THE PRINCIPAL-AGENT PROBLEM AND PRO-GOVERNMENT MILITIAS: CASES FROM COLOMBIA AND PERU
By Rachel M. Althouse, Lieutenant, U.S. Navy
Editor's Note: Lieutenant Althouse's PhD Dissertation won the FAO Association writing award at the U.S. Naval Postgraduate School. Because of the length of the full dissertation (over 100 pages), we have extracted the Introduction and Concluding Remarks to publish here, without research documentation. To see the full dissertation, please visit www.faoa.org. The Journal is proud to bring you this outstanding scholarship.
Introduction
States (principals) frequently employ pro-government militias (agents) in low intensity conflicts with mixed results. In some cases, principal and agent interests diverge or the principal loses control over its agent, which devolves into an autonomous terrorist, rebel, warlord, or criminal organization. For example, the Rondas Campesinas in Peru is considered a relative success for integration, control, and strategic effect, while the Self- Defense groups of 1980s Colombia is a worst-case scenario of pro-government militias growing out of control and even threatening the stability of the sponsor government. Looking at historical cases of Latin American pro-government militias from Colombia (Self-Defense groups, Convivirs, and Hometown Soldiers) and Peru (Rondas Campesinas), I examined the principal-agent problem in the context of state-sponsored, pro-government militias and answered the following question: How do sponsor states succeed or fail in maintaining positive control and influence over pro-government militias? States fail when they grant too much autonomy and firepower to militias and they succeed when they limit militias’ autonomy by subjecting them to a tailored combination of control mechanisms: monitoring, screening, sanctions, and rewards. Cases from the Colombian and Peruvian civil wars demonstrate that the best way to control militias and employ them in counterinsurgency is by incorporating them as legitimate auxiliaries of the armed forces. Incorporation of a militia into the Armed Forces greatly reduces the principal-agent problem, the associated risk of diverging interests and objectives, and the future pain of demobilization. As the reviewed cases demonstrate, arming militias with restricted use weapons, outsourcing their financing to private benefactors, and granting them too much autonomy is a guaranteed recipe for disaster.
From review of the Colombia and Peru Cases, I determined the following factors have a notable effect on how well a state implements control mechanisms:
State Capacity: Control mechanisms require personnel and resources. Neither Colombia nor Peru were able to successfully apply control mechanisms until they saw improved state capacity.
Opportunities for illicit gain: In a narco environment, control mechanisms will have less overall effect over militias, especially monetary incentives.
Private Benefactors: A state should avoid privatization of militia sponsorship at all costs or it will lose its leverage over the militia and give power over the militia to another non-state actor.
Legislation: Criminalizing militias without suppressing them first will make them more difficult to rein in. Small sanctions will have little to no effect in an informal/illicit relationship.
Offensive vs. Defensive Role: Because civil-defense forces are meant to be defensive and part of a clear-hold COIN strategy, they should only receive training in, and take part in defensive operations.
Weapons: Civil-defense forces need weapons to defend themselves and prevent guerrillas from returning to cleared areas, but the state should place limitations on the type and quantity of weapons it issues.
Pro-government militias (PGMs), also called pro-government armed groups, paramilitary forces, vigilante groups, or self-defense forces, increasingly play a role in low intensity conflicts. Presently, Syria uses them to fight Rebel-Coalition forces and the Islamic State, Mexico employs village self-defense forces to counteract drug cartels, Nigeria uses them to combat Boko Haram, Ukraine used them to counter pro-Russian unrest, and Iraq uses Shia militias to combat the Islamic State. The use of PGMs by both strong and weak governments is not a new phenomenon and the relationship between government sponsors and these groups is often unsteady due to moral hazard (conflicting actions, goals and interests) and adverse selection (asymmetric information).
There are many reasons why governments use PGMs; some of the most common reasons include plausible deniability, force multipliers, and a counterinsurgency isolation strategy. However, more often than not, the government may lose control of these groups, which turn against the government or devolve into terrorist or criminal organizations, or even warlord para-states, interrupting military operations, government reforms, or the peace process. Human rights abuses by these groups can further inflame tensions and push the population away from the government. In one of the most contemporary, ongoing cases, there is evidence that the self-defense militias in Mexico are transforming into criminal organizations in their own right. Recent skirmishes with the government indicate that these groups may take the same trajectory as the Colombian paramilitaries, further complicating Mexico’s conflict with the cartels.
From the Cold War to the drug wars, nearly every country in Central and South America has employed PGMs in areas where security forces were weak and the U.S. provided billions of dollars in aid and an advisory role in a majority of these low intensity conflicts. In fact, both countries covered in this thesis received USAID and International Military Education and Training (IMET) funds. PGMs add a level of risk and uncertainty to U.S. assistance as the principal-agent dilemma not only affects the domestic government, but also extends to the United States. PGMs may engage in unwanted behavior, have conflicting goals, or inadvertently sabotage population-centric counterinsurgency operations. For example, during the Salvadoran Civil War, some self-defense patrols dually acted as rightwing death squads and sabotaged U.S. led COIN efforts by assassinating land reform administrators.
The literature on pro-government militias tends to focus on why states delegate security and violence to PGMs and the impact of these relationships on human rights. I seek to bring more understanding to an aspect of the principal–agent (PA) relationship least approached in the PGM literature: the how factor. Few works focus on how states try to control militias and manage the high risks associated with sponsorship. Carey and Mitchell acknowledge this gap in the literature: “we need to know more about the details of the ‘contract’ between these groups and their respective governments…we need to know [their] lines of communication. Is there a functional equivalent of the chain of command found in regular forces? When are militia members most likely to shirk?” Moreover, the majority of literature deals with cases in the Middle East, Africa, and South Asia. I intend to fill a gap in the literature by analyzing PGMs in Latin America, which is a gold mine for scholars who wish to study how states interact with, and how much they exert control over pro-government militias. PGMs have been a staple of the Latin American political landscape since Spanish colonialism, when territory was divided between caudillo oligarchs, who ran feudal agricultural and ranch estates with the protection of private militias. Thus, reaching into the wealth of case studies from Latin America, I intend to contribute to the small, but growing body of literature on the principal-agent problem and pro-government militias.
Principal-Agent (PA) Problem. The PA problem applies to a wide range of areas from economics, to legislation, to civil military relations. According to the Encyclopedia of Power, the PA problem “represents situations in which [one party] (principal) delegates a task to another [party] (agent), who has specialized knowledge about the task; but, also has objectives that are different from those of the principal.” Underpinning the core challenge of divergent objectives, a root cause of the PA problem is asymmetric information. Some of the most common situations that arise in the PA relationship are as follows: A) Moral Hazard occurs when the principal cannot control or observe the agent’s actions and the agent may take unnecessary risks or act in his own interest, assuming the principal will incur the costs.
B) Adverse Selection occurs when the agent has information unknown to the principal prior to entering the relationship or contract. C) Agent shirking is when “an agent minimizes the effort it exerts on its principal’s behalf.” D) Agency slack is an “independent, undesired action by an agent.”
All of these issues risk culminating into conflicts of interest and the principal must implement appropriate control mechanisms (“mechanism design”) in order to manage these agency costs. According to the principal-agent theory, in order to reduce moral hazard, adverse selection, agency shirking, and agency slack, principals should practice screening and monitoring agents and should provide incentives to agents for not shirking or slacking.
The PA Problem in National Security Studies. Because PA analysis was developed in the purview of economics and management, the main “unit of analysis is the contract.” As legally binding contracts do not often exist in illicit or covert relationships between states and militias, PA analysis runs into a gray area where control mechanisms are harder to implement, the agent’s perceived value of incentives may shift, and thus, agents are more difficult to control than, say, a financial manager. According to Byman
and Kreps, the “motivation that differs most noticeably between licit and illicit forms of delegation is that of plausible deniability,” where principals delegate to agents in order to avoid accountability for an illegal action. Therefore, in the absence of a physical legal contract, I studied interactions between Colombia and its illegal paramilitaries to determine what mechanisms it used to influence and control illicit agents, and I determined that arms and impunity were the mechanisms of choice in a constrained relationship.
Pro-Government Militias (PGMs). Schneckener defines pro-government militias as “irregular combat units that usually act on behalf of, or are at least tolerated by, a given regime.” I will cover several variations of PGMs: paramilitaries, civil-defense forces (CDFs, also known as self-defense patrols), and vigilante groups. Scholars and journalists often use these terms interchangeably and have not yet reached a consensus on typology. PGMs tend to have an amorphous, non-static quality, changing armed group identities for authorized or unauthorized reasons. For example, the Autodefensas Unidas de Colombia (AUC) is, by name, a self-defense force conglomerate, but has been labeled a paramilitary, pro-government militia (PGM), right-wing terrorist group, drug cartel, and a death squad at various stages of its existence. Therefore, for the sake of simplicity, I use the acronym PGM to frame this thesis, but when referring to the groups in the Colombian and Peruvian case chapters, I use the terms self-defense groups and
paramilitaries, keeping in line with prevailing primary sources and security studies literature on those countries.
Most of the literature focuses on why states delegate to PGMs in low intensity conflicts. Identifying the “why factor” is important to gauging the nature of the relationship and its constraints on control. The most frequently cited reasons are as follows: First, PGMs make a cheap force multiplier, especially when states are weak and do not have reach in the greater part of their territory. Second, most scholars assert that states delegate violence to PGMs in order to maintain plausible deniability for human rights abuses and other illegal acts. Third, states may delegate security to PGMs as part of a counterinsurgency strategy, where PGMs—specifically self-defense forces with specialized local knowledge—are useful for intelligence gathering, and isolating and protecting villages from the insurgency. Fourth, PGMs are more flexible forces, as they are less constrained by organizational bureaucracy. Nonetheless, the “why factor” directly impacts the type of relationship between state and militia. For example, a state that delegates human rights abuses to a militia in order to maintain plausible deniability and retain conditional aid from the United States would most likely choose a covert and unofficial relationship.
The literature also considers how the nature of the relationship influences mechanism design. There are several ways to define a relationship between a state and a pro-government militia: covert, overt, semi-official, or informal (Carey et al. use two types: semi-official and informal). The type of relationship between the state and the armed group is important as it denotes the level of public accountability that the state has for the armed group’s actions or the level of deniability it may have if the relationship is illicit and covert. For example, sponsorship of a PGM for intelligence collection and village protection in a counterinsurgency effort would most likely merit an overt relationship, meaning the government has more leeway in how it designs control mechanisms. Moreover, states may have more difficulty applying control mechanisms to covert groups for fear of exposing their involvement in the group’s activities, and due to a higher level of group autonomy.
Literature Review. A small, but growing body of literature exists that analyzes the relationship between states and PGMs using PA analysis. The majority of these works seek to address the question of why, and in what context, states delegate security or violence to PGMs and their functional role in unconventional warfare. Several works also focus on the link between militias and human rights violations as a form of moral hazard. Several authors identify the most common problems that can arise from states delegating to PGMs (moral hazard, adverse selection, conflict of interest, etc.), but do not spend a lot of time exploring the question of how states seek to prevent or manage these problems. Overall, very few works focus on the question of how states delegate to and control their agents (mechanism design) or what external factors impede their ability to control their agents. Nonetheless, the literature broadly suggests that states use the main control mechanisms presented by PA theory: monitoring, screening, rewards, and sanctions. The literature also maintains that autonomy and control have a negative relationship. In my thesis, I argue that governments must apply a tailored combination of these control mechanisms to reduce the PA problem.
Close monitoring is the first commonly suggested control mechanism in the PGM literature. Placing PGMs (specifically CDFs) under military and police leadership, and thus increasing monitoring, is an effective way to maintain control and reduce shirking or agency slack. Semi-official and overt ties to the government also facilitate easier monitoring of agents. Carey et al. draw a correlation between relationship type and control; overt and semi-official relationships correlate with higher monitoring and control, while covert and non-official relationships correlate with lower monitoring and control. Of note, a state’s decision on whether to have a semi-official, overt relationship versus an unofficial, covert relationship may result from external constraints. Conditional aid from a democratic patron such as the United States, or international pressure from an IGO, could influence states to deny involvement with militias, while still using them to achieve their end game in a counterinsurgency.
A fourth suggested hypothesis asserts that states may implement a robust sanction- reward system to control militias, otherwise known as the carrot and stick approach. For example, according to Peic, in the Guatemala Civil War, where 20% of self-defense forces collaborated with the enemy, “infiltration by insurgents and covert defection are ever- present risks of Civil Defense Force deployment.” Since CDFs are formed to protect villages from insurgents, they may be composed of civilians who are not necessarily loyal to either side and therefore will entertain both sides in order to play it safe when either the army or the guerrillas arrive at their doorstep. Peic finds that CDFs that are underpaid and undertrained may be more likely to shirk and give information to the insurgents. Moreover, without economic incentives, militias may seek other sources of income by changing sides or engaging in economic opportunism (drugs, arms dealing, etc.). Connable and Libicki, on the other hand, play with the notion that unpaid, part time militias may be more dependable than paid, full-time militias: “since the paid militia members are primarily motivated by money, they are also more vulnerable to infiltration, bribery, desertion, and defection.”
The problem of shirking by Guatemalan CDF’s also suggests that states should more carefully vet militia leadership. Screening or vetting agents is a commonly cited way that governments can lower risk of adverse selection. Understanding the agent’s priorities (economic, ideological, etc.) can help states determine which incentives or rewards and which sanctions they should choose to apply to their agents. Information gathering also allows principals to ensure that agents do not have conflicting ideological or strategic interests. In some cases, states may also want to choose agents with a similar demographic (social, ethnic or religious background) depending on the nature of the conflict. For example, Stanton observes patterns indicating that militia members are less likely to follow state orders to kill insurgents or noncombatants from their same demographic background.
The most widely accepted and most basic assumption is that autonomy and control have a negative relationship. For example, Peic draws a connection between level of autonomy and frequency in shirking; the more autonomous the PGM, the less control a government has over their actions. Similarly, Stanton reveals that data from all civil wars between 1989 and 2010 show that the level of autonomy and control are inversely related. Carey and Mitchell also consider the autonomy-control hypothesis as key; “the more loosely connected the group is to the state, the more likely it is that PGM members use violence for their own benefit, for example to loot or to settle private disputes.” Overall, the literature concurs that more autonomy reduces state control, and, as a result increases the likelihood of a militia’s goals and interests diverging from those of the state.
Next, state capacity ties in with several hypotheses as weak states with fewer resources may have a difficult time designing and implementing control mechanisms such as vetting, training, monitoring, and a robust sanction-reward system. However, Staniland suggests that capacity may have a limited impact on a state’s ability to control its agents due to factors outside of the state’s control: “The politics of coercive deployment and restraint may be more important than raw state capacity.” Böhmelt and Clayton even argue that state capacity is not crucial to sustaining PGMs. Nonetheless, I determined in my examination of the Colombia and Peru cases that state capacity has a long-term impact, as at least some resources are necessary to implement “coercive deployment and restraint.”
Lastly, some authors assert that states may use no control mechanisms at all. For example, Mitchell, Carey, and Butler identify plausible deniability for accountability avoidance and outsourcing of violence in relationships as the most logical reason for why states would sacrifice part of their monopoly on violence to PGMs. The authors assert that states either “can’t control or won’t control” their agents. “If challenged by other state or non-state actors, they can claim a simple agency problem and lack of control, passing the costs to the ‘out of control’ or ‘bad apple’ agents.” In other words, governments may find that the strategic benefits of yielding control outweigh the costs. The possibility that some governments do not have the capacity or resolve to control their agents may indicate that they are willing to absorb serious risk in entering relationships with these groups.
The PA Problem and other non-state actors. Literature on relationships between states and terror groups, or rebel groups in proxy wars, also yield valuable insight on control mechanisms. For example, Byman and Kreps discuss how states attempt to control sponsored terror groups and how the nature (covert or overt) of the relationship affects their ability to maintain positive control. Similarly, Salehyan examines the relationship between foreign sponsors and rebel groups in proxy wars. The authors list “modifying scope of authority” to the agent, monitoring and reporting behavior, ex-ante information gathering (screening), and a sanction/reward system as the primary mechanisms that states have used to control agents. Each mechanism has a variation depending on the nature of the relationship between principal and agent. Salehyan notes that “principals walk a fine line between empowering rebels enough to ensure that they can impose costs on the target government, but not so much that the rebels do not fear abandonment, nor be able to turn against the patron.” In sum, relationships between states and terrorist, or rebel organizations, are similar to relationships between states and PGMs due to the challenges of the illicit nature to managing the principal-agent problem.
Hypothesis. For my hypothesis, I tested the classic principal-agent theory control mechanisms: screening, monitoring, sanctions and incentives/rewards. These mechanisms are not mutually exclusive and states should use them in a tailored combination at varying degrees. Moreover, the incentives should come from the state, not private actors, like landowners or businesses—I learned that when the Colombian military permitted private funding of militias, it soon lost control thereafter. These mechanisms are also very difficult to implement when a country has low state capacity. As soon as a state is able, it should incorporate a militia into its security apparatus and take away the militia’s autonomy.
Screening: If a state uses screening in order to weed out criminals and determine if the PGM has compatible interests and ideology, it may reduce the chance of adverse selection and future conflicts of interest, thereby increasing its ability to influence the PGM leadership.
Monitoring: If a state monitors its agent, it may reduce the chances of moral hazard, shirking, slack, and asymmetric information; and, the state may more quickly and effectively apply sanctions or rewards when necessary. Moreover, monitoring prevents the militia from becoming too autonomous, especially in hard-to-reach geographical areas.
Sanctions: If the state applies sanctions when the PGM engages in shirking and slacking, one of three outcomes may occur: A) the sanctions could backfire and the PGM could seek other means of support or turn against the state. B) The sanctions could have no effect at all. C) The sanctions will coerce the PGM to submit to the state.
Rewards: If the state uses material incentives or rewards, the PGM may be less likely to seek economic opportunities elsewhere. Conversely, material incentives may be ineffective if the militia discovers a better source of support elsewhere, such as the drug trade or another criminal enterprise. Monetary incentives have little effect in a narco environment.
State capacity has an exogenous effect on control. If a state has limited military and government capacity, and is not present in certain geographic areas to monitor militia activities, or is unable to produce incentives for a period of time, the militias could obtain more autonomy, which could also decrease the state’s future ability to resume applying control mechanisms such as rewards and sanctions.
In examining the above, I took into consideration the relationship type, which is a policy choice that has important implications for control. PGMs in a semi-official relationship with the state generally have more oversight from military or police than their counterparts in informal relationships, and therefore lower autonomy. Part of this has to do with public accountability; if the state has announced their semi-official relationship with the PGM, it may be more motivated to enforce control mechanisms. Moreover, it is difficult for a state to apply sanctions to a group with which it denies involvement for risk of exposing the relationship. The unofficial relationship between Colombia and the illegal paramilitaries is illustrative of this point.
Research Design. I compared case studies from protracted counterinsurgency conflicts in Colombia and Peru which have some of the most prolific examples of state sponsorship of pro-government militias in Latin America. Table 1 shows the similarities and differences between militias in both countries. I built on and updated data from Carey, Mitchell and Lowe’s PGM Database, and gave each PGM a ‘level of control’ number on a scale of one to five, with five being the highest level of control relative to the other cases. This number is an abstract estimate based on the sum of my findings. All militias I examined started as civil defense forces in a classic clear-hold counterinsurgency strategy during the Cold War and diverged in different directions at various points during the conflicts (Table 1). Both states saw marked improvement in state capacity toward the end of their conflicts, which enabled them to incorporate self-defense groups as part of their armed forces.
In a detailed analysis of each conflict, I extracted evidence that the state used control mechanisms (screening, monitoring, sanctions, rewards) from declassified state department memos, declassified intelligence reports, subject country government documents, archived interviews, news reports, periodicals, court investigations, truth commission reports, and human rights reports. I also measured state capacity by GDP, military expenditure, tax revenues (as % of GDP) and level of territorial fragmentation.
I found archived newspapers and declassified U.S. State Department documents through a search (spanning 1964-present for Colombia and 1989-present for Peru) on ProQuest Historical Newspapers, ProQuest National Security Archive, ProQuest Congressional, George Washington University National Security Archive, and the CIA Electronic Reading room. The majority of the Colombian government and truth commission documents and historical newspapers were obtained through Google search. Lastly, I referred to books and some journal articles for historical background. The non-transparent, and sometimes illicit nature of these relationships has meant that primary sources available can be spotty and even biased, particularly on the military side. As such, I am aware that I do not have a complete picture of all transactions and agreements between principal and agent during the protracted conflicts of Colombia and Peru. To fill in some gaps, I drew from quality secondary sources, such as particularly useful books and articles by Phillipe Dufort, Mario Fumerton, Jacobo Grajales, Maiah Jaskoski, and Julie Mazzei. These authors spent valuable time in the field conducting interviews and gathering data from Peruvian and Colombian government archives.
Table 1. A comparison of pro-government militias in Colombia and Peru.
Concluding Remarks. When the United States plays a COIN advisory role or foreign internal defense role for a host nation, it should consider the above conclusions when determining how aid is allocated, and how much funding and resources the host nation requires to control PGMs. More often than not, PGMs play either supporting or center stage roles in civil conflicts, and should be suppressed and, or incorporated before they grow too powerful and autonomous. Governments should also consider the long-term repercussions of using illegal PGMs for plausible deniability. This type of cooperation damages the government narrative and civil society, and may present barriers to war termination should a powerful militia fail or refuse to demobilize properly (the AUC’s post-demobilization transformation into BACRIMs illustrates this point). Coopting or absorbing a militia is a better alternative to a painful demobilization process down the road.
The government should also act as early as possible before militias become too out of hand. The example of Ukraine illustrates this point as well as my point about incorporation: after 2014, Ukraine grappled with some 30 out-of-control PGMs that had proliferated in response to the Crimea Crisis. “The government drew a line: all independent pro-Ukrainian paramilitary groups would either join the official armed forces or face demobilization by any means necessary. The Ukrainian government acted at the right time—that is, when its army was strong and public opinion was on its side.” During the integration process, the government screened members to root out Neo-Nazis and criminals. Men that did not pass background checks were given the alternative of joining a “civilian volunteer corps to help the war effort; these corps assisted police, cleared snow…and even worked on a public radio.” The government took creative approaches to groups that refused incorporation by exploiting internal rifts, and offering fighters military service with full benefits in exchange for deserting their militias, or otherwise face prison. While Ukraine still struggles to suppress splinter groups and other rising far-right vigilante groups (like the “National Militia”), it avoided certain disaster by incorporating most of the militias when it did.
In conclusion, pro-government militias or civil-defense forces are a useful way to extend the state’s security presence into ungoverned spaces, access valuable intelligence, and hold areas cleared by the military in counterinsurgency operations, but they need to be under the close supervision of the military in order to prevent paramilitarization on the scale of the Colombian AUC. The cases of Peru and Colombia demonstrate that a state cannot implement a civil-defense program in a counterinsurgency just by handing over arms to civilians without strict control mechanisms (monitoring, screening, sanctions and rewards) and a short tether. A weak state with a corrupt socio-economic environment, and a myriad of alternative actors that could provide better incentives than the military such as Colombia’s should never organize PGMs, or give weapons to an independently organized PGM without resources to apply strict control measures, and a very short tether.