an opportunity to influence the population as a whole. Thus, over time, the beneficial effects of identity change at the elite level may be translated to the larger population, beyond just a resolution of the conflict.

Identity and mediation

Mediation can take numerous forms, and a shared identity aspect can enhance several of them. To make the connection between the benefits of shared ID and mediation techniques, we use the same categorization of techniques as Beardsley et al. (Reference Beardsley, Quinn, Biswas and Wilkenfeld2006; see also Touval and Zartman Reference Touval and Zartman1985; Princen Reference Princen1992; Bercovitch and Houston Reference Bercovitch and Houston1996; Wilkenfeld et al. Reference Wilkenfeld, Young, Quinn and Asal2005). This categorization separates mediation techniques into three useful groups: facilitation, consisting of information revelation and trust building; formulation, consisting of the active proposal of and coordination on bargaining outcomes and procedural tactics designed to focus the negotiation (Bercovitch and Houston Reference Bercovitch and Houston2000); and manipulation, consisting of pressure, rewards, and punishments. In this categorization, negotiating in the context of a shared ID directly enhances the efficacy of facilitation and formulation, and can indirectly aid with manipulation. The enhancement of each operates under the causal mechanisms described in the previous sections, on two levels. A shared ID between disputants and a shared ID between disputants and the mediator(s) both improve the probability of successful conflict resolution.

Effective information revelation in mediated negotiations is hampered by the credibility of both the other party and the mediator. Without some reason for the disputants to believe that the information provided is truthful, the words of all other actors are no more than cheap talk. In the Beagle Channel case, the Pope's stature among Catholics as a ‘moral authority’ garnered his emissary credibility. Noting this, scholars have directly addressed the role of credibility and bias in mediation, concluding that some degree of bias of the mediator toward one party can be beneficial, although bias toward peace on its own is not (e.g. Kydd Reference Kydd2003, Reference Kydd2006; Rauchhaus Reference Rauchhaus2006; Savun Reference Savun2008). Bias toward a party provides more reason for the favored party to consider the information provided by the biased mediator, for preferences of the two will be aligned; likewise, the non-favored party is likely to accept the terms of a resolution if that party believes that a biased mediator is arbitrating evenhandedly despite its bias (Carnevale and Choi Reference Carnevale and Choi2000).

As detailed in the third section, a shared ID promotes information revelation in several ways. First, it increases the credibility of all actors via increased trust in the other side's positive intentions (Karim and Pegnetter Reference Karim and Pegnetter1983; Kelman Reference Kelman1997; Kriesberg Reference Kriesberg2001; Kydd Reference Kydd2006).Footnote 18 Of course, increased trust itself is also an element of facilitation. Second, shared ID induces the perception of shared preferences with the other party or parties. Third, it can actually shift the preferences of the actors closer together. Both of these act as does explicit bias, providing disputants more reason to consider information. Note that, in the former case, a mediator using a shared identity appeal can effectively be ‘biased’ toward both sides at once, as perceptions of shared preferences need not always match reality.Footnote 19 Fourth and finally, the improved communication occasioned by a shared ID facilitates information flow directly. All these benefits accrue to information revelation both between mediator and disputant, and between disputants themselves, if the latter share a salient identity aspect.

Formulation strategies are also enhanced in the presence of a shared identity. One way in which formulation works is for the mediator to discover and help both disputants focus on a favorable settlement that they are not themselves seeing. A shared identity aspect aids in this endeavor in three ways: (i) it can shift the preferences of the disputants; (ii) it can shift each party's perception of the other party's preferences; and (iii) it enhances trust in the mediator and the perception that the mediator's interests are aligned with each disputant sharing its ID. The first two expand the bargaining zone, increasing the number of possible settlements acceptable to both sides. The third way better enables the mediator to focus on the particular settlement it desires, as the mediator is better able to convince the parties that it has their best interests at heart.

Manipulation strategies rely on forms of pressure, rewards, or punishments that effectively change each party's utility from a settlement. A shared ID can have the same effect, albeit indirectly, through the pressure mechanism we identified in the previous section. If the mediator is able to induce the same increase in commonality and trust in third parties upon whom the disputants rely for support – for example, a democratic populace – then the pressure that these external parties will apply to the disputants to settle will change the utility the disputants gain from a settlement just as other forms of manipulation would. In essence, the mediator can use shared ID to induce constituents to engage in manipulation. Further, third-party mediators also have constituencies; if they possess military power as well, then disputants might be more likely to believe that the third party would enforce any agreement, potentially improving the likelihood of peaceful settlement (Favretto Reference Favretto2009).

Thus, salient shared ID aspects, both between the disputants and between the mediator and the disputants, can aid in successful conflict resolution, in a manner complementary to those previously identified in the mediation literature. Given this, the practical question is: How can we maximize the salience of any shared identity aspects, particularly in the context of conflict? In the remainder of this section, we detail how to choose a mediator to take advantage of shared ID, and what strategies should be used by the appropriate mediator.

Concluding remarks

This essay breaks new ground in several respects. By integrating the rational-choice framework of conflict bargaining with psychological models of social identity, we synthesize perspectives from two seminal literatures – literatures that seldom speak to one another – to develop a novel theory of identity formation. Beyond simply offering the old maxim that ‘identity matters’, this paper develops a systematic argument to explain why it matters and, more importantly, how identity manifestation can be utilized in bargaining.

The applicability of our theory to the study of conflict negotiation is our second major contribution. Discussion of the role of identity in conflict typically focuses on its detrimental aspects. Cleavages between claimants’ identities make it easy for each to view the other in an ‘us vs. them’ context that renders bargaining more difficult. However, as we have argued, shared aspects of identity can also be used to help resolve conflict. When claimants share some aspect of their identities, even when this aspect is not initially salient, a third-party mediator can help to enhance its salience and reduce the barriers to successful settlements. The theory we present traces the causal chain behind this claim and details how the salience of identity aspects endogenously changes with interpersonal and intergroup interactions. By enhancing trust, commonality, and communication between actors, shared identity moves disputants’ preferences and their perceptions of the other disputant's preferences closer together, and allows both disputants to make better use of the information revealed during the negotiation. This widens the bargaining zone and makes successful conflict resolution more likely.

Third parties making concerted ‘identity plays’ can exploit this palliative role of shared identity in order to improve bargaining outcomes. Our theory thus predicts a complementary role for third-party mediation in international conflict to that extant in the literature. Consistent with the literature on bias in mediation (e.g. Kydd Reference Kydd2003, Reference Kydd2006; Rauchhaus Reference Rauchhaus2006; Savun Reference Savun2008), using shared identity increases the utility of information, reducing bargaining uncertainty. With shared identity, however, not only do both disputants get more use out of information, but more potential settlements are also possible due to a shifting of preferences. Consequently, we foresee a powerful role for identity plays in third-party mediation, and believe their use should be encouraged. To that end, we have offered some suggestions for choosing the mediator and the mediator's strategy. We also believe the potential importance of this mechanism suggests an increased attention to the dynamics of mediation, as data on tactics used over time would be necessary for any true empirical test of the theory we present.

In addition to the future research that it can potentially generate, this paper informs long-standing conversations about the use of identity as a form of socio-political power. Clearly, identity bargaining can be viewed as a ‘power play’ in the sense that clever mediator choice and negotiation tactics can influence disputants’ preferences and thus ultimately their decisions (Lukes [1974] Reference Lukes2005). When considered from a rationalist perspective, appeals to common identity can help third-party mediators bring about favorable negotiation outcomes. Perhaps more importantly, such appeals can also change the incentive structures of disputants, which facilitates conflict resolution (Dowding Reference Dowding1996). Overall, the ideas expressed in this article speak to the relational nature of power, for we argue that the process of identity bargaining involves not only choosing an appropriate identity in a given context, but also manipulating the identities of those in conflict.