Community-based identification of banking networks

We present a statistically-based analysis of community characteristics of two networks in the context of a stylized model of the overnight interbank network. The networks analyzed are simulated interbank lending networks, which are formed on the basis of simple random criteria or on the basis of familiarity, or past history, between banks. We use a measure of betweenness centrality, coupled with a topological community network partition, to define a metric partition dubbed the Global Community Average Betweenness. Based on simulation data, this metric shows promise in distinguishing familiarity-based networks from randomly-selected networks, particularly when applied to the largest topological community in the network.