Russia in Global Affairs

For citation, please use:
Maltsev, A.N., 2026. Allies, Partners, and Rivals: A Three-Dimensional Classification of Military-Political Alignments. Russia in Global Affairs, 24(2), pp. 102–127. DOI: 10.31278/1810-6374-2026-24-2-102-127

The formation, maintenance, and operation of formal alliances is viewed as a paradoxical phenomenon in contemporary IR theory. States’ power and intentions are uncertain over the long run, but a legally-binding agreement mandates long-term cooperation. Although the fulfillment of security obligations is in no way guaranteed (Waltz, 1979), their very existence demonstrates states’ desire to formalize the convergence of their interests, willingness to cooperate, and preparedness for collective defense.

Alliance formation is usually explained by the balance of power. Middle and great powers unite against more powerful forces (Waltz, 1979) or greater threats (Walt, 1987) (balancing)—like they did against Germany in the World Wars—while middle and minor powers tend to align with greater ones (bandwagoning)—like the U.S. and the USSR during the Cold War. Alliances’ duration depends on the stability of the world order and balance of power. In times when states rapidly acquire/lose their military power (e.g., in war), coalitions are supposed to quickly disintegrate and reform, while long periods of stable rivalry promote the maintenance and strengthening of alliances.

However, the end of the Cold War and bipolar system called into question the basic principles underlying the Realist theory of alliances. The principle of balancing expected NATO’s abolition after the collapse of its adversaries. This view was shared by Kenneth Waltz, who argued that “NATO’s days are not numbered, but its years are” (Waltz, 1993, p. 76), and by Mearsheimer, who repeatedly stated that “NATO ceases to function as an alliance” (Mearsheimer, 1990, pp. 6-8). Yet NATO endured and actually expanded, while maintaining vigorous military activity in various regions, from the Balkans to the Middle East. To explain this, structural Realists turned to the concept of bandwagoning, claiming that old and new NATO members thus expressed support for unipolarity and the U.S.’s military dominance (Walt, 1997, p. 171; Waltz, 2000, p. 19). But why did the UK, France, Germany, etc. bandwagon instead of building an EU-based military bloc as a counterweight? Obviously, balancing and bandwagoning are complex and require scrutiny of states’ interests and motivations in security cooperation (Müller, 2013, pp. 611-612).

Patricia Weitsman (2004) shares the fundamental Realist postulate that in anarchic international relations all countries pose a military threat to each other. Yet the level of immediate threat to each country and the number of potential aggressors among neighbors vary historically, which makes identification of balancing’s or bandwagoning’s specific goals difficult. So, a state’s readiness for and intensity of security cooperation depend on the level of threat, and this correlation is not linear. Weitsman (2004) proposed two additional strategies: hedging and tethering. Against a weak but highly aggressive threat, states hedge, maintaining and formalizing limited military partnerships with as many states as possible. This enhances the transparency and predictability of the strategic environment, facilitates a transition to full alliance as necessary, reduces dependency on any particular third country, and (especially in the case of minor and middle powers) supports foreign policy autonomy. Against a more powerful but less aggressive threat, states tether each other with more formalized obligations, deterring the threat by increasing the costs of unexpected aggression, and creating a framework for cooperation to freeze existing disagreements.

Per Weitsman, the failure of hedging and tethering leads to balancing. When war is seen as inevitable and a compromise with the threat as impossible, partnerships are expanded and institutionalized as full alliances. Conversely, bandwagoning is adopted when potential allies are completely absent (or distant) and the threat is obviously superior.

Weitsman’s classification clearly illustrates the dynamics of alliances in the late 19th and early 20th centuries, and hedging and tethering are particularly helpful regarding defense cooperation in modern Europe (Ringsmose and Webber, 2020) and East Asia (Jackson, 2014).

However, discussions regarding the rationale behind military partnerships continue. Some authors believe that Western European countries’ heavy dependence on the U.S. resembles bandwagoning more than hedging or tethering (Cladi, 2022). Interestingly, as some Russian theorists have noted, flexible, multilateral, and ‘de-centered’ security cooperation is more typical of non-Western countries (Russia, China, India) and often accompanied by public criticism of formal alliances. Such criticism is clearly anti-hegemonic and often directed at the U.S. (Istomin and Baikov, 2020, p. 20).

Reliably verifying the hypotheses about alliances necessitates defining the subject’s boundaries.

From Alliances to Alignments: Problems of Concepts and Terms

The conceptual framework of modern alliance theory simultaneously, and often interchangeably, uses a wide range of overlapping notions. English-language academia employs the terms alliance, alignment, coalition, entente, defense cooperation, security cooperation, security community, security organizations, and strategic partnership. Of these, the foremost is the (military-political) alliance, which is defined broadly as “a formal or informal relationship of security cooperation between two or more sovereign states” (Walt, 1987, p. 1), and more narrowly as a “formal association of states for the use (or nonuse) of military force, in specified circumstances, against states outside their own membership” (Snyder, 1997).

This creates confusion, in that ‘alliance’ may refer to all forms of security cooperation (Walt, 1987, p. 1) or more specifically to formal collective defense commitments (Wilkins, 2012). Modern empirical research gravitates towards a narrower, formal-agreement interpretation (Istomin, 2017, p. 94). Although this contradicts Realist prioritization of the real over the nominal, the public status of such declarations makes them convenient for study.

Today, the most comprehensive source of information on international security treaties is the Alliance Treaty Obligation and Provisions (ATOP) database, first published in 2002. As part of the ATOP project, a team of scientists led by Brett Ashley Leeds analyzed and classified more than 200 alliances, dividing them into categories depending on the specific commitments assumed by their members. ATOP distinguishes between defensive and offensive alliances, non-aggression and neutrality pacts, and agreements on military consultations, taking into account the circumstances (political, geographical, etc.) surrounding their initiation.

However, most ATOP-based studies consider only defensive and offensive alliances that obligate joint military action during war.

Per the ATOP database, of the 162 such military alliances formed in 1945-2018, at least 48 are still active. Amongst those signed over the past 25 years, the survival rate is over 70%.

Fig. 2 shows that obligatory defense agreements are concluded and terminated rarely. Approximately 1 to 3 are formed each year. Therefore, the narrow definition of alliances excludes most interstate defense cooperation. Non-obligatory agreements are concluded much more frequently. Once again, the definition of ‘alliance’ used by Leeds in the ATOP database distinguishes between formal alliances and informal non-binding alignments. Such conceptualization does not allow for variation in allied relations, reducing them to a fixed dichotomy. Meanwhile, even formal ATOP alliances indicate that states are likely to conclude agreements that do not envision commitments to direct involvement in military defense (Fig. 3).

A strict distinction between alliances and alignments was first proposed by Glenn Snyder (1990, pp. 104-105) in the early 1990s. In his seminal work Alliance Politics, ‘alignment’ is a generalized concept that includes various types of allied relations in a broad sense (Snyder, 1997). At the same time, Snyder distinguished alliances (legally binding agreements) from looser alignments. Both are based on strategic bargaining: states maximize security interests by choosing between the risk of entrapment in an ally’s undesirable conflict (greater under alliances) and the risk of being abandoned in confrontation with a more powerful adversary (greater under alignments). Bipolar systems are dominated by rigid alliances, while multipolar and unipolar orders lend themselves to non-binding alignments and more pronounced dynamics as states constantly maneuver between the risks of entrapment and abandonment. However, Snyder left the definition of ‘alignment’ vague, and it continues to vary amongst different scholars.

For example, Russian studies of alliances/alignments (Istomin and Silaev, 2018, p. 47; Istomin and Baikov, 2019, p. 35) almost always interpret alignments as ad-hoc coalitions. Western scholars may define them as the convergence of foreign policy priorities (Brooks and Wohlforth, 2005, p. 98), institutionalization of defense cooperation (Cladi, 2022, pp. 631-632), joint participation in armed conflicts (Werner and Lemke, 1997), military-technical cooperation (Yarhi-Milo et al., 2016), etc. The vast majority of recent studies use ‘alignment’ as something self-evident that does not require detailed conceptualization (Erkomaishvili, 2019; Niou and Zeigler, 2023). In English-language scholarship, ‘alignment’ often refers not to a status but to a process of aligning.

This definitional confusion generates false dichotomies, in which certain areas of defense cooperation, such as military-technical cooperation (Yarhi-Milo et al., 2016) or joint military operations (Weitsman, 2010), are viewed as the main alternative to formal alliances. It also inhibits empirical research: military cooperation is often studied separately from the structure of the international system as a whole (Willardson and Johnson, 2022).

Recent years have seen a shift towards Snyder’s interpretation of alignment as a spectrum ranging from various forms of cooperation to legally binding formal alliances. Zachary Selden (2013, p. 340) proposed a spectrum from ‘soft’ informal alignment to ‘hard’ treaty-based alliances, reflected by joint exercises, joint operations, permanent bases, and formal collective defense obligations. Igor Istomin (2021, pp. 35-36) and a team led by Ivan Fomin (et al, 2019, p. 103) similarly propose classifying alignments along the axes of actual cooperation and formal obligation, with the three ideal types of nominal, informal, and genuine alliances.

Thomas Wilkins (2012, p. 57) proposed the categories of alliances, coalitions, security communities, and strategic partnerships, as well as (less commonly) concerts, ententes, and non-aggression pacts.

Wilkins seeks to encompass common conventional terms and to categorize cooperation based on parameters like purpose, nature of membership, level of cooperation, legal formalization of obligations, cohesion of participants, and duration of cooperation.

In practice, however, only two criteria are effectively used, which limits the model’s discriminatory power, to differentiate between the four types: a common military threat (long-term for alliances and short-term for coalitions) and the nature of institutionalization (transnational for security communities and interstate for strategic partnerships).

Although Wilkins’s model is undoubtedly the most complete, it has a number of problems. Firstly, the boundaries between types are unclear, and descriptions are sometimes contradictory. For example, Wilkins argues that strategic partnerships are not directed against a specific threat but may be used for asymmetric balancing (Wilkins, 2012, p. 68). It is unclear how to distinguish security communities from semi-formal multilateral strategic partnerships (e.g., BRICS).

And the boundaries are unclear between ‘strategic partnership’ and ‘mutual understanding,’ and between ‘regime,’ ‘concert, and ‘security community.’

Secondly, identifying the purpose of cooperation is complicated by the uncertainty of states’ intentions and the discrepancy between rhetoric and action. For diplomatic reasons, alignments (such as the AUKUS trilateral pact, the QUAD security dialogue, NORDEFCO, and OCCAR) are rarely named as directed against their targets. Such organizations’ categorization per Wilkins seems unclear and arbitrary.

While all the typologies reviewed above, from Snyder’s dichotomy to Wilkins’s multidimensional model, incorporate multiple features, none empirically verifies the independence of their classification dimensions. They rely on a single assumption that the strategic need to contain a common threat inevitably leads to increased cooperation and institutionalization of relations. The empirical analysis described below shows that this assumption is wrong—the proposed dimensions are statistically independent, which proves the need for a genuinely multidimensional measurement.

Actual Defense Cooperation as a Dimension of Alignment

Based on the above considerations and following the leading Realist theories of alliance (Snyder, 1990; 1997; Selden, 2013; Wilkins, 2012), this article suggests using the term ‘alignment’ as encompassing all (not directly coerced[1]) military-political cooperation, including but not limited to ‘alliances’ (legally formalized defense obligations).

More broadly, ‘military alignment’ will mean all forms of defense cooperation pursued by states in the strategic context of national security interests but not involving direct forcible coercion by one participating state.

The intuitive perception of the term ‘military alignment’[2] presupposes mutual agreement of the parties involved.

I propose three axes (dimensions) to capture alignment’s various forms and contexts: 1) strategic orientation of the states in question; 2) intensity of their actual military cooperation; 3) degree to which their interactions are institutionalized.

Strategic orientation refers to states’ real strategic positions relative to one another in the international system. This is close to Wilkins’s ‘alignment-against-threats,’ when a state identifies its potential military opponents and its potential partners or allies against those opponents. But states may also align with the potential aggressor as part of tethering—formalizing obligations with a moderately threatening neighbor to constrain its behavior—(Weitsman, 2004) or as part of a détente.

Strategic orientation ranges from overlapping interests to clashing interests (the latter potentially resulting in economic warfare, proxy war, and—at the extreme—actual war). On one end of this axis, states strategically align with each other to achieve common goals, while on the other end, they interact amid a foreign policy conflict that may ultimately lead to a strategic rupture. In its most acute form, a strategic rupture may manifest itself in indirect or limited armed conflict (including a proxy war, where a state provides direct military assistance to a warring party involved in an armed conflict, while at the same time staying away from openly confronting it militarily) and then escalate to a full-scale war. States with strongly clashing interests, confronted with strategic nuclear deterrence or the risk of unacceptable conventional damage, may seek limited cooperation to contain escalation. Conversely, even states with overlapping interests may limit military cooperation in order to preserve some domestic and foreign policy autonomy, since extreme alignment can lead to the diminishment or even loss of sovereignty.

The middle of the strategic orientation axis contains various forms of strategic partnership between states with partly overlapping interests and an absence of critical disagreements. Such partnerships feature compartmentalization of military cooperation to select areas (Kalinin, 2023; Mizin, 2025, p. 54). Failure to compartmentalize a strategic partnership leads either to a strategic rupture, as disagreements make all cooperation impossible (e.g., Russia-U.S. arms control in 2026), or to even greater alignment (bandwagoning), in which military cooperation (against a common enemy, which might have previously even been a partner) becomes the key to security due to significantly overlapping interests.

Empirical indicators of strategic orientation can be found in the Militarized Interstate Disputes (MID) database, the Global Sanctions Database (GSDB), and UNGA statistics on convergence/divergence of the states’ voting (ideal point distance) (Bailey et al., 2017). While comprehensive operationalization of strategic orientation is impossible in this article, the following formulae are used for proof-of-concept.

Conflictualness of relations (0-1 scale):

F = 0,30 · x̃_MID + 0,30 · x̃_{san_t} + 0,20 · x̃_{san_c} + 0,20 · x̃_IP

where x̃ = the value’s rank normalization (see below); MID = number of militarized disputes (Palmer et al., 2022); san_t = targeted sanctions; san_c = complex sanctions (GSDB; Morgan et al., 2023); IP = ideal point distance (Bailey et al., 2017).

Strategic orientation (1-5 scale):

C_orient = 1 + 4 [0,60 · (1 − F) + 0,40 · x̃_{jt_MID}]

where F = conflictualness of relations (see above); x̃_{jt_MID} = normalized value of joint military operations per MID.

Future research should account for systemic network effects on strategic orientation, since changes in any one relationship may reverberate through the entire international system. However, this article focuses on dyadic analysis as the first necessary step.

The second axis—intensity of actual military cooperation—characterizes the level of actual defense (military) cooperation between countries. Modern literature applies this term mainly to: supplies of arms and materiel (often called ‘military-technical cooperation’ in Russian (Istomin, 2021)) (Selden, 2013; Fomin et al., 2019) (Yarhi-Milo, Lanoszka, Cooper, 2016; Willardson and Johnson, 2022); the deployment of military bases; joint participation in joint military exercises (JMEs) and multinational military exercises (MMEs); joint military operations. In practice, these things tend to overlap. Transferred arms are usually accompanied by training, maintenance and service, supply of spare parts, etc. Joint military exercises involve the deployment of personnel and infrastructure on the host’s territory—sometimes for quite a while (e.g., for training pilots of modern jet fighters).

Basically, all military cooperation is a combination of 1) joint activity of the armed forces and other security agencies, and 2) provision of material resources and infrastructure. This article’s approach excludes economic, financial, and diplomatic partnerships (e.g., support of UNGA resolutions, which some Russian authors have taken as indicating military-political cooperation (Fomin et al., 2019; Istomin and Baikov, 2020)). Partnership in such spheres may reflect broader political alignment but is not directly related to national security.

As actual military cooperation expands and deepens, the aligning states’ armed forces not only closely interact but begin to rely on each other in various tactical, operational, and strategic tasks. With a high level of such cooperation, states pursue a common security and defense policy, create a common military infrastructure, and integrate their troop command and control systems. Naturally, such structures are often asymmetrical and built around the most powerful state. In this case, the armed forces of dependent countries are tied to the hegemon’s infrastructure and may become an appendage of it. However, as WWII demonstrated, deep integration does not necessarily mean a complete loss of strategic autonomy, and the hegemon may also need some assistance from its allies (strategically important military bases, resources, etc.). Despite NATO’s unified operational and strategic command, unprecedentedly high standardization of arms and materiel, and integration of the U.S. and Western European military-industrial complexes, some members (France, Turkey, and Greece) continue to pursue independent military policies in some spheres, approaching or distancing themselves from the U.S. depending on the situation. Nevertheless, many states prefer to limit military cooperation to certain areas.

 Actual defense cooperation is here measured in arms and materiel transfers, foreign military deployments, and joint military exercises. Both bilateral relations (e.g., permanent military bases) and multilateral ones (e.g., multinational military exercises and deployments) are accounted for.

Data Normalization Procedure

Rank normalization:

x̃i = rank(x_i)

                                                                                                                 N    

where x̃_i = rank normalization of i to an interval of 0-1; rank(x_i) = the observation’s rank in the sample; N = number of observations.

Scaling to 1–5:

S = 1 + 4 · ∑ w_k x̃_k

where S = the final value of a dimension on a scale of 1-5; w_k = expert-weighted coefficients (∑w_k = 1); x̃_k = rank-normalized indicator values.

Actual military cooperation (1-5 scale):

C_coop = 1 + 4 [0,30 · x̃_CIPRI + 0,30 · x̃_IMDT + 0,15 · x̃_JMEb +0,15 · x̃_deploy + 0,10 · x_JMEm]

SIPRI = cumulative (i.e., summed over the period) TIV of arms transfers between the two states (SIPRI Arms Transfer DB); IMDT = average annual personnel of foreign troop deployments between each of the states to the other (Maltsev et al., forthcoming); JME_b = number of bilateral military exercises; deploy = presence of a permanently deployed military contingent (binary: 0 or 1); JME_m = number of jointly attended multinational exercises (Bernhardt, Jordan, 2021).

As can be seen in Fig. 4, the cases form several general clusters: NATO allies (high values on both axes), strategic partnerships (medium on both), and rivalries (low on both).

Aggregation: rank normalization + weighted mean → scale [1 to 5]. Note: The markers denote strategic orientation on a continuous scale (red → yellow → blue). Source: Compiled by the author on the basis of SIPRI Arms Transfer DB, IMDT V2.2, ATOP V5.1, DCAD, SP DB, CoW MID V5.0, GSDB, UN GAVD.

Fig. 5 shows that formally similar relations (for example, China’s nominal ally—the DPRK, and the U.S.’s true ally—the UK) occupy contrasting positions on the axes of actual cooperation.

Aggregation: rank normalization + weighted mean → scale [1 to 5]. Dyads are grouped into three categories—ALLIANCES, PARTNERSHIPS, and RIVALRIES

Fig. 6 provides insight into which specific forms of cooperation shape various profiles.

Note: raw indicators (JME bilateral/multinational, SIPRI TIV, IMDT personal, DCA agreement, SP partnership, ATOP defense, UN dist UNGA) are shown for the same 30 cases.

Level of institutionalization. The third axis of alignment is states’ legal/formal institutionalization of their security relations, thus establishing and formalizing rules and norms of military interaction. The main instruments of institutionalization are treaties and intergovernmental organizations. High institutionalization does not necessarily imply allied relations: arms-control agreements and non-aggression pacts, for instance, may be concluded between opposing states seeking strategic stability. Conversely, de facto allies with overlapping interests may cooperate closely but informally.

Institutionalization publicly affixes the aligning countries’ positioning vis-à-vis each other, their allies, partners, and potential opponents and establishes explicit obligations and interaction mechanisms. It generally contributes to greater transparency and predictability, but it may also antagonize excluded countries. Formal alliances may also entrap a state in unwanted conflicts (Kim, 2011), limit its ability to pursue other partnerships, and ultimately diminish the independence of its foreign policy.

A state may eschew institutionalization if its military and political leadership seeks strategic ambiguity. Such an approach deters opponents by threatening to expand military cooperation with third parties if forced to do so. The absence of formal and public commitments broadens a state’s options, allowing the state to distance itself from de facto allies with minimal reputational damage, and to develop military cooperation with a wide range of countries without attracting unwanted attention or provoking unwanted responses.

Note that secret formal agreements are another matter entirely and require separate study (Istomin, 2017).

Besides formal defense commitments (Alliance Treaty Obligations and Provisions, ATOP v5.1; Leeds et al., 2002), formal defense cooperation agreements (Defense Cooperation Agreements Database, DCAD; Kinne, 2020) and strategic partnership declarations (Strategic Partnership Database) also indicate institutionalization of defense cooperation.

Institutionalization is calculated as follows:

Strict obligations (0-1 scale):

IATOP =      2d + 0,5c + 0,5n

                  max(2d + 0,5c + 0,5n)

where d = defense pact obligations; c = consultation obligations; n = neutrality obligations (ATOP v5.1) (Leeds et al., 2002).

Soft obligations (0-1 scale):

I_SOFT = 0,40 · x̃_DCA + 0,40 · x̃_SP + 0,20 · x̃_orgs

where DCA = bilateral defense cooperation agreements (Kinne DCAD, 2018); SP = strategic partnership declarations (Chang-Liao, 2025); orgs = joint-security organizations.

Institutionalization (1-5 scale):

C_instit = 1 + 4 [0,50 · I_ATOP + 0,50 · I_soft]

where I_ATOP is the strict obligation index; I_soft is the soft obligation index.

Fig. 7 shows the discrepancy between formal status and actual orientation, including cases of ‘paper allies’ (high institutionalization with cold strategic orientation). The framework also creates space for identifying de facto alliances (friendly strategic orientation with low institutionalization), though such cases are rare in the current sample (the closest example being Japan–Australia).

Fig. 8 shows that formal alliance status (per ATOP) does not always correlate with intensity of actual military cooperation.

Fig. 9 shows that no single indicator determines relationship type. Ideal types are defined by varying configurations along all three axes.

Note: Profiles are based on 7 raw indicators (JMEs, SIPRI TIV, IMDT, joint MID operations, formal ATOP alliance, DCAD agreements, proximity at the UNGA) for U.S. alliances, Russian alliances, and non-Western alliances. Indicators are grouped into color-coded supra-categories (military cooperation, institutionalization, orientation).

Measurements’ Independence as an Argument for the Multidimensional Approach

The key empirical evidence for the proposed multidimensional approach is the statistical partial independence of alignment indicators. Within dyads for 2006-2016, the JME, SIPRI, IMDT, Kinne DCAD, and Strategic Partnership DB databases correlate with one another on average at a rate of .13, with the highest correlation (between exercises and agreements) reaching .37. Deployment of military bases (IMDT) barely correlates with anything. Thus, again, formal defense alliances do not always reflect the nature of actual defense cooperation or strategic orientation. Interestingly, various indicators of actual defense cooperation do not always correlate with one another, either, revealing the diversity of possible military alignments.

This confirms the thesis that no single measurement is sufficient to characterize military cooperation. Only consideration of multiple independent axes can identify ideal types (Fomin et al., 2019, p. 103).

The trajectories in Fig. 10 clearly illustrate degradation (Russia—U.S.), convergence (Russia—China), and sharp convergence (Russia—North Korea after 2022).

*  *  *

IR theory explains security interactions between states primarily in terms of alliances, but recent studies are re-conceptualizing relations within the broader category of ‘alignment.’ The dynamics of different alignment types’ formation/disintegration may not agree with the main theories of alliance formation (balancing, bandwagoning, hedging, and tethering).

This article proposes a classification of military alignment along the axes of strategic orientation (overlapping interests to clashing interests), actual military cooperation (from minimal situational interactions to deep integration), and institutionalization (from informal to legally formalized). This classification yields various forms of cooperation, partnership, and alliance.

Naturally, the proposed taxonomy does not claim to cover all forms of military alignment, yet it permits the identification of the main archetypes of alliance that are described by modern alliance theory. Foreign policy decisions on interstate military cooperation should be taken with consideration of its multidimensional context (including strategic orientation and legal framework).

Deeper conceptual analysis and classification remain necessary. Further research should account for asymmetry, economic interdependence, and the influence of domestic politics. Nevertheless, the main finding of this research—that not all alliances are created equal, and that none of the three dimensions can classify them on its own—lays the foundation for a more rigorous and empirically-founded taxonomy of military-political alignments.