Dealers and Brokers in Civil Wars: Why States Delegate Rebel Support to Conduit Countries

External state support to non-state armed groups is commonly presented as a direct relationship between a state sponsor and a rebel group, often through the lens of principal-agent theory.¹ This theoretical lens has yielded significant insights and attained a paradigmatic status in the conflict delegation debate.² Yet, the theoretical depiction of the state-rebel binary does not mirror the complex reality of the phenomenon, obscuring key actors and aspects of the process of conflict delegation.³ In this article, we present a theory about intermediary states, conceptualized here as secondary, subordinate principals that are part of extended chains of “dual delegation.” We illustrate how principal-intermediary-agent chains of delegation affect wars by proxy. In doing so, we first provide greater clarity regarding how state sponsors select, endow, control, and manage rebels. Second, we explain how intermediaries can empower or constrain rebel organizations’ use of political violence, and how they can affect broader conflict dynamics. We argue that conflict delegation is not a linear, state-rebel process. Rather, it often includes one or more third parties—intermediaries—whose involvement in proxy wars is important, yet underexplored.

The delegation of war to non-state armed groups is neither rare nor new.⁴ Belgin San-Akca estimates that almost two-thirds of all rebel groups active since World War II have received resources from outside states,⁵ and data on external support to insurgencies over the last two centuries demonstrate that the odds of groups receiving aid have increased from about 20 percent to about 80 percent.⁶ Even though states often use intermediaries to channel their support to non-state armed groups, this practice has been ignored by theory-driven case study research. It is also largely invisible in large-N research on the topic, since global datasets assume a direct relationship between state sponsors and rebel groups to facilitate quantitative analysis.⁷ This omission is surprising, given the apparently widespread distribution of external support through intermediaries.⁸

Superpower involvement in many African insurgencies during the Cold War required operating “through the medium of neighbouring African states through which their aid to insurgents could be channelled.”⁹ In the early stages of the Angolan Civil War, China used Zaire and the United States used both Zambia and Zaire as conduits to distribute military support, while the Soviet Union channeled its military assistance through the Republic of the Congo.¹⁰ In Biafra's secessionist war, France channeled support to rebels through Côte d'Ivoire and Gabon.¹¹ Similarly, Ethiopia and Eritrea provided support to each other's rebel groups through Somalia,¹² while Uganda distributed weapons to Sudanese rebels on behalf of Israel.¹³ In Asia, the United States trained Hmong soldiers in camps in Thailand¹⁴ to fight in Laos and North Vietnam, and Tibetan guerrillas operating out of neighboring Nepal. In Latin America, the infamous Brigade 2506 behind the Bay of Pigs fiasco was trained in Guatemala with the support of the Carlos Castillo Armas regime, whose own rise to power in the 1954 coup d’état benefited from Central Intelligence Agency (CIA) backing and operating bases in Honduras.¹⁵ Similarly, the United States’ support of the Contras was enabled by safe havens in Honduras, El Salvador, and Costa Rica, whereas Argentina and Panama facilitated the training and arming of the rebels.¹⁶ In 1994, the United States approved the establishment of an Iranian arms channel through Croatia to support the Bosnian Muslims fighting the Serbs.¹⁷

More recently, Qatar's support to the Dawn faction in the Libyan civil war involved coordinating with Turkey and using Sudan as an intermediary,¹⁸ while Iranian arms reached Houthi rebels reportedly through Oman.¹⁹ Finally, the United States’ backing of the opposition to Syrian President Bashar al-Assad involved training bases in neighboring Jordan and Turkey; these bases were also transshipment points for lethal support.²⁰ All of these empirical observations raise a set of questions: Why are intermediaries used in conflict delegation? How do intermediaries affect the provision of external support? Why do some intermediaries deviate from the state sponsor's preferences while others remain compliant? These are the questions that we seek to answer in this article.

In focusing on the role of intermediaries, this article offers a novel extension of principal-agent theory as applied to the study of conflict delegation. We elaborate on the practice of “dual delegation,” which we define as a process in which a state (principal) delegates authority to one agent (intermediary), which then further delegates this authority to another agent (proxy/rebel group). Specifically, we suggest that powerful states often face a double principal-agent problem when providing material support to rebel groups: the difficulties and problems associated with controlling or rewarding the agent are also reflected in the delegation dynamics between principal and intermediary. We demonstrate that intermediaries exercise agency and can affect several fundamental dynamics of support provision: the selection of agents, the allocation of resources, and the management and directing of agent operations. Consequently, the intermediary needs compensation, reassurances, and, potentially, sanctioning to distribute the resources as intended.

We present two ideal types of intermediaries, which we refer to as dealers and brokers. A dealer has an independent agenda and actively uses its position between the principal and the agent to further its own interest following a competitive logic. A broker, on the other hand, has goals that are much more closely aligned to that of the principal, and hence it sees little reason to deviate from stipulated policy objectives. The broker follows a cooperative logic. In our empirical analysis, we stress the value of knowing the type of intermediary to understand its effect on support provision. We show that the dealer is highly involved in distributing support by selecting proxies, allocating resources, and directing rebel operations. Conversely, brokers are less involved in providing support because they have minimal influence in selecting proxies, less interest in determining how support is provided, and they only partly attempt to direct operations.

This article makes three larger contributions. First, by acknowledging the strategic use of intermediaries in proxy warfare, we open up a novel and extensive research agenda across a set of subfields. This allows us to fundamentally reconsider how to conceptualize international interventions and conflict processes more broadly. Second, we highlight that the type of intermediary is central for understanding the distribution of support. Powerful states that want to channel support through others must consider the convergence/divergence of interests or risk strategic failure. Third, our results point to a need to strengthen the international legal framework in relation to proxy wars in order to assign legal responsibility more adequately. For instance, understanding that both control and direct instructions from states could be channeled through intermediaries is consequential because it invites a reconsideration of the legal thresholds for holding governments accountable for violating the non-intervention principle under customary international law.

The article proceeds as follows. We elaborate on existing research on state sponsorship of non-state armed groups, highlighting how researchers have used principal-agent theory to understand conflict delegation. This section also briefly outlines the core features of principal-agent theory and contrasts it to more complex forms of conflict delegation models involving more than two actors. We then introduce our theory of dual delegation in proxy warfare and our typology of intermediaries. Next, we present the research design before turning to two empirical cases from the 1980s: U.S. support to the Mujahideen channeled through Pakistan, and U.S. support to the National Union for the Total Independence of Angola (UNITA) channeled through Zaire.²¹ The article presents a summary of the main results and ends with some concluding remarks, suggested paths for future research, and a brief discussion of how its findings might inform future policy.

Research on state support to non-state armed groups is a rapidly growing enterprise that has been developing across several subfields.²² Our contribution is located at the intersection of four research areas in international relations: external support in civil wars,²³ state sponsorship of terrorism,²⁴ proxy warfare,²⁵ and covert action and secrecy.²⁶ Despite the sophistication and speed at which these literatures have advanced—both separately and jointly—they overlook how third-party states act as intermediaries in the principal-agent relationship.

The literature on external support mainly focuses on explaining why support is provided and its impact on various conflict dynamics.²⁷ Theoretical accounts assume a direct relationship between state supporters and rebels, while quantitative research designs predominantly render intermediaries indiscernible in empirical tests. Milos Popovic provides what is perhaps the most detailed set of observations on intermediaries. In a discussion on rebel defection, he identifies short and long variants of chains of delegation, the latter including “many autonomous agents,”²⁸ and elsewhere he assesses the implications of multiple sponsors on rebels and rebel alliance formation.²⁹ This analysis of the collective action problems emerging from multiple, parallel principals³⁰ does not, however, consider multiple, subordinate principals, which we discuss in this article.

Similarly, the state sponsorship of terrorism literature focuses on the motives for states utilizing terrorist groups, but it stresses its consequences for the sponsor³¹ or its impact on specific groups.³² Although conceptual work elaborates on the principal-agent model,³³ it mainly focuses on demonstrating its utility rather than extending or revising the theory. The emerging proxy war literature only briefly notes the role of intermediaries in detailed microlevel case studies.³⁴ While this literature understands proxy wars as strategic and interactive bargains,³⁵ intermediaries are not the focal point of analysis nor is their involvement in the process of support distribution theorized. Finally, recent literature on covert action and secrecy predominantly focuses on the state's ability to hinder conflict escalation³⁶ and to increase (im)plausible deniability.³⁷ Despite the key role that intermediaries may play in affecting states’ plausible deniability, they only figure in the periphery of this literature. An exception is Armin Krishnan's analysis on success and failure of paramilitary activity, which acknowledges the role of regional and strategic partners in covert action as part of “multiple layers of agents” involved in such operations.³⁸ Across all these debates, researchers’ discussions about intermediaries are infrequent, largely anecdotal, and insufficient.³⁹

Across the four strands, the principal-agent framework dominates how researchers conceptualize proxy relationships, despite recent criticism.⁴⁰ Even authors who do not explicitly mention principal-agent theory in their analyses adhere to many of its core assumptions implicitly.⁴¹ This tendency is understandable given that the theoretical framework points to several key aspects of conflict delegation, such as the different interests between principal and agent,⁴² problems likely to emerge between the two actors,⁴³ or understanding the process of delegation as part of broader dynamics of substitutability of foreign policy options.⁴⁴ Yet, by elaborating on the role of intermediaries, we tap into a gap across a set of literatures that explore issues ranging from causes to consequences of state sponsorship, both qualitatively and quantitatively.

We propose a paradigmatic shift away from two-actor accounts of delegation of war to proxies to theorizing complex conflict delegation. Our discussion of intermediaries introduces a specific type of complex conflict delegation, namely “dual delegation,” in the first attempt at challenging the flawed assumption that conflict delegation involves exclusively one principal and one agent. In fact, the number of parties in conflict delegation is a variable and not a given, and the variation in conflict delegation patterns—and its puzzles—remains a conspicuous blind spot in the debate.

Delegation is at the core of the principal-agent relationship, and it represents a conditional grant of authority from a principal to an agent that empowers the latter to act on behalf of the former.⁴⁵ The logic assumes that a principal will delegate to an agent a set of responsibilities under certain conditions. First, the exchange marries the principal's potential of making (greater) gains with the agent's strategic advantages. The principal's gains range from ensuring cost-effectiveness and deniability, to avoiding retaliation, to consolidating chances of success by relying on the agent's greater knowledge of the local context.⁴⁶ Second, the principal-agent relationship is contractual, be it formal or informal. The principal agrees to provide support in exchange for the agent assuming responsibility over the delegated task.

But sponsorship should not be conceived of as the mere provision of resources.⁴⁷ In fact, the act of delegation is limited in time and scope, and the principal has the power to both grant authority and revoke it, exercising some degree of control over agents. State sponsors are likely to influence the aims, strategies, and tactics of rebel groups,⁴⁸ which effectively use their own autonomy and agency despite forfeiting their full independence of action. When agents have different interests, they deviate from the principal's task, leading principals to incur costs. This independent action by the agent that diverges from the principal's interest is often referred to as agency slack or moral hazard.⁴⁹ Since the agent has more information than the principal, the principal cannot always ensure that the agent is acting in the principal's best interest. This problem is known as adverse selection, and it is at the heart of the principal-agent problem.⁵⁰

To summarize, principal-agent theory currently proposes that it takes two for delegation to happen, and that it is simultaneously a trade-off and a high risk–high reward gamble. We argue that delegation often involves layered and subordinate delegation relationships, with some agents becoming principals with each added delegation process. We contend that some principals will favor delegating war not to rebels directly but to one or more secondary principals in complex interactive chains. The next section elaborates on these processes, which we call “complex conflict delegation.”

Starting with economics,⁵¹ principal-agent theory is applied to assess more complex forms of delegation involving more than two actors. Such an extension of the principal-agent model is used to explore a diverse set of relationships within political science, business, law, and sociology, rendering different accounts about double, dual, complex, multiple, and (inter)mediated delegation. Notwithstanding the inconsistent use of these labels, the model is used to understand congressional delegation and the separation of powers,⁵² third-party settlement mechanisms,⁵³ subcontracting,⁵⁴ governance and regulation,⁵⁵ as well as the delegation to bureaucracies,⁵⁶ nongovernmental organizations,⁵⁷ and international organizations.⁵⁸

Principal-agent theory has made significant advances in accounting for the complexity of delegation across these literatures. First and foremost, they contain valuable insights from delegation discussions about triadic relationships, premised on three-level hierarchy, involving third-party agents as intermediaries, supervisors, or regulators. For example, research on governance and electoral politics suggests that intermediaries’ employment conditions are shaped by the principal's lack of direct access or capabilities⁵⁹ and the intermediary's local knowledge and influence.⁶⁰ Such reasoning likely informs the decision-making of principals seeking to wage war indirectly. Similarly, theories of financial intermediation reveal a set of functions that range from contracting and monitoring to agenda-setting and coordination,⁶¹ while institutional accounts highlight problems resulting from empowering and disempowering intermediaries.⁶²

These findings are relevant to our theoretical interest in dual delegation; in this situation, the principal-agent relationship is mediated by an intermediary. We do not claim that powerful states always act through intermediaries and that they will never provide direct support to rebel groups. Direct support regularly occurs—either in parallel with or instead of dual delegation. But existing research has only been concerned with the direct transfer of military resources, which obscures key aspects of the process of conflict delegation and their consequences for both the armed opposition and the principal.

Dual delegation is a distinct form of complex delegation in which a principal uses an intermediary to delegate a set of tasks to an agent. In our conceptualization, the principal-intermediary-agent relationship is defined by three salient features: redundancy, sequence, and externality. The relationship is redundant because principals purposefully increase transaction costs by engaging an intermediary rather than directly delegating to the agent. Dual delegation is also hierarchical, occurring sequentially within the chain (from a principal to an intermediary and then to another agent) rather than simultaneously. Finally, the intermediary is an external actor in the principal-intermediary-agent relationship, maintaining distance from the principal and proximity to the agent.

Based on this formulation, we specify three assumptions that inform our theory. First, we analytically separate the type of intermediary under observation from intermediaries that are internal to the principal, such as organizations within the foreign, defense, and security apparatuses. Second, we acknowledge that the power distribution between principals is unequal and that the most powerful actor serves as the principal. We argue that more powerful states use less powerful states strategically to further their own interests and to increase their own security.⁶³ Third, the delegation chain consists of two (i.e., “dual”) sequential transfers of authority.⁶⁴Figure 1 visualizes the key logic of this chain of delegation. In our model, the principal corresponds to the main state sponsor, which we assume will be the most powerful state. External support is then channeled through an intermediary (the conduit state) and received by the agent (the rebels).⁶⁵

The main incentives for a state to become an intermediary would be to deepen a relationship to a particular great power and to reap economic or political benefits. Intermediaries will try to distribute resources so that they promote their own goals, which may not be identical to those of the principal. If the war effort is not going according to plan, intermediaries and agents—or both—may conceal, withhold, or distort information that is harmful to their interests in order to protect their relationship with the principal and the opportunity to benefit from future rewards.⁶⁶

Why would principals channel support to non-state armed groups through intermediaries? While acknowledging there is “no definite list of incentives,”⁶⁷ we suggest that this decision is mainly based on three considerations: (1) informational advantages, (2) proximity, and (3) plausible deniability. The importance of each of these will vary depending on what the principal identifies as the key problems to be addressed in a specific context. Therefore, under certain circumstances, one, several, or all motivations might inform a principal's decision to work with an intermediary.

“Informational advantages” refer to an intermediary's ability to reduce one of the principal's main strategic deficits: the lack of local, specialized, or expert knowledge. Without some gains from specialist knowledge held by an agent, there is little reason to delegate anything to anybody.⁶⁸ First, by enlisting an intermediary, a principal gains access to local knowledge and expertise that help it to overcome limitations related to the strategic context of wars by proxy. Second, because principals “may lack time and task-specific expertise to carry out all required operations,”⁶⁹ they can maximize efficiency by using an agent's expertise. In doing so, principals can access otherwise unavailable or costly information. An intermediary's ability to minimize informational asymmetries is, therefore, a function of possessing and leveraging superior information that lowers the setup costs for wars by proxy when combined with local access. Intermediaries can also monitor the agent and report back to the principal. Since intermediaries are often located in or at the borders of conflict zones, principals benefit from more fine-grained intelligence and local knowledge about actors, military operations, and conflict dynamics. Thus, intermediaries help the principal deal with one of the core problems in the principal-agent relationship: the enforcement problem.⁷⁰

Proximity refers to the intermediary being located geographically closer to the agent than the principal. Proximity offers considerable advantages when it comes to the provision of external support since access to the agent and routes for distributing resources may be restricted. Certain types of support are likely to be of higher value to a rebel group if provided closer to the conflict zone, such as cross-border safe havens.⁷¹ In some strategic contexts, a principal's only option for delegating war to an agent is to establish a relationship with a geographically proximate intermediary state.

Finally, plausible deniability refers to principals’ desire to maximize deniability and minimize attribution. Operating through an intermediary makes it even harder to trace support back to the original state sponsor. Concealing principals, commonly understood as plausible deniability, is a key strategic advantage of waging war indirectly.⁷² The logic of plausible deniability builds on the notion that leaders prefer to establish as much distance as possible between the decision-maker and the supported forces, since this helps states avoid retaliation by other states at the same time as it limits the influence of domestic political actors that seek to influence foreign policy choices.⁷³ It allows senior officials to neither confirm nor deny responsibility.⁷⁴ It might seem counterintuitive for principals to “take cover” in longer chains of delegation, not least because of the increased opportunities for exposure. But conflict delegation often hides in plain sight. The role of intermediaries as discussed here is not merely secrecy, for it is “usually easy to find out who is backing whom, and even when support for a proxy force is an official secret the information is likely to be uncovered at some point.”⁷⁵ Rather, it is about significantly enhancing deniability. Extended deniability chains involving an intermediary decrease the confidence with which others can make statements linking principals to proxies and increase the delay in potential responses. A principal therefore finds strategic utility in intermediaries because their participation induces further uncertainty as part of the principal's efforts to deny involvement and disclaim responsibility in wars by proxy.

Taken together, intermediaries help to alleviate several distinct problems for the principal: limited information and local expertise, restricted access, and the risk of retaliation/sanctioning by other states. Based on this discussion, we identify three hypotheses for why states would use intermediaries.

In standard conflict delegation, if states delegate action to rebel groups, they risk losing agency and autonomy over the objectives and means of the war effort.⁷⁶ Principals face several known problems, including that agents may hide information or conceal actions taken against the principal's interest, and that they may turn against the principal.⁷⁷ New problems are likely to arise in dual delegation processes since the intermediary can also have interests that diverge from the interests of the principal.⁷⁸ Accordingly, the principal faces a double problem of control: the difficulties and problems associated with the agent are also reflected in the delegation dynamics between principal and intermediary. Working through an intermediary state exacerbates the risks involved since both the intermediary and the rebels can act in ways that conflict with the state sponsor's interests.

The logic of delegation assumes that principals should appoint agents who share preferences⁷⁹ and that “the alignment of interests, or objectives between the principal and the agent is of paramount importance.”⁸⁰ In practice, however, this might be too difficult, costly, or impossible,⁸¹ and because the interests of principal and agent do not overlap perfectly, “there will always be agency slippage between what the principal wants and what the agent does.”⁸² This logic applies to intermediary states as well. Since intermediaries are often deeply embedded in the regional context, these states are most likely to also have strategic objectives that may significantly diverge from the principal's goals. Depending on the divergence of interests between the state sponsor and the intermediary, the state sponsor needs to exercise different levels of control. Whether the intermediary acts on its own behalf or adheres to the will of the state sponsor depends on the degree of divergent interests between the state sponsor and the intermediary. Based on this, we differentiate between two ideal types of intermediaries: brokers and dealers.

The defining feature of a broker is that its strategic objectives are aligned with the state sponsor's objectives. Brokers thus act more as simple conduits and exercise little independent agency that might otherwise shift the aims behind the provision of external support. In contrast, dealers have objectives that differ from the state sponsor's and hence increase agency costs by bringing those strategic objectives to the table. Dealers strive to shape how aid is distributed and monitor how it is being used.⁸³ On this basis, we distinguish between two different strategic logics of intermediary behavior in proxy wars: cooperative and competitive. Providing support in a cooperative manner means that the intermediary's actions complement those of the principal. Whereas brokers follow the will of the principal in selecting receivers, allocating resources, and managing and directing operations, dealers follow their own interests in making such decisions. We contend that both types of intermediaries influence the provision of external support, although to varying degrees: dealers are highly involved in key decisions related to support provision compared with brokers, who function more as interlocutors between the principal and the agent. Based on this reasoning, we derive a set of additional hypotheses:

We conducted an initial test of our theoretical argument by leveraging a research design that combines both within-case and cross-case analysis. This enabled us to draw some early inferences about the validity of dual delegation, which are that intermediaries influence the provision of external support, and their behavior differs depending on goal alignment with the principal. Our argument has three scope conditions. First, we assume that the process of dual delegation builds on an inherent asymmetric power relationship between the principal and the intermediary. Although there could be cases in which states of roughly equal power coordinate distributing aid, this would most likely warrant another theoretical lens. Second, we focus on external state support to rebels because delegation dynamics related to security assistance to states may be different. The latter is often public rather than covert and based on formal security agreements or alliance commitments rather than an informal relationship. Third, we restrict our focus to states acting as intermediaries, although our theory of dual delegation could likely be extended to non-state actors.

Methodologically, we first traced the process of conflict delegation in our two empirical cases.⁸⁴ This analysis highlights the distribution chain and shows that intermediaries are indeed present in these types of situations and that they exercise agency. Second, we conducted a structured, focused comparison⁸⁵ to assess our claim that the double principal-agent problem is most acute when the principal and the supervisor have divergent interests. We adhered to a most similar systems design by holding several important features constant between the cases: principal characteristics, period, Cold War context, and type of support provided.⁸⁶ The cases have different types of intermediaries (the independent variable)—a broker (Zaire) and a dealer (Pakistan)—and we assess how this impacts their involvement in the provision of external support (the dependent variable).

This analysis highlights how intermediaries have agency and that, depending on the degree of divergent interests, they do not act as simple transmission points that channel the will of the principal. More specifically, we look at how intermediaries can influence three key dynamics of support provision: (1) the selection of receivers, (2) the allocation of resources, and (3) the management and direction of rebel operations. We define selection as the intermediary's ability to decide who should receive support among individual commanders or factions within a specific rebel group or among various rebel groups,⁸⁷ and it includes the option not to select any of the available groups, factions, or leaders.⁸⁸ Allocation refers to the degree to which the intermediary can determine how to provide the support from the principal. Finally, management and direction of operations refer to the intermediary's ability to directly influence a group's strategy and tactics, such as when and where to conduct attacks.

In each case we focus on one specific instance of dual delegation.⁸⁹ Holding many factors constant allows us to concentrate specifically on the different roles of the intermediaries. Although this approach comes with strong analytical leverage, one drawback is that it limits the generalizability of our findings because we only observe these delegatory relationships from the perspective of one principal at a given point in time. Another limitation is that we cannot systematically control for confounding factors related to the intermediaries, such as capacity or prior experience of distribution. We leave for future research to evaluate whether other powerful states have a similar approach to intermediaries and to further unpack additional intermediary characteristics. Table 1 provides an overview of the most similar systems design.

Empirically we draw on a rich set of archival sources and memoirs by key decision-makers.⁹⁰ The behind-the-scenes reflection provided by archival material as a database of information implies a necessary trade-off concerning access, bias, and selection limitations, with some official documents still classified.⁹¹ To minimize this inherent bias, we triangulate memoirs, public papers, newspaper coverage (ProQuest and LexisNexis), and secondary literature.

After a military coup, the communist party (the People's Democratic Party of Afghanistan) took control over the state apparatus in Afghanistan in April 1978. The following year, U.S. President Jimmy Carter authorized a covert CIA program named “Operation Cyclone” to finance and arm the various Afghan resistance fighters known as the Mujahideen who had started to oppose the communist regime. On Christmas Eve of 1979, the Soviet Union began its invasion of Afghanistan, prompting a significant expansion of the initial U.S. support commitment.⁹² It was clear from the outset that Pakistan, and especially its intelligence agency Inter-Services Intelligence (ISI), would play a central role as an intermediary. Senior officials within the U.S. administration emphasized early on that Pakistan was “the key to ‘doing something’ inside Afghanistan,”⁹³ and National Security Advisor Zbigniew Brzezinski considered Pakistani support absolutely crucial for the Afghani resistance.⁹⁴ CIA Director Stansfield Turner recalls that Carter was easily convinced about the need to provide covert support to the Mujahideen: “The primary thing we discussed was how we were going to pull it off, and that meant Pakistani cooperation. I explained to him how we were going to send Soviet-made weapons [through Pakistan] because we didn't want knowledge of our own involvement to get out.”⁹⁵

Following the Soviet invasion, President Carter personally telephoned Pakistan's President Muhammad Zia-ul-Haq and sent a high-level mission to Pakistan to increase support for the Mujahideen.⁹⁶ Dual delegation was later institutionalized by the Ronald Reagan administration, drawing on the then-established practice of working with President Zia-ul-Haq and the ISI. Secretary of State George Shultz put it bluntly: “We must remember, without Zia's support, the Afghan resistance … is effectively dead.”⁹⁷

The United States and Pakistan had different reasons for wanting to support the Mujahideen. The United States’ primary objective was to make the war as costly as possible for the Soviet Union, with the ultimate aim to force the Soviets to withdraw all military personnel from Afghanistan.⁹⁸ As Brzezinski wrote to Carter, support of the Afghan resistance would help “to keep the Soviets bogged down.”⁹⁹ This goal, maintained by the Reagan administration, was specified in National Security Decision Directives 75¹⁰⁰ and 166.¹⁰¹ In addition, the United States wanted to prevent spillover effects, worrying that increased Soviet influence would further the spread of communism in the region.¹⁰² Pakistani¹⁰³ and Soviet¹⁰⁴ diplomatic accounts underscore this understanding of U.S. strategic goals. Pakistan, on the other hand, wanted to install a friendly regime in Kabul that could suppress Pashtun separatism and limit Indian influence in Afghanistan.¹⁰⁵ Specifically, Pakistan wanted “an outright military victory and the establishment of an Islamic government in Kabul.”¹⁰⁶

This aim was pursued hardest by General Akhtar Abdur Rahman Khan, the chief of the ISI from 1979 to 1987.¹⁰⁷ Pakistan's strategic priorities balanced the threat posed by the Soviet invasion in Afghanistan with the long-standing threat posed by India, which informed the differences in goal preferences between Pakistan and the United States: “Pakistan's perception of the [Soviet] threat is at total variance with the U.S. perception. For Pakistan the prime danger is political disintegration not military invasion undertaken for strategic purposes … [Pakistan's] primary threat … is from India and not from the Soviet Union.”¹⁰⁸ Therefore, the Soviet invasion of Afghanistan was a strategic vulnerability for Pakistan because it could have contributed to a Moscow-Kabul-New Delhi axis that would encircle Pakistan.¹⁰⁹ Pakistan was a dealer, with different strategic objectives than the United States and a will to act independently. As such, it actively controlled how aid was distributed and imposed significant restrictions on the armed resistance that did not favor the long-term strategic objectives of the principal. In an early letter to Zia-ul-Haq, President Carter wrote: “It is of greatest importance that the United States and Pakistan share a common appreciation of the situation and build a basis of trust for dealing with each other.”¹¹⁰ The United States, however, was dependent on Pakistan.

There were three primary reasons for why the United States needed Pakistan as an intermediary. First, deterred by recent memories of the Vietnam War, the United States was keen to limit its own footprint in Afghanistan, while trapping the Soviet Union in its own Vietnam as payback.¹¹¹ Delegating to Pakistan thus became a way for the United States to increase the level of plausible deniability, in addition to carefully selecting types of arms—from China, eastern Europe, and Egypt—that would not be traced back to U.S. involvement.¹¹² Second, with minimal CIA presence in Afghanistan (mostly confined to Kabul), the United States lacked local intelligence and had no direct means of funneling arms and equipment into the country.¹¹³ Pakistan's geographic proximity was, therefore, key to any form of U.S. involvement, and especially given that the other two land routes into Afghanistan were from the Soviet Union and Iran. The United States also relied heavily on Pakistani intelligence for strategic assessments. The ISI became the main source of information about resistance groups’ politics, and the agency was assumed to have a good understanding of the local context.¹¹⁴ Third, the CIA and the ISI had worked together successfully in the past,¹¹⁵ and the ISI had already forged important links to some of the Mujahideen commanders. Thus, there was already an existing intelligence network in Afghanistan that the CIA could readily tap into.¹¹⁶

The U.S. covert aid program came to rely heavily on Pakistan as an intermediary to provide support to the Mujahideen.¹¹⁷ Aware of Pakistan's key role, President Zia-ul-Haq insisted that the ISI would have the exclusive responsibility of channeling funds, arms, and equipment to the Mujahideen. As Pakistani accounts make clear, it was “a cardinal rule of Pakistan's policy that no Americans ever become involved with the distribution of funds or arms once they arrived in the country.”¹¹⁸ The United States largely agreed to this, and the ISI distributed resources without any significant oversight from the CIA.¹¹⁹ This was detrimental to U.S. interests, as the ISI was eager to influence the fighting and hoped to use the aid that it distributed to further Pakistan's own national agenda in Afghanistan.¹²⁰

The Mujahideen was fragmented, consisting of a variety of groups with individual leaders who sometimes cooperated and sometimes fought against one another.¹²¹ At first, Pakistan used its role as a dealer to favor selected clients and to marginalize and exclude some factions, notably Pashtun nationalist groups and movements loyal to the former king, Mohammed Zahir Shah.¹²² Pakistan favored those Mujahideen commanders who shared an ideological connection with the Pakistani leadership and who accepted significant oversight and control by the ISI,¹²³ which filtered the aid to the Afghan Mujahideen and allocated it to those Mujahideen groups that cooperated with and participated in ISI's regional strategy.¹²⁴

The ISI directed most of the support to the more radical-fundamentalist Hezbi Islami faction led by Gulbuddin Hekmatyar. Hekmatyar had little grassroots support inside Afghanistan and was highly dependent on outside support.¹²⁵ Despite Hekmatyar espousing an anti-Western and particularly harsh anti-U.S. rhetoric, it is estimated that roughly half of all resources provided by the CIA went to his faction.¹²⁶ Moreover, Hekmatyar consistently placed the long-term goal of Islamic revolution above resistance to the Soviets or the Afghan regime, with the ultimate aim to ensure that his party became the voice for all Islamic forces. Consequently, he had a reputation for killing more fellow Afghans—both competitors and civilians—than Soviet troops.¹²⁷

Pakistan's selection of the Hezbi Islami faction was not the one that the United States would have made, but the ISI's favoritism for fundamentalist Muslim groups foreclosed support to the moderate and more militarily effective opposition.¹²⁸ The ISI sidelined the Mujahideen factions led by Ahmad Shah Massoud and Abdul Haq that enjoyed more support inside Afghanistan and that had consistently proved to be more militarily efficient.¹²⁹ This was, for instance, evident in the ISI-planned Jalalabad offensive of March–May 1989, during which at least 3,000 guerrillas were killed in an attempt to install ISI's favored client.¹³⁰

Not only did the ISI determine to whom support was distributed, but it also largely dictated policy and directly controlled the guerrilla forces’ training and operations. Bruce Riedel describes the details of the U.S.-Pakistani arrangement: “All training of the insurgents was done by Pakistani soldiers. If the CIA provided equipment that required special training (like the Stinger missiles), CIA personnel trained Pakistanis, who in turn trained the Mujahideen.”¹³¹ Mohammad Yousaf, the head of ISI's Afghan Bureau from 1983 to 1987, described the bureau's total control over the resistance: “This department controls the allocation of arms and ammunition; their distribution to Mujahideen leaders and commanders; the training of Mujahideen in Pakistan; the allocation of funds from the US and Saudi Arabian governments; and the strategic planning of operations inside Afghanistan.”¹³² The ISI planned operations, picked targets, and promised commanders and parties extra weapons for carrying out the operations. Pakistan also sent advisers into Afghanistan to oversee key operations.¹³³ Teams from the Pakistan Army's Special Services Group were clandestinely inserted across the border to direct Afghan guerrilla operations. According to Yousaf, there were at least two teams with soldiers from the Special Services Group at any given time during his tenure—one of which was captured by Afghan authorities in April 1989.¹³⁴ Although a few of the Mujahideen commanders in the field were able to preserve their independence and sustain their own operations, most were constrained by the ISI, which regarded itself as being in charge of the war effort, planning and commanding the actions of the Mujahideen. The ISI also operated training camps and provided logistical support inside Afghanistan. Seven major resistance groups were headquartered in the Pakistani city of Peshawar, and the ISI provided both unrestricted access to the refugee camps for recruitment and almost complete freedom of movement across the border.¹³⁵

Pakistan leveraged its role as a dealer in Afghanistan. The United States viewed Pakistan's cooperation as the sine qua non for opposing Soviet aggression in Afghanistan,¹³⁶ and Pakistan argued that it risked retaliation from the Soviet Union for supporting the Mujahideen. When President Zia-ul-Haq met President Carter in the White House on October 3, 1980, Zia-ul-Haq said: “Pakistan should continue to fight in Afghanistan, but Pakistan must be safe. It is in the interest of both Pakistan and the U.S. to support the freedom fighters.”¹³⁷

From the outset, Pakistan strived to keep most of its compensation from the United States covert to avoid risking further escalation. President Zia-ul-Haq rejected one of the first major aid packages from the United States, calling the $400 million economic and military aid offer “peanuts.”¹³⁸ This was interpreted in Washington as a move to build credibility and to “ratchet up the offer.”¹³⁹ Once Ronald Reagan became president, Pakistan renegotiated not just its compensation but also the entire relationship with the United States.¹⁴⁰ Shortly after the U.S. presidential inauguration, Pakistan received arms worth $3.2 billion from the Reagan administration, which also disregarded its clandestine nuclear program and worked toward significant debt relief. Pakistan also got special deals on advanced military matériel such as F-16 fighter jets. Before President Zia-ul-Haq's visit to the United States in November 1982, a National Intelligence Estimate described the delivery of the first of the F-16 aircrafts as “the most viable symbol” of the new U.S.-Pakistan relationship.¹⁴¹ Importantly, the U.S. administration provided these fighter aircrafts to Pakistan knowing that doing so would likely worsen relations with India. The initial aid package was not only renewed in 1986 but increased to $4.02 billion.¹⁴² That same year, President Reagan began certifying that Pakistan did not possess nuclear devices and that U.S. aid “reduced significantly” the risk that Pakistan would acquire nuclear capabilities.¹⁴³

Pakistan's role as a dealer was significant. Because of a lack of control and oversight from the principal, ISI chose to distribute resources in ways that furthered its own objectives, with little sanctioning from the principal. Pakistan played a key role in the selection of recipients, which led to the bulk of support being channeled to the more radical Mujahideen factions instead of more moderate and militarily efficient groups mainly based inside Afghanistan. This also hindered the emergence of a more unified Afghanistan rebellion, which was in line with Pakistan's strategic interest of achieving a favorable regime in Kabul. During the Afghan war, “the Americans and the Pakistanis were partners, not allies. Their relationship was a marriage of convenience.”¹⁴⁴ Pakistan was effectively diverting U.S. resources, pursuing a “military strategy that contradicted U.S. interests and policies.”¹⁴⁵ In hindsight, Pakistan's involvement also planted the seeds for the emergence of the Taliban regime and al-Qaeda.

The civil war in Angola emerged out of the Angolan War of Independence (1962–1975) as the People's Movement for the Liberation of Angola (MPLA)—backed by the Soviet Union and Cuba—took control of the state apparatus and formed the first Angolan government. This government was contested by UNITA, led by Jonas Savimbi, while a former rebel group, the National Front for the Liberation of Angola (FNLA), was slowly disintegrating. In the run-up to independence, UNITA had received considerable state support, especially from the United States and South Africa.¹⁴⁶ In December 1975, however, the U.S. Congress passed the Clark Amendment prohibiting the use of funds for paramilitary operations and effectively ended the U.S. presence in Angola. The ban remained in place until its repeal a decade later.

With the repeal of the Clark Amendment in 1985, “a debate began to rage: should the United States now provide aid to Savimbi's UNITA?”¹⁴⁷ In supporting UNITA, U.S. goals extended beyond negating Soviet military gains to helping the group overcome the political and military impasse with the MPLA.¹⁴⁸ While intelligence about UNITA praised its fighting capabilities, organization, and leadership, it concluded that an outright military victory was improbable.¹⁴⁹ With this in mind, the United States’ main goal was to “bring enough pressure on the Angolan government to enter into negotiations and for all the parties to find a political solution.”¹⁵⁰ At the same time, the United States’ strategic interests concerned the wider region, and it sought to resolve the Namibian independence issue and the withdrawal of Cuban troops from Angola.

As with Pakistan and the Mujahideen, supporting UNITA required consolidating a partnership with a neighboring state. As Secretary Shultz writes: “The point was that the aid had to be delivered, and to obtain the cooperation of an acceptable neighboring state, delivery had to be deniable.”¹⁵¹ For this, the United States turned to Zaire, which became not only an indispensable partner in sponsoring UNITA but also a cooperative broker whose goal alignment with the United States secured UNITA's compliance as a proxy. Zaire was a key intermediary acting as a transmission point of lethal support to UNITA. “How are you going to get the aid delivered?” Shultz asked in the White House Cabinet Room in November 1985. “Zaire and Zambia cannot openly support insurgents in another African state. And the aid had to go through there! If the aid isn't delivered, it's worthless to Savimbi.”¹⁵² Reagan's diary entries also acknowledge Zaire's core role in mounting “a covert operation for real help.”¹⁵³

Extending the delegation chain to an intermediary had several advantages for the United States. Much like in the Pakistan case, working with and through Zaire provided plausible deniability and local access. This made sponsoring UNITA less risky by expanding the range of support that it could receive and bypassing South Africa, whose involvement would have violated a U.S. embargo on arms shipments.¹⁵⁴ Using Zaire as an intermediary also raised the costs of waging war for the MPLA regime and its Soviet and Cuban backers, while lowering the risks of retribution. In addition, Zaire became a source of intelligence for U.S. policymakers, albeit a less useful one than the Pakistani ISI given that Zaire's intelligence services focused chiefly on maintaining President Mobutu Sese Seko in power.¹⁵⁵ Finally, the United States was able to draw on a long-standing cooperative relationship with Mobutu's Zaire as an intermediary. Since the 1960s, the United States had been providing Zaire with economic and military aid in exchange for security commitments, with Zaire becoming the go-to broker for U.S. military aid to other African insurgents.¹⁵⁶

Zaire's main strategic goal—securing UNITA's access to a power-sharing agreement to prevent further Angolan support for Zairean rebels—overlapped with that of the United States. Additionally, it sought to increase Mobutu's international prestige as a power broker and bolster Mobutu's regional ambitions to secure resources, such as those of Angola's oil-rich Cabinda Province.¹⁵⁷ Taken together, these strategic goals aligned with Reagan's strategy for the region and drew on the country's history of direct and indirect involvement in the conflict. To this end, Zaire repeatedly signaled its commitment and capacity to contribute to the United States’ African policies.¹⁵⁸

Zaire's aims were a function of the country's proximity to the ongoing civil war and its externalities (e.g., economic impact, refugees, wider regional instability). Specifically, the MPLA government posed a direct threat to Mobutu because it had supported Katangese rebels during their two incursions into the Shaba Province in 1977 and 1978. Even though CIA intelligence reports repeatedly estimated a low possibility for a successful third Shaba insurrection, its potential created a permanent existential security threat.¹⁵⁹ Subsequent defense agreements succumbed to issues of credible commitments because they obligated Angola and Zaire to pledges not to retaliate by supporting each other's rebels.¹⁶⁰

While the overlap between the principal's and the intermediary's goals in this case removed the competitive environment that shaped the U.S.-Pakistan goal divergence, Zaire's alignment assured Mobutu of an enhanced bargaining position.¹⁶¹ Zaire's facilitation of lethal aid transportation and commitment of its territory to UNITA operations were integral to the United States’ Angola policy. This made compensation through economic and security benefits, rather than sanctioning, the key to U.S. policymakers’ success in managing the relationship with Zaire. Reagan's 1987 national security directive on Angola set out strategic objectives “acceptable to key African partners whose support [was] essential” and explicitly mentioned the United States’ commitment to Zaire through continued development, security assistance, and humanitarian relief programs.¹⁶²

Yet Zaire found in its broker role an opportunity to seek additional compensation by demanding increases in U.S. assistance. A 1986 National Intelligence Estimate directly links Mobutu's position to “a 21-year record of support to U.S. policies, particularly his current support for UNITA.”¹⁶³ Reagan's deputy assistant for national security affairs, Donald Fortier, described the U.S. relationship as a catch-22: “Our need to cooperate with a partner in this fashion limits what we can do…. We have to be sensitive to his weakness and vulnerabilities.”¹⁶⁴ To a certain degree, the United States actively compensated Zaire, giving credence to its security grievances while acknowledging Mobutu's opportunism. Assistance to Zaire grew from $40 million in 1983 to $79 million in 1986.¹⁶⁵ In fact, an overview of the United States’ Angola policy for Director of Central Intelligence William Casey explained that Zaire was given priority in assistance, “which demonstrated our appreciation of their support and shared interests.”¹⁶⁶ Further, Reagan embarked on sustained efforts to “court” Mobutu¹⁶⁷ while pushing France and Belgium to “heighten” the visibility of their support to Mobutu.¹⁶⁸ Despite some mild posturing from Zaire—expelling ambassadors and adjusting its relationship with Moscow—that the CIA described as “attention getters,”¹⁶⁹ Zaire never undermined or seriously threatened U.S. policy in Angola because “the implications of perceived U.S. support for Zaire are more valuable to Mobutu than the actual assistance he receives.”¹⁷⁰

Therefore, as a broker, Zaire's involvement in the provision of support to the rebels in the Angolan civil war on behalf of the United States created numerous positive payoffs. UNITA funding began in February 1986 following arguments over whether the support should be covert or overt. Specifically, while Congress proposed public nonlethal aid, Secretary Shultz saw “far greater importance of covert and lethal assistance.”¹⁷¹ More importantly, in December 1985, CIA Director Casey traveled to Zaire to set out the parameters under which Mobutu's regime would become a conduit of weapons for UNITA.¹⁷² In many ways, Zaire's brokerage was a continuation of Mobutu's own support of UNITA with other and better means. Before U.S. involvement, Zaire's support to UNITA consisted of access to the country's territory as a sanctuary and an area of operations,¹⁷³ as well as “intelligence support, the transit of personnel and equipment, and access to facilities for rest and training.”¹⁷⁴

Both the United States and Zaire viewed UNITA and its leader Savimbi as the right choice and coordinated to prevent leadership disputes and factional splintering. In fact, after the MPLA defeated the FNLA, Mobutu switched to allocate aid to UNITA and allowed Savimbi to re-group and organize in Zaire, just as the Carter administration's National Security Council stopped “advising friendly countries against aid to Savimbi.”¹⁷⁵ Zaire, acting as a compliant broker, simply followed the principal's wish regarding selection: “After the 1976 Clark Amendment barring further US covert aid in Angola, US intelligence officials were determined to find other ways to continue the campaign against the MPLA. They had been convinced, moreover, that Savimbi was a more viable client than [Holden] Roberto. Despite Mobutu's closer links with Roberto, he was encouraged to provide access in Zaire for Savimbi.”¹⁷⁶ Zaire also used its own contacts with the United States to shape UNITA's strategic narrative, including by frequently allowing journalists to travel from Kinshasa to the remote areas of southeastern Angola.¹⁷⁷

Zaire's ability to direct UNITA's strategic and tactical decisions presents a more nuanced picture: while influencing issues ranging from operations to cessation of hostilities, Zaire's never sought to deviate UNITA from its U.S.-supported goals. Because Mobutu was “favorably disposed toward Jonas Savimbi,” he regularly met with UNITA leadership.¹⁷⁸ Similarly, Zaire effectively used its territory to support the Angolan insurgents. Zaire initially controlled UNITA operations through access to territory.¹⁷⁹ This occurred in close cooperation with the United States, as it was “generally understood that American policy [was] to get UNITA to work from Zaire in northern Angola, in order to allow the United States to control them better.”¹⁸⁰ While decisions not to constrain access prevailed, evidence of hidden UNITA bases in Zaire near Cabinda led Mobutu to expel some members of the UNITA leadership in 1986.¹⁸¹

As a broker for the United States, Zaire used its territory to allocate resources and to allow UNITA to shift between conventional and unconventional attacks. Starting in 1986, the abandoned military base at Kamina in southwest Zaire was the key point for airlifting arms for UNITA into Jamba, Angola, where the CIA “formed a special task force to administer the program.”¹⁸² Beginning in 1987, weekly C-141 flights started carrying Stinger antiaircraft missiles,¹⁸³ whose impact contributed to UNITA's fighting, albeit not as decisively as in Afghanistan.¹⁸⁴ In November 1989, a CIA plane carrying military and other equipment crashed near Jamba, killing U.S. personnel.¹⁸⁵ Resources were allocated jointly with the United States and, according to a UNITA supply officer captured in December 1987, members of UNITA traveled to the Zairean airbase for “training by the Americans in intelligence and in the use of Stinger anti-aircraft missiles and TOW antitank weapons.”¹⁸⁶ Zaire also influenced UNITA's tactical choices by helping Savimbi's rebels overcome logistical problems¹⁸⁷ and by allowing them to expand their areas of operation.¹⁸⁸

Arms support to UNITA through Zaire led Savimbi to escalate the war successfully. Secretary Shultz recalled that “US-supplied antitank and antiaircraft missiles played an important role, and I was glad that covert U.S. lethal aid had gone forward.”¹⁸⁹ With support to UNITA growing to $40 million in 1988, the protracted conflict gave Zaire the chance to broker not just Savimbi's relationship with the United States but an end to the conflict itself. By 1988–1989, the United States had committed to ending the war in Angola and had taken a series of key steps that included the New York agreements on the independence of Namibia and the phased withdrawal of Cuban troops.¹⁹⁰ This diplomatic strategy pressured the United States’ African partners, especially Zaire. Mobutu organized a meeting in Gbadolite intended to reach a cease-fire that would lead to a political solution to Angola's civil war, but Mobutu's mediation abilities were discredited when this cease-fire broke down shortly after. As punishment, Zaire cut off CIA aid to UNITA, and it took President George H. W. Bush to restore Zaire's relationship with UNITA and resume support.¹⁹¹

The intermediary efforts failed because Mobutu's strategy was ill-defined, poorly communicated, and imposed unacceptable demands on both parties. In fact, President Bush told Savimbi that the United States did not see “any feasible alternative to working through Zaire, and through Mobutu.”¹⁹² Moreover, a CIA estimate on Zaire's prospects speculated that removing Mobutu would eliminate “a potential interlocutor for negotiated settlements in the region.”¹⁹³ Zaire's role as a broker demonstrated strategic behavior consistent with our theory. As a closely aligned partner, the broker bought into the principal's goals and willingly served as a conduit to provide support. Zaire therefore merely “augmented” the covert assistance program as an important ally and “the logical conduit for U.S. support to UNITA.”¹⁹⁴ In contrast to the United States’ partnership of convenience with Pakistan, Zaire “proved a natural ally to stem a hostile tide in the region.”¹⁹⁵ More importantly, attempts to manage and influence the agent overlapped with Zaire's own interests in mitigating insurgent prospects and collecting a security dividend in a strong UNITA ready to meet Zaire's security challenges regarding a possible Shaba insurgency.

Zaire and Pakistan were key intermediaries in the United States’ proxy wars against the Soviet Union in Angola and Afghanistan. The intermediaries pursued different goal alignment strategies, which largely determined their levels of involvement in the provision of external support. As a dealer, Pakistan followed a competitive logic. It prioritized enhancing its own bargaining position in the broader, regional context over supporting the United States’ determination to defeat the Soviet Union. On the contrary, Zaire followed a cooperative logic, using its brokerage mandate to contribute to both removing Soviet and Cuban influence in southern Africa and bolstering its own security and regional ambitions. Table 2 provides a summary of our main findings.

In terms of the principal's’ motives for using intermediaries, our analysis suggests that informational advantages (H1a) were somewhat less important than geographic proximity (H1b) and plausible deniability (H1c). Decision-makers in both cases made explicit references to plausible deniability and geographic proximity. Interestingly, our analysis shows that deniability is important not only for letting the principal avoid international or national sanctioning but also for the intermediary's calculus. Pakistan was mainly concerned about retaliation from the Soviet Union, whereas Zaire sought to balance escalation in relation to the Angolan government.

Proximity appears to be key, as both cases indicate that the United States needed a local partner to gain access and to distribute resources. The findings about informational advantages are more mixed. In the case of Zaire, there were no explicit references to this motive—potentially because the intelligence was not as valuable to the United States, which could rely on other local partners such as South Africa and Zambia for local information. In Pakistan, on the other hand, there was stronger support for the United States having used Pakistan to gain informational advantages relating to the Mujahideen, as the United States clearly valued and relied on intelligence from the ISI. In addition to the stipulated hypotheses, our analysis also demonstrates that the principal seemed to prefer basing its selection of intermediaries on prior cooperation.

Zaire's and Pakistan's different alignments with U.S. strategic aims inform two pathways of involvement in the provision of external support. Our analysis presents evidence on Zaire's cooperative “low” involvement and Pakistan's competitive “high” involvement, and how actions either consolidated or subverted U.S. interests. It is clear that intermediaries influence the provision of external support (H2), and that the double principal-agent problem is most acute when the principal and the intermediary have divergent interests (H3). Furthermore, our findings offer the opportunity to think more creatively about intermediaries as being located on a spectrum of involvement concerning their roles in allocating support, screening recipients, and directing operations.

As a dealer, Pakistan was highly involved in selection. It both allocated resources and directed the rebels’ operations. The United States provided little oversight and largely left it to President Zia-ul-Haq and Pakistan's ISI to control the distribution of military assistance. Consequently, Pakistan channeled resources to its own preferred factions of the Mujahideen and directly controlled rebel operations by rewarding behavior that conformed to its own interest and by punishing deviations.

As a broker, Zaire also channeled aid, but it did so by working with the CIA teams to transport, distribute, and coordinate aid. Zaire's involvement overlapped with the United States’ goals to employ UNITA as a proxy and to empower Savimbi rather than to cause factions to splinter or to create other factions, as was the case in Afghanistan. Zaire's ability to direct operations was in line with U.S. positions on UNITA, namely, to translate battlefield successes into political power at the negotiating table.

Finally, Zaire did little to control UNITA the way that Pakistan did with groups like Hezbi Islami. When Zaire cut off aid to UNITA because of the reputational costs that it (the broker) incurred as a failed mediator, the principal brought both the intermediary and the agent to Washington to repair relations. Conversely, Pakistan's involvement presented no real opportunities for control or oversight, highlighting once more the effects of goal divergence/convergence on intermediaries’ ability to shape wars by proxy.

One important difference adds weight to our dealer-broker typological assessment and provides partial evidence for H4—that principals compensate and sanction the intermediary to establish control. Pakistan's divergence of interests required additional efforts from the United States to secure alignment and to consolidate trust (e.g., not sanctioning the ongoing nuclear program, providing advanced fighter jets, and debt relief), even if these did not yield the desired results. This indicates that some intermediaries—that is, dealers—impose high costs, driving a hard bargain over which the principal has little power in the short term. Conversely, Zaire's repeated alignment signaling created credible expectations for cooperation that permitted a confident assessment of Zaire's commitment, even as Mobutu attempted to bargain for increased compensation. Counterintuitively, our findings suggest that more aligned brokers may be less successful than divergent dealers at influencing the compensation mechanisms subtending their intermediary partnerships.¹⁹⁶ H4 only receives partial support, however, because compensation seemed to prevail over sanctioning, with both intermediaries being rewarded with economic and military aid. In addition, the United States appeared to exercise more control over Zaire than Pakistan. It had a more direct CIA presence in Zaire to monitor the intermediary's’ actions, and it brought both the intermediary and the proxy to Washington when mediation plans were unsuccessful.

In sum, our analysis highlights the agency of intermediaries and the extent to which they consolidated or subverted the principal's strategic goals. Exactly how principals deal with the double principal-agent problem and exercise control warrants a study of its own, but this illustrative comparison demonstrates the value of dual delegation and the relevant insights that can be gained by an extension of standard principal-agent theory.

External support to non-state armed groups is usually seen as a direct relationship between a state sponsor and a rebel group. But we have highlighted how powerful states often rely on intermediary states to distribute military aid. Because intermediaries are likely to have their own separate agendas, powerful states often face a double principal-agent problem when providing material support to rebel groups: the difficulties and problems associated with controlling the agent are reflected in the delegation dynamics between principal and intermediary. To adequately capture these dynamics, we have offered an extension of principal-agent theory, what we call “dual delegation.” Our theory includes intermediaries as secondary, subordinate principals that are part of longer chains of delegation, and it shows that the principal may need to also provide control and oversight of the intermediary. Importantly, states that want to channel support through others must consider the convergence/divergence of interests or risk strategic failure.

Our illustrative case studies focus on two types of intermediaries, which we call “dealers” and “brokers.” Not only do intermediaries matter in the provision of external support, but their roles differ depending on goal alignment with the principal. Pakistan—acting as a dealer for U.S. support to the Mujahideen in Afghanistan—was highly involved in the provision of external support and followed a competitive logic. Pakistan helped to select proxies, allocate resources, and direct rebel operations. Zaire on the other hand—acting as a broker for distributing U.S. support to UNITA in Angola—was less involved in the provision of external support. Following more of a cooperative logic, Zaire was less concerned with the selection process, it allocated resources jointly with the United States, and it only partly attempted to direct operations.

Our article invites a larger discussion on intermediaries in conflict research more broadly. The findings related to proxy warfare may inform the emergence of a new research agenda centered on their role and influence on a wide range of conflict processes. The diversity of remaining puzzles surrounding dual delegation underlines its theoretical, methodological, and empirical neglect. Under what conditions do states use intermediaries, and when do they decide to act unilaterally? What are the wider consequences of intermediary involvement for conflict dynamics and the prospects for peace? Do states assume different intermediary roles for different principals? Do states reemploy or change intermediaries because of past experiences? How do intermediaries affect the post-conflict context?

Future research should examine how state sponsors coordinate and jointly channel support—and how this affects rebel groups. So far, such research is limited to how state sponsors might foster unity or encourage fragmentation within armed movements.¹⁹⁷ But our empirical analysis suggests that there are likely to be more consequences connected to external support provision. Priority should also be given to including intermediaries in empirical data, since they are rendered practically invisible in all major datasets on the topic.¹⁹⁸ This would enable a larger systematic analysis on the role of intermediaries in proxy wars.

The article has also presented insights into the issue of control in the delegation of war to non-state armed groups, a key topic the debate has yet to address systematically. There is great scope in assessing the micro-foundations of control in dual delegation because it identifies three interconnected dynamics of control: principal-intermediary, intermediary-agent, and principal-agent. Our empirical analysis points to compensation being the main form of influence employed in the principal-intermediary relationship, but this should be explored further in different settings. Finally, some of our key theoretical insights can easily be expanded by abandoning the canonical focus on states. Because the provision of state support to rebels resembles a complex network of relations more than a direct chain of command from principal to agent, the category of actors capable of assuming the roles of principal, intermediary, and agent should be expanded. Recent literature presents persuasive arguments about the political power of non-state armed actors as principals,¹⁹⁹ and there is evidence that non-state intermediaries such as Hezbollah play a key role in training Syrian and Iraqi militias for Iran.²⁰⁰

Finally, this article's findings have two important policy implications. First, states engaging in counterterrorism need to widen the scope of their analysis beyond sponsors of terrorism and explore the role of all states involved in the process of conflict delegation. A recent overview of U.S. policy proposed locating sponsorship “along different spectrums, such as informal versus formal support and direct aid for violence versus incitement.”²⁰¹ A focus on intermediaries adds nuance to such attempts and presents an opportunity to design more robust sanctioning regimes that consider the complexity of the sponsorship process. Second, there is a need to strengthen the international legal framework in relation to proxy warfare to assign legal responsibility more adequately. That principals could use intermediaries when providing support to non-state armed groups indicates that holding principals accountable for violating the nonintervention principle under international law should be reconsidered.

The authors thank Rory Cormac, Kristine Eck, Quint Hoekstra, Adam Humphreys, Erin K. Jenne, Michel Wyss, and participants at the War Studies Seminar at the Swedish Defence University and the Conflict & Change Seminar at University College London for insightful comments on how to improve this manuscript.