2019 Election Results

Tracking the Count – Updated

By Peter Attard Montalto 9 May 2019
Caption
The results board at the IEC National Results Operation Center (ROC) on Thursday 9 May 2019. Photo: Leila Dougan

Daily Maverick and Intellidex analyst Peter Attard Montalto are joining up to track results as they unfold - barring glitches like printer melt-downs and such - in this, South Africa’s most contested elections 25 years into democracy. The pollsters have had their say, and on 8 May it’s up to the 26, 736,803 registered voters to cast their ballots before 9pm when voting stations close. And then the count’s on. We will be looking at forecasts that can be made, if any, from early partial results to the final result, looking through the party swings in certain areas and important difference between votes and actual seats, especially in tight provincial contests.

Noon

The model is still stable around 50.45% for the ANC in Gauteng despite the latest haul of results. The National model level for the ANC is also stable just above 57%. Of most importance is that seat count projections are not changing under different assumptions.

Around 76% of the projected national turnout has now been counted which means our model at a national level moves very subtly and is now broadly stable at just above 57% for the ANC nationally and the EFF settling just below 10.5%. Seat allocations will change but only by a seat here or there as smaller parties drop in and out of the projected new parliamentary makeup. The ANC losing around 19 seats is not great but hardly a disaster for them assuming some caucus coherence. If, of course, the caucus becomes more raucous as the fightback inside the party intensifies after these elections – so a 30 seat majority down from a 49 seat majority is a risk.

Focus is now very much on Gauteng with a small avalanche of votes in the province coming in against the ANC and dragging them down to 49.77%. Most of these, however, came from traditional non-ANC voting areas and as such the actual swing in these voting districts taken together was minimal and so the overall model for Gauteng has not really shifted for the ANC (it has marginally for smaller parties but not enough to be a major narrative). As such, we are still sitting at 50.45%.

With only 57.7% of the votes counted in Gauteng, however, there is still room for an upset – indeed our model looks at the standard deviation of VD level results in order to gauge uncertainty (and calculate the margin of error around the prediction) – and by this measure, uncertainty has been slowly rising overnight. This means that whilst on aggregate surprises have balanced each other out, actually at a VD level there is a lot of volatility vs 2014 ongoing. With a number of key marginal votes ongoing (DA to ANC, ANC to EFF, DA to VF+ etc) this is to be expected maybe.

With our error margin still at +/-0.96% for Gauteng, it still cannot be definitively called, though as we mentioned in our 1am update we are more confident on seats.

Below we look at seat scenarios for the ANC on its raw result level in Gauteng at 49.77% and the model’s 50.45%. It gets the key 37 in both cases. As we’ve mentioned many times before, the PR system can throw up profound oddities depending on where smaller parties vote shares are, however it looks like, still, we can say with most confidence that the ANC will have 37 seats vs 37 needed and that the vote share will probably be around 50.45% but could be a little lower.

An interesting question is to think about what the minimum level of ANC vote share would be to still retain 37 seats. Doing this is almost impossible given so many moving parts in a PR system with so many smaller parties. However, currently assuming that a drop in the ANC’s vote share was assigned proportionately to other parties, we think the vote share could fall to as low as 49.16% with the ANC still keeping 37 seats. A level to watch then, albeit with a pinch of salt.

Given the results have slowed to a trickle now we will keep updating the model in the background and post next if there is any meaningful change.

Live national and Gauteng results detail

Source: Daily Maverick, IEC

National ANC predicted share and error bands

Source: Daily Maverick, IEC

Gauteng ANC predicted share and error bands

Source: Daily Maverick, IEC

Projected change in absolute number of votes received and turnout statistics

Source: Daily Maverick, IEC

Gauteng Provincial Legislature Seat scenarios

Source: Daily Maverick, IEC

1am, Friday, 10 May

As large metros start reporting the model is on the move, but only subtly. The numbers are shifting very slightly, but the narrative is not. We still see the ANC on around 57.1% nationally (230 seats) and 50.5% in Gauteng (37 seats — the minimum number).

The introduction of the IEC’s new audit process and the usual lull in results before the metros start pouring in has all meant that we are sitting at 64.1% of the votes projected counted nationally and 50.36% counted in Gauteng. This threshold is a good opportunity to check in with Gauteng, however, especially.

Greg Mills and others here on Daily Maverick commented earlier that there seems now to be more of a consensus that the ANC will still just keep a vote (and seat) majority in Gauteng. The DA’s James Selfe said their internal models put the ANC at 50.2%. Nevertheless, we have been surprised by the unusually reserved nature of ANC types in both public and private on the Gauteng question.

Our latest model saw the ANC tick back very marginally, but overall it has been broadly stable for the ANC. By contrast, the DA has improved marginally at the expense of the EFF and smaller parties in Gauteng. In terms of projected provincial legislature seats, things remain remarkably stable.

We have been calling a seat majority all along, even with the ANC below 50% given the PR allocation of seats we saw. We have called for an outright majority, though of varying size, from 5.7% of the vote being counted in the province. We think we can now say with some certainty that the ANC will get a majority of the vote, though given that they are still on the nose with only 37 seats vs 37 required for a majority — so anything could still technically happen with seats, depending on the makeup of small parties’ support.

It looks likely, however, that the ANC will get an outright majority of seats if the underlying swing seen in seats highly correlated with Soweto and similar Gauteng areas comes through. It is still not possible to “call” Gauteng really — given the margin of error is still +/-1%. However in the morning, with more results, it should then be possible, barring any major slowdown in declaration speed.

Nationally we have seen a larger tickback in ANC vote share projected from 57.49% to 57.09% as a large number of second cities started reporting. The DA took up the slack here, increasing to 21.31%. Overall, we think we can now call (given a margin of error of +/-0.6%) that the ANC will get around 57%.

Nationally, turning attention to projected seats, we can now see the IFP and FF+ swopping between 12/11 seats and this will continue to shift in the final straight. The UDM has rallied back somewhat to three seats now, given their urban-centric base, but they are still underperforming 2014 significantly. The ATM and AIC still make it into Parliament, just, but could easily fall off again.

We have particularly concentrated on the Gauteng provincial race here given its closeness. The EFF leapfrogging into official opposition in Limpopo and so on are more fixed narratives that had a higher degree of certainty much earlier on. Given the uncertainty over the Gauteng results, however, we will publish more updates in the morning until we can “call” it.

For now, the national results would seem to suggest that notional “mandate” levels for President Ramaphosa are not met (unless those goal posts are shifted yet again and as such markets and investors will remain sceptical of reforms to come).

6pm

The model is still broadly stable around 57.5% nationally, now a tick higher in Gauteng but still holding in around 50.5%.

After having climbed very slightly as turnout dropped the ANC vote share in our model is now stable again around 57.5% with a +/-0.9pp error margin. A more mixed density of different demographic voting districts reporting around this time with 46.5% of the projected national vote counted means that the underlying swings can be more stable.

The national turnout is now looking to eventually settle around 65% which is a historically low level and would imply that the ANC lost around 1.6million voters (net) vs 2014. The DA will have lost around 620k and the EFF gaining ‘only’ 645k with smaller parties gaining 296k. We had originally thought the EFF would gain around 900k voters on higher turnout. The shock here to the ANC loss of support in raw number terms is seismic and will have deep implications in the part and its branch structure. It also means that as few as 27.5% of eligible voters may actually have ended up voting for the ANC with less than half the eligible population turnout out to vote. We look at both these issues in the new chart below.

In Gauteng, as the pace of larger voting districts audited results starts to pick up so we’ve seen actually a very marginal increase in the projected endpoint for the ANC now at 50.56%. The seat allocation has been changing, however,  the ANC is still sitting on the magic 37 majority point.

The DA has dropped back both nationally and in Gauteng in our projection but actually, EFF has seen a split, strengthening marginally in Gauteng while dropping back nationally. As we commented on in the last update we don’t think this is a bad result at all for them and their outcome may well have been over-hyped beforehand by both the media and the party itself.

Forecast seat allocations are problematic given the key factor in this kind of PR system is the shape of the support curve across intermediately supported parties. As such around this 40-60% of the count mark, things can become very volatile. Hence, in between, these posts updating our model has seen NFP and PAC fall in and out of national assembly seats. VF+, however, seems more stable now around 12 seats (six extra), though much will depend on how they turn out their vote in large urban areas reporting last vs earlier semi-rural areas that have already reported.

In Gauteng, it seems unlikely that any smaller party will be able to catch ACDP or IFP as the smallest parties to get seats and so the interplay here is them keeping their second round seat allocations of one each.

Error bands are now starting to close around the forecasts fast, but last minute surprises are still possible as major urban centres start reporting in larger size. As such, whilst notionally our model says there is a 79% chance the ANC get an outright majority in Gauteng, we do take this with a pinch of salt.

Live national and Gauteng results detail

Source: Daily Maverick, IEC

National ANC predicted share and error bands

Source: Daily Maverick, IEC 

Gauteng ANC predicted share and error bands

Source: Daily Maverick, IEC

Changes in number of votes and turnout statistics

Source: Daily Maverick, IEC

1pm

ANC is showing a small rally back in our adjusted numbers to 57.2%, but still flatlining just on top of 50% in Gauteng. We don’t see the EFF as weak as their raw numbers.

With the 20% mark of expected votes reached in Gauteng and nearly 30% overall here is another update of our model.

The ANC continues to grind very marginally higher at national level but is flatlining in Gauteng just on the north side of 50%.

Overall the ANC looks set still to get around 57.2% of the vote at the end nationally and lose around 20 seats (shifting smaller party results is causing a deviation of around +/- seats here on updates). The EFF now is projected to gain one or two extra seats than ANC loses which is interesting.

VF+ is still the other big gainer at a projected six seats increase though as more results for smaller parties comes out where they have pockets of support coming through. This is the complication with a PR system like this, marginal changes in smaller parties can make a big difference. What we can say is that whilst some like PAC and NFP and UDM are losing ground it could come down to the wire if they stay or not. ATM still looks set to get at least one seat. GOOD has now accelerated ahead somewhat to gain around 2-3 projected seats though overall seems to be struggling vs initial expectations.

In Gauteng the model is still stable, sitting just above 50% of the vote share but with such low returns still from key areas of the province, confidence bands are still wide. Even as the smaller parties vote shares get added, and ADCP may gain a seat (though on the edge), the ANC is sticking there with right on the number of 37 required for a majority. This clearly could be easily lost and so we still find Gauteng too close to call – albeit the current indication is that the ANC can just make it come.

Two key questions:

Is EFF weak?

No, we see them at 8.9% in raw results and 10.18% in our adjusted model given the underlying swings and type of voting districts declaring. This is a little on the lower side of our pre-election expectations but not outrageously so.

Why is turnout falling in the model?

The model uses a global turnout assumption which was 72%, then 71% yesterday as we got feedback. Further feedback this morning saw us adjust this to 70%. However, as actual turnout is recorded on completed wards (not voting district) so the model shifts away from the global turnout assumption towards actual. This is now happening.

Live national and Gauteng results detail

Source: Daily Maverick, IEC

National ANC predicted share and error bands

Source: Daily Maverick, IEC

Gauteng ANC predicted share and error bands

Source: Daily Maverick, IEC

9.00am

A small amount of ANC momentum since the last update, EFF doing quite a bit better than the raw results, overall looking like ANC will still just hold onto a working majority in Gauteng and Nationally at 56.5%.

The IEC has progressed further in the last two hours to now count some 15.7% of projected total votes nationally and 8.3% in Gauteng of projected votes there.

Our model has been remarkably stable as results come in showing that, accounting for differences in voting district type and turnout etc, the underlying shifts in support are quite consistent. Error margins are still wide however given the still small number of votes counted – they are now around +/- 2.58pp.

Turnout estimates have dropped further now to 68.8% on the latest results. Feedback that we have received during voting, especially in townships, seemed to indicate that there was not a particularly strong turnout there vs 2014 and so these headline turnout figures are unlikely to rise much, we think.

Nationally the ANC has ticked up very marginally from the last update to 56.48% whilst in Gauteng has also leapfrogged the 50% level to 50.36%. Other party’s support levels are broadly stable, however. There are still wide spreads vs the raw results, especially for the ANC being over-represented in the raw results and the EFF underrepresented.

The seat allocations with the current adjusted results have not moved significantly with the ANC still said to gain a majority of seats in Gauteng even on a close vote. The smaller party parliamentary results are fascinating, in particular, the major gains of EFF and VF+, but also the fact that UDM, NFP, PAC and others are being wiped out. ATM is currently showing in our model as getting 2 seats though this will be diluted away towards the end and they may be lucky to get one.  A large cliff edge in the results after 6th position in Gauteng means that the ANC is unlikely to find any bedfellows for coalitions in the province – if required – apart from EFF or IFP.

Live national and Gauteng results detail

Source: Daily Maverick, IEC

National ANC predicted share and error bands

Source: Daily Maverick, IEC

Gauteng ANC predicted share and error bands

Source: Daily Maverick, IEC

First live results model

5.30am

We currently see the end result for the ANC nationally at 56% on 69% turnout, however with very low confidence given the low number of votes counted. We see the ANC in Gauteng at 49.6% but it could still get a majority of seats just.

Calling an election from live, partial, results is a serious challenge in South Africa at the best of times. But given how contested these 2019 elections we wanted to give some flavour of how the results are shaping up and so cutting through the hype that can take over social media and parts of the media when individual (tiny voting station) results are seen.

What are we trying to achieve?

Our model tries to look at how results that come out differ from 2014 elections and from signals sent by the 2016 elections and subsequent recent by-elections. The model treats partially released results from given wards sceptically (i.e assigns them a lower weight than fully complete wards). The model works off projected turnout at a per vote district level and a range of other assumptions which in turn shift through the process of releasing the results. At each stage of the results, it will give a ‘best guess’ and a confidence limit around that estimate. That confidence limit is important and reflects the uncertainty created by partially released results. As more results get released so that uncertainty cone contracts around the results and things will settle towards the final result.

The first batch of results have now come in and we present the results below. There are enough results to make a prediction but with a wide margin of error here.

How to read the data?

Our tracking model is currently showing with 6.26% of the votes counted the ANC nationally at 56%, losing 24 seats. Note this is an adjusted score based on the underlying trends in the data whilst the raw live results are actually are printing at 52%. This is why adjusting the results and looking at the underlying trends is important to account for the type of VD declaring when through the results process.  EFF is the big winner gaining 22 seats with 10.77% and the DA broadly flat. Note the error bands are very wide given the tiny fraction of the vote. The bands will narrow as more results come in. The confidence bands will tighten markedly particularly as we get to 20% of the votes counted.

For Gauteng our estimates so far, though only with a small 3.1% of the projected votes available, we see the ANC only getting 49.6% of the vote but that it can still get a majority of seats (37 of them, there are 73 seats in total so the threshold is 37) given the distribution of smaller parties.

Live national and Gauteng results in detail

Source: Daily Maverick, IEC

National ANC predicted share and error bands

Source: Daily Maverick, IEC

Gauteng ANC predicted share and error bands

Source: Daily Maverick, IEC

DM

 

Gallery

In other news...

July 18 marks Nelson Mandela day. All over the country, South African citizens devote 67 minutes to charitable causes in memory of Madiba. It's a great initiative and one of those few occasions in South Africa where we come together as a nation in pursuit of a common cause. An annual 67 minutes isn't going to cut it though.

In the words of Madiba: "A critical, independent and investigative free press is the lifeblood of any democracy."

Every day Daily Maverick investigates and exposes the deep rot of state capture and corruption but we need your help. Without our readers' support we simply won't survive. We created Maverick Insider as a membership platform where our readers can become part of our community while ensuring that we can keep doing the investigations that we do and, crucially, that our articles remain free to everyone that reads them. Sign up to Maverick Insider this Mandela Month and make that meaningful contribution last longer than 67 minutes.For whatever amount you choose, you can support Daily Maverick and it only takes a minute.

Support Daily MaverickPayment options


Comments - share your knowledge and experience

Please note you must be a Maverick Insider to comment. Sign up here or if you are already an Insider.

SCORPIO

Cash cows and computers – Ace Magashule’s laptop deals with Guptas’ Vrede dairy man; Joemat-Pettersson’s Saxonwold meetings

By Pieter-Louis Myburgh

The hacking tools used in the Matrix were real actual tools.