Journal of Quantitative Criminology

, Volume 30, Issue 4, pp 651–676 | Cite as

The Impact of Leadership Removal on Mexican Drug Trafficking Organizations

Original Paper

Abstract

Objectives

Has the Mexican government’s policy of removing drug-trafficking organization (DTO) leaders reduced or increased violence? In the first 4 years of the Calderón administration, over 34,000 drug-related murders were committed. In response, the Mexican government captured or killed 25 DTO leaders. This study analyzes changes in violence (drug-related murders) that followed those leadership removals.

Methods

The analysis consists of cross-sectional time-series negative binomial modeling of 49 months of murder counts in 32 Mexican states (including the federal district).

Results

Leadership removals are generally followed by increases in drug-related murders. A DTO’s home state experiences more subsequent violence than the state where the leader was removed. Killing leaders is associated with more violence than capturing them. However, removing leaders for whom a $30m peso bounty was offered is associated with a smaller increase than other removals.

Conclusion

DTO leadership removals in Mexico were associated with an estimated 415 additional deaths during the first 4 years of the Calderón administration. Reforming Mexican law enforcement and improving career prospects for young men are more promising counter-narcotics strategies. Further research is needed to analyze how the rank of leaders mediates the effect of their removal.

Keywords

Leadership Violence Murders Drug trafficking 

Introduction

Has the Mexican government’s policy of capturing and killing drug-trafficking organization (DTO) leaders led to reduced or increased violence? The debate over this question centers around two competing arguments. One side argues that leadership targeting has contributed to rising violence by inciting retribution. The other side argues that violence would have increased anyway, so the Calderón administration’s policies have not been counterproductive. Scholars have taken both sides of the debate, but existing answers to this question are largely limited to case studies.1 At its core, this debate is focused on a counterfactual—what would have happened if the government of Mexico had not captured or killed DTO leaders. While it is impossible to actually examine such an alternate world, we can explore this counterfactual using theory, qualitative evidence, and quantitative analysis.

This section presents a brief background on violence in Mexico, discusses findings in related research, and outlines the remainder of the paper. The section “A Model of DTO Leadership Removal” builds upon existing findings regarding both violent criminal and political organizations to develop a theory of leadership removal in DTOs. This theory predicts that leadership removals will be followed by increased violence due to the willingness of group members to signal their comparative advantage in violence. The section “Varieties of Leadership and Violence in DTOs” offers qualitative evidence to substantiate the assumptions of the model. Based on government documents and testimony of former DTO members, there are several reasons to expect the theoretical mechanisms to apply to DTOs in Mexico. Leadership in DTOs is well organized, often with several levels of command, each with their own geographic and strategic areas of responsibility. DTO violence is often thoroughly planned and employed strategically, and a reputation for violence is valuable for DTO members and the group as a whole.

After laying out the theoretical mechanisms for why violence in Mexico may increase following leadership removals, the section “Research Design, Data, and Key Definitions” describes the empirical approach and introduces a new data set of leadership removals. This data allows for a quantitative test of how violence changes following disruptions in DTO leadership. The section “Leadership Removal and Violence” presents findings from cross-sectional time-series negative binomial modeling of drug-related murders in 32 Mexican states (including the federal district) from December 2006 to December 2010. There are three main results: removing DTO leaders is associated with additional murders, with a larger increase in the DTO’s home state than the state in which the leader is removed; killing leaders is associated with more violence than capturing them; and the removal of leaders for whom a $15m or $30m peso bounty was offered are associated with less violence, although there is less certainty about this claim. During the period under consideration 415 additional murders were associated with leadership removal (95 % confidence interval (CI) 240 and 592). The article concludes by offering suggestions for further research and policy recommendations, including reforming the Mexican judiciary and increasing economic opportunities for young men to prevent them from joining DTOs.

Background

According to government figures, nearly half of all murders in Mexico are related to drug violence (Rios and Shirk 2011). For the 4 years from 2007 to 2010 alone, drug-related killings were four times higher than in the entire 6-year administration of Calderón’s predecessor, Vicente Fox. In response, the Calderón administration targeted DTO leaders for removal. However, this policy may have intensified the conflict. In fact, earlier leadership removals in Mexican DTOs, such as the Tijuana cartel’s Ramón and Benjamín Arellano Félix in 2002 or the Gulf cartel’s Osiél Cárdenas Guillén, may have been the spark that ignited more recent violence (Freeman 2006; González 2009; Shirk 2011b). The discovery of civilians in mass graves, the targeting of journalists and political officials, and numerous accounts of seemingly indiscriminate civilian murders suggest that violence has spread well beyond active members of the drug trade (Olson et al. 2010; Rios and Shirk 2011; Currie 2011).

Paradoxically, some observers assert that the increase in violence means the current strategy is working.2 While this claim rightly points out that levels of violence have to be considered separately from levels of drug trafficking, it makes two mistakes. First, it turns “success” into a non-falsifiable claim: under this logic, what would we expect to see if the strategy was not working—a decrease in violence? Second, disaggregating arguments over violence from drug trafficking does not necessitate that violence take a back seat. Indeed, growing numbers of civilian casualties present a serious concern for the Mexican government. From the government’s perspective, the main benefit of suppressing drug trafficking may be reducing funds that cartels can use to buy weapons. As Freeman (2006) pointed out long before violence reached its current levels, “Some Mexican and US officials attempt to paint a positive picture of Mexico’s counter-drug campaign, even claiming that rising violence is an indicator of success\(\ldots\). But an increase in murdered traffickers doesn’t translate into fewer drugs entering the US”. Whether or not removing leaders has an impact on the level of drugs crossing the border (and there is currently no evidence that it does), the increase in violence is large enough to be troubling.

Previous Research on Leadership Removal in Violent Organizations

This article takes the view that crime and insurgency are points along a spectrum, with Mexican DTO violence somewhere in between (Sullivan 2012). While the goal is not to overthrow the Mexican state, there is an undeniably political aspect to some acts of DTO violence (Bailey and Taylor 2007; Currie 2011; Williams 2012). By drawing on both criminological and political violence research, the theory here is intended to show the mutually beneficial relationship between these two fields of study.

Bringing security studies into dialogue with criminological research helps to give a full portrait of DTOs’ structure and behavior. Other research along these lines includes Kenney (2007), which shows both the benefits and dangers of unifying counter-narcotics and counter-terrorism strategies (cf. Boyle 2012). Like insurgent groups, DTOs are concerned with controlling specific territorial areas (e.g. border crossings and key production locations; Flanigan 2012). DTOs have also shown an interest in manipulating the Mexican government and the public through coercion and intimidation (González 2009; Kellner and Pipitone 2010).

In addition to these similarities, it is important to note the contrasts between DTOs and other violent groups. Foremost among these differences is that the former do not seem to be ideologically driven.3 Moreover, DTOs’ desire to control specific territories is primarily motivated by commercial goals (Salazar and Olson 2007; Williams 2012). The psychological effect of gruesome attacks, while similar to terrorism, is not foreign to organized crime. Within the criminal underworld a reputation for violence is an important currency-in-trade, as we will see in the next section (Gambetta 2009; Varese 2005; Skarbek 2011). A third important difference is the presence of financial incentives for DTO members. Although some terrorists and insurgents are motivated by financial gain for themselves or their families, DTOs are far more profitable, especially for their leaders (Kellner and Pipitone 2010).

Friman (2009) argues that the use of violence by drug trafficking organizations is selective rather than an essential facet of their existence. However, several characteristics of DTOs make violence probable. First, one of the key functions of government—to enforce contracts and property rights—is unavailable to them due to the illicit nature of their business and the goods they sell (Schelling 1971). Second, the government often attempts to operate in direct opposition to these groups, requiring them to maintain at least some capability for violence in order to survive. Thirdly, government enforcement increases uncertainty between actors in the drug industry (organizations as well as individuals, both buyers and sellers) since their interactions give each party information that can later be leveraged by the government to cause one side to inform on the other. This puts stronger parties at a particular disadvantage since the government is more likely to “turn” their subordinates against them (Gambetta 2009).

Reuter’s (2009) typology suggests three main sources of violence in drug markets: internal factors, successional conflicts, and fights between competing DTOs. Recent events in Mexico reflect all three categories, but the key insight for this study is worth quoting at length:

Managerial succession is complicated by the specificity of reputation within the organization. A promising mid-level manager cannot readily provide evidence of performance to another potential employer; as a consequence higher-level managers get weaker market signals and may withhold deserved merit increases. This gives incentives to lower-level agents to use violence for upward mobility.

When a leader is removed, the incentive to use violence is further increased due to its use as a signal of the subordinate’s qualifications.

Similar conditions apply within terrorist and insurgent groups, and leadership targeting has been a popular strategy for combating them. However, there is a growing body of evidence suggesting that this strategy is counter-productive against such organizations (Asall et al. 2013; Byman 2006; Hafez and Hatfield 2006; Honig 2007; Lawrence 2010; Miller 2012; Turbiville 2007, 2010; Wilner 2010). Jordan (2009) finds that groups whose leaders are removed tend to disintegrate more slowly than others, although that could be the result of selection bias: it is possible that only stronger organizations in the sample are targeted. The current study overcomes this obstacle by analyzing levels of violence rather than organizational survival. Furthermore, all major Mexican DTOs were subject to leadership targeting during the sample period, so it is unlikely that a selection effect biases the results.

Some insurgency scholars have been more supportive of targeted removal (Tiernay 2013; Wilner 2010). In one of the most comprehensive analyses to date, Johnston (2012) finds that leadership decapitation increases the probability of war termination and government victory, while reducing the intensity and frequency of insurgent violence. That study avoids many of the problems that have plagued previous empirical work by including attempted as well as successful leadership removals. As more data becomes available in the future, adding attempted DTO removals could be a helpful complement to the current analysis. Johnston estimates that the lethality of insurgent violence generally decreases following a leadership decapitation. However, the use of OLS modeling in that paper was inappropriate for the binary dependent variable (conflict termination), which casts doubt on its conclusions.

It should be emphasized that although this article draws on previous research regarding terrorist organizations, it does not equate DTOs with those groups, which may conceal more than it reveals (Williams 2012; Flanigan 2012). Rather, the argument suggests that groups engaging in political violence share some internal characteristics that help explain their reaction to counter-narcotics or counter-terrorism policies. If this claim holds up (as the analysis below suggests that it does), it represents an important contribution to the study of criminal and political violence. The key point at this stage is to recognize that the secretive and illicit nature of both types of organizations complicates leadership succession, and that scholars can gain leverage on understanding the effects of targeted leadership removal by focusing on these characteristics. The next section builds upon these findings to develop a theory of DTO leadership removals and violence.

A Model of DTO Leadership Removal

The Impact of Leadership Removal

The model proposed here consists of three parts. First, removing DTO leaders leaves a power vacuum within the group that must be filled. Once a leader is removed, individuals within the group vie for promotion to the newly available position. In this competition for promotion, violence serves as a signaling mechanism for these prospective leaders. Thus, the removal of DTO leaders is followed by an increase in violence.

This argument rests on several assumptions drawn from previous research on organized criminal organizations. The first is that there are no “outside managers” who will be brought in to lead the organization after a removal (Reuter 2009). Promotion from the inside helps to foster the loyalty of members, and solves some of the information asymmetries discussed below (Berman and Laitin 2008; Iannaccone 1992). Recent research has suggested that gang members’ embeddedness in the criminal network correlates with their loyalty to the group (Pyrooz et al. 2013).

The next assumption is that everyone in the group has a comparative advantage in violence relative to the average member of the population, but that there are asymmetries to this advantage within the group (Skarbek 2011; Varese 2005). The final assumption is that the group does not enjoy perfect information about its members’ leadership capabilities (Gambetta 2009; Kalyvas et al. 2006; Varese 2005). One way to solve this information problem is through the use of signaling mechanisms. Ascending to a new position of leadership requires demonstrating one’s willingness and ability to commit violence on behalf of the DTO, similar to the process observed in some terrorist groups (Hafez and Hatfield 2006; Lawrence 2010).

The ability to commit and/or direct violent attacks to signal one’s qualifications for promotion to the other members of the group is therefore additionally incented when a leader is removed. The competition between subordinates to replace a leader who has been captured or killed leads to a temporary increase in violence relative to its average level. Violence works as a clear signal that effectively covers the organizational distance across which they need to communicate their qualifications to replace the removed leader. Such costly signals are commonly required by both violent and non-violent organizations to alleviate the free-rider problem and enhance member quality (Berman 1998; Berman and Laitin 2008; Iannaccone 1992; Skarbek 2011).

The Impact of Removal Method on Violence

How does the method of removal amplify or dampen the changes in violence afterward? What if a leader is captured rather than killed? The mechanisms described above are argued to work in both cases. However, the uptick in violence should be stronger when a leader is killed than when he is captured, because a captured leader has a small but non-zero chance of returning to the group.4 Furthermore, captured DTO leaders have sometimes been able to maintain links with their group while incarcerated. This does not mean that there will be no violence in the competition between members to replace a captured leader; rather, we would expect the increase in violence to be smaller since they are competing for what may be a temporary position.5

Evaluating the Model

The theoretical model focuses on the different types of leadership within a DTO and the actions necessary to signal qualifications for promotion in the wake of a superior’s removal. Qualitative evidence to substantiate the model’s assumptions is provided in the next section. The predicted effects of leadership removal will be tested below using quantitative evidence. What will be more difficult is providing evidence for the dynamics of information flows within the organization. However, there is one empirical test that can be used as a proxy for this: we can compare the changes in violence after a leader’s removal in his organization’s home state versus the state in which he was actually removed.6 If the increase in violence is strictly due to a desire for revenge on behalf of his fellow group members, we should expect them to direct this at the security forces and/or civilians in the state where he was removed. If, on the other hand, violence is used as a tool to signal qualifications for promotion to other members of the group, then we should expect these effects to be larger in the state where the DTO is based. This is because information problems in communicating violence from the state where the removal took place to the DTO’s home state would be greater. Although this study does not explicitly employ spatial analysis in the empirical model, it does speak to a growing interest in geographic patterns of violence among criminologists (LaFree et al. 2012; Tita and Radil 2011).

In summary, the model yields three hypotheses to be tested:
H1

Targeted leadership removals generally predict an increase in violence.

H2

The predicted level of violence after a leader is killed will be higher than when a leader is captured.

H3

When removals are followed by an increase in violence, that increase will be larger in the organization’s home state than the state in which the removal occurred.

One important qualification should be mentioned before we proceed. The model presented here predicts a marginal increase in violence following a leadership removal. This article does not argue that Mexico’s policy of removing DTO leaders is solely responsible for all or even most of the violence observed during Calderón’s presidency. It does, however, argue that this policy has not decreased violence, and on the whole has contributed to an increase in violence.

The predictions here draw on criminological research both within and outside of Mexico, based on patterns observed among violent organizations. A closer look at characteristics of Mexican DTOs in the next section will help to illustrate how leadership and violence are organized within those groups, and why we should expect the hypotheses above to be valid in this case.

Varieties of Leadership and Violence in DTOs

Do DTOs and their members behave in a strategic manner, as implied by the theory above? If so, several characteristics should be true of these organizations. Leadership should be well organized and several levels or types of specialization should be present. Violence should be employed by the group and its members (at least some of the time) in carefully planned attacks, rather than randomly. A reputation for violence should also be viewed as a valuable signal both for individuals and the organization.

Evidence for these traits will be drawn from recently released US and Mexican government documents, court testimony of former cartel members, and Wikileaks cables. Of course, these documents do not constitute a random sample of the evidence available and are not intended as a substitute for hypothesis testing. However, since the testimonies were given under oath and the government documents were composed as open communication between individuals with security clearances, they do provide an opportunity to hear from DTO members and government officials in their own words.7 Similar evidence has been used by other criminological studies to determine the network structure of criminal organizations (Natarajan 2000, 2006).

There may be some question as to the actual number of distinct DTOs, which is here taken to be six.8 This is the same number identified by Rios and Shirk (2011), and used for operational purposes by the US Government (e.g. Bureau of Alcohol, Tobacco, Firearms, and Explosives 2010). The group more commonly referred to as the Sinaloa cartel (aka “Pacific”, “Federation”, or “Golden Triangle” cartel) was identified in this dataset as the Guzmán cartel in order to avoid confusion with the Beltrán-Leyva cartel, which splintered from it and is also based in Sinaloa. Los Negros was not identified as its own cartel, as it is a sub-organization of the Guzmán cartel. Las Zetas was treated similarly vis-à-vis the Gulf cartel, although they switched allegiance to the Beltrán-Leyva cartel in February 2010. Arellano-Félix is taken to be the proper name of the Tijuana cartel, given that it encompassed the Oaxaca cartel in 2003, before the period under consideration here. While these alliances and mergers offer interesting and potentially fruitful cases for historical and sociological researchers and perhaps students of network analysis, this study is focused on the how removing leaders is associated with changes in violence rather than the overall group dynamics.

Sinaloa/Guzmán DTO

According to the US Department of Homeland Security’s (DHS) Office of Intelligence and Analysis,9 the Sinaloa/Guzmán DTO is organized into distinct branches under the top leadership: “The DTO leaders that direct the Sinaloa cartel independently run the day-to-day business of the cartel, but pool resources when convenient—to target a rival cartel, for example”. These sub-organizations within DTOs are commonly called ‘plazas’, designating a semi-autonomous branch responsible for trafficking in a certain region of Mexico.10 The Guzmán DTO is thought to be resilient to leadership removals: “The Mexican military’s 19 July 2010 killing of Sinaloa kingpin Ignacio ‘Nacho’ Coronel Villareal during an arrest raid\(\ldots\) may hurt his subordinate organization, but the Sinaloa cartel will likely rebound quickly”.

This document also describes the roles of secrecy and brutality in DTOs. Secrecy measures include the use of “coded language and other means” to hide “the drug trafficking activities of the conspiracy, and to avoid detection and apprehension by law enforcement authorities”. Gaining a reputation for brutality among other criminals is an important way to preclude betrayal (Gambetta 2009; Kalyvas et al. 2006; Varese 2005). A reputation for violence can also help DTO members’ chances of promotion, as the indictment discusses in detailing the careers of Joaquin (“El Chapo”) Guzmán and Juan Jose Esparragoza Moreno.

Beltrán-Leyva DTO

Evidence regarding the Beltrán-Leyva DTO provides a unique opportunity to examine the reaction of both DTO members and law enforcement officials to a high-level DTO leadership removal, due to the high-profile nature of Arturo Beltrán-Leyva (ABL). Although this evidence may be biased given Arturo’s notoriety, it is illustrative of the general mechanisms at play. A US State Department cable speculates as to the potential effect of ABL’s removal on DTO violence:

A spike is probably likely in the short term as inter- and intra-cartel battles are intensified by the sudden leadership gap in one of the country’s most important cartels. With all the Beltran-Leyva brothers likely dead or in prison, there are a number of other cartel functionaries likely to vie for the leadership slot (emphasis added).

This informed speculation speaks directly to the theoretical mechanism for violence proposed above. Retaliatory violence directed at security forces is specifically mentioned as part of the DTO’s strategy: “At the very least, efforts to clean the Beltran Leyva house and root out suspected informers will be bloody, and retaliation by the organization against Mexican law enforcement or military officials is not out of the cards”. Over the longer term, however, the cable’s author remains optimistic that ABL’s removal will result in lower levels of cartel violence, suggesting reunification with the Guzmán DTO as one possibility. His optimism is cautious, however, noting that “ABL’s death will certainly not resolve Mexico’s drug problem”. As we will see, this caution turned out to be well-founded.

Gulf/Zetas DTO

The leadership structure of the Gulf Cartel (and, by association, the Zetas) is outlined in a chart from the U.S. Department of Treasury’s Office of Foreign Assets Control.11 The document mentions President George W. Bush’s June 2007 decision to classify the Gulf Cartel “as a Tier I drug kingpin pursuant to the Foreign Narcotics Kingpin Designation Act”. On April 15, 2009, President Obama added the Zetas to that category. The chart identifies a few top leaders and a larger number of mid-level leaders and“senior members” (about 35) of the organization, many of whom have been detained. In 2007 one of the organization’s top leaders, Osiel Cardenas-Guillen, was extradited to the US where he was sentenced to a 25-year prison term.

Tijuana/Arellano-Felix DTO

The Arellano-Felix DTO operates primarily in the western Mexican states of Baja California and Baja California Sur. Consequently, the organization and tactics of this group have been of particular interest to state and federal law enforcement in California. In November, 2010, the San Diego Law Enforcement Coordination Center (SDLECC) released an analysis of ambushes that the DTO’s members had conducted against police in Tijuana.12 The investigation provides insight into the level of strategic planning that goes into DTO operations, stating that “the tactics, techniques, and procedures used by the Tijuana Cartel\(\ldots\) suggest a high level of training and preoperational planning”. The DTO’s repeated interactions with police have given members an opportunity to refine their counter-surveillance techniques, “such as varying vehicle speeds, switching lanes, and rotating vehicles in the convoy, to identify the vehicles following the drug shipment”. Other techniques include three distinct methods for ambushing rivals or law enforcement, one of which is known as the “complex mobile ambush”. Often the materiel used for the ambush—vehicles, weapons, and body armor—was found abandoned near the ambush location, suggesting a willingness on behalf of the DTO to use costly equipment for even a single revenge attack. The analysis of these attacks indicates that DTO violence is not random, but follows a strategic logic when directed at enemies of the cartel.

La Familia DTO

Like the Guzmán Cartel, the Michoacán-based DTO known as “La Familia” also adheres to a structure based around plazas. There are between two and four dozen plaza bosses in that state, each of whom is responsible for one sizable city and its surrounding municipalities. An anonymous Mexican journalist has also described a number of factions within the organization that are responsible for different aspects of the business, such as methamphetamines or pirated films and software (cited in Grayson 2010). Grayson describes the organization’s tendency to “heap retribution on authorities who pursue their leaders”, including revenge attacks for the arrest of Rueda Medina, as a means of maintaining morale among their members. After that arrest the DTO engaged in 4 days of “mayhem” that concluded with the capture and execution of a dozen officers of Mexico’s Federal Police, whose bodies were left with a note saying, “Come for another [of our leaders], we are waiting for you”.13

Juarez DTO

The case of Guillermo Eduardo Ramirez-Peyro illustrates the use of violence as a signaling device. Ramirez-Peyro was a member of the Juarez DTO who became an informant for US authorities and later sought immunity from removal when he was found to be living illegally within the US. His case was primarily based on his fear that he would be killed by members of the cartel—or corrupt police officers associated with them—for the information he gave, which led to approximately fifty arrests. In his court testimony, Ramirez-Peyro describes incidents in which informants were killed after being reported to the DTO by the AFI (Mexico’s Federal Agency of Investigation). He also describes an incident in which he was sent to jail and was not harmed by other criminals, suggesting that the cartels have power within incarceration facilities in Mexico akin to the position held by the Mexican Mafia in Los Angeles County (Skarbek 2011). The transcript of the cross examination by Kevin Lashus, Assistant Chief Counsel for the Department of Justice in the case, reads:

Q. So you didn’t get (sic) any harm because the cartel was protecting you.

A. Yeah, in several jobs\(\ldots\) I would emphasize my relation, you know, with the cartel, and that would stop the people to try to harm me or to have some kind of an argument with me. Because to harm me was like having a hand of Santillán.

Q. Right. And these were, these were criminals who trying to go after—these were criminals that did not belong to the cartel and they feared the cartel.

A. Of course, of course.

Ramirez-Peyro’s credibility was reconfirmed when the Eighth Circuit Court of Appeals heard his case in 2009.14

Mexico’s policy of leadership targeting implicitly recognizes the important role of leaders in criminal organizations. This view is demonstrated in a number of government documents describing the internal structure of DTOs. Many of these documents support the sketch of DTO structure presented above. DTO violence is coordinated by leaders behaving strategically, and violent acts can serve as signals of these leaders’ capabilities. This evidence can increase confidence in—but not prove—the mechanisms of the theoretical model. The data and methods for testing the model’s predictions are described in the next section, after which results of the quantitative analysis are presented.

Research Design, Data, and Key Definitions

DTO Leadership Removal Data

Before proceeding with the analysis, the hypotheses offered from the model above contain a number of key terms that must be defined for the sake of clarity. “Leadership removal” refers to an instance in which a high-ranking DTO member lost his position in the group due to actions by government security forces. (As discussed in the section “A Model of DTO Leadership Removal”, neither voluntary exits nor internally initiated removals were observed during the data collection process). The removal method is the means by which the individual lost his position within the group, either by being captured or killed.

Instances of leadership removal in Mexican DTOs were collected by the author for the time period from 2006 to 2010. The cases were collected by searching open-source databases15 for the terms “Mexico”, “cartel”, “leader”, and “captured”, “killed” or “arrested”. The individual’s name, the cartel that he belonged to, the date of his removal, the city and state in which it occurred, and his role within the organization were all ascertained to the greatest degree possible and recorded in the dataset.

Individuals qualified as DTO leaders if they fell into one of three categories. The first category includes individuals who founded their DTO or were the sole top leader at the time they were removed. For example, Arturo Beltrán-Leyva was the single highest-ranking leader of his group when he was killed in a firefight by security forces in December, 2009. Leaders in the second category had control over a portion of their group but not all of it, such as the plaza leaders mentioned in the section “Varieties of Leadership and Violence in DTOs”. Alfredo Beltrán-Leyva was serving in this capacity for the Guzmán (Sinaloa) Cartel when he was arrested in Culiacán in January, 2008. The third type of leadership is control over trafficking routes or other logistical functions of the DTO. José Gerardo Álvarez-Vázquez played such a role in the state of Guerrero. He was captured by state security forces after a firefight in the town of Huixquilucan in April, 2010. The three tiers are mutually exclusive and exhaustive for the domain of interest. A preliminary analysis indicates that the leader’s role can mediate the effect of his removal on the group, although that finding is not discussed in detail here. As a robustness check the models are also estimated without the inclusion of third-tier leaders, which does not change the substantive results (Table 7 in “Appendix”).

The date and location of each removal was ascertained as precisely as available accounts would allow. Given that the data was aggregated into state-months, this did not present a difficulty. Removals are analyzed in all but one model by the home state of the removed leaders organization since that is where the theoretical mechanisms can be examined most clearly: the state in which the DTO is headquartered is where it is most capable of revenge against security forces, where individuals can most clearly demonstrate their willingness and ability to commit violence in their aspiration for promotion, and where competing organizations have the most to gain from striking at the recently decapitated organization in an attempt to gain territory. The theoretical mechanism works not because of the levels of personnel and materiel in the area, but because information flows quickly from the base state to other branches of the organization. Coding of DTOs’ home states was based on reports by Stratfor and the Congressional Research Service.16 Figure 1 displays the location of the leadership removals in the dataset.
Fig. 1

Location of DTO leadership removals, 2006–2010

The governments of Mexico and the US have identified important DTO figures by placing $30 million peso bounties (about US $2.5 million) on 24 leaders and $15 million peso rewards (about US $1.2 million) on another 13. Of these, nine “$30 million peso men” have been captured or killed, all of whom are included in this dataset. Two of those with $15 million peso bounties who have been captured were also included. In all, 25 unique cases of DTO leadership removal were collected for use in this analysis. Table 2 shows a subset of the data set to give a sense of the variety in cases. Table 1 provides summary statistics for the variables of interest.
Table 1

Leadership removal examples

Organization

Date

Removal state

Name

Type

Reward

Arellano-Felix

Oct. 2008

Baja California

Eduardo Arellano Felix

Captured

 

Guzman

Mar. 2009

Distrito Federal

Vicente Zambada Garcia

Captured

 

Juarez

Apr. 2009

Distrito Federal

Vicente Carrillo Leyva

Captured

 

La Familia

Apr. 2009

Michoacán

Dionicio Loya Plancarte

Captured

$30m

Gulf

Jan. 2010

Tamaulipas

Victor Mendoza

Killed

 

Beltrán-Leyva

Sep. 2010

Puebla

Sergio Villarreal Barragan

Captured

$30m

Table 2

Summary statistics (\(n=1{,}568\) state-months)

 

Min

Median

Mean

Max

SD

Murders

0

6

22.7

452

50.5

Removal (base St)

0

0

0.007

1

0.08

Removal (rem St)

0

0

0.015

1

0.12

Captured

0

0

0.012

1

0.11

Killed

0

0

0.003

1

0.06

$30m bounty

0

0

0.007

1

0.08

$15m bounty

0

0

0.001

1

0.04

Violence Data

For this study, violence was measured by monthly murders tied to DTO violence in the Reforma dataset, which covers DTO-related killings from December 2006 to December 2010. This is one of three datasets available for drug-related violence in Mexico; the others are government figures from the National Human Rights Commission (CNDH) of the Mexican Attorney Generals Office (PGR) and statistics compiled by Calderons National Public Security System (SNSP). Because Reforma’s dataset uses a specific, clear methodology and has been verified (in some cases corrected) by researchers from the Trans-Border Institute at the University of San Diego and the Guardian in the UK, it is considered to be the most reliable.17

Authors from the Trans-Border Institute note that although “the absolute numbers and distribution of killings reported by Reforma and other media sources vary greatly, due to differing methodologies and definitions of ‘drug related’ violence” the sources available “mostly agree on the general direction of mortality trends” (Rios and Shirk 2011). In fact, upon visual inspection the three series exhibit noticeable similarity, although Reforma’s counts are lower than the CNDH or SNSP counts, which should help protect against Type I errors. In addition to total murders, the Reforma dataset contains information on aggregate monthly levels of executions and confrontations, but total murders is most logical as the main dependent variable because it captures the whole picture of violence—between competing organizations, as well as organizations versus security forces and civilians—before and after the leader’s departure.

Of course, no data set is perfect. Reforma relies on its pool of correspondents across Mexico to report drug-related killings in their regions each week. Homicides committed by DTOs are regularly reported, whether the targets are their own members, a rival group, law enforcement or political officials, or a member of the public. Unfortunately there is not a clear mechanism for back-coding homicides that come to light many weeks after the fact, nor does the newspaper release detailed case-by-case information (Rios and Shirk 2011). Nevertheless, this data set is considered preferable to the government records which lack transparency and tend to have higher estimates of violence. Presumably the results would be even stronger using such sources, but to avoid overstating the case the more conservative Reforma estimates are used (following, Freeman 2006; Ingram and Shirk 2010; Kellner and Pipitone 2010; Sabet and Rios 2009; Salazar and Olson 2007). In total, the data set consists of 32 states (including the federal district) from December 2006 to December 2010, for a total of 1,568 state-months.

Statistical Method

Because the dependent variable of interest is a count variable whose variance is greater than its mean (i.e. overdispersed), cross-sectional time-series negative binomial regression was employed. To avoid simultaneity bias, the independent variable is lagged by 1 month unless stated otherwise. Two-way (temporal and cross-sectional) fixed effects are included to control for unobserved heterogeneity between state-months.18 The next section discusses these results and their substantive interpretation. The estimates are robust to a number of additional tests such as Poisson modeling, the inclusion of additional lags of the independent variables, omission of borderline cases, and checks for reverse causality.19

Leadership Removal and Violence

How does violence change following DTO leadership removals? The theoretical argument above predicts an increase in murders, especially in the organization’s home state. Evidence from government documents and court testimony by former DTO members indicated that DTO violence is employed strategically, and that the use of violence can be an important signal for potential replacement leaders. This section tests the theoretical predictions using the data and methods described in the preceding section.

The results below estimate that a total of 415 additional deaths are associated with this strategy (95 % CI 240–592). The majority of these additional deaths are in the state where the DTO is based, which is consistent with the use of violence as a signaling mechanism rather than as revenge on security forces or civilians in the area where the leader was removed. To estimate these results, monthly drug-related murders by state were regressed on predictors for various types of removal, lagged by 1 month to avoid simultaneity bias. The findings support all three hypotheses derived from the model of leadership removal above. In Mexican DTOs leadership removals are followed by violence, likely due to signaling efforts on the part of surviving members.

The models here are robust to a number of alternative specifications. Similar estimates are present when using a 6- or 12-month lag structure. Using Poisson models or omitting the lowest-level leaders in the dataset did not change the results. Lagged murders were not a significant predictor of removal in either the state where it occurred or the DTO’s base state. These robustness checks are discussed in more detail after the presentation of the results, and exact estimates of these additional tests can be found in the “Appendix”.

Results

This subsection briefly discusses the direction and magnitude of the results shown in Table 3, with the bulk of substantive interpretation left until the section “Simulating Substantive Effects”. Table 3 displays the results for four negative binomial regression models using removal variables lagged 1 month. Table 4 presents models that extend the lag structure to 12 months for the two independent variables of primary interest. A visual representation of these results is shown in Fig. 2.
Fig. 2

Coefficient plot for negative binomial models in Table 3. NotesThick lines indicate the 95 % CI; thin lines do the same for the 99 % CI. Where the same name appears more than once, M\(x\) denotes the model in Table 3 from which the coefficient is taken. These results indicate that leadership removals are generally followed by increased murders the following month, but that certain types of removals amplify or dampen these effects

Table 3

DTO leadership removals and violence, 2006–2010

 

Model 1

Model 2

Model 3

Model 4

 

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

Removal (base St)

1.04

0.09

    

1.26

0.06

Removal (rem St)

  

0.52

0.10

    

Captured

    

0.40

0.39

  

Killed

    

1.20

0.69

  

$30m bounty

      

−0.75

0.14

$15m bounty

      

−0.92

1.20

N

1,536

1,536

1,536

1,536

AIC

11,697.67

11,702.41

11,699.67

12,505.24

\(\log \mathcal {L}\)

−5,845.84

−5,848.20

−5,845.84

−6,247.62

All models shown are cross-sectional time-series negative binomial models. The dependent variable for each model is total murders. The independent variables are lagged 1 month to avoid simultaneity bias. Variables are measured at the state-month level, and both state and temporal fixed effects are included in all models

Table 4

Long-run estimates of violence following DTO leadership removal, 2006–2010

  

Model 1

Model 2

  

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

\(\hat{\beta }\)

\(\sigma _{\hat{\beta }}\)

Removal (base state)

Lag 1

0.83

0.05

  

Lag 2

1.21

0.09

  

Lag 3

1.44

0.11

  

Lag 4

1.00

0.06

  

Lag 5

0.84

0.05

  

Lag 6

0.30

0.06

  

Lag 7

0.28

0.03

  

Lag 8

0.43

0.03

  

Lag 9

0.80

0.06

  

Lag 10

1.19

0.09

  

Lag 11

1.77

0.10

  

Lag 12

1.36

0.12

  

Removal (removal state)

Lag 1

  

0.35

0.07

Lag 2

  

0.23

0.21

Lag 3

  

0.49

0.14

Lag 4

  

0.50

0.14

Lag 5

  

0.18

0.12

Lag 6

  

0.24

0.07

Lag 7

  

0.40

0.13

Lag 8

  

0.42

0.25

Lag 9

  

0.35

0.13

Lag 10

  

0.46

0.53

Lag 11

  

0.59

0.85

Lag 12

  

0.86

0.10

N

 

1,184

1,184

AIC

 

9,792.981

9,859.67

\(\log {\mathcal {L}}\)

 

−4,882.49

−4,915.835

All models shown are cross-sectional time-series negative binomial models. The dependent variable for each model is total murders. The independent variables are lagged as indicated. Variables are measured at the state-month level, and both state and temporal fixed effects are included in all models

Models 1 and 2 help to evaluate hypotheses H1 and H3. In general, it appears that removing DTO leaders is accompanied by an increase in violence at the state level. This effect is predicted to be larger in the state where the DTO is based rather than where the removal took place. In the 12-month lag models, the positive impact of removal in Model 1 is stable throughout all twelve lags, while the estimate in the state of removal is only statistically significant (at the 0.05 level) in some of the models. These results suggest strong support for both of these hypotheses: removals are followed by more murders in the state in which a leader’s DTO is based. This increases our confidence in, but cannot confirm, the signaling mechanism as a motivation for increased violence.

Model 3 directly speaks to H2, that killing DTO leaders predicts larger levels of violence than when leaders are captured. These two removal methods are mutually exclusive and completely partition the independent variable: there were no other types of removal observed. (As noted above, however, cases in which a leader was expelled from his organization may come to light in the future). The results here tentatively support H2: levels of violence are predicted to be higher after a leader is killed than if he were captured. These coefficients are only marginally significant and await further testing, but they are stable in the 6-month lagged models in the “Appendix” (Table 5).

Model 4 provides an opportunity to analyze another aspect of the Mexican government’s strategy: leaders for whom a $30m or $15m peso bounty was offered. The removal of $30m peso leaders is followed by a smaller increase in violence. There is too much uncertainty in the estimate for $15m peso leaders to interpret this result with confidence (recall that there were only two removals of this type in the dataset). Estimates suggest that removing the Mexican government’s prime targets was still followed by increased violence, but to a lesser degree than other removals. The relative difference here is consistent in the models with additional lags. More information about the government’s bounty policy is necessary to determine why this might be the case.

Simulating Substantive Effects

Inspecting a plot or table of coefficients can give some sense of the size and magnitude of the results, but analyzing substantive effects can provide a clearer picture of a model’s predictions. Figure 3 displays four panels, one corresponding to each of the models in Table 3. This allows us to clearly see the difference between various removal strategies in the average state-month. To generate these figures, 10,000 predictions were made for each of the scenarios displayed using the models estimated above. For each prediction, the model covariates were drawn from a multivariate normal distribution parameterized by the point estimates and standard errors from the negative binomial models. This incorporates the model uncertainty into the presentation, allowing readers to see how precise each estimate is and where they overlap.
Fig. 3

Predicted substantive effects of removing DTO leaders. Note Models in each plot correspond to Table 3. Simulations were performed by making 10,000 draws of each model coefficient from a multivariate normal distribution parameterized by the point estimates and SEs from the negative binomial models. Thus, each plot represents the uncertainty inherent in the estimation procedure

Model 1 shows the average predicted increase in violence in a DTO’s base state when one of its leaders is removed. Without a removal, the predicted level of violence is about three deaths. When a removal occurs, the predicted number of deaths the next month is four to five. Thus, about one additional death is attributable to the leadership removal the month after it occurs. The removal is also associated with nine to 13 additional murders in the 12 subsequent months. For the 25 removals in the dataset this is a predicted effect of 287 more deaths (95 % CI 245, 328).

Model 2 displays a slightly smaller predicted change in the state where the removal occurred rather than where the DTO is based. The size of the substantive effect is slightly lower. This model predicts one additional death. The long-run impact in the removal state is five more murders than average. For all leadership removals the total predicted effect is 128 additional murders, but the 95 % CI includes zero (−5 to 264). Total deaths associated with leadership removals in both the DTO’s home and the removal state, then, are estimated at 415 (95 % CI 240, 592).

Model 3 compares the predicted effects of killing versus capturing a DTO leader. When a DTO leader is killed, the estimated number of deaths is between four and five, up from a baseline of three. This suggests an increase in violence, but with a high degree of uncertainty. When a leader is captured, however, predicted violence increases by at most one additional death and may slightly decrease. Recall that these estimates are only marginally significant and should be interpreted cautiously. Given the lack of statistical significance, the long-run effects of this estimate are not presented.

Model 4 evaluates whether predicted levels of violence when a leader for whom a bounty has been offered differ from those accompanying a generic leadership removal, and finds that they do. A generic removal leads to a predicted increase of about two deaths. When a “$30m peso leader” is removed, this level is slightly lower: only one additional death is predicted in the month after the removal. Estimates for the removal of a “$15m peso leader” are not shown due to the high degree of uncertainty (the line would be essentially flat over the entire interval shown).

Case Study

The arrest of Flavio Mendez Santiago in January, 2011, was discovered during the initial data collection process, but was outside the temporal range of the statistical analysis above. After estimating the models presented above, data was obtained for the first 3 months of 2011, allowing for an out-of-sample examination of violence following a leadership removal.20

Santiago was a senior member of Los Zetas, ranking high enough to be included on a list of Mexico’s most-wanted fugitives. The Mexican government offered a $15m peso reward for his capture. His arrest took place in the state of Oaxaca. For the purposes of this study, the base state of Los Zetas was taken to be Tamaulipas, due to Zetas ties to the Gulf cartel.

One difference in the out-of-sample data is that it is disaggregated to the weekly level. Figure 4 plots the data for Oaxaca and Tamaulipas at the weekly level for 5 weeks: the week that Santiago was arrested, and 2 weeks before and after.
Fig. 4

Deaths in Oaxaca and Tamaulipas, January 2011. Note The vertical line is placed at about January 18, the date on which Flavio Mendez Santiago was arrested. Weekly levels of violence in Tamaulipas are indicated by grey points; unfilled black points represent violence in Oaxaca. Violence increased in both states following the leadership removal, but to a greater degree in the home state of Mendez’s organization

In each case, violence increased following the arrest of Santiago Mendez, to an even greater degree than that predicted by the regression models. While it would be foolish to make too much of a single data point, this case provides a modest degree of additional confidence in the main empirical result. Further robustness checks are discussed below.

Robustness Checks

Several additional models were estimated to check the robustness of the results above: additional lags, Poisson modeling, omitting certain cases of removal, and checking for reverse causality. These checks are presented in “Appendix” and briefly discussed here. First, extending the lag structure to 6 months allows us to see how long increased levels of murders persist after a removal (Table 5). In the base state this effect is significant for all six lags, while in the removal state it is not. This further supports the theoretical argument that the increase is due to competition within the DTO rather than retribution against security forces. The other main findings presented above—more violence after killing leaders than capturing them, and the smaller increase when $30m peso leaders are removed—are also robust to the inclusion of additional lags.

Poisson models of violence after leadership removal also show similar results (Table 6). All the substantive effects are identical, and the estimates are more precise. However, fit statistics show that these models are less appropriate for the data at hand, as would be expected due to overdispersion in the dependent variable.

As discussed in the section “Research Design, Data, and Key Definitions”, three types of leaders were included: top-level leaders, those with control over a portion of the group (i.e. a plaza), and leaders who control trafficking routes and perform other logistical functions. Excluding the lowest level of leaders allows us to see how sensitive the results are to the particular cases included (Table 7). Although the estimates are slightly different from those above, the substantive interpretation remains unchanged.

Finally, an important question could be raised about the direction of causality: were leadership removals more likely to occur in states that were already experiencing high levels of violence? In the main results above this was addressed in two ways: by lagging removals to avoid simultaneity bias and by including state- and month-level fixed effects. An additional test is to predict the likelihood of removal using lagged murders as the explanatory variable. Table 8 shows that murder levels in the DTO’s home state or the state of removal up to 6 months beforehand are not helpful predictors. These checks provide additional confidence in the results presented above: DTO leadership removals are followed by increased violence. The next section discusses the implications of these findings for policy and future research, and concludes the paper.

Discussion

Did the Calderón administration’s targeted removal of Mexican DTO leaders help prevent violence or exacerbate the conflict? Experts have taken both sides in this debate (Currie 2011; Freeman 2006; González 2009; Guerrero-Gutiérrez 2011; Olson et al. 2010; Rios 2011; Shirk 2011a, b). This article reviewed research on criminal organizations, presented several documents that have recently come to light regarding the internal workings of DTOs, and conducted an empirical analysis of murder counts following removals to offer an answer. In total, the models here suggest that leadership targeting has been associated with 415 additional murders. Although this figure is a small fraction of the total drug-related deaths in Mexico over the last several years, these deaths were avoidable. A strategy intended to combat narcotics trafficking instead worsened the plight of Mexico’s civilians caught in the crossfire.

Why might this have occurred? Criminal organizations use violence as a strategic tool to signal their viability to outsiders, which may be especially useful after the organization suffers a disruption. Violence also signals a member’s commitment to the group. Related research on similar criminal organizations suggests that leadership removal gives subordinates an incentive to commit additional violence in order to signal their qualifications for promotion. Qualitative evidence regarding DTOs indicated that we should expect these patterns to hold in Mexico as well. In keeping with these characteristics, violence increased more in the states where DTOs are headquartered than it did in the state where the leader was captured or killed. If the increase in violence were solely motivated by revenge, this should not have been the case. This suggests that DTOs value violence as a qualification for leadership, and that candidates for promotion to a vacated position are willing to conduct violence to signal their commitment to the group.

The method of removal and offer of a bounty served as potentially mitigating or exacerbating effects. Killing leaders was followed by more violence than capturing them, likely because a captured leader can potentially return to the group or perform leadership functions while imprisoned. With a smaller chance of replacing the removed leader, lower level members have less incentive to commit additional violence. Removing leaders for whom a $30m peso bounty was offered was also followed by an increase in murders but to a lesser extent than other removals.

This study raises new questions to be addressed by future research. Why is the removal of $30m peso leaders associated with a smaller increase in violence than average? Perhaps the choice to target these leaders resulted from a different decision-making process than the targeting of leaders for whom no bounty (or a smaller one) was offered. The specific role played by different leaders could also mediate the effects of their removal on the group. A disaggregation of leadership by function or tier could help address this question. The addition of more removal cases over time and information on the choice to capture or kill certain leaders could also illuminate our understanding of the connection between leadership removal and violence. More information about those killed would help to determine the relative distribution of murders between DTO members, law enforcement and government officials, and civilians.

The findings here yield four policy recommendations, two of which are specific to counter-narcotics policy and two that require broader change. First, Mexico’s government could further encourage other North American countries (especially the US) to consider modifying their drug policies. The recent decriminalization of cannabis in several US states may impact drug trafficking, though this remains to be seen and previous research suggests it may not reduce violence (Freeman 2006; Skaperdas 2001; Snyder and Duran-Martinez 2009). Second, since targeting $30m peso leaders appears to have a different impact on violence the government may want to publicize the logic for its targeting decisions. Law enforcement could also use “carrots” rather than “sticks” against lower level DTO leaders, by offering rewards for information on their superiors. This would give potential replacement leaders an incentive to inform on one another rather than competing to ascend to the vacated position, although the benefit of doing so would have to overcome the severe costs imposed on those who betray the DTO.

A broader policy recommendation is further reform of the Mexican judiciary and security forces. This move could help to restore Mexicans’ confidence in their government’s ability to deal with organized crime and decrease the perceived need for extrajudicial killings (Olson et al. 2010; Rios and Shirk 2011; Sung 2004). The rise of vigilante groups to combat DTOs is a particularly troubling trend as it further undermines the Mexican government’s monopoly on violence.21 Development-oriented strategies to reduce youth unemployment and provide other opportunities could also shrink the recruiting pool from which DTOs draw their members (Villarreal 2002). President Peña Nieto has publicly pledged to make reducing violence his top priority. Initial efforts to do so include programs to address widespread poverty and inequality. By improving prospects for Mexico’s citizens, and young men in particular, the government can reduce the relative gains associated with DTO membership. In the meantime, leadership targeting as it was conducted under President Calderón will not help to reduce violence.

Footnotes

  1. 1.

    For a detailed discussion of this debate, see Guerrero-Gutiérrez (2011), Rios (2011) and Shirk (2011a).

  2. 2.

    This claim is made by experts quoted in Olson et al. (2010), Currie (2011) and elsewhere. It was also the position taken in a conversation that the author had with Mark Wells (Director of the Office of Americas Programs for the State Department’s Bureau of International Narcotics and Law Enforcement) and Ambassador Patrick Duddy, who served throughout Latin America during his career.

  3. 3.

    La Familia is one possible exception to the non-ideological argument. See Williams (2012) for a thorough discussion of this point.

  4. 4.

    All of the DTO leaders in the sample were male, so the gendered pronoun is used—advisedly—throughout this section and the remainder of the paper

  5. 5.

    An objection could be raised here that rational leaders might anticipate their own demise and leave the group voluntarily upon news of an attempted removal. Alternately, they could be pushed out of the group by their own members. However, neither voluntary nor internally initiated removals were observed in the sample. The absence of the former is likely due to the high level of profits involved in drug trafficking, some of which can be used to buy off corrupt officials and finance greater security. The absence of the latter form of removal is not theoretically necessary, and cases of internal removals may become public in the future. Thanks to an anonymous reviewer for raising this point.

  6. 6.

    See section “ Research Design, Data, and Key Definitions” for a discussion of how DTOs’ base states were identified and coded in the data set.

  7. 7.

    The documents cited throughout this paper were obtained in one or more of the following three ways: through Freedom of Information Act (FOIA) requests; as part of the Wikileaks releases; or as another release of hacked documents. Wherever possible, hacked documents (including Wikileaks cables) were requested through FOIA and confirmed by multiple versions and sources. Thanks to Glen Courtney at US Customs and Border Patrol and other agency contacts throughout the Department of Homeland Security for their prompt responses to requests and phone calls.

  8. 8.

    Reducing this number as low as four or increasing it as high as eight, roughly the plausible range for the time period considered here, could introduce biased estimates but should not affect their sign.

  9. 9.

    “Mexico: The Sinaloa Drug Cartel”, 28 October 2010.

  10. 10.

    Second Superseding Indictment, USA versus Joaquin Guzman-Loera, Ismal Zambada-Garcia, Jesus Vicente Zambada-Niebla, Alfredo Guzman-Salazar, Alfredo Vasquez-Hernandez, Juan Guzman-Rocha, German Olivares, Felipe LNU, Tomas Arevalo-Renteria, Pedro Flores, and Margarito Flores, US District Court, Northern District of Illinois, Eastern Division, No. 09 CR 383, April 5, 2011.

  11. 11.

    “Perpetrators of Mexican Drug Trafficking Violence”, March 2010.

  12. 12.

    SDLECC Gang Team, Intelligence Bulletin 10-002, November 17, 2010.

  13. 13.

    Vengan por otro, los estamos esperando (Ibid).

  14. 14.

    Ramirez-Peyro versus Holder, US Court of Appeals for the Eighth Circuit, No. 08-2657.

  15. 15.

    Helpful sources included the Lexis-Nexis database as well as ongoing projects to document drug-related violence in Mexico by the LA Times, New York Times, and La Reforma newspapers.

  16. 16.

    Stratfor, “Mexico Security Memo”, May 17, 2010; Beittel, J.S. 2009. “Mexico’s Drug Related Violence”. Congressional Research Service. For visualizations of DTO-controlled areas, see http://www.npr.org/templates/story/story.php?storyId=126890893, http://www.nytimes.com/interactive/2009/03/22/us/BORDER.html?ref=us, and http://www.economist.com/blogs/graphicdetail/2012/11/mexican-drug-war. Most DTOs’ primary areas of influence in 2010 were limited to single states, although the Guzmán cartel was more expansive as was the Gulf-Los Zetas DTO before their split.

  17. 17.
  18. 18.

    For an overview of seasonal patterns of violence, which are present in the Mexican crime rate, see McDowall et al. (2012).

  19. 19.

    Thanks to the anonymous reviewers for their input on modeling strategies and checks of robustness.

  20. 20.

    Thanks to Marvin McNeese Jr. for providing this data, obtained from Nicole Ramos of the Trans-Border Institute.

  21. 21.

    According to Reforma vigilante groups are now active in 13 states (http://www.reforma.com/estados/articulo/691/1380952/). For an overview on vigilante groups and other non-government responses to violence in Mexico see this summary http://www.insightcrime.org/news-briefs/growth-of-mexican-vigilante-groups-cause-concern by Miriam Wells.

References

  1. Asall V, Gill P, Rethemeyer R, Horgan J (2013) Killing range: explaining lethality variance within a terrorist organization. J Confl Resolut. doi:10.1177/0022002713508927
  2. Bailey J, Taylor M (2007) Evade, corrupt, or confront? Organized crime and the state in Mexico and Brazil. In: Latin American Studies Association’s Annual Meeting, Montreal, CanadaGoogle Scholar
  3. BATF (2010) Project gunrunner: a cartel focused strategy. Bureau of Alcohol, Tobacco, Firearms, and ExplosivesGoogle Scholar
  4. Berman E (1998) Sect, subsidy, and sacrifice: an economist’s view of ultra-orthodox jews. Technical report, National Bureau of Economic ResearchGoogle Scholar
  5. Berman E, Laitin D (2008) Religion, terrorism and public goods: testing the club model. J Public Econ 92(10–11):1942–1967CrossRefGoogle Scholar
  6. Boyle MJ (2012) Progress and pitfalls in the study of political violence. Terror Polit Violence 24(4):527–543CrossRefGoogle Scholar
  7. Byman D (2006) Do targeted killings work? Foreign Aff 85:95–111CrossRefGoogle Scholar
  8. Currie D (2011) Mexico agonistes. National review onlineGoogle Scholar
  9. Flanigan ST (2012) Terrorists next door? A comparison of Mexican drug cartels and middle eastern terrorist organizations. Terror Polit Violence 24(2):279–294CrossRefGoogle Scholar
  10. Freeman L (2006) State of siege: drug-related violence and corruption in Mexico. Washington office on latin America special report, JuneGoogle Scholar
  11. Friman H (2009) Drug markets and the selective use of violence. Crime Law Soc Change 52(3):285–295CrossRefGoogle Scholar
  12. Gambetta D (2009) Codes of the underworld: how criminals communicate. Princeton University Press, PrincetonGoogle Scholar
  13. González F (2009) Mexico’s drug wars get brutal. Curr History 108(715):72–76Google Scholar
  14. Grayson GW (2010) La familia drug cartel: implications for US-Mexican security. Strategic Studies Institute, USAGoogle Scholar
  15. Guerrero-Gutiérrez E (2011) Security, drugs and violence in Mexico: a survey. In: Prepared for the 7th North American Forum, Washington, DCGoogle Scholar
  16. Hafez M, Hatfield J (2006) Do targeted assassinations work? A multivariate analysis of Israeli counter-terrorism effectiveness during Al-Aqsa uprising. Stud Confl Terror 29:359–382CrossRefGoogle Scholar
  17. Honig O (2007) Explaining Israel’s misuse of strategic assassinations. Stud Confl Terror 30(6):563–577CrossRefGoogle Scholar
  18. Iannaccone L (1992) Religious markets and the economics of religion. Social Compass 39(1):123CrossRefGoogle Scholar
  19. Ingram M, Shirk D (2010) Judicial reform in Mexico: toward a new criminal justice system. Trans-Border Institute special reportGoogle Scholar
  20. Johnston P (2012) Does decapitation work? Assessing the effectiveness of leadership targeting in counterinsurgency campaigns. Int Secur 36(4):47–79CrossRefGoogle Scholar
  21. Jordan J (2009) When heads roll: assessing the effectiveness of leadership decapitation. Secur Stud 18(4):719–755CrossRefGoogle Scholar
  22. Kalyvas S et al (2006) The logic of violence in civil war. Cambridge University Press, CambridgeCrossRefGoogle Scholar
  23. Kellner T, Pipitone F (2010) Inside Mexico’s drug war. World Policy J 27(1):29–39CrossRefGoogle Scholar
  24. Kenney M (2007) From Pablo to Osama: trafficking and terrorist networks, government bureaucracies, and competitive adaptation. Penn State Press, PennsylvaniaGoogle Scholar
  25. LaFree G, Dugan L, Xie M, Singh P (2012) Spatial and temporal patterns of terrorist attacks by eta 1970 to 2007. J Quant Criminol 28(1):7–29CrossRefGoogle Scholar
  26. Lawrence A (2010) Triggering nationalist violence: competition and conflict in uprisings against colonial rule. Int Secur 35(2):88–122CrossRefGoogle Scholar
  27. McDowall D, Loftin C, Pate M (2012) Seasonal cycles in crime, and their variability. J Quant Criminol 28(3):389–410CrossRefGoogle Scholar
  28. Miller E (2012) Patterns of onset and decline among terrorist organizations. J Quant Criminol 28:77–101CrossRefGoogle Scholar
  29. Natarajan M (2000) Understanding the structure of a drug trafficking organization: a conversational analysis. Crime Prevent Stud 11:273–298Google Scholar
  30. Natarajan M (2006) Understanding the structure of a large heroin distribution network: a quantitative analysis of qualitative data. J Quant Criminol 22(2):171–192CrossRefGoogle Scholar
  31. Olson E, Shirk D, Selee A (eds) (2010) Shared responsibility: U.S.-Mexico policy options for confronting organized crime. The Woodrow International Center for Scholars and the Trans-Border Institute, University of San Diego, Washington, DC/San DiegoGoogle Scholar
  32. Pyrooz D, Sweeten G, Piquero A (2013) Continuity and change in gang membership and gang embeddedness. J Res Crime Delinq 50(2):239–271Google Scholar
  33. Reuter P (2009) Systemic violence in drug markets. Crime Law Soc Change 52(3):275–284CrossRefGoogle Scholar
  34. Rios V (2011) Understanding Mexico’s drug war. Harvard University, manuscriptGoogle Scholar
  35. Rios V, Shirk D (2011) Drug violence in Mexico: data and analysis through 2010. Trans-Border Institute, San DiegoGoogle Scholar
  36. Sabet K, Rios V (2009) Why violence has increased in Mexico and what can we do about it? Technical report, unpublished working paper, November 28, 2009. As of October 4, 2010. http://www.gov.harvard.edu/files/SabetRios09_VersionPostedOnline.pdf
  37. Salazar M, Olson E (2007) A profile of Mexicos major organized crime groups. Woodrow Wilson International Center for Scholars, Mexico InstituteGoogle Scholar
  38. Schelling T (1971) What is the business of organized crime? J Publ L 20:71Google Scholar
  39. Shirk D (2011a) Drug violence and state responses in Mexico. University of San Diego, manuscriptGoogle Scholar
  40. Shirk D (2011b) The drug war in Mexico: confronting a shared threat. Council on Foreign Relations Press, New YorkGoogle Scholar
  41. Skaperdas S (2001) The political economy of organized crime: providing protection when the state does not. Econ Gov 2(3):173–202CrossRefGoogle Scholar
  42. Skarbek D (2011) Governance and prison gangs. Am Polit Sci Rev 105:702–716CrossRefGoogle Scholar
  43. Snyder R, Duran-Martinez A (2009) Does illegality breed violence? Drug trafficking and state-sponsored protection rackets. Crime Law Soc Change 52(3):253–273CrossRefGoogle Scholar
  44. Sullivan J (2012) Criminal insurgency: Narcocultura, social banditry, and information operations. Small Wars J 3(4):30Google Scholar
  45. Sung H (2004) State failure, economic failure, and predatory organized crime: a comparative analysis. J Res Crime Delinq 41(2):111–129CrossRefGoogle Scholar
  46. Tiernay M (2013) Killing Kony leadership change and civil war termination. J Confl Resolut (in press). http://jcr.sagepub.com.proxy.lib.duke.edu/content/early/2013/08/21/0022002713499720
  47. Tita G, Radil S (2011) Spatializing the social networks of gangs to explore patterns of violence. J Quant Criminol 27(4):1–25Google Scholar
  48. Turbiville G (2007) Private security infrastructure abroad: Criminal-terrorist agendas and the operational environment. Technical report, DTIC DocumentGoogle Scholar
  49. Turbiville G (2010) Firefights, raids, and assassinations: tactical forms of cartel violence and their underpinnings. Small Wars Insurg 21(1):123–144CrossRefGoogle Scholar
  50. Varese F (2005) The Russian Mafia: private protection in a new market economy. Oxford University Press, USAGoogle Scholar
  51. Villarreal A (2002) Political competition and violence in mexico: hierarchical social control in local patronage structures. Am Sociolog Rev 67:477–498Google Scholar
  52. Williams P (2012) The debate over mexican drug trafficking violence. Terror Polit Violence 24(2):259–278CrossRefGoogle Scholar
  53. Wilner A (2010) Targeted killings in Afghanistan: measuring coercion and deterrence in counterterrorism and counterinsurgency. Stud Confl Terror 33(4):307–329CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2014

Authors and Affiliations

  1. 1.Department of Political ScienceDuke UniversityDurhamUSA

Personalised recommendations