Please click here if you are unable to view this page.
TOPIC:
ROBUST OPINION AGGREGATION AND ITS DYNAMICS
ABSTRACT
We consider a robustification of the DeGroot's linear model of social learning. By relaxing the assumption of quadratic utility of the agents, we obtain an opinion aggregator that is normalized, monotone, and translation invariant. We directly link these properties to natural conditions of the microfoundation. In addition to the less demanding assumptions on the payoff function of the agents, the opinion aggregator allows for several economically relevant patterns ruled out by the linear model. For instance, agents can feature homophily, dislike (or attraction) for extreme opinions as well as discard information obtained from sources that are perceived as redundant. We also show that under this weaker assumptions it is still possible to explore the standard questions addressed by the linear model, such as convergence of limit opinions, and the properties of consensus and wisdom for this limit.