Bank runs in a local interaction model

We propose an agent-based computational model of bank runs. In the model, synchronization effects, which generate bank runs, can arise from imitation among depositors. There are three important interacting factors which influence the patient agents’ strategies (withdraw or wait), the proportion of patient agents, the activation threshold, and the interaction neighborhood of agents. We study the question of how frequently bank runs might occur or of what features of interaction mechanism might minimize their frequency.