Truth-telling Outcomes in a Reputational Cheap-talk Game with Binary Types

Abstract

Experts with different abilities of information acquisition who receive multiple pieces of signals over time can choose the timing of recommendation and whether to be truthful in a later period, when a recommendation is made in an earlier period. Giving inconsistent recommendations may be seen as a sign of a poor information acquisition ability, but it can also work as a “safety net” that prevents the worst reputation. This study uses a simple binary-ability framework to capture this aspect and proposes equilibriums where all information is delivered truthfully on the path. I examine when such an equilibrium exists, and compare such equilibriums with those where only partial information is delivered; it is found that the former brings higher expected payoffs to the expert than the latter under a certain range of parameters when the utility function is strictly convex in the reputation.