Spatial Analysis of Criminal Collaboration in the American Mafia
Clio Andris, Daniel DellapostaThe American Mafia is a violent network of criminals who engaged in drug trafficking, physical violence, racketeering and illegal activities. In this work, we analyze a historical spatial social network of 680 members of the American Mafia geolocated to a known household address across 15 major U.S. cities, and concentrated in New York City. Connections between members represent `known associates' found through a federal crime investigation by the U.S. Federal Bureau of Narcotics in the 1960s. The network, both spatially and socially, is heavily dictated by Mafia family structure.
Using existing and new spatial social network (SSN) methods and geographic information systems (GIS) data, we explore how the mafia network is distributed as a portfolio of nearby and distant ties.We uncover a mixture of spatial and network proclivities and strategies used by organized crime families.
We find that the network has bi-coastal ties within families and strangers who lived on the same block. We also uncover a strategic mixture of family members in open cities like Miami, and especially South Beach, likely sent to facilitate supply chain management for gambling, and drug, alcohol, cigars, etc. imports from Cuba. We identify high-degree individuals' tendency to live near the center of their respective families, as well as near docks and waterways, ostensibly to oversee overseas trade of olives, cheese and other goods from Italy. Per the security-efficiency tradeoff theory and characteristics of small world networks, we find that family-based ties acted more efficiently than non-family based ties, indicating some evidence of top-down organization.
We also advance methodologies for spatial social networks (SSNs) that can be generalized to other types of spatially-embedded social networks. These new metrics, such as the cluster/cluster matrix and network hotspot analysis reveal patterns that would otherwise be difficult to detect and quantify. The cluster/cluster matrix reveals variation in families' network density and spatial proximity so that not all cliques are highly clustered and not all sparse networks are dispersed. The network hotspot analysis shows that areas that would mathematically be considered as hot spots of criminals by their clustered location only, may not be as important as the hot spots where people were connected. These metrics highlight potential missed opportunities (e.g., living nearby but not connected could be a strategy to avoid detection or a missed opportunity for collaboration), and the logistics needed for families to meet. We measure the route efficiency for each node and find that families tend to be closer, but are still intermixed with other families. Lastly, in New York, we use traditional GIS spatial joins and with Census data to show that members did not live in wealthier neighborhoods than the average New Yorker.