THE CRITERION REPORT VOL 1; No 3

10 December 2018

The Criterion Report is home to the findings and insights drawn from market research conducted by the South African Institute of Race Relations (IRR) into voter preferences, attitudes and the nature of the South African political landscape.

December IRR Snap Poll

In line with the IRR’s objective to become an authority on political market research, the Institute has undertaken a “snap-poll”, focusing exclusively on party political support, to supplement the IRR’s first full survey on the electoral landscape, carried out in September 2018.

The IRR “snap-poll” was in the field during the last week of November and the first week of December 2018. It comprised three sets of questions: Voting intention, party favourability (ANC, DA, EFF) and, finally, a question on the strength of support for Patricia De Lille’s proposed new political party, when forced to choose between it and the four biggest political parties.

The IRR will undertake a full political survey every quarter. The next such full quarterly survey will be conducted in February 2019, ahead of the May 2019 election.

What follows is a two-part summary of the key political findings in the IRR December “snap poll”: Part A comprises an objective overview of the data and what it says. Part B comprises the IRR’s own analysis of those findings and some of what it believes are the key insights to be drawn from them.

The banner headline findings are:

- The ANC is on 56%, up four percentage points from September and, on a projected 69% turnout scenario, comes out with 59%.

- The DA is on 18%, down five percentage points from September and, on a projected 69% turnout scenario, comes out with 22%.

- The EFF is on 11%, down two percentage points from September and, on a projected 69% turnout scenario, comes out with 10%.

- That pattern is reflected in the party favourability ratings, which has 54% of respondents somewhat or very favourable towards the ANC (up from 53% in September); 28% somewhat or very favourable towards the DA (down from 34% in September) and 21% somewhat or very favourable towards the EFF (down from 28% in September).

- In Gauteng, no party holds a majority. The ANC comes in with 48%, the DA 25% and the EFF 12%. The ANC has increased by two percentage points from September, the DA declined by three percentage points and the EFF has declined by five percentage points from September. On a projected 73% turnout scenario, the ANC comes in at 50%, the DA at 27% and the EFF at 10%.

- The potential market for Patricia de Lille’s new party seems to be primarily among undecided voters, some of whom are likely alienated from the DA and the ANC. After that, she draws support directly from respondents who had previously selected the DA or the ANC.

Methodology:

The poll was conducted between 26 November 2018 and 4 December 2018. The sample was fully demographically representative and comprised only registered voters. A total of 1,017 respondents were questioned. The margin of error is 3.9%. For the Gauteng sample, the margin of error is 6.4%. The confidence level is 95%. The poll was conducted telephonically, using a single frame, random digit-dialling sampling design. Briefly: The sampling frame consists of every potential cell phone number in existence in South Africa, from which a probability sample is drawn. This approach ensures that every number stands an equal chance of being included in the study, which is the most basic condition that must be met for survey results to be generalizable to the population from which a sample is drawn. A fuller explanation of the methodology is available on request. Those who were “undecided” in response to the voting intention question (around 10%) were assigned a preference based on their response to party favourability questions. The poll was conducted by Victory Research.

Turnout:

As part of the poll, we generated a number of potential turnout scenarios. Any poll, in and of itself, represents a 100% turnout scenario, as every respondent’s opinion is recorded. But, on Election Day, inevitably not every voter goes to the polls. In 2014, there was a 73.5% turnout. In 2019, turnout is likely to be lower, a trend which is typical of maturing democracies and of electoral environments with high levels of discontent (which drives up voter apathy). We have thus generated four additional turnout scenarios for our numbers: 69%, 71%, 75% and 78%. This was done by assigning a probability of voting to each respondent based on their responses to several questions designed to measure their likelihood of voting, and excluding those least likely to vote.

Interpretation:

This poll is not a prediction. It is a snapshot in time, in this case of the electoral market between 26 November 2018 and 4 December 2018. Likewise, the numbers presented in the poll are not absolutely definitive. A 3.9% margin of error means, for example, the DA – which comes out with 18% – could be on 21.9% or 14.1%. A confidence level of 95% means we are confident 95% of the time the findings will never vary more than 3.9 percentage points up or down from reality. When reporting on the poll, it is important to bear these parameters in mind. Finally, the most valuable aspect of any poll is its ability to identify trends and patterns, particularly over time. One should thus avoid ascribing absolute authority to any given single, isolated finding.

Background:

The IRR’s first poll (September) put the ANC on 52%, the DA on 23% and the EFF on 13% (once undecided voters had been allocated). Since then, a number of significant events will have had some impact on the electoral marketplace. Among them: The South African economy has since slipped into a technical recession, Patricia de Lille parted ways in acrimonious fashion with the DA and established a new political party, the EFF has been caught up in a range of controversial funding allegations involving the now liquidated VBS Mutual Bank and, in the background, the Zondo Commission of Inquiry into “State Capture” has produced a steady stream of disturbing revelations. None of these, in and of themselves, are likely to have impacted definitively on the support for any one party – precedent suggests it takes much time for a single scandal or event to fully manifest in a change in voting behaviour – but each is likely to have had some impact.

PART A Findings:

1. Party Political Support:

Purpose of Question: Voting intention is designed to give an indication of where political support stands for each respective party, at a given moment in time. Turnout scenarios tell you what happens as the number of voters who actually vote on Election Day drops. As the number drops, you can see which party gains or loses support and, thus, how committed or enthusiastic each party’s support base is.

1. All Voters [Undecided Voters Allocated]: The poll put national support for the ANC at 56%, the DA at 18% and support for the EFF at 11%, once all undecided voters were allocated. [See Table 1.1.] From the IRR’s September poll, this represents an increase of four percentage points for the ANC, and a decrease of five and two percentage points for the DA and EFF respectively.

2. Turnout [Undecided Voters Allocated/All Voters]: Declining turnout appears to favour both the ANC and the DA, but has a negligible effect on the EFF. At a projected turnout scenario of 69%, the ANC is projected to get 59% (compared to 56% on 100% turnout), the DA 22% (compared to 18%) and EFF 10% (down a percent from 11%). [See Table 1.2.]

3. All Voters [Without Undecided Voters Allocated]: The poll put national support for the ANC at 51%, the DA at 17% and support for the EFF at 11%, without undecided voters, who account for 10% of the sample, being allocated [See Table 1.3.]. From the IRR’s September poll, this represents an increase of four percentage points for the ANC, a decrease of three percentage points for the DA and an increase of one percentage point for the EFF.

4. Key Finding: The poll suggests a shift in support towards the ANC since September. The opposite trend applies to the two largest opposition parties, with both the DA and the EFF down significantly from September. On a projected 69% turnout scenario, overall support for the ANC would increase from 56%, on 100% turnout, to 59%. The DA would also appear to benefit from low turnout. If an election were held tomorrow, on a projected 69% turnout scenario, the DA would return to 22%. However, that would not be an increase from its 2014 result. Lower turnout has a negligible effect on support for the EFF, which remains significantly higher than where it stood in 2014.

Table 1.1: Voting Intention [All Voters/Undecided Voters Allocated]

Table 1.2: Voting Intention [All Voters/Undecided Voters Allocated/By Turnout Scenario]

Table 1.3: Voting Intention [All Voters/Undecided Voters Not Allocated]

2. Party Political Support by Race:

Purpose of Question: By breaking down party political support by race it becomes possible to see how diverse the support for various political parties is and, over time, should there be a change in support levels, whether those shifts are particular to one demographic group or cut across demographic indicators.

1. Black Voters: Among black voters, national support for the ANC stands at 69%; that represents a six percentage points increase from September, when it stood at 63%. The DA’s support among black voters has fallen four percentage points, from 10% in September, to 6%, approximately where it stood in 2014. EFF support among black voters fell by two percentage points, from 16% in September to 14%. [See Table 2.1.]

2. Minority Voters: Support for the DA among minority voters fell ten percentage points, from 71% in September to 61%. ANC support among minority voters fell two percentage points, from 14% in September to 12%. Support for the EFF among minority voters is statistically insignificant. [See Table 2.1.]

3. Finding: The DA continues to dominate the electoral share of the market among minority voters, but it did lose significant ground in this market. Likewise, the party experienced a significant drop in support among black voters. ANC growth appears to be built on an increase in its dominant market share among black voters, as it lost some small ground among minority voters, off a low base. The EFF’s support base, almost universally comprising of black voters, shrank significantly, also off a low base.

Table 2.1: Voting Intention [Black Voters/Minority Voters]

3. Party Political Support by Province:

1. All Voters: The majority of ANC voters come from KwaZulu-Natal (21%), Gauteng (21%) and the Eastern Cape (19%). The majority of DA voters come from Gauteng (31%), the Western Cape (24%) and KwaZulu-Natal (12%). The majority of EFF voters come from Gauteng (32%), Limpopo (18%) and KwaZulu-Natal (12%). [See Table 3.1.]

Table 3.1: Voting Intention [All Voters/Province/ANC/DA/EFF Voters]

4. Party Favourability:

Purpose of Question: Favourability tells you about potential markets. If a voter is somewhat or very favourable towards a particular party, they are generally willing to listen to its message and are open to persuasion. Not everyone favourable towards a political party may vote for it, but are those voters who represent the pool of potential from which the party is mostly likely to draw increasing support. The bigger the pool, the better for the party in question.

1. ANC Voters: The ANC’s favourability improved fractionally from September, with 54% of voters somewhat or very favourable towards the party, compared to 53% previously. Its unfavourability dropped by four percentage points, with the number of voters somewhat or very unfavourable towards the party down from 33% in September, to 29%. [See Table 4.1.]

2. DA Voters: The DA’s favourability declined by six percentage points from September, with 28% of voters somewhat or very favourable towards the party, compared to 34% previously. Its unfavourability increased by two percentage points, with the number of voters somewhat or very unfavourable towards the party up from 42% in September, to 44%. [See Table 4.1.]

3. EFF Voters: The EFF’s favourability declined significantly, by seven percentage points from September, with 21% of voters somewhat or very favourable towards the party, compared to 28% previously. Its unfavourability increased fractionally, with the number of voters somewhat or very unfavourable towards the party up from 51% in September, to 52%. [See Table 4.1.]

4. Key Finding: Favourability towards the ANC remains relatively stable but would seem to mirror the party’s current support levels, suggesting a small pool from which to grow further. Both the DA and EFF have relatively bigger pools of potential than the ANC. However, those pools have shrunk since September. The shift would seem to be towards indecision, with “Don’t know enough to rate”, up for all parties.

Table 4.1: Party Favourability [All Voters/ANC/DA/EFF Voters]

5. Gauteng

1. All Voters: The ANC has improved marginally in Gauteng, up from 46% in September to 48%. Both the DA and the EFF have declined marginally. The DA declined three percentage points, from 28% to 25%, while the EFF declined by five percentage points, from 17% to 12%. [See Table 5.1.]

2. Turnout: Using a projected 73% turnout model, the ANC’s share of the vote in Gauteng increases to 50%, as does the DA’s (to 27%), while the EFF’s share of the vote declines to 10%. [See Table 5.1.]

3. Key Finding: As in September, no party holds a majority in Gauteng. The national trends have been replicated in the province, with the ANC improving its vote share and both the DA and the EFF declining. Even on a 73% turnout scenario (and bearing in mind the 6.4% margin of error for the Gauteng sample), while the ANC does reach 50%, as things stand, an outright majority for the ANC hangs in the balance.

Table 5.1: Voting Intention – Gauteng [All Voters/Undecided Voters Allocated]

6. De Lille’s New Party

Purpose of Question: This is what is called a “forced question”. It is not designed to accurately measure total electoral support, because it only provides respondents with five options, but rather to see how party support breaks, when the choice is limited in this way.

1. All Voters: When given only five choices – ANC, DA, EFF, IFP or De Lille’s new party – 4% of voters chose De Lille’s new party. It would appear, in comparison to each party’s actual electoral support [See Table 1.1.], support for De Lille’s new party comes primarily at the expense of the ANC, which drops into the 40% range, while the DA, EFF and IFP remain relatively unaffected. [See Table 6.1.]

2. Turnout: The percentage of voters who chose De Lille’s new party when presented with a closed list of five options remains constant at 4% through all five turnout models.

3. Breakdown of Support: By looking at which parties those respondents chose before being presented with the forced question, it is possible to see which political parties the 4% of respondents who chose De Lille’s new party had previously selected. [See Table 6.2.] Before undecided voters were allocated, 39% of potential De Lille voters were undecided in some fashion (they chose “other”, “undecided”, “won’t vote” or “refused”), 28% had chosen the DA and 12% the ANC. After undecided voters were allocated, 39% had chosen the DA, 22% were still unsure (“other” or “won’t vote”) and 15% had chosen the ANC.

4. Key Finding: Support for De Lille’s new party seems primarily to lower the ANC’s percentage of the vote, which drops from the mid-fifties in the voting intention question to the mid-forties, when respondents are forced to choose. The DA and EFF remain relatively unaffected. The majority of these voters were uncertain of which party to vote for prior to the question. After that, the majority of De Lille supporters had selected either the DA or the ANC.

Table 6.1: Support for Patricia de Lille’s New Party [All Voters]

Table 6.2: Breakdown of Support for Patricia de Lille’s New Party [Voting Intention/De Lille’s New Party]

PART B: Analysis

Summary: The ANC has improved its national percentage by four percentage points from September, to 56%. Both the DA (down from 23% to 18%) and the EFF (down from 13% to 11%) tracked down from September. This pattern is reflected in the party favourability ratings. The number of respondents somewhat or very favourable to the ANC increased fractionally (up to 54%), while the number of respondents somewhat or very favourable towards the DA declined (from 34% to 28%), as was the case for the EFF (down from 28% to 21%). On a band of turnout scenarios ranging from 69% to 78%, the ANC’s percentage of the vote generally increases into the upper fifties, occasionally touching 60%. The DA also benefits from lower turnout, generally getting into the lower twenties. The EFF seems generally unaffected by lower turnout, staying at around 11% and dropping to 10% at 69%.

The loss in DA support seems to cut across racial demographics. It is down from 10% among black voters in September to 6% and, among minority voters, down from 71% in September to 61%. ANC growth came on the back of an increase in black voter support (up from 63% to 69%). In turn, the EFF, which relies almost exclusively on black voters, lost two percentage points in this market, down from 16% to 14%.

The most likely explanation for these trends is not current affairs but a general move back towards the ANC by those voters previously alienated from the party, although issues like the advent of Patricia de Lille’s new party and the VBS Mutual Bank scandal would have acted to retard the ability of both the DA and EFF to counter this. It is important to understand that there are still a great many variables at play. The impact of load- shedding, for example, has yet to be tested. Likewise, political parties have yet to properly enter into “election mode”, and the ability of bigger parties like the ANC and DA, with more money, to “squeeze” smaller parties during the campaign has also yet to be seen.

It is also important to understand that those voters alienated from the ANC (around 10% from 2014, some of whom appear to be returning to the party) are fluid, and still available to both the DA and the EFF. All three parties are capable of winning them back, depending on circumstances, over the next five months. Thus, the trend of support towards the ANC and away from the DA and EFF is not necessarily a permanent one, and the ANC will have to work hard, under a great deal of pressure as it is on a wide variety of fronts, to retain them until Election Day.

In Gauteng, no party holds a majority, although the ANC has improved its position slightly, while both the DA and the EFF have declined. The EFF’s decline in Gauteng is significant. 32% of EFF support comes from that province. It is the backbone of its national support and it is likely its decline in Gauteng has been largely responsible for the decline in its national percentage. A coalition between the ANC and the EFF would likely constitute a stable, two-party majority in the province and, given the pressure and tension informal DA/EFF coalitions are currently under, this would seem a possibility both the ANC and EFF will now be well aware of.

The potential market for Patricia de Lille’s new party seems to be primarily among undecided voters, some of whom are likely alienated from the DA. After that, she draws support directly from respondents who had previously selected the DA or the ANC. Her new brand – the Good Party – was, however, not directly tested, as it had not yet been formalized before the poll went into the field. For a true and accurate reflection of her national support, we will have to wait until February 2019, when the IRR’s next full survey goes into the field and the party has had a chance to properly establish itself in the public mind.

Gareth van Onselen

Head of Politics and Governance

SA Institute of Race Relations

Issued by the IRR, 10 December 2018