Early Warning and Early Response, by Susanne Schmeidl and Howard Adelman (eds.)

 

16. Deterrence Strategies and Conflict Prevention:
States versus Institutions

David Carment
NPSIA, Carleton University

Frank Harvey
Dalhousie University

There is no definitive evidence on when and under what conditions different types of third parties should intervene in a coercive way to prevent different types of crises from escalating out of control, or how to manage these disputes when they do. The literature is unclear about the conditions under which deterrent and compellent threats (or any form of coercive diplomacy) succeed and fail across a variety of contexts. Without an empirical base to evaluate the conditions under which coercive diplomacy is likely to succeed and fail, answers to pressing questions about the onset, escalation and resolution of ethnic violence, the most common form of violence today, will remain elusive.

The paper investigates the question of whether and how perceptions of commitment, credibility and resolve are influenced by (a) the type of intervener (states versus international organizations), (b) the type of conflict (ethnic versus non-ethnic), and (c) the location of the dispute (interstate versus intrastate). Several propositions are tested against crisis data. The findings indicate, among other things, that (a) ethnic and intrastate crises are more violent than their counterparts, and (b) multi-state coalitions are more likely than unilateral interveners, and unilateral interveners more likely than international organizations (IOs) to control hostilities in both ethnic and intrastate crises. The implications are discussed in the context of growing opposition to unilateral state intervention into ethnic crises.

 

1. Introduction

Ethnic conflict is the most common form of armed strife in this decade. As Michael Ignatieff argued in 1993, “huge sections of the world’s population have won the right of self determination on the cruellest possible terms: they have been simply left to fend for themselves. Not surprisingly, their nation-states are collapsing” (1993: 8). Robert Kaplan’s highly influential article on the “The Coming Anarchy” offered a more bleak assessment of state failure in Africa. In this neo-Malthusian perspective—the world—especially the South—is beset by increasing conflict and crises generated by fast-growing populations, demographic changes and weakening state capacity to regulate conflict.

Between 1945 and 1980, there was a gradual increase in ethnic rebellion and non-violent protest among ethnic groups. As new, institutionally weak and divided states emerged on the geopolitical map in the 1960s the upward trend in violent ethnic conflict began (Gurr 1994). The trend is most closely associated with decolonization and contention for state power in Africa and Asia. However, the collapse of the Soviet Union and other communist states has since provided the basis for ethnic rebellions in these transitional states. As a consequence, overall levels of ethnic violence in the international system reached their highest levels in the 1990s. ₁

Ironically, these tensions are arising at a time when barriers to third party intervention by major powers and multi-state coalitions appear to be falling. It remains unclear, however, whether this new environment improves the capacity of third parties to manage ethnic disputes, or whether it exacerbates ethnic divisions and violence. The outlook is not promising, to say the least. Evidence from the most recent cases of protracted ethnic violence in the former Yugoslavia, Somalia and Rwanda suggests that international organizations and multi-state coalitions are not well suited to handle these newly emergent threats. Even a brief, four-point review of contemporary global politics confirms this suspicion. First, technological change sustains the trend toward more rapid escalation of crises. Decision makers are pressured to make critical decisions in progressively shorter time intervals and that is especially problematic when underlying conflicts concern explosive issues connected to ethnic identity.

Second, international institutions and individual states experience a great deal of pressure to manage and even resolve ethnic conflicts. This perceived responsibility for action to preserve human rights and political stability is unlikely to diminish. As Haas (1983) and Diehl (1996) have shown, in areas where state interests converge, states have generally provided strong support for the application of institutionalised forms of conflict management. In the absence of cooperation among states, institutions, especially the United Nations, have proved far less effective (Roberts 1996, Betts 1996). The most notable aspect of Haas’ analysis is UN ineffectiveness when “metaissues” around which state interests coalesce (eg. post-war decolonisation and self-determination in Asia and Africa) are absent.

Third, a possible long-term decline in the utility of unilateral interventions towards cost effective multilateral preventive strategies creates the need to coordinate military actions with diplomatic communications, information and analysis. This kind of cooperation is made difficult by a limited ability to demonstrate clear resolve and identify limited objectives, especially when the number of actors involved expands beyond a very few. In reality, there is little agreement within the discipline on the kinds of strategies necessary for the management of armed conflicts. Critics have suggested that coercive and unilateral forms of intervention serve only to erode international norms of mutual restraint among states (Arend and Clark 1993; Roberts 1996). Others, more provocatively, have argued that the quickest way to terminate an internal conflict with fewest casualties is to favour the stronger side in any conflict (more often than not the state-centre) (Betts 1996, Regan 1996, Licklider 1995).

Fourth, and finally, the size and diversity of the modern world creates problems for civilian control over military forces stationed abroad. These complications may even outpace improvements in communications and transportation, which makes multilateral militarized options a more risky and potentially unattractive venture with the passing of time.

It is clear that the UN, the world’s leading international organization, is not equal to the challenges created by the preceding four factors. In their assessment of the failure to respond to Rwanda’s 1994 genocide, Adelman and Suhrke argue that one of the greatest problems in UN intervention is policy incoherence (Adelman and Suhrke 1996). Adelman suggests that within multilateral situations “...assignation of roles encourages waffling because a decision must be made and responsibilities accepted” (Adelman 1996: 34). The commensurate bureaucratic pulling and hauling within and between institutions stifles immediate response and leads to ambiguous half-measures.

Today, the pace of events is faster than ever before, pressure exists for help in multiple regions, coordination of diplomatic and military strategies is crucial and ethnic conflict occurs in a larger and more diverse international system. The demands on the UN’s abilities to manage and resolve ethnic conflict seem overwhelming when juxtaposed with the difficulties that the organization currently faces (Neack 1997) Debates occur within the ranks of the UN over division of labour and burden sharing. The UN is still recovering financially from a recent phase of activity and must deal with the problem of unpaid dues from its members. Perhaps most important of all is the intense conflict with the US, by far the most important member of the organization, over a wide range of important issue. ₂

With these issues in mind, we address several important questions in this paper: What are the conditions under which deterrent and compellent threats succeed and fail in violent conflict settings? More specifically, how is credibility, commitment and resolve influenced by the type of intervener, the type of conflict, or the location of the dispute? Are some third-party interveners more capable than others of deterring ethnic challenges? Are some third-party interveners more capable than others of controlling violence if and when deterrence fails?

Deterrence is crucial in the management of today’s conflicts, but rarely is there a collective political will to act prior to the outbreak of violence. On average, states are better equipped to respond to and prevent certain conflicts because they can overcome some problems related to political will. However, certain kinds of conflicts, specifically ethnic ones, make satisfying the prerequisites for successful deterrence difficult. When leaders are convinced that they cannot satisfy these prerequisites they will simply ignore early warnings of impending violence. Analysis of the conditions under which deterrence succeeds in different contexts is becoming crucial at a time when theory and policy on early warning are being criticised for their lack of cumulation and their inability to overcome analytical gaps. Without an empirical base to evaluate the conditions under which coercive diplomatic strategies are likely to succeed in different settings, answers to pressing questions about the onset, escalation and prevention of the most common form of violence today will remain elusive. ₃

Apart from this introduction, the paper unfolds in six sections. In the next section, we develop a model of deterrence. In the third section we discuss key assumptions underlying the model and specify the propositions to be tested. The fourth and fifth sections of the paper explain our methodology and test the propositions against the evidence. The conclusion summarizes the findings and identifies areas for further research in the context of early warning.

 

2. A Model of Deterrence

Using the term “preventive diplomacy” , Lund captures the underlying theme of this study:

Preventive diplomacy, or conflict prevention, consists of governmental or non-governmental actions, policies and institutions that are taken deliberately to keep particular states or organized groups within them from threatening or using organized violence, armed force, or related forms of coercion, such as repression, as the means to settle interstate or national political disputes, especially in situations where the existing means cannot peacefully manage the destabilizing effects of economic, social, political and international change (1996b: 379). In its most robust form, deterrence (a particular form of preventive diplomacy) includes a full range of political, diplomatic and military instruments (Kaufman 1996; Lund 1996a). Among the primary goals of this type of coercive diplomacy is to act prior to the outbreak of armed violence to reduce conflict between groups and to prevent regional adventurism.

John Ruggie (1993) developed a theory of deterrence to clarify the strategic dimensions of third party military intervention. Addressing the problem within the context of UN activity in Bosnia, Ruggie argued that there is a need to fill the ‘doctrinal void’ between peacekeeping and peace-enforcement. Missions which involve goals clearly beyond those of traditional peacekeeping, such as seeking to neutralize local forces and to push belligerent parties towards the negotiating table, require different strategies. Ruggie proposed that international forces be given the means and mandate to “deter, dissuade and deny” the use of force by local protagonists (Ruggie 1993). If deterrence of violence fails, Ruggie argues, deployed forces can attempt to dissuade parties from continuing military activities. Failure on this level necessitates the use of force to deny any one side military victory in a conflict.

Building on work by Ruggie, Carment and Rowlands (1996:11) observe that by presenting a credible military threat, “a third party force seeks to convince all conflictual parties that violence will not succeed. International force is brought to bear to neutralize the local forces and to reduce the expected gains of continued fighting. The political objective is to prevent local force from becoming the successful arbiter of disputes and to persuade combatants that they have no viable alternative but to reach a third party assisted negotiated settlement.” The intended effect is to deny victory to any one group in order to create the military stalemate on which negotiated settlements often depend. To ensure credible effectiveness, coalition forces must not only decide whether to escalate their intervention, but also must consider the degree of coercion to apply.

Using the same theoretical insights as Ruggie, Harvey (1997) examines third party intervention in the former Yugoslavia through the prism of deterrence theory. Harvey argues that, in almost every major encounter with the Bosnian Serbs between April 1993 and April 1994, US and European (NATO) officials failed to satisfy even the most basic strategic requirements of deterrence. Instead, external powers either diluted or intentionally qualified most of their retaliatory threats. As failures mounted, Bosnian officials simply ignored subsequent efforts to control hostilities. In contrast, satisfaction of all four prerequisites for effective deterrence ultimately re-established credibility and enabled the US and NATO to control fighting and obtain cooperation on key demands. Eventually, NATO/UNPROFOR did create the incentive for the Serbs to sign the Dayton Peace Accord by making it costly for the Serbs to persist in rejecting demands. Consistent with Ruggie and Harvey’s assessments, NATO eventually escalated its involvement in 1995 through a week-long series of selective air strikes against Serb positions after Serb forces laid siege to Sarajevo. The attacks were intended to force the Serbs to remove their artillery aimed at UN designated safe areas.

Whether the lead role is played by an institution or state actor, deterrence theory stipulates that a retaliatory threat will succeed if leaders (A) define the unacceptable behaviour, (B) communicate to challengers a commitment to punish violations in a way that denies them the objectives sought, (C) possess the means (capability) to do so, and (D) demonstrate the resolve to carry through with the threat (Lebow 1981: 65–69).

Resolve is most effectively demonstrated through costly signals—i.e., any action, statement or condition that increases the political, economic or military costs associated with the status quo, while lowering the costs of responding to a challenger’s probes. Resolve is enhanced in one or more of the following ways: D1) actions (e.g., deployment of air, sea or ground forces; evacuation of peacekeepers from safe-havens, thus allowing a more decisive air-strike response to probes); D2) statements (e.g., public announcements (promises) of impending retaliation; using explicit ultimatums and deadlines; public displays of unity among coalition members in support of response); D3) domestic support for retaliation (e.g., public opinion; positive domestic and international press coverage). If these conditions (A–D1–3) are satisfied, the expected net costs of the threatened sanction (to the challenger) should be greater than the expected net gain of non-compliance, because the punishment (if carried out) would prevent the challenger from achieving intended goals. If these four conditions are met, but the behaviour still occurs, that would constitute a case of failure—both in theory and strategy. On the other hand, if one or more of the conditions are not satisfied, the theory predicts failure in most cases. In other words, even clear and credible threats will fail if the challenger believes the challenge is worth the risks and political, military or economic costs incurred by triggering the threatened response.

Our model builds on Regan’ s (1996) study of utility estimates of challengers faced with a threat of third party intervention. According to Regan, when contemplating a challenge, a typical challenger’ s preference ordering is:

U_s >  U_bd > U_i.

Where:
Ubd	= utility (payoff) to challenger for backing down in the face of 3rd party threat. 4
z	= probability that the third party intervener will intervene if challenge takes place.
Ui	= utility (payoff) to challenger if intervention takes place. 5
Us	= utility for a successful challenge (with no intervention)

If U_bd > z(U_i) + (1–z)(U_s) then the challenger is expected to back down and deterrence succeeds. Based on this preference order, Regan (1996: 14) argues that the critical risk for the challenger is z > U_s–U_bd/U_s–U_i —“...when the right hand side of the inequality is greater than z, the challenger will not be deterred by the threat (that is, violence will escalate); when the right hand side of the inequality is smaller than z, the challenger will succumb to the threat (deterrence succeeds). By implication, the intervener can manipulate the challenger’ s critical risk on two dimensions: by increasing the utility to the challenger of backing down in the face of a threat (U_bd), thus decreasing the value of the numerator in the equation; or by decreasing the utility to the challenger of the intervention (U_i), thus increasing the value of the denominator.” ₆

We make two interrelated arguments in this context. First, the challenger’ s utility estimates are directly related to the type of third party intervener (state versus international organization) and the type of conflict (intrastate versus interstate, ethnic versus non-ethnic)). Both of these factors have an impact on the challenger’ s utility estimates (and critical risk), because the prerequisites for successful deterrence are more or less present depending on the type of intervener and dispute in question. Second, in line with conventional wisdom, we argue that crises occur through phases (Kriesberg 1997, Dixon 1996, Lund 1996b). While Regan’ s model is very useful for understanding a challenger’ s decision calculus during the onset phase (that is, before intervention takes place), a slightly different set of utility (and critical risk) estimates are likely to be made by the challenger once an intervention takes place, during the peak or crisis phase of the conflict. The revised preference order for the challenger, then, is:

U_si >  U_s > U_bd >  U_i > U_bdi.

Where:
Ubdi	= utility to challenger for backing down in the face of 3rd party intervention
Us	= utility to challenger of a successful challenge without intervention
Ui	= utility to challenger of not backing down following third-party intervention
psi	= probability that 3rd party will succeed in preventing escalation
Usi	= utility to challenger of successfully challenging 3rd party
Uui	= utility to challenger of unsuccessful challenge

The challenger will back down during the peak or crisis phase when U_bdi > p_si(U_ui) + (1–p_si)(U_si). ₇ The revised critical risk for the challenger following intervention (based on above preference order) is z > (U_si–U_bdi) / (U_si–U_ui). Once again, there are two ways to improve the chances of deterrence success at this stage of the crisis: (a) by increasing the numerator (that is, the utility of backing down after intervention), and/or (b) by decreasing the denominator (that is, decrease the utility of an unsuccessful challenge). ₈ As deterrence stipulates, both of these can be accomplished if the third-party intervener satisfies the four key prerequisites for successful deterrence, as described above. However, the ability to satisfy these conditions varies with the dispute profile—the type of conflict and type of intervener. ₉ High cost tolerant actors such as the Bosnian–Serbs had an incentive to escalate at lower levels early on in the conflict. If the opponent backed down, the Bosnian–Serbs would obtain a greater net benefit than what would have resulted from escalating at higher levels. On the other hand, if the opponent refuses to concede, high cost tolerant players are willing to pay additional escalation costs to press their advantage. ₁₀ In essence, peacekeeping is an improvised technique that has no legal basis in the UN Charter and is consequently not addressed as such (Durch 1993). The rules surrounding peacekeeping have been built by practice over the years, much like common law. The legal basis for operations is the mandate of each separate mission. The UN Security Council has, by its actions, established a broad body of de facto principles for traditional peacekeeping, but no measures exist specifically for ethnic conflict. In general, Security Council resolutions and agreements between the UN and individual nations are the legal basis for the initiation and implementation of peacekeeping operations. A coalition of forces may have the de facto right to intervene once the Security Council labels events as aggression (although that term is not clearly defined in the Charter) and any state that has ratified the Charter agrees to its interpretation by the Security Council.

Six criteria apply: (a) Whether a situation exists the continuation of which is likely to endanger or constitute a threat to international peace and security; (b) Whether regional or subregional organizations and arrangements exist and are ready and able to assist in resolving the situation; (c) Whether a cease-fire exists and whether the parties have committed themselves to a peace process intended to reach a political settlement; (d) Whether a clear political goal exists and whether it can be reflected in the mandate; (e) Whether a precise mandate for a United Nations operation can be formulated; (f) Whether the safety and security of United Nations personnel can be reasonably ensured, including in particular whether reasonable guarantees can be obtained from the principal parties or factions regarding the safety and security of United Nations personnel. ₁₁ The next section outlines key assumptions and propositions derived from the model.

 

3. Propositions

In order to compare the success rates of third-party state versus international organization intervention into ethnic conflict, we begin with the following two assumptions:

1. the probability of successful third-party intervention depends on the ability of the third-party to satisfy all four prerequisites for successful deterrence, as stipulated above: communication, commitment, capability and resolve.
2. states are more capable than international organizations of satisfying more of these prerequisites, more often.

Successful intervention is measured with reference to controlling violence. Regan (1996a: 17) makes a compelling case for why violence is an appropriate indication of success or failure: stopping the violence usually is a key motivating factor for most interveners. Decisions to intervene are often based on some perceived political need for immediate results. Most declarations and public statements tend to focus on stopping the violence, a common goal of third parties in most crises. As Regan (1996: 341) argues, “[t]he key to any intervention strategy is to alter the calculations by which the antagonists arrive at particular outcomes...[t]he goal is to make it too costly for the combatants to continue fighting.” Based on these assumptions, and the basic tenets of deterrence, the following propositions about the effectiveness of third-party intervention are tested:

Proposition 1:	ethnic crises are more likely than non-ethnic crises to be violent (Carment 1993, Davis, Jaggers and Moore, 1997).
Proposition 2:	intrastate crises are more likely than interstate crises to be violent (Regan 1996, Dixon 1996, Licklider 1995).
Proposition 3a/b:	states are more likely than IOs to control hostilities in both ethnic and non-ethnic crises.
Proposition 4a/b:	states are more likely than IOs to control hostilities in both intra- and interstate crises.
Proposition 5a/b:	state and IO intervention is more likely than state-only or IO-only intervention to control hostilities in ethnic and non-ethnic crises (Lund 1996a).
Proposition 6a/b:	state and IO intervention is more likely than state-only or IO-only intervention to control hostilities in intrastate and interstate crises.

These propositions are not intended to be exhaustive. They serve only as a list of key expectations found in the literature and derived from the model of deterrence outlined above. Moreover, the evidence is intended to be indicative rather than absolute. For example, empirical support for these propositions can help to estimate the probability of successful intervention across different conflict profiles. The following typology is used to illustrate one possible rank ordering of conflict scenarios that could be produced through these tests. They are listed in order of probability of success.

Table 1: Probability of Success

State intervention into interstate non-ethnic conflict ------ highest success State intervention into intrastate non-ethnic conflict,   IO intervention into interstate non-ethnic conflict,   IO intervention into intrastate non-ethnic conflict,   State intervention into interstate ethnic conflict,   State intervention into intrastate ethnic conflict,   IO intervention into interstate ethnic conflict,   IO intervention into intrastate ethnic conflict ------ lowest success

The logic associated with these propositions and particular rank ordering can be explained thus. If the conflict is interstate and, for example, based on territory, we would expect decisive outcomes, because there are clearly defined rules and procedures in place to help states negotiate these kinds of settlements—that is, states are recognized actors with legal personality (Dixon 1996). A decisive outcome is one in which the starting and end points are more autonomous. Crises are discrete and outcomes are clearer. Victory and defeat may be more readily identifiable and accepted as such (Brecher and Wilkenfeld 1997).

At the other end of the scale, intrastate ethnic conflicts are usually associated with ambiguous and indecisive results, like compromise or stalemate, symbolizing an unresolved conflict (Licklider 1995). Subsequent crises and violence are anticipated by all adversaries. Ethnic conflicts are more difficult to deter than non-ethnic conflicts, given restraints on satisfying the capability and communication prerequisites for deterrence (Regan 1996). Non-combatants are at risk of large-scale violence, armed militias generally operate outside the purview of authority structures and many internal conflicts spillover into the international domain adding complexity to overburdened management strategies.

Further confounding the resolution of ethnic crises is the fact that they often encompass dynamic processes with at least five stages of escalation. These include a latent stage in which differences between ethnic groups are made salient but there is no overt conflict; an onset phase whereby a “trigger” creates the conditions for violence; a peak point leading to large scale confrontation between groups leading to repression or domination; a de-escalation phase, including perhaps some form of third-party involvement and finally a termination phase resulting in the resolution or transformation of the conflict. Ethnic conflicts can last months, years or decades. The most salient are protracted conflicts, fluctuating in intensity over the course of several decades and involving entire communities (Kriesberg 1997).

Viewed as high politics, the collective claims in the literature on the relationship between security and ethnic conflict are that ethnic strife can lead to violence in three different ways: (1) ethnic diversity compounds existing political and economic problems within states leading to intensified competition for resources and a weakening of the state leading to state failure (Zartman 1995); (2) ethnic conflict carries serious risks of contagion and diffusion through processes know as horizontal escalation (Lake and Rothchild 1995, Davis, Jaggers and Moore 1997); (3) ethnic conflict leads to vertical escalation culminating in interstate confrontation and war (Zartman 1992; Carment and James 1997).

Collective identities are particularly conflict prone because they are derived from fundamental, incontrovertible and non-negotiable values such as language, territory, history and religion (Licklider 1995). Under such conditions a perceived higher threat to core values in certain kinds of internal conflicts, specifically ethnic ones would be anticipated (Carment & James 1995). Bercovitch has argued that the identity and characteristics of the adversaries are important factors that influence the effectiveness of the third party (1993). Disunity and lack of cohesion within the ranks of the adversaries makes it difficult for the adversaries and the third party to engage in any meaningful form of conflict negotiation, because the leader’s lack the power or authority to take decisions or make concessions (Kaufman 1997).

In support of these arguments, Brecher (1993: 164) and Brecher and Wilkenfeld (1997) found that crises are more likely to escalate to violence when (a) three or more actors are involved and (b) they take place within a protracted conflict. Third-party involvement in an ethnic conflict is likely to stimulate interest among more actors (and further complications) and the presence of an intermediary may signal that the events are unfolding within some larger process, such as a protracted conflict.

The fact that most internal conflicts involve a multiplicity of non-state actors adds even greater complication to the mix. Non-state actors, such as ethnic minorities, lack the legal personality of states and therefore have difficulty relating to international norms and procedures that were designed exclusively for states. Brecher (1993: 164) also discovered that crises with violent triggering events, and target responses of equal or higher severity, had a much greater chance of resulting in serious clashes or war.

Not surprisingly, many ethnic conflicts examined by Brecher and Wilkenfeld (1997) have violent triggers. This points to the inherent problem in deterring any internal conflict; the clear lack of enforcement of any solution and the lack of restraints on violence (Lund 1996b). However costly and irrational it appears in human and material terms, violence is a means of regulating behaviour, forging identities, creating a shared history and safeguarding long term political and economic outcomes. Therefore commitment to a negotiated solution as opposed to violent measures is highly dependent on the problem of “agency.׆ In ethnic disputes in particular, leaders must be convinced of a settlement’s long term viability and they must also convince their followers of its merits.

Under such circumstances, third parties have a primary interest in stopping the killing and in containing and reversing the spread of internal unrest. Where consent is absent, violence is widespread and groups are at risk, third party force may come into play. Under such conditions, third parties are likely to be required to take on a multiplicity of functions, including peacekeeping and possibly peace enforcement. Missions that involve goals clearly beyond those of traditional peacekeeping, such as seeking to neutralize local forces and push belligerent parties towards the negotiating table, require different strategies and tactics. Using force entails a different kind of bargaining strategy with hostile militias. Such strategies are suited to relatively narrow time frames within each phase of a conflict (Lund 1996a). The receiving end needs to reflect on the action being undertaken and the costs associated with challenging that action. ₁₂

 

4. Methodology and Operationalization of Variables

This study isolates the influence of three interrelated factors on the level of violence in crises: 1) type of third party intervener, 2) type of conflict, and 3) location. It relies on the actor level data from the International Crisis Behaviour (ICB) Project (Brecher and Wilkenfeld, 1991, 1997). ICB data provide a wide range of cases and variables, that permit an evaluation of crises as compared to a data set that focuses exclusively on violent events, most notably war. Indeed, many crises are managed successfully without recourse to violence. Among other important indicators, the data also provide the number of states involved, the nature of the threat, and issues over which the crisis arose. Such information is vital in determining whether the conflict is intrastate or interstate, ethnic or non-ethnic.

A two-stage content analysis for each crisis was carried out. Each crisis was coded as either intra or interstate respectively and ethnic or non-ethnic respectively. In each instance, the case must include a foreign policy crisis for at least one state (please see Brecher and Wilkenfeld 1997, for coding procedure). The ICB Project’s definition refers to “a situation with three individually and collectively sufficient conditions, deriving from changes in a state’s internal or external environment. All three perceptions are held by the highest-level decision-makers of the actor concerned: a threat to basic values, awareness of finite time for response to the value threat and a high probability of involvement in military hostilities” (Brecher and Wilkenfeld, 1991). Brecher and James (1988:91) have argued that internal conflicts are defined by political, economic and social upheavals at the domestic level. In other cases “external crises have fuelled the fires of internal disruption.” A conflict was deemed internal if it involved a single crisis actor and the trigger to the conflict came from within that state, or it involved multiple crisis actors and the trigger to the conflict came from within a single state (The full list of cases are available from the authors upon request).

Each crisis was also coded on the basis of whether ethnicity was deemed to be a salient factor in the conflict. For the purposes of this paper ethnic conflicts were identified as one of three kinds, secessionist, distributional or irredentist. A secessionist conflict is the formal and informal aspects of political alienation in which one or more ethnic groups seek a reduction of control or autonomy from a central authority. A distributional conflict is one where a group or groups seek redistribution of resources through political and military means. Such conflicts may or may not involve (1) the use of force and (2) politically mobilized, well organized, ethnic insurgency movements. An irredentist conflict is one which triggers a foreign policy crisis for a through an internal challenge supported by a redeeming state (see Brecher and Wilkenfeld 1997).

The non-ethnic conflict category consists of political, military, and ideological wars. There is some consistency in recent research to justify this separate category. For example, Licklider (1995) distinguished between identity civil wars and political/economic wars. Similarly Regan (1996) using three categories separates religious wars from ethnic wars and ideological wars. Of course some ideological conflicts carry with them a heavy ethnic component and similarly not all ethnic conflicts are always about differences in identity. ₁₃

The independent variable, type of intervener, ₁₄ is identified by whether the highest crisis management technique was implemented by a state or by an institution (regional or global). ₁₅ The ICB variables that focus on global organization and regional involvement identify the content of UN or Regional Organizational involvement during the course of an international crisis. ₁₆ A similar coding was utilised in identifying state involvement.

In this analysis, third party involvement was taken to mean a perceptible form of action. Both Dixon (1996) Regan (1996) provide insight into the impact of the intervener’s actions on the party against whom action is directed. Interveners were judged not to be involved by simply expressing concern, raising the issue at the UN or verbal expressions of support or condemnation of one crisis actor. ₁₇ More active forms of involvement were necessary. This included sending a mediation or investigative team (not necessarily a successful one), active participation in negotiations, sending troops, or overtly threatening to become involved if measures weren’t taken.

It should be noted that a fair test of the propositions requires a comparison of the relative effectiveness of different interveners (state vs. IO) when one and/or the other is highly involved in the crisis. Dixon notes that it is difficult at best to separate out the influences of multiple third party techniques (1996). We would also add that multilateral and state-based intervention can occur contemperaneously. Without controlling for “high” intervention, it would be difficult to know whether the variation on the dependent variable (cessation of violence) is the result of the intervener or the size/potency of the intervention (low, medium, high). By selecting only those cases in which the intervener is highly involved, it was easier to make straightforward comparisons without having to deal with issues related to intensity. To determine which actor was most involved, the ICB profiles were assessed to get a measure of the level of UN/RO involvement. This coding was compared to the level of state involvement described in both the ICB case summaries and journal articles (e.g., from the Economist and the NYT Index) relevant to the cases. Each crisis was assigned an intervener code based on which third party was judged to be most active in the crisis. ₁₈

With respect to the dependent variable, we use violence as an indication of success/failure—operationalized using ordinal variables from ICB. Violence is assessed in terms of the intensity (or severity) from no or low clashes to full scale war. For the purposes of this analysis we have dichtomized the severity of violence in order to identify those crises which managed to avoid severe violence and those that did not.

 

5. Analysis of Data

We begin by comparing levels of violence in different crisis settings. Propositions one and two are supported by the data, described in Table 2. When compared with non-ethnic and interstate cases, the percentage of crises experiencing severe violence increases for disputes with an ethnic or intrastate dimension. Compared to the overall percentage of violent crises (58.3), the percentage of violent ethnic (69.0) and intrastate (86.4) disputes is higher, while the percentage of violent non-ethnic (50.4) and interstate crises (53.9) is lower (p < .001).

Table 2: Violence and Conflict Type

	All Crises	Ethnic	Non-ethnic	Intrastate	Interstate
Violence
No Violence	41.7 (265)	31.0 (83)	49.6 (182)	13.56 (11)	46.1 (251)
Violence	58.3 (370)	69.0 (185)	50.4 (185)	86.4 (70)	53.9 (293)
sig. p < .001

 

Table 3: Change in Percentage of Violent Crises
(derived from Table 2)

Conflict Type
Ethnic	10.7% higher
Non Ethnic	7.9% lower
Intrastate	28.1% higher
Interstate	4.4% lower

 

Table 4: Percentage of Violent Crises and Third Party Involvement

	All Crises	Ethnic	Non-ethnic	Intrastate	Interstate
Third Party Most Involved
State and IO 1	35.5	54.2	23.7	85.7	29.6
State	43.7	51.9	38.6	80.0	38.3
UN	77.3	81.4	73.6	100.0	73.6
RO	53.8	70.0	45.2	79.2	48.3
X2	51.5	13.32	37.73	47.83	5.56
p <	.0000	.005	.0000	.0000	.1
Notes: 1. “State and IO” refers to crises in which both a state and an international organization (either the UN, an RO or both) were involved.

 

Table 5: Percentage of Non-Violent Crises and Third Party Involvement

	All Crises	Ethnic	Non-ethnic	Intrastate	Interstate
Third Party Most Involved
State and IO 1	64.5 (62) 2	45.8 (24)	76.3 (38)	14.3 (7)	70.4 (54)
State	56.3 (71)	48.1 (27)	61.4 (44)	20.0 (10)	61.7 (60)
UN	22.7 (143)	18.6 (97)	26.4 (110)	0 (24)	26.4 (178)
RO	46.2 (207)	30.0 (50)	54.8 (93)	20.8 (24)	51.7 (118)
No Third Party	47.4 (152)	37.1 (70)	56.1 (82)	18.8 (16)	50.7 (134)
All	41.7 (635)	31.0 (268)	49.6 (367)	46.1 (544)	13.6 (81)
Measures of Association 3
X2	51.5	13.32	37.73	47.83	5.56
p <	.0000	.005	.0000	.0000	.1
Cramer’s V	.3265	.25944	.3638	.3415	.2924
Gamma	.1549	.2211	.1425	.1595	-.19174
Notes: 1. “State and IO” refers to crises in which both a state and an international organization (either the UN, an RO or both) were involved. 2. Numbers in parentheses show the total number of crises for the given category. 3. The measures of statistical association excludes the category “No Third Party” since the focus is on the comparative effectiveness of third parties.

 

Propositions three, four and five also appear to be supported by the data (described in Tables 4 and 5). Consistent with expectations about the relative capacity of different interveners to deter hostilities, the number of violent crises is lower when states and IOs are involved together—35.5 (state and IO), 43.7 (state-only), 53.8 (RO-only), and 77.3 (UN-only). The same trend is produced when controlling for conflict type (see Table 2), with the exception of intrastate cases. In these cases, RO interventions alone appear to be much less likely to experience violence than any other form of third-party involvement—79% (RO), 80% (state-only), 85.7% (state and IO), and 100% (UN-only). Sole involvement by the UN is the least effective form of third-party intervention into these types of conflicts. Compared to state-only intervention, for example, the frequency of violent cases actually increases by 33.6% (from 43.7 to 77.3 across all cases) when the UN is the sole intervener, by 29% in ethnic crises, by 34% in non-ethnic crises, by 20% in intrastate cases, and by 35% in interstate cases. On the other hand, when states join the UN with a significant contribution to the multilateral effort, the frequency of violent conflicts tends to drop—by 41.8% (from 77.3 to 35.5) across all crises, by 27% in ethnic crises, by 49.9% in non-ethnic crises, by 14.4% across intrastate cases, and by 44% in interstate disputes. State involvement appears to add another, more credible and potent dimension to the deterrent threat. UN involvement alone is especially problematic for ethnic and intrastate cases—protracted ethnic conflict in Bosnia serves well to illustrate how the UN, and associated UN charter limitations, may (and often do) undermine efforts by external states to deter hostilities or resolve conflicts. When external states provide military backing in support of UN threats and resolutions, violence is less likely to be severe.

The next stage in the analysis compares the zero-order correlation between type of intervener and outcome with the results produced when “ethnicity” and “type of conflict” are introduced as controls. We rely on Cramer’s V and Goodman and Kruscal’s gamma statistics, commonly used for drawing inferences about the relationship between nominal and ordinal variables, respectively. Cramer’s V transforms X2 into a single measure between 0–1 which, although crude, offers a basic measure of association between type of intervener and violence. The gamma statistic represents the proportion (percentage) of errors reduced when predicting values on the dependent variable, violence, given information on the independent variable, type of intervener.

The result for both sets of statistics (outlined at the bottom of Table 5) are strong and significant across every category. A more meaningful assessment of propositions three to five, however, can be obtained by observing the influence “ethnicity” and “conflict type” have on the strength and direction of the relationship between intervener and violence. The results indicate that third-party intervention into “ethnic” crises are more likely to function as per standard applications of deterrence theory—the relationship between intervener and outcome (as predicted by the rank ordering on the independent variable) is stronger for “ethnic crisis” (gamma = .22; p < 005). In other words, state and IO intervention together is more likely than state-only or IO-only intervention to control violence, but more so when the crisis is characterized by ethnic conflict.

zero order	gamma = 0.16	p < .0001
Non-ethnic	gamma = 0.14	p < .0001
Ethnic	gamma = 0.22	p < .005
Interstate	gamma = 0.16	p < .0001
Intrastate	gamma = -0.19	p < .1

With respect to comparing interstate and intrastate crises, the results appear, once again, to support the proposition that intrastate crises are more difficult than interstate crises to manage in line with the deterrence propositions outlined earlier. As we move from interstate to intrastate cases, our ability to predict violence levels based on the type of intervener actually increases, but in the opposite direction—we can eliminate 19% of the error when predicting violence levels if we reversed the rank ordering on the independent variable from the order specified by the deterrence model; regional organizations are likely to play a more productive role than external states or IOs when attempting to control intrastate violence. Of course, the test may be picking up a selection effect—less intense disputes are more likely to be addressed by ROs, while the global organization is burdened by weightier issues. On selection effects see: Dixon 1996.

While the cross tabulations and correlations of each variable effect, in isolation, show their impact to be fairly strong, the analysis should not stop there. Since all of the factors operate simultaneously to determine the level of violence, the influences might be confounded by these correlations. More precisely, the relative magnitudes of each effect are hard to determine by examining only one factor in isolation.

The third stage of our analysis applies logistical regression to compare probabilities of violence across categories of third-party intervention. The logit regression aspires to explain violence as the result of all influences operating simultaneously. The logit procedure also allows for assessment of the statistical strength of the relative influences. ₁₉ The findings, specified in Table 6 and summarized in Tables 7 and 8, support the rank ordering stipulated in propositions three to five. The probability of violence is lowest, for example, when states and IOs intervene together, and highest when IOs are involved without significant state military or political intervention. The relative magnitudes of the parameters are of some interest. The parameter estimates show that the factor most strongly associated with an increase in the level of violence is whether a crisis is internal or external. The characteristic most strongly associated with decreased levels of violence is joint mediation by both states and IOs. The significance levels indicate that state mediated crises don’t differ appreciably from unmediated crisis, nor do RO mediated crises.

Table 6: Probability of Severe Violence
Logit Regression Parameter Estimates

Variable	B	S.E.	Wald	df	Sig	R	Exp(B)
State and IO	-.7296	.3296	4.8991	1	.0269	-.0584	.4821
State-only	-.3952*	.3074	1.6527	1	.1986	.0000	.6735
UN-only	1.1299	.2423	21.7379	1	.0000	.1524	3.0954
RO-only	.0127*	.2486	.0026	1	.9592	.0000	1.0128
Intrastate	1.8390	.3491	27.7451	1	.0000	.1740	6.2905
Ethnic	.7666	.1823	17.6818	1	.0000	.3580	2.1524
Constant	-.3858	.1897	4.1347	1	.0420
Note: * denotes a parameter that is not significant at the 5%-level.

	-2 Log Likelihood		741.929
	Goodness of Fit		616.389

		Chi-Square	df	Significance
	Model Chi-Square	108.112	6	.0000
	Improvement	108.112	6	.0000

 

Table 7: Probability of Severe Violence
Logit Regression Parameter Estimates

Control Characteristics		Probability of Severe Violence
Reference Crisis (Constant):		41.0
Third Party Interveners:
State & IO Mediated Crisis		25.1
State Mediated Crisis		31.8
UN Mediated Crisis		68.2
RO Mediated Crisis		41.3
Crises Characteristics:
Internal Crises		81.4
Ethnic Crises		59.9

 

Table 8: Logit Regression Estimated Change in Probability of Severe Violence

Control Characteristics		Change in Probability of Severe Violence
Reference Crisis (Constant):		0.0
Third Party Interveners:
State & IO Mediated Crisis		-15.9
State Mediated Crisis		-9.1
UN Mediated Crisis		27.3
RO Mediated Crisis		0.3
Crises Characteristics:
Internal Crises		40.4
Ethnic Crises		18.9

Table 7 uses the logit parameters to make some inferences about the probabilities of severe violence in crises with certain characteristics, based on the following formula for converting the logit parameters into probabilities—i.e., Prob of Violence = 1/(1+e–(constant + controls)); where e=2.7182. For instance, for the reference crisis, the probability of severe violence would be is 1/(1+e–(–0.3658)), or 40.96%. For a state & IO-mediated crisis, the probability of severe violence falls to 25.06%—1/(1+e–(–0.3658 – 0.7296)). An analogous series of calculations produced the other percentages, which reflect the signs and magnitudes of the logit parameters—internal crisis are strongly associated with violence while state and IO mediated crises more associated with non-violent action. ₂₀

The percentages generated in the Table 7 were used for Table 8—which specifies the difference in the probability of violence for the various categories relative to the reference case. For example, we know the probability of violence for the reference crisis (unmediated, non-ethnic, external) is 41.0%, and the probability of violence for a state and IO mediated crisis is 25.1%. The increase in violence associated with moving from an unmediated (non-ethnic, external) crisis to a state and IO mediated (non-ethnic, external) crisis is 41.0 – 25.1 = 15.9%. This table most directly shows the relative influences of each factor in controlling severe violence. Among other things, the results, once again, indicate that internal and ethnic conflicts appear to be managed more effectively by states than IOs, especially when acting alone.The first set of logit runs are very helpful with respect to isolating the relative influence of three core variables (intervener, conflict type and location) on the level of violence. But these runs examine “intervener,” “ethnicity” and “location” as separate factors in the model. Of course, an equally relevant, and important, question is this—how do third parties control violence in different settings. Put differently, we know the relative importance of, say, state and IO intervention compared to ethnicity (–0.7296 vs. 0.7666) across all cases), but we do not yet know the relative importance of state and IO intervention compared to, say, state-only intervention in ethnic cases, regardless of how important ethnicity is to violence more generally. This requires analysis of the parameter estimates for each of type of intervener when controlling for both “ethnicity” and “location” separately.

The results from a second series of controlled logit runs, described in Tables 9–11, support many of the findings described throughout the paper: (a) ethnic and intrastate conflicts are more difficult to manage, (b) state involvement, alone or with the UN, is associated with less violence and escalation across all types of conflict, (c) state-only intervention appears to be more effective than state and IO intervention together, and (d) the UN, acting in isolation as a global organization without significant economic and, especially, military support from the larger member states, appears to be unable to control violence in most settings.

Table 9: Probability of Severe Violence
Logit Regression Parameter Estimates

	Parameter Estimates
Control Characteristics	1 All	2 External	3 Internal	4 Ethnic	5 Non-ethnic
Reference Crisis (Constant): No Third Party, External, Non-ethnic	-0.3858	-0.3526	0.7431 *	0.3218 *	-0.3456 *
Third Party Interveners: (External, Non-ethnic)
State & IO Mediated Crisis	-0.7296	-0.8494	0.8028 *	-0.2775 *	-1.1154
State Mediated Crisis	-0.3952 *	-0.4414 *	0.3721 *	-0.4196 *	-0.3821 *
UN Mediated Crisis	1.1299	1.062	8.9882 *	0.9794	1.2199
RO Mediated Crisis	0.0127 *	0.0343 *	0.148 *	0.2402 *	-0.113 *
Crises Characteristics
Internal Crises No Third Party, Non-ethnic	1.839			1.9304	1.8457
Ethnic Crises No Third Party, External	0.7666	0.7593	1.1918
Note: * denotes a parameter that is not significant at the 5%-level.

 

Table 10: Probability of Severe Violence
Logit Regression Parameter Estimates

	Parameter Estimates
Control Characteristics	1 All	2 External	3 Internal	4 Ethnic	5 Non-ethnic
Reference Crisis (Constant): No Third Party, External, Non-ethnic	40.47	41.28	67.77	57.98	41.44
Third Party Interveners: (External, Non-ethnic)
State & IO Mediated Crisis	24.69	23.11	82.43	51.11	18.83
State Mediated Crisis	31.41	31.13	75.31	47.56	32.57
UN Mediated Crisis	67.79	67.03	99.99	78.60	70.56
RO Mediated Crisis	40.78	42.11	70.91	63.69	38.73
Crises Characteristics No Third Party, Non-ethnic
Internal Crises	81.05	41.28	67.77	90.48	81.76
Ethnic Crises	59.41	60.03	87.38	57.98	41.44

 

Table 11: Logit Regression Estimated Change in Probability of Severe Violence

	Parameter Estimates
Control Characteristics	1 All	2 External	3 Internal	4 Ethnic	5 Non-ethnic
Reference Crisis (Constant): No Third Party, External, Non-ethnic	0	0	0	0	0
Third Party Interveners: (External, Non-ethnic)
State & IO Mediated Crisis	-15.79	-18.16	14.66	-6.87	-22.61
State Mediated Crisis	-9.06	-10.14	7.54	-10.42	-8.88
UN Mediated Crisis	27.32	25.75	32.23	20.63	29.12
RO Mediated Crisis	0.31	0.83	3.14	5.72	-2.71
Crises Characteristics No Third Party, Non-ethnic
Internal Crises	40.58			32.51	40.31
Ethnic Crises	18.93	18.75	19.61

 

6. Conclusions

Optimism about the ‘New World Order’ between 1989 and 1992 drew attention to the conflict management role of the UN, culminating in the Boutros-Ghali’s An Agenda for Peace. (Since then the OSCE, OAU, ASEAN and OAS also have developed mechanisms for conflict prevention.) The UN’s long-term role in international politics, however, is regarded pessimistically because of apparent failures in the former Yugoslavia, Somalia, Rwanda and Angola.

We postulated that relative to states, UN performance would be unimpressive because it lacks effective leverage relative to most adversaries and is poorly equipped to act as an interface between states and non-state actors. The evidence confirmed our suspicions that the UN is a poor conflict manager. Others have found similar results (Diehl et, al 1996).

What was a little more surprising was the fact that violence is more often prevented when ROs are involved. In the context of our deterrence model, ROs are more likely than the UN to satisfy the four requirements and control violence. This may be as much due to the low intensity conflicts that ROs are more likely to encounter on average. Nevertheless, regional organizations offer several advantages in pursuit of conflict prevention, most notably, familiarity with the history of the locale and parties to an impending dispute. These organizations often have the most at stake and therefore generally are more willing to get involved. By their proximity to a conflict, regional organizations almost inevitably are involved because their members must deal with refugee-related problems and other consequences. Finally, states that hesitate to refer a local dispute to the United Nations—for fear that it will no longer be under their control—may be more willing to see the matter addressed at a regional level.

More generally, the findings suggest that third party interventions succeed and fail for reasons that can be linked to the central tenets of deterrence and preventive diplomacy. Ethnic and intrastate conflicts leading to crises call for superior leverage and leadership that is capable of directing, containing and reducing the level of violence and spillover. Only a few states possess these characteristics.

The analysis set out a framework for understanding deterrence failure and success in the context of contemporary conflicts. Alone, such analyses can never furnish any guarantee of success. Governments still carry the heavy burden of having to choose to act in a timely and responsible fashion in response to a perceived threat. Our findings, while certainly not conclusive, appear to suggest that more work needs to be done in linking analysis to action with respect to early warning, coalition building, and conflict salience, respectively.

First, with respect to early warning, does it really matter whether early warning capabilities are enhanced if managing conflict depends on who ultimately intervenes? Early warning is about anticipation and responding to prevent likely events from occurring (Adelman, 1996: 32). Adelman argues that humanitarian realism is the appropriate frame of reference for developing and enhancing response to early warning. This perspective means that equal weighting be given to both state values and state interests in the formation of preventive policies. In the short run, more effective strategies may mean that international organizations will need to behave more like states. ₂₁

Second, major powers are not likely to become heavily involved in preventing ethnic conflicts until it is clear that substantial political, military or humanitarian benefits will be gained from the intervention. This finding is consistent with Alexander George’s argument that the essence of statecraft is to develop and manage relationships with other states in ways that will protect and enhance one’s own security and welfare. Lund has argued that a state will act in a way that favours its own particular interests and often a state may be unable to act until it has secured the support of its public and political elite (Lund 1996b). Thus, proposals on deterring conflict are often more farsighted than current circumstances permit. This is because the thinking of interested governments has yet to move sufficiently far to reshape their approach to the many ways in which peace is sustained. In other words, to effectively anticipate and address impending conflicts states need not only to consider revising the existing provisions and mechanisms for maintaining peace, but changing the attitudes of its users and public.

Third, rare is the direct proof that preventive efforts are responsible for accomplishing anything significant (Stedman 1995). Even if they have, leaders are confronted with a particularly difficult task when trying to mount large scale preventive efforts. The dilemma deterrence theorists face when trying to identify successes (that is, proving that a retaliatory threat prevented a challenge) applies equally well to preventive action; it succeeds when nothing happens. The problem is obvious—if there is no war to stop, or, for that matter, any other concrete measure of preventive success, leaders are not likely to make the first (often essential) move, even if early involvement is cost effective over the long run.

Fourth, compounding these issues is the fact that most ethnically-based civil wars are local affairs attracting little interest from extra-regional actors. The particularities of ethnic conflicts tend to make them self-limiting as a basis for undermining the sovereignty of other states. In the post-Cold War era, claims that give rise to ethnic violence have expanded across borders in so far as there are kindred groups who might be interested in providing support. In essence this means that most conflicts do not spread like “wild-fire” . Even neighbouring states who might have every reason to undermine a rival have generally shown surprising restraint in expressing unconditional support for secessionist claims. ₂₂

Progress is possible only if coercive diplomacy is nested in a broader set of policies and strategies More specifically, sanctions, mediation and international condemnation may be necessary—but not sufficient—conditions for the management and prevention of today’s conflicts. It may be essential to augment these traditional approaches with those suited to the particularities of ethnic and other internal disputes.

 

References

Adelman, Howard. 1996. “Responding to Failed States” Paper prepared for a conference on Canada and Global Issues, Ottawa, October.

Adelman, Howard and Astri Suhrke, 1996. Joint Evaluation of Emergency Assistance to Rwanda (Norway).

Arend, Anthony Clark and Beck, Robert J. 1993. International Law and the Use of Force. London: Routledge.

Bercovitch, Jacob. 1993. “The Nature of the Dispute and the Effectiveness of International Mediation” Journal of Conflict Resolution, vol. 37, no. 4. pp. 70–691.

Betts, Richard K. 1994. “The Delusion of Impartial Intervention” Foreign Affairs Vol. 73, No. 6, Nov./Dec., 20–33.

Brecher, M. and Wilkenfeld, J. 1997. “The Ethnic Dimension of International Crises” in Carment, David, and Patrick James, eds., Wars in the Midst of Peace: The International Politics of Ethnic Conflict. Pittsburgh: University of Pittsburgh Press.

Brecher, M. 1993. Crises in World Politics (London: Pergamon Press).

Brecher, Michael & Jonathan Wilkenfeld. 1991. International Crisis Behaviour Project Code Book. McGill University and University of Maryland.

Brecher, Michael; Patrick James & Jonathan Wilkenfeld. 1990. “Polarity and Stability: New Concepts, Indicators and Evidence”, International Interactions, vol. 16, no. 1, Winter, pp. 49–80.

Brecher, Michael & Jonathan Wilkenfeld (with Patrick James). 1989. Crisis, Conflict and Instability. Oxford: Pergamon Press.

Brecher, Michael; Jonathan Wilkenfeld & Sheila Moser. 1988. Crises in the Twentieth Century vols. 1 and 2. Oxford: Pergamon Press.

Carment, David, and Patrick James, eds. 1997. Wars in the Midst of Peace: The International Politics of Ethnic Conflict. Pittsburgh: University of Pittsburgh Press.

Carment, David and Patrick James, eds. 1997. The International Politics of Ethnic Conflict: Prevention and Peacekeeping. (University of South Carolina).

Carment, David and Dane Rowlands. 1996. “Explaining Multilateral Intervention in Ethnic Conflict: A Game Theoretic Approach” ISA Paper, San Diego.

Carment, David. 1993. “The International Dimensions of Ethnic Conflict: Concepts, Indicators and Theory”, Journal of Peace Research vol. 30, pp. 137–150.

Davis, D. Jaggers, K and Moore, W. 1997. “Ethnicity, Minorities and International Conflict” in Wars in the Midst of Peace: The International Politics of Ethnic Conflict. Pittsburgh: University of Pittsburgh Press.

Diehl, Paul F. 1993. International Peacekeeping Baltimore: John Hopkins University Press.

Diehl, Paul, F., Reifschneider, J. and Paul R. Hensel, 1996. “United Nations Intervention and Recurring Conflict” International Organization 50:4.

Diehl, P. and Goertz, G. 1997. “Risky Dyads” in Schneider, G. and Weitsman, P. eds Enforcing Cooperation (London: MacMillan).

Dixon, W.J. 1996. “Third Party Techniques for Preventing Conflict Escalation and Promoting Peaceful Settlement” International Organization, Vol. 50:4.

Durch, William J. 1993. The Evolution of UN Peacekeeping: Case Studies and Comparative Analysis, New York: St. Martins Press.

George , Alexander 1991. (ed) Avoiding War: Problems of Crisis Management. Boulder/San Francisco: Westview Press.

Gurr, Ted, Robert and Michael Haxton. 1996. “Minorities Report 1 Ethnopolitical Conflict in the 1990s: Patterns and Trends” Department of Government and Politics University of Maryland, College Park and Minorities at Risk Project Center for International Development and Conflict Management.

Gurr, Ted Robert. 1994. “Peoples Against States: Ethnopolitical Conflict and the Changing World System”. International Studies Quarterly 38: 347–377.

Gurr, Ted, Robert. 1992. “The Internationalization of Protracted Communal Conflicts Since 945: Which Groups, Where and How”. Midlarsky,Manus,I. ed. The Internationalization of Communal Strife. London: Routledge, pp.4–24.

Gurr, Ted, Robert. 1991. “Minorities at Risk: The Dynamics of Ethnopolitical Mobilization and Conflict, 1945–1990”. International Studies Association Annual Meeting, Vancouver.

Gurr, Ted, Robert. 1990. “Ethnic Warfare and the Changing Priorities of Global Security”, Mediterranean Quarterly, no.1, winter. pp. 82–98.

Harvey, Frank. 1997. “NATO and Deterrence Theory Revisited” in Carment, D. and Patrick James, eds. The International Politics of Ethnic Conflict (USC Press , forthcoming).

Harvey, Frank. 1995. “Rational Deterrence Theory Revisited: A Progress Report,” Canadian Journal of Political Science (XXVIII:3, pp. 403–436)

Harvey, Frank. 1997. “Rigor Mortis or Rigor, More Tests: Deterrence and Compellence in Protracted Crises,” International Studies Notes, (Spring, 1997).

Huth, Paul (1990) “The Extended Deterrent Value of Nuclear Weapons.” Journal of Conflict Resolution 34: pp. 270–290.

_________ and Bruce Russett (1990) “Testing Deterrence Theory: Rigor Makes a Difference.” World Politics 42: pp. 466–501.

Hindell, Keith. 1996. “An Interventionist Manifesto” International Relations, Vol. XIII, No.2. pp. 23–35.

Ignatieff, Michael (1993) Blood and Belonging (Toronto: Praeger).

Kaplan D. Robert “The Coming Anarchy” Atlantic Monthly (February 1994).

Kaufman, Stuart. 1997. “Conditions for Successful Peacekeeping” in Carment, D. and Patrick James, eds. The International Politics of Ethnic Conflict (Columbia: USC Press).

Kriesberg, Louis. 1997. “The Phases of Destructive Communal Conflicts” in Carment, D. and Patrick James, eds. The International Politics of Ethnic Conflict (Columbia: USC Press).

Kugler, Jacek (1984) “Terror Without Deterrence.” Journal of Conflict Resolution 28:pp. 470–506.

Lake, D. A. and Rothchild, D. 1996. “Containing Fear:The Origins and Management of Ethnic Conflict” International Security Vol. 21:2.

Lebow, Richard Ned and Janice Gross Stein (1989) “When Does Deterrence Succeed and HowDo We Know?” Paper presented at the Annual Meeting of the International Studies Association.

Lebow, Richard Ned (1981) Between Peace and War. London: Johns Hopkins University Press.

Lund, M.S. 1996a. “Early Warning and Preventive Diplomacy” in Hampson, F. Crocker, C. and Aall, P. eds. Managing Global Chaos (Washington, D.C.) pp. 379–403.

Lund, M.S. 1996b. Preventing Violent Conflict: A Strategy for Preventive Diplomacy Washington D.C.: United States Institute of Peace Press.

Mackinlay, John. 1993. “Defining a Role Beyond Peacekeeping.” Military Implications of United Nations Peacekeeping Operations ed. William H. Lewis. Washington: Institute for National Strategic Studies, McNair Paper 17 June.

Mackinlay, John. 1993. “Problems for U.S. Forces in Operations Beyond Peacekeeping” Peacekeeping: The Way Ahead ed. William H. Lewis. Washington: Institute for National Strategic Studies, McNair Paper 25.

Mandell, Brian. 1996. “Getting to Peacekeeping” . Journal of Conflict Resolution (Spring, 1996).

Neack, L. 1997. “Peacekeeping and Risky States” in Schneider, G. and Weitsman, P. eds Enforcing Cooperation (London: MacMillan).

Picco, Giandomenico. 1994. “The UN and the Use of Force: Leave the Secretary-General Out of It.” Foreign Affairs Sep/Oct, 14–18.

Regan, Pat. 1996. “Conditions for Successful Third Party Intervention in Intrastate Conflicts” in Journal of Conflict Resolution, vol. 40, no. 2.

Regan, Pat. 1996. “Interventions into Intense Intrastate Conflicts: Minimizing the Uncertainty in the Decision Process” ISA Paper, San Diego.

Roberts, Adam. 1996. “The Crisis in UN Peacekeeping” in Crocker, C. Hampson, F and All, P. eds Managing Global Chaos (Washington, D.C. USIP Press).

Ruggie, John Gerard. 1993. “Wandering in the Void: Charting the UN’s New Strategic Role” Foreign Affairs, Nov/Dec., 26–31.

Ruggie, John Gerard. 1993. “The United Nations: Stuck in a Fog Between Peacekeeping and Enforcement” Peacekeeping: The Way Ahead ed. William H. Lewis. Washington: McNair Paper 25.

Ruggie, J. 1994. “The New U.S. Peacekeeping Doctrine”. The Washington Quarterly. vol.17:4.

Stedman, S. J. 1995. “Alchemy for a New World Disorder: Overselling Preventive Diplomacy”, Foreign Affairs (May/June).

Weiss, Thomas G. 1994. “Intervention: Whither the United Nations”. The Washington Quarterly 17: 109–128.

Wendt, D. 1994. “The Peacemakers: Lesson of Conflict Resolution for the Post-Cold War Era” The Washington Quarterly, vol. 17:3.

Whitman, Jim and Bartholomew, Ian. 1994. “Collective Control of UN Peace Support Operations” Security Dialogue. Vol. 25, 77–92.

Wurmser, David and Bearg Dyke, Nancy. 1994. “The Professionalization of Peacekeeping: A Study Group Report” Washington: The United States Institute of Peace, Jan 1994, 1–67.

Zartman, I. William. 1989. Ripe for Resolution: Conflict and Intervention in Africa. Second edition.New York: Oxford University Press.

Zartman, I. William. 1992. “Internationalization of Communal Strife: Temptations and Opportunities of Triangulation” in Midlarsky, Manus, I. ed. The Internationalization of Communal Strife. London: Routledge, pp.4–24

Zartman I. William, ed. 1995. Collapsed States: The Disintegration and Restoration of Legitimate Authority (Boulder, CO: Lynne Reinner).

 

---

Endotes

Note 1: By 1996, the total number of ongoing very serious conflicts (those with 1000 battle-field fatalities or more) had levelled off, either because of military defeat, government concessions, or some form of concerted third party intervention. Some long-standing conflicts were either settled or at least achieved a truce during which the ethnic cleansing and bombing campaigns were interrupted. Among the most notable settlements were those in Bosnia, Northern Ireland and the Philippines. These successes contrast with failed attempts of governments to end internal wars in the Sudan, Burma, India, Iraq and Sri Lanka (Gurr 1994). Back.

Note 2: Despite these shortcomings, UN-based deterrence strategies are potentially useful for two kinds of tasks: preventing an internal conflict from becoming violent as in the Golan Heights (Mandell 1996) or spreading into an adjacent territory as in Macedonia (Kaufman 1997). However, apart from the presence of consent—an important precondition for any peacekeeping operation—preventive deployment is generally not thought to be consistent with current thinking on UN peacekeeping operations. Given the ambitious goals of preventive deployment, which is essentially a form of limited deterrence, it can be effective only if both sides to a conflict agree to and cooperate with the mission. Thus, success is more likely if adversaries also recognize the limitations in using violence to resolve differences and are open to the conditions for peace presented to them. If such acceptance and commitments are lacking are UN-based deterrence operations likely to be effective? Back.

Note 3: For example, Ernst Haas’ now decade-old, but still relevant, study found that the UN has undergone a transformation in regime coherence with intervening periods of success (Haas 1983). The most notable aspect of his analysis is the decline of UN effectiveness due to the reduction in the number of “metaissues” around which state interests coalesce (eg. post-war decolonisation and self-determination in Asia and Africa). In the past, when an internal ethnic conflict involved a varied set of issues and protagonists, there was a sense of collective responsibility to maintain international stability; severe solutions (for instance, complete partitioning of a state) took a back seat to more reasonable power-sharing solutions and conflict management by one or both superpowers or the United Nations. (For example, between 1945 and 1989, only Bangladesh successfully managed to challenge the status quo, due almost exclusively to Indian intervention). Back.

Note 4: This is usually very low, unless there never was an intention to challenge, but this is difficult to determine. It could be high if political leaders have a preference for demonstrating the military’ s weakness in an effort to purge the military of rogue leaders and regain control over foreign policy (as in Chechnya). This too is also difficult to isolate, since “no action” may or may not be interpreted as backing down, while and some “actions” may or may not represent a decision to back down, depending on the provocation or the demands outlined in the third-party ultimatum. Back.

Note 5: This may be high or low, depending on challenger’ s objectives, intentions and capabilities. Bosnian Serb leaders actually preferred certain kinds of interventions by NATO, as long as they were limited. This gave them an opportunity to demonstrate successes against an international coalition of forces sponsored by NATO and UN. The point is that U_i often depends on the type of intervention that is to take place, which is directly related to propositions regarding the relative success of state versus international organization intervention. Back.

Note 6: This may be high or low, depending on challenger’ s objectives, intentions and capabilities. Bosnian Serb leaders actually preferred certain kinds of interventions by NATO, as long as they were limited. This gave them an opportunity to demonstrate successes against an international coalition of forces sponsored by NATO and UN. The point is that U_i often depends on the type of intervention that is to take place, which is directly related to propositions regarding the relative success of state versus international organization intervention. Back.

Note 7: If the challenger backs down, the intervener receives a greater net benefit than had they initially intervened high. The problem is that the challenger is aware of this gap between credibility and capability and will attempt to force the issue by testing the third-party’ s commitment. The net result is that there is a gradual escalation of the conflict until that point when the belligerent has more to lose by continuing to fight. We are concerned with identifying conditions that facilitate this transfer point. The speed at which this occurs depends on the capacity of the intervener to increase the costs of a challenge. Once again, this varies depending on the profile of the intervener and the type of conflict. Moreover, for the multilateral to intervene high at the outset requires an estimation of the opponent’s willingness to continue fighting. There are instances where that willingness might be low (low expected gains) and some where it might be quite high. That willingness is a function of not only the capabilities of the intervener but the other ethnic group (ie Serbs vs Bosnian Muslims / Croatians). b) therefore the capability of the belligerent determines its willingness to back down in addition to expected gains from continued fighting. There may be instances where low capability plus low gains mean that a multilateral is effective in a low intensity intervention(Rwanda). The utility function of the challenger needs to include these factors. If the unilateral is even more likely to control escalation (prevent a challenge) than a multilateral under the same “low capability / low gains” situation, then our expectations regarding the relative capacity of unilaterals and multilaterals to satisfy coercive diplomatic requirements would be supported. In any case, the utility function of the challenger (as noted above) does include these factors in the form of U_i and U_bd. Back.

Note 8: If we combine the utility estimates from each phase, that is, if we assume that a challenger’ s overall utility for backing down at the onset of the crisis is based on evaluation of both pre- and post- intervention scenarios—then the new expected utility model for the challenger’ s decision to back down during the onset stage is U_bd > (z(U_i) + (1-z)(U_s). Back.

Note 9: Events in Bosnia identify the central problem for a third party in managing intrastate conflicts: The benefit of a negotiated outcome decreases for both actors when one of them decides to escalate (Diehl 1993). The further apart the two sides are on a negotiated settlement, the higher the costs associated with disagreement (or the higher the value of a negotiated settlement). An actor concedes, as did the Bosnian-Serbs in late 1995, if the costs of enduring further escalation outweigh the benefits of giving in to the opponent. Back.

Note 10: Knowing this, a third party force may therefore seek to raise the costs borne by one or more of the combatants by engaging in more aggressive operations. These operations, however, also impose a cost on the intervener (MacKinlay 1993b). Theoretically, depending on the objective at hand and resistance met, third parties do have an incentive to escalate. As Schelling (1960) points out, escalation is the coercive side of negotiating a peace plan in which the fear of even greater cost imposition motivates actors to make concessions at the bargaining table. In any intrastate conflict a third party coalition can, in theory, assist in guaranteeing agreements between disputants. Back.

Note 11: The task is made easier if the groups have reached their own self-imposed “hurting stalemate”. Then the purpose of the intervention is exclusively to separate the forces and keep the peace. At this stage a viable low intensity mission requires territorial demarcation as well as some minimal agreement between enemies. Back.

Note 12: Third party coalitions usually begin the bargaining process by articulating proposals for a negotiated solution coinciding with a low intensity mission. This proposal can be either accepted or rejected by the belligerent. This initial action requires no force on the part of the third party coalition. If one side accepts the terms for agreement, then both sides receive the benefits they associate with the proposed outcome. If an offer is rejected, then the bargaining process continues and neither player receives any benefits until one of the sides concedes to a demand. The problem facing both actors is to provide the opponent with an incentive to make concessions with limited costs to themselves. In general, the negotiation process is begun by one or both actors to reconcile their positions over some issue in dispute. In the case of intrastate conflicts these consist of: a) control over territory; b) power sharing arrangements and: c) a ceasefire (Regan 1996. Diehl et. al. 1996). Back.

Note 13: For the purposes of this paper, political, military and ideological conflicts are those where groups refuse to recognize the existing political authorities, which can trigger a foreign policy crisis for the state in question. The result is an internal challenge leading to potential conflict, crisis and war. Conflicts of either the ethnic or non-ethnic variety and inter or intrastate variety that have yet to produce international crises are not included in the analysis. The ICB collection is large enough, with sufficiently diverse spatial and temporal characteristics, to justify its use. Furthermore, given ICB’s selection rules and procedures for identifying cases, it is very improbable that any intense ethnic conflict would be overlooked during data assembly. This time frame covers the history of the UN from its inception to the transition phase into the post-Cold War era. Back.

Note 14: Crises with IO involvement almost always have state involvement as well, but we separate interventions into three categories (listed below). The reader should note that we are interested only in whether the IO or a state had the highest level of involvement. If we were unable to determine whether the IO or the state had the highest level of involvement we coded it as ambiguous and put it into category three: (3) IO with state involvement -mixed (UN peacekeeping AND US/NATO air strikes later on); (2) no IO, only third party state involvement (Chechnya); (1) no state, only IO involvement. Back.

Note 15: In broad terms, UN intervention in ethnic conflict includes good offices, mediation, peacemaking, peacekeeping, protection of human rights, humanitarian assistance and stigmatization of rogue governments. Call for condemnation by the General Assembly would fall under the rubric of intervention even though the UN need not take any follow-up action. Back.

Note 16: UN and RO activity includes the following ICB categories: Security (or Regional) Council discussion without resolution, resolution without action, resolution with authorized members active, and General Assembly (or Regional Assembly) discussion only. More direct UN and RO activity includes fact-finding mission, mediation, arbitration, and good offices resolution,. Coercive or high forms of intervention include: sanctions, embargoes, threats, observer forces, emergency military forces and peace enforcement (see Brecher and Wilkenfeld 1989, 1997). Back.

Note 17: A correlation matrix for the logit runs was produced to check for multicollinearity among independent variables. The matrix suggests that multicollinearity was not a problem. The few relatively high correlations (.3–.4 range) were between variables we would expect to be partially related—e.g., UN-ACTIVE and ST–ACTIVE. High UN involvement obviously requires some form of state involvement, but states can be involved in a crisis without UN backing. Back.

Note 18: The second form of the intervener variable focused exclusively on high involvement of third parties. In determining whether third party involvement was high, the deployment of forces and/or coercive diplomacy defined as sanctions, embargoes and threats was the criteria for selection. Back.

Note 19: Significance levels—while informative, are not crucial when working with a universe of relevant cases. Significance levels measure the probability that the results obtained from a sample of cases were obtained by chance or measurement error, and offer a preliminary indication of how likely it is that results apply to the universe of cases under investigation. Since our investigation examines the universe of crises int he relevant time period, significance levels assigned to Chi square statistics should be interpreted with this in mind. Back.

Note 20: The raw parameter estimates of any logit regression are difficult to interpret directly. Parameter estimates show the increase in the log-of-the-odds-ratio associated with a one unit increase in a given regressor. (Suppose there is an 80% chance of winning a lottery—then the odds ratio would be 4:1 or 4—then you’d take the log of 4 to find the log-of-the-odds-ratio). Back.

Note 21: Other possibilities would be to shift responsibilities to coalitions of the willing or possibly the development of dedicated and capable rapid reaction force capable of responding quickly and efficiently. There is a need to build regional capacity whereby a hub of states, presumably those with the most at stake, would take a lead in responding. Back.

Note 22: Several policy and theoretical questions need to be addressed in future research. For example, does intervention frequency and success vary depending on historical circumstance? In other words, are the prerequisites for deterrence success more or less difficult to satisfy pre- or post-Cold War—as per Bosnia, Rwanda, Chechnya, Somalia. Are some prerequisites more or less important (relevant) than others when controlling escalation of ethnic conflict? There are also questions about the relationship between why an intervention takes place (an issue we do not address in the paper) and why interventions are challenged. Another objective is to extend the work on phases (Kriesberg 1997, Fisher 1995, Lund 1996). Is the timing of intervention related to the probability of success? Perhaps less intense interventions are more likely to succeed if they take place during the onset phase, while larger interventions are more likely to succeed during the escalation or crisis phase. Weak interventions during the crisis escalation (or peak) phase, for example, may actually provoke violent responses, because challengers take advantage of the opportunity to demonstrate their superiority (Harvey 1996, 1997). The Bosnian Serb refusal to comply with weak UN threats, and subsequently escalate attacks on UN declared safe havens within hours of the passage of several UN resolutions, serve well to illustrate the point. Back.

Early Warning and Early Response