Share

Why we should ignore polls that predict election results

There are 'opinion polls' and then there are polls that serve as propaganda tools, which just happen to be called 'opinion polls'. Both can be dodgy, write Ivor Sarakinsky and Ebrahim Fakir.

There are sound reasons why democratic societies hold elections. They enable the establishment of the will of the people and consequently constitute representative government. 

Regular free and fair elections bring democratic government into office. In a democracy, elections are conducted on an inclusive and independent basis, taking place in a spirit of mutual respect, tolerance, personal safety, vigorous debate and vibrant campaigning. 

A secret ballot is the cornerstone of the integrity of the electoral process and an important factor in determining how fairly and equally the electoral process treats its citizens, the principals, rather than the candidates and parties, the mere agents of citizens.  

READ: The important message the upcoming election results will send South Africans

Elections are process driven where there is equality of opportunity in the exercise of political choice.  The deliberative dimension of elections also enables public reasoning on the challenges facing society, where citizens exercise voice in identifying the issues prior to exercising their choice. In this way, citizens have the opportunity to set the agenda and influence the outcome of the elections.

Elections systematise and manage the contestation and conflict over the issues and the distribution of resources that may be at stake in society. Depending on levels of voter participation (turnout), an election creates the density of social capital and creates the possibility for the construction of a civic culture of social reciprocity, solidarity and co-operation in society. 

Elections are critical for promoting both the legitimacy and credibility of a government, which in turn catalyses citizen consent to be governed. An election is a mechanism to ensure that there is the potential for alternation in power and for its periodic reconstitution, and ensures that government is accountable to the public while simultaneously also delineating the government's powers, authority and responsibility in making policy and enacting legislation, for which they are to be answerable. 

All of this confers legitimacy on government formation and credibility to the accountability, oversight and transparency demand roles that opposition parties and other role players such as citizens, citizen groups and the media, play. Hypothetically, policy formulation is triangulated from and lent the requisite authority and credibility emergent from the aggregation of the people's will, the legitimate formation of government and its subsequent credible operation and administration.  

In aggregating the people's will through an election, there are good reasons for a secret ballot. It protects the citizens' right to free political choice individually and in community. In a democracy the result is only known once the final tally is computed. This then begs the question, as to why actual elections are so strongly prefigured or pre-empted by opinion polls and surveys, and why opinion polls are taken so seriously by the media, political parties and commentators? 

Opinion polls are for the trolls 

Make no mistake, we are not opinion poll naysayers. In marketing and advertising, opinion polls can be useful in testing a market's appetite and propensity for new products, services and brands. Opinion polls on public institutions and leaders might be useful indicators of citizen trust and confidence in public institutions but this kind of "research" does not translate well, if at all, into making predictions in the political realm. In reality they are akin to alchemy with pollsters acting as the alchemists of old, striving to do the impossible. 

Our primary concern here is not with the methodological flaws and problems concerning election related opinion polls and their predictive power. The methodological problems are well known and range from sample size, through to sample stratification and representativity, sample randomisation, confidence levels, the nature of the actual questions posed and the fact that respondents can be less than honest when polled on sensitive questions of "political choice". 

Yet, pollsters believe, like mystical oracles, in the validity and accuracy of their epiphanies. But if this faith is to be vindicated then the various different polls, applying similar methods, should ordinarily produce similar results. However, when the different opinion polls in South Africa at present are triangulated, the results are divergent, casting serious doubt on their predictive, or other value. The manner in which they have been so seriously taken, and the vehemence with which their results are defended by the respective pollsters, should obviate the need for public elections. 

Elections are onerous, costly, time consuming, inordinately expensive and resource intensive and above all, incept risk in already fraught socio-political contexts, exacerbating underlying tensions that competitive political contests introduce, that could easily spill over into overt conflict. Perhaps, if the pollsters are to be taken at their word and believed, there is no need for societies to actually hold elections, and instead, societies can simply poll about 5 000 people every now and then to ascertain how and by whom society should be governed, since pollsters appear to believe that this would be sufficient to harness the expression of the people's political will. 

This is patently absurd and undermines all of the very sound reasons why societies actually hold elections. 

(Ab)uses of opinion polls 

We know that there are 'opinion polls' and then there are polls that serve as propaganda tools and the marketing, advertising and public relations as part of party campaigns, which just happen to be called 'opinion polls'. Both can be dodgy. 

Political parties often conduct internal polls to identify community issues, establish where their support bases are and then strategically decide on the deployment of scarce campaign resources. Such instruments are pragmatic and generally more useful than opinion polls executed by marketing companies or special interest, politically aligned, lobby groups.

Internal polls conducted by political parties need to be differentiated from those conducted by media and marketing agencies. Special interest lobby groups often publish polls in order to influence and sway the electoral result in favour of the party they may be aligned to, or to bolster or undermine one or other faction of that party. Such action may even include a cognitive bias to undercount their political opponents' support to demoralise their voters to enable voter conversion. 

At the same time, political parties are usually reluctant to publish their internal polling for fear of exposing their perceived strengths and weaknesses to their competitors and opponents. The palpable excitement generated through a glimpse at a political party's internal poll is often unquestioned, but is obviously the result of a strategic leak, to send a specific message as part of the parties' electoral arsenal and campaign in order to alert, encourage and incentivise its potential voters who might otherwise stay away due to lethargy, apathy or anger to turn out to vote.  

ALSO READ: Best estimates - Dawie Scholtz on where the votes could go, and why

A different approach to interpreting the political context of elections and the resulting trends, is one that considers previous election results and voter turnout and behaviour. These actual numerical trends can be triangulated with longitudinal national and provincial election results data, including municipal elections, and recent municipal by-elections. These need to obviously account and adjust for both the variable and distinct electoral and voting systems used at national and provincial elections from local ones, as well as account for differential voter turnout and participation rates. 

Nonetheless, these trends are based on large sample sizes (the whole electorate), and also reflect real voting choices and decisions, as opposed to hypothetical and artificial choices influenced by multiple different variables, in opinion polls. 

For these reasons alone, this technique is bound to be more reliable than artificial and ersatz polls. The limit of this technique would be assessing and estimating the extent to which current events in the close run-up to an election impact on voter choices and voter behaviour.  

Opinion polls are therefore for the political trolls. There is no real value to, or need, for them. The real and actual poll result will be known once the ballots are counted and tabulated anyway. This is the only poll that counts. Should an opinion poll come close to predicting a result, this will be due to fortuitous coincidence, rather than methodological rigour or predictive accuracy.  

- Ivor Sarakinsky is associate professor at the Wits School of Governance. Ebrahim Fakir is director of programmes at the Auwal Socio-Economic Research Institute (ASRI) and an associate of the Democracy Works Foundation.    

Disclaimer: News24 encourages freedom of speech and the expression of diverse views. The views of columnists published on News24 are therefore their own and do not necessarily represent the views of News24.

We live in a world where facts and fiction get blurred
Who we choose to trust can have a profound impact on our lives. Join thousands of devoted South Africans who look to News24 to bring them news they can trust every day. As we celebrate 25 years, become a News24 subscriber as we strive to keep you informed, inspired and empowered.
Join News24 today
heading
description
username
Show Comments ()
Voting Booth
Do you think corruption-accused National Assembly Speaker Nosiviwe Mapisa-Nqakula will survive a motion of no confidence against her?
Please select an option Oops! Something went wrong, please try again later.
Results
No, her days are numbered
41% - 496 votes
Yes, the ANC caucus will protect her
59% - 702 votes
Vote
Rand - Dollar
18.91
+0.1%
Rand - Pound
23.87
+0.1%
Rand - Euro
20.37
+0.3%
Rand - Aus dollar
12.31
+0.2%
Rand - Yen
0.12
+0.2%
Platinum
908.05
0.0%
Palladium
1,014.94
0.0%
Gold
2,232.75
-0.0%
Silver
24.95
-0.1%
Brent Crude
87.00
+1.8%
Top 40
68,346
0.0%
All Share
74,536
0.0%
Resource 10
57,251
0.0%
Industrial 25
103,936
0.0%
Financial 15
16,502
0.0%
All JSE data delayed by at least 15 minutes Iress logo
Editorial feedback and complaints

Contact the public editor with feedback for our journalists, complaints, queries or suggestions about articles on News24.

LEARN MORE