Strategic Experimentation with Random Serial Dictatorship
49 Pages Posted: 22 Apr 2017 Last revised: 13 Sep 2019
Date Written: April 17, 2018
We consider one-sided matching problems with costly information acquisition about preferences. As information is costly, agents are willing to acquire information only if it is relevant to their decision making. Hence, whether agents acquire information crucially depends on their beliefs about their choice set (i.e., the set of objects each agent can obtain through some type reports). In this study, we fix the assignment rule to be the random serial dictatorship, and investigate the information design for choice-set disclosure policies. In a stylized environment, we demonstrate that the full-disclosure policy, which perfectly discloses each agent's choice set, is typically ex ante Pareto inefficient, because it fails to exploit the positive externality effect of information acquisition. Then, we illustrate that by disclosing less information about choice sets, we can induce more efficient information acquisition. Finally, we show that, in the worst case, the loss of the full-disclosure policy relative to the optimal one is large.
Keywords: One-sided matching, Information acquisition, Bayesian persuasion, Strategic experimentation
JEL Classification: C78, D47, D83
Suggested Citation: Suggested Citation