Toward Performance-based Politics: Swing Voters in South Africa’s 2016 Local Elections

ABSTRACT

Who are the swing voters in South Africa’s elections? This study is among the first to systematically investigate the correlates of the swing vote in South Africa. The paper argues that race, cohort, performance, and partisan networks influence the likelihood that an individual is a swing voter. To investigate these arguments, this study uses original exit poll survey data from South Africa’s 2016 local elections. The results indicate that swing voters are those who have weaker racial identities, weaker attachments to their racial group’s party, are ‘born free’, have lower assessments of ANC performance, and have fewer friends and family who support their preferred party. The paper also predicts what drives swing voters to support a certain party. The results present key implications for race and identity-based voting in South Africa and dominant regimes across the continent and suggests that South Africa’s elections are not clearly a racial census.

Acknowledgement

The author would like to thank the Program on Governance and Local Development (GLD) at the University of Gothenburg for funding the project.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 The paper does assume, in line with Przeworski’s (2015) contention, that a transfer of power is an indicator of a strong democracy. Further in the South African context, this assumption suggests that the erosion of ANC electoral support and thus more black swing voters will strengthen democracy.

2 The term ‘born free’ is contested with some preferring ‘post-apartheid generation’; however, I use the former term as this is the term most used in the academic literature.

3 Coloured support is of course more nuanced than presented here (i.e. Coloureds in the Western Cape tend to support the DA at higher rates compared to elsewhere in the country), but for our purposes here, I focus on the general tendencies of groups, which tends to be the approach of racial census arguments.

4 Further, when coding partisanship purely as the party one identifies with, closeness to the ANC, but not to the DA, reduces the probability that one will be a swing voter. Clearly ANC partisanship has an effect on swing voter status regardless of race, but, given this coding, it is difficult to know how race influences partisanship and swing voter status.

5 These expectations are in line with the broader literature on rational voting based on performance records (i.e. Miller and Wattenberg 1985).

6 Tshwane municipality did see a number of violent protests in the run-up to the elections regarding the imposition of an ANC mayoral candidate that many people disliked, which further illustrates it uniqueness.

7 Only 8% of the survey respondents did not reveal how they voted in the 2016 elections; therefore, bias due to non-response is likely limited. Those who did not reveal their preference are more likely to be white, less educated, less embedded in partisan networks, and have lower assessments of the ANC’s performance. The model controls for these variables.

8 I characterise the EFF as a viable non-white alternative to the ANC as it and the ANC are the only black parties in Parliament with more than 5% of the vote. It is not viable in the sense that it can defeat the ANC but rather that it is a party black voters feel they can support if they are unhappy with the ANC and/or are unwilling to vote for a ‘white’ party as many perceive the DA to be.

9 The exact wording was: ‘People identify themselves in a variety of different ways; how do you identify yourself?’ 23% identified in religious terms, 22% in racial terms, 11% in ethnic terms, 10% in terms of nationality (i.e. South African), and 9% in terms of gender. Less than 8% identified in terms of a single other identity such as occupation, class, education, region, etc.

10 See the appendix for the questions used to measure linked fate.

11 Of course, voters may abstain for a variety of reasons (illness, unable to pay to travel to the polling station, etc.), but from over 20 qualitative interviews conducted throughout the research, it seems that many ANC supporters, in the face of corruption scandals, abstain rather than vote for another party which does suggest they are relatively more up for grabs than those ANC supporters who do turn out to vote. I therefore assume that most abstainers are in fact deciding to abstain and that those who were unable to vote for other reasons are relatively few. This seems reasonable given that the plurality (49%) of those who abstained in 2014 voted for the ANC in 2016.

Additional information

Funding

This work was supported by Program on Governance and Local Development (GLD), University of Gothenburg, Sweden.