SciELO - Scientific Electronic Library Online

 
vol.23 número1Conflicto, violencia y democracia en América LatinaControl territorial y decisión de voto en Colombia. Un enfoque multinivel índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados

Revista

Articulo

Indicadores

Links relacionados

  • No hay artículos similaresSimilares en SciELO

Compartir


Política y gobierno

versión impresa ISSN 1665-2037

Polít. gob vol.23 no.1 México ene./jun. 2016

 

Articles

Federalism, drugs, and violence. Why intergovernmental partisan conflict stimulated inter-cartel violence in Mexico *

Guillermo Trejo** 

Sandra Ley*** 

** Profesor de Ciencia Política en la Universidad de Notre Dame. Kellogg Institute for International Studies, 312 Hesburgh Center, Notre Dame, Estados Unidos IN 46556. Tel: +1 574 631 9303. Correo-e: gtrejo@nd.edu

*** Profesora visitante en la División de Estudios Políticos del Centro de Investigación y Docencia Económicas (CIDE). Carretera México-Toluca 3655, 01210, México, D.F., México Tel: +52 (55) 57 27 98 00. Correo-e: sandra.ley@cide.edu


Abstract:

The dominant view of the dramatic increase of criminal violence in Mexico following the 2007 federal intervention in the War on Drugs suggests that inter-cartel violence became particularly intense in subnational regions where the president could not coordinate the federal government's actions with subnational opposition rulers but came under control where the president worked with his co-partisans. In this article we challenge the "coordination" argument and claim that in contexts of acute political polarization between Left and Right -like the one Mexico experienced before the War on Drugs- partisan conflict can motivate federal authorities to develop cooperative military and policing interventions in regions where the president's co-partisans rule, but to deliberately neglect effective assistance to the president's main political rivals and then blame the violence on them. Based on an original dataset of inter-cartel violence in Mexico (2006-2012), we show that while criminal violence was more intense in municipalities from states ruled by opposition parties, it was five times greater in cities ruled by the Left -the president's political nemesis. We use case studies to show how Mexico's conservative federal government followed differentiated strategies to deal with spirals of drug violence: it worked together and protected subnational co-partisans (PAN); partially cooperated with centrist opposition authorities (PRI); but confronted leftist governors and mayors (PRD) and left them at the mercy of drug cartels. Our results are consistent with findings in conflict studies showing that state agents do not always seek the monopoly on violence and sometimes tolerate violence to punish their political enemies.

Keywords: federalism; inter-governmental conflict; partisanship; inter-cartel violence; Mexico

Resumen:

La visión dominante del dramático aumento de la violencia criminal en México tras la intervención federal en la guerra contra las drogas en 2007 sostiene que la violencia del narcotráfico se intensificó en regiones donde el presidente no pudo coordinar las acciones del gobierno federal con las autoridades locales de oposición, pero fue significativamente menor en zonas donde se coordinó con sus correligionarios. En este artículo desafiamos el argumento de la "coordinación" y afirmamos que en contextos de aguda polarización política -como el que privó en México antes de la guerra contra las drogas- el conflicto partidista puede motivar al gobierno federal a desarrollar intervenciones militares y policiales cooperativas en regiones donde gobierna el partido del presidente y a negar asistencia efectiva en regiones gobernadas por sus enemigos políticos, para después culparlos de la violencia criminal. A partir de una nueva base de datos sobre la violencia del narcotráfico en México (2006-2012), mostramos que aunque la violencia criminal fue más intensa en los municipios de los estados gobernados por la oposición, fue cinco veces mayor en ciudades gobernadas por la izquierda -la némesis política del presidente-. A través de estudios de caso, mostramos cómo el gobierno federal utilizó estrategias diferenciadas ante la violencia del narcotráfico: coadyuvó y protegió a sus correligionarios (PAN), cooperó parcialmente con la oposición de centro (PRI), pero confrontó y desprotegió a la oposición de izquierda (PRD). Nuestros hallazgos son consistentes con estudios que muestran que los agentes del Estado no siempre buscan el monopolio de la violencia y que incluso a veces toleran la violencia para castigar a sus enemigos políticos.

Palabras clave: federalismo; conflicto intergubernamental; partidismo; violencia criminal; México

One of the most widely accepted assumptions in the study of governance and political violence is that states will always seek to hold the monopoly of violence within a given territory (Weber, 1946). This Weberian presupposition has led social scientists to expect that when states confront major waves of violence from non-state armed actors -e.g., rebel groups, terrorist organizations, mafias or organized criminal groups- they will use their military and police power to suppress violent attacks and bring state challengers and illegal actors back under the law.

A series of important works on governance in middle-income countries has recently challenged the canonical Weberian assumption and shown that government officials often make a strategic use of law enforcement and stimulate, tolerate, or simply "manage" violence rather than suppressing it. Staniland captures the gist of this critique when he observes that "the interests of a government are not always synonymous with the state's monopoly of violence (Staniland, 2012)". A growing stream of research in economics (Acemoglu, Robinson and Santos, 2013), sociology (Auyero, 2006) and political science (Wilkinson, 2005; Staniland, 2012) has shown in a wide variety of settings and world regions that government officials will seek to monopolize violence and punish violent entrepreneurs only to the extent that this yields electoral benefits.

This article focuses on the electoral incentives for policing and law enforcement in federal countries experiencing major waves of violence. We assess a major federal government intervention in Mexico between 2007 and 2012, in which national authorities launched a massive military and police campaign against drug trafficking organizations (DTOs) to put an end to twelve years of inter-cartel wars for the control of drug trafficking routes into the United States. Rather than reducing violence, the intervention resulted in a six-fold increase in inter-cartel violence and in the expansion of DTOs and their criminal associates into new criminal markets, including extortion, kidnapping for ransom, human smuggling, and the looting of natural resources. Violence and criminality, however, did not spread evenly across Mexico's territory -it became particularly intense in some parts of the country.

Scholars and Mexican government officials have insistently argued that the different trajectories of violence were partly the result of coordination problems between the federal government and subnational opposition authorities.1 According to this view, elected officials at all three levels of government -national, state and municipal- shared the goal of suppressing the illegal actions of cartels and eliminating inter-cartel violence, but they failed to coordinate because they had different understandings of security policy and the president was unable to discipline subnational opposition officials to work with him leaving their policy differences aside.

On the basis of a strategic approach to policing and law enforcement in federations, in this article we challenge the coordination argument and suggest, instead, that intergovernmental partisan conflict explains variation in levels of inter-cartel violence. In a context of unprecedented political polarization following the contested 2006 presidential election -in which the conservative candidate defeated the leftist candidate but the losing party refused to concede defeat and challenged the result in the courts and in the street- the federal intervention was heavily influenced by the dynamics of partisan conflict between Right and Left.

We distinguish four dimensions of the federal intervention: 1) the deployment of military and federal police forces throughout the Mexican territory and their degree of cooperation with subnational police forces; 2) the judicial investigation and prosecution of subnational authorities suspected of collusion with organized crime; 3) the communication strategy to expose and denounce corruption and collusion of subnational authorities with organized crime; and 4) the crisis management assistance to help subnational authorities confront spirals of criminal violence in response to the military intervention.

Our central claim is that Mexican national authorities developed coordinated interventions in subnational regions under the control of the president's conservative party but adopted confrontational strategies in states ruled by leftist subnational authorities, who belonged to the party that persistently denied the president's legitimacy as an elected authority, bitterly opposed the president's market-oriented legislative program, and had become the president's main electoral challenger.

While the military and the federal police were initially deployed to the country's most conflictive regions regardless of partisanship, we argue that federal authorities adopted radically different strategies to confront the criminal backlash and the unprecedented rise in violence that followed the military intervention. In confronting the escalation of criminal violence, partisanship became a crucial factor.

The federal government actively sought to support subnational co-partisans confronting major spirals of criminal violence and took credit for the coordinated policy response. Federal authorities protected mayors from criminal attacks and assisted them in purging their police forces; coordinated military and federal police operations with local officials and police and shared intelligence information with them; removed corrupt co-partisans from office but did not prosecute or publicly exposed them; and actively worked together with subnational co-partisan authorities in the provision of public goods and services in areas where DTOs recruited young men from street gangs. These coordinated actions weakened local cartels, encouraged outside cartels to abandon the state, discouraged others from entering the state, and contributed to halting the epidemics of violence.

In leftist subnational regions, in contrast, the federal government did not cooperate with subnational authorities and instead sought to punish them by leaving governors and mayors alone to confront the escalation of criminal violence and then blamed the violence on them. Federal authorities unilaterally designed military and federal police interventions without sharing any information with local authorities; prosecuted subnational leftist authorities and actively exposed them in national media as corrupt and inept -even with inconclusive evidence and prior to a judicial trial; deliberately declined to protect leftist mayors and their personnel against criminal attacks; and not only failed to promote any meaningful social policy intervention but actually reduced federal transfers. These confrontational strategies weakened leftist local governments and allowed DTOs to capture municipal governments and target civilians via extortion and kidnapping.

We argue that the uneven spread of criminal violence following the federal intervention was not the result of coordination problems in which the national government and subnational opposition leaders shared the same objectives to curb violence but differed on the policy means; we suggest, instead, that the conspicuous intensification of criminal violence in leftist regions was the byproduct of a bitter partisan conflict between two antithetical political forces in which federal authorities used the War on Drugs to undermine the electoral base of the president's political nemesis- the Left.

Testing the coordination argument against the partisan conflict hypothesis is a straightforward exercise. If the coordination argument is correct, following the federal deployment of the army and the federal police throughout the country we should observe the most intense levels of inter-cartel violence in subnational regions ruled by opposition parties -regardless of whether they are from the leftist Party of the Democratic Revolution, PRD, or the centrist Institutional Revolutionary Party, PRI. The coordination argument holds if we find no significant differences between PRI-and PRD-ruled regions. In both cases, violence should be equally higher than in regions ruled by the president's party, the conservative National Action Party (PAN). However, if the partisan conflict argument is correct, we should observe the most intense levels of criminal violence in subnational regions ruled by the leftist PRD. The conflict argument holds if we find systematic differences between PRI and PRD regions; that is, if criminal violence is higher in leftist regions than in both PRI and PAN regions.

Drawing on the Criminal Violence in Mexico (CVM) Dataset, an original newspaper-based databank that we constructed, we show that while municipalities from states ruled by either opposition party did indeed experience greater levels of violence than those from states ruled by the president's co-partisans, the most intense levels of violence took place in Mexican cities located in states ruled by leftist PRD governors. Compared to a situation of unified governance, where the president's party ruled at all levels of government, our statistical analyses unambiguously show that municipalities in states ruled by the Left experienced five times more drug violence than municipalities in states ruled by other opposition parties. Controlling for economic, demographic, social, geographic and spatial factors, our statistical findings provide strong support for the partisan conflict hypothesis and are robust to a wide variety of alternative explanations.

To understand how partisanship shaped the federal government interventions in the Mexican War on Drugs and why these varied interventions reduced or stimulated criminal violence, we conduct three case studies in cities that shared many relevant characteristics except for the vertical distribution of political power: Tijuana in the state of Baja California (both ruled by the president's co-partisans); Apatzingán in the state of Michoacán (both ruled by the Left); and Ciudad Juárez in the state of Chihuahua (both ruled by the PRI).

The article is structured into five sections. We first discuss different theoretical approaches to the motivations driving intergovernmental cooperation for policing criminal and political violence in federations and outline the paper's main propositions about the politicization of law enforcement under conditions of ideological polarization. Focusing on Mexico's deeply contested 2006 presidential election, in the second part we describe the context of unprecedented political polarization between Right and Left that served as background for the federal intervention in the War on Drugs. In the third section we discuss results from a wide variety of statistical models testing the coordination and the partisan conflict hypotheses, and in the fourth part we present the case studies. The paper concludes with a discussion of the implications of our findings for the study of the state and of electoral incentives for law enforcement and governance in federations.

Policing crime and violence in federations: What are the drivers of intergovernmental cooperation?

Organized crime in the world today has become a global industry in which criminal groups operate in a web that connects illicit actors and organizations from different countries and subnational locations. Drug trafficking, for example, is a global chain of local operations in which DTOs operate at global, national and local levels. This multi-level feature demands that any government intervention to confront drug trafficking and drug-related violence involves multiple authorities working at different geopolitical and spatial levels. While scholars and international institutions have emphasized cooperation between the international community and national governments, intergovernmental subnational relations are a major omitted variable in the analysis of the drug industry and of inter-cartel wars. This is particularly relevant in federations.

One of the defining features of federations is that national and subnational governments have different administrative jurisdictions, responsibilities and institutional capacities. In most federal countries, national authorities have jurisdiction over issues of national security and organized crime and local police forces are mainly responsible for petty crime. This division of labor results in different policing capacities. Whereas national authorities have access to military forces as well as federal police forces -which have the best available weaponry- subnational authorities command police forces with less sophisticated equipment and training. These differences entail that when mayors face particularly violent situations, they are dependent on protection from federal authorities. In this asymmetric relationship, intergovernmental cooperation is crucial for peace and for the survival of local authorities in violent federations.

What factors drive intergovernmental cooperation in federations?

An important tradition in legal and economic research suggests that intergovernmental cooperation is a question of institutional efficiency (Oates, 1972). When federal institutions efficiently allocate policy responsibilities to the most suitable and capable level of government (e.g., national security to the federal government and garbage collection to local governments), intergovernmental relations are marked by cooperation rather than conflict. In this approach, government officials are assumed to be benevolent social planners and intergovernmental cooperation is not a political but a technical question of clearly defining "property rights" over policy across different levels of government.

An influential alternative explanation in political science suggests that partisan politics is a defining factor of intergovernmental cooperation in federations. In his seminal study of federalism, Riker suggests that parties and party systems condition the actions of government officials in federations (Riker, 1964). Because federalism is a political bargain, the distribution of political power across different levels of government can be a crucial determinant of intergovernmental relations. When national leaders have several co-partisans ruling at subnational levels, vertically integrated political systems ("harmony") empower national leaders to discipline local representatives, facilitating intergovernmental cooperation and policy coherence. In contrast, when national leaders have few subnational co-partisans in power, vertically divided governance ("disharmony") can lead to uncooperative relations and policy incoherence (Riker and Schaps, 1957).

Following Riker's partisanship approach to intergovernmental relations, scholars of drug violence in Mexico have suggested that president Calderón's War on Drugs and the federal deployment of the army to control inter-cartel violence was more effective under conditions of unified governance where the president's PAN co-partisans ruled at the gubernatorial and municipal levels but less effective under conditions of partisan vertical plurality.

According to Urrusti's argument, partisan vertical alignment allows authorities to have more effective territorial controls (Urrusti, 2012). Federal, state and municipal law enforcement agents can act in coordination to secure subnational territories, and their fluid communication and exchange of information facilitate the implementation of effective attacks against drug cartels. Thus, criminal violence is less intense under vertically integrated governments, because cartels do not try to contest drug trafficking routes where political power is coherently aligned. Under Rios's alternative interpretation, partisan vertical alignment empowers the president to discipline his (or her) subnational co-partisans, producing coherent and consistent security policies that dissuade drug cartels from competing for territorial controls (Ríos, 2015).

One important limitation in Riker's study of intergovernmental relations -and in studies that apply his approach- is that it does not consider the possibility that in vertically divided governments higher-level authorities may use their power to strategically reward their subnational co-partisans and punish their enemies. As Weingast has suggested for the study of fiscal decentralization, national incumbents in vertically divided federations may have incentives "to encroach on decentralization as a means of weakening the ability of [subnational political opponents] to succeed in their policy goals" (Weingast, 2014).2 This strategic behavior has been underscored in classic studies of distributive politics in which authors claim that governments can provide transfers, subsidies, tariff protections, or pork barrel projects to core supporters or to swing voters. In fact, in their influential core-voter model Cox and McCubbins explicitly propose that governments can reward their co-partisan supporters and punish voters from rival political parties (Cox and McCubbins, 1986).3

Building on models of distributive politics, scholars of political violence have made a bold move to assess how electoral incentives can shape policing decisions in federations. Instead of looking at the allocation of resources from higher- to lower-level authorities as a technical problem, they analyze the deployment of military and police forces to confront subnational outbreaks of violence as a strategic problem in which national government officials may order security forces to suppress violence in some subnational regions but tolerate or even instigate violence in others, depending on local electoral conditions.

For example, in his influential study of religious riots in India, Wilkinson shows that in a context of acute polarization between Hindus and Muslims, state-level authorities more actively deployed police forces to stop Hindus from attacking Muslim neighborhoods in districts where Muslim voters could become pivotal voters and decide the outcome of elections (Wilkinson, 2005). In contrast, where Muslim voters were not a decisive electoral group, state-level authorities allowed Hindu gangs to break into Muslim neighborhoods and victimize residents.

In his insightful study of poor people's urban riots following the major 2001 macroeconomic collapse in Argentina, Auyero shows that in a context of acute class polarization, opposition subnational authorities associated with the populist Justice Party (PJ) ordered local police forces to prevent looting in areas where their business constituents would be affected but to strategically allow looting where businesses were associated with the center-right incumbent party (Auyero, 2006).

In their important study of the linkages between elected officials and paramilitary forces in Colombia, Acemoglu, Robinson and Santos show that right-wing paramilitary forces played a key role in the election of conservative politicians to the Colombian congress and that these representatives, in turn, passed legislation that was favorable to the interests of paramilitaries (Acemoglu, Robinson and Santos, 2013). As Steele persuasively shows, the de facto protection that national political actors provided to paramilitaries in Colombia proved to be very effective when these specialists in violence used their weapons to force the geographic relocation of leftist voters and union members to other villages and electoral districts (Steele, 2011).

Contrary to the classic Weberian assumption, which suggests that states will axiomatically seek to have the monopoly on violence and to suppress non-state violence, these three studies show that state leaders in developing countries can choose to tolerate or even stimulate violence for their own electoral advantage. The three cases also show that under conditions of acute social or political polarization, government authorities can strategically manipulate law enforcement to reward their political allies and punish their enemies.

To formulate our hypotheses about violence in the War on Drugs in Mexico, we use Riker's partisan approach to intergovernmental relations as our starting point. Beyond Riker, however, following strategic studies of distributive politics and of political violence, we argue that under conditions of acute political polarization federal authorities can adopt a strategic approach to policing criminal violence by which they effectively deploy security forces to protect subnational regions where co-partisans rule but tolerate criminal violence in areas dominated by their main political rivals. We suggest that federal incumbents can adopt a "punishment" regime by which they deliberately leave their subnational political rivals unprotected in conflictive regions to undermine their credibility and electoral base. This punishment strategy also entails persuading local voters to attribute the responsibility for criminal spirals of violence to his (or her) subnational political rivals.

We assess two alternative arguments: the coordination and the partisan conflict hypotheses. Following Riker, the coordination argument would expect that:

H.1. Inter-cartel violence will be more intense in subnational regions where power is vertically fragmented -regardless of the ideology of opposition parties- than in regions where power is unified and all three executive authorities belong to the same party.

Beyond Riker, the partisan conflict argument would propose that in a context of political polarization:

H.2. Inter-cartel violence will be more intense in subnational regions ruled by elected officials from the president's main ideological rival than in subnational regions ruled either by the president's co-partisans or by opposition parties that are ideologically closer to the president's party.

Note that while criminal violence in H.1 results from the inability of national and subnational opposition parties to coordinate their anti-drug actions, violence in H.2 results from the president's strategic decision to punish his (or her) subnational political rivals by leaving local government officials without protection against major criminal threats. Coordinated action contributes to effectively suppressing local criminal wars and preventing rival criminal groups from entering a region, but bitter intergovernmental conflict attracts rival cartels to contest local territories through violence.

We use the Mexican federal intervention in the War on Drugs to test the coordination and partisan conflict hypotheses. We test H.1 and H.2 using information from inter-cartel violence across 2 108 Mexican municipalities from 2007 and 2012. Before conducting our statistical testing, however, we need to establish the extent and nature of the political polarization that led to the politicization of Mexico's Drug Wars.

Mexico's Drug Wars in the shadow of political polarization

The 2006 presidential election

After seven decades of uninterrupted rule in presidential power, the defeat of the PRI in 2000 opened an unprecedented era of political polarization in Mexico between the conservative PAN and the leftist PRD. During the PRI rule, despite their sharp differences in economic policy, the PAN and the PRD shared a common goal for free and fair elections and for democratization. Figure 1a illustrates the spatial distribution of parties on the economic (Left vs. Right) and political (authoritarianism vs. democracy) dimensions during the last decade of PRI rule (Klesner, 2005). Although the PAN and the PRD never created a unified opposition front against authoritarian rule and even though the strategic coalition of the PAN with the PRI to enact major market-oriented reforms slowed down the pace of political liberalization, the opposition's shared desire for free and fair elections always left some open space for cooperation between Left and Right.

Source: Klesner (2005).

Figure 1a Ideological position of Mexican parties under authoritarian rule, 1990s 

The 2000 presidential victory of the PAN and the concurrent victory of the PRD in Mexico City, however, led Mexico into an era of bitter inter-partisan conflict. In the absence of a visible PRI leader, president Vicente Fox and the leftist mayor Andrés Manuel López Obrador -two charismatic leaders- rapidly became Mexico's leading political figures and personified the country's political polarization. As Figure 1b shows, in the post-authoritarian period the economic and sociocultural dimensions became the battleground for the confrontation between Left and Right. On the economic front, president Fox and the PAN sought to enact an ambitious program of second-generation market-oriented reforms, which López Obrador and the PRD bitterly opposed (Moreno, 2010). On the sociocultural front, the PRD in Mexico City took up the banner of abortion and rights for same-sex couples, which Fox and his party adamantly condemned. In Mexico's post-authoritarian ideological geography, the PAN and the PRD became antithetical forces. When the two charismatic leaders took these ideological battles to the mass media, the PAN-PRD rivalry quickly became the country's daily headline news.

Figure 1b Ideological position of Mexican parties under democracy, 2000s 

Political polarization reached a high point when president Fox's attorney general accused López Obrador of violating a court order to halt the construction of a hospital access road over private property. The case escalated; López Obrador became subject to impeachment by the Mexican Congress and was stripped of his immunity from prosecution. This meant that the mayor had to face trial and would no longer be eligible to run for presidential office, despite leading the polls on vote intentions.4 But an unprecedented wave of mass protest in favor of López Obrador and a shift in public opinion against the trial led President Fox to drop all charges, enabling his rival to run for president in 2006.5

The 2006 presidential campaign became a direct confrontation between Left and Right and the election was a plebiscite on López Obrador (Schedler, 2007; Langston, 2007). Mexico went through months of bitter negative campaigning. The panista candidate, Felipe Calderón, ran a campaign framing López Obrador as the "Mexican Hugo Chávez", an "authoritarian populist" leader who represented a major "threat to Mexico". President Fox ran his own campaign asking Mexicans not to take López Obrador's populist road. And López Obrador's campaign called the PAN's private-sector allies a group of "white collar criminals". Flouting the Federal Electoral Institute's order that the parties removed their negative ads and that president Fox stop intervening in the campaign, all parties sidestepped the law and continued their bitter media confrontation.

The victory of the conservative candidate by a razor-thin margin of 0.6 percent took elite polarization to unprecedented levels.6 López Obrador did not concede defeat, denounced the result as a major electoral fraud, and contested the election in the courts and in the street. After the Electoral Tribunal reviewed the legal complaints and confirmed the victory of the PAN candidate, despite "the president's unlawful intervention... which endangered the legality of the election", López Obrador launched a major campaign of nonviolent resistance and called for the creation of a parallel "legitimate" government.7

Amidst major demonstrations, Felipe Calderón was sworn into office at a military base. With the endorsement of his party and the PRI, the next day president Calderón went to Congress to make his assumption to power official but was violently removed from the tribune by a crowd of leftist legislators. At the same time, at a mass rally in downtown Mexico City, López Obrador was sworn in as "Mexico's legitimate president".

Facing a situation of dual sovereignty in Mexico City, and a conspicuous rise of inter-cartel violence in central and northern states, president Calderón began his administration with a radical policy announcement intended to overcome the political crisis. Addressing the country on national television, he called upon all Mexicans to transcend political rivalries and focus their energies on fighting the real enemy: DTOs. Even though antinarcotic policies had been barely mentioned during the campaign, the incoming president declared a "War on Drugs" and ordered the deployment of the military throughout the country's territory to quell growing inter-cartel violence (Presidencia de la República, 2006).

Inter-cartel wars and the evolution of violence

Over the course of the next six years, ending inter-cartel wars and bringing drug violence under control became the administration's central policy concern. Inter-cartel wars in Mexico had first broken out in the early 1990s but reached a high peak and became a serious challenge to the state between 2004 and 2006, the two years prior to the election of president Calderón (Trejo and Ley, 2014). The wars initially involved Mexico's four main DTOs: the Tijuana, Juárez, Sinaloa and Gulf cartels. By the early 2000s, a new cartel, La Familia Michoacana, emerged in the western state of Michoacán as a major contender for control over drug trafficking corridors.

Using information from the Criminal Violence in Mexico (CVM) Data-set,8 Map 1a illustrates the geography of inter-cartel violence in 2000-2006 and identifies three major areas of conflict: 1) northwest, 2) northeast, and 3) southwest. Inter-cartel wars between 1990 and 2006 involved mainly dyads of DTOs and their private militias fighting for control over drug trafficking routes.9 Since 1989, when the PRI lost power in Baja California and the Tijuana Cartel found itself without immediate access to state police protection, Mexican cartels began developing their own private militias to safeguard their drug trafficking routes and renegotiate state police protection with incoming opposition authorities. The proliferation of private militias following every new cycle of subnational alternation in gubernatorial power in the 1990s and early 2000s led to the outbreak of multiple inter-cartel wars in the northwest, northeast and southwest (Trejo and Ley, 2014).

Source: Criminal Violence in Mexico Dataset (Trejo and Ley).

Map 1a Geography of inter-cartel violence in Mexico (murders/executions), 2000-2006 

Although the president believed the federal intervention would be a relatively easy military victory that would reunite the country and help his government overcome the major post-electoral crisis, between 2007 and 2012 inter-cartel violence increased by a factor of six. By the end of the Calderón administration more than 70 000 lives had been lost (Reforma staff, 2013); 22 000 people were missing; over 300 local authorities, political candidates and party activists had been victims of assassination attempts or had been murdered (Trejo and Ley, 2014a); and drug trafficking had expanded into new criminal markets, including extortion, kidnapping for ransom, human smuggling, and the looting of natural resources (Guerrero, 2011; Grillo, 2012). These multiple forms of violence, however, did not spread evenly; they became more intense in some regions but not in others.

Map 1.b shows the spread of inter-cartel violence between 2007 and 2012 and identifies three new focal points of violence: 4) Gulf, 5) center-north, and 6) center. By 2012, Mexico's drug trafficking landscape had shifted from an industry dominated by five cartels to one in which over fifty organizations were actively involved in drug trafficking and related criminal activities (Guerrero, 2011; Rios, 2012). Over the course of six years of War on Drugs, inter-cartel violence had evolved into conflicts involving numerous criminal organizations fighting for control over cities, villages, and neighborhoods. These were no longer wars for the transshipment of drugs but conflicts over control of multiple criminal markets and administrative-territorial jurisdictions, particularly municipal governments.

Source: Criminal Violence in Mexico Dataset (Trejo and Ley).

Map 1b Geography of inter-cartel violence in Mexico (murders/executions), 2007-2012 

Against the backdrop of acute political polarization during the 2006 presidential election and the intensification of violence after the federal intervention, in the next section we test whether intergovernmental coordination failure or bitter political conflict between the president and his leftist subnational political rivals had any impact on the uneven spread of inter-cartel violence between 2007 and 2012.

Intergovernmental partisan conflict and inter-cartel violence: A statistical analysis

To test whether intergovernmental partisan conflict had any impact on the intensity of inter-cartel violence in Mexico's subnational regions, we use information from CVM10 and analyze the temporal and spatial evolution of 30 000 murders and executions perpetrated by DTOs across 2 108 Mexican municipalities from 2007 to 2012.11 We use a count of dto-murders in municipality i and year t as an indicator of our dependent variable. We retest all our models using information from the Mexican government database on deaths attributed to organized crime from 2007 to 2010 (not shown), and the main results remain unchanged.12

Coordination versus partisan conflict

We test the coordination (H.1) and conflict (H.2) arguments using information on the partisan affiliation of governors and mayors as reported by Mexico's federal and state electoral commissions. Mexico's three major political parties (PAN, PRI, and PRD) compete for office at three levels of government (federal, state, and municipal).13 During the 2007-2012 period, the conservative PAN held the presidency but gubernatorial and municipal powers were dispersed across the three major parties. For example, the president's party, the PAN, held 22.9 per cent of Mexico's thirty-one governorships; the PRI 59.4 per cent; and the PRD 17.7 per cent.14

To test for the coordination argument, following Urrusti, we create four "coordination" variables: Coord1 identifies with a dummy variable all cases of unified governance where the president's party rules at all levels of government (PAN, PAN, PAN). For purposes of notation, the party identified on the first row in the bracket is the party in presidential power, the second is the party in the gubernatorial office, and the third is the party ruling the municipality. Coord2 describes cases in which the governor is the only opposition force (PAN, opposition, PAN); Coord3 identifies cases in which the mayor is the only opposition force (PAN, PAN, opposition); and Coord4 captures cases in which both governors and mayors belong to the opposition (PAN, opposition, opposition). Note that these measures do not identify the partisan affiliation of the opposition; following the coordination argument, these measures simply identify parties in power that are different from the incumbent PAN -the president's party. When we test for Coord2-4, we use Coord1 as the reference category.

To test for the conflict hypothesis, we introduce partisanship and create three sets of variables that distinguish between the ideological orientations of the different opposition parties in government.

First, we focus on the governor's partisan affiliation and create measures of intergovernmental conflict regardless of who rules at the municipal level. Confl1 identifies with a dummy variable all cases in which the governor belongs to the president's party (PAN, PAN, regardless); Confl2 are cases of centrist opposition governors (PAN, PRI, regardless); and Confl3 are cases of leftist opposition governors (PAN, PRD, regardless). Given the weak policing capacity of mayors, the assumption here is that governors are the only relevant actor for intergovernmental cooperation in the War on Drugs. We use Confl1 as the reference category.

Second, we create a set of variables to identify the layering or vertical distribution of parties at all levels of government. The assumption here is that, despite their weakness, municipalities play an important role in the operation of drug trafficking organizations (via informal government protection networks) and in any governmental attempt to defeat organized crime (through the information on local conditions that uncorrupted mayors and municipal police forces can provide).

As Table 1 illustrates, we identify nine different combinations of vertical partisan fragmentation. As the information on the table's right-hand column shows, political power in Mexico's federation was considerably fragmented vertically.15 After six decades of monopolistic unified governance under the PRI, Mexico's "federalist" transition to democracy in the 1990s (by which the PRI began losing power in cities, then in states, and finally at the national level) yielded a complex mosaic of territorial power diffusion and pluralistic intergovernmental relations that continued to develop in the post-authoritarian era under the PAN in the 2000s.

Table 1 Layering of parties in Mexico's three levels of government, 2007-2012 

Source: Local electoral institutes. Note: When a party ran for office in alliance with another party, we gathered information about the party origins of the candidate and the strength of the party in the locality under study.PAN = incumbent (conservative); PRI = opposition (center); PRD = opposition (left).

We use the information in Table 1 in two alternative ways. We collapse all nine combinations into a single ordinal measure, Juxtaposition Index, which ranges from unified governance (PAN-PAN-PAN) at the lowest level to subnational leftist opposition (PAN-PRD-PRD) at the highest level. We used the ideological distance between parties as guidance to rank the different combinations, from Right to Center to Left. The fact that the PAN-PRD conflict was the main source of polarization in the 2000s makes us confident that our ranking identifies major partisan differences that actually go beyond ideology. We subsequently test for the independent effect of the nine categories using unified governance (PAN-PAN-PAN) as the reference category. This disaggregated measure of the distribution of partisan power along Mexico's federation provides us with the most accurate measure to test implications of the coordination hypothesis and to more directly test the conflict hypothesis.

Alternative explanations and controls

We control for a number of political, law enforcement, sociodemographic and geographic variables that have been tested in studies of crime and organized crime.

A number of important studies of criminal violence in economics and political science associate the outbreak and intensification of inter-cartel violence in Mexico with subnational democratization and electoral competition. Authors argue that the spread of electoral competition destroyed protection networks previously developed under the PRI and opened a new era of inter-cartel wars for territorial controls and the renegotiation of protection. We control for State and Municipal Electoral Competition and for the Alternation of political parties in gubernatorial and municipal power.16

Studies in political science also underscore the importance of law enforcement as a key causal factor in limiting criminal violence. Consistent with the political violence literature (Fearon and Laitin, 2003), scholars have argued that the state's policing and judicial capacity can deter criminal violence. We control for state capacity using the number of Prosecutors per 10000 population17. One of the most significant arguments about the intensification of violence in Mexico after the federal intervention in 2007 suggests that the government's policy of eliminating the top leaders of the drug cartels fueled an escalation of inter-cartel violence. According to Guerrero's pioneering analysis, leadership decapitation led to the fragmentation of DTOs, to the rise of new leaders and groups, and to the escalation of conflict over the control of the drug trafficking corridors (Guerrero, 2011).18 We control for the count of cartel chiefs and their deputies arrested or killed, Leadership Decapitation, as a result of government operations or inter-cartel conflicts, and assign the count to all of the state's municipalities where the event took place.19

Following the extensive literature on the sociology of crime, we control for a number of variables associated with social context and family structure that have been shown to be important predictors of criminal violence (Sampson, 1985; Villarreal, 2002), including a Poverty index (Conapo) (2006, 2011) which is a composite index of access to public goods and services at the municipal level, the municipal Sex Ratios, the proportion of Female-Headed Households, Population Age 15-35, and the municipal population size, lnpop (INEGI, 2000; 2010).

We include seven geographic regions (North, North-Center, Center, Pacific, Gulf, South, and Southeast, using Southeast as the reference category) to control for unobserved regional characteristics and a one-year lag of Violence in neighboring municipalities to control for the spatial dispersion of violence.

For statistical testing we use negative binomial models -the most appropriate modeling technique for count data when observations are non-independent and over-dispersed. We use random effects (RE) models instead of fixed effects (FE) because our key independent variables do not vary for several consecutive years, rendering FE inappropriate. We transform re coefficients into incidence rate ratios (IRR) to facilitate substantive interpretation.

Results

The results, summarized in Table 2, show that, consistent with the conflict hypothesis, intergovernmental partisan rivalry between the president and the leftist opposition party was an important predictor of the intensity of inter-cartel violence between 2007 and 2012. While there is evidence that there might have been problems of administrative or logistical coordination between the federal government and subnational opposition authorities, our results show that the underlying problem was a bitter partisan conflict between Left and Right. Our findings suggest that by omitting partisanship from the analysis, the coordination argument yields biased estimators and underpredicts overall variance in inter-cartel violence.

Table 2 Intergovernmental conflict and inter-cartel violence in Mexico, 2007-2012 (random effects negative binomial models) 

Source: Authors' calculations. Standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.10. PAN = incumbent party in presidential power (conservative); PRI = opposition (center); PRD = opposition (left). Note that party in first cell is the president's; the second one is the governor's and the third is the mayor's. PAN-PAN-PAN is the reference category. IRR = incidence rate ratio.

Models 1 and 2 test the coordination argument. The results in Model 1 show that unified governance under the PAN did offer an advantage: compared to the rest of the country, violence was 17.5 per cent lower in municipalities that were part of a vertically unified scheme of governance (Coordl, IRR = 0.825). Model 2 shows the flip side of the coin: compared to the case of PAN unified governance, inter-cartel violence was 56.3 per cent higher in cities with both opposition governors and mayors (Coord4, IRR = 1.563).

Models 3-5 reveal, however, that distinguishing between different opposition parties significantly improves the models' overall explanatory power and adds crucial information about the geographic variation of criminal violence following the 2007 federal intervention.

Compared to municipalities in states ruled by the president's party, results in Model 3 show that municipalities in states ruled by PRI governors (Confl2) experienced 31.8 per cent more violence (IRR = 1.318), but municipalities in states ruled by leftist PRD governors (Confl3) experienced 162.4 per cent more violence (IRR = 2.624). This striking five-fold difference between the PRI and PRD states suggests that, contrary to the coordination argument, something unique was taking place in leftist subnational regions.

The results in Models 4 and 5 provide important additional evidence about the impact of inter-partisan conflict on criminal violence.

The results in Model 4, in which we test for an ordinal index of juxtaposition, reveal that, holding all else constant, as a municipality moves away from a situation of unified conservative governance (PAN-PAN-PAN) in the direction of leftist subnational governance (PAN-PRD-PRD), the intensity of violence is likely to increase by 13.1 per cent (IRR = 1.131). The cumulative effect means that a leftist municipality in a leftist state experienced 104.8 per cent (13.1 x 8 layers) more violence than one under unified PAN governance.

The results in Model 5 introduce important nuances about the nature of intergovernmental partisan cooperation and conflict. To facilitate the substantive interpretation of results, Figure 2 provides a graphic representation of the model's main findings. Four findings are worth stressing.

Source: Authors' calculations. Notes: Incidence Rate Ratios (IRR) from Model 5 were rescaled for purposes of interpretation. Original IRRs were subtracted one unit to show the direct impact of juxtaposition on violence. PAN = incumbent party in presidential power (conservative); PRI = opposition (center); PRD = opposition (left). Note that party in first cell is the president's: the second one is the governor's and the third is the mayor's. PAN-PAN-PAN is the reference category.

Figure 2 Impact of vertical fragmentation of power on criminal violence in Mexico, 2007-2012 

First, the null statistical significance associated with the PAN-PAN-PRI and PAN-PAN-PRD cases shows that PAN-dominated states experienced lower levels of violence than the rest of the country regardless of the party in control of the municipality. This suggests that coordination between the president and his co-partisan governors was crucial to keep criminal violence under control -as the case of Baja California emblematically shows.

Second, compared to states ruled by panista governors, states under the PRI did not always experience higher levels of violence. It was only in cases with the PAN-PRI-PRI structure that violence was slightly higher. Compared to a case of unified governance under the PAN, PRI municipalities in states with PRI governors experienced 23.7 per cent more criminal violence. This suggests that president Calderón was able to reach some level of contingent cooperation with PRI governors to fight criminal violence -as the case of Chihuahua attests.

Third, results unambiguously show that municipalities in states ruled under the leftist PRD experienced significantly higher levels of criminal violence than the rest of the country. Both PRI and PRD municipalities in leftist states (PAN-PRD-PRI and PAN-PRD-PRD) on average experienced 179.6 and 125.9 per cent more violence than municipalities under PAN unified governance. As Figure 2 strikingly shows, the most violent municipalities were opposition municipalities in leftist states. These results suggest that president Calderón did not reach any significant cooperation agreement with leftist governors and that his confrontation with leftist subnational authorities and his decision to strategically abandon them in confronting major waves of criminal attacks resulted in more intense criminal violence -as the case of Michoacán exemplifies.

Finally, results in Model 5 and Figure 2 provide a straightforward test of the coordination and partisan conflict arguments. If the coordination argument is correct in claiming that the intensification of violence was the result of party fragmentation -regardless of which party was in subnational office- we should expect that the two cases in which power was more vertically fragmented (PAN-PRI-PRD and PAN-PRD-PRI) would both experience the highest levels of criminal violence. But this is not the case: The PAN-PRD-PRI layering (IRR = 2.796) yields nearly nine times more criminal violence than the PAN-PRI-PRD layering (IRR = 1.208).

This comparison strongly suggests that in the War on Drugs president Calderón's relations with Mexico's two major opposition parties differed significantly: whereas the president might have achieved some degree of cooperation with the PRI, the party that did not contest the president's electoral victory and that was a likely legislative ally for his economic agenda, the relationship with the Left, the president's nemesis, was marked by confrontation, conflict, and lack of intergovernmental cooperation.

The results from the control variables show that other factors beyond intergovernmental partisan conflict also mattered. Consistent with findings in the sociology of crime, results across models show that municipalities with more young males and a larger percentage of female-headed households -where DTOs recruit foot soldiers from street gangs to fight their wars- experienced more violence. Contrary to the Weberian view that a greater state presence reduces violence, our findings show that municipalities with more state judicial presence (more public prosecutors) tend to be more violent. This potentially speaks of the collusion between local authorities and crime. Finally, consistent with the literature on leadership decapitation, we find that the federal government's success in removing narco leaders stimulated, rather than deterred, inter-cartel violence.

In summary, controlling for a wide variety of political, sociodemographic, law enforcement, and geographic factors, our results unambiguously show that criminal violence was more intense in cities in states ruled by leftist governors -the main political rivals of the conservative president. The fact that our measures of vertical partisan fragmentation remain significant in the presence of these controls indicate that criminal violence in municipalities from leftist states was not the result of particular social characteristics present in these subnational regions.20 The empirical question now becomes why and how intergovernmental partisan conflict between Left and Right led to more intense criminal violence in Mexico's War on Drugs.

Unpacking the federal intervention: Why intergovernmental partisan conflict stimulated criminal violence

While our empirical findings show that inter-cartel violence during the War on Drugs was disproportionately higher in cities in states ruled by the Left, the statistical models do not explain why. To explain this connection, we need to understand the nature of the federal intervention, how intergovernmental partisan conflict impacted the War on Drugs, and the cartel's violent responses to intergovernmental conflict.

The War on Drugs: Explaining the federal intervention

President Calderón's War on Drugs and the federal intervention can be divided into two phases: the initial deployment of the army and the federal police to the country's most conflictive regions in 2006-2007 and the development of a more comprehensive security intervention to respond to the cartel's virulent response from 2008 until 2012.

The initial deployment of the armed forces in December 2006 responded to a governance imperative. By declaring war on the country's DTOs, the president was hoping to shift public opinion away from the polarizing postelection crisis toward the country's rising security crisis. Rallying the nation behind the armed forces and their commander-in-chief to destroy the country's powerful drug cartels seemed to be a "valence issue" that would help the president overcome the political crisis. Hence, the initial deployment of the armed forces did not discriminate along partisan lines. It began in the leftist state of Michoacán -the president's home state- and then spread to the rest of the country's most conflictive regions. President Calderón believed that the War on Drugs would be a relatively easy military victory and a quick fix to a major post-election crisis (Aguilar and Castañeda, 2009).

However, the military/federal police interventions to recover territorial control and undermine DTOs resulted in an unexpected criminal backlash and the dramatic escalation of violence (Merino, 2011; Espinosa and Rubín, 2015). The government's success at decapitating the leadership of the country's main DTOs led to major spirals of intra and inter-cartel violence and to attacks against security officers and local public authorities. As violence skyrocketed and several cities and regions experienced the outbreak of "violence epidemics", it became evident that the War on Drugs was not a valence but a divisive issue.

By the end of 2008 "managing violence", rather than suppressing it, became the new government objective. President Calderón and his team were no longer Weberian state officials trying to establish the monopoly of violence within a given territory but opportunistic politicians trying to adopt a new strategy of damage controls: accepting responsibility for the violence in areas where president Calderón could persuade his subnational co-partisans to follow him in confronting DTOs while elsewhere blaming the violence on his political rivals21 -particularly leftist subnational authorities, who did not recognize him as a legitimate president, bitterly opposed his economic agenda, and were poised to become his main electoral challengers in the 2009 midterm legislative election.

The key question is whether the partisan turn in the War on Drugs affected the federal intervention and whether intergovernmental partisan conflict stimulated criminal violence. To address this question we compare the intervention in three states, focusing on three cities: Baja California (Tijuana), Michoacán (Apatzingán), and Chihuahua (Ciudad Juárez). These cases respectively represent three different structures of partisan vertical integration: PAN-PAN-PAN, PAN-PRD-PRD, and PAN-PRI-PRI. These cities are large urban centers that simultaneously function as major drug trafficking hubs and home to powerful DTOs. In all three cases the federal government launched a military intervention to quell inter-cartel wars.

We assess how partisanship shaped the federal intervention along four dimensions: 1) military/police cooperation; 2) judicial prosecution of local authorities; 3) the federal government communication strategy of attributing responsibilities for growing criminal violence; and 4) the federal assistance to help subnational authorities deal with the multiplication of violence. Our goal is to assess whether variation on these four dimensions contributed to quell or stimulated criminal violence.

Baja California: Tijuana

The federal (PAN) intervention in the city of Tijuana (PAN) in the northwestern state of Baja California (PAN) illustrates how effective intergovernmental coordination between federal and subnational co-partisans resulted in law enforcement actions that facilitated the control of an unprecedented spiral of inter-cartel violence (Sabet, 2012; Durán-Martínez, 2015). The intervention in Baja California also shows that intergovernmental coordination is not an automatic or unproblematic process; it reveals, however, that shared electoral incentives can motivate co-partisans from different levels of governments to effectively work together and to recover the state's authority over territories previous under the control of drug cartels.

Home of the Tijuana Cartel, Baja California experienced the outbreak of the first major inter-cartel war in Mexico after the PAN gubernatorial victory in 1989.22 Since then, the Sinaloa Cartel has fought a protracted war against the leaders of the Tijuana Cartel, the Arellano Félix brothers, for the control of Tijuana, the main entry point into California. Violence grew to new levels in 2007 after the federal government deployed the army to control inter-cartel violence and then reached levels never before experienced after the arrest of Eduardo Arellano Felix, then Tijuana's criminal boss, in 2008 (Guerrero, 2012). Besides a surge in inter-cartel executions, the state experienced a major wave of kidnappings, car robberies, and the assassination of state and local authorities and police officers throughout 2008.

Although the federal government had deployed thousands of military personnel to Baja California in 2007, the incoming PAN governor Jose G. Millán (2007-2013) approached general Sergio Aponte, the chief of the second military zone stationed in Baja California, for additional assistance in confronting the state's new security crisis. Aponte agreed and named high-ranking members of his battalion to serve as state-level and municipal-level police chiefs. After a few months in office, however, Aponte published a manifesto in the state's leading news outlets denouncing widespread corruption and collusion with organized crime in the state attorney's office, the state judicial police, and the secretariat of public security and in the police forces of Tijuana and other major cities. The manifesto named over fifty high and mid-ranking officials, mostly from PAN administrations.23

Facing the potential of a major political scandal in a PAN stronghold, the federal government developed a comprehensive security intervention for Baja California. To protect his co-partisans, president Calderón transferred general Aponte to another state and appointed general Alfonso Duarte as new military zone commander. Governor Millán, in turn, appointed general Duarte as coordinator of all of the state's police forces and with the support of the federal government removed corrupt state-level officials without prosecution to prevent any major media scandal. Jorge Ramos, the panista mayor of Tijuana and his secretary of public security, colonel Julián Leyzaola, followed a similar strategy when they purged the city police. Coordination in these cases meant the strategic and selective management of corruption and the silent removal of corrupt officers linked to PAN administrations; the goal was to avoid voters' punishment for corruption and to signal to the Tijuana and Sinaloa cartels that the authorities were unified in their actions.

With resolute federal support, general Duarte and colonel Leyzaola adopted a coordinated iron-fist policy involving 1) the arrest or elimination of drug lords and the leaders of private militias, 2) major seizures of drug shipments, and 3) the confiscation of weapons. Yet, as different human rights ngos and the Human Rights National Commission (CNDH) reported, Baja California's militarized police systematically relied on torture and violated the due process in conducting many of these arrests and investigations (Sabet, 2012). Despite accusations of systematic and generalized human rights violations, the federal government supported Duarte's and Leyzaola's actions and provided renewed military assistance when their civilian bosses became target of criminal attacks.

With full federal assistance, including generous economic transfers, governor Millan and successive Tijuana mayors from the PAN and the PRI launched a series of economic and social investment programs targeted at the city's youth. The goal was to restore economic activity in Tijuana, rebuild the city's public image, open new business opportunities for the youth, and prevent young men from continuing to join the gangs that were eventually absorbed by DTOs to fight turf wars.

The federal intervention in Baja California and Tijuana succeeded in reducing a major spiral of criminal violence to pre-crisis levels because federal, state and local authorities were able to engage in coordinated action to implement a broad security agenda (Durán-Martínez, 2015). The military intervention was effective because a new and relatively cleaner Tijuana police provided crucial information to military and federal police officers that facilitated multiple arrests and drug seizures. The federal protection the Tijuana mayor received when his administration had to confront a major criminal backlash prevented DTOs from capturing the local government and empowered the mayor to conduct additional police purges. Finally, the coordinated social policy actions provided opportunities for economic and social mobility to young males whose only attractive option had been to join the cartels' private militias.

Without widespread local police protection, the Tijuana Cartel became weaker. Without extensive access to the local youth, the cartel's private militias became less effective. Aligned partisan authorities working together encouraged rival cartels to leave the city and the state and avoid further turf wars. Although this coordinated strategy deactivated an epidemic of violence, Tijuana continued to be a very violent city because the drug smuggling business into the United States remained profitable, assault weapons from the United States continued to flow into Mexico (Dube, Dube, García-Ponce, 2013), and the Tijuana Cartel continued to enjoy some amount of state-level protection from units that the federal and state governments left unpurged in order to avoid a major political scandal.

Michoacán: Apatzingán

The case of Michoacán on the Mexican Pacific coast and the city of Apatzingán -the center of economic and political activity in the southern part of the state- illustrate how intergovernmental conflict between federal (conservative) and (leftist) subnational authorities and the partisan use of law enforcement by the federal government led to previously unknown levels of criminal violence and facilitated the rise of the Knights Templar -a leading local cartel- as the de facto authority in a large swath of the state.

Inter-cartel wars in Michoacán began after the first (leftist) opposition victory in the state in 2002, when the Gulf Cartel and its private army, the Zetas, ventured into Michoacán to contest the drug trafficking monopoly of the Valencia brothers24 -and their allies, including the Arellano Félix brothers- whose expansion had taken place under the protection of government officials from the PRI administration of Victor Manuel Tinoco Rubí (Maldonado, 2012). After the Zetas and their local allies, La Familia Michoacana, defeated the Valencia brothers and forced them out of the state, La Familia and the Zetas went to war with each other for the state's criminal hegemony. Between 2005 and 2006 violence escalated to new levels and Apatzingán and its surrounding areas -the Apatzingán Valley and the Tierra Caliente region- became the epicenter of turf wars.

The first federal intervention of president Calderón's administration occurred at the request of leftist governor Lázaro Cárdenas Batel in December 2006. This shows that despite political rivalries PRD subnational officials did not hesitate to demand federal assistance when they faced major criminal threats. In fact, cooperation between the federal government and governor Cárdenas (and later with his leftist successor, governor Leonel Godoy) facilitated an initial military success. This joint military/police effort weakened the Zetas, whose many members had to return to the northeastern state of Tamaulipas to defend their own territory from another military intervention (Guerrero, 2011), and undermined key players in La Familia's extensive drug production and distribution networks.

After the initial military success, however, La Familia launched a major counterattack and in 2008 perpetrated a terrorist attack against the civilian population gathered in downtown Morelia, the state's capital, during the Independence Day celebrations and then blamed it on the Zetas. Even though governor Godoy had a good working relationship with president Calderón throughout 2007, and even though he and former governor Cardenas Batel had shared with the president their concern that 10 per cent of the state's municipal authorities were on La Familia's payroll,25 intergovernmental cooperation broke down after the terrorist attack. From that point on, the federal government did not share any security information with the state government and the military did not coordinate security actions with state or municipal police forces and not only did not launch any joint economic or social intervention in the state but actually significantly reduced the state's budget.26

Intergovernmental conflict reached a high point when the federal government unilaterally launched a mega arrest of twelve mayors from the three main political parties - most of them from the Apatzingán Valley and the Tierra Caliente regions - and twenty-three top- and mid-level officials of Governor Godoy's security cabinet in May 2009, seventy-two hours before the beginning of the campaign for the mid-term federal election. These officials were charged with protecting La Familia. Governor Godoy did not find out about the operation until federal police forces were breaking into the state's capitol. Following the arrest, the federal government launched a major national media campaign to frame the governor and the Left as corrupt and inept. A year later, however, all but one of the allegedly corrupt officials had been liberated, and a report by the National Human Rights Commission (CNDH) concluded that most of the arrests had been illegal and had violated the due process (El Economista, 2009).

Security officials argue that the federal government did not inform governor Godoy because they knew his half-brother was on La Familia's payroll and had facilitated the cartel's infiltration of state and local authorities.27 Research for this intervention took place throughout 2008 and by January 2009 federal authorities were ready to conduct the arrest, but president Calderón asked them to wait until they were sure that the cases rested on sound judicial grounds. Federal authorities claim that although every case was backed by solid evidence, corrupt judges let all but one of the indicted officers go free.28

After the mega arrest, intergovernmental cooperation between the armed forces and the federal police and the state and local authorities broke down, and federal and subnational authorities entered into an era of bitter conflict. The federal intervention in the Apatzingán Valley and Tierra Caliente was no longer a joint effort. According to national officials, the federal government decided not to cooperate at all with corrupt authorities who had been liberated by a corrupt judicial system.29 As reported by governor Godoy, after the mega arrest the relationship with the federal government was marked by distrust, personal animosity, and political rivalry.30

Genaro Guízar, the leftist mayor of Apatzingán -one of twelve mayors arrested in 2009- bitterly bewailed the absence of any intergovernmental cooperation after his return from prison: "When a major kidnapping occurred in the city, [the military and the federal police] took charge of the situation and didn't bother informing me" (Animal Político, 2011). His recollection of the military occupation of Apatzingán in December 2010, when the federal government allegedly killed Nazario Moreno, La Familia's top leader, is harshly critical. He claimed that "The federal police raped young girls and violated human rights" (Ferrer y Martínez, 2010). Without proper local information, the federal government believed, and announced on national television, that they had killed Nazario Moreno. However, the drug lord had survived the attack, gone underground and revamped La Familia under a new name: the Knights Templar.

After the mega arrest amid the growing tensions between the federal security forces and leftist subnational authorities, the Knights Templar took advantage of the increasing vulnerability of mayors in the Apatzingán Valley and the Tierra Caliente region and sought to capture local governments through lethal coercion and establish new forms of criminal governance in the cities, seizing control of municipalities and their local budgets and taking control of local businesses (e.g., lime and avocado producers) and intimidating citizens via extortion and kidnapping. From 2009 onward the state plunged into a wave of criminal attacks against local authorities that reached an unprecedented level of forty attacks during the 2011 state elections.31

During the contested 2011 gubernatorial and municipal elections, in which Luisa María Calderón, the president's sister, ran for gubernatorial office, polarization between federal and leftist state authorities became particularly acute. The Knights Templar took advantage of this conflict and through coercion and the assassination of political candidates and party activists tried to influence the election campaigns and the election outcome. After the election, the Knights were able to capture the state government and a large number of the state's municipalities; they looted municipal coffers, expanded criminal taxation to businesses and local citizens, and demanded social obedience. When leftist mayors who did not want to surrender their budgets to the Knights Templar requested federal protection, national authorities simply did not respond. One striking case is that of Ygnacio López, a reputed town doctor, social activist, and leftist mayor of Santa Ana Maya, who was killed after he went on a hunger strike to protest the bankruptcy of his municipality.32

The breakdown of intergovernmental cooperation meant that federal authorities no longer had access to local information and that local authorities no longer had proper federal protection against criminal attacks. La Familia and the Knights Templar capitalized on these conflicts and rapidly moved to take over local governments and populations via lethal violence. Unlike in Baja California, where intergovernmental cooperation allowed the federal government to partially recover the state's monopoly on violence and reestablish order, in Michoacán the punishment strategy the federal government used against leftist subnational authorities opened the way for the Knights Templar to capture state and municipal powers and become the state's de facto rulers.

The case of Michoacán is not unique. The federal intervention in the leftist state of Guerrero followed a similar logic of punishment against sub-national authorities. When the leftist mayor of Acapulco, Félix Salgado (2005-2008), requested federal assistance after receiving death threats from criminal organizations, federal authorities accused him on national television of protecting organized crime (Pacheco, 2007). Although they publicly retracted two weeks later, Salgado, and his leftist successors became vulnerable to criminal attacks. Confronted with the federal government's selective judicial prosecution and without proper federal protection, PRI and PRD mayors in Guerrero were incapable of defending their municipalities from criminal takeovers (Trejo and Ley, 2015).

In the northern state of Zacatecas, the federal government did not assist the leftist governor, Amalia García, in avoiding the spillover effect from the epidemics of violence in neighboring Tamaulipas.33 Unlike in Aguascalientes and Guanajuato, where federal authorities worked closely with PAN governors to avoid spillover effects from violence in neighboring states, federal authorities accused the Zacatecas governor of corruption and ineptitude when federal prisoners who had been relocated in the state escaped. The Zetas -the lethal dto from Tamaulipas- easily gained control over important parts of the state.

Chihuahua: Ciudad Juárez

The federal intervention in Ciudad Juárez (PRI) in the northern state of Chihuahua (PRI) illustrates a case of contingent cooperation: one in which exceptional circumstances drove the federal government to collaborate with the PRI but in which president Calderón did not adopt a strategy to punish PRI subnational authorities -as he did against the Left in Michoacán. It shows that contingent intergovernmental cooperation contributed to bring down an epidemic of criminal violence to pre-crisis levels in Juárez.

After more than a decade of turf wars between the Juárez and the Sinaloa cartels, criminal violence spiked in 2008, leading governor José Reyes Baeza and Juárez's Mayor José Reyes Ferriz to request federal intervention (Zubia, 2008). As in other parts of the country, the federal intervention led to a major criminal backlash and to the unprecedented escalation of criminal violence that turned Juárez into the world's most dangerous city. Federal and subnational authorities blamed the escalation of violence on each other and the initial intergovernmental cooperation broke down (González, 2008).

Yet, the mishandling of a civilian massacre in the working-class neighborhood of Villas de Salvarcar, in which president Calderón publicly criminalized a group of fifteen high school students who had been tragically killed by a group of cartel hit men who confused them with a rival gang, triggered a wave of civil outrage that forced Calderón to refashion the federal intervention (Aziz, 2012). Although the president initially bypassed state and local governments in developing the early stages of "Todos Somos Juárez" (We Are All Juárez) -a key social intervention that involved a major expansion of public infrastructure, transportation, day care centers, and cultural opportunities for young students- mounting pressure from civil society forced the federal government to cooperate with the PRI subnational authorities in the implementation of the program.34

Even though the national attorney had gathered important evidence that governor Reyes Baeza and the state attorney had provided protection to the Juárez Cartel (Lagunas, 2012), the federal government strategically declined to prosecute them or leak information to the media -as it did in Michoacán- and decided instead to cooperate with the incoming PRI state governor -as it did in Baja California. In this new scheme of cooperation, federal authorities were instrumental in persuading Héctor Murgma, the new mayor of Juárez, to appoint colonel Julián Leyzaola -the former Tijuana police chief- as head of the Juárez police. After a rough start in which Murgma, his security staff, and Leyzaola survived separate attacks from federal forces (Luján, 2011) and the municipality's federal transfers for security reform were suspended, the federal government eventually supported Leyzaola's effort to purge the local police and worked together with city authorities to overcome the crime epidemic through a combination of iron-fist policies, police reforms, and an extensive social intervention.

While civilian mobilization played an important role in demanding intergovernmental coordination, cooperation between federal and subnational authorities took place because it involved governors and mayors from the PRI -the party that did not question president Calderón's election and that was a likely legislative partner for the president's agenda of market-oriented reforms- rather than from the leftist PRD. This contingent cooperation paid off: The intelligence information shared between reformed local police forces and the military and federal police facilitated the decapitation of cartels and private militias and the seizures of drug shipments and weapons, and the adoption of an extensive program of public goods provision contributed to depleting the cartel's private militias. These actions weakened the Juarez and Sinaloa cartels and empowered local governments to resist the violent attempts to capture local governments and civil society- as the Knights Templar succeeded in doing in Michoacán. It must be said, however, that although the contingent cooperation between PAN and PRI authorities brought the epidemic of violence to an end in Juárez, the city remained a relatively violent place for the same reasons that Tijuana remained a conflictive urban center.

The contingent cooperation between PAN federal authorities and PRI subnational officials that contributed to the de-escalation of an epidemic of violence in Juárez, also took place in Monterrey, in the northern state of Nuevo León, where powerful business associations and different civil society groups forced federal and PRI subnational authorities to cooperate in bringing violence under control (Conger, 2014). Note that in other states where there was no civil society pressure, PAN federal authorities did not assist PRI governors and mayors in confronting dtos -as the case of Tamaulipas emblematically shows. While there was no cooperation in Tamaulipas, it is noteworthy that the federal government did not punish PRI subnational officials through selective prosecution or exposing them as corrupt in national media, despite widespread evidence that linked former PRI governors and mayors with the Gulf Cartel or the Zetas.

Conclusions

A little over fifty years ago, in his seminal study of federalism, William Riker introduced the importance of political parties into the study of intergovernmental cooperation and suggested that vertical party alignment produced more coherent and efficient policies than vertically fragmented federations. Building on this argument, students of drug violence in Mexico have argued that the inability of federal and subnational opposition governments to logistically coordinate their actions to fight drug cartels explains the six-fold increase in violence since the 2007 federal intervention in the War on Drugs.

In this article we have challenged the coordination argument and suggested that the uneven spread of inter-cartel violence after the federal intervention was not the result of logistical differences between governments from different parties. We have presented extensive quantitative and qualitative evidence showing that intergovernmental partisan conflict between Right and Left led conservative federal authorities to develop cooperative military and policing interventions in subnational regions where the president's co-partisans ruled but confrontational and biased interventions where his main political enemies were in power. Through case studies we have shown that the partisan use of the army and federal forces and the politicization of law enforcement not only contributed to a major increase of inter-cartel and criminal violence in leftist subnational regions but also made opposition local authorities and municipalities more vulnerable to criminal attacks and even criminal capture of governments and societies.

Our empirical findings have three important theoretical implications for the study of governance and criminal violence in federations.

First, by introducing partisanship and partisan conflict to our understanding of intergovernmental relations, we have shown that electoral incentives can define policy outcomes in federations. While students of distributive politics have long recognized the important role of electoral incentives in the allocation of public spending, scholars of security and policing have only recently begun to recognize the partisan use of law enforcement. Consistent with theories of distributive politics and with Wilkinson's important finding of the partisan use of police forces to control inter-religious violence in India, we have provided extensive evidence of the partisan use of law enforcement in Mexico's War on Drugs -of the political incentives that led Mexico's president to deploy federal forces to protect his subnational co-partisans facing major surges in inter-cartel violence but to punish and leave unprotected his enemies facing similar conditions.

Second, our findings challenge the widely held Weberian assumption that government officials will always want to adopt policies that maximize the state's monopoly of violence. Consistent with Wilkinson's and Staniland's findings for South Asia and with those of Acemoglu, Robinson, and Santos and Auyero for South America, our results show that under conditions of acute political polarization between Right and Left, conservative federal authorities in Mexico sought to "manage" criminal violence for electoral purposes. They purposefully devised military and police, judicial, and communications interventions to protect the president's subnational co-partisans and coordinate security policy with them but deliberately left the president's political enemies unprotected, refused to cooperate with them, and then publicly blamed them for corruption and ineptitude. This finding strongly suggests that the generalized assumption that state agents will always seek to monopolize violence is unwarranted: under conditions of acute political polarization state agents can reward their allies but tolerate private violence to punish their political enemies.

Finally, beyond theory, our findings have important policy implications for Mexico. Contrary to the coordination argument, which has led scholars and policymakers to cast doubt on the efficacy of Mexican federalism and emphasize a centralist approach to policing, our findings point to the crucial need for effective checks and balances to constrain the discretionary use of military and police forces and law enforcement agencies by national executives in federations. Whereas the centralist temptation in countries such as Mexico has led national officials to single out municipal authorities and local police forces as corrupt enemies in the War on Drugs, our findings suggest that unchecked federal power contributed in important ways to the intensification of inter-cartel violence, to the rapid institutional erosion of municipal governments, and to the capture of local institutions by drug lords and organized crime. Imposing stricter controls on Mexican federal authorities to prevent presidents and federal prosecutors from ever again using the federal police and the army to punish political rivals and to assist co-partisans would be an absolutely essential institutional reform in establishing the rule of law and ending Mexico's unprecedented waves of criminal violence.

REFERENCES

Acemoglu, Daron, James A. Robinson y Rafael J. Santos (2013), "The Monopoly of Violence: Evidence from Colombia", Journal of the European Economic Association 11(1), pp. 5-44. [ Links ]

Aguilar, Rubén y Jorge Castañeda (2009), El narco: La guerra fallida Mexico, Punto de Lectura. [ Links ]

Animal Político (2011), "Los medios están satanizando a Apatzingán: Genaro Guízar, alcalde", 5 de agosto, disponible en: Animal Político (2011), "Los medios están satanizando a Apatzingán: Genaro Guízar, alcalde", 5 de agosto, disponible en: http://www.animalpolitico.com/2011/08/%E2%80%9Clos-medios-estan-satanizando-a-apatzingan%E2%80%9D-genaro-guizar-alcalde/ [fecha de consulta: 7 de enero de 2015]. [ Links ]

Aparicio, Javier (2009), "Análisis estadístico de la elección presidencial de 2006: ¿Fraude o errores aleatorios?", Política y Gobierno (volumen temático), pp. 225-243. [ Links ]

Auyero, Javier (2006), "The Political Makings of the 2001 Lootings in Argentina", Journal of Latin American Studies 38(2), pp. 241-265. [ Links ]

Aziz, Alberto (2012), "Violencias en el norte de México: El caso de ciudad Juárez", Revista Iberoamericana 12(48), pp. 143-155. [ Links ]

Bruhn, Kathleen y Kenneth Greene (2007), "Elite Polarization Meets Mass Moderation in Mexico's 2006 Elections", PS: Political Science and Politics 40(1), pp. 33-38. [ Links ]

Calderón, Felipe (2014), Los retos que enfrentamos México, Random House. [ Links ]

Calderón, Gabriela, Gustavo Robles, Alberto Díaz-Cayeros y Beatriz Magaloni (2015), "The Beheading of Criminal Organizations and the Dynamics of Violence in Mexico," Journal of Conflict Resolution (en prensa). [ Links ]

Consejo Nacional de Población (Conapo) (2006), Indice de marginación 2005 Colección Índices Sociodemográficos, México, Conapo. [ Links ]

Consejo Nacional de Población (Conapo) (2011), Indice de marginación 2010 Colección Índices Sociodemográficos, México, Conapo. [ Links ]

Conger, Lucy (2014), "The Private Sector and Public Security: The Cases of Ciudad Juarez and Monterrey", disponible en: Washington, Woodrow Wilson Center, Conger, Lucy (2014), "The Private Sector and Public Security: The Cases of Ciudad Juarez and Monterrey", disponible en: Washington, Woodrow Wilson Center, https://www.wilsoncenter.org/sites/default/files/conger_private_sector.pdf [fecha de consulta: 3 de enero de 2015]. [ Links ]

Cox, Gary y Mathew McCubbins (1986), "Electoral Politics as a Redistributive Game", Journal of Politics 48(2), pp. 370-389. [ Links ]

Dell, Melissa (2011), "Trafficking Networks and the Mexican Drug War", American Economic Review (en prensa). [ Links ]

Dickenson, Matthew (2014), "The Impact of Leadership Removal on Mexican Drug Trafficking Organizations," Journal of Quantitative Criminology 30(4), pp. 651-676. [ Links ]

Dube, Arindrajit, Oeindrila Dube y Omar García-Ponce (2013), "Cross-Border Spillover: U.S. Gun Laws and Violence in Mexico", American Political Science Review 107(3), pp. 397-417. [ Links ]

Durán-Martínez, Angélica (2015), "To Kill and Tell? State Power, Criminal Competition, and Drug Violence", Journal of Conflict Resolution (en prensa). [ Links ]

El Economista (2009), "Detención de alcaldes michoacanos fue ilegal: CNDH", 29 de octubre. [ Links ]

Espinosa, Valeria y Donald Rubin (2015), "Did the Military Interventions in the Mexican Drug War Increase Violence?", The American Statistician69(1), pp. 17-27. [ Links ]

Fearon, James y David Laitin (2003), "Ethnicity, Insurgency and Civil War", American Political Science Review 97(1), pp. 75-90. [ Links ]

Ferrer, Eduardo y Ernesto Martínez (2010), "Oleada de asaltos en Apatzingán..." La Jornada 14 de diciembre. [ Links ]

Gibson, Edward (2004), "Federalism and Democracy", en E. Gibson (ed.), Federalism and Democracy in Latin America Baltimore, Johns Hopkins University Press. [ Links ]

González, Félix A. (2008), "Estado a la federación: Hay que esclarecer ejecuciones", Norte de Ciudad Juárez 19 de agosto. [ Links ]

Grillo, Ioan (2012), El Narco: Inside Mexico's Criminal Insurgency Nueva York, Bloomsbury Press. [ Links ]

Guerrero, Eduardo (2011), "La raíz de la violencia", Nexos 1 de junio. [ Links ]

______ (2012), "Epidemias de violencia", Nexos, 1 de julio. [ Links ]

INEGI (Instituto Nacional de Estadística y Geografía) (2000), XII Censo General de Población y Vivienda, México, INEGI. [ Links ]

______ (2009), Censo de Gobierno, Seguridad Pública y Sistema Penitenciario, México, INEGI. [ Links ]

______ (2010), Censo de Población y Vivienda, México, INEGI. [ Links ]

______ (2013), Censo de Gobierno, Seguridad Pública y Sistema Penitenciario, México, INEGI. [ Links ]

Klesner, Joseph (2005), "Electoral Competition and the New Party System in Mexico", Latin American Politics and Society 47(2), pp. 103-142. [ Links ]

Lagunas, Icela (2012), "Otro gobernador en la mira", Reporte Indigo 30 de julio, disponible en: 30 de julio, disponible en: http://www.reporteindigo.com/reporte/mexico/otro-gobernador-en-la-mira [fecha de consulta: 7 de enero de 2015]. [ Links ]

Langston, Joy (2007), "The PRI's 2006 Electoral Debacle", PS: Political Science and Politics 40(1), pp. 21-25. [ Links ]

Lessing, Benjamin (2012), "The Logic of Violence in Criminal War", tesis de doctorado, University of California, Berkeley. [ Links ]

Ley, Sandra (2014), "Citizens in Fear: Political Participation and Voting Behavior in the Midst of Violence", tesis de doctorado, Duke University, Durham. [ Links ]

Luján, Francisco (2011), "Ataque a Leyzaola rompe relación municipio-federación", Norte de Ciudad Juárez 27 de julio. [ Links ]

Magaloni, Beatriz (2006), Voting for Autocracy Cambridge, Cambridge University Press. [ Links ]

Maldonado, Salvador (2012), "Drogas, violencia y militarización en el México rural. El caso de Michoacán", Revista Mexicana de Sociología 74(1), pp. 5-39. [ Links ]

Martínez, Julieta (2008), "Carta íntegra del comandante Sergio Aponte Polito", El Universal 23 de abril. [ Links ]

Merino, José (2011), "Los operativos conjuntos y la tasa de homicidio. Una medición", Nexos junio 1. [ Links ]

Moreno, Alejandro (2010), "Changing Ideological Dimensions of Party Competition in Mexico, 1990-2006", en Y. Esmer y T. Pettersson (eds.), The International System, Democracy and Values Uppsala, Uppsala University. [ Links ]

Moreno, Alejandro y Roberto Gutiérrez (2005), "Prevén triunfo perredista", Reforma 30 de mayo. [ Links ]

Oates, Wallace (1972), Fiscal Federalism Nueva York, Harcourt. [ Links ]

Osorio, Javier (2013), "Hobbes on Drugs: Understanding Drug Violence in Mexico", tesis de doctorado, University of Notre Dame, Notre Dame. [ Links ]

Pacheco, Ossiel (2007), "Diputados perredistas critican amenazas de muerte contra Salgado Macedonio", La Jornada Guerrero 11 de febrero. [ Links ]

Presidencia de la República (2006), "Presentación del Gabinete de Seguridad", 30 de noviembre, disponible en: http://calderon.presidencia.gob.mx/2006/11/presentacion-del-gabinete-de-seguridad/ [fecha de consulta: 7 de febrero de 2015]. [ Links ]

Reforma/redacción (2006), "Libra amlo el desafuero," Reforma 5 de enero. [ Links ]

Reforma (2013), "Pide edil ayuda y lo matan", 8 de noviembre. [ Links ]

Remes, Alain (1999), "Gobiernos yuxtapuestos en México: Hacia un marco analítico para el estudio de las elecciones municipales", Política y Gobierno VI(1), pp. 225-253. [ Links ]

Riker, William (1964), Federalism: Origins, Operations and Significance Boston, Little Brown. [ Links ]

Riker, William H. y Ronald Schaps (1957), "Disharmony in Federal Government", Behavioral Science 2(4), pp. 276-290. [ Links ]

Ríos, Viridiana (2012), "Why did Mexico Become so Violent? A Self-reinforcing Violent Equilibrium Caused by Competition and Enforcement", Trends in Organized Crime 16(2), pp. 138-155. [ Links ]

______ (2015), "How Government Coordination Controlled Organized Crime: The Case of Mexico's Cocaine Markets", Journal of Conflict Resolution (en prensa). [ Links ]

Sabet, Daniel (2012), Police Reform in Mexico Stanford, Stanford University Press. [ Links ]

Sampson, Robert (1985), "Neighborhood and Crime: The Structural Determinants of Crime Victimization", Journal of Research in Crime and Delinquency 22(1), pp. 7-40. [ Links ]

Schedler, Andreas (2007), "The Mexican Standoff: The Mobilization of Distrust," Journal of Democracy 18(1), pp. 88-102. [ Links ]

Snyder, Richard y Angélica Durán-Martínez (2009), "Does Illegality Breed Violence? Drug Trafficking and State-Sponsored Protection Rackets", Crime Law and Social Change 52, pp. 253-273. [ Links ]

Staniland, Paul (2012), "Beyond the Monopoly of Violence: Militancy and the State in Pakistan", manuscrito, Chicago, University of Chicago. [ Links ]

Steele, Abbey (2011), "Electing Displacement: Political Cleansing in Apartado, Colombia", Journal of Conflict Resolution 55(3), 423-445. [ Links ]

Trejo, Guillermo y Sandra Ley (2014), "Votes, Drugs, and Violence: Subnational Democratization and the Onset of Inter-Cartel Wars in Mexico, 1995-2006", manuscrito, University of Notre Dame. [ Links ]

Trejo, Guillermo y Sandra Ley (2014a), "Mexico's Drug Wars and the Re-making of Local Order: Why Criminal Organizations Murder Local Officials", en Agustina Giraudy, Eduardo Moncada y Richard Snyder, Subnational Analysis in Comparative Politics (en prensa). [ Links ]

Trejo, Guillermo y Sandra Ley (2015), "Municipios bajo fuego", Nexos 1 de febrero. [ Links ]

Urrusti Frenk, Sinaia (2012), "La violencia como consecuencia de la falta de coordinación política", en J.A. Aguilar (coord.), Las bases sociales del crimen organizado y la violencia en México México, Secretaría de Seguridad Pública. [ Links ]

Villarreal, Andrés (2002), "Political Competition and Violence in Mexico", American Sociological Review 67(4), pp. 477-498. [ Links ]

Weber, Max [1918] (1946), "Politics as a Vocation", en H.H. Garth y C. Wright Mills (eds.) Essays in Sociology Nueva York, Macmillan. [ Links ]

Weingast, Barry (2014), "Second Generation Fiscal Federalism: Political Aspects of Decentralization and Economic Development", World Development, 53, pp. 14-25. [ Links ]

Wilkinson, Steven (2005), Votes and Violence: Electoral Competition and Ethnic Riots in India Cambridge, Cambridge University Press. [ Links ]

Zubia García, Ángel (2008), "Confía Reyes Baeza en el apoyo federal", Norte de Ciudad Juárez 20 de febrero. [ Links ]

*We presented preliminary versions of this paper at the 2014 Midwest Political Science Association Meeting and the 2014 American Political Science Association Meeting. We are grateful to Regina Bateson, Matthew Kocher, Fausto Hernández, Mario Chacón, and three anonymous reviewers for insightful suggestions. Sebastián Garrido provided statistical advice. Funding from the Kellogg Institute for International Studies (University of Notre Dame) and from the Trent Foundation (Duke University) enabled our data collection. Mario Moreno and Valeria Ramírez provided superb research assistantship and the Parametría team invaluable support. We benefitted from Caroline Domingo's editorial assistance and from the able guidance of the editors of this special issue. Errors and omissions remain our own.

1For scholarly arguments, see Urrusti (2012) and Ríos (2015). For the government's view, see Calderón (2014).

2For a critical review of Riker, see Gibson (2004).

3For a pioneering test of the core-voter model in the allocation of antipoverty resources in Mexico, see Magaloni (2006).

4In May 2005, López Obrador had 36 per cent of vote intentions, Roberto Madrazo (PRI) 25 per cent, and Santiago Creel (PAN) 24 per cent. See Moreno and Gutiérrez (2005).

5Approximately one million citizens marched in downtown Mexico City to contest the impeachment of López Obrador. See Reforma staff (2006).

6Based on elite and general population surveys, Bruhn and Greene (2007) show that political elites became bitterly polarized during Mexico's 2006 election but voters did not.

7Although president Fox and all political parties consistently violated Mexico's electoral laws during the electoral campaign, there was no compelling evidence of fraud in the counting of the votes. For a thorough analysis of the quality of the election count, see Aparicio (2009).

8CVM contains information on drug-related violent events reported in the Mexican daily newspaper Reforma. Based on Mexico City and Monterrey, and with extensive coverage of central and northern Mexico, Reforma is the most specialized source of daily information on drug trafficking in Mexico. While CVM does not provide a census of drug-related violence, its focus on the main conflict zones minimizes sources of geographic bias.

9For a detailed assessment of these conflicts, see Grillo (2012).

10Unlike official homicide statistics, which do not distinguish murders undertaken by criminal groups from those undertaken by individual actors, CVM only reports murders that can be attributed to DTOs. When the newspaper report did not explicitly attribute a murder to a DTO, we relied on three indicators to decide whether this was actually a case of organized crime or not: 1) the use of assault weapons; 2) signs of torture and brutal violence (e.g., bodies wrapped in a rug or mutilated); and 3) written messages left on the bodies.

11We exclude 418 municipalities from Oaxaca that select their mayors through indigenous customary practices and where political parties do not participate in municipal elections.

12We rely on CVM because the government database only covers the first four years of the Calderón administration (2007-2010).

13Four small parties played important political roles during this period. Because all except for one party fielded candidates for office in coalition with the three big parties, we subsumed the small parties into the three major ones (e.g., the Green Party under the PRI and the Workers' Party under the PRD). In the few cases of PAN-PRD coalition, we tracked the candidate's own party and assigned it as the incumbent force.

14Mexico's Federal District, which has a special administrative status, is excluded from the analysis.

15For the pioneering analysis on vertically fragmented power in the Mexican federation, see Remes (1999).

16Villarreal (2002); Dube, Dube and García-Ponce (2013); Dell (2011); and Osorio (2013) have tested for the impact of municipal electoral competition on homicides and organized crime violence, and Snyder and Durán-Martínez (2009) and Trejo and Ley (2014) have emphasized the importance of state-level electoral competition and alternation in power. For electoral data we rely on cidac and on official electoral data published by Mexico's former state electoral institutes.

17We base our calculation on the national censuses by INEGI, in 2000 and 2010. Relying on the Government, Public Security, and Justice Censuses conducted by INEGI between 2009 and 2013, we also ran models using the number of police per 10 000 population and results remain unchanged. We use prosecutors because the police information is incomplete.

18See also Dickenson (2014), Durán-Martínez (2015) and Calderón et al. (2015). For important alternative views of the impact of law enforcement on criminal wars, see Lessing (2012) and Osorio (2013).

19This information is based on a systematic keyword search in four Mexican national newspapers (Reforma, El Universal, El Financiero, and Excelsior) and sixteen local dailies available through the news database ISI Emerging Markets.

20Is it possible that inter-cartel violence drove party fragmentation rather than the other way around? A simple bivariate statistical analysis (not shown) reports no association between inter-cartel violence and partisan turnover in municipal elections from 2007 to 2012.

21 Ley (2014) notes that 45 per cent of the public blamed the ongoing violence on the government and 71 per cent of them blamed it specifically on president Calderón. In subnational races, PAN candidates sided with the president and emphasized the long-term benefits of his security policies, while opposition candidates highlighted the failure of the president's militarized strategy.

22Interview with Ernesto Ruffo, governor of Baja California (1989-1995).

23The letter was reproduced by the national and local press. See Martínez (2008).

24Interview with Lázaro Cárdenas Batel, governor of Michoacán (2002-2008).

25Interview with Leonel Godoy, governor of Michoacán (2008-2012).

26Ibid.

27Anonymous interview with federal security official.

28Interview with Guillermo Valdés, former director of Mexico's secret service agency (CISEN) under President Calderón.

29Ibid.

30Interview with Godoy.

31Trejo and Ley (2015) provide evidence of this wave of attacks in Michoacán and the rest of the country. Their statistical analysis shows that mayors in leftist states were significantly more vulnerable to criminal attacks than those in PAN-ruled states.

32Anonymous interview with López's close collaborators. See also Reforma (2013).

33Interview with Amalia García, governor of Zacatecas (2004-2010).

34Anonymous interviews with two leading social activists in Juárez.

Received: January 15, 2015; Accepted: July 14, 2015

Creative Commons License Este es un artículo publicado en acceso abierto bajo una licencia Creative Commons