POLITICS

We underestimated how low turnout would go – IRR

Gareth van Onselen says ANC managed to squeeze opposition (EFF) support, DA failed to do same

IRR Post-Election Poll Analysis - media statement by Gareth van Onselen, IRR Head of Politics and Governance

12 May 2019 

Ahead of the 8 May general election, the IRR commissioned five polls on the electoral landscape, focusing primarily on the national ballot, Gauteng and the Western Cape. The results of the final poll, which was in the field from 1-4 May, were published on Monday 6 May. All our polling was conducted by Victory Research.

Polls by nature are not predictive, but rather describe the electoral market at a given point in time, and at a particular turnout level. There is much that political parties can do to move the electoral market between the time of polling and election day.

However, given that the last poll came out of the field with just four days to go until the election, it is worth assessing how well it aligned with the final election results.

The National Ballot

National Ballot

MOE

IRR Final Polling: 1-4 May 2019

ANC

DA

EFF

3%

Baseline [100% Turnout]:

50.9%

22.7%

14.3%

70.0% Turnout:

53%

24%

14%

74.6% Turnout* [2014 Parameters]:

54%

23%

13%

Remodelled for 65.0% Turnout:

57%

23%

11%

 

Actual Result [65.99% Turnout]:

57.50%

20.77%

10.79%

Gauteng Provincial Ballot

Gauteng Provincial Ballot

MOE

IRR Final Polling: 1-4 May 2019

ANC

DA

EFF

5.8%

Baseline [100%]:

39.0%

31.4%

15.4%

71.1% Turnout:

41%

33%

14%

75.7% Turnout* [2014 Parameters]:

42%

31%

14%

Remodelled for 65.0% Turnout:

53%

25%

16%

 

Actual Result [68.28% Turnout]:

50.19%

27.45%

14.69%

Western Cape Provincial Ballot

Western Cape Provincial Ballot

MOE

IRR Final Polling: 1-4 May 2019

DA

ANC

EFF

6.7%

Baseline [100%]:

55.7%

24.2%

5.4%

71.7% Turnout:

57%

21%

5%

72.9% Turnout* [2014 Parameters]:

54%

22%

7%

Remodelled for 65.0% Turnout:

54%

26%

7%

 

Actual Result [66.28% Turnout]:

55.45%

28.63%

4.04%

Turnout

The most important takeaway from the poll and election is the record low turnout.

Turnout stood at 73% in 2014. The poll modelled for scenarios around the 70% mark, as a 3-percentage point drop was in line with historical trends. In 2019, it dropped to 65.99%, a dramatic decline of 7 percentage points from 2014.

No one saw this coming. Thus, we generated no 65% turnout scenario. In future we aim to model a larger range of turnout scenarios, including the highly improbable: from 50% turnout to 100%. We have included in the assessment above a category where we have remodelled our final numbers for a 65% turnout. Doing so reveals numbers more in line with the actual result, given that a 65% turnout aligns with the turnout on election day. Overall, the different turnout levels provide a summary of what the election results would have looked like depending on how many South Africans had voted.

Analysis

Influence of the ANC squeeze campaign on the EFF

The final poll stated, with four days to go until the election, that “there is some fluidity in the market, and the last few days will likely see some small movement. If history is a guide, it will be towards the bigger parties.”

The poll noted that the EFF’s vote was “very fragile above 10%”, and suggested the ANC was mostly likely to capitalize on that. Consistent in the poll analysis was that low turnout helps the ANC, and hurts the EFF.

In the final analysis, the ANC did squeeze the EFF very hard in final few days, driving its vote up to 57% and the EFF down to 10%. When we remodelled our numbers for 65% turnout, that same trend played out (ANC 57%; EFF 11%).

Failure of the DA’s national squeeze campaign

The DA’s performance ended up well within our margin of error (20.7% vs 23%). However, its squeeze campaign failed dramatically. It polled as stable in the final week, with the prospect of some small growth (23%), but on the day it declined. This does not happen even in our remodelled numbers for a 65% turnout, and was, no doubt, a result of significant growth for the FF Plus. We put the FF Plus on 1.9% in our final poll, which was within the margin of error (3%) and it clearly had a higher potential, driving down the DA’s vote share.

Gauteng in the balance

The final poll stated: “Gauteng is set to be hung, barring a last-minute surge from the ANC.” The ANC did surge, as the trends we saw nationally were at their most intense in Gauteng; the ANC squeezed the EFF and the DA’s decline was at its most pronounced. The combination drove the ANC up just over 50%. Although, when remodelled for a 65% turnout scenario, our final numbers would have put the ANC on 53%, the DA on 25% and the EFF on 16%, our lowest turnout model was 71.1% in May. This meant there was a significant discrepancy between our final numbers (42%) and the ANC’s final percentage (50.19%).

DA majority in the Western Cape safe

With regard to our final numbers for the Western Cape, we stated: “The DA should hold the Western Cape relatively comfortably, despite its recent difficulties in the province.” The trend lines held in the Western Cape, and the DA ended up on 55%. If the final numbers are remodeled for a 65% turnout, the DA comes out with 54% and, more significantly, the ANC with 26%.

The future

There is, in the IRR’s estimation, a dearth of high-quality public polling in South Africa. It is a gap we wish to fill as an institution.

The IRR’s polls helped accurately identify some important trends (the ANC’s decline, the DA’s stagnation, the EFF’s growth, the knife-edge situation in Gauteng and the fight to death between the ANC and the EFF for an ever-vacillating 5% of alienated ANC voters). These were supplemented by a range of other indicators: voter attitudes to service delivery, a tax revolt, leader favourability, coalition preferences and the biggest issues.

Ahead of 2021, we aim to adapt and improve our approach and bring you even better and more in-depth polling in the run-up to the local government election – which is set to be as complex as it will be compelling.

*This turnout scenario was developed exclusively for the final day of the Tracking Poll. It was set by using the turnout for black, white, Coloured and Indian voters for the 2014 national elections, and applying those parameters to the final baseline figures.

Issued by Michael Morris, Head of Media, IRR, 12 May 2019