Events

Prof. Jussi Keppo
Professor
Department of Analytics & Operations
National University of Singapore

ABSTRACT

This paper investigates the effect of crowd composition on the predictions made by individuals with private information. The individuals are heterogeneous in their external motivations, either tending towards conformity or contrarianism. We find that strategic tendencies become more extreme as the crowd becomes more conformist, and that a weakly contrarian crowd leads to the best performance for a randomly selected individual. When using simple averaging to combine individual predictions, a diverse crowd is never superior to a homogeneous one. We propose a clustering-based method to identify subgroups within the crowd and use weighted averaging to combine their predictions, taking into account the extent of private information sharing influenced by the different strategic motives. We demonstrate the effectiveness of the proposed method through a simulation study.