Political Violence and Social Instability

A set of agents decides whether or not to start or join a protest. Each individual is endowed with a threshold so that she will participate if and only if the overall number of activists exceeds this threshold. Chwe (1999) provides a static model in which individuals to not know all thresholds; instead agents are connected by an undirected graph and know the thresholds of their neighbours. The present paper shall provide a dynamic version of this model: at the beginning of each period all agents observe the behaviour of all their neighbours in the past. They use this information to update their beliefs about the number of (potentially unobserved) active agents, before they revise their decision. The model leads to two key observations: (i) denser networks lead to more active agents, (ii) an increase of thresholds might lead to less or more active agents in the long run. The second, somewhat surprising, observation originates from the two-fold role of thresholds. An agent with a high threshold will become active only very late (if at all); but once she is active, all her neighbours know that her high threshold has been crossed, that is, that the number of active agents must be high even if the latter cannot be observed. In a variation of the model, we allow random agents to coordinate their behavior. Then an increase of thresholds has the effect that small protests will be observed less frequently, while the probability that a protest (once it has started) turns into a full grown revolution increases. We illustrate this finding using global data on mass protests, revolutions, and the political terror scale in a time window from 1976 to 2014.