We exploit a quasi-experimental setting provided by an election day with multiple polls to estimate the effect of voters' turnout on the spread of new COVID-19 infections and to quantify the policy trade-off implied by postponing elections during high infection periods. We show that post-poll...