Collective Search in Networks

Abstract

I study social learning in networks with information acquisition and choice. Rational agents act in sequence, observe the choices of their connections, and acquire information via sequential search. I characterize equilibria of the model by linking agents' search policies to the probability that they select the best action. If search costs are small enough, an improvement principle holds. This allows me to show that asymptotic learning obtains in sufficiently connected networks in which information paths are identifiable. If search costs are bounded away from zero, even a weaker notion of long-run learning fails, except in special networks. When each agent observes a random number of immediate predecessors, the rate of convergence, the probability of wrong herds, and long-run efficiency properties are the same as in the complete network. Network transparency has short-run implications for welfare and efficiency and the density of indirect connections affects convergence rates.

Keywords: Social Networks; Rational Learning; Improvement and Large-Sample Principles; Speed of Learning; Search; Bandit Problems; Information Acquisition and Choice.

JEL Classification: C72; D62; D81; D83; D85