Poster presented at 8th Behavioral Macroeconomics Workshop (Vienna, 2026).

Abstract: We analyze a setting of duopolistic learning where two managers compete to invest the pooled resources of society, offering strategies that balance the exploration of new investment opportunities with the exploitation of known ones in the canonical Multi-Armed Bandit setting. We demonstrate that though it is socially optimal to explore new investment opportunities, competitive pressures force managers to offer purely exploitative strategies.  This results a suboptimal learning process that enters inefficient absorbing states, highlighting a market failure in a sequential learning setting with parameter uncertainty.