Gradual Learning from Incremental Actions
-Micro Theory Seminar
Abstract: *Joint with Tuomas Laiho and Pauli Murto
We study an experimentation problem where feedback from actions arrives gradually over time. Because current choices have persistent effects, the problem has two state variables: a summary of past actions and the current belief on the state of the world. We solve the decentralized equilibrium, where small agents choose when to take an irreversible action, and the socially optimal policy, which takes into account the social value of information. There is a novel informational tradeoff on the social level as acting today speeds up learning but postponing actions means more informed choices. We then show how different experimentation patterns - including the socially optimal policy - can be implemented as a decentralized equilibrium by using simple transfers that only depend on the cumulative stock of past actions. Applications include durable good monopoly and investments under uncertainty for which we provide new economic insights.