82

Belgian and Swiss media regularly interfere in French elections by releasing polls and predictions before polling stations close. We exploit the unusual timing and degree of confidence of polls in the second round of the 2017 presidential election to investigate their effect on voter turnout. Our analysis compares turnout rates at different times on the election day, in the first and second round, and with respect to the 2012 and 2022 elections. We observe a significant decrease in turnout after the exit polls’ publication. The effect amounts to 1.1 percentage points in the triple-differences comparison with the 2022 election and is stronger in departments close to the Belgian border. We also find suggestive evidence of a small underdog effect, which may have reduced the winning margin by up to 1 percentage point. JEL Codes: D72.