Talking Points: A Bandit Theory of Political Agency


Politicians frequently need to announce their positions on issues before they have access to relevant information. If subsequent information gathering leads to evidence that contradicts the initial statement, the politician may have to admit to mistake. For politicians concerned about their reputation, this creates an incentive to avoid producing such evidence. This paper presents a dynamic model of a reelection-seeking politician with unknown expertise who chooses, in each period, which policy to collect information on from a set of policies with unknown payoffs. Reelection concerns lead the politician to diverge from an efficient information collection strategy and result in sub-optimal policies.