Do Polls Change Votes? Evidence from Elections in the US, the UK, and Canada

• In Event: Polling, Measurement, and Survey Effects

Abstract

The potential for pre-election polls to influence voting behavior has long interested both political scientists and political analysts alike (Ansolabehere & Iyengar 1994, Ceci & Kain 1982, Fujiwara & Sanz 2017). Both behavioral and rational theories of politics suggest channels through which the information conveyed in publicly released polls should cause certain voters to alter their decision to vote and whom to vote for. Bandwagonning theories, for instance, suggest a psychological disposition to vote for the expected winner as signaled by pre-election polls (Nadeau et al. 1994). Theories of strategic voting, on the other hand, suggest that voters should vote for their most preferred viable candidate, where one major informant of expectations over candidate viability is political polls (Cox 1997). Existing empirical research on this subject ignores fundamental causal identification challenges or relies on lab experiments with limited external validity. We improve upon existing methods by conducting an interrupted time series analysis using data from recent elections in Canada, the United Kingdom, and the United States. By using both publicly released polls to measure the information the mass public is receiving about candidates’ electoral fortunes as well as private survey measures of vote choice that were not released during the election, we are able to measure the influence of public polling on self-reported vote choice. We explore whether there is evidence for bandwaggoning and strategic voting as well as whether these behaviors vary based on electoral context across countries and electoral context. We also evaluate whether the influence of public polls on vote choice is mediated by the amount of news coverage the polls receive, as measured using Twitter data.

Authors

• Margaret Schwenzfeier, Harvard University

• Georgina Evans, Harvard University