ALSA 2025 meeting

Framing corruption as a relational crime, this paper uses social network analysis to reconstruct public procurement corruption networks in Indonesia and to explore how their structure can inform network-based enforcement. A purposive sample of Supreme Court final judgments on procurement offences is coded to extract legally established relations among defendants, public officials, firms, and intermediaries, including bribe payments, bid rigging agreements, and collusive meetings. These events are modelled as one mode actor networks and a bipartite buyer supplier graph. Using Gephi, we compute degree, betweenness, closeness, and eigenvector centrality, together with density and Freeman centralization, to locate positional power and structural vulnerabilities.
The analysis shows that central figures in procurement corruption are often not formal decision makers, but brokerage actors, typically consultants or service providers, who bridge structural holes between bureaucratic and private clusters. The networks exhibit relatively high density and moderate centralization, indicating cohesive trust based structures that combine efficient coordination with resilience against law enforcement disruption. Simulated removal of popular actors with many direct ties produces limited structural change, whereas removing high betweenness brokers fragments the network and sharply reduces overall connectivity.
Methodologically, the study develops a replicable forensic workflow that converts narrative court records into analyzable graphs and operational indicators for investigative prioritization, such as high betweenness brokers, cut points, and structurally equivalent firms. Substantively, it shows how corruption in public procurement is embedded in durable, trust based social relations and argues for a shift from individual centric enforcement to network aware strategies, including risk-based audit targeting and early warning mechanisms that aim to disrupt the relational infrastructure of corrupt procurement ecosystems.