The Blame Game: Militias, civilians, and the states’ accountability-effectiveness tradeoff

Abstract

In civil wars, why do states sometimes outsource violence against civilians to Pro-Government Militias (PGMs), while at other times rely on the official armed forces of the state? Scholars have proposed that PGMs function as sources of plausible deniability for shameful violence. Yet, states that delegate repression to militias sometimes also engineer human rights abuses that are perpetrated by regular counterinsurgent troops. If militias are already present, why not delegate? Moreover, why do militias sometimes appear in states where governments are not subject to institutional accountability? In this dissertation, I examine the variation in the co-occurrence of militias and official state forces in contemporary civil wars and theorize that state leaders' decisions to delegate to militias or regular forces reflects competing strategic and accountability pressures from within the state. The decision to outsource to militias is influenced by these factors, which I theorize include rebel territorial control, rebel capacity, information environments, and the preferences of civilian audiences. I test this theory through cross-conflict analysis, a sub-national quantitative case study in Colombia, and a survey experiment. I argue doing so helps illustrate the extent to which governments respond to civilian demands when engaging in one-sided violence. In Chapter 2, I develop a framework to understand the choice between militias and regular forces, which I call the accountability-effectiveness tradeoff. Drawing from the rationalist literature of civil war violence, I begin with the assumption that states believe selective violence can be a viable way to coerce civilian support. I argue that state leaders expect that uniformed troops can more effectively influence the support of civilians. However, because civilians can easily attribute responsibility to state leaders, the use of regular troops for repression increases the chances of backlash effects among a broader domestic audience. Outsourcing to militias can help mitigate these backlash effects, but introduces command and control problems, muddied signals, and moral hazard. State leaders anticipate that these problems will reduce the effectiveness of coercive violence at shaping civilian loyalty. As such, I argue that the decision of whether or not to outsource reflects competing strategic and accountability costs. I argue that outsourcing to PGMs allows for states to deny responsibility among the broader domestic audience, but introduces risks that can undermine strategic goals. Alternatively, if the value of strategic incentives exceeds costs of accountability, state leaders will prioritize control by keeping repression `in-house.' I test the observable implications of this argument using cross-national, sub-national, and experimental evidence. In Chapter 3, using an original dataset, I capture the degree of coordination between militias and regular state forces across modern civil wars. I argue that decreased coordination reflects strategies of outsourcing, while increased coordination reflects cases where states attempt to increase control over militias. Cross-nationally, I find that an increase in factors associated with accountability costs results in decreased coordination, suggesting that governments seek to distance themselves from militias in exchange for plausible deniability. Of these factors, the most influential is a country's domestic information environment, indicating that civilian monitoring capacity plays a key role in state leaders' delegation decisions. Chapter 4 shows these patterns hold sub-nationally. I examine the composition of one-sided violence at the municipal level in Colombia between 1997-2006. I leverage municipal-level data on radio access to test how information flows shape delegation. I find that when civilians in surrounding areas lack the ability to monitor the site of violence, the composition of violence shifts towards official forces, while an increased monitoring capacity results in an increased share of PGM violence. This effect is amplified when citizens in neighboring municipalities had previously voted for the incumbent, implying that state leaders are especially concerned about backlash effects from their own supporters. Alternatively, information dynamics have no conditional impact on the composition of violence in municipalities where insurgents pose a more credible threat, revealing that state leaders may prioritize strategic gains over plausible deniability. Finally, in Chapter 5, I examine why militias supply plausible deniability at the micro-level using an original survey experiment from Colombia. The experiment compares attributions of responsibility and preferences for punishment across different perpetrator groups. I find that civilians are significantly less likely to attribute PGM violence to the state than violence perpetrated by regular forces. This difference is not moderated by increased information about PGMs or broader information consumption, helping us understand why state leaders prefer to rely on militias in areas of increased monitoring. Taken together, these findings demonstrate that state violence against civilians is shaped not only by strategic considerations, but also by leaders’ sensitivity to domestic accountability. By highlighting how information environments and civilian perceptions structure the use of repression, this dissertation contributes to broader debates on militia deployment, the rationalist signaling logic of political violence, and bottom-up constraints on wartime behavior.