Social Interactions with Cognitive Hierarchy

Abstract
In this paper, we propose a novel discrete-choice model for social interactions based on cognitive hierarchy. Compared to the rational expectation model in the existing literature, the cognitive hierarchy model does not require the agents to have comprehensive knowledge of the network or to solve complicated fixed-point problems. We discuss the interpretation and implications of the cognitive hierarchy model in the context of social interactions, conduct Monte Carlo simulations to show the proposed model can successfully detect the sophistication level of the agents and the true value of the spillover effect, and apply it to study the smoking behavior of middle school students. We find that, for most students, the smoking decisions are not swayed by their friends. For those whose smoking decisions are influenced by their friends, it is primarily by their immediate friends.