In May 2019, as part of the rollout for his national crime-fighting strategy, President Andrés Manuel López Obrador recognised the existence of 37 “cartels”, many of them factions of larger organisations. This figure is a significant underestimate. But in trying to give a precise number, the government highlighted the sheer difficulty of mapping Mexico’s underworld, which has fragmented under the pressure of previous governments’ militarised “war on drugs”. Without a better understanding of the scale and character of this splintering, the government will have a hard time designing effective policies for curbing drug-related conflict. No one, for instance, knows the full answers to basic questions about exactly how the “war on drugs” has fractured organised crime; how in turn this phenomenon relates to the greater intensity of violence; or how the government might calibrate its policies to take on criminal groups of various sizes and structures. Until the government knows when and why new groups form, and how this process affects conflict, it will face major challenges in trying to end the cycle of fragmentation and violence.
The Challenges of Counting
Understanding these criminal groups – how they operate, how they relate to one another, how they respond to government policies – is central to explaining the spike in homicides in Mexico. But attaining accurate information on them poses a huge practical challenge. Typically, researchers would rely on the press, but crime reporting is extremely difficult. Mexico is the most dangerous country for journalists in the Western Hemisphere, despite federal protection efforts, and self-censorship generated by the threat of reprisal makes the quality of reporting on the drug war suspect, particularly in the most conflict-affected regions.
Complicating things further, in 2011 several large Mexican media outlets, agreed “to omit and dispose of information coming from criminal groups for propagandistic purposes”. The intention was to make it harder for these groups to intimidate and coerce the public via the press, but in practice it meant that some outlets stopped reporting all their statements. As a result, the public often does not know when new groups have arrived, because smaller or emerging organisations often use propaganda, such as banners hung in streets or statements posted on Facebook, to announce themselves.
In addition, many criminal groups in Mexico operate at such a small scale that the media may deem them too insignificant to cover at all. As factions become increasingly local, or increasingly specialised in extortion or trafficking of particular commodities, they tend to fall off the radar. Yet these small-timers play a large role in Mexico’s rising rates of violence and hold sway over many people’s lives.
Crisis Group has developed a better method of identifying which and how many criminal groups are operating in Mexico: analysis of “narcoblogs”, anonymously run websites that aggregate news of cartel activities from both mainstream media outlets and ordinary citizens. Narcoblogs, The Guardian reports, lay bare “day after day, the horrific violence censored by the mainstream media”. Like professional journalists, the operators of these websites face considerable danger. The creator of an early narcoblog fled Mexico after her partner called her to say only one word: “run”. But because they are anonymous, the blogs can avoid self-censorship, and because they rely on citizen testimony – as well as press accounts – they can offer a fuller picture of the drug war than traditional media.
To build a dataset of violent groups operating in Mexico, Crisis Group began by “scraping” major websites, that is, automatically downloading the text of posts in a way that can be processed into data. We relied in particular on Borderland Beat, an English-language narcoblog run by Mexican-Americans, because it has the longest-running and most consistent coverage of drug-related violence in Mexico. We identified the names of possible crime rings using natural language processing (computational techniques designed to simplify and make sense of complex text), and we then hand-coded the possibilities to confirm a list of operating organisations.
In total, we identified 463 criminal groups operating in Mexico between mid-2009 and
2019. Only about half of these appeared in El Universal, one of the country’s largest newspapers. Figure 1 shows the number of groups mentioned in Borderland Beat and El Universal since 2010, the first full year of the narcoblog’s posts. These figures could fall if it emerges that certain organisations operate under various names, or if some local outfits are excluded, but they nevertheless point to a steep rise in criminal activity. In 2019 alone, Borderland Beat mentioned 198 groups, up from 98 in 2010. We see a similar doubling in El Universal’s coverage, even though the existence of smaller groups tends to go unreported in mainstream media. We included an organisation based on whether it operated under a unique name, a signal that it was working at least semi-independently. Fifty-two groups were identified solely on the basis of a leader’s name.
Identifying these criminal groups, even those affiliated with or allied to large cartels, is important because their proliferation shows how multi-sided Mexico’s drug war has become. Two bands called Los M and Gente Nueva warred in Durango state, though at the time both were flying the flag of the Sinaloa Cartel. Los 28 were operating as independent hit men before Jalisco New Generation Cartel recruited them to kill El Chapo’s children. An organisation led by El Teo, known as Los Teos, was initially allied with the Arellano Felix Cartel before realigning with the Sinaloa Cartel.
Given the complexity of relationships among these groups, we additionally tracked whether a group is allied with, splintered from or belongs to another organisation. Of those identified, at least 135 were cells of large cartels. We are expanding this data using outside sources to better identify connections among criminal actors. Thirty-one of the groups identified were autodefensas, local crime-fighting vigilantes, indicating that some of these outfits have criminal ties themselves. This figure is likely an underestimate, since many autodefensas operate under generic names that our research would not capture.
Our next step is to link these groups to municipalities where they operate. The new Crisis Group report on Guerrero state includes a map of the number of armed groups in each municipality in 2018 and 2019, drawing on El Blog del Narco, Mexico’s most popular narcoblog. To ensure these findings’ validity, we hand-coded each post related to Guerrero for whether a group operated in a given municipality or region. To replicate this method for the entire country would be time-consuming, however, and we are exploring ways to scale up our coding or improve the algorithm linking groups and municipalities.
Some preliminary maps for the whole of Mexico are below, though these should be interpreted with caution. Currently, we treat a group as operating in a municipality simply if the names of the group and locality are mentioned in the same post, likely yielding many “false positive” relationships. Our data do not yet take into account regions and cities mentioned instead of specific municipalities, which means that groups may be undercounted in certain areas. We also do not yet adjust for the fact that groups’ presence may not be reported in every year they operate in a municipality: if an organisation is linked to a region in 2016 and 2018, it may follow that they were also operating in 2017.
What the Data Shows
To our knowledge, this project is the first attempt to document all the non-state armed groups in Mexico, the vast majority of which are tied to violent crime. If anything, the 463 groups identified are an underestimate, and there are important limitations as to what is included. Criminal groups tend to be covered on narcoblogs when they engage in violence, suffer arrests, announce their activities or fall under government scrutiny. The data is more likely to include groups that seek some form of territorial control, and to miss some organisations that discreetly traffic in drugs. We hope to minimise the problem of missing groups by mining other narcoblogs, some of which are now defunct. For now, the estimate is best understood as tracking violent criminal groups that are fighting over territory.
This data can help explain the dynamics of criminal violence and group formation in Mexico, and the challenges facing conflict resolution strategies at the regional and national levels. For example, it can shed light on how government security policies have affected fragmentation, and how this process in turn relates to heightened levels of violent crime, particularly homicide. The data can also be used to understand how economic variables, such as shifts in commodity prices or the emergence of new trafficking routes, may affect the entry or exit of groups. In recent years, for example, increased consumer demand has led cartels to battle for control of the avocado trade. Above all, given the number of small and local outfits now involved in drug-related conflict, the data shows that focusing on dismantling major cartels is insufficient to reduce violence. It suggests that new policies – such as targeted recruitment prevention efforts, regional intervention plans, or reintegration and disarmament initiatives – may be better suited to the task. Narcoblogs also serve as a rich source for tracking other features of criminal violence, such as cartel propaganda and the growth of self-defence groups.
Our data shows a considerable rise in the number of criminal actors in Mexico over the last ten years. As more groups operate, violence among them becomes more likely and attempts at negotiation – like those in Guerrero – increasingly difficult to coordinate. Understanding the causes and consequences of this fragmentation will be essential for addressing the roots of Mexico’s crisis.