Ipsos pre-election forecasting on the mark nationally
|Issued by: Ipsos|
[Johannesburg, 23 August 2016]
Ipsos in South Africa again delivered a very close forecast on the final local election results. At its latest release of its poll findings (released Monday 1 August), Ipsos suggested that the ANC would garner 54% of the votes, the DA 26% and the EFF 8%.
(1) Summary of possible party support countrywide:
With regards to the metros, Ipsos forecast that the results would be too close to call: "The survey results in the City of Johannesburg and the City of Tshwane are indicating that the possible result in these cities are definitely too close to call. However, it seems as if the DA is doing very well and is ahead in the race to win the control in Nelson Mandela Bay."
Metro areas real outcome:
With the increase in voter support for the DA, the results of the Ipsos polling in partnership with the eNCA came under heavy scrutiny and criticism in the South African media and even amongst some of the prominent political leaders.
Mari Harris, Director of Public Affairs and a Political Analyst at Ipsos South Africa said she trusted their process.
"I think it's important that every step of the way you need to say with your hand on your heart, this is the best we can do," said Harris in an interview with eNCA.com.
By the weekend after the voting, major results were well within the margin of error, predicting no outright winners in three of the major metropolitan areas.
"As a research company with one of our specialisations being public affairs and social opinion, we could not be more pleased that our results have aligned so closely with the actual election results" states Country Manager of Ipsos South Africa, Nick Coates. "It speaks to the thoroughness and accuracy of our research and fieldwork operations."
Ipsos's forecasting was based on two studies conducted ahead of the elections:
Findings are from two series of studies: the series of pre-election polls conducted for the eNCA and the findings of Pulse of the People studies.
This project has been undertaken in the eight weeks before the Local Government Elections in 2016. A panel of about 3 000 voters were recruited using RDD (Random Digit Dialling), representative of eligible voters in Johannesburg, Tshwane and Nelson Mandela Bay. Each week at least 1 500 of them were phoned back for a five-minute interview. All interviews were conducted using CATI (Computer Assisted Telephone Interviews) with a representative sample of residents of the three metropolitan areas who have access to a mobile phone.
Interviews were conducted on the Monday and Tuesday of each week, data processing happened on the Wednesday and results were published on the Thursday. The purpose of these studies was to measure the "political climate" in the country during the pre-election period, measure the influence of campaigning and things that happened during this period (like the unrest and violence in Tshwane) and contribute to the political discourse in the country.
(1)Pulse of the People:
This study is undertaken annually by Ipsos every six months, however, in election years a third round is often added (this happened this year). A total of 3 861 interviews were conducted from 17 June to 18 July with a representative sample of South Africans. The margin of error for this sample size is between 0.7% - 1.6%.
Results are then filtered by those 18 years and older and the question on whether the respondent is registered to vote. This resulted in a total of 3 142 respondents, with the margin of error confirmed at 0.8% - 1.7% (for this sample as a whole).
Some of the questions in the Pulse of the People study were asked since 1976 and questions about the opinions on political leaders and parties were included since 1990.
This study is done by face-to-face interviews using CAPI (Computer Assisted Personal Interviews) in the homes and the home languages of respondents. A process of stratified random sampling is used to determine the choice of respondent and interviewers do not have any influence on this process.
Some of the other forecasting, along with the actual results after counting, is included in the table below.