Should Birds of a Feather Learn Together? An Experimental Study on How Group Composition Affects Social Learning

Abstract
People learn by collecting their own private signals and observing others. The effectiveness of social learning relies on what information people can observe, how they aggregate information, and with whom they interact in their social group. We conduct a series of laboratory experiments by varying the type of information that subjects can observe and the composition of their social group. We find that people learn differently in homogeneous and heterogeneous groups. When one’s private signals indicate a wrong state, one benefits from learning in both types of groups and learns at a faster speed in a homogeneous group. However, when one’s private signals indicate a correct state, one is more likely to be misled in a homogenous group. The analysis of belief updating processes yields further evidence that subjects in a homogenous group put more weight on others’ information and thus react more. Furthermore, we find that wrong consensuses are more likely to be reached in a homogeneous group.