Full Text

For voters, estimating beforehand which candidates will receive many or few votes in an upcoming election is valuable information. Probabilistic election forecasts help voters brace themselves for adverse outcomes or mobilize themselves to bring about favored ones, and constitute a critical prerequisite for strategic voting: By estimating which candidates stand a chance, voters can adjust their vote choices between multiple acceptable options, trying to maximize the impact of their vote. Especially in two-round voting systems such as the French presidential elections, tactical estimations of candidates’ chances are critical for making one’s vote count.
While strategic voting has been widely studied, we know little about how voters rely on the news and other sources of information to gauge candidates’ differential chances at receiving a sufficient share of votes. In this study, we distinguish two main mechanisms that may explain voters’ differential probability estimates: On the one hand, ongoing news coverage informs voters’ expectations, as journalists, pollsters, pundits and other commentators give visibility to those candidates deemed most relevant and expressly discuss their respective chances. On the other hand, voters have relatively stable, intrinsic reasons for believing in the viability of candidates’ bids based on their political party preference, para-social sympathies and other forms of motivated reasoning.
To determine how voters gauge candidates’ probabilistic electoral prospects, we examine how French voters in the 2022 presidential elections predict candidates’ respective chances and adjust them over time to accommodate new developments. We draw upon a four-wave panel survey (N=1,294) coupled with a large-scale news content analysis of nine opinion-leading outlets, using a longitudinal linkage-study design, which links voters’ expectations to the media contents that they are exposed to. Focusing first on voters’ candidate-specific probabilistic estimates, a series of dynamic panel models (DPM) documents a dominant influence of wishful thinking, while information updating accounts for systematic adaptations only with regard to a handful of key contenders. Subsequently, predict voters’ alternative predictions regarding which candidates are expected to advance to the second, run-off round of elections, and eventually win the presidency. Two multi-modal dynamic panel models (estimated using GEE) overall confirm the prevalence of wishful thinking-based predictions, but identify important between-candidate dynamics, wherein exposure to hopeful news about one candidate significantly diminishes the chances attributed to specific rivals, but may still boost the chances attributed to third candidates within the same political camp.
Our study contributes to the scholarly debate on the electoral effects of opinion polling by considerably expanding the range of information capable of informing voters’ predictions. Our simultaneous assessment of wishful thinking and information effects adds perspective to the prevalent focus on news coverage, identifying important contingent influences. Finally, we present a powerful statistical approach for modeling the interdependent dynamics of electoral predictions in complex multi-party elections, wherein new information about one candidate may entail differential implications for other contenders. We discuss lessons for election research and wider public opinion scholarship.