From “Frozen Conflict” to Enduring Rivalry: Reassessing the Nagorny-Karabakh Conflict

Laurence Broers. Nationalities Papers. Volume 43, Issue 4, July 2015.


The Armenian-Azerbaijani conflict in Nagorny Karabakh (NK) was the first of several territorial conflicts that emerged in the Soviet Union’s twilight years, and the most instrumental in contributing to the Soviet downfall. Compared to others in its cohort, the Karabakh conflict featured a wider scope and variety of violence, encompassing communal violence and pogroms in 1988-1990; a small-scale Soviet civil war involving opposed army and irregular units in 1991; an all-out war between two newly established sovereign states in 1992-1994; the massive forced displacement of hundreds of thousands of refugees and internally displaced persons; and the extensive occupation of territories far beyond those originally under dispute. It was also unique as the only scenario where direct rule from Moscow was instituted in 1988-1989 as a solution to the problem. This ultimately unsuccessful experiment in higher level jurisdiction served only to refract across the Union the center’s inability to manage the dispute. An estimated 25,000 people lost their lives and over a million Azerbaijanis, Armenians, Muslim Kurds, and others permanently lost their homes. Two closed borders ensuing from the Karabakh war, one winding between Armenia and Armenian-controlled territories and Azerbaijan, and the other between Armenia and Turkey, slice across the South Caucasus and preclude its emergence as a coherent regional space. One recent analysis concludes that the Armenian-Azerbaijani conflict is indeed “the Gordian knot preventing the transformation of this region” (Oskanian, 104).

The extent to which the Karabakh conflict has subsequently receded from the public eye and the focus of research on Eurasian conflicts is therefore surprising. This can be partially explained by the lack of headline news—of either renewed war or a breakthrough in negotiations. Other conflicts in the neighborhood have repeatedly overshadowed it: the Chechen wars in the 1990s, the August War in South Ossetia in 2008, and the events of 2014 in Crimea and eastern Ukraine. By contrast, the 20-year peace process mediated by the Minsk Group of the Organization for Security and Co-operation in Europe (OSCE) has generally failed to make news (or indeed seek publicity). Since its inception in 1992, this process has produced five different peace proposals and while reportedly achieving significant progress in the late 1990s and a near miss in 2001 (at Key West), it is often seen as being too top-down to generate sufficient legitimacy for controversial outcomes (Freizer).

The receding of the Karabakh conflict’s salience can be further explained by three widely circulating conceptual deficits. The first is the inscription of stability into perceptions of the conflict, by deploying a semantics of suspension, stasis, and abeyance when referring to it. This trend is especially salient among international policy-makers, although it also penetrates academic analysis, reflected in the ubiquity of the terms “frozen conflict,” and the similar “no war, no peace.” Although Morar and others have usefully unpacked the problems associated with this terminology, it remains popular; its resilience can perhaps be explained by its convenience. Analytically, however, “frozen conflict” is an oxymoron that creates a blind spot between “freezing point” and “boiling over.” This is what allows small wars, in Ronald Asmus’s formulation, to “shock the world,” after it has failed to recognize dynamism in situations it has been conditioned to regard as frozen (Asmus). Intentionally or otherwise, the semantics of suspension emphasize the anachronism, manageability, and, ultimately, irrelevance of the Karabakh conflict.

A second deficit is the tendency among area studies specialists to research the NK conflict in terms of its typological affinity with the other conflicts in the vicinity, namely those in Abkhazia, South Ossetia, and Chechnya. It is of course useful and legitimate to situate the NK conflict within this wider cohort of conflicts, especially for questions related to onset and process, and several fine studies have done so (Walker; Zürcher; Cheterian; Souleimanov). However, this retrospective lens unfortunately emphasizes 1990s commonalities rather than 2014 differences. This is important because in terms of the balance of power between belligerents, sustainability at the interstate level and the deployment and scale of armed forces on the ground, the Karabakh conflict today offers a very different picture from other Eurasian conflicts of its generation.

Third, there is the growing tendency over the last decade to link South Caucasus conflicts to great power geopolitics in a “new Great Game.” Although geopolitical analysis has always been present in works on the Karabakh conflict (Croissant ), its ever-growing salience over the last decade poses problems. First, as Souleimanov (2-3) and others have argued, South Caucasus conflicts long preceded the interactions of great powers, generating a causal cart and horse problem. Second, overemphasis on geopolitics is fundamentally disempowering of societies in the region. This is a less appealing aspect of “geopolitical peace,” that is, the notion that peace will only be possible when great powers align. This chimerical, and in any case demonstrably false, idea may be contrasted with the “state-society peace,” that is, emphasis on the centrality of political modernization, participatory politics, and democratization to the long-term settlement of conflict. A sustained body of policy-oriented literature produced by local and international peace-building organizations has over the last decade increasingly highlighted the importance of the “state-society peace:” through a focus on elite-society relationships (Conciliation Resources), the potential of a democracy-conflict transformation nexus, the cost of unresolved conflict for the modernization of society (Federal Foreign Office; Alieva), and the capacity of civil society activism to impact community security and conflict dynamics (Saferworld; International Alert). This literature does not underestimate the problems confronting resolution of the Armenian-Azerbaijani conflict, but proposes that transformation in methods and approaches is a pre-requisite to achieving it.

In charting these three deficits, my aim is not simplistic criticism. My point is more that collectively, like the circles of a Venn diagram that fail to meet in the center, these tendencies delimit an analytical void around the Karabakh conflict today. The dimensions of this void may be understood through two fundamental and related questions. First, why has this conflict proved so resistant to resolution? How can we conceptualize its resilience as an active process, a result of purposive agency and action, rather than as an absence of war, a conveniently “frozen conflict” or a result of over-determining geopolitics? Second, in what ways has the conflict changed over the intervening 20 years? What are the sustaining mechanisms and practices by which the conflict continues today, as opposed to its initial causes? How has the arc of causality evolved? The posing of these two questions in itself invites a third: how does the answer to the second question relate to that of the first? In other words, what is it in the way that the conflict has evolved over time that contributes to its intractability?

This article attempts to open up new ways of answering these questions. In so doing, I stress the exploratory and scoping nature of the discussion here: many of my claims are speculative, conjectural, and in need of further empirical elucidation, and I ask more questions than I answer. The article proceeds by summarizing the situation on the ground today. I then propose a new framing of this conflict drawing on international relations theory, before exploring applications of this framing to different layers of the conflict. I also identify a number of directions for future research.

The Karabakh Conflict Today

Summarizing more complex debates, I present here a brief survey of the conflict parties’ positions today. From an Azerbaijani perspective, NK is the unfinished business of the Soviet collapse in 1991 and the re-emergence of an independent Azerbaijani republic. Regaining jurisdiction over NK and the surrounding territories is a foundational moment of contemporary Azerbaijani identity, without which this identity—and Azerbaijani statehood—will remain incomplete. Azerbaijan defines the conflict as an irredentist one and its quarrel is with Armenia, not the Armenians of NK, who are parsed in Azerbaijani thinking as one of two equal communities in NK, alongside Karabakh Azeris. Azerbaijan has borne the cost of a massive humanitarian disaster in the form of hundreds of thousands of displaced persons from NK and the adjacent occupied territories, as well as refugees from Armenia. Furthermore, under a consistent stream of criticism regarding its human rights record, the Karabakh conflict is a rare and emblematic moment where international legal opinion and the claims of the Aliyev government are congruent.

From an Armenian perspective, the self-determination of NK as an entity separate from Azerbaijan finally rectified, through popular will, the injustice of decisions taken by totalitarians in the 1920s, of which there can be no rehabilitation. Karabakh is part of the Armenian heartland and, it is argued, ceding it to Azerbaijan would risk a second annihilation of an Armenian population in their historical homeland. Communal and republican-level violence against Armenian communities in Soviet Azerbaijan in 1988-1991, resulting in their mass expulsion, are popularly cited to attest to this. Armenians see possession of Karabakh as a small victory against the larger backdrop of human and territorial losses in Western Armenia in the early twentieth century; it is part of Hay Dat, the Armenian national idea uniting Armenians across the globe and a transcendent ideology that cannot be negotiated. A distinct Karabakh Armenian perspective exists, marked by greater radicalism vis-à-vis control over occupied territory beyond Soviet NK. In a tactical concession to the international disapproval of irredentism, a unificationist discourse has transformed into one of de facto statehood. The de facto Nagorno-Karabakh Republic is nonetheless dependent on and informally integrated with Armenia.

These positions evolve in the context of a number of dynamic processes, contrasting sharply with both the sense of obscurity attached to the Karabakh conflict and the sense of impasse associated with the peace process. The most striking and visible of these processes is militarization and accompanying tensions. Military expenditures and arms procurements have become a regular feature of reporting on this conflict, as both Armenia and Azerbaijan have become importers of substantial quantities of conventional weapons, and both are reporting efforts to also develop domestic arms production industries (Holtom). Increased tensions along the 160-mile LOC have accompanied this process of rearmament, notably in the summer months of 2014 (Broers; Oskanian). Although there have always been ceasefire violations, these are more prominently reported than in the past, and their timing to coincide with important diplomatic initiatives has also drawn attention (Giragosian). A third tandem dynamic is the increased public affirmations of the conflict and its consequences. In Azerbaijan, this has taken the form of a number of high-level statements reifying Armenian-Azerbaijani antagonism and securitizing dialogue. On the Armenian side, one may note the normalization of occupation as a result of which occupied territory has become routinely referred to as “liberated.” Collectively, militarization, LOC tensions, and the rising temperature of rhetoric have led to much-increased speculation on the possibility of renewed war—a warning consistently flagged, for example, by the International Crisis Group.

Another dynamic process on the ground is Armenia’s accelerating integration with a Eurasian, rather than European, space. In a process integrally linked to militarization, Armenia has been solidifying its alliance with Russia, agreeing in 2010 to host the Russian base at Gyumri until 2044 and in September 2013 to enter the Eurasian Customs Union, dispensing with two years of intense diplomatic work toward a EU Association Agreement (Popescu; Grigoryan). The relative absence of domestic resistance to this decision to close off Armenia’s European path is an indication of the extent to which the security imperatives created by the conflict take primacy.

Less obvious is a third, underlying dynamic: the sublimation of a multi-tiered conflict into a single, overriding confrontation between two states. As Huseynov argues (23-24), full analysis of the Karabakh conflict as it emerged between 1988 and 1994 requires engagement with inter-communal, intra-state, and interstate levels. The inter-communal level describes conflict emerging at the level of local communities in Armenia, Azerbaijan, and in NK itself. This level of the conflict exhausted itself with mass mutual expulsions, following pogroms in Sumgait, Baku, Ganja, and other cities in Azerbaijan in 1988-1990, less well documented yet still evident violence against Azeris in Armenia in 1987-1990 (de Waal, 19-20), and the expulsion of Azeris from NK during the war. By intra-state level, I refer to conflict between Karabakh Armenians and the Azerbaijani state, although I acknowledge that in Armenian perspectives this level of the conflict is made redundant by the discourse of de facto statehood in NK. A cumulative trend over the last 20 years is the submerging of these inter-communal and intra-state layers into a primary interstate dynamic. A resulting paradox is that resolution of these layers to the conflict, essential for the legitimacy of any overall peace agreement, has been effectively deferred until an initial framework agreement is reached. These “lower” layers to the conflict are now visible primarily in civil society debates on approaches to justice for displaced communities and possible strategies for engagement and the reconfiguration of relations between Karabakh Armenians and the Azerbaijani state (e.g. Conciliation Resources).

The Enduring Rivalry Framework

The task, then, is to devise a new conceptualization of the situation, somehow avoiding the traps outlined at the outset and capable of incorporating the above-mentioned dynamics. The particular theoretical framework that I propose for this purpose is that of the enduring rivalry (ER). Now the language of rivalry is already commonly used as a figure of speech to describe Armenian-Azerbaijani relations, but I am using it here in the more limited and specific sense associated with the ER literature in international relations. Appearing implicitly in the 1960s, the concept of ER reached its fullest elaboration in literature produced primarily by US and Israeli scholars in the 1990s and 2000s, boosted by the termination of the “mother of all enduring rivalries,” the Cold War. I now offer a brief overview of the ER concept, before highlighting a number of dividends and problems attending its deployment for a specific case study.

ER theory was premised on the finding that rather than an equal or random distribution of wars among states over the last 200 years, the same small minority of states had fought a wildly disproportionate number of wars, and largely against each other (Maoz and Mor, 3). This finding of repeated and causally recursive conflict between the same set of belligerents held an obvious empirical importance in addressing the concentration of warfare in a small minority of states—what Paul Diehl in an interview referred to as the “career criminals” of the world of interstate war (Weisskopf Bleill). Diehl and Gary Goertz found that if only 5% of interstate rivalries became “enduring,” these accounted for an astonishing 49% of all wars between 1816 and 1992 (Diehl and Goertz, 50-61). ERs count among their number some of the most notorious of militarized competitions: US-USSR, Israel-Arab states, India-Pakistan, and Greece-Turkey. For Stinnet and Diehl (718), enduring rivalries are “the most dangerous form of interstate interaction.” Beyond this empirical relevance, ER theory held out for its advocates the possibility of overcoming the dichotomy between war (the unit of event-centered analysis) and peace (conceptualized by default as a non-event, non-war). By shifting the unit of analysis to rivalry, both the unit of analysis and its termination in peace could be explained (Diehl and Goertz, 71-76).

Diehl and Goertz provided perhaps the fullest elaboration of the concept. Distinguishing between isolated, proto-, and enduring rivalries, and acknowledging the ambiguities of a tight definition, they defined ER as a militarized interstate competition lasting for at least 20 years and involving at least six militarized disputes (44). They found that an escalatory, or what they call “volcano,” model failed to capture the dynamics of the rivalries in their dataset. Instead, they proposed the “punctuated equilibrium model,” asserting that enduring rivalries came into being and ended rapidly, with long intervening periods of stasis (132-141). They conceptualized this stasis as coalescing around what they termed a “basic rivalry level,” a “normal” and extremely stable range of hostile relations between the rivals (165-167). Somewhat pessimistically, Diehl and Goertz highlighted the role of external shocks in terminating rivalries. In a contrasting approach, Maoz and Mor defined ER as: “a persistentfundamental and long-term incompatibility of goals between two states” (4), and took a less structural, more game-theoretic approach to ER, stressing the strategic interdependence of the rivals and the evolution of successive varieties of supergame.

In a literature developing through the 1990s to the mid-2000s, scholars of ERs explored a variety of questions associated with the onset, dynamics, and termination of enduring rivalries. Methodologically, these studies adhered to a strict bifurcation between a majority trend toward multivariate analysis against a large-n data set of rivalries, and a minority trend using game theory. By the late 2000s, however, the ER concept appeared to be in desuetude, which can perhaps be explained by the post-Cold War and especially post-9/11 salience of intra-state conflict and non-state actors in international conflict. These influenced the study of conflict and security in ways that made ER theory less relevant.

How can the concept of ER, drawn from a perhaps outmoded tradition in international relations, be relevant and applicable for the Karabakh conflict today? Some problems need to be highlighted. ER theory consciously defined itself in opposition to what its advocates saw as empirically and texturally rich, yet non-theoretical and non-generalizable, case studies of individual rivalries. ER was conceived more as a set of statistical thresholds for hypothesis testing in order to isolate and delimit patterns across large datasets; its proponents were statisticians and game theorists. The elaboration of ER in the literature remained, therefore, conceptually disembodied, as an abstracted, dyadic, interstate relationship far removed from contextual specificities of history or geography. Can this concept be taken out of its “natural habitat” in international relations and profitably re-contextualized for an area studies approach?

I will argue that with a loose application and some adaptation, it can, and that indeed an ER framework offers considerable new analytical horizons—especially compared with some of the approaches already mentioned. Caveats notwithstanding, I see the analytical dividends as being fivefold. First, and most straightforwardly, the idea of an ER, or to use another term in the literature, a “dangerous dyad” (Bremer), explicitly acknowledges the danger inherent in rivalries of this kind. This displaces the convenience and complacency inherent in the language of “frozen conflict” without, however, predetermining future violence—since enduring rivalries can end without war. Second, ER theory explicitly addresses the interstate dynamics, which, as I have argued, is now the preeminent level of the Armenian-Azerbaijani conflict demanding analysis. It is therefore an interstate theory of competition that is needed; however, to fully operationalize the concept for a specific case, some adaptations and additions will be necessary in order to identify rivalry effects in the domestic sphere, to which I will return below. A third analytical dividend is the opening up of a whole new set of creative and relevant comparisons: the wider family of enduring rivalries. Of course, such comparisons need to avoid crude transpositions and “apples and oranges” problems. Yet, there are numerous striking and intuitive parallels, between the Armenia-Azerbaijan situation and several noted ER relationships (see below). Fourth, ER theory offers a reconfiguration of stasis not as an absence of conflict, but as a dynamic outcome in itself worthy of explanation. Rather than an event-centered focus on wars, it invites a process-driven focus on different levels through which rivalry is sustained. This requires a wider analysis of dynamics at multiple levels over the longer term than only the period immediately preceding the outbreak of actual hostilities. Investigation of these multiple levels offers analytical openings to interpret the deeper long-term social, political, and cognitive processes that actively generate the outcome of stasis. This offers the prospect of understanding the logic of rivalry at each of these levels, how they reinforce one another, and, ultimately, how they may be challenged. Finally, the application of ER theory to the Armenian-Azerbaijani case implies a number of follow-on hypotheses relating to how enduring rivalries come to an end.

Does the Armenian-Azerbaijani conflict fit the definition of an ER? Looking again at Diehl and Goertz’s definition given above, it is evident that the archetypal ER is very suggestive of the Armenian-Azerbaijani conflict today. The conflict has lasted (in different forms) since 1988, and at an interstate level for 24 years. The stipulation of six militarized disputes is more contentious; rather than new wars or resumptions of large-scale violence, the Armenian-Azerbaijani “basic rivalry level” has hardened into consistent low-intensity violence along the LOC. Yet, if actual militarized disputes may be (thankfully) few, as noted already, militarization is a dominant feature of contemporary Armenia and Azerbaijan, and threats or affirmation of the use of force is a constant in the discourse of their leaders. There is no doubt that Armenia and Azerbaijan also exhibit another three features of ERs listed by Maoz and Mor (5): an outstanding set of unresolved issues, strategic interdependence, and psychological manifestations of enmity. The pervasiveness of the antagonism for both states is evident; as Deriglazova and Minasyan note, “the confrontation defines the content and landmarks for foreign policy and the dynamics of internal social and political processes” (45). All in all, the Armenian-Azerbaijani relationship has settled into a relatively low-intensity, yet deepening, ER.

ERs are notable for the totality of their reach. Although formally conceptualized in international relations as an interstate level phenomenon, it seems that for contemporary ERs (as opposed to historical rivalries when some states could act relatively independently of their populations), it is precisely in the convergence of systemic, interstate, societal, and decision-maker-level dynamics that an answer to the first question (why has this conflict proved so resistant to resolution?) can be found. While emphasizing the preliminary nature of the remarks below, I now examine how these levels might be conceptualized and empirically illuminated for the Armenian-Azerbaijan case, identifying both factors embedding rivalry and areas where more research is needed.

The Armenia-Azerbaijan Rivalry

System-Level Factors

Analysis of the roles of system-level factors in enduring rivalries needs to tread a fine line between an overdeterminism recasting local rivalries as proxy wars, and insufficient appreciation of how wider systemic configurations cause or sustain rivalries. The prominence of geopolitical discourse as an explanation of South Caucasus conflicts has been noted above. Great power geopolitics dominates policy debates on the region’s conflicts, and national elites and their supporting analytical communities actively promote geopolitics as the primary level of causality. A preoccupation with Russian motives and actions stands at the center of many accounts of conflict in the South Caucasus, an approach that has seemingly only been vindicated by Russia’s interventions in Ukraine in 2014. At its bluntest, this analysis suggests that Russia drives South Caucasus conflicts and rivalries, which are only refractions of the real rivalry between Russia and the West. Russian-Western rivalry can in this account also “explain” the ineffectiveness of the Minsk Group on account of the conflicting interests of the mediators.

ER theory itself is ambiguous in its treatment of system-level factors. It attributes great importance to external shocks as contributors to the formation and termination of ERs. Perhaps because the “normal” surrounding geopolitical conditions for any rivalry are so context-dependent (place matters!), the processual impacts on rivalries of wider geopolitical configurations are less clear. Yet, analytically, it is clear that there are two questions that are commonly conflated but which it is necessary to separate: What is the role of external actors in the onset of a rivalry? And, what is the role of external actors in sustaining a rivalry once it has begun? With regard to the first question, as already noted, Armenian-Azerbaijani conflict long preceded the context of geopolitical rivalry in the South Caucasus. Indeed, the most egregious periods of Armenian-Azerbaijani violence in the twentieth century (1918-1920, 1988-1994) have occurred at moments of center power collapse. This supports ER theory’s emphasis on external shocks, as it was precisely this kind of shock, the collapse of the Russian Empire and then the Soviet Union, that enabled already contentious sets of issues to take center stage in the (re-)emergence of Armenian and Azerbaijani statehood, in a context where security had broken down.

With regard to the second question, in a necessarily brief overview here, I suggest that systemic factors are enabling, rather than determining, the Armenian-Azerbaijani rivalry. A number of points support this assessment. First, the salience of discourse and rhetoric about great powers in the region is itself not a given but a factor in need of explanation. Great powers and external interventions play central roles in the geopolitical imaginaries of the rivals. Razmik Panossian has demonstrated how benevolent intervention by external powers is a recurring, long-term trope in the Armenian geopolitical imaginary (Panossian, 117, 189-194). This has its contemporary echoes on the Azerbaijani side in consistent invocations of the four United Nations Security Council resolutions of 1993, and the view that the Karabakh conflict will be decided “when the international community makes up its mind.” In a conceptually rich account of the South Caucasus as a regional (in-)security complex, Oskanian has pointed to the importance of constructed insecurities (securitizations) across a triadic structure between and among hostile states, rebellious minorities, and meddling great powers. In his account of what he calls “great power penetration” of the South Caucasus, Oskanian stresses the need to go beyond superficial macro-material indicators and to engage with subjective and inter-subjective elements (53). The resulting account is not one of local proxies and manipulating great powers, but a more complex mutual constitution of geopolitical threats, allegiances, and power relations. In other words, great power penetration is foreseen and enabled by local geopolitical imaginaries, and is not autonomous of them. This is a deus in machina, intrinsic to the rivalry and not imposed by the system in which it currently operates.

A second point is that both the leverage and goals of great powers with regard to Armenia and Azerbaijan are more diffuse than in the conflicts in Georgia or Ukraine; compared to Abkhazia, South Ossetia, Crimea, or eastern Ukraine, Russia has fewer direct levers of influence in Karabakh. It has no peacekeeping forces, no ethnic Russian minorities or passportized “citizens” on the ground, and no direct territorial access into the theater of conflict. Furthermore, neither Armenia nor Azerbaijan straddles the line of east-west competition as they were drawn in Georgia in the period preceding the 2008 war and in the 2014 Ukrainian crisis (a central state authority intent on removing Russian influence and actively seeking alignment with Euro-Atlantic structures). For Russia, this takes a number of ideological and security red lines out of the Karabakh context and allows a more transactional approach. As noted already, one of the rivals, Armenia, has demonstrated (an albeit coerced) compliance with Russia’s Eurasianist ambitions, and has also permitted a deep Russian penetration of its economy. Azerbaijan is also locked into important commercial and strategic relationships with Russia through the Baku-Novorossiisk pipeline and ongoing arms purchases. Hence, the reification of the conflicts in South Ossetia and Abkhazia as prisms of bipolarity and microcosms of Western-Russian rivalry seems less relevant for NK.

In parallel with the more diffuse set of indirect levers at its disposal, Russia does not appear to have a clear overriding strategic goal vis-à-vis the Armenian-Azerbaijani rivalry. This is reflected in the multiplicity of roles it plays. Russia has been, or continues to be, ceasefire broker, mediator, and arms supplier to both parties and security guarantor for one of them. In certain periods, there have also been a number of Russian state actors operating simultaneously, sometimes working against each other. Russia has also on different occasions invested considerably in mediation efforts when these have been seen in Moscow as compatible with Russian interests. Yet, despite having a unique convening power among the external actors in this rivalry, unilateral Russian diplomatic efforts have yielded little. By contrast, as I discuss below, Russian provision of arms to the rivals is critical in enabling the rivalry to continue. Hence, Russia appears in this conflict as enabler and spoiler, but not decider.

The OSCE-mediated peace process may also be seen as a systemic-level actor influencing Armenian-Azerbaijani rivalry. The impacts of rivalries between mediating great powers have been reported with reference to the early period of the Karabakh peace process, for example, in the initiation of duplicate negotiations tracks and turf wars between mediators (Maresca). As the process matured, however, evidence of competition within it has declined: former US and French co-chairs active in the 2000s and 2010s attest to the professionalism and constructiveness of their Russian colleagues. The notion that conflicting mediator agendas have been important in prolonging the rivalry further ignores two countervailing realities. First, there have been moments when the Minsk Group has been the focus of a well-coordinated push for a breakthrough, never more so than at Key West in 2001. Second, compared to its analogues in Abkhazia and South Ossetia, the Karabakh peace process has been remarkably productive in terms of the multiplicity of proposals, the longevity of key concepts, and apparent acquiescence to the broad outlines of a solution (Freizer). There is no shortage of ideas over the last 20 years which Russian, US, and French mediators have been able to agree on, and propose to the presidents. This relatively high margin of cooperation and tacit conceptual consensus suggests caution regarding the explanatory power of great power competition.

Other systemic-level variables that can influence Armenian-Azerbaijani rivalry are outcomes in other conflicts deemed analogous in structure. For example, the outcomes of unipolar recognitions of previously de facto states in Kosovo, Abkhazia, and South Ossetia have been much discussed with regard to their impact on self-determination conflicts more generally. Two contradictory trends emerge from this situation. A first trend is a popular perception in NK (and other de facto entities) that the international community is becoming more tolerant of secession overall, and will eventually acquiesce to legalization of Karabakh’s separation from Azerbaijan. On the other hand, as former Armenian Foreign Minister Vartan Oskanian has argued, selectivity and inconsistency in the way that the international community addresses self-determination conflicts allow belligerents to draw the lessons they prefer (Oskanian). In the wider international system, secessionist conflicts are being dealt with in different ways: while Armenians might make comparisons with South Sudan, Azerbaijanis make comparisons with Sri Lanka. Here again, then, systemic influences appear to be inconclusive and offer support to rival visions.

This admittedly cursory round-up of systemic-level factors suggests that they are permissive of, rather than directly engendering, Armenian-Azerbaijani rivalry. This seems especially true of Russia’s ambiguous “balancing act” seemingly aimed at supporting deterrence as the best alternative to a peace on its terms. As Kevork Oskanian reminds us, NK is the “ … one issue where great powers do cooperate … an exception to the otherwise competitive rule” (Oskanian, 90). The Armenian-Azerbaijani rivalry has nevertheless been able to resist moments of geopolitical conjunction, when a “geopolitical peace” seemed within reach, notably in 2001. This points to the fact that key drivers are not systemic, but at “lower” levels. For as long as these persist, Armenian-Azerbaijani rivalry will outlast specific geopolitical constellations as, for example, the India-Pakistan and North and South Korea rivalries have done.

Interstate Level Factors

This article takes this level to have become the analytically salient layer to Armenian-Azerbaijani relations today, and indeed, the entry point making the idea of ER relevant. When it comes to explaining the origins of the interstate rivalry between Armenia and Azerbaijan, the centrality of territory is clear. This makes it a typical ER: Tir and Diehl report that 81% of all ERs involve territory (Tir and Diehl). What seems less clear is the relationship between a set of unresolved territorial issues and the processes making territorial issues salient in domestic-level politics, and indeed creating red lines that constrain decision-makers and prevent compromise.

Although territory is prominent across the ER literature, it should not be excessively reified as the cause of rivalry. Some very long-running rivalries with a strong territorial component have dissolved (Greece-Turkey), while others have survived without one (the US-Soviet Cold War). In a very useful synthesis of the ER literature with that on asymmetric conflict, Paul has argued that the structure of power differentials between rivals can provide a powerful explanation of the rivalry’s persistence. Paul’s argument addresses the paradox of long-term, sustainable rivalry between rivals exhibiting asymmetry in power. This paradox contradicts the hypotheses of ER scholarship that approximate power parity as necessary in order to sustain rivalry over the long term (Vasquez). Realist expectations would after all suggest that a stronger rival would be able to overcome a weaker one. Yet, the existence and persistence of a number of power-asymmetrical rivalries pose a challenge to realist assumptions: the USA and Cuba, for example, or more pertinently, India and Pakistan. Paul shows that while Pakistan is on a variety of material parameters (size, population, and conventional military capability), the weaker party vis-à-vis India, it has been able through strategy, external alliances, and weapons procurement to challenge India consistently since 1947 (Paul).

Across numerous material parameters, Armenia and Azerbaijan exhibit considerable asymmetry. At 9.3 million, Azerbaijan’s population is three times the size of Armenia’s, reported at just over 3 million. Azerbaijan’s GDP in 2013, at $77.2 billion, was more than seven times that of Armenia’s $10.3 billion. Per capita, this works out to $8300 for Azerbaijan, and $3040 for Armenia. Azerbaijan has further entrenched the economic asymmetry in its favor by ensuring that all regional oil and gas pipelines (the Baku-Tbilisi-Ceyhan and Baku-Supsa oil routes and the Baku-Tbilisi-Erzerum and Trans-Anatolian Natural Gas Pipeline gas routes) circumvent Armenia.

In terms of military capability, the relationship is ambiguous for several reasons. First, data are politicized and there appear to be substantial discrepancies between public rhetoric on military expenditure and the actual sums spent. Second, there is ambiguity about the actual extent of Armenian military capability in NK and the adjacent occupied territories, which is not counted in state-focused indexes and reports such as the Military Balance. Third, material military superiority stands in an uncertain relationship with the intangibles of war: resolve, morale, and discipline. These factors change over time and are often not revealed until war actually happens. On both sides of the conflict, revelations and protests regarding non-combat deaths have been consistent in recent years, suggesting a reality gap between the discourse of military procurement and rearmament and actual readiness for combat. What is indeed striking is that non-combat-related deaths exceed those arising from ceasefire violations. Finally, the possibility of “total war,” involving the targeting of civilians, oil and gas installations, and the Medsamor nuclear power station, suggests that Armenia and Azerbaijan may be locked in a situation of mutually assured destruction, and hence that an effective deterrence dynamic is at work (Minasyan).

These points notwithstanding, asymmetry in two dimensions of military capability may be noted. First, there is little doubt that Azerbaijan’s oil and gas revenues allow it greater military purchasing power. This is reflected in reports of big arms deals with Russia, reports of military procurement relationships being developed with Israel (drones), Turkey (armored personnel carriers—APCs—and rocket launchers), South Africa (APCs), and South Korea, and investment in the domestic arms production industry (Aslanov; Giragosian; Kucera). On paper, Azerbaijan has a larger army than Armenia, and superior airborne offensive power (Military Balance, 170-173); the extent to which this is compensated for by superior combined Armenian and Karabakh Armenian territorial defense capability is a key question driving rhetorical brinkmanship today. Second, military asymmetry has been adopted and reified as a rhetorical strategy: In 2011, President Ilham Aliyev announced that Azerbaijan’s military spending budget at more than $3 billion now surpassed Armenia’s entire state budget, and Azerbaijan has committed to spending at least this emblematic amount since then. I leave aside the controversies over what actually is included in this amount, but certainly, the reproduction of these statistics across numerous sources highlights a new subjective sense of military asymmetry.

Yet, despite its asymmetric disadvantage, it seems that in ways usefully comparable to Pakistan, Armenia has been successful in reducing the asymmetry through strategy, international advocacy, and alliances with outside powers. A number of factors are significant here. First, in another parallel with Kashmir (Paul 13; 617), military asymmetry is not evident in the actual theater of conflict, the LOC between Armenian and Azerbaijani forces. The forces facing each other across the LOC are broadly equal, and Armenian forces draw advantage from the terrain, holding defensive positions on higher ground. Among other justifications, the notion that Armenian occupation of the territories adjacent to that originally under dispute reduces the length of the front, and hence power asymmetry, is an article of faith among many Armenian observers. The Armenian-Azerbaijan case strongly reinforces Paul’s assessment that “local power matters tremendously in making an asymmetric rivalry enduring” (628). In the context of Karabakh, local power parity counterbalances Azerbaijan’s global (overall) power advantage.

Second, certainly for the first 10 years after the ceasefire, Armenian international advocacy drawing on the human and financial capital of the diaspora was much more organized and effective than Azerbaijani efforts. Armenian campaigning succeeded, for example, in securing US Congressional approval of Section 907a of the Freedom Support Act, which prohibited the allocation of US aid to Azerbaijan for 10 years between 1992 and 2002. Diaspora advocacy efforts continue today in the form, for example, of lobbying for the legislatures of US states to recognize the Nagorno-Karabakh Republic (to date, Rhode Island, Maine, Massachusetts, Louisiana, and California legislatures have done so). To some extent, however, in recent years, Azerbaijan has closed this gap as a number of very active Azerbaijani public relations-oriented organizations, such as the European Azerbaijani Society, and networks have emerged.

Third, Armenia has effectively “borrowed power” through its alignment and alliance with Russia. Armenia’s compliance with Russia’s Eurasian project has already been noted. Armenia is also a member of the Collective Security Treaty Organization, and hosts the only remaining Russian military presence south of the Caucasus outside of Abkhazia and South Ossetia, the 102nd Military Base at Gyumri, part of the Commonwealth of Independent States (CIS) integrated Air Defense system, under the command of the Russian North Caucasus Military District, and an air facility at Erebuni. An August 2010 defense agreement commits Russia to assisting Armenia in securing modern and compatible hardware, and Armenia reportedly enjoys preferential prices and access to purchases of Russian military hardware. Again, it seems that Azerbaijan is able to close this gap, buying from Russia at full price what Armenia gets at a discount. But Armenia also undoubtedly benefits at least as much from subjective Russian power projection, and the unambiguous willingness of Russia to deploy force beyond its borders powerfully reinforces Armenia’s capacity to sustain its rivalry with Azerbaijan.

Overall then, the ostensibly weaker party, Armenia, has been successful in mitigating its power asymmetry with Azerbaijan. These have allowed the weaker rival and status quo power to mitigate its material disadvantages and maintain the territorial status quo. This leads me to suggest Paul’s (12) concept of truncated asymmetry as being equally applicable to the Armenian-Azerbaijani rivalry, and indeed, as it is in the Indian-Pakistani context, a key source of the rivalry’s persistence. Hence, in arriving at answers to the second fundamental question posed above (In what ways has the conflict changed over the intervening 20 years?), truncated asymmetry offers some inroads. A multi-layered conflict has become conflated into a peculiarly asymmetric interstate rivalry, in which one rival sees itself as being ultimately able to win through its preponderance, yet the other rival has been able to effectively counter through power parity in the theater of conflict, alliances, and deployment of resources in the international system. This dynamic in the deep structure of the rivalry both lends it considerable stability and embeds it within long-term state policies.

Domestic-Level Factors

The domestic drivers of ER, which lie beyond the traditional (neo-)realist purview of international relations scholars, are under-specified in the ER literature. Several of the key theoretical innovations, such as the basic rivalry level and the punctuated equilibrium model, are largely silent on the domestic dynamics and processes supporting them. This is therefore an area where the ER framework needs adjustment and additions, and on account of the limited space available here, I restrict myself to a few comments pointing in two directions for future research.

The first concerns the already-mentioned role of territory. On the one hand, the role of territory as an outstanding unresolved issue is obvious. On the other, the processes through which territory is suffused with intangible, non-negotiable values, and the ways in which these are then imbued with popular appeal to become politically salient incentives and constraints for decision-makers are less clear—yet typically assumed. The international relations literature, while identifying the centrality of territory to ERs and conflict more generally (Huth; Vasquez and Henehan), appears at times unnecessarily restrained by disciplinary boundaries in its reluctance to consider the domestic salience of territory. The securitization approach taken by Oskanian is certainly useful, but is sketchy in articulating securitization to salient frames and practices relating specifically to territory.

In what might be an unorthodox methodological coupling, future research might profitably draw on perspectives from critical geopolitics. As a methodological response to grand geopolitical narratives (especially that of the Cold War itself), critical geopolitics provides a number of perspectives and tools useful in deconstructing the very notion of ER itself. The realist premise of ER after all unquestioningly imbibes an essential binarism between two rivals and homogenizes a complex array of actors, voices, and audiences into a dyadic structure. The post-structuralist ethos of critical geopolitics might usefully disturb “the objectivist perspectivism” (Ó Tuathail and Dalby, 5) inherent in the language of ER in order to “insist on the situated, contextual and embodied nature of all forms of geopolitical reasoning” (6)—in this case, reasoning that claims the logic, inevitability, and incompatibility of Armenian and Azerbaijani territorial claims. For the purposes of a specific case study, then, critical geopolitics perspectives can achieve the domestic embodiment of ER as a set of strategies, practices, and discursive frames deployed and performed by domestic actors that reproduce rival territorialities and make them routine in the daily life of citizens. Space does not allow discussion of this theme here, but I will simply note that the post-structuralist method of critical geopolitics stresses multiple performative acts—territorializations, visualizations, representations—offering useful methodological openings, for example, in the analysis of cartography, discourse, and everyday images through which territoriality is embodied in contemporary Armenia and Azerbaijan.

A second direction for future research is the relationship between ER and regime type. ER research has found that although the causal dynamics are complex, democratic dyads are less likely to enter rivalries, and that rivalries are more likely to end when both rivals are democratic than when one or both of them are not (Diehl and Goertz, Chap. 6). Diehl and Goertz warn against simplistic allusions to democratization as the “answer.” They explore (107-127) what they call the “rivalry” approach to the democratic peace, which allows them to reformulate a number of important questions. How often and in what kind of rivalry do democracies get involved? What is the role of democratization in affecting the trajectory of a rivalry? Is democracy sufficient to terminate a rivalry? They report broad support for the democratic peace theory, in that democratic dyads are less likely to get into rivalries, and rivalries are more likely to end when both rivals are democratic. But they find no clear evidence about the direction of causality: The onset of rivalry does not appear to decrease democracy, nor do states move rapidly to democracy when a rivalry ends (127). These findings are useful and suggestive in terms of the cross-sectional salience of the democracy variable. The direction of causality in this regard can also be questioned specifically for the Armenian-Azerbaijani conflict: After all, the Karabakh conflict took hold at a time when Armenia and Azerbaijan were more democratic than they are now. Yet, this does not refute the relevance of investigating the dynamics of governance in Armenia and Azerbaijan over the last two decades for the continuation of the rivalry between them.

This leads to a number of questions. To what extent does the rivalry endure because of the type of governments in place in Armenia and Azerbaijan? Would movement toward a different kind of regime lead to the attenuation of the rivalry? What is it about the current regime type that enables rivalry? An opening observation is that semi-authoritarian regimes tend to depend more on symbolic forms of legitimation, which can make them more dependent on emotive, symbolic issues such as territorial conflict. Institutionally, they also lack checks and balances that constrain more democratic regimes from competitive risk-taking displays, civil-military balances favoring the latter, and so on. As Tremblay and Schofield argue for the Indian-Pakistani case (and drawing on the long tradition of the democratic peace theory), strong functional institutional compatibility can attenuate rivalry. Yet, if the direction of causality is inverted, we can hypothesize that a relatively weak and dysfunctional institutional compatibility can embed rivalry through the satisfaction of key elites in each rival benefitting from, for example, arms manufacture and procurements, or monopolies on key commodities arising from closed borders. As accounts by Kevork Oskanian and Behlül Özkan have argued, both Armenia and Azerbaijan exhibit actors and features characteristic of wider post-Soviet governance, or what Oskanian calls “vertical incoherence:” neo-patrimonialism, fusion of political and economic elites, vested interests in the maintenance of monopolies and cartels made possible by closed borders, and an overwhelming securitization of politics (Özkan; Oskanian). These rather schematic remarks highlight the need for further research on the domestic drivers of rivalry, and their interactions with other levels.

Decision-Maker-Level Factors

Scholars working in closely related fields to ERs have highlighted the role of leadership learning as a factor. Russell Leng, for example, has stressed the role of “dysfunctional learning” among decision-makers. Bringing this perspective to Indian-Pakistani learning, Leng has argued that Indian and Pakistani decision-makers have learned the “wrong lessons” from their various crises: Indians have learned that power politics works and Islamabad cannot be trusted, while Pakistanis have learned that since they are not powerful enough to conquer Kashmir, they need proxy wars and great power interventions to support them (110-114). Leng terms these the lessons of Realpolitik and coercive diplomacy. He highlights the role of “cognitive schemata”—”the generic concepts that enable individuals to classify and categorize individuals, events and situations according to familiar types” (Leng, 15), and distinguishes causal learning (about consequences and causality) from diagnostic learning (about the attributes of the other party).

Leng’s social-psychological perspective offers ways into questions (2) and (3) above. This is again a topic for further research, and I limit myself to a few further specifications of the relevant issues. Applying Leng’s perspectives to the Armenia-Azerbaijan rivalry involves the identification of each side’s influence or inducement strategies and how these have changed over time. By implication, this implies identification of the cognitive schemata informing Armenian and Azerbaijani decision-maker thinking; here, the above-mentioned critical geopolitics perspectives can be useful. Second, what are each side’s responses to the other’s inducement strategies, are there tactical adjustments, and what kind of causal and diagnostic learning is the result? Even the specification of these questions suggests a number of hypotheses regarding the role of coercive bargaining and military strategies, and not diplomacy, as the primary source of Armenian and Azerbaijani learning about each other. These points are important because Leng and the wider ER literature stress the significance of imaginative leadership in ending rivalries.

Temporal Factors

I add a fifth dimension aimed at emphasizing the enduring in ER. This is the temporal dimension necessary to capture diachronic change in a conflict context, fully address question (3), and hence avoid the conceptual trap of “frozen conflict.” In a perceptive and suggestive account of Israeli policy and perceptions toward the occupied Golan Heights, Hassner has specified the impacts of longevity for unresolved territorial conflicts. Hassner argues that prolonged stasis results in significant changes in dispute perception and specifically the value accorded to disputed territory: “as these conflicts mature, the perceived cohesion of the disputed territory rises; its boundaries are perceived as becoming more clearly defined; and the availability of substitutes for the territory appear to decline” (110). Material encroachments (construction of transport, infrastructure, energy, and communications networks linking the disputed territory to the homeland) combine with what Hassner terms “symbolic entrenchments” (the construction of new symbolic sites, religious architecture, and archaeological finds establishing links between today’s population and the homeland/the past) to constitute “a process of institutionalization in which disputes take on a life and a causal power of their own” (113). This process encroaches on and eventually excludes the fungible character of territory, insidiously converting it from a chip in a bargaining game into a non-negotiable aspect of collective identity.

Hassner’s observations chime with investigation of these issues vis-à-vis Armenian occupation of territories surrounding the former Nagorno-Karabakh Autonomous Oblast (Huseynov; Broers and Toal). Huseynov draws direct parallels between Armenian policies of supporting settlement and the opening of archaeological excavations in the occupied territories, and Israeli and Turkish policies in Palestine and Northern Cyprus respectively, and argues that these policies narrow the chances for a negotiated settlement (Huseynov, 42). Even if claims of ethnic engineering in the Armenian-occupied territories seem over-stated, given the limited numbers of people alleged to have settled in them, what seems less contestable is that symbolically the passage of time has marked the slow and almost imperceptible transformation of a territorially defined self-determination struggle into a wider and much messier spatial contest, embedding rival territorializations ever deeper into Armenian and Azerbaijani spatial imaginaries.

For Azerbaijan (as for Syria in the case of the Golan Heights) as the party losing control over territory, temporal effects are (even) more insidious, since they are inscribed in an elusive human geography, defined by generational change among displaced communities. This is captured in the title of a 2011 Brookings report written by Yulia Gureyeva-Alieva and Tabib Huseynov Can You Be an IDP for 20 Years? This issue is especially sensitive owing to the troubling perception that without a constituency of returnees, political claims to territory lose their human face, and consequently, legitimacy.

Sustained occupation is transforming and deepening the territorial issues at stake. Whereas the original disagreement concerned the question of to whom post-Soviet NK should accede, whether self-determination was the appropriate framework, and whether an agreed process had been followed, the current situation admits the potential for deeper disagreement over what the territory to be demarcated actually is. This deficit intersects with a second disconnect, that between the territorial categories deployed in Armenian nationalist consensus and those of the Minsk Group (Broers and Toal, 33). Hassner’s warning that “when it comes to territorial disputes, time is not on the peacemakers’ side” underscores the embedding of rivalry in both the physical attributes of territory and symbolic-emotional attachments to it (Hassner, 138).


The Armenian-Azerbaijani confrontation is in urgent need of re-conceptualization. It is no longer a post-Soviet conflict; it is neither only a self-determination conflict, nor only an irredentist one; and it probably never was a “frozen conflict.” This article has argued that the framework of ER, drawn from international relations, offers several new analytical and comparative horizons for the deeper understanding of the Armenian-Azerbaijani conflict as it exists today. Specifically, this framework emphasizes the totality of rivalry as a convergence of conflict dynamics across disparate systemic, interstate, domestic, and decision-maker levels, each of which exhibits change and dynamism over time. All levels are shaped further and in less obvious ways by the inscription of time itself in the physical and human geography of the Karabakh conflict.

An initial and tentative conclusion of this article is that the underlying sources of intractability are to be found in the interactions at the interstate, domestic, and decision-maker levels of the rivalry, rather than in external/systemic factors. The latter enable, rather than determine, Armenian-Azerbaijani rivalry. Particularly important is the asymmetric power dynamic in this rivalry, in which the aggregate material advantage of one of the rivals, Azerbaijan, is countered by power parity in the theater of conflict itself and Armenia’s strategies of alliances and diaspora mobilization. Truncated asymmetry generates a sufficient sense of relative equality for the rivalry to endure. The primacy in this analysis given to internal drivers of the rivalry suggests that the impetus to mitigate it by peaceful means needs to come from within, as well as being enabled from outside. As in the case of other rivalries, imaginative leadership is needed, leadership capable of room for maneuver independent of both the vested interests of political and economic elites, and ethno-nationalist narratives. Transformation of leadership, whether through turnover or ideational change, is necessarily a domestically driven process.

These closing remarks underscore the analytical benefit of the ER approach outlined here, which is its ability—with some adaptation—to accommodate a multifactor analysis, accommodating different theories, addressing different actors and levels of causality. This acknowledges that only a multifactor analysis can address the questions raised at the beginning of this article, and indeed the wider set of questions raised by the Karabakh conflict and other Eurasian conflicts in its cohort. As Ramet (756) remarks at the end of her exploration of the roots of the Yugoslav wars, it would indeed be a pity to rely on just one theory or factor in explaining such complex phenomena—nor is there any reason to do so.