Dynamic coordination with switching costs

We study dynamic coordination with costly switching, where players are initially uncertain regarding how to coordinate. We model two players searching for compatible platforms in the presence of two kinds of frictions: noisy outcomes and costly switching. The model predicts that players remain on their current platforms whenever their common belief is sufficiently optimistic; otherwise, they randomize their actions, switching with higher probability whenever they are more pessimistic about compatibility, switching is less costly, or success conditional on compatibility is more likely. In an experiment, we vary (i) whether success is Deterministic or Stochastic when players are on compatible platforms and (ii) the cost of switching. We allow subjects to choose explicit mixed strategies to provide a better test of the theory. The data are consistent with most comparative statics predictions, particularly in the Deterministic case, but differ from some of the other theoretical predictions. To explain the disparities between theory and the data, particularly when learning is noisy and switching is costly, we discuss the impacts of several factors, including cognitive ability, inattention, sensitivity to recent failures, and experience.