In the build-up to the 2019 national and provincial elections there has been much focus on polling results and their predictive value. With the two biggest pollsters reporting such divergent numbers it’s easy to dismiss the whole industry as a waste of time.
Regardless of where you stand on polling, most people would agree on a hierarchy of usefulness when it comes to the numbers. The best predictors of future election results are previous elections, and general elections are more useful than by-elections, which in turn are more useful than opinion polls. The disagreements are in the details; just how much less predictive are by-elections when compared with general elections?
The greatest strength of general elections is that they are near-universal; they cover every corner of SA at the same time. Opinion polls sample a much smaller segment of voters, typically between 1,500 and 3,000 people at a time. By-elections, over time, can cover the length and breadth of the country but not its depth: over a five-year period less than 5% of SA's wards will hold by-elections.
Apart from the lack of universality, by-elections suffer from self-selection. A well-designed opinion poll may still canvass potential voters at random, but by-elections are far from random. By-elections are disproportionately held in competitive wards and municipalities, and their outcomes are far more likely to change the composition of municipal councils.
This reflects the power held by politicians and parties: defections, expulsions and resignations by councillors are the leading reasons for holding by-elections. Even by-elections held due to the deaths of councillors are affected by political assassinations in a nonrandom way.
The self-selection extends beyond the by-election itself to the parties contesting it. The DA, for example, has trumpeted its success in by-elections but what is not mentioned is that the party has only contested about 70% of by-elections since 2016. Between November 2016 and June 2018, the party skipped 41 of the 139 by-elections that were held. Any party or candidate has the prerogative to participate in or to ignore a by-election, but you can’t gauge the overall health of that party if they only pick the easier battles.
Self-selection results in other confounding factors. If only two or three parties contest a by-election in a ward where 10 or more parties scrapped for votes in the general election, how should we interpret the results? How should we apportion the votes that party A receives in a narrower field — with far lower voter turnout to boot?
These three challenges — nonuniversality, self-selection and lower turnout — make the job of evaluating new trends difficult but not impossible. Provided that a person is prepared to think a little bit differently about by-election results, there is a lot of useful information to be gleaned from them. It’s just that more of this information is descriptive, rather than strongly predictive.
The EFF’s overall performance in by-elections, for example, has been positive since 2016. The party has contested almost every by-election held over the last three years and has increased its share of the vote in over half of these by-elections. It’s also seen its share reduced in about a third of all by-elections.
In terms of raw voting numbers, the party has increased its votes tally in about half of all by-elections. This is more significant than its share of the vote, as turnout levels in by-elections can fall far below the turnouts in general elections. For the party to maintain its votes as turnout levels fall by between 20% and 40% points suggests a very high degree of loyalty from its 2016 supporters, or strong growth of new voters, or both.
So, while the by-election results don’t allow for a finely calibrated prediction in the 2019 elections they do allow us to claim authoritatively that the EFF has been in ruder health over the past three years than the ANC or the DA, and that its growth and appeal is more broad-based now than it was in 2014 or 2016.
Another trend of interest is the fluidity of certain voting blocs in by-elections, suggesting that lower-stakes elections lead people to vote less strategically and more ideologically. The FF Plus polled better in certain wards during by-elections than it did in 2016, suggesting that there is a shift among some voters from the DA to the FF Plus. This is in line with similar international results: the Ukip does better in elections for the European Parliament than it does in the UK’s general elections.
Again, this isn’t conclusive evidence of a trend away from larger parties to smaller ones, and the DA is expected to win back most of these voters in 2019.
Finally, by-elections can add to the narrative around our political and electoral structures and motivate for changes and reforms beyond our choice of political representation. When the overwhelming majority of by-elections are triggered by party machinations and the need for political control, rather than by the need for improved service delivery and the needs of voters, maybe it’s time for us to consider introducing greater checks and balances into law, such as recall mechanisms for underperforming councillors.
By-election results may not yield the clear evidence we seek on the growth or decline of individual parties, but they still contain valuable clues about voter choices and intentions, and they can help us paint a more nuanced picture about the strengths and weaknesses of our form of representative democracy.
• Berkowitz is the director of Edgis, a researcher and consultant
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.