Lying by Omission: Verifiable Communication in Large Populations
Abstract: We study a model of motivated communication in large populations. Agents meet randomly. In each meeting, the sender has some hard information about the state of the world and chooses how much to share with the receiver. Information received from other agents can in turn be shared in later meetings. Agents' preferences are misaligned due to private biases; this tempts senders to manipulate their messages by cherry-picking signals, i.e., lying by omission. We ask: how much information is transmitted in equilibrium, and what types of signals are more likely to be disseminated? We show that, if initial information and meetings are abundant enough, most agents become informed, but this requires much stronger parameter assumptions than if agents were to communicate truthfully. In addition, signals that are close to the mean are much more likely to be disseminated. The reason is that extreme signals are communicated in a predictable direction, which stifles their propagation: e.g., high signals are only shared to lower-bias agents to increase their beliefs, but low-bias agents are unlikely to pass them on further. On the other hand, moderate signals can travel in both directions. However, this tendency for moderate information to predominate may be overturned if agents can personalize their news sources, since biased agents would rather obtain extreme signals to use them as ammunition.
Abstract: We study a dynamic model of elections where biased candidates compete through ability investments and platform choice. Good politicians facing weak competition extract policy rents, which lowers welfare. Term limits alleviate this problem in two ways. First, they make the current election less relevant for continuation utility; this makes inferior candidates less unattractive, which helps discipline the winners. Second, incumbents exacerbate rent extraction by deterring challenger entry; term limits prevent this by creating open elections. However, as term limits lower incumbent quality, their overall impact is ambiguous. We find that strong limits are better when politicians are more biased, and challengers' entry cost is intermediate. We also analyze the optimal structure of term limits. If agents are myopic, classic term limits--given by a cap on reelections--are effective. But they are inefficient if agents are strategic, as they incentivize voters to discriminate against challengers, which discourages entry. Instead, in that case stationary term limits, which trigger open elections randomly, are best.
Research in Progress
Ideological Drift in Organizations
Abstract: We consider a model of dynamic policy selection in organizations with free entry. There is a population of potential members with heterogeneous policy preferences. At the start of each period, the existing members vote to choose a new policy. Afterwards members can exit the club and outsiders can join, given this new policy; the process repeats at the end of the period. We characterize the set of equilibria and show that all equilibrium paths converge to the same steady states. If the willingness to belong to a club is a function of policy distance, stable steady states correspond to modes of the preference distribution, leading to centrist organizations. However, if extremist agents are more willing to belong to a moderate club than vice versa, they can “poison” and take over the club despite being less numerous. In addition, a small change in the distribution of preferences may radically alter the set of steady states, leading the organization to drift to a very different policy. Thus, a constitution that limits the pool of members or discriminates against some subset with distinct preferences may have a large impact on the long-term evolution of the club. Finally, we consider a variant of the model where agents have common values but heterogeneous beliefs, and can learn over time. Learning brings most agents in the game to an agreement. However, counterintuitively, in this case the organization may be taken over by a progressively shrinking set of members with extreme beliefs, an even starker result than under the assumption of heterogeneous preferences.
Information and Coordination in Multicandidate Elections