Daniel Monte (Sao Paulo School of Economics – FGV)

“Information Design under Markovian Rules”

Abstract: An uninformed long-run designer wishes to persuade short-lived agents to invest in a project of fixed, but unknown, quality. Motivated by privacy regulations and limited record keeping, we assume a designer restricted to use Markovian communication rules. Messages serve a dual purpose: they incentivize actions and enable learning. We identify conditions under which the optimal policy is a direct recommendation (two messages), as well as conditions under which the designer’s payoff increases with the number of messages. The optimal policy qualitatively depends on the conditional payoff distributions. If failures are very informative, a two-state rating system approximates the first-best payoff. In contrast, if successes are more informative, the designer will not approach the upper bound.