Methodological Summary

This visualization presents an interactive heat map of Organized Criminal Groups in Mexico geo-referenced at the municipality-day level between 2000 and 2019. 

To generate the data, the project relies on the computerized identification of actors and locations using Eventus ID (Osorio & Reyes 2017).  The raw information comes from a large collection of news stories from local and national Mexican newspapers and press releases from government agencies.

​

To select the relevant stories, the methodology first relied on a team of trained human coders who identified news stories by strictly following the codebook guidelines. The team of research assistants has an intercoder agreement of  90.4% and a Fleiss’ Kappa of 0.704. In order to automatically update the selection of relevant news stories, the methodology used the manually gathered data to train a Machine Learning (ML) model. After comparing the performance of different ML models, the Linear Regression model performed with the highest accuracy with an F1 of 0.949. The final collection of selected news stories became the input corpus for computerized event coding.

Based on a dictionary of Organized Criminal Groups containing more than 7,000 names of criminal groups and individuals, Eventus ID reads the content of the corpus in order to identify the actors and places mentioned in the news stories. The coding protocol implements additional routines of geographic disambiguation and deduplication of coded actors. The result of the coding process is a database of Organized Criminal Groups geo-located at the municipality-year level.

Each data point in the database represents one observation of Organized Criminal Groups in a given municipality-day as detected in the corpus The exact location of the incidents at the XY coordinate level is not determined in the newspaper narratives, so the methodology assigned the coordinates of the municipality center to their respective observation. The map presents a heat map displaying a higher intensity color range in areas concentrating a higher density of Organized Criminal Groups.