Aligning Voices: Shared Election Polls on Social Media Reflect Voter Wishes

Pre-election polls are published almost continuously, particularly in the run-up to elections. Increasingly, polls are commissioned and published by the media and gain traction through their propagation on social media. But polls differ along many dimensions, including in their estimated vote shares and the corresponding portrayal of the state of public opinion. They also have substantial differences in methodological aspects. We rely on recent evidence about how people interpret poll information in mixed-poll environments to argue that these factors are likely to affect people’s willingness to give credence to election polls and share them with their networks on social media. To test this, we conduct two survey experiments, one in Spain and one in the USA. In the first, we expose Spanish voters (n=830) to a random selection from 78 recent, real, 2023 Spanish General Election vote intention polls. We study several dimensions of these polls: estimates for the then six biggest national parties, b) the date of publication, c) the polling firm, d) the media company that sponsored the poll, e) the party leading in the poll, f) the lead margin, and g) the sample size. While our first study emphasizes realism and ecological validity at the cost of causal identification (randomization is between polls and not features) and internal validity, our second study flips this compromise on its head: an unconstrained conjoint analysis presents US voters (n=500) with a selection of abstract, hypothetical 2024 Presidential Election polls. This allows us to estimate the causal effects of each poll feature because here all features for each poll are randomized. Across both studies, we find no evidence of any significant effect of polling firms, fieldwork dates, or sample sizes on perceptions of credibility and intentions to share polls. However, we find some evidence that the publishers (i.e., media outlets) of the polls could have an influence. Above all, our results clearly indicate that the only factor consistently affecting voters’ proclivity to share polls with their networks is the result of the poll itself: voters are much more likely to share polls that are more favorable to their preferred parties or ideological dispositions. Our findings are consistent with previous evidence that suggest that voters’ reception of poll results may be driven more by directional goals than accuracy goals, and that motivations to affiliate with audiences drive sharing of information on social media. We discuss the implications of our results concerning research on partisan-motivated reasoning and standard Bayesian forms of information processing.