To Vote or Not? Testing Micro-level Theories of Voter Turnout in South Africa’s 2014 General Elections

ABSTRACT

Empirical studies rarely examine the reasons why some South Africans choose to vote at elections and others not. In South Africa’s early democratic elections scholars assumed that voting was an expression of group loyalties or strong party loyalties. However, the decline in turnout in recent elections appears to contradict these socio-psychological explanations. Using 2014 post-election public opinion data, the paper addresses this research lacuna in the South African literature by testing several theories of voter participation. It finds that the influence of one’s social networks and, to a lesser extent, psychological engagement with politics are central to voter turnout. Trade unions and churches, and political parties, despite their obvious efforts to mobilise voters, make little impact on the decision to vote. Media exposure is a weak mobiliser but perceptions of media bias are important, suggesting that people filter out or select among media sources to suit their partisan proclivities, and this exercise in itself encourages turnout.

Disclosure statement

No potential conflict of interest was reported by the author.

ORCID

Collette Schulz-Herzenberg http://orcid.org/0000-0002-6039-801X

Notes

1 Turnout/VAP is the standard measure for voter turnout in cross-national research. See Franklin 2004, 86–7; Norris 2002, 41; Powell 1980, 7.

2 ‘Young voters’ are conceptualised differently across numerous studies but usually understood to comprise voters between the ages of 18–24 and certainly under the age of 30 years of age.

3 The survey was conducted nationally following the 2014 elections and includes 1,300 personal interviews. The samples were drawn using multi-stage, stratified, area cluster, probability sampling. The CNEP is a multi-national project that studies political communication and social structure within the context of election campaigns using compatible research designs and a common core of survey questions.

4 Multinomial regression treats categorical predictors as factors in the regression thus highlighting important differences among groups within an individual predictor. Binary logistic regression does not provide these decomposed effects.

5 Nagelkerke R² is similar to the R² in linear regression in that it provides a gauge of the substantive significance of the model. See Field 2009, 269.