Strategic Inaccuracy

• In Event: Information Transmission

Abstract

Political attacks are the results of strategic decisions by the actors who choose how much to learn about the issue at hand and whether to politicize it. I present a formal model in which two parties with opposite preferences about an outcome decide whether to attack each other and reveal the truth about a scandal or remain silent. Importantly, they receive a public signal before making the decision, and one party - the designer - can strategically choose how accurate this public signal is.

The model shows that when the status quo division is sufficiently even, the designer optimally chooses a completely uninformative signal that leads to both parties remaining silent in equilibrium. As will be detailed later, despite this being a very adversarial context, both parties often prefer this silence over receiving perfect information about the state of the world. Additionally, parties in equilibrium may ironically attack the other party more as the status quo becomes more favorable. This occurs when the outcome is less likely to be in favor of the parties. It would seem intuitive that such conditions will only decrease a party's incentives to attack the other party since one would prefer to keep the status quo and avoid a disadvantageous gamble. However, the designer who foresees this probability optimally sets a lower accuracy for herself (and higher accuracy for the other player) that ultimately increases the probability of conflict. Endogenous accuracy can therefore reverse the relationship between the status quo division and the likelihood of political attacks. Third, parties can have preferences for equality in the status quo division. Parties sometimes prefer to have less initial share of the division, as a higher status quo division can lead to the designer optimally choosing some signal inaccuracy that leads to worse payoffs for the parties. Such preferences for a lower share of the pie allow parties to sometimes agree on a division; a weakly Pareto optimal division exists when the designer’s status quo division is sufficiently large. In an extension, I find that when parties in equilibrium agree on a new division, they always prefer a more equal division which leads to both parties staying silent in equilibrium. Lastly, in another extension of the model, as in the recent literature on Bayesian persuasion and information design, I endow the designer with full flexibility in her choice of information. I find that this additional freedom can increase the probability of silence despite the innate divergence of interests between the parties.

Author

• Jenny Seoyeon Kim, Emory University