The effect of crime on South Africans’ voting behaviour

ABSTRACT

What effect does crime have on South Africans' electoral behaviour? Do crime victims and citizens who fear crime enter or exit the electoral arena? If they do vote, do they sanction the ruling African National Congress (ANC)? We use two rounds of Afrobarometer data to provide answers. We corroborate earlier findings that crime victimisation does not appear to shape electoral behaviour. Further, we find that sociotropic perceptions of crime do influence vote choice, although not always in the predicted direction. We explore who tend to be South African crime victims to cast light on these results. We find that opposition parties, specifically the Democratic Alliance (DA), have benefitted electorally from citizens' crime perceptions. We contend that the party's significant focus on crime help account for their ownership of the issue. Lastly, relative to other government performance evaluations, crime does matter, and its salience and importance have increased over time.

KEYWORDS:

• Crime
• voting behaviour
• South Africa
• issue ownership

Acknowledgement

We are very grateful to the two anonymous reviewers for their thoughtful suggestions that improved the paper. We also wish to thank Andrew Wyatt for his excellent feedback that helped shape the final product.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 As a series of robustness checks, we ran the same models with these variables as ordinal. The results do not change.

2 Similarly, to the above, we ran the same models with these variables coded as binary. The results in the models do not change.

3 We do not include survey respondents’ partisan preferences in our modelling of vote choice. We do so for three reasons. First, we are interested in the specific effect of crime and the fear of crime on voters’ behaviour. Disentangling that effect from race required significant effort. Given our limited word count, partisan identity would require additional efforts. Second, our multinomial models already have limited responses for non-ANC vote choices. The addition of party attachment, in a context of decreasing partisan identity, would likely absorb much of our variation. Third, including a party identity variable would almost certainly introduce endogeneity into our models. It is likely that many voters were attracted to some party options due to their policy offerings or solutions on crime. This serves as part of the focus in our discussion section.

4 We calculated variance inflation factors (VIF) for all explanatory variables in our models. As a rule of thumb, we set a score of 4 or more to be the threshold for multicollinearity (Ferree & Horowitz, 2010). None of the variables in any model registered a score greater than 1.72.

5 One possibility was that these variables, given the salience of crime, lacked enough variation to allow for statistical testing. We ran a series of histograms to check their distributions and each possible value for both questions had at least 10.7 per cent of the responses, reflecting considerable variation across variables.

6 The manifestos all differ in length and increase in pages between 2014 to 2019. As a check that our results were not just driven by document length, we also examined the mentions of corruption, land and jobs. As one would expect, the DA focused on corruption to a similar degree as crime, while the EFF focused to a greater extent on land and jobs. The ANC gave roughly equal share to all four issues, but slightly fewer mentions of both corruption and jobs in 2019. This leads us to conclude that the number of mentions is an adequate measure, albeit imperfect, of issue salience and is not just driven by document length. Rather, the numbers reflect a prioritisation of crime most strongly by the DA, and the EFF to a lesser extent.

Additional information

Notes on contributors

Robert Nyenhuis

Robert Nyenhuis is an Associate Professor of Political Science at Cal Poly Pomona.

Thomas Isbell

Thomas Isbell is a Postdoctoral Research Fellow for Afrobarometer. Their research considers questions of voting behaviour, the appeals that candidates make during elections, and questions of democratic quality.