Prosecution of complex criminal networks: a multilevel ERGMs approach to CICIG's judicial cases

TL;DR

The paper addresses how prosecutors can deter complex criminal networks by leveraging a multilevel ERGM framework to connect criminal actors, offenses, and prosecutorial imputations across eight CICIG-FECI cases. By modeling three interlinked networks—criminal, legal framework, and prosecution—it tests deterrence-based configurations and builds prosecution-building blocks to evaluate strategic charging. Findings show partial support: high-severity offenses are prioritized; multi-offenders are often linked to offenses across different laws, and co-offender imputations frequently involve triangular configurations, while cross-law and offense-type mixing yield mixed or non-significant results. The study argues for a shift toward a multi-case, data-driven imputation framework to systemically disrupt criminal networks, with practical implications for targeting central actors and balancing corruption and non-corruption charges.

Abstract

Prosecutors are essential in combating organized crime, making key decisions about prosecution, target selection, and structuring imputation strategies. Despite their importance, the configuration of these strategies remains empirically underexplored. This study engages with that premise by considering the cases investigated by the International Commission Against Impunity in Guatemala (CICIG) and focusing specifically on the role of prosecutors, aiming to uncover how their discretionary decisions translated the CICIG mandate into operational practices intended to achieve systemic deterrence, and to what extent, can we talk about deterrence effectiveness. The research employs a multilevel Exponential Random Graph Model (ERGM) analysis, integrating three networks: the criminal network of actors involved in illegal activities, the legal framework network that represents offenses, and the prosecution network that connects actors to offenses. The model assesses whether the observed network aligns with deterrence-based theoretical assumptions and examines how punishment severity can be effective when it disrupts functional ties that sustain criminal activity -both through long-term sanctions and by increasing the perceived threat of punishment among co-offenders. This approach underscores the need for prosecutorial strategies to evolve beyond a case-by-case model toward a multi-case, multi-offender imputation framework that fully integrates intelligence and data-driven analysis to dismantle criminal networks.

Figures (4)

• Figure 1: Multilevel network structure consisting of: (1) a criminal network capturing within-actor ties, (2) a legal framework representing within-law connections between offenses, and (3) a prosecution network based on affiliation ties linking actors to offenses. Node attributes are encoded as follows: offense nodes are labeled by type (0 = non-corruption, 1 = corruption) and by severity (0 = low severity, 1 = high severity); actor nodes are labeled by sector (0 = non-public, 1 = public).
• Figure 2: Prosecution challenges and building blocks. Red circles depict criminal actors, while blue squares denote offenses.
• Figure 3: Prosecution building blocks. This figure illustrates the key network motifs used in the analysis. Red circles represent criminal actors, blue squares denote offenses. Each building block is linked to a specific hypothesis and expected estimation result.
• Figure 4: Multilevel network. This figure depicts the integrated multilevel network that combines three distinct but interrelated layers. The upper layer represents the criminal network (red circles and red ties), the lower layer captures the legal framework (blue squares and blue ties), and the cross-level layer, shown with grey ties, constitutes the prosecution network.