DRUGS, VIOLENCE, AND STATE-SPONSORED PROTECTION RACKETS IN MEXICO AND COLOMBIA

Richard Snyder, Angélica Durán Martínez

Profesor de Ciencia Política en la Universidad Brown, Providence, Estados Unidos. richard_snyder@brown.edu

Candidata a PhD en Ciencia Política de la Universidad Brown, Providence, Estados Unidos. angelica_duranmartinez@brown.edu

ABSTRACT

Illegality does not necessarily breed violence. The relationship between illicit markets and violence depends on institutions of protection. When state-sponsored protection rackets form, illicit markets can be peaceful. Conversely, the breakdown of state-sponsored protection rackets, which may result from well-meaning policy reforms intended to reduce corruption and improve law enforcement, can lead to violence. The cases of drug trafficking in contemporary Mexico and Colombia 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.

KEYWORDS
protection rackets - drugs - violence - Mexico - Colombia.

DROGAS, VIOLENCIA Y REDES EXTORSIVAS CON APOYO DEL ESTADO EN MÉXICO Y COLOMBIA

RESUMEN

La ilegalidad no necesariamente engendra violencia. La relación entre mercados ilícitos y violencia depende de la existencia de instituciones de protección. Cuando se forman redes extorsivas con apoyo estatal, los mercados ilícitos pueden ser pacíficos. En cambio, el desplome de estas redes —que puede ser resultado de reformas políticas bienintencionadas planeadas para reducir los niveles de corrupción y para mejorar el cumplimiento de la ley— puede generar violencia. Las dinámicas recientes de trafico de drogas en México y Colombia muestran que un enfoque en la aparición y desplome de redes extorsivas con apoyo estatal ayuda a explicar las variaciones en los niveles de violencia que existen dentro y a través de los mercados ilícitos.

PALABRAS CLAVE
redes extorsivas - drogas - violencia - México - Colombia.

Recibido el 4 de mayo de 2009 y aceptada el 20 de Octubre de 2009.

We thank Peter Andreas, Sukriti Issar, Stephen Kosack, Crystal Linkletter, Sebastián Mazzuca, and Joel Wallman for helpful suggestions on this material.

Angélica Durán Martínez's research in Mexico in 2008 was supported by a Summer Fieldwork Fellowship from the Graduate Program in Development (GPD) at Brown University.

What is the relationship between illegality and violence? According to the conventional wisdom, illegality and violence are directly connected: the less legal an activity, the more likely it is to be associated with violence, whereas the more legal the activity, the lower the likelihood of violence. This view is validated by high-profile illicit and violent markets, such as illegal drugs in Colombia and Afghanistan. Still, licit markets are not necessarily peaceful. As seen in Diagram 1, diamond mining in the West African countries of Sierra Leone and Liberia and emerald mining in Colombia are associated with high levels of violence.¹ Likewise, illicit markets are not necessarily violent. Whereas in the Democratic Republic of Congo, Kenya, and Zimbabwe, wildlife poaching is carried out mainly by armed gangs and military units and produces frequent violence, in Namibia and South Africa poachers employ violence only sporadically (Warchol, Zupan and Clack 2003). Why is the same illicit activity, poaching, associated with frequent violence in some countries but not others? Moreover, the level of violence associated with an illicit activity can vary widely across time in the same country. In Burma (Myanmar), a major expansion of the narcotics industry in the 1990s occurred during a period that also saw a sharp reduction in violence.² And in Mexico, a recent and dramatic upswing in narcotics-related violence was preceded by a long period of relatively low violence. Why are illicit markets associated with low levels of violence in one period and high levels in another?

Together, these examples not only challenge the conventional understanding of the relationship between legality and violence. They also highlight an important limitation of existing research: the lack of a theory that explains variation in levels of violence across different illicit markets and also within the same market over time. This article contributes to building such a theory. Focusing on illicit drugs, we argue that institutions of protection, especially what we call state-sponsored protection rackets, help explain varying levels of violence within and across illicit markets. Where state-sponsored institutions of protection exist, levels of violence will likely be low. Conversely, the breakdown of state-sponsored protection rackets, which may result from well-meaning reforms, intended to strengthen and improve law enforcement, can ironically lead to large increases in violence.

The next section develops a theory of state-sponsored protection rackets by exploring the conditions under which these rackets form and persist as well as their effects on violence. A subsequent section analyzes the cases of drug-tracking in Mexico and Colombia, showing how the theory of state-sponsored protection rackets provides a stronger understanding of variation in the level of violence across and within illicit markets. A concluding section summarizes the argument about the pacifying effects of state-sponsored protection and then raises questions for future research.

A THEORY OF STATE-SPONSORED PROTECTION RACKETS

State-sponsored protection rackets are informal institutions through which public officials refrain from enforcing the law or, alternatively, enforce it selectively against the rivals of a criminal organization, in exchange for a share of the profits generated by the organization. The central role played by public officials differentiates state-sponsored from private rackets, which have been the focus of most research on protection.³ In addition to sharing profits, criminal organizations that join state-sponsored rackets may also be expected to provide information about rivals and comply with certain behavioral expectations, for example, refraining from violence in situations where peace and order are in the interest of state officials, or helping control "public hazards," such as common criminals and drug consumers (Gambetta 1996) or class enemies (Stanley 1996). As Lupsha (1991) notes in his analysis of the reciprocal expectations that underpinned state-sponsored protection rackets in Mexico's illicit drug economy, trackers were expected not only to share profits with officials. They were also obliged to provide information about "dealings, associates and competition, especially about those who sought to traffic without permission. The trafficker was expected to assist the police and the political system by providing grist for the judicial mill as well as public relations materials to give the US drug enforcers. Thus, while a trafficker could gain protection and warning information; the police could gain credit, praise, and promotions; the political system gained campaign monies and control; and the us, statistics, to justify a job well done."

As seen in the analysis below of Mexico, the emergence of state-sponsored protection rackets can have a powerful pacifying effect in illicit markets. Conversely, the breakdown of these rackets can cause an increase in violence. Before turning to the empirical material, we first explore the conditions under which these rackets form and persist.

The State: The Capacity to Enforce

Officials looking to forge protection rackets require the capacity to make a credible commitment to enforce the law: without a credible threat of enforcement, it makes little sense for criminal organization to pay for non-enforcement. If the state lacks the power to enforce the law, illicit actors may prefer to bear the costs of haphazard and weak enforcement to paying of state officials to refrain from enforcement.⁴ Power is, of course, a relational concept, and the power of state officials to render a credible threat of enforcement, thereby making non-enforcement something worth purchasing, depends on the power of illicit economic actors. The stronger the illicit actors, the stronger the state has to be to induce them to participate in state-sponsored protection rackets.

The ability of state officials to construct protection rackets is strengthened when they can credibly commit not only to non-enforcement of the law for their criminal "partners," but also to apply enforcement selectively against the rivals of their partners. Hence, protection in state-sponsored rackets has two faces: on one hand, state officials supply selective non-enforcement, that is, protection from the state itself; on the other hand, they also supply selective enforcement against rivals, that is, protection from competitors. The capacity of state officials to offer predictable and selective non-enforcement and enforcement of the law is a key prerequisite for state-sponsored protection rackets.⁵

Criminal Organizations: The Capacity to Comply

From the standpoint of criminal organizations, their ability to join state-sponsored protection rackets depends crucially on their capacity to offer a credible guarantee to (1) make payments to state officials, and (2) comply with any agreed on behavior, for example, refraining from violence, sharing information, and controlling "public hazards." If the criminal organization is incapable of refraining from violence against rivals, or even its own members, state officials may face pressure generated by unwanted media attention, jittery international investors, foreign powers and multilateral agencies to stem the violence by enforcing the law, thereby making it more difficult for state officials to participate in protection rackets. To behave as reliable partners in state-sponsored protection rackets, criminal organizations thus require a certain level of internal command, control, and coherence. Moreover, they must be able to signal reliably to state actors looking for criminal partners that they are capable of delivering on their promises. At the same time, the capacity to withdraw compliance, for example by cutting of bribes to officials who do not deliver on their side of the bargain, is also a necessary component for stable protection rackets.

Protection in Time and Space

The likelihood that state-sponsored protection rackets will form depends on the time horizons of public officials and the relationship between the spatial organization of law enforcement, or the geography of enforcement, and the spatial organization of illicit markets, or the geography of criminality. The longer the time horizons of state officials and the greater the congruence between the geographies of enforcement and criminality, the easier it is for state-sponsored protection rackets to develop and persist.

The Shadow of the Future: Trust, Reputation, and Information For state-sponsored protection rackets to emerge and endure, the time horizons of public officials need to be long. If officials are constantly rotated or purged, then stable deals with criminal organizations are hard to cut. Long time horizons increase both the reciprocity and credibility of transactions, because the repeated interactions that are possible when officials have long-term appointments are more likely to generate trust and a reputation for compliance. Research on state-business relations shows how long time horizons can increase the probability that the state and legal business associations will have strong collaborative relations (Haggard, Mayfield and Schneider 1997). Likewise, trust and a reputation for compliance can promote durable collaboration between state officials and illicit business organizations.

The length of the time horizons of state officials also affects the durability of state-sponsored protection rackets by influencing the amount of office-based knowledge and information officials can acquire. Such knowledge, which may include the most common routes used for drug trafficking, the likelihood of finding other officials who are looking to profit by offering protection, and even information about potential buyers, can have an important impact on the ability of officials to supply protection to criminal organizations. Hence, the longer the time horizons of officials, the more likely they are to have the information required to make protection work.

Boundary Issues: The Geography of Enforcement and Criminality
States and criminal organizations operate in and across territory, and it is therefore important to consider the territorial dimension of law enforcement and crime when assessing how state action affects levels of violence in illicit markets. Law enforcement can be organized spatially in a variety of ways. For example, in federal political systems, national and sub-national government agencies often have overlapping jurisdiction over the same territory, thereby increasing the number of potential "protectors," that is, officials looking to offer protection, available to criminal organizations. Even in unitary systems, which lack potentially autonomous jurisdictions at the sub-national level, responsibility for enforcing the law and administering justice is often shared across multiple government agencies, and the country may be divided into distinct, territorially-defined jurisdictions of varying sizes.

In federal and unitary systems alike, the size and configuration of the territorially-defined jurisdictions through which the law is enforced and justice dispensed can have an important impact on the likelihood that state-sponsored protection rackets will form. For example, large jurisdictions, such as a single nationwide district as might be found in a unitary system, would likely encompass multiple criminal organizations. This "one protector, many organizations" scenario is depicted in Figure 1a. By contrast, if jurisdictions are small, the territorial scope of a single criminal organization may cross the boundaries of several jurisdictions, resulting in a "many protectors, one organization" situation, as seen in Figure 1b. Depending on the territorial reach of criminal organizations, small or medium-sized enforcement districts may result in a "one protector, one organization" scenario, as illustrated by Figure 1c. Finally, as seen in Figure 1d, redrawing the jurisdictional map, for example, by decentralizing law enforcement, can alter the ratio of protectors to criminal organizations, resulting in a "many protectors, many organizations" situation. These distinct territorially-defned scenarios have contrasting consequences for the stability of state-sponsored protection rackets, and, hence, for the likelihood that illicit markets will produce violence. This can be seen by exploring the preferences of protectors and criminal organizations over these scenarios.

Protectors want to maximize the price of protection.⁶ This can best be achieved by operating as "the only game in town," that is, by becoming a monopoly supplier of protection. Protectors also prefer to maximize the number of criminal organizations in their jurisdiction. This serves to increase their income, because the more criminals they can protect, the higher their cut of the total criminal revenue. Competition among rival organizations for the protector's favor drives up the price of protection, and protecting multiple organizations reduces dependence on any single source of income. Still, protectors face countervailing pressures to restrict the number of protégés. First, as the number of protected organizations increases, so do the monitoring and other transaction costs faced by protectors.

According to a businessman interviewed by Gambetta (1996) in his study of the Sicilian mafia, "It is easier to eat from the plate of three who cover the whole market than from that of thirty-three who cover the same market." As the number of protected organizations grows, it becomes more costly and difficult to police all the transactions to make sure that clients are not engaging in "tax evasion." This can increase the risk that the protector's reputation as a force to be feared and a credible source of protection will be undermined (Gambetta 1996). Moreover, too many organizations can generate excessive competition that dissipates profits and thus reduces the rents available to protectors.

Criminal organizations aim to minimize the price of protection, ideally dispensing with it altogether. They prefer to have a monopoly on illicit business activities, and they prefer multiple protectors, because the competition among protectors drives down the price of protection. Moreover, having multiple protectors allows criminal organizations to reduce their dependence on any single source. However, criminal organizations also face a countervailing drive, anchored in transaction costs, to limit the number of protectors.

The different scenarios in Diagram 2 have distinct implications for violence. In the "one protector, many organizations" situation preferred by protectors, criminal organizations operating in the same jurisdiction are driven to try to eliminate rival organizations in order to lower the price of protection. Coupled with competition over market share, this rivalry may spark violence between organizations.⁷ Still, if the state has the capacity to cut durable deals with the organizations, it can mitigate struggles over market share and thus potentially maintain peace. Conversely, in the "many protectors, one organization" scenario preferred by criminal organizations, violence may occur between protectors as they compete to control the income generated by a single organization.

The "one protector, one organization" scenario is the least likely to generate violence, because there is no competition among either rival organizations or protectors. Moreover, the behavior of each player in the transaction is predictable, provided they have repeated interactions. Predictability, and hence stable protection rackets, is more difficult to achieve in the other scenarios because of the coordination problems posed by having multiple protectors and organizations. When there are multiple protectors, an organization may cut a deal for protection in the jurisdiction of one protector and then have to negotiate a separate deal in a different jurisdiction. This situation is further complicated by the possibility that the willingness of officials to break the law by providing protection may vary across jurisdictions. Under these conditions, not only would a protector be unable to guarantee that the criminal organization would not have to make another payment in the next jurisdiction, it would be unable to guarantee that it would not face prosecution. Likewise, if there are multiple organizations, competition among them for market share and the protector's favor may undermine their ability to make a credible guarantee not to use violence. Although "one protector, one organization" may therefore be the least likely situation to generate violence, it is the first choice of neither protectors nor criminal organizations, but, as seen in Diagram 2, a second-best outcome for both. Hence, it is prone to instability, because protectors and criminal organizations alike have incentives to defect and try to get their first choice by increasing either the number of organizations or protectors.

An interesting objection regarding our specification of the preferences of protectors and criminal organizations merits consideration.⁸ A single criminal organization with monopoly control over the market should be able to generate more profits than multiple competing organizations. If, as we argue, protectors' profits are a function of organizations' profits, should not protectors stand to earn the highest amount in the face of a monopoly and therefore prefer a "one protector, one organization" over a "one protector, many organizations" situation?⁹

Although overall profits in the criminal sector may indeed be highest when there is a single organization that enjoys monopoly control and prices, protectors will not necessarily receive more income, because their capacity to drive up the price of protection depends crucially on the presence of rival organizations. In the absence of rival organizations, the credibility of the protector's threat to enforce the law and drive the criminal monopolist out of business is weak, because there are no alternative sources of protection income. This inability to render a credible threat of enforcement, in turn, lowers the price of protection. Protectors thus prefer multiple organizations to one. This can be seen by exploring the price of protection in two distinct scenarios: monopoly and duopoly.

Consider first a monopoly. The price of protection will be determined by the amount of income, E, the protector expects to earn by enforcing the law and shutting down the criminal organization.¹⁰ E is likely to be a "single-shot" reward in the form of praise and a promotion. Because collecting E is the only option the protector has to increase its income, the criminal organization knows it can purchase non-enforcement of the law by paying any amount above E. Hence, under monopoly conditions, the price of protection will be E + 1.¹¹

Now, consider a duopoly. Here, the price of protection depends not just on the rewards protectors can reap by enforcing the law (E), but also on the amount, P, that a criminal organization is willing to pay to outbid its rival and stay in business. P is a function of how much income the organization would expect to lose if it were forced out of business. It is also a function of how much the rival organization is willing to pay to stay in business (i.e., its estimate of how much income it would lose if it were forced out of business). This amount is likely to far exceed the value (E) of the raise or promotion protectors can earn by enforcing the law. Hence, even though a monopoly will likely generate the most criminal profits and thus the biggest "pie," protectors resist monopolies because the size of their slice of the pie is larger when there is more than one organization.¹²

An intriguing corollary to our argument concerns the exclusivity of protection. The value of exclusive protection to criminal organizations should be greater than the value of shared protection. Hence, organizations ought to be willing to pay a premium price, P*, for exclusive rights to protection. Yet the logic of our argument suggests it is against the interests of protectors to supply exclusive protection. Even though an organization may offer a premium price (P*) for exclusive protection, it cannot credibly guarantee that it will not later renege by lowering its payment from P* to E + 1 after the protector forces all its rivals out of business, at which point E, the profits that can be reaped by closing down the last criminal organization, would be the protector's only alternative source of income. Because organizations cannot make this guarantee, it is not in the interest of a protector to use its power of enforcement to produce a monopoly.¹³ Criminal organizations seeking monopoly control will thus have to create and enforce their own monopoly not only against the interests of competitors, but also against the anti-monopolistic impulses of the protector. Moreover, if competition among rival criminal organizations turns violent and threatens to destroy organizations, resulting either in a monopoly, or, from the protector's standpoint, the even worse outcome of a criminal extinction, where there are no surviving organizations from which to extract any protection rents, the protector will face strong incentives to intervene to try to reduce violence to "non-lethal" levels that do not eliminate organizations. To avoid a monopoly or organizational extinction, and hence a large drop in the price of protection, the protector could even be driven to play a more robust pacifying role by brokering and enforcing peace agreements among organizations.¹⁴

In sum, depending on the capacity of state officials to enforce, criminal organizations to comply, and the temporal and spatial factors considered, state-sponsored protection rackets may form. When these institutions persist, they can have a strong pacifying effect that lowers the level of violence in illicit markets.

STATE-SPONSORED PROTECTION RACKETS IN MEXICO AND COLOMBIA

To show how the theory of state-sponsored protection rackets helps us better understand why levels of violence vary widely across and within illicit markets, we analyze two cases, drug trafficking in Mexico and Colombia. In Mexico, a state-sponsored protection racket formed during the 1940s and endured until the late 1980s, resulting in relatively low levels of violence during this period. The breakdown of these institutions of protection in the 1990s, partly as a consequence of administrative reforms aimed at reducing corruption among state officials, led to a sharp rise in violence. Conversely, in Colombia a combination of political decentralization, weak control by the central state over the national territory, and especially the pressures generated by competitive electoral politics impeded the consolidation of a stable state-sponsored protection racket. The absence of a durable protection racket helps explain the high levels of violence associated with drug trafficking in Colombia during the 1980s and 1990s. These cases thus demonstrate how a focus on state-sponsored protection rackets helps explain variation in the levels of violence across illicit markets.

The Breakdown of Protection in Mexico: From Stable Pacts to Strategic Violence

Until the late 1980s, the Mexican state was able to elicit relatively peaceful behavior from drug traffickers. The monopoly of power held by a hegemonic political party, the Institutional Revolutionary Party (pri), and the centralization of enforcement, despite the federal character of the political system, provided the capacity the state required to deliver a credible threat of enforcement. More-over, trafficking organizations had enough internal command and coherence to be reliable partners in the state-sponsored protection racket. The result was a "one protector, many organizations" situation.¹⁵ Beginning in the late 1980s, the protection racket in Mexico weakened and eventually broke down because (1) political competition increased, severely hampering the PRI's capacity to control the enforcement and non-enforcement of the law, (2) ongoing reforms intended to reduce corruption within the Attorney General's Office (pgr) transformed the geography of enforcement, altering the ratio of protectors to organizations, and significantly shortening the time horizons of public officials, and (3) an influx of Colombian cocaine traffickers coupled with changes inside Mexican criminal organizations made the task of coordination among protectors and organizations more difficult. Together, these factors caused a breakdown of the state-sponsored protection racket, resulting in an escalation of drug-related violence.

The origins of drug trafficking in Mexico date to the late eighteenth century. As prohibitions tightened in the 1920s and the focus of enforcement shifted from public health to public security, the illicit market became increasingly linked to the political system under the control of the ruling pri (Astorga 2004, Florez 2005, Serrano 2007). Traffickers cut a wide range of deals with the state that included the purchase of licenses to operate from local politicians and police, active participation in trafficking ventures by government agencies, such as the Federal Security Directorate (dfs) and the Federal Judicial Police (pjf), and protection of trafficking by high-level officials (Astorga 2005, Lupsha 1991).

The PRI's monopoly facilitated the protection racket because the party's hierarchical control extended across all the sub-national political units that composed Mexico's federal system. The unchallenged dominance of the pri since the end of the Mexican Revolution provided the stability necessary to guarantee predictable enforcement and, crucially, non-enforcement of the law. As a result, stable pacts between traffickers and corrupt state officials proliferated. Violence occurred during the period of pri hegemony, but it was mostly the result of retaliation by traffickers against competitors, and it never reached the levels seen in other illicit drug markets, such as Colombia's. This relatively peaceful situation changed, however, in the mid-1980s, when the pri started to lose its monopoly of political power and a series of reforms to the pgr undermined both the spatial and temporal stability required for a protection racket.

Starting in 1989, when the National Action Party (pan) won the governorship of the state of Baja California, heightened political competition led to a growing number of sub-national political units governed by opposition parties.¹⁶ The resulting heterogeneous political landscape increased the number of potential protectors available to traffickers and undercut the coordination that the shared allegiances of public officials to a single party had previously made possible. The fragmentation of the pri's control, in turn, made it more difficult for the central state to provide the credible guarantees of enforcement and non-enforcement needed to sustain the old protection racket.

Alongside these political transformations, the pgr initiated a series of reforms in the mid-1980s that also made it harder to sustain state-sponsored protection by shortening the time horizons of state officials in three key ways: first, the reforms unleashed an ongoing process of creation and elimination of ofces¹⁷; second, they mandated relocation of personnel; and, third, they resulted in mass fringes of corrupt officials, as entire units, like the dfs in 1985, were shut down. The reforms of the pgr increased the rotation of police offers and civilian officials. Since the term of President Carlos Salinas de Gortari (1988-94), rotation at the highest level of command intensifed, because Attorneys General began a new practice of appointing fresh teams of collaborators when they entered office. Together, purges, turnover and rotation reduced the time horizons of public officials, who were increasingly uncertain about both the length and location of their appointments. The capacity of public officials to acquire the information necessary to provide protection to criminals was thus curtailed. Moreover, the prospect of rapid removal led corrupt officials to extract as much as possible from single transactions, thereby privileging ad hoc extortion over building institutions of protection (Resa Nestares 2003).

The reforms of the pgr also changed the geography of enforcement in ways that made it more difficult to sustain state-sponsored protection. Until 1996, the pgr's enforcement scheme was based on (1) central control by federal authorities, (2) a functional, rather than territorial, division of power across sub-agencies, and (3) a division of the pgr's fled offices into three contiguous territorial zones (North, Center and South). Under this scheme, the pgr's field offices as well as sub-national government had little power compared to the agency's central command. This centralized framework allowed the pgr's officials to offer credible and stable protection to trafficking organizations across Mexico (Resa Nestares 2003).

The centralized scheme was replaced in 1996 with the creation of three new Assistant Attorneys General for Prosecution (subprocuradurias penales). As seen in the map of Mexico in Figure 2, each Assistant Attorney General's office controlled a set of non-contiguous states.¹⁸ This new non-contiguous scheme increased the responsibilities and autonomy of sub-national governments and pgr field offices. From the standpoint of criminal organizations, this meant an increase in the number of actors who needed to be bribed, and it became far more difficult to determine whom to bribe in order to guarantee the transit of drug cargoes across the country. Thus, the old scenario of one enforcer, anchored in the hegemony of the pri and the centralization of the pgr, was supplanted by a scenario of many enforcers, that is, the pgr's field offices, the state governments, and the pgr's central ofce.¹⁹ This new, territorially-fragmented enforcement scheme was incongruent with the spatial organization of drug trafficking enterprises, the so-called Tijuana, Sinaloa, Gulf and Juarez cartels, which operated mainly along the lines of the old North, Center, and South divisions.²⁰

The heightened uncertainty generated by this new enforcement framework gave criminals strong incentives to acquire their own means of protection. Indeed, the first paramilitary group created by the Gulf Cartel dates to 1997, the year after the pgr's centralized scheme was reformed. The actions of these paramilitary groups likely contributed to higher levels of drug-related violence. Moreover, the uncertainty caused by the transformation of the geography of enforcement led to conflicts among criminal organizations over market share. In the absence of a state-sponsored protection racket, violence increased because criminal organizations needed both to scare away law enforcement and to defend their markets from encroachment by competitors. Instead of being an episodic response by traffickers to failed transactions, violence thus became the dominant strategy of survival.

The feasibility of state-sponsored protection rackets was also undercut by the appearance of Colombian cocaine traffickers. In the early 1980s, the us Government intensified its pressures on drug trafficking routes running from Colombia through the Caribbean to South Florida. As a result, cocaine flows shifted toward inland routes running through Mexico, and Colombian traffickers became increasingly reliant on Mexican smugglers (Andreas 2000, 45-53). The entrance of Colombian traffickers destabilized the state-sponsored protection racket in four ways. First, it multiplied the number of organizations, thus making it harder to achieve coordination among protectors and organizations. Second, it introduced foreign players who not only lacked the local knowledge and networks necessary to participate in the protection racket, but were also allegedly more violent than their Mexican counterparts. Third, it generated new conflicts among Mexican traffickers over whether or not to collaborate with the foreigners. Finally, the higher proftability of cocaine compared to marijuana significantly raised the stakes of drug trafficking and produced more sophisticated and powerful criminal organizations (Andreas 2000).

Key changes inside trafficking organizations as a result of massive extraditions and imprisonments of their leaders by the Mexican government further destabilized the protection racket by making it harder for the organizations to behave as reliable partners of corrupt state ofcials.²¹ For example, the capture of Benjamin Arellano Felix from the Tijuana Cartel in 2002 and the extradition to the United States of Osiel Cárdenas Guillén of the Gulf Cartel in 2007, shifted lines of command and reduced the internal coherence of their organizations.²² These internal changes, in turn, weakened the ability of the organizations to send a credible signal to officials looking for criminal partners that they could be trusted to deliver on their promises.

If our argument about the pacifying effects of state-sponsored protection in Mexico is correct, then we should observe an increase in levels of drug-related violence over the course of the 1990s, as the pri's political monopoly eroded, and especially after 1996, when the pgr's enforcement scheme was decentralized. Moreover, this increase in drug-related violence should occur in those sub-national units where the reform of the pgr resulted in new jurisdictional boundaries that cut across territory controlled by multiple trafficking organizations. Unfortunately, existing data on violence in Mexico do not provide a firm basis for testing the argument, because disaggregated statistics on drug-related violence are available neither at the national or sub-national level during the period in question. Although data on overall homicide rates do exist for the 1990s, showing a decrease in violence during this period, valid inferences about trends in drug-related violence are difficult to draw, because drug-related slayings do not represent a stable proportion of total homicides across time or sub-national units. For example, as seen in Table 1, the proportion of homicides linked to illicit drugs in the two years for which these data are, in fact, available (i.e., 2006 and 2007) shifted from 20% to 25% over the course of just one year. Moreover, the proportion of homicides related to drugs ranges from 0% to 85% across Mexican states.

Despite these limitations, the available data do allow us to test one key aspect of our argument about the effects of state-sponsored protection. We hypothesize that the pri's monopoly of power in conjunction with the centralized scheme of the pgr resulted in coordinated law enforcement across sub-national units. If this hypothesis is correct, then prior to the 1990s we should observe similar levels of violence across neighboring sub-national units as a consequence of this centralized coordination of enforcement. Over the course of the 1990s, as the pri lost power, the pgr shifted to a decentralized enforcement scheme, and the state-sponsored protection racket fragmented, we should see less convergence in levels of violence across neighboring sub-national units. Using spatial analysis techniques we find evidence of just such a trend.²³ Between 1981 and 1996, a clear pattern of spatial clustering of violence exists, with neighboring states having similar homicide rates.²⁴ In 1997, this pattern of convergent levels of violence across neighboring states disappears, as confirmed by the lack of statistical significance of the indicators of spatial clustering between 1997 and 2003.²⁵ Moreover, in the 1990s in regions where drug trafficking organizations are concentrated, highly violent states start to appear next to states that are far less violent. For example, until 1989, Baja California Norte and Baja California Sur, two adjacent states controlled by the same trafficking organization, the Tijuana Cartel, had similar homicide levels but then diverged sharply in the 1990s. Homicide rates also diverged in the neighboring states of San Luis Potosi and Tamaulipas, controlled by the Gulf Cartel, and in Sinaloa and Nayarit, controlled by the Sinaloa Cartel. This pattern of divergent levels of violence across neighboring jurisdictions controlled by the same trafficking organizations is the type of outcome we would expect from the fragmentation of the old protection racket. Still, because disaggregated statistics on drug-related violence for the 1990s are not available, we do not know whether this pattern of diverging homicide levels across neighboring states also pertains to drug-related murders. Hence, these results should be taken at most as tentative evidence of the effects of state-sponsored protection.

In sum, we argue that democratization and anti-corruption reforms in Mexico shortened the time horizons of public officials and altered the geography of enforcement in ways that not only undercut the state-sponsored protection racket forged under pri hegemony, but also made it difficult to replace the old centralized scheme with a new one. Violence thus supplanted state-sponsored protection as the main survival strategy of drug traffickers.

Unstable Protection in Colombia: Competitive Politics, Drugs, and Violence

In Colombia, complex geography, the territorially uneven reach of the central state and, especially, competitive electoral politics have prevented the emergence of a stable state-sponsored protection racket. The absence of stable state-sponsored protection, in turn, helps explain the very high levels of violence in the illicit drugs market. Colombia is a unitary state, but since the late 19^th century it has been characterized by the extreme influence of regional political leaders. Administrative and political practices were heavily centralized until the decentralization process started in 1986; these reforms shifted the distribution of power between the central government and the regions, favoring the latter (Falleti 2005). Governors and mayors started to be elected democratically precisely as the drug business was burgeoning. Thus, as in Mexico a decade later, political decentralization shortened and made less predictable the time horizons of public officials at the same time that it reduced the congruence between the geographies of enforcement and criminality, hindering the emergence of a protection racket. The resulting scenario was thus one of "many protectors, many organizations" and highly prone to violence. By the end of the 1990s, however, the drug trade in Colombia shifted toward a less violent scenario because of (1) the consolidation of paramilitary groups as the main players in the drug trade, and (2) the ability of these groups to gain control of governmental power across vast areas of the country. These circumstances contributed to the emergence of a new configuration approximating a "one protector, one organization" ²⁶ scenario whereby paramilitaries obtained political power and control over regional governments and resources in exchange for eliminating the electoral competitors of politicians and confronting guerrilla groups. Yet, these arrangements rooted in controlling political competition proved unstable, and therefore did not allow the consolidation of a state-sponsored protection racket.

Several authors have asserted that in Colombia, unlike Mexico, the state has lacked the capacity to centralize and control its relations with drug traffickers (Duncan 2005, Resa Nestares 2001), thus making traffickers less dependent on the political establishment (Flores 2005). According to this argument, in Colombia, unlike Mexico, the drug trade has not developed close linkages to national political elites because of the inability of the state to extend its power across vast areas of the Colombian territory coupled with the strength of local elites. Still, it is implausible that Colombian traffickers were fully independent of national-level officials, as credible rumors of elected politicians receiving money from drug traffickers go as far back as 1978, when the us Government questioned the credentials of some of President Julio Cesar Turbay's collaborators for their links to drug traffickers (Toumi 2002). Rather, the crucial difference between the two countries concerns the fact that protection rackets in Colombia have proven less predictable and durable than in Mexico under the pri because they operate in a more competitive and decentralized political system. In turn, the kind of pacifying effect that protection rackets produced in illicit markets in Mexico under the pri has been weaker in Colombia.

Traffickers challenge the state: The rise and fall of "narco-terrorism"
The origins of drug trafficking in Colombia can be traced to the 1940s when marijuana fields started to appear in the north of the country. By the early 1970s marijuana trafficking groups had consolidated in the Guajira region, sparking what became known as the "marijuana boom" (la bonanza marimbera). By the early 1980s the marijuana boom receded as a result of the us eradication campaign, a reduction of the price, and the increasing preference of consumers for the seedless variety of marijuana grown in Mexico and starting to be cultivated in California (Camacho and López 2001). As marijuana production and trafficking receded, Colombians grew increasingly important as intermediaries between Andean producers and Cuban traffickers of cocaine. By the late 1970s Colombian cocaine traffickers started to separate from Cubans, and the names of prominent traffickers became more visible. In 1976, Pablo Escobar was arrested for the first time for drug trafficking. By 1978 Carlos Lehder had consolidated a network of cocaine trafficking both with us and Colombian citizens, and by the early 1980s two organizations, based in the cities of Medellin and Cali, controlled most cocaine exports (Camacho and López 2001). Since the early 1980s the Medellin traffickers began to employ increasingly violent methods. Tis violence was the result of three interrelated factors: (1) the inability of Medellin traffickers to successfully penetrate the political establishment, (2) the government's decision to confront traffickers by approving an extradition treaty with the United States, and (3) the relative centralization and internal coherence of Medellin traffickers under the leadership of Pablo Escobar. The inability of the Medellin traffickers to penetrate the political establishment became evident in 1982, when Escobar was elected to represent Medellin in the Lower Chamber of Congress. Escobar's election generated a strong negative reaction among a wide range of political elites, who opposed the public presence of a trafficker in Congress and successfully pushed for Escobar's loss of political immunity and expulsion from Congress in 1983 (Camacho and López 2001). These events motivated Escobar to react violently against political "oligarchs", who in turn publicly declared war on traffickers by approving an extradition treaty with the us that included narcotics offences. As a reaction to this policy, traffickers led by Escobar created the group "Los Extraditables," responsible for initiating the period of "narco-terrorism" by engaging in strategic violence against the state, targeting high level politicians and carrying out terrorist attacks against the civilian population in an effort to push the government to refrain from making extradition effective. The first sign of Escobar's war against the state was the assassination of the Minister of Justice Rodrigo Lara Bonilla in 1984, and, as of 1990, the violence of Medellin traffickers had claimed the lives of some 500 police offers in Medellin, hundreds of civilians in terrorist attacks in Bogota and Medellin, and prominent politicians, including presidential candidate Luis Carlos Galán.

Some authors have argued that high violence was the result of Escobar's excessive political ambition (Camacho and López 2001), which made traffickers unnecessarily visible. Indeed, running for office, which was likely the result of Escobar's megalomaniac personality, proved a strategic blunder. Escobar underestimated the high costs that politicians in Medellin would bear in order to force him out of office. Furthermore, although Escobar's electoral fate illustrates the difficulty that Colombian traffickers faced in penetrating and making stable connections with the political establishment, the event does not by itself explain the highly violent methods of Escobar's organization, which preceded his brief transit through Congress and persisted well after it. The relative centralization and coherence of Escobar's organization were also crucial factors that help explain its employment of highly organized and brutally violent methods.

By the mid 1980s, Medellin traffickers had consolidated a security structure by professionalizing and controlling about 300 previously fragmented groups of common criminals²⁷ in Medellin and by introducing specialized armed structures for protection²⁸. In 1982 members of the "Ochoa clan" in Medellin created the group mas (Death to Kidnappers), the first paramilitary organization funded by drug traffickers. The emergence of these groups illustrates the traffickers' ability to organize coherent structures that deployed violence not only as an instrument for retaliation, but also as a powerful tool for intimidation. For example, the young hit men (sicarios) paid by traffickers deployed fairly sophisticated methods, such as the "paseo", the practice of kidnapping enemies for a short period of time, torturing and killing them, then leaving the bodies with explicit messages.²⁹ Recent research has questioned the dominant notion that Colombian trafficking groups during the 80s were highly centralized and organized, pointing to the networked and fragmented character of trafficking operations (Kenney 2007). Still, without some internal command and discipline the trafficking organizations dominant during the 1980s would have not been able to carry out the terrorist attacks and the organized violence which characterized narco-terrorism.

The period of narco-terrorism partially receded when, as a response to the extreme violence besieging the country, extradition was banned from a new Constitution approved in 1991 during the government of President César Gaviria. Within the framework of a policy of voluntary surrender to justice ("sometimiento a la justicia") Pablo Escobar handed himself over to the authorities in June 1991 on the condition of not being extradited. Tus, Escobar's extreme tactics and intimidation fnally seemed to pay by preventing his extradition. Escobar was able to obtain highly favorable terms for his detention, but as public opposition towards his luxurious living conditions in prison increased, the government was obliged to announce his prison transfer, and Escobar escaped in July 1992. The government deployed a vast amount of military and police forces in order to defeat Escobar who was finally killed in 1993. During 1992 and 1993 violence increased although this time mainly as a result of the war between the Medellin traffickers and their Cali counterparts, who helped to dismantle Escobar's organization, rather than as a result of narco-terrorism. After Escobar's death, homicide rates in Medellin started to decline and terrorist attacks by trafficking groups were over.

Failed collusion between the state and illicit business organizations
While Pablo Escobar engaged in an open confrontation with the state, the Rodríguez Orejuela brothers from the Cali trafficking group pursued a different strategy, co-opting rather than confronting the state (Camacho and López 2001, Duncan 2005, Skaperdas 2001, Toumi 2002). The situation in Cali resembled a state-sponsored protection racket, where traffickers obtained enforcement prerogatives (lighter sentences) and legislative advantages in exchange for providing regular payments to politicians and police officers and helping them prosecute their Medellin rivals.

Documents from the "Proceso 8000," a judicial process that started in 1995 when former Presidential candidate, Andrés Pastrana, reported that the campaign of the incumbent president, Ernesto Samper, had received money from the Cali traffickers, show how this group built a protection agreement with the state by funding campaigns and creating a payroll system which employed more than 2000 people, including politicians and members of law enforcement institutions (Castillo 1996, Flores 2005, Torres and Sarmiento 1998). The agreements brought substantial benefits for the traffickers. For example, in 1984 Gilberto Rodríguez Orejuela was detained in Spain, and after two years in prison he was returned to Colombia by the Spanish government given the lack of evidence. Upon his return Rodríguez was quickly released by a Colombian judge and magistrates who were later investigated for this sudden decision. Trough their influence on several legislators the Cali traffickers where also able to shape key pieces of legislation, such as those concerning the properties seized from drug traffickers. The collusive arrangement between the traffickers and the state was strengthened as they collaborated in the government's efforts to capture Escobar, especially after his prison escape in 1992. In that year, the Cali traffickers, together with paramilitary leaders who had previously collaborated with Escobar, created the "Pepes" ("People Persecuted by Pablo Escobar"). This group contributed to Escobar's capture and death at the hands of the Colombian police in 1993 by eliminating his partners, destroying his property (Camacho and López 2001) and providing information to the government. The fact that homicide rates, although extremely high, were significantly lower in Cali than in Medellin in the period 1984-1991³⁰, and that the violent methods employed by Cali traffickers were not targeted against the state but against their Medellin rivals, provides some evidence of the pacifying effect of this protection racket.

The protection racket established between the Cali traffickers and some sectors of the state broke down as the Process 8000 scandal forced the government to adopt a tough stance in response both to public opinion and pressure from the United States. In 1995, the principals of the Cali trafficking organization were jailed, and although initially given light sentences, they were eventually extradited to the us in 2004 and 2005. This process underscores the point that although protection rackets may form in situations where politics is highly competitive, these rackets are prone to instability because of the incentives for malfeasance generated by competitive elections. Scandals may easily emerge as competitors look for an electoral advantage, creating pressures for "crack downs" and shifts in public opinion that destabilize the racket. By contrast, in Mexico under the pri, where political competition was far more limited, there were weak incentives to publicize or manufacture scandals as a way to gain electoral advantage. This, in turn, helped make deals between corrupt government officials and drug traffickers more predictable and stable.

By the mid 1990s the main trafficking organizations in Colombia had been dismantled, and, according to several analysts, replaced by smaller, more fragmented and less visible organizations (Camacho and López 2001). Although there is little information and research on the specific configuration of the trafficking market during the second half of the 1990s, it likely approximated a "many protectors, many organizations" scenario with more fragmented and less internally coherent organizations. The lack of internal coherence of organizations made them less able to enter protection agreements and decreased their fire power, thus making them less violent than their predecessors. Yet, by the late 1990s and early 2000s another major transformation in drug trafficking took place in Colombia: the consolidation of paramilitary groups as the main actors controlling the production and distribution of drugs. These groups consolidated their political control over vast regions of the country by making deals with politicians that included contributing to their electoral success by eliminating competitors or coercing populations to vote, and helping them control guerilla groups. These paramilitary groups, like their Cali predecessors, were able to create a protection racket with state officials that probably made it easier to extend their domination over the drug trade without resorting to violence.³¹

Information unveiled during the process of demobilization of paramilitary groups, which started in 2002, and evidence obtained from a computer seized from the paramilitary leader Jorge 40 in 2006, show how paramilitaries became crucial players in processing and distributing drugs, the stages of the drug trade most prone to oligopolistic control (Duncan 2005). Paramilitary leaders reportedly established arrangements for smuggling drugs to Europe with the collaboration of the antinarcotics police in the Atlantic Coast of Colombia.³² Since 2001, prominent paramilitary leaders, such as the now-extradited Salvatore Mancuso, cut deals with the Italian Mafia for the distribution of drugs to Europe using the Venezuela-Africa route.³³ This information suggests that after a period of fragmentation of trafficking groups following the collapse of the Medellin and Cali cartels, paramilitaries emerged as a central organization in the drug market, with the coherence and capacity to cut the kinds of stable deals required for a state-sponsored protection racket.

In 2006, the Supreme Court of Justice and the Attorney General's Office initiated an investigation of the links between paramilitaries and politicians, known as the "parapolitica" scandal. The proceedings, which resulted in the investigation of more than 50 legislators and several mayors and governors, revealed how paramilitaries resorted to armed pressure in order to guarantee their success in the 2002 legislative elections. One of the most telling aspects of the paramilitaries' strategy to extend their influence over the political system was the agreement known as the "Pacto de Ralito" signed by paramilitary leaders, 4 representatives of the Lower Chamber, 4 Senators and 5 mayors in 2001, with the aim of "re-founding the patria" and forging a new social contract. Although it is clear that the paramilitaries' objectives while constructing their own racket went beyond drug trafficking to include their territorial interests as key actors in the armed conflict (Duncan 2005), their strategy allowed them to obtain protection, at least temporarily, from prosecution for their drug-trafficking offences. For example, while participating in the demobilization process, prominent paramilitaries and traffickers, such as Salvatore Mancuso, were able to avoid extradition to the United States (Duncan 2005). However, as in the case of their Cali predecessors, these leaders were eventually extradited, as the government responded to the pressure created by the parapolitica scandal.³⁴ This outcome highlights again how competitive electoral politics posed a barrier to the formation of stable protection rackets in Colombia. Not only did electoral competition provide strong incentives for political actors to publicize scandals that might undermine their opponents by linking them to illicit criminal activities. It also made incumbent officials more responsive to the public pressures generated by such scandals when they arose. Together, these factors made it difficult for the Colombian government to offer traffickers a credible guarantee of protection.

In sum, the case of drug trafficking in Colombia shows how competitive politics, in conjunction with territorially uneven control by the central state, can pose an important barrier to the formation of state-sponsored protection rackets. Although protection rackets may emerge in the face of competitive politics, they are prone to destabilization by scandals and the other uncertainties created by electoral competition, including how long the tenure of the incumbents will last. The infeasibility of state-sponsored protection rackets means that the levels of violence generated by illicit markets may be higher in places with more competitive politics.

CONCLUSIONS AND QUESTIONS FOR FUTURE RESEARCH

Illegality does not necessarily breed violence. The relationship between illicit markets and violence depends on institutions of protection: if state-sponsored protection rackets form, illicit markets can be peaceful. Conversely, the breakdown of state-sponsored protection rackets, which may result from well-intentioned reforms aiming to improve law enforcement, can lead to violence. In Mexico, a state-sponsored protection racket formed in the 1940s and endured until the late 1980s, resulting in low levels of violence linked to drug trafficking. The break-down of the racket, due to an increase in political competition, anti-corruption reforms, and new entrants into the illicit market that, together, shortened the time horizons of state officials and altered the geography of enforcement, led to a sharp increase in violence. In Colombia, competitive politics, in conjunction with the territorially uneven control of the central state, hindered the formation of stable state-sponsored protection rackets. Although the Cali traffickers and later the paramilitary groups enjoyed state protection for a while, these rackets were short-lived, proving unable to withstand the political pressures generated by scandals. The infeasibility of durable state-sponsored protection rackets in Colombia helps explain the high levels of violence generated by the narcotics market.

This article poses several challenges for future research on illicit markets and violence. First, the state-sponsored protection racket in Mexico, operated in the context of a non-democratic political regime, and democratization played an important role in the breakdown of the protection racket. Moreover, in Colombia, the uncertainty generated by a fundamental component of democracy, competitive elections, posed an important barrier to stable institutions of state-sponsored protection. This raises the question, are state-sponsored protection rackets feasible in democratic systems? If not, are new democracies that emerge, as in Mexico, in the face of large illicit markets doomed to high levels of violence? More generally, what is the relationship between the type of political regime and violence in illicit markets? Because democracy is "by definition a government pro tempore, a regime in which the electorate at regular intervals can hold its governors accountable and impose change" (Linz 1990), durable institutions of protection may prove harder to sustain in democratic systems.³⁵ Hence, the likelihood that illicit markets will generate violence may be higher in democracies than in non-democracies.

Second, we have shown how "bringing geography in" by focusing on the spatial organization of enforcement and crime provides a stronger understanding of violence in illicit markets. As seen most vividly in Mexico, shifts in the "ft," or congruence, between the geography of enforcement and the geography of criminality made it harder to sustain state-sponsored protection in the narcotics market, which, in turn contributed to a large increase in violence. The dealignment of the geographies of enforcement and criminality in Mexico resulted partly from policy reforms that deliberately redrew the jurisdictional map. A key task for future research involves exploring other factors besides reforms, such as changing market forces and shifts in the capacity of states to project power, that can alter the alignment between the geographies of enforcement and criminality and thus affect levels of violence in illicit markets.

Finally, we focus on one key mechanism, state-sponsored protection rackets, with a pacifying effect on illicit markets. Are there other, perhaps more normatively appealing, mechanisms that could produce a similar violence-mitigating effect? If, as suggested above, state-sponsored protection rackets are infeasible in democracies, what alternative methods are available to democratic governments for reducing violence in illicit markets? Intriguing clues about other pacifying techniques can be found in the cases of poaching in nineteenth-century Britain and crack cocaine markets in the contemporary United States. Poaching in Brit-ain in the nineteenth century took both violent and non-violent forms. Peaceful poaching occurred when poachers (1) were local residents seen by the community as semi-legitimate "social criminals" who were poaching not for proft, but "for the pot" (i.e., for food), (2) had the local knowledge to ascertain when game keepers were away, whether they were armed, and whether they were likely to put up a fight to protect the game, and (3) faced light punishment if caught (Archer 1999). By contrast, poaching led to violence when poachers were outsiders linked to urban-based gangs and stigmatized as part of the "dangerous classes" emerging in industrializing cities, lacked the local knowledge required to avoid confrontations with game keepers, and faced severe punishment. Under these conditions, poachers "preferred to fight their way out of trouble if they met up with a posse of keepers" (Archer 1999). The case of poaching thus points to several conditions that may affect levels of violence in illicit markets: cultural factors, that is, whether or not criminals are seen as legitimate actors by the communities in which they operate, informational factors, especially the capacity of criminals to acquire reliable information about the probability of getting caught, and the severity of punishment.

Crack cocaine markets in the United States provide further insights about pacifying mechanisms. According to Reuter (2009), violence in crack markets declined considerably as a result of a combination of technological and demo-graphic factors. The introduction of cell phones and pagers meant that drug deals were increasingly carried out in locations agreed on by the buyer and seller, such as apartments, restaurants, and offices, instead of exposed street corners. This increased flexibility in the locus of illicit transactions as a result of new technologies led to a drop in violence by reducing both the vulnerability of buyers to robbery and territorially-motivated conflict among sellers. Moreover, the aging of the population participating in crack markets likely had a further pacifying efect.³⁶ Demographic conditions, like the age of the population involved in illicit markets, may prove difficult to change through public policy, whereas technological parameters are probably more susceptible to policy interventions. Although a government program to distribute cell phones and BlackBerries to drug dealers and consumers might raise a public outcry, the evidence from crack markets suggests that this kind of measure could lead to a reduction in violence.

Studies that address questions such as these will provide a far stronger under-standing of the complex relationship between illegality and violence.

Comentarios

¹ On the complex relationship between so-called "blood diamonds" and violence, see Snyder and Bhavnani 2005 and Snyder 2006.

² For an analysis of the relationship between illicit drugs and political stability in Burma, see Snyder 2006.

³ See, for example, Gambetta (1996).

⁴ We leave aside such cases of non-enforcement "by default," that is, where the state is either absent or lacks the capacity to enforce the law even if it wanted to.

⁵ Dal Bó, Dal Bó and Di Thella (2006) use a similar logic to explain how pressure groups extract policies more easily from government officials when they can use both transfers, such as requests for bribes, and threats at the same time in order to get policy favors.

⁶ For the sake of simplicity, we assume there is no collusion among protectors or among criminal organizations.

⁷ It bears emphasis that deploying violence is costly to criminal organizations, because it can destroy wealth and jeopardize profits by bringing unwanted scrutiny from the public and law enforcement (Schelling 1980). A more fully specifed model of state-sponsored protection rackets should include as an endogenous factor the cost to both organizations and protectors of using violence.

⁸ We thank Sebastián Mazzuca for calling this matter to our attention.

⁹ Moreover, dealing with a single organization should be attractive to protectors because it lowers their transaction costs.

¹⁰ We assume there is just one protector.

¹¹ We assume the protector has the capacity to put the criminal organization out of business. A more complex model would relax this assumption by considering the uncertainty protectors face about whether their efforts to shut down the organization will, in fact, succeed. Failed attempts to drive a criminal organization out of business may result not in praise and a promotion, but in criticism and a demotion or worse, especially if these failed efforts generate violence and negative publicity. A more complex model of state-sponsored protection could also allow for "incremental enforcement," which would permit protectors to alter the costs of doing business faced by criminal organizations without necessarily destroying them.

¹² The optimal number of criminal organizations from the standpoint of a protector is not clear. Two are better than one, but are three better than two? As noted, monitoring and other transaction costs, which rise as the number of organizations increases, pose an important constraint on the number of organizations preferred by protectors. Moreover, too many organizations can generate excessive competition that would dissipate profits and thus reduce the rents available to protectors. Still, as long as competition among organizations does not reduce the rents available to a protector to an amount less than E + 1, that is, to a level just above the value of the raise or promotion the protector can earn by enforcing the law, then it should prefer multiple organizations to one.

¹³ By the same logic, no organization is willing to pay a premium price (P*) for exclusive rights to protection, because a protector cannot make a credible commitment that, after taking the premium payment, it will actually deliver on the promise to drive all the organization's rivals out of business.

¹⁴ Still, protectors may prefer some level of violence, because it can drive up the value of protection.

¹⁵ However, Astorga (2005) notes that many rival drug trafficking organizations had common origins in the state of Sinaloa and thus to some extent "all of them emerged from the same root." As a result, the boundaries among the organizations were difficult to draw precisely.

¹⁶ Mexico is a federal system with thirty-one states.

¹⁷ For example, in 1985, the DFS was shut down and replaced by another centralized agency, the Center for Research and National Security (CISEN). Then, in 1993, another agency, the National Institute to Combat Drug Trafficking, was created alongside CISEN.

¹⁸ The office of Assistant Attorney General "A" controlled the states of Aguascalientes, Campeche, Distrito Federal, Durango, Guerrero, México, Morelos, Nuevo León, Sonora and Veracruz; the office of Assistant Attorney General "B" controlled Baja California Sur, Chihuahua, Colima, Guanajuato, Hidalgo, Jalisco, Oaxaca, Tabasco, Tamaulipas, Yucatán and Zacatecas; and the office of Assistant Attorney General "C" controlled Baja California, Chiapas, Coahuila, Michoacán, Nayarit, Puebla, Querétaro, Quintana Roo, San Luis Potosí, Sinaloa and Tlaxcala.

¹⁹ The reforms thus correspond to a shift from scenario A to D in Figure 1.

²⁰ The offices of the Assistant Attorneys General for Prosecution were dismantled in 2002, and the main argument for eliminating them was precisely that "the current zones comprise discontinuous territories; the same delegation can include states in the North, South and Center of the country [...] it is necessary to reform the structure in order to achieve greater coordination in the fight against crime." Justification to the Organic Law for the Attorney General's Office (LOPGR), 23 April 2002.

²¹ In January 2007, Felipe Calderón initiated his term as President of Mexico with the mass extradition of 15 prominent drug traffickers.

²² Moreover, in some cases, the seconds-in-command, who often replaced their captured or extradited leaders, had stronger military orientations than their predecessors and were thus more prone to use violence.

²³ We use indexes of spatial autocorrelation (Moran's I and LISA maps) and simple mapping techniques that we do not report here because of space constraints. These analyses are available from the authors by request.

²⁴ The most significant and stable cluster of high violence during this period was located in southwest Mexico, in the states of Oaxaca, Guerrero, Chiapas, and Puebla. This cluster was likely caused by political instability unrelated to drug-trafficking dynamics.

²⁵ Between 1981 and 1996 the value of the Moran's I statistic, an indicator of spatial clustering, ranged from –0.13 to 0.48, with a mean of 0.29, and its significance level ranged from p = 0.001 to p = 0.032, with a mean of 0.02. Between 1997 and 2003, the Moran's I ranged from –0.005 to –0.26 with a mean of –0.18, and its significance level ranged from p = 0.03 to p = 0.58, with a mean of 0.17. It is important to note that the Moran's I statistic is not usually compared across time, and, hence, the values reported here should be taken as descriptive measures and not as statistical trends. Still, the notable change in the value of the Moran's I statistic and its significance levels, coupled with a mapping of homicide rates over time, provide strong suggestive evidence of a lack of spatial correlation in levels of violence after 1996.

²⁶ It is necessary to note that there are analytical limitations in considering paramilitary groups as a unified organization. Modern paramilitary groups are highly complex and fragmented as their emergence was linked to varied state and societal actors (sectors of the military, landlords, and drug traffickers) and closely tied to regional power dynamics. In 1997 paramilitary groups became unified in a single federation called Autodefensas Unidas de Colombia (AUC) under the leadership of Carlos Castaño, which provided more coordination and central control over diverse factions. Yet, the story of paramilitary groups has been characterized by cycles of breakdown and restructuring (Baron and Gutierrez, 2005) among highly regionalized factions that diverge in their assessment of the most appropriate military, political and funding methods. Yet, by the end of the 1990s the nature of paramilitary groups had changed notably: their number of soldiers grew, they extended social networks, their discourse and military methods were increasingly more sophisticated (Romero 2002). Thus, the paramilitaries became a far more coherent federation than in the previous decade.

²⁷ Hasta el fin de los dias. Revista Semana. January 23, 1995.

²⁸ Yet it is important to note that common criminal groups engaged in their own war while competing for the jobs outsourced by drug traffickers. By the end of 1986 massacres among common criminal groups had become very frequent in Medellin. "Radiografía del sicariato en Medellín". In: El colombiano. January 25, 1990.

²⁹ "En Fredonia y Santa Bárbara le dieron "el paseo" a dos sujetos". In El Colombiano. March , 1981.

³⁰ The average homicide rate in Cali in the period 1984-1991 was 61 per 100.000 compared to 219 per 100.000 inhabitants in Medellin (Source CIDSE, Jaramillo Ana Maria, DIJIN). Homicide rates increased substantially in Cali in 1992 and 1993 (89 and 117 respectively), precisely the years in which the war of the "Pepes" against Escobar was at its peak.

³¹ This does not mean that paramilitaries did not engage in high levels of violence and extreme forms of social control. Still, this violence seems to have been driven more by the objectives of advancing territorial and political control than by the goal of advancing the drug trade per se. Because of the persistent armed conflict in Colombia, it is problematic to claim that general violence levels have gone down. Still, bearing in mind this complication, during the government of President Alvaro Uribe (2002 to present) some indicators related to armed violence have, in fact, declined.

³² Revista Semana. El Cartel de 40. # 1277. October 21, 2006.

³³ El Espectador. El ex jefe del bloque Catatumbo estaría involucrado en complejas redes de narcotráfico. March 17, 2009.

³⁴ From the standpoint of some analysts, victims of paramilitaries, and human rights organization, extradition was actually an advantageous result for paramilitaries because it meant they would not face prosecution for their crimes against humanity in Colombia (see Revista Cambio "Con extradición de Mancuso quedan sin esclarecer mas de 5000 muertes en el Catatumbo" November 2, 2008). There is evidence that in 2001 the paramilitaries in fact considered establishing an arrangement with politicians explicitly related to drug trafficking called the "Plan Birmania" inspired by the arrangements between the military and drug traffickers in Burma. The "Plan Birmania", was conceived as an Alliance of illegal forces, funded by the drug trade, with the aim of consolidating short-term control over political power, initially at the local, and then at the national level. This plan was allegedly opposed by the paramilitary leader Carlos Castaño, who resigned from the direction of the AUC (Autodefensas Unidas de Colombia) when this plan was envisioned. El Espectador. "El pepe Mayor" September 13 2008.

³⁵ Still, if state officials, especially those involved in law enforcement, enjoy autonomy from elected politicians and thus are not directly subject to "democratic control," their time horizons may be long, and, hence, they may have the capacity to forge durable institutions of protection.

³⁶ Reuter (2009) notes that "rates for violent crime peak early, at about ages 18-22."

References

Andreas, Peter. 2000. Border games: Policing the US-Mexico divide. Ithaca: Cornell University Press.        [ Links ]

Archer, John. E. 1999. Poaching gangs and violence: the urban-rural divide in nineteenth-century Lancashire. British Journal of Criminology 39 (1): 25–38.        [ Links ]

Artz, Sigrid. 2003. La militarización de la procuraduría general de la república: riesgos para la democracia mexicana. usmex 2003-04 Working Paper Series.        [ Links ]

Astorga, Luis. 2004. El siglo de las drogas. México: Plaza & Janés.        [ Links ]

Camacho, Álvaro, and López, Andrés. 2001. From smugglers to drug lords, to "traquetos": Changes in the Colombian illicit drugs organizations. In Peace, democracy, and human rights in Colombia, eds. Christopher Welna, and Gustavo Gallón, 60–89. Notre Dame: University of Notre Dame Press.        [ Links ]

Castillo, Fabio. 1996. Los nuevos jinetes de la cocaína. Bogotá: Oveja Negra.        [ Links ]

Dal Bó, Ernesto, Pedro Dal Bó, and Rafael Di Thella. 2006. "Plata o plomo?": Bribe and punishment in a theory of political influence. American Political Science Review 100 (1): 41–53.        [ Links ]

Duncan, Gustavo. 2005. Narcotrafcantes, mafosos y guerreros. Historia de una subordinación. In Narcotráfco en Colombia: Economia y violencia. Bogotá: Fundación Seguridad y Democracia.        [ Links ]

Falleti, Tulia. 2005. A sequential theory of decentralization: Latin American cases in comparative perspective. American Political Science Review 99 (3): 327–346.        [ Links ]

Flórez, Carlos Antonio. 2005. El estado en crisis: Crimen organizado y política: Desafos para la consolidación democratica. PhD dissertarion, Universidad Nacional Autónoma de Mexico.        [ Links ]

Gambetta, Diego. 1996. The Sicilian Mafa: The business of private protection. Cambridge: Harvard University Press.        [ Links ]

Gutiérrez, Francisco, and Mauricio Barón. 2005. Re-stating the State: Paramilitary territorial control and political order in Colombia, 1978–2004. Crisis State Programme, Development Research Center, Working paper 66.        [ Links ]

Haggard, Stephan, Silvia Mayfield, and Ben Ross Schneider. 1997. Theories of business and business-State relations. In Business and the State in developing countries, eds. Silvia Mayfield and Ben Ross Schneider, 36–60. Ithaca: Cornell University Press.        [ Links ]

Kenney, Michael. 2007. From Pablo to Osama: Tracking and terrorist networks, government bureaucracies and competitive adaptation. University Park: Penn-sylvania State University Press.        [ Links ]

Linz, Juan. 1990. Perils of presidentialism. Journal of Democracy 1 (1): 51–69.        [ Links ]

Lupsha, Peter A. 1991. Drug lords and narco-corruption: the players change but the game continues. Crime, Law and Social Change 16 (1): 41–58.        [ Links ]

Resa Nestares, Carlos. 2001. El estado como maximizador de rentas del crimen organizado: El caso del tráfico de drogas en México. Documento 88, Instituto Internacional de Gobernabilidad.        [ Links ]

———. 2003. El comercio de drogas ilegales en México: La nueva policía mexicana. http://www.uam.es/personal_pdi/economicas/cresa//nota0103.pdf.        [ Links ]

Reuter, Peter. 2009. Systemic violence in drug markets. Crime, Law and Social Change 52 (3): 275–284.        [ Links ]

Romero, M. (2002). Democratización política y contrarreforma paramilitar en Colombia. Política y Sociedad 39 (1): 273–292.        [ Links ]

Schelling, Tomas C. 1980 (1967). Economics and criminal enterprise. In The economics of crime, eds. Ralph Andreano and John Siegfried, 377–394. Cambridge: Schenkman.        [ Links ]

Serrano, Mónica. 2007. Narcotráfico y gobernabilidad en México. Pensamiento Iberoamericano 1: 251–278.        [ Links ]

Skaperdas, Stergios, and Constantinos Syropoulos. 1995. Gangs as primitive states. In The economics of organized crime, eds. Gianluca Fiorentini and Sam Peltzman. Cambridge, New York: Cambridge University Press.        [ Links ]

Snyder, Richard. 2006. Does lootable wealth breed disorder? A political economy of extraction framework. Comparative Political Studies 39 (8): 943–968.        [ Links ]

Snyder, Richard and Ravi Bhavnani. 2005. Diamonds, blood, and taxes. A revenue-centered framework for explaining political order. Journal of Conflict Resolution 49 (4): 563–597.        [ Links ]

Stanley, William. 1996. The protection racket state: Elite politics, military extortion and civil war in El Salvador. Philadelphia: Themple University Press.        [ Links ]

Thoumi, Francisco. 2002. Illegal drugs in Colombia: From illegal economic boom to social crisis. Annals of the American Academy of Political and Social Science 582 (1): 102–116.        [ Links ]

Torres, Édgar and Armando Sarmiento. 1998. Rehenes de la mafia: En las entrañas del cartel. Bogotá: Intermedio.        [ Links ]

Transborder Institute. TBI Mapping Project. University of San Diego. http://www.sandiego.edu/peacestudies/tbi/projects/current_projects/mapping_project.php.        [ Links ]

Warchol, Greg, Linda Zupan, and Willie Clack. 2003. Transnational criminality: An analysis of the illegal wildlife market in Southern Africa. International Criminal Justice Review 13(1): 1–27.        [ Links ]