Strategic experimentation in patent races

We study a patent race between two …rms as a two-armed bandit model. The …rst …rm that successfully completes two phases (R&D) acquires a patent license. Each …rm can learn from its rival’s actions and outcomes. We show that two possible ine¢ ciencies can be observed in equilibrium. On the one hand, spill-overs of information reduce the expected pro…tability of the patent and therefore …rms may invest inadequately in , in the fear of releasing good news to the market. On the