A RANDOM CLUSTER PROCESS APPROACH TO COLLECTIVE MARKET DYNAMICS WITH LOCAL INTERACTIONS

In this paper, we adopt an agent-based approach to model collective market dynamics when interactions between the agents in a market are significant. Our model has two special features. First, social groups are formed in a random cluster process which, we believe, mimics the actual formation of social circles. Second, the process is shown to have an equilibrium distribution which gives the highest probability to the market configurations with the maximum total likelihood. With this model we are able to catch some key characteristics of collective dynamics emerged from agent interactions. These characteristics include, for example, a heavy-tailed distribution for the market returns.