Information on the Polls, Beliefs, Turnout and Party Votes in a Close Election: Experimental Evidence from South Africa

  • 14 Nov 2017

    17:00-18:00, Clay Room, Nuffield College

  • Political Science Seminars   Add to Calendar
Speaker: Kate Orkin

Merton College, University of Oxford

This event is part of the Political Science Seminar Series.

Abstract: How does increased electoral competition affect citizens’ participation, vote choices and engagement? Theory and lab experiments predict turnout is higher in competitive elections where votes are more pivotal, but are ambiguous on how information about likely results from pre-election polls affect partisanship: people may swing behind the leading party (bandwagon effects) or the party which is behind (underdog effects). There is limited well-identified evidence in real elections, none in the developing world, and none in contexts where voters are experiencing meaningful electoral competition for the first time. In South Africa’s 2016 municipal elections, the dominant party, the ANC, suffered a surprise upset, losing Johannesburg and two other major cities by small margins. I conducted a field experiment with 2,023 participants from around Soweto, the largest black township in Johannesburg. Pre-election, few believed the ANC would lose or the election would be close. I provided two treatment groups with face-to-face information that the election would be close from a (poorly publicised) IPSOS-Mori poll. One group were told the challenger party was ahead and one that the incumbent party was ahead. A placebo group got a get-out-the-vote message and all other information contained in the treatment messages, except for information on how close the race was. I collect face-to-face survey data from a baseline survey and two follow-ups, one conducted after treatment and another after the election. I examine how information on competitiveness affects verified measures of voter turnout and partisanship and both self-reported and behavioural measures of political engagement. I can examine alternative mechanisms through which information affects behaviour, through detailed data on voters' expectations about electoral outcomes, party identification, ratings of party viability, and political efficacy.

The Political Science Seminar Series is convened by Geoff Evans, Elias Dinas and Sergi Pardos Prado. For more information on this or any of the seminars in the series, please contact