Connectors and Influencers

Abstract

This paper presents an experiment on activity in networks and network formation. The theory predicts a star network. In small groups, the equilibrium is a pure influencer outcome (only the hub chooses positive efforts). In large groups there exists, in addition, a pure connector outcome (hub chooses zero effort, peripheral nodes positive efforts). We test these predictions on a new experimental platform with asynchronous activity in continuous time with small and large groups –ranging from four to hundred subjects. We find that, in all groups, subjects form a generalized star network with small distance, low indegree, and high link inequality. In small groups, subjects choose efforts in line with the pure influencer outcome. In large groups, hubs choose very large efforts (in excess of pure influencer outcome) or they choose very low efforts (close to a pure connector outcome) in response to information feedback on others’ payoffs. We suggest that departures from equilibrium are caused by the complexity of the network environment.