This article analyzes anomalous patterns of agent, adjuster, and producer claim outcomes and determines the most likely pattern of collusion that is suggestive of fraud, waste, and abuse in the federal crop insurance program. Log–linear analysis of Poisson-distributed counts of anomalous entities is used to examine potential patterns of collusion. The most likely pattern of collusion present in the crop insurance program is where agents, adjusters, and producers nonrecursively interact with each other to coordinate their behavior. However, if a priori an intermediary is known to initiate and coordinate the collusion, a pattern where the producer acts as the intermediary is the most likely pattern of collusion evidenced in the data. These results have important implications for insurance program design and compliance.