Polls, Forecasts, and Voters’ Perceptions of Uncertainty

• In Event: Challenges Associated with Media Coverage of Polls and Election Forecasts

Abstract

This paper uses a conjoint design to examine the relationship between common journalistic practices around the presentation of polling data and the degree to which news audiences may become overconfident that a given candidate is likely to win a hypothetical election. The survey experiment, conducted on a nationally representative population in the U.S., tests how people process polling data when it is expressed as a single result versus an average of recent polls versus a probabilistic forecast—across a variety of vote share conditions and levels of undecided voters. Journalistic strategies thought to mitigate tendencies toward overconfidence and misinterpretation are also examined including language about polls’ margins of sampling error and visualizations illustrating ranges of possible outcomes. Moderating factors including numeracy and education are also examined. Our results help to empirically ground discussions about ways to improve coverage of survey data and public opinion by accounting for how the public actually understands and makes sense of such numbers.

Authors

• Michael Decrescenzo, University of Wisconsin, Madison

• Benjamin Toff, University of Minnesota

• Zach Warner, Cardiff University