Searching for Breakthroughs

Abstract

We consider a decision maker (DM) who engages in costly search for breakthroughs between two uncertain alternatives, with the option to search the alternatives sequentially or in parallel, or to stop at any time and adopt one alternative or a safe status-quo outside option. Alternatives are either good or bad, and only good ones promise breakthroughs. We show that searching the less promising alternative is always preferred under high search costs and a low outside option. Low search costs (a high outside option) prompt the DM to search the more promising alternative when she is optimistic (pessimistic). The option of searching both alternatives in parallel can also make the more promising alternative more preferred, and it can be optimal to search in parallel even when the DM’s beliefs about the two alternatives are dissimilar.

Keywords: exponential bandits, Bayesian learning, sequential search, parallel search, R&D

JEL Classification: D83, L15, M31