Regime change in large information networks

Abstract

We study global games of regime change within networks of truthful communication. Agents can choose between attacking and not attacking a status quo, whose strength is unknown. Players share private signals of the state with their immediate neighbors. Communication with neighbors introduces local correlations in posterior beliefs, and also pools information. In order to isolate the latter effect, we provide sparseness conditions on networks that allow for asymptotic approximations that eliminate covariances from equilibrium strategies. We ask how changes in the distribution of connectivities in the population affect the type of coordination in equilibrium as well as the likelihood of successful rally. We find that, without public signals, strategic incentives align and the probability of success remains independent of the network. With a public signal, the distribution of degrees unambiguously affects the probability of success, although the direction of change is not monotone and depends on the cost of attacking.