Popularity Bias in Anonymous Matching Markets : Efficiency, Errors, and Fake Views

Abstract:Popularity bias -- the bandwagon effect -- is the tendency to select the most popular from alternative(s). We incorporate this into a parsimonious dynamic model of an anonymous bipartite economy with heterogeneous agents and show that anti-egalitarian networks (where the smallest possible number of members of the long market side are maximally connected) arise. Our results prove that long-run market efficiency is ensured only when we allow mistakes in creating and severing links, mistakes which also favor popular nodes.We find that a winner-take-all network can occur if popularity is a tie breaker, with occasional mistakes ensuring a maximum number of connections. We discuss implications for search algorithm mechanisms and “fake” views and reviews on social media