Eight years of homicide evolution in Monterrey, Mexico: a network approach

Mexico has experienced a dramatic surge in homicide rates since the declaration of the war on drugs in 2007. The Monterrey Metropolitan Area (MMA), a major urban center and business hub, has been significantly affected by this violence, experiencing hundreds to thousands of casualties. While various methods exist for analyzing crime data, the lack of accurate spatiotemporal data often leads to incomplete descriptions of homicide evolution. This research addresses this gap by applying a novel network-based approach to analyze geo-located homicide data from the MMA. The study aims to understand the spatial and temporal dynamics of homicide violence, identifying patterns of spread and clustering, and investigating the influence of geographic factors, such as major roadways, and socioeconomic factors on the distribution of homicidal events. The MMA provides a unique case study due to its high homicide rates during the period, linked to the conflict between the Cartel del Golfo and Los Zetas, and the availability of a detailed, geo-located homicide database from the newspaper El Norte. This study’s findings are crucial for informing crime prevention strategies and resource allocation in the MMA and other areas experiencing similar levels of violence.

Existing research on conflict and gang-related violence employs diverse methods, including literature-based approaches, data-mining network inference, reaction-diffusion equations, and combined methods. Studies often focus on property crimes like burglary and robbery due to data availability. Some research has used homicide data, but the granularity (municipality or state level) limits detailed analysis of spatial and temporal patterns. Previous network approaches have analyzed gang-related homicides using gangs as nodes and homicides as links but often neglected spatial factors. Other work has explored the role of geographical boundaries in gang conflicts using models like the Lotka-Volterra model to predict gang territories. However, few studies have analyzed homicide data with the level of geo-spatial precision and temporal resolution achieved in this research, offering a more nuanced understanding of crime dynamics during a period of intense drug-related violence.

This study used a geo-located homicide database (El Norte newspaper daily-updated database, ENDB) covering the period from January 2011 to February 2018. The ENDB provides daily homicide records with casualty numbers, longitude and latitude coordinates. Shapefiles containing neighborhood data for the MMA were obtained from datamx. The study conducted several analyses: 1) Time series analysis of homicides at municipality, locality, and neighborhood levels to examine temporal trends. 2) Network construction: Neighborhood networks were constructed where nodes represented neighborhoods, and edges linked adjacent neighborhoods with at least one homicide within a specific time window (entire period and yearly). Co-occurrence networks were constructed linking neighborhoods with homicides in the same week, reflecting temporal correlation. 3) Spatial correlation analysis between neighborhoods using a metric from Alves et al. (2015) to determine distance dependency. 4) Correlation of homicide locations with high-speed roads/highways from OpenStreetMap data to investigate the role of urban infrastructure in shaping homicide patterns. The analyses were performed using pandas, geopandas, networkx, and Cytoscape. Statistical significance for correlations was assessed using Z-scores derived from a null model created by reshuffling casualty data.

The study revealed significant heterogeneity in homicide patterns across the MMA. Time series analysis showed variation in homicide rates among municipalities, with some consistently experiencing high rates (Monterrey, Apodaca, Cadereyta Jiménez, San Nicolás de los Garza, Guadalupe) and others (Santa Catarina) displaying more stable, lower rates. Analysis of correlations between municipalities indicated that proximity was not the only factor determining correlation; some distant municipalities were strongly correlated. At the neighborhood level, a notable lack of distance-dependent correlation was found, particularly between 2011 and 2016. Yearly co-occurrence networks illustrated the spatial dynamics of homicides over time. The networks revealed multiple edges connecting outer neighborhoods to central neighborhoods (2011-2012), which decreased over time. A network depicting the whole period showcased a star-like structure, with a central polygon in downtown Monterrey showing the most frequent co-occurrences with distant peripheral neighborhoods. This suggests that control over downtown Monterrey was central to the overall homicide pattern. The analysis found that a high number of homicides occurred near the 85th freeway, which connects MMA to the US border. Finally, socioeconomic barriers, particularly those between municipalities with contrasting GDP per capita and development, appeared to segregate areas of high and low homicide rates. The proximity to highways also influenced the distribution of homicide events, with shorter distances in years of peak violence and longer distances in following years.

The findings demonstrate the value of the network approach for analyzing complex spatiotemporal crime patterns, particularly during periods of intense drug-related violence. The lack of distance dependency at the neighborhood level contradicts some existing research and emphasizes that social factors and strategic control of key locations (like downtown Monterrey and the 85th freeway) might be more important drivers of homicide patterns than simple proximity. The star-like network structure, with frequent connections between the city center and peripheral areas, suggests a strategic pattern of violence possibly related to the control of drug trafficking routes and key infrastructure. The influence of socioeconomic barriers underscores the role of social inequality in shaping crime patterns. The spatial distribution of homicides related to the highway system suggests its use as a route for criminal activity and potentially as a physical boundary separating areas with varying levels of violence. This study supports the hypothesis that the high homicide rates in MMA were not simply the result of random or diffuse violence but were strategically driven by the control of central areas and crucial transportation arteries.

This study makes significant contributions to the understanding of homicide dynamics in urban areas impacted by drug-related violence. The network approach presented provides a powerful tool for analyzing complex spatiotemporal data. The findings suggest that strategies to reduce violence in such settings should focus not only on physical proximity but also on the control of key locations and the addressing of socioeconomic inequalities. Future research could explore the directionality of homicide events, investigate the role of specific criminal organizations in shaping the network, and perform null model analyses to assess the significance of observed segregation by socioeconomic and geopolitical boundaries.

The study's reliance on a single newspaper's database might limit the generalizability of the findings. The lack of information on victim and offender characteristics prevents a detailed analysis of network directionality and underlying mechanisms driving homicide patterns. Additionally, the spatial resolution is limited by the granularity of the neighborhood-level data, potentially missing finer-scale patterns of homicide clustering.