Davide Cantoni (LMU Munich)

“Do Polls Affect Elections? Evidence from Swiss Referenda”

Abstract In recent years, the outcomes of several high-stakes votes have not been correctly predicted by public opinion polls. We propose that polls themselves may shape election outcomes by affecting voter turnout. We first present evidence that polls have a causal effect on voter turnout, with greater turnout in response to close polls. To identify this effect, we exploit the precise day-level timing of the release of poll results for federal referenda, and a dataset on daily mailin voting for the canton of Geneva. The release of a closer poll causes turnout to sharply rise
immediately after poll release, with cumulative turnout remaining significantly higher through election day. Turnout rates are no different in the days prior to the release of close polls, suggesting that the information contained in the polls was not anticipated, issue types were not perceived as different, and the political “supply side” was not differentially active. We generalize our findings by specifying a model of voter turnout depending on locally available information and on the existence and closeness of polls. We confirm the predictions of the model using municipality-level data from across Switzerland, exploiting the introduction of national polls in 1998. We then provide evidence that this turnout effect can shape election outcomes. We find that the effect of close polls is heterogeneous, with an asymmetric effect leading to a higher vote share for the underdog, and effects varying with newspaper coverage of polls. The effect sizes we estimate are large enough to flip high-stakes election outcomes.