How Electoral Context Shapes the Spread of Misinformation

• In Event: The Diffusion of Misinformation

Abstract

This paper develops and tests a formal model examining the conditions under which misinformation spreads more rapidly in the context of a social network.

Standard models of the spread of misinformation and group polarization assume that agents have identical preferences and all prefer an equilibrium in which the population learns the truth. However, in the context of an election, individual preferences for candidates vary and citizens prefer that information supporting their preferred candidate spreads while negative information about their candidate does not. In this paper, we construct and analyze a model of the spread of true and false information in which individual voters hold such preferences. In particular, voters are less likely to share information that hurts their preferred candidate, even if they believe that information. We show that even when no one knowingly shares false information, population beliefs can shift to more extreme positions. Beyond this, we examine the contexts under which misinformation is most likely to spread.

We test model using a survey experiment that examines when people choose to share new information they receive in an electoral context. We vary three elements of the context: (1) the two candidates’ relative positions, (2) whether the upcoming race is a primary or general election (and whether the primary is partisan or not), and (3) the election’s competitiveness. In addition, we vary whether how negative the information is for a given candidate as well as the likelihood the information is false. We expect that the probability of sharing negative information about a person’s opposing party is positively related to the size of the stakes in the election – either for an individual (measured as the difference in the distance between their two alternatives and their likelihood of being pivotal) or for the election (measured by the overall distance and whether it is a general election or primary), regardless of how likely the information is to be false. (The experiment is currently under IRB review and expects to be in the field in late January and early February.)

The paper has implications for understanding misinformation and party polarization. Notably, the model suggests that closed primary elections may lead to greater polarization and more misinformation than alternative systems, such as the top-two or top-four primaries used in California and Alaska.

Authors

• PJ Lamberson, UCLA

• Georgia Kernell, UCLA