Dead Ends

Abstract

Evidence suggests that individuals and firms are more innovative when subject to low-powered incentives. I offer an explanation based on a characteristic feature of creative work: hitting dead ends. An agent works on successive ideas, each of which may lead to a breakthrough with some probability. At any time, the agent may abandon her current idea, incurring delay to come up with a new one. Larger rewards and greater impatience cause the agent to spend more time on each idea and to work on lower quality ideas. I subsequently consider a planner who can choose tax or subsidy policies to maximize the value of research spillovers. If spillovers are large and relatively certain, then optimal policy favors the use of subsidies. If spillovers are concentrated among the highest quality ideas, then optimal policy favors taxes. The results highlight why policy makers should care about the structure of incentives for innovation beyond simply encouraging more overall investment.

Keywords: Innovation, Experimentation, Exponential Bandits

JEL Classification: D81, D83