It is a well-known fact to Mexican citizens, journalists and academics that, beginning 2007 when former President Felipe Calderon declared war on the drug cartels, Mexico became an extremely violent country: going from an average of 9,000 people killed a year to over 27,000. While the country’s murder rate still lags behind that of some other Latin American countries, such as Venezuela, Honduras and Colombia, nevertheless Mexico’s murder rate is eight times higher than the average rate of the member countries of the Organisation for Economic Co-operation and Development (OECD) and over 2.3 million Mexicans live in municipalities whose homicide rate is higher than that of the country with the highest homicide rate in the world, El Salvador (which is a little over 116 homicides per 100,000 people). Between 2007 and 2015, over 27,000 Mexicans disappeared, and over 180,000 were murdered in the country.

For a long time, these numbers made little impact on either the international or the national media. Events like the massacres in San Fernando in Tamaulipas and Villas de Salvarcar in Ciudad Juarez, as well as attacks like the one on Casino Royale in Monterrey, didn’t manage to stay in the Mexican news cycle for more than a week or two, let alone be discussed in the international media. The government’s institutional response, and even sometimes the media’s, to the majority of these mass killings was to present the murdered not as victims but as members of drug cartels or rival gangs, thus inhibiting any possible empathy with these individuals and shifting the blame from the state to the narcos. To illustrate the type of responses referred to above, we can take a look at the initial coverage of the killing of two students in Monterrey in Nuevo Leon in 2010. These two students were caught in a confrontation between the Mexican Army and the narcos and initial reports labeled them as sicarios, or hitmen, and police even went as far as to state that they were found carrying guns.[i]

Yet the very Manichean and fairly abstract speech about violence used by the government and the media changed radically with the disappearance of the 43 students from the Ayotzinapa Normal School in Iguala in Guerrero on September 26, 2014. These students were on their way to Mexico City to participate in a peaceful student demonstration commemorating the Tlatelolco Massacre (the killing of close to 300 students and civilians by the military and police on October 2, 1968); they had to enter the municipality of Iguala to pick up a third bus to take more students to the demonstration. The municipal police and other armed men ambushed the buses, attacking two of them and ordering all the students off the third bus. Two students were killed and one was shot and taken to the hospital during the attack; one student who fled the scene was discovered the next day by police forces, dead, without facial skin and eyes; 43 normalistas (a term used for high school students who are training to become teachers in rural Mexico) that were taken into police custody are still missing today.

Contrary to events like the massacres mentioned above, the disappearance of the 43 normalistas has been part of the Mexican news for more than two years now and it had a large impact in international media as well (the BBC, the New York Times, CNN, PRI, La Nacion, and Univision, among other news sources, still continue to publish multiple articles related to Ayotzinapa). The reason for this, I argue, is that from the very beginning these young men were portrayed, by their parents and by civil society, as students and not as members of gangs, as the government would’ve wanted. This particular framing contributed to the massive demonstrations we saw in the aftermath of the crisis in Mexico City, as well as other cities in the U.S. and Mexico, and the pressing of civil society for answers.

Mexico’s Attorney General, Jesus Murillo Karam’s immediate response went from the detention of over 100 people (including 22 municipal police officers), to the incarceration of the mayor of Iguala and his wife. After a four-month investigation, the Office of the Attorney General (or PGR) concluded that the 43 normalistas were killed, their bodies were incinerated near the town dump of Cocula and their remains were bagged and thrown into the river. Yet, this version of the story has now been widely disputed by two international groups of forensic experts. First, the sicarios who testified that they had incinerated the bodies presented signs of torture. Second, there is no evidence that the students were ever taken to Cocula and the forensic experts couldn’t find any human remains related to the students in the site. Third, it was raining that day and, if they would’ve incinerated the bodies, we would’ve seen a massive smoke cloud over Iguala, which we didn’t. Fourth, according to the government they found over 40 bullets in Cocula that prove the students were killed there, nonetheless the experts determined these bullets do not match the guns the hitmen said they used. Fifth, there is evidence that the PGR lost, altered and destroyed evidence.[ii] Therefore, two years later we are still nowhere close to knowing what happened to the missing students.

In the midst of this crisis, two months after the disappearance of the normalistas, the President of Mexico, Enrique Peña Nieto, proposed a series of security, justice and economic measures to focus on particular conditions that fostered these barbaric acts. Some of these measures included: ending corruption;  reducing poverty, marginalization and inequality levels in Chiapas, Guerrero, and Oaxaca; proposing new human rights reforms; as well as creating the 911 call center.[iii] Most importantly, among these measures, the President proposed reintroducing to Congress the constitutional reform (which still hasn’t been passed yet), to substitute the municipal police forces with a centralized state police force called “Mando Único Policial.” While some states had already begun to do so since this reform was first introduced in 2010, this time all 32 states would be obligated to change their policing structures.

In this article, I will look specifically at the effects of this policy on two measures of narco violence in Mexico: homicides and disappearances.

 

Reviving the “Mando Único Policial”

The idea behind Mando Único is to substitute, at the state level, the country’s 1,800 local police units with centralized police forces. This reform modifies article 115 of the Mexican Constitution, which specifies that the municipalities, and not the states, are in charge of providing public security (and only allows for state and federal intervention in critical events). Peña Nieto’s logic, based on a thorough analysis of numerous security speeches made during his presidential campaign in 2012[iv], is as follows: centralizing police forces would resolve problems of coordinating among the different security agencies and would make police agents less susceptible to corruption, therefore leading to decreases in crime and disappearances.[v]

This kind of policy recommendation, which centralizes the police forces, is not new: former President Felipe Calderón made a similar proposal in 2010, claiming that the model of local police units was obsolete given the units’ weakness vis-à-vis organized crime groups and drug cartels.[vi] President Calderón said that Mexicans were in desperate need of an honest and professional police force and that municipalities (especially smaller ones) did not have the necessary financial resources to maintain such forces. Yet, the federal government under President Calderon was unable to negotiate the terms of the reform with the 32 governors (a lot of mayors refused to lose federal funding for their security tasks), and the reform wasn’t approved in the Senate.

Even though both reforms share the same name, there are key differences between Calderon and Peña’s proposed policy of Mando Único Policial. First, the creation of a centralized state police unit will be obligatory for all states under Peña’s policy, while it would only have applied to municipalities with weak and corrupt security forces under Calderón. Second, Peña proposes greater coordination between the state and federal legislative and executive powers, while Calderon would have given exclusive decision power to the Secretary of Public Security regarding whether to intervene in critical municipalities. Third, Peña’s reform will not include the creation of metropolitan police units, while Calderon’s did. Fourth, and most important, Peña’s reform does not detail the specific steps municipalities must take in order to centralize their police units, while Calderon’s proposal listed the procedures of implementation.[vii]

Despite the differences between Peña’s and Calderon’s policy, experts in security matters, like Alejandro Hope, still emphasize that Peña’s policy will not solve the really problematic features of police forces, like training and salaries.[viii] Hope stated that President Peña’s policy was nothing but a simulation: once again the Mexican government proposed a policy for the sake of proposing something as a response to the Ayotzinapa crisis and not because centralizing police forces is a necessary condition to bring crime rates down. While I agree with Hope that the debate on Mando Único Policial is still centered exclusively on opinions about which types of reforms are needed to strengthen the police vis-à-vis the drug cartels and not facts, and that Peña’s policy falls short in addressing other important aspects of police reform like salaries and supervision, we can’t jump to conclusions on the effectiveness of this policy without gathering data first.

 

The arduous task of gathering data

There is little documentation on the implementation of this policy in the 2,457 municipalities in Mexico. Because of the previously-mentioned vagueness regarding the operation procedures of Peña’s Mando Único Policial reform, neither Mexican citizens nor international observers really even know what to look for in terms of documentation and implementation of the reform to determine its success in reducing crime. Furthermore, the use of official statistics can be very tricky. While the Secretary of the Interior, Miguel Angel Osorio Chong, has claimed repeatedly that over 85% of municipalities have signed an agreement with their state to implement the Mando Único Policial,[ix] there is no systematic public information regarding which municipalities have done so or when they signed that agreement. Besides, the problem with statements like Chong’s is that they are only counting cases in which the mayor signed an agreement with the governor to implement this policy, yet these agreements are not binding. Even with the mayor’s approval, the administrative council (cabildo) of the municipality still has to approve the agreement by simple majority. While this vote makes the agreement binding, the municipalities still face resource allocation disputes that make its implementation even harder.

To best exemplify the difference between binding and non-binding agreements we need not look further than Iguala, the municipality were the normalistas of Ayotzinapa disappeared. In June 17th, 2013, over a year before the attack on the normalistas, the mayor of Iguala signed an agreement with Guerrero’s governor Angel Aguirre to establish the Mando Único Policial.[x] After intense media attention was trained on the municipality, it became clear that the policy was never set in motion. Mayor Jose Luis Abarca signed the agreement with Guerrero’s governor, but later refused to implement it.[xi]

Despite all the problems of finding information of this policy and the vagueness of the federal government’s reference to its implementation, through a Public Information Request to the National System of Public Security (SNSP) of Mexico, I obtained the list of municipalities whose administrative council has approved the agreement; the SNSP stated they do not have the data regarding the dates of the municipalities’ signatures. Combining this data set with the 2014 and 2015 National Census of Government, Security and the Penitentiary System (conducted by INEGI), as well as my own extensive research in local, state and national newspapers and legislative records on approval of agreements at the cabildo level, I was able to build a data set that accounts for all the municipalities that have signed and approved the Mando Único agreements and the date of approval.[xii] This is a time-series cross-section data set that accounts for status of Mando Único of all the 2,457 Mexican municipalities for the 2006-2014 period.[xiii]  The data set includes two dichotomous variables: a “treatment” variable, which codes the municipality-year as 1 if it is part of the treatment group (it has approved Mando Único Policial reform) and 0 otherwise; and a “policy in place” variable, which codes the municipality-year as 1 if the agreement is now place and 0 otherwise. For example, if the municipality signed the agreement in October of 2012, then it will be coded as 1 for the full 2006-2014 period in the “treatment” variable and will be coded as 1 for 2013 and 2014 in the “policy in place” variable.

It is important to keep in mind that for now I only have information regarding the agreements; unfortunately, I do not have enough information to address at what stage of implementation of the Mando Único Policial individual municipalities have reached. Knowing the stage of implementation of the policy would be the ideal scenario, given that one could easily expect that Mando Único would have a different effect on narco violence in municipalities that have only signed an agreement versus those that have actually enforced the policy.[xiv]

While only having information on the agreement is a setback in the evaluation of the policy, I argue this is an important step for two main reasons: it is the first time a systematic analysis of the security policy has been conducted using data that no one else has gathered, and it will provide an important stepping stone in understanding whether the Mando Único Policial has had the desired consequences, of reducing crime and violence, delineated by the President himself.

 

What do the data say and what can we conclude?

I compare the violence trajectory (average homicide levels from 2006-2014) between municipalities that have signed the agreement (the treatment group) and municipalities that have not signed Mando Único Policial (the control group) by running regressions of the form:

cpr-equation
In this equation, (y) is the outcome or dependent variable of interest (I am testing this relationship using two different violence measures: the count or total number of homicides first, and the number of disappearances in a municipality in a given year second); (x1) is a dichotomous variable for the treatment group, (x2) is a dichotomous variable for the years in which the agreement has been in force, (x1x2) is the interaction between treatment and policy (years in which the agreement has been in force for the municipalities in the treatment group); (gamma) are other control variables. In all the regressions I control for levels of urbanization, income and education and past violence levels and adjust the standard errors for within-state correlation.[xv]

The form of this regression is determined by several factors. First, because the distribution of homicides and disappearances is highly skewed to the right, meaning a majority of the municipalities recorded either 0 or 1 homicides/disappearances in the years between 2006 and 2014, I needed to specify the model as a zero-inflated negative binomial regression that takes into account the distribution of my outcome variable. This specification is a common practice among research that studies crime, gun violence and homicides. Additionally, zero-inflated negative binomial regression requires the use of a count outcome variable (we sum up the number of events in a specific space and time), but this creates a potential problem because the number of homicides or disappearances is highly correlated to population size. One can expect more homicides the larger the population, and for this reason most studies also tend to use rates per 100,000 people. Nonetheless, we can adjust for this issue by adding the logarithm of the population to the right hand side of the equation (as an independent variable) and isolate the effect of violence from population size.[xvi]

Second, while I can’t randomize which municipality signs the agreement and which municipality doesn’t, I can approximate this ideal laboratory scenario by exploiting variation in the agreement of Mando Único Policial in time and space. I can use a difference-in-difference (DD) strategy to estimate the relationship for two reasons. First, there is no systematic bias in which municipalities adopted the policy and which didn’t (meaning it is not the case that all the municipalities that adopted the agreement had high levels of crime and violence in the first place). Second, the relation between homicides or disappearances a year before the adoption of the agreement is indistinguishable from zero.

DD is one of the most popular tools in economics to evaluate the effects of public interventions on relevant outcome variables. A DD strategy requires one to identify which is the group affected by the policy change (treatment) and which is the group not affected (control) and allows you to estimate the trajectory (in my case the levels of homicides or violence for the period) of each group once the policy in place. If there are systematic differences between both groups after the policy is in place, then it is possible to conclude there is a relationship (although not necessarily a causal one) between the outcome variable and the policy implementation.

As a cautionary note, I must emphasize once more that these are preliminary results given that I am only looking at municipalities that have signed the agreement; I am not factoring in the stage of implementation of the actual security policy. Nonetheless, because signing an agreement is one type of commitment strategy, commonly used in game theory models, in which the municipality agrees to implement Mando Único in the future, this commitment changes the type of message sent to other players (corrupt policemen and politicians as well members of gangs or drug cartels) by the local government and therefore can change the equilibrium payoffs of engaging in criminal activities. For this reason, we can expect municipalities that have signed the agreement to behave differently from those that didn’t, which makes these initial results both empirically and theoretically relevant.

Specifying disappearances as the dependent or outcome variable, I find that among those municipalities that have at least one missing person, moving from a municipality without Mando Único Policial to a municipality that has signed the agreement, and this agreement is in place, increases the expected number of missing people by 510%, holding other variables constant at their means. Plus, I find that the approval of the agreement is not statistically related to the odds of a municipality staying with a count of zero missing people.

The fact that we observe a positive relationship between Mando Único Policial and disappearances should not be that surprising given the timing of the agreements and therefore the results should not be interpreted as: “the implementation of the policy caused more disappearances”. Contrary to the trend in homicides, while there was a temporary fall in disappearances in 2012, the number of missing people has slightly increased for the last three years.[xvii] This increase is correlative in time to the biggest wave of agreements of implementing the security policy (in 2013 and  2014), which explains the positive relationship between both. Additionally, the increase in disappearances of these last three years appears to reflect a trend that the Mando Único, which was suggested specifically in response to cartel-related violence, might not be equipped to address: gender-related violence and trafficking. Women disappear at a younger age than men (at an average age of 23 years old, compared to 33 years old for men) and in different locations than men (in states not known for high narco-violence like Guanajuato, Estado de Mexico and Puebla). If the Mando Único was set in motion to exclusively solve violence associated with drug cartel activities, as was stated by the President himself,[xviii] and female disappearances have a stronger link to sex trafficking than to narco violence, then it is difficult to see how the policy will have a significant effect in reducing female disappearances in the future.

When using homicides as the dependent or outcome variable, however, I find that among those municipalities that have at least on homicide, the implementation of Mando Único decreases the expected number of homicides by 52%, holding other variables constant at their means. Plus, the approval, once more, is not statistically related to the odds of a municipality staying with a count of zero homicides for the 2006-2014 period.

The negative relationship between the centralized police force agreement and homicides is more interesting as a causal story because crime and policing theories suggests there are possible causal mechanisms driving this association. First, it could be that citizens see this new police force as more trustworthy and thus a new trust equilibrium is emerging that is bringing crime rates down. It could also be that, as violence expert Eduardo Clark has suggested, the substitution of the municipal police force for a centralized unit could implies that the narco now need to establish agreements with a smaller number of actors and therefore it is easier to maintain a non-violent equilibrium between the government and the cartels.[xix] Or it could be that these police forces are simply better at their jobs, and we could in the near future see a larger number of cartel and gang members being prosecuted and incarcerated. Only time and process-tracing analysis will reveal which of these mechanisms, or even others, explain the results found.

As a conclusion, there is preliminary evidence that the Mando Único Policial is associated with decreases in some types of violence (homicides) but not others (disappearances). The next logical step in the analysis of the implementation of the security strategy is to build a data set that differentiates the types of agreements each municipality has signed and their stage of implementation. Only through these data will we have a clearer understanding of this new security policy and its effect on narco violence in Mexico.


References:

Image: Flickr/Presidencia de la República Mexicana.

[i] URL: http://www.jornada.unam.mx/2010/03/21/politica/009n1pol

[ii] URL: http://www.animalpolitico.com/2016/04/que-sucedio-con-los-43-normalistas-tras-los-ataques-esta-es-la-segunda-parte-de-la-cronologia/

[iii] URL: http://mexico.cnn.com/nacional/2014/12/03/mando-unico-policial-y-otras-10-claves-del-nuevo-plan-de-seguridad-de-pena

[iv] URL: http://www.animalpolitico.com/blogueros-causa-en-comun/2014/02/21/la-seguridad-publica-esta-en-el-discurso-de-pena-nieto/

[v] The Mexican Federal Criminal Code, in art 215, defines forced disappearances as cases where a public official is involved in the crime.

[vi] URL: http://mexico.cnn.com/nacional/2010/10/06/calderon-formaliza-la-peticion-para-crear-una-policia-de-mando-unico

[vii] URL: http://www.oem.com.mx/elsoldedurango/notas/n3637092.htm

[viii] URL: http://www.noticiasmvs.com/#!/noticias/apoya-causa-en-comun-el-establecimiento-del-mando-unico-575

[ix] URL: http://www.milenio.com/policia/poblacion-policias-mando-unico_0_595140492.html

[x] URL: http://www.unotv.com/noticias/estados/suroeste/Formalizan-mando-unico-policial-en-seis-municipios-de-Guerrero-302035/

[xi] URL: http://eleconomista.com.mx/sociedad/2014/11/02/guerrero-muestra-fracaso-mando-unico

[xii] Three states were excluded from the sample because I could not verify via legislative records if their agreement was approved in the local administrative councils or there was contradictory information in the sources I consulted (specifically between SNSP and INEGI); these are Tlaxcala, Baja California and Chihuahua. I also have missing observations for a number of municipalities in Oaxaca because I could not access the state’s legislative records and therefore could not verify if the agreement was approved in the administrative council for some municipalities. Lastly, states that have a de facto centralized force but do not have a de jure agreement, like Nuevo Leon, were coded as zeroes (as part of the control group).

[xiii] The only reason I do not include the year 2015 is because homicide data is not available, it is usually published around October of the following year.

[xiv] By effect I mean the correlation, and not causality given that with regression analysis I cannot prove a causal relationship.

[xv] Data sources: INEGI, RNPED, SINAIS, CONAPO.

[xvi] For a more detailed explanation on this, refer to Long and Freese (2014).

[xvii] Because data on disappearances is constantly being updated, more recent years could witness a fall in the cases of missing people in the near future.

[xviii] URL: http://www.excelsior.com.mx/nacional/2016/03/01/1078037

[xix] URL: http://noticieros.televisa.com/foro-tv-sin-filtro/  (August 11th, 2013)


Jessica Zarkin

Jessica Zarkin is a second year PhD student in Comparative Politics at the Cornell Government Department. Her research focuses on democratic development and accountability mechanisms in violence-ridden countries in Latin America. Before coming to Cornell, she was a data analyst at Data4, a Mexican firm that works to make data accessible to more people and provides data solutions to an array of clients including media outlets, NGOs, and government branches. During her time at Data4, she co-authored several pieces on violence in Mexico, including a weekly blog in AnimalPolitico and two in-depth articles in Nexos.

Written by Jessica Zarkin

Jessica Zarkin is a second year PhD student in Comparative Politics at the Cornell Government Department. Her research focuses on democratic development and accountability mechanisms in violence-ridden countries in Latin America. Before coming to Cornell, she was a data analyst at Data4, a Mexican firm that works to make data...
Read more