Technological progress is typically a result of trial-and-error research by competing firms. While some research paths lead to the innovation sought, others result in dead ends. Because firms benefit from their competitors working in the wrong direction, they do not reveal their dead-end findings. Time and resources are wasted on projects that other firms have already found to be dead ends. Consequently, technological progress is slowed down, and the society benefits from innovations with delay, if ever. To study this prevalent problem, we build a
tractable two-arm bandit model with two competing firms. The risky arm could potentially lead to a dead end and the safe arm introduces further competition to make firms keep their dead-end findings private. We characterize the equilibrium in this decentralized environment and show that the equilibrium necessarily entails significant efficiency losses due to wasteful dead-end replication and a flight to safety - an early abandonment of the risky project. Finally, we design a dynamic mechanism where firms are incentivized to disclose their actions and share
their private information in a timely manner. This mechanism restores efficiency and suggests a direction for welfare improvement.