Politikon

South African Journal of Political Studies

ISSN: 0258-9346 (Print) 1470-1014 (Online) Journal homepage: https://www.tandfonline.com/loi/cpsa20

The Unconvinced Vote: The Nature and

Determinants of Voting Intentions and the
Changing Character of South African Electoral
Politics

Benjamin J. Roberts, Jarè Struwig, Steven L. Gordon & Yul Derek Davids

To cite this article: Benjamin J. Roberts, Jarè Struwig, Steven L. Gordon & Yul
Derek Davids (2019): The Unconvinced Vote: The Nature and Determinants of Voting
Intentions and the Changing Character of South African Electoral Politics, Politikon, DOI:
10.1080/02589346.2019.1687120

To link to this article: https://doi.org/10.1080/02589346.2019.1687120

Published online: 11 Nov 2019.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=cpsa20
POLITIKON
https://doi.org/10.1080/02589346.2019.1687120

The Unconvinced Vote: The Nature and Determinants of

Voting Intentions and the Changing Character of South
African Electoral Politics
Benjamin J. Roberts , Jarè Struwig , Steven L. Gordon and Yul Derek Davids
Democracy, Governance & Service Delivery (DGSD), Human Sciences Research Council (HSRC), Durban, South
Africa

ABSTRACT
In the lead up to South Africa’s sixth National and Provincial Elections in May 2019, the recent
performance and leadership dynamics within the country’s major political parties raised
fundamental concerns about the potential impact on voter turnout. These concerns were
not unfounded, given that the 2019 General Elections recorded the lowest voter turnout
since 1994, with only 49% of the voting age public participating. Despite this, relatively
little remains known about the factors that differentiate decided voters from abstainers,
undecided voters and undisclosed voters. To contribute further to the understanding of the
determinants of planned electoral participation in the country, this article tests several
dominant theoretical accounts of turnout using cumulative data from sixteen annual
rounds of the South African Social Attitudes (SASAS) series conducted between 2003 and
2018. Specifically, the relative influence of key socio-demographic attributes, psychological
engagement and regime evaluations is examined. The results point to psychological
engagement variables playing a decisive role in separating different categories of voter,
with age and education also exerting an influence. The article concludes by reflecting on
the role of the unconvinced vote in the 2019 Elections, teasing out the implications for
future elections in the country.

Introduction and background

This article explores the question of what drives voting intentions in South African elec-
tions, using data covering the period between 2003 and 2018, and reflects on the chan-
ging nature of the electorate in the country. Despite a national voter outreach
campaign organised by the Electoral Commission of South Africa, the country’s sixth
General Elections had the lowest voter turnout of any national and provisional election
since 1994. In addition, there is mounting concern over a growing disinclination to register
among younger members of the electorate. The 2019 General Elections represented a
pivotal point in the history of electoral democracy in the country. After all, the election
came after a decade of increased intra-party factionalism, a diminishing margin of
victory for the ruling African National Congress (ANC) in local government elections,
high-profile corruption scandals as well as growing discontent with the quality of govern-
ance and pace of socio-economic change. Against this background, understanding voter
behaviour has become critically important. However, relatively little is currently known
about the factors that differentiate decided voters from abstainers, undecided voters
and undisclosed voters.

CONTACT Benjamin J. Roberts broberts@hsrc.ac.za

© 2019 South African Association of Political Studies
2 B. J. ROBERTS ET AL.

The key narrative of Election 2019 arguably relates to declining participation and it is
important to consider this decline in more detail. Despite concerted efforts, voter
turnout in the 2019 General Elections was considerably lower than in previous national
and provincial elections. Electoral participation among registered voters was 66% in
2019, signifying a drop of seven percentage points relative to the 2014 NPE results (see
Table 1). Moreover, turnout as a share of the voting age population (VAP) fell from 57%
in 2014 to 47% in 2019. Many young people did not even register to vote ahead of the
2019 NPE. Voter registration among 18–29 year-olds declined from 58% to 49%
between the 2014 and 2019 elections, while among 18–19 year-olds the reduction fell
more precipitously from 33% to a mere 19%, which translates into a 42% decline
(Schulz-Herzenberg 2019b). Despite lower turnout, the ANC secured an electoral victory
with 58% of the vote. This represented a decline from the 62% achieved in 2014 and is
the third successive drop in a NPE contest since the high of 70% recorded for the party
in 2004.
Since the late 2000s scholars expressed concern over this pattern of declining public
participation in South African elections (Kersting 2007; Schulz-Herzenberg 2007; Kimmie,
Greben, and Booysen 2010), and the aforementioned 2019 statistics are likely to be the
source of substantive focus as the country moves to future elections. In many respects,
this parallels the growing acknowledgment of, and disquiet over, declining turnout
rates in democracies worldwide (Hooghe and Kern 2017). Turnout as a percentage of
VAP in South Africa is lower than many advanced industrial countries, though the
country still fares better than democracies such as France and Switzerland (International
Institute for Democracy and Electoral Assistance 2019). The diminishing voting rate
raises critical questions about what social and political mechanisms might be influencing
voting intentions in South African elections. In this article, we examine the impact of both
socio-demographic attributes and attitudinal factors on individual predispositions towards
electoral participation in South Africa, drawing in particular on select theoretical accounts
advanced in international turnout literature. It should be stated that quantitative (multi-
variate) examinations of electoral behaviour in the country remain surprisingly rare,
although there is a burgeoning body of evidence on both turnout and voting choice
(Roberts, Struwig, and Grossberg 2012; Struwig, Roberts, and Gordon 2016; Schulz-Herzen-
berg 2019a). The present study differs firstly by focusing on voter turnout decisions ahead
of the 2019 General Elections, not only as a binary choice between participation and
abstention, but also accommodating voter indecision and undisclosed preferences and

Table 1. Electoral participation in South Africa during national and provincial elections, 1999–2019
(Count ‘000 and % of the eligible voting age population).
Registered
Voting age population
population (VAP) (RP) Total votes cast Spoilt votes Total valid votes
% % % % % % %
Count Count VAP Count VAP RP Count VAP RP Count VAP RP
1999 22,589 18,173 80 16,480 73 91 251 1 1 16,228 72 89
2004 27,437 20,675 75 16,114 59 78 251 1 1 15,864 58 77
2009 29,957 23,182 77 18,159 61 78 239 1 1 17,920 60 77
2014 32,688 25,388 78 18,907 58 78 252 1 1 18,655 57 73
2019 35,868 26,757 75 17,924 50 67 252 1 1 17,672 49 66
Source: Electoral Commission of South Africa (IEC); Schulz-Herzenberg 2019b.
 POLITIKON 3

how these constituencies vary in profile. It also looks at the recent election against a
broader period of time, by relying on data covering late 2003 to late 2018, as a means
of locating current tendencies against emerging dynamics in political culture. In so
doing, the study aims to extend certain aspects of the modelling undertaken by Schulz-
Herzenberg (2019a) based on data relating to turnout in the 2014 General Elections.
Classic studies of electoral participation suggest that certain socio-economic character-
istics are associated with electoral participation (see, for instance, Campbell et al. 1960;
Lipset 1960; Parry, Moyser, and Day 1992; Verba, Schlozman, and Brady 1995). The
socio-economic status (SES), material wealth and educational attainment of individuals
is said to shape the level of resources (time, money and civic skills) available to facilitate
participation in political activities, such as voting (Brady, Verba, and Schlozman 1995). Evi-
dence on the predictive power of SES variables associated with the resources model
remains weak. Education is nonetheless a stronger predictor of turnout decisions, while
age, and gender to a lesser degree, are other significant socio-economic factors informing
individual differences in turnout (Blais 2000; Blais, Massicotte, and Dobrzynska 2003; Blais,
Gidengil, and Nevitte 2004; Nevitte et al. 2009). Specifically, the weight of evidence
suggests that electoral participation increases with age and higher levels of education,
and is more common among men than women.
One of the most widely recognised determinants of electoral participation is psycho-
logical engagement with politics. Democracy theorists consider low levels of engagement
to be damaging to voter turnout. In the seminal study by Almond and Verba (1963), Civic
Culture, it is maintained that one of the central features of a model citizen in normative
democratic theory is engagement in politics (see also Campbell et al. 1960; Verba, Schloz-
man, and Brady 1995; Carpini and Keeter 1996). In most empirical research, such engage-
ment has been characterised by the degree to which individuals think and talk about
politics. Political interest is consequently considered a good measure of how much politi-
cal information and knowledge an individual possesses (Norris 2000; Inglehart and Welzel
2005; Dalton 2006). A sense of duty to vote and belief in the efficacy of voting is also seen
to play a role (e.g. Blais 2000; Blais, Massicotte, and Dobrzynska 2003; Blais, Gidengil, and
Nevitte 2004; Blais and Achen 2019). Research in South Africa shows that an individual’s
voting efficacy is, to a degree, a product of being on the winning political side (Gordon,
Struwig, and Roberts 2018). This finding demonstrates the influence that political partisan-
ship has on an individual’s beliefs about the political world. The results suggest further
questions on how partisanship may drive political attitudes in the country.
Apart from psychological engagement, Almond and Verba (1963) contended that
certain civic attitudes might further motivate political participation, identifying evaluative
orientations as important. The associated valence politics or regime evaluations model
centres on the idea that ‘citizen involvement will vary according to levels of (dis)satisfac-
tion with the performance of political leaders, the incumbent government and the wider
political system’ (Clarke et al. 2009, 244–245). For Almond and Verba, citizens’ negative
attitudes towards the political system could result in alienation, which erodes their willing-
ness to participate in the system. In a comprehensive analysis of public opinion trends in
South Africa, Struwig, Roberts, and Gordon (2016) found that a considerable segment of
the public lost faith in the political class. Satisfaction with democracy and political insti-
tutions was low, and public trust in key political institutions (such as national government)
had significantly eroded. Such discontent, particularly among the poor, seems to be linked
4 B. J. ROBERTS ET AL.

in large part to widespread public dissatisfaction with government’s efforts to create jobs
and keep people safe (also see Gouws and Schulz-Herzenberg 2016). Disillusionment with
politicians and political parties may lead citizens to embrace what some have termed ‘anti-
politics’ or ‘anti-formal politics’, involving a rejection of conventional forms of political par-
ticipation and a turn towards alternative forms of political action, such as protest action
(Dalton 2006). Concomitantly, a number of studies have been quite critical of the sup-
posed relationship between disillusionment and political behaviour (for, example, Saun-
ders 2014). The question therefore remains whether such discontentment with aspects
of regime performance in the country ultimately sways decisions regarding electoral
participation.
While a sizeable number of other factors have been proposed as potential influences on
turnout, such as mobilising agencies and networks, as well as social context (see Schulz-
Herzenberg 2019a), this article focuses on testing of the individual resources, psychologi-
cal engagement and regime evaluations hypotheses. Following an outline of the data and
methods employed for the study, we systematically examine the role of these socio-demo-
graphic and attitudinal correlates on planned electoral participation using multivariate
modelling. This leads into a concluding discussion that teases out the implications for suc-
cessive elections in the country.

Methodology
The data used to examine the patterns and determinants of electoral participation derives
from the South African Social Attitudes Survey (SASAS), a repeat cross-sectional survey series
conducted by the Human Sciences Research Council on an annual basis since 2003. The
series was designed to be nationally representative of the adult population aged 16 years
or older living in private residence. The sample for each round of surveying consisted of
500 Population Census Small Area Layers (SALs) as primary sampling units, stratified by pro-
vince, geographical sub-type and majority population group. In each of the sampled
localities, seven visiting points were randomly selected for interviewing, followed by the
random selection of a single, age-eligible member in each household using a Kish grid.
Questionnaires were administered using face-to-face interviewing in the respondent’s
language of choice. The realised sample size was for each survey round ranges between
2500 and 3300. Data for this study is available for the period 2008–2014. The research instru-
ments and protocols were approved by the HSRC Research Ethics Committee.
For the present study, the main measure used to examine electoral predispositions is a
constructed categorical variable deriving from a question on the planned intention to
vote. Each SASAS round since 2003 has asked respondents the following: ‘If there were
an election tomorrow, for which party would you vote?’ Those specifying that they
would vote for a specific political party were coded as ‘decided voters’, while those indi-
cating that they would not vote were classified as ‘abstainers’. Two additional categories
were identified based on the pattern of responses, namely ‘undecided voters’ who were
uncertain of their electoral choice, and ‘undisclosed voters’ who refused to voice a clear
preference when asked the voting intention question. Data from the sixteen survey
rounds conducted between 2003 and 2018/19 were combined into a cumulative file, per-
mitting an examination of trends over time and the conducting of multivariate analysis
that covers all years of data while controlling for year effects. The study is confined to
 POLITIKON 5

those age-eligible to vote in each survey round, irrespective of registration status. Trends
in electoral intentions and select relevant independent variables are presented over the
full period from 2003 onwards, while the modelling is restricted to the 11 rounds con-
ducted between 2008 and 2018/19 due to certain key variables not being fielded
during the first five rounds of the SASAS series. The sample size for the combined
2003–2018 data is 47,882, while the 2008–2018 modelling is based on a sample of
33,648 cases. A description of the coding of the independent variables used to test the
resources, psychological engagement and regime evaluations hypotheses is presented
in an Addendum to the article.

Trends in electoral intentions since the early 2000s

Figure 1 presents trends in relation to responses to the intention to vote question over
sixteen rounds of annual SASAS surveying between 2003 and 2018/19. For ease of
interpretation, the pattern for the four category of responses – decided voters (those
who will vote), abstaining voters, undecided voters and undisclosed voters – are presented
on separate line graphs. Firstly, in relation to decided voters, who unequivocally declare an
intention to vote for a particular political party if there were to be an election tomorrow,

Figure 1. Trends in electoral intention among the voting age public, 2003–2018/19 (%). Source: HSRC
SASAS 2003–2018.
Notes: (1) Data is weighted to be nationally representative of the voting age population (18 years and older) living in South
Africa, (2) the vertical axis scale in the decided voters (‘will vote’) graph differs from the other three graphs, with the latter
on a finer grained scale in order to better depict dynamics for these unconvinced voters.
6 B. J. ROBERTS ET AL.

there have been fluctuations over time, with a generally declining trend. In the survey
rounds conducted approximately six months prior to General Elections (the 2003, 2008,
2013 and 2018 rounds), there are discernible dips in the share of the voting age public
openly declaring an intention to vote. Downturns of this type in the pre-election cycle
ranged from 10 to 20 percentage points, and are likely to reflect a combination of influ-
ences. This may include growing introspection about electoral choice in the build-up to
a specific electoral contest as well as a heightened sensitivity to disclosing electoral par-
ticipation preferences during such periods. The share reporting an intention to vote in
the lead-up to the 2019 General Elections (58%) was the lowest recorded over the exam-
ined period.
In contrast with decided voters, the tendency to report planned abstention has been
increasing on aggregate over time. A careful examination of the year-on-year patterns,
shows that there have been ebbs and flows that correspond with political terms of
office and contextual events. During former President Thabo Mbeki’s second term of
office, from 2004 until his ‘resignation’ in September 2008, there is an observable surge
in reported abstention, rising from 5% in late 2004 to 12% in late 2008. This declined
again to 7% in the aftermath of the 2009 General Elections, and during the first term
and most of the second term of the Zuma administration, this fluctuated within a 7–
12% range, with a slight inclining pattern. The results of the 2018 survey, conducted
between November 2018 and February 2019, show a rapid upswing in planned absten-
tion, from 11% in late 2017 to 18%. This is the highest recorded share indicating that
they would not vote if an election were held tomorrow. This corroborates official regis-
tration and turnout statistics, which also displayed a notable downturn, especially
among younger age cohorts (Schulz-Herzenberg 2019b). The decline in the share of
decided voters and a corresponding upswing in planned abstention remains a defining
factor of the 2019 General Elections. Growing absenteeism raises fundamental questions
about the underlying determinants and future trajectory of this emerging pattern of
behaviour.
Beyond the behavioural extremes of confirmed participation and abstention lie the
other two categories of voters presented in Figure 1, namely undecided and undisclosed
voters. Expressions of electoral indecision we especially high in the lead-up to the 2004
and 2009 general elections, accounting for 18% and 14% of the voting age population
(18 years and older) respectively. The 2004 General Elections posed the electorate with
the difficult choice of a second term under the leadership of President Mbeki in the
context of growing discontent about the pace and nature of post-apartheid reconstruction
and development. The 2009 General Elections, however, occurred barely six months after
the recalling of President Mbeki, the rise in electoral fortunes of President Zuma, the out-
break of widespread xenophobic riots and an intensification of protest events around the
country. The run up to the 2009 General Elections was also the period of the global
financial crises and often considered by many economists as the most serious financial
crises since the Great Depression of the 1930s. In a context of complex electoral and
broader political dynamics, the observable spikes in voter indecision is perhaps somewhat
unsurprising. These spikes did not recur in the context of the 2014 and 2019 General Elec-
tions, and with the exception of the 2003 and 2008 survey rounds, the share of undecided
voters has remained relatively unchanged, ranging narrowly between 3% and 7% in the
other 14 years of observation.
 POLITIKON 7

The undisclosed vote is the most difficult group among the electorate to profile. Do
these adults with clear party allegiances but a reticence to openly discuss such identifi-
cation with interviewers? Are they voters with weak levels of partisanship, or even
voters with no particular sense of party attachment at all? Could they even be discerning
swing voters who decide on their electoral choices following campaigning or on voting
day itself based on a careful weighing-up of different factors? It could also be they are a
group encompassing a blend of all the above. What is again apparent is that there are
upswings in the share of undisclosed voters in each of the survey years prior to general
elections falling within the period under examination (the 2003, 2009, 2013 and 2018
survey rounds). Furthermore, like abstention, the share of undisclosed voters shows a
upward trend over time, with the highest reported shares falling into this category in
both 2013 and 2018/19 (17%). Exploring the factors informing lack of disclosure of elec-
toral intentions, and the manner in which this group differs or approximates decided
and abstaining voters would also help to enrich our understanding of the South African
voter.

Testing the relative influence of different theoretical models

Using pooled, annual data from the SASAS survey series covering the period from late
2008 through to early 2019, we tested the theoretical assumptions underlying the individ-
ual resource, psychological engagement and regime evaluation models. Given the nature
of our categorical dependent variable measuring voting intention, we used a multinomial
regression approach to mode voting intention. The base category for analysis are decided
voters, meaning that the models essentially test for the similarities and variances in the
predictors of planned electoral turnout relative to the categories of abstaining, undecided
and undeclared voters. The approach employed is similar to that adopted by Schulz-Her-
zenberg (2019a) in her examination of the determinants of turnout in the context of the
2014 General Elections. Specifically, we begin by independently testing the predictive
effect of the three theoretical explanations. This is followed by a fully specified model
that includes all clusters of indicators as a means of ascertaining which factors and
specific indicators predominate when entered simultaneously. Since the modelling is
based on combined data from eleven consecutive annual SASAS rounds, we include a
control in the models for year of surveying, in order to account for year effects. This is
important given the swings in the relative electoral predispositions of the adult public
observed over this period of time.

Individual resources (socio-demographic traits)

The results of the initial testing of the individual resources hypothesis are presented in
Table 2. Compared to voters that express a clear intention to vote, abstaining voters are
more inclined to be younger on average, Indian or coloured adults, and tend to reside
in formal urban areas rather than in rural, former homeland areas. No gender or strong
educational effects are apparent. Similarly, the main discernible differences between
decided and undecided voters is that the latter are slightly more likely to younger, and
Indian or coloured adults rather than black African adults. Again, no statistically significant
difference is evident in relation to gender or educational attainment, and there is only a
8 B. J. ROBERTS ET AL.

Table 2. Multinomial logistic regression results testing the impact of individual resources on voter
predispositions in South Africa, pooled 2008–2018/19 data.
Abstainer Undecided voter Undisclosed voter
Individual resources model Coef. Sig. Coef. Sig. Coef. Sig.
Female (ref. male) −0.088 n.s. 0.119 n.s. −0.168 **
Age (in years) −0.024 *** −0.015 *** 0.001 n.s.
Population group (ref. Black African)
Coloured 0.426 *** 0.651 *** 0.364 ***
Indian/Asian 1.178 *** 1.215 *** 0.739 ***
White −0.014 n.s. 0.135 n.s. 0.317 ***
Type of geographic location (ref. Urban formal)
Urban informal −0.136 n.s. 0.070 n.s. −0.200 n.s.
Rural traditional authority areas −0.273 ** −0.132 n.s. −0.306 ***
Rural farms −0.200 n.s. −0.275 * −0.325 **
Educational attainment (ref. Post-secondary)
Junior primary schooling or less 0.319 * −0.046 n.s. −0.343 *
Senior primary schooling 0.222 n.s. −0.011 n.s. −0.416 ***
Incomplete secondary schooling 0.199 n.s. 0.024 n.s. −0.375 ***
Complete secondary schooling 0.146 n.s. 0.033 n.s. −0.200 *
Year of survey 0.056 *** −0.025 * 0.040 ***
Constant −113.429 *** 48.013 * −81.475 ***
Notes: (1) Data is weighted to be nationally representative of the voting age population (18 years and older) living in South
Africa, (2) the base outcome is ‘Decided voters’ and (3) the regression model controlled for an individual’s province of
residence. ***p < .001. **p < .01. *p < .05. n.s. = not significant.

marginal effect on type of geographic location. Finally, the undisclosed voter displays a
greater tendency to be male, a coloured, Indian or white adult, and more likely to have
a tertiary education than a completed secondary or lower level of educational attainment.
These findings therefore suggest that, in terms of socio-demographic attributes, the
factors differentiating decided voters from both undecided and abstaining voters are
broadly equivalent. Age and race are the main determinants, coupled with a modest loca-
tional influence. Younger Indian and coloured voters are accordingly more likely than to
voice a preference for abstention or uncertainty about electoral participation. By contrast,
in profile the undisclosed voter seems characteristically better educated, more decidedly
urban-based, and has a greater chance of being male than the decided voter. Whether
these patterns remain, modify or fall away altogether once other attitudinal factors are
taken into account for will be determined a little later on, under the discussion of the
fully specified model presented in Table 5.

Psychological engagement
The psychological mood of the voting age public ahead of the 2019 General Elections
could be regarded as rather sombre in character, especially in perspective of the trends
observed over more than a decade (Figure 2). Levels of political interest have remained
fairly stable at a relatively low level, with those declaring that they are ‘very’ or ‘quite’ inter-
ested in politics varying between a modest 28% and 38% over the 2008–2018 period. In
late 2018, the figure stood at 30%, which is below the all-year average of 34%, and in a
similar range to that observed over the 2016–2018 period (28–30%). By contrast, the
sense of political efficacy expressed by the electorate has shown an appreciable decline
since the early 2000s, with 2018 representing a pre-electoral low-point. The political
efficacy index represents a combination of internal and external efficacy measures,
 POLITIKON 9

Figure 2. Trends in psychological engagement variables among the voting age public, 2003–2018/19
(%). Source: HSRC SASAS 2003–2018.

suggesting that the public is losing faith in the power of their vote and the accountability
of elected officials.
South Africans have tended to display a resolute sense of the duty to vote, a trait that is
a likely reflection of the hard-won struggle for the franchise, which was secured for the first
time in the country’s history in 1994. While this remains true for a clear majority of the
voting age population, there has nonetheless been a clear downward trend in this
belief, falling from a highpoint of 87% in 2004 to a low of 69% in 2018. Strength of parti-
sanship shows a less distinctive pattern over time, with the share declaring non-partisan-
ship fluctuating between 22% and 35% since 2008, with 2018 falling at the upper limit of
this range. Taken together with the other indicator trends, it suggests that entering into
the 2019 electoral context, not only was the electorate expressing a lower degree of par-
tisanship than in preceding years, but there was also an unprecedented low level of pol-
itical efficacy and sense of duty to vote, coupled with a low general level of political
interest. The question remains what bearing this has on electoral behavioural intentions,
and what it might mean for future electoral contests if the observed patterns continue
along the current trajectory.
The multivariate analysis presented in Table 3 tests the combined effect of the psycho-
logical engagement factors on the voting age public. The results suggest that political inter-
est, political efficacy and party closeness all yield a significant effect to some degree. In
contrast with decided voters, abstaining voters possess significantly lower political interest,
political efficacy and duty to vote. They are also less likely than decided voters to have
stronger party ties, and instead display a tendency towards non-partisanship. Like abstai-
ners, undecided voters also demonstrate weak partisanship, but fundamentally differ in
10 B. J. ROBERTS ET AL.

Table 3. Multinomial logistic regression results testing the impact of psychological engagement factors
on voter predispositions in South Africa, pooled 2008–2018/19 data.
Abstainer Undecided voter Undisclosed voter
Psychological engagement model Coef. Sig. Coef. Sig. Coef. Sig.
Political interest −0.161 *** −0.021 n.s. 0.112 *
Political efficacy −0.171 *** −0.024 n.s. 0.019 n.s.
Duty to vote −0.199 *** 0.028 n.s. 0.031 n.s.
Party closeness (ref. Non-partisan)
Very close −4.277 *** −4.633 *** −5.444 ***
Quite close −3.842 *** −3.881 *** −5.490 ***
Not close −3.608 *** −3.654 *** −5.129 ***
Year of survey 0.066 n.s. 0.008 n.s. 0.081 ***
Constant −131.439 n.s. −16.532 n.s. −162.929 ***
Notes: (1) Data is weighted to be nationally representative of the voting age population (18 years and older) living in South
Africa, (2) the base outcome is ‘Decided voters’ and (3) the regression model controlled for an individual’s province of
residence (not shown) and year of survey. ***p < .001. **p < .01. *p < .05. n.s. = not significant.

that political interest, political efficacy and duty to vote have no bearing in differentiating
them from decided voters. As for undisclosed voters, they tend to exhibit marginally higher
political interest than undecided voters but there is no distinguishing political efficacy or
duty to vote effect. Most notably, undisclosed voters demonstrate a weaker degree of par-
tisanship than abstainers and undecided voters.

Regime evaluations
Moving from psychological engagement to confidence in the democratic regime, we
examine three factors of performance and trust. Before moving to the predictive power
of such variables among voting-age citizens, we again examine the general trends for
the select indicators used in the multivariate analysis (Figure 3). Despite fluctuations
over time, satisfaction with the functioning of democracy in the country has displayed a
general downward tendency. Although there was a slight recovery in the transition
from the Zuma to Ramaphosa administrations between late 2017 and 2018, barely a
third (34%) of the electorate was satisfied with democracy ahead of the 2019 election.
Evaluations of specific areas of government performance have not demonstrated the
same degree of downturn across the 2003–2018 period. Nonetheless based on a compo-
site index of performance in six areas of government service delivery performance (water
and sanitation, electricity, low-cost housing, access to health care, crime reduction and job
creation), it is evident that satisfaction with service delivery has fallen sharply since 2016 to
an all-period low in 2018, slightly below that recorded in 2003. Trust in core political insti-
tutions (national and local government, as well as Parliament) has not been immune to the
critical evaluations provided for other political support measures. The mean political trust
index score fell from a high of 3.47 (on the 1–5 scale) in 2004 to a low of 2.60 in 2018, with
scores below the scalar midpoint (of 3.0) evident every year since 2012. There has again
been a modest increase in trust between 2017 and 2018, though the level of expressed
confidence remains worryingly low.
From this review of trends, it is evident that the 2019 electoral context was one in which
greater psychological disengagement among the voting-age public was accompanied by
decidedly harsher views on the supply of democracy, the basic performance of the demo-
cratic regime, and the level of confidence vested in political institutions. Yet, do these
 POLITIKON 11

Figure 3. Trends in regime evaluations variables among the voting age public, 2003–2018/19 (%).
Source: HSRC SASAS 2003–2018.

critical voices matter for planned electoral behaviour? We find that these regime evalu-
ations measures exert an influence on the electoral inclinations of the voting age
public, though arguably to a lesser extent than psychological engagement variables
(Table 4). Views on the functioning of democracy in South Africa, evaluations of service
delivery performance, and confidence in political institutions are significant factors that
distinguish decided voters from other voters. Those more satisfied with overall democratic
functioning and voicing more confidence in key political institutions display a greater like-
lihood of being a decided voter than an abstaining, undecided or undisclosed voter. By
contrast, decided voters are less likely to be content with government’s performance in
delivering select basic services on average than undecided and undisclosed voters,
though no significant effect is present in the case of abstaining voters.

Table 4. Multinomial logistic regression results testing the impact of psychological engagement factors
on voter predispositions in South Africa, pooled 2008–2018/19 data.
Abstainer Undecided voter Undisclosed voter
Regime evaluations model Coef. Sig. Coef. Sig. Coef. Sig.
Satisfaction with democracy −0.115 *** −0.122 *** −0.130 ***
Service delivery index 0.020 n.s. 0.133 ** 0.112 **
Political trust index −0.251 *** −0.158 *** −0.130 ***
Year of survey 0.036 *** −0.034 ** 0.033 ***
Constant −72.859 *** 66.415 ** −66.798 ***
Notes: (1) Data is weighted to be nationally representative of the voting age population (18 years and older) living in South
Africa, (2) the base outcome is ‘Decided voters’ and (3) the regression model controlled for an individual’s province of
residence. ***p < .001. **p < .01. *p < .05. n.s. = not significant.
12 B. J. ROBERTS ET AL.

It is worth mentioning that when the three regime evaluations measures are modelled
individually rather than jointly, the pattern of results remains unchanged in relation to sat-
isfaction with democratic functioning and political trust, but there are slight differences in
relation to service delivery evaluations. We find firstly that the coefficient on this measure
for abstainers is statistically significant (p < 0.001) and negative. This implies that abstai-
ners are less satisfied with government’s service delivery performance than decided
voters on average. Secondly, we find an absence of service delivery performance effect
for undecided and undisclosed voters. However, when the three measures are included
jointly in a single multinomial regression model, the service delivery effect for abstainers
cancels out, while this variable becomes significant (p < 0.01) for undecided and undi-
sclosed voters.

Full model
The results of the full model, which jointly includes all sets of measures corresponding to
the different theoretical hypotheses explaining electoral intentions, are presented in
Table 5. In this final analysis, we find that abstainers differ from decided voters in that

Table 5. Fully specified multinomial logistic regression results testing the impact of individual
resources, psychological engagement and regime evaluations factors on voter predispositions in
South Africa, pooled 2008–2018/19 data.
Abstainer Undecided voter Undisclosed voter
Coef. Sig. Coef. Sig. Coef. Sig.
Individual resources
Female (ref. male) −0.071 n.s. 0.072 n.s. −0.166 *
Age (in years) −0.020 *** −0.011 ** 0.005 n.s.
Population group (ref. Black African)
Coloured 0.334 ** 0.560 *** 0.298 *
Indian/Asian 0.576 *** 0.698 *** 0.121 n.s.
White −0.335 * −0.150 n.s. 0.033 n.s.
Type of geographic location (ref. Urban formal)
Urban informal −0.103 n.s. 0.274 n.s. 0.050 n.s.
Rural traditional authority areas −0.101 n.s. 0.007 n.s. −0.148 n.s.
Rural farms −0.201 n.s. −0.320 n.s. −0.415 **
Educational attainment (ref. Tertiary)
Junior primary schooling or less 0.020 n.s. −0.255 n.s. −0.796 ***
Senior primary schooling 0.028 n.s. −0.025 n.s. −0.475 **
Incomplete secondary schooling 0.074 n.s. −0.014 n.s. −0.459 **
Complete secondary schooling 0.013 n.s. −0.040 n.s. −0.284 *
Psychological engagement
Political interest −0.136 ** −0.004 n.s. 0.095 *
Political efficacy −0.167 *** −0.010 n.s. 0.020 n.s.
Duty to vote −0.167 *** 0.037 n.s. 0.032 n.s.
Party closeness (ref. Non-partisan)
Very close −4.473 *** −4.660 *** −5.445 ***
Quite close −3.927 *** −3.867 *** −5.508 ***
Not close −3.651 *** −3.625 *** −5.125 ***
Regime evaluations
Political trust index −0.090 * −0.033 n.s. 0.025 n.s.
Satisfaction with democracy −0.020 n.s. −0.077 n.s. −0.061 n.s.
Service delivery index 0.063 n.s. 0.087 n.s. 0.009 n.s.
Year of survey 0.055 *** 0.002 n.s. 0.089 ***
Constant −108.392 *** −4.008 n.s. −177.394 ***
Notes: (1) Data is weighted to be nationally representative of the voting age population (18 years and older) living in South
Africa, (2) the base outcome is ‘Decided voters’ and (3) the regression model controlled for an individual’s province of
residence. ***p < .001. **p < .01. *p < .05. n.s. = not significant.
 POLITIKON 13

they are younger, more likely to be a coloured or Indian adult than a black African adult.
Select psychological engagement factors continue to separate abstainers from decided
voters, with the former more inclined to exhibit lower levels of political interest, political
efficacy and a belief in the civic duty to vote than the latter, together with much lower
levels party closeness. The only regime evaluations effect that can be observed in the
full model is in relation to political trust, with abstainers generally displaying slightly
lower confidence in core political institutions than decided voters.
In the case of undecided voters, in common with abstainers they tend to be younger
and more likely to be a coloured or Indian adult than a black African adult than
decided voters. The predominant basis of variation is party closeness, with undecided
voters less likely to display a stronger degree of partisanship than decided voters. The
other psychological engagement factors remain statistically insignificant, and continue
to be the traits that differentiate abstainers from the electorally undecided. None of the
regime evaluations measures retain statistical significance once the individual resources
and psychological engagement measures are controlled for.
Lastly, in comparison to decided voters, undisclosed voters tend to be to be better edu-
cated, while also displaying a modest tendency towards being male, coloured, and based
in a formal urban area. As with undecided voters, the main psychological engagement
attribute that is statistically significant is party closeness. However, in this instance, undi-
sclosed voters demonstrate a lower strength of partisanship than undecided voters and
abstainers. Undisclosed voters also tend to be marginally more politically interested
than decided voters. The lack of observable effect on regime evaluations measures
repeats for undisclosed voters.

Concluding discussion
The analysis presented in this paper comes at a time of earnest reflection about Election
2019 and the state of South Africa’s electoral democracy, the changing patterns of regis-
tration, turnout and abstention, and what this might presage for future electoral contests
(Schulz-Herzenberg and Southall 2019). It has drawn on nationally representative, survey-
based data covering a fuller sweep of recent history in an attempt to provide deeper
insight into the changing persuasions of the South African electorate leading up to the
2019 General Elections, and test the extent to which three distinct theoretical constructs
inform planned electoral behaviour. The results suggest that these theoretical models
have a significant though differential degree of influence over electoral intentions.

Psychological engagement
Psychological engagement with politics clearly plays a dominant motivating role in deter-
mining whether South Africans turn out to vote, stay away from the ballot box, display
electoral indecision or refuse to disclose their preferences. The extent and strength of par-
tisanship is a common factor that separates decided voters from abstainers, undecided
voters and undisclosed voters alike. An increasing sense of party closeness therefore
serves to raise the likelihood of reporting an intention to vote, whereas weaker party
attachment or non-partisanship conversely sways individuals towards uncertainty or a dis-
inclination to vote. The survey series points to party non-identifiers accounting for
14 B. J. ROBERTS ET AL.

approximately a third of the voting age population by early 2019. As Schulz-Herzenberg

(2019a, 15) maintains, this represents a ‘large potential pool of (swing) voters’ that political
parties should try and persuade to vote, though convincing them to do so is inevitably
going to be an exceptionally challenging undertaking.
Apart from party closeness, other psychological orientations to politics variables, such
as political interest, sense of efficacy and belief in the civic duty to vote, have a strong
bearing in differentiating abstaining from decided voters specifically. Compared to
decided voters, abstainers are less interested in politics, less convinced that their vote
makes a difference or that the elected are responsive to their needs, and possess a
weaker sense of moral obligation to vote. This is perhaps one of the most disconcerting
findings from the study, considering the reduced turnout recorded in the 2019 General
Elections, as well as the declining trends in political efficacy and duty to vote observed
since the early 2000s. Compounding this picture is the effect of age, with abstainers gen-
erally presenting as younger on average than decided voters. It should also be mentioned
that psychological engagement is the main factor that distinguishes abstainers from unde-
cided voters. On face value, voter indecision emerges as a seemingly unimportant aspect
of electoral participation dynamics in the country, reported by a fairly stable 3%–7% of the
electorate in general over time. Nonetheless, as the multivariate results signal, if indecision
combines with a mounting sense of psychological disengagement, there remains a real
possibility that this could fuel the rising tendency towards electoral abstention in future
elections.

Regime evaluations
In contrast with psychological engagement in politics, the study findings indicate that
regime evaluations do not appear to have a particularly decisive role in motivating elec-
toral intentions in the country. Admittedly, abstainers, undecided voters and undisclosed
voters voice less confidence in core political institutions and more general discontent with
the functioning of democracy than decided voters. Yet, apart from a modest inverse effect
on satisfaction with democracy among abstainers, these influences fell away in the full
multinomial logistic regression model, especially following the addition of psychological
engagement factors. This is likely due to the fact that those with lower political interest,
efficacy, duty and partisanship are also likely to express more critical views on the perform-
ance of the democratic regime and institutions post-1994.

Socio-demographic factors
As for the socio-demographic traits that remain salient from the modelling, age, race
and education to a lesser degree seem to play a role. Age has already been mentioned
as a factor differentiating decided from abstaining voters. Similarly, undecided voters
are younger on average than decided voters. The fact that abstention and voter
indecision is more likely among younger age eligible adults is worrying given the demo-
graphic composition of the electorate. As of 2019, 18–29 year-olds represented approxi-
mately a third (33%) of all eligible voters, with a further quarter (25%) in their thirties
(Schulz-Herzenberg 2019b). Although a majority of youth still vote, an emerging
pattern of lower registration and turnout rates among youth, coupled with generational
 POLITIKON 15

replacement over time, is likely to have a bearing on the level of participation in future
elections.
It is interesting that race remains a significant factor underlying electoral intentions in
the country, even after controlling for all the other socio-demographic, attitudinal and eva-
luative variables in the modelling. The results suggest that black adults are more likely to
be decided voters, and less likely than coloured and Indian adults to be abstainers or unde-
cided voters. While controlling for psychological engagement and regime evaluation
factors has largely removed the significance of race in separating decided from undi-
sclosed voters, the same is not true for abstainers and undecided voters. The implies
that there remain explanatory factors that have not as yet been accounted for that differ-
entiate black adults from other South Africans in motivating electoral participation.
Educational attainment was only significant in explaining differences between decided
and undisclosed voters, with the latter tending to be better educated in general. This,
taken together with a greater propensity than decided voters towards being urban-
based, displaying slightly higher political interest and being appreciably lower levels of
partisanship suggests that undisclosed voters might possibly be opposition supporters
reluctant to openly state their affiliation with a particular opposition party, or even stra-
tegic voters that postpone their vote decision until well into the campaigning period in
the lead-up to Election Day.

Limitations and future areas of research

The article examined the relative influence of two of the three core components of the
civic voluntarism model of political participation (Verba and Nie 1972; Verba, Schlozman,
and Brady 1995) in the South African context, namely individual resources and psychologi-
cal engagement with politics. It has not tested the role of recruitment networks that bring
people to vote, including social networks and news media. Recently, Schulz-Herzenberg
(2019a) showed that this might be a particularly salient explanation of turnout in the
country, based on data from 2014. Moreover, recent quantitative election research has
also point to the existence of a culture of voting, with past electoral participation emerging
as a strong determinant of voting intention (Roberts, Struwig, and Grossberg 2012,
Struwig, Roberts, and Gordon 2016). Future South African electoral research should con-
tinue to monitor the salience of these and other emerging theoretical explanations for
electoral behaviour.

Implications for electoral democracy in South Africa

From a broader electoral democracy perspective, how distinctive are the conclusions
about the character of the South African voter (and non-voter) described above? In
many respects, the decline in electoral turnout, the changing civic norms away from elec-
toral participation among youth, and the rise of critical citizens who are discontent with
politics and democratic performance are broadly observed trends in many established
liberal democracies (Dalton 2008; Hooghe and Kern 2017). This has given rise to discus-
sions of a global democratic recession or retreat (Schenkkan and Repucci 2019). We
would contend that, despite the parallels, these democratic reversals matter more funda-
mentally in the context of young democracies such as South Africa. In such instances,
16 B. J. ROBERTS ET AL.

continued patterns of political disengagement and electoral abstention have the potential
to undermine democratic legitimacy and frustrate ongoing efforts at democratic consoli-
dation. In addition, the youthful population structure of the country is a key element that
has to be considered. Accordingly, any tendency towards abstention and declining
psychological engagement in politics among younger members of the electorate may
mean that the process of electoral non-participation and democratic retreat is likely to
occur at a faster pace than has been recorded in other more established democracies,
which are generally typified by older or ageing electorates.
Further research is needed to disentangle the relative contribution of generational and
life-cycle effects underlying changes in electoral attitudes and behaviour in the local
context. The South African electorate has clearly become more critical over time, disillu-
sioned with the quality of governance and party politics, and questioning the efficacy
of electoral participation. It may however be premature to state that democracy is in
crisis. Instead, the findings could be interpreted as an appeal for viable party alternatives
and greater representation, as well as a demand for greater accountability and integrity in
inter-election periods. The degree to which elected representatives heed this appeal will
be decisive in shaping the evolving character of South African political culture.

Disclosure statement
No potential conflict of interest was reported by the authors.

ORCID
Benjamin J. Roberts http://orcid.org/0000-0002-0607-5447
Jarè Struwig http://orcid.org/0000-0003-2410-654X
Steven L. Gordon http://orcid.org/0000-0002-6393-2118
Yul Derek Davids http://orcid.org/0000-0002-8226-8937

References
Almond, G. A., and S. Verba. 1963. The Civic Culture: Political Attitudes and Democracy in Five Nations.
New York: SAGE Publications.
Blais, A. 2000. To Vote or Not to Vote? The Merits and Limits of Rational Choice Theory. Pittsburgh, PA:
University of Pittsburgh Press.
Blais, A., and C. H. Achen. 2019. “Civic Duty and Voter Turnout.” Political Behavior 41 (2): 473–497.
Blais, A., E. Gidengil, and N. Nevitte. 2004. “Where Does Turnout Decline Come from?” European
Journal of Political Research 43 (2): 221–236.
Blais, A., L. Massicotte, and A. Dobrzynska. 2003. Why Is Turnout Higher in Some Countries Than in
Others? Montreal: Environment Canada.
Brady, H. E., S. Verba, and K. L. Schlozman. 1995. “Beyond SES: A Resource Model of Political
Participation.” American Political Science Review 89 (2): 271–294.
Campbell, A., P. E. Converse, W. E. Miller, and D. E. Stokes. 1960. The American Voter. Ann Arbor, MI:
University of Michigan Press.
Carpini, M. X. D., and S. Keeter. 1996. What Americans Know About Politics and Why It Matters. New
Haven, CT: Yale University Press.
Clarke, H. D., D. Sanders, M. C. Stewart, and P. F. Whiteley. 2009. Performance Politics and the British
Voter. Cambridge: Cambridge University Press.
 POLITIKON 17

Dalton, R. J. 2006. Citizen Politics: Public Opinion and Political Parties in Advanced Industrial
Democracies. 5th ed. New York: SAGE Publications.
Dalton, R. J. 2008. “Citizenship Norms and the Expansion of Political Participation.” Political Studies 56
(1): 76–98.
Gordon, S. L., J. Struwig, and B. J. Roberts. 2018. “The Hot, the Cold and the Lukewarm: Exploring the
Depth and Determinants of Public Closeness to the African National Congress.” Politikon 45 (2):
163–180.
Gouws, A., and C. Schulz-Herzenberg. 2016. “What’s Trust Got to Do with It? Measuring Levels of
Political Trust in South Africa 20 Years After Democratic Transition.” Politikon 43 (1): 7–29.
Hooghe, M., and A. Kern. 2017. “The Tipping Point between Stability and Decline: Trends in Voter
Turnout, 1950–1980–2012.” European Political Science 16: 535–552.
IDEA (International Institute for Democracy and Electoral Assistance). 2019. Voter Turnout Database.
Accessed August 13, 2019. https://www.idea.int/data-tools/data/voter-turnout/.
Inglehart, R., and C. Welzel. 2005. Modernization, Cultural Change, and Democracy: The Human
Development Sequence. Cambridge: Cambridge University Press.
Kersting, N. 2007. “Electoral Reform in Southern Africa: Voter Turnout, Electoral Rules and
Infrastructure.” Journal of African Elections 6 (1): 134–151.
Kimmie, Z., J. M. Greben, and S. Booysen. 2010. “The Effect of Changes in Registration and Turnout on
the Results of the 2009 South African Election.” Politeia 29 (1): 101–123.
Lipset, S. M. 1960. Political Man. The Social Bases of Politics. Garden City, NY: Doubleday.
Nevitte, N., A. Blais, E. Gidengil, and R. Nadeau. 2009. “Socioeconomic Status and Nonvoting: A Cross-
National Comparative Analysis.” In The Comparative Study of Electoral Systems, edited by H. D.
Klingemann, 85–108. Oxford: Oxford University Press.
Norris, P. 2000. A Virtuous Circle: Political Communications in Postindustrial Societies. Cambridge:
Cambridge University Press.
Parry, G., G. Moyser, and N. Day. 1992. Political Participation and Democracy in Britain. Cambridge:
Cambridge University Press.
Roberts, B., J. Struwig, and A. Grossberg. 2012. “A Vote of Confidence: Gender Differences in Attitudes
to Electoral Participation and Experience in South Africa.” Journal of African Elections 11 (2): 7–36.
Saunders, C. 2014. “Anti-politics in Action? Measurement Dilemmas in the Study of Unconventional
Political Participation.” Political Research Quarterly 67 (3): 574–588.
Schenkkan, N., and S. Repucci. 2019. “The Freedom House Survey for 2018: Democracy in Retreat.”
Journal of Democracy 30 (2): 100–114.
Schulz-Herzenberg, C. 2007. “A Silent Revolution: South African Voters, 1994–2006.” In State of the
Nation: South Africa 2007, edited by S. Buhlungu, J. Daniel, and R. Southall, 114–145. Cape
Town: HSRC Press.
Schulz-Herzenberg, C. 2019a. “To Vote or Not? Testing Micro-Level Theories of Voter Turnout in South
Africa’s 2014 General Elections.” Politikon 46 (2): 139–156.
Schulz-Herzenberg, C. 2019b. “Trends in Voter Participation: Registration, Turnout and the
Disengaging Electorate.” In Election 2019: Change and Stability in South Africa’s Democracy,
edited by C. Schulz-Herzenberg, and R. Southall, 44–65. Johannesburg: Jacana Media.
Schulz-Herzenberg, C. and R. Southall, eds. 2019. Election 2019: Change and Stability in South Africa’s
Democracy. Auckland Park: Jacana Media (Pty).
Struwig, J., B. J. Roberts, and S. L. Gordon. 2016. “Amandla Awethu: Public Attitudes, South African
Democracy and Electoral Participation.” In State of the Nation: South Africa 2016 Who Is in
Charge? Mandates, Accountability and Contestations in South Africa, edited by D. Plaatjies, C.
Hongoro, M. Chitiga-Mabugu, T. Meyiwa, and M. Nkondo, 140–159. Cape Town: HSRC Press.
Verba, S., and N. H. Nie. 1972. Participation in America: Political Democracy and Social Equality.
New York: Harper & Row.
Verba, S., K. L. Schlozman, and H. E. Brady. 1995. Voice and Equality: Civic Voluntarism in American
Politics. Cambridge, MA: Harvard University Press.
18 B. J. ROBERTS ET AL.

Addendum
Coding of indicators
Individual resource measures
Gender of respondent: (1) Male (2) Female. Age of respondent in years (at time of last birthday). Inter-
viewer recorded respondent’s race: (1) Black African (2) Coloured (3) Indian or Asian (4) White. Type of
geographic location: (1) Urban formal (2) Informal urban settlement (3) Rural traditional authority
areas (4) Rural formal (farms). What is the highest level of education that you have ever completed?
(1) Junior primary schooling or less (2) Senior primary schooling (3) Incomplete secondary schooling
(4) Complete secondary schooling (5) Post-secondary.

Psychological engagement measures

How interested would you say you are in politics? (0) Not at all interested (1) Hardly interested (2)
Quite interested (3) Very interested. Political efficacy is measured through an index constructed
based on responses to the following three statements - ‘Whether I vote or not makes no difference’;
‘After being elected all parties are the same, so voting is pointless’; ‘Voting is meaningless because no
politician can be trusted’. Responses were captured using a 5-point agreement scale: (1) Strongly
agree (2) Agree (3) Neither agree nor disagree (4) Disagree (5) Strongly disagree. The items combined
using an averaging approach to produce a reliable scale, with a Cronbach alpha of 0.8272, and the
original 5-point scale was reversed and transformed into a 0–10 score, with (0) representing low pol-
itical efficacy and (10) high political efficacy. ‘Do not know’ responses were coded as missing. Duty to
vote is measured based on responses to a single attitudinal item: ‘It is the duty of all citizens to vote’ –
5 point scale, reversed for analysis (higher values = higher sense of duty): (1) Strongly disagree (2)
Disagree (3) Neither agree nor disagree (4) Agree (5) Strongly agree.
Party closeness is a constructed measure based on the pattern of responses to two consecutive
survey questions: ‘To which political party do you feel most close?’ and ‘How close do you feel to this
party?’. The first question included a coded set of 15 political parties followed by an ‘other party’
option. Additional categories were included for ‘no party’, ‘refusal’ and ‘do not know’. If a respondent
identified with a specific political party, they were asked the follow-up strength of partisanship. The
variable was coded as: (0) Non-partisan (1) Not close (2) Quite close (3) Very close.

Regime evaluation measures

How satisfied or dissatisfied are you with the way democracy is working in South Africa? – reversed 5-
point scale: (1) Very dissatisfied (2) Dissatisfied (3) Neither satisfied nor dissatisfied (4) Satisfied (5)
Very satisfied. ‘Do not know’ responses were coded as missing.
The service delivery index is a measure combining responses to six questions evaluating govern-
ment performance in one’s local area of residence. How satisfied or dissatisfied are you with the way
that the government is handling the following matters in your neighbourhood? ‘Supply of water and
sanitation’, ‘Providing electricity’, ‘Affordable housing’, ‘Access to health care’, ‘Cutting crime’, ‘Creat-
ing jobs’ – 5 point scale: (1) Very dissatisfied (2) Dissatisfied (3) Neither satisfied nor dissatisfied (4)
Satisfied (5) Very satisfied. ‘Do not know’ responses were coded as missing. The items combined
to produce a reliable scale, with a Cronbach alpha of 0.7150, making use of a reversed satisfaction
scale ranging from 1 to 5, where (1) represents strong dissatisfaction and (5) strong satisfaction.
Political trust was measured using an index constructed from three measures capturing confi-
dence in core political institutions. Indicate the extent to which you trust or distrust the following
institutions in South Africa at present: National government, Parliament, Local government. (1)
Strongly trust (2) Trust (3) Neither trust nor distrust (4) Distrust (5) Strongly distrust. The items com-
bined to produce a reliable scale, with a Cronbach alpha of 0.7580, making use of a reversed trust
scale ranging from 1 to 5, where (1) represents strong distrust and (5) strong trust. ‘Do not know’
responses were coded as missing.