Impact of mobile money fraud detection mechanisms on fraud incidence in Somalia’s telecom sector

Mobile money has expanded financial access in Somalia, where over 75% of adults rely on telecom-led services for daily payments and remittances. Transactions exceed 60% of GDP, with monthly volumes surpassing USD 2 billion. However, weak consumer protection and fragmented oversight have contributed to unauthorized reversals, SIM swapping, and agent collusion, while under-reporting obscures the true scale of fraud. This study examined whether operator fraud detection mechanisms, from manual reconciliation to real-time machine learning analytics, are associated with lower actual fraud incidence in Somalia's telecom sector. Using a comparative quasi-experimental design across five operators over 12 months, the study triangulated a survey of 353 technical staff, transaction logs, and complaint records. Forensic sampling of 60,000 transactions estimated detection rates and adjusted for reported fraud. Negative binomial models with operator and month fixed effects showed that medium and high sophistication were associated with markedly lower fraud incidence. Fraud increased during Ramadan and with conflict intensity, whereas cross-border partnerships were associated with lower incidence independently of sophistication. Advanced detection is associated with substantially lower fraud in low-sanction environments, though weaker in conflict-affected regions. These findings support tiered minimum detection standards, compulsory incident reporting, and measures targeting collusion and informal governance constraints.