Illegality does not necessarily breed violence. The relationship between illicit markets and violence depends on institutions of protection. When statesponsored protection rackets form, illicit markets can be peaceful. Conversely, the breakdown of state-sponsored protection rackets, which may result from wellmeaning policy reforms intended to improve law enforcement, can lead to violence. The cases of drug trafficking in contemporary Mexico and Burma show how a focus on the emergence and breakdown of state-sponsored protection rackets helps explain variation in levels of violence both within and across illicit markets.

Violence is commonly viewed as an inherent attribute of the drug trade. Yet, there is dramatic variation in drug violence within countries afflicted by drug trafficking. This article advances a novel framework that explains how the interaction between two critical variables, the cohesion of the state security apparatus, and the competition in the illegal market determines traffickers’ incentives to employ violence. The analysis introduces a generally overlooked dimension of violence, its visibility. Visibility refers to whether traffickers publicly expose their use of violence or claim responsibility for their attacks. Drawing on fieldwork in five cities in Colombia and Mexico (Cali, Medellin, Ciudad Juárez, Culiacán, and Tijuana), 175 interviews, and a new data set on drug violence, I argue that violence becomes visible and frequent when trafficking organizations compete and the state security apparatus is fragmented. By contrast, violence becomes less visible and less frequent when the criminal market is monopolized and the state security apparatus is cohesive.

Over the past decade, drug consumption has increased in Colombia and Mexico, countries traditionally concerned with drug production and trafficking. Governments and observers have associated this growth with spikes in violence. Drawing on drug consumption surveys and fieldwork in four cities, this study argues that contrary to this perception, there is no automatic connection between domestic drug markets and violence. Violence depends on whether large drug-trafficking organizations (DTOs) control low-level street dealers and on whether those DTOs have a market monopoly at the local level. When dealers are independent, violence might be sporadic, but when DTOs control dealers, violence can explode (given competition between DTOs) or implode (if one organization holds a monopoly). Control over dealers provides DTOs not only income but also informants and armed muscle. This article also shows that domestic drug markets are not new, and have grown incrementally in the past two decades.

In 2012, the two major street gangs in El Salvador, Mara Salvatrucha 13 and Barrio 18, struck a truce credited with reducing homicide rates by more than 50% in one year. Although the gang truce held for only 18 months, the significant reduction in homicides puzzled observers that believed youth gangs were unable to coordinate violence, especially considering that previous and similar efforts aimed at striking agreements did not achieve similar results. This article addresses a question posed by the puzzle of the Salvadoran truce’s success: under what conditions do negotiations between or with criminal organizations effectively reduce criminal violence? By comparing truces and criminal pacts in El Salvador and in Medellin, Colombia, we argue that criminal pacts can reduce homicides when (a) they directly involve the state as an administrator of incentives to reduce violence and (b) criminal organizations have achieved organizational cohesion and leadership that facilitate territorial control and strategic dependability. These conditions allow organizations to regulate violence. Our argument highlights the importance of how violence is performed and, more importantly, its visibility, to fully understand criminal behavior within pacts. It also points to potential lessons for countries seeking alternatives to reduce criminal and political violence in Latin America and elsewhere.