Cocaine Trafficking: Link Prediction and Global Network Reconstruction

• In Event: Drug Markets and Drug Policy: Methodological Advances and Theoretical Perspectives

Abstract

Transnational cocaine trafficking constitutes a complex global network. Social network analysis holds the potential to unravel the intricacies of cocaine trafficking, thereby informing effective policy measures. However, achieving a comprehensive mapping of the cocaine trafficking network remains elusive due to inconsistent data reporting across countries. In this study, we leverage a suite of nine link prediction models to reconstruct the global trafficking network using incident-level seizure data provided by the UNODC. Demonstrating high predictive accuracy, with AUROC scores reaching 0.970 in external validations, these models identify both known and previously unobserved transnational connections. The analysis reaffirms the centrality of major hubs along the Atlantic route, including Venezuela, Brazil, Ecuador, Spain, and Portugal. Moreover, it highlights underestimated cocaine flows toward the United States and suggests significant underrepresentation of routes through Western Africa, Eastern Europe, and the Balkans. Overall, the proposed approach offers an advancement in the current representation of global cocaine market, paving the way for a new generation of empirical studies on the determinants and consequences of cocaine trafficking.

Author

• Alberto Aziani, Università Cattolica del Sacro Cuore - Transcrime