The Poll Bludger

Click here for full display of South Australian election results.

Saturday night

A batch of postals in Narungga broke 21-18 in favour of One Nation over Liberal, putting their lead at 47 – which is significant because it may well conclude the count (notwithstanding that a solitary vote at the Moonta booth got shifted from One Nation to Labor, for one reason or another). The deadline for late arrival of postals was the close of business yesterday, and full preference distributions will apparently begin in some cases on Monday. SUNDAY MORNING UPDATE: The Kadina EVC tally has been revised to remove 21 votes from One Nation, reducing their lead to 26, which together with the Moonta change suggests we may not be done with rechecking. Heysen’s small number of Electoral Visitor/Mobile Declaration Votes increased the Liberal lead over Labor from 264 to 288, but there remain uncounted vote categories in that case. Nothing worth noting changed in Morphett, and we’re closing in on a week since a new vote got added to the count in MacKillop.

Friday night

ECSA’s preference throw between One Nation and Labor finally swung into action in Hammond today, and it can now be confirmed as One Nation’s second seat, with the two parties evenly placed on the primary vote and One Nation getting about 60% of preferences. Yesterday I registered surprise that the number of “polling day declaration votes” was as high as 1698 – I’m not told this was very likely a data entry error and these votes belong in a different category (polling day absent ordinary votes would be my guess).

With all sorts of new vote types added to the count, One Nation’s lead over Liberal in Narungga today was cut from 147 to 46, almost entirely as a result of a 159-57 break in electoral visitor/mobile declaration votes. I imagine that doesn’t leave much: probably only one last batch of postals. Each of the three batches of these has slightly favoured the Liberals, in slightly dimishing degree: 512-457 on Wednesday, 598-572 on Thursday and 220-214 today.

Similarly, various categories of vote were added in Morphett, suggesting that loose ends are now being tied up and we are near a final result, leaving the trailing candidate hoping for some sort of anomaly turning up in the full preference distribution. In this case it’s Liberal member Stephen Patterson, whose deficit against Labor candidate Toby Priest narrowed today from 293 to 259.

The Liberal lead over Labor in Heysen is out from 156 to 254 after they got the better of a 1054-1010 split in early voting absents broke 1054-1010 and 250-186 from a second batch of postals. That probably sees off any chance of a Labor win, and my system is the only thing extant that’s allowing for the possibility of preferences allowing the Greens to overhaul first Labor and then Liberal.

A curiosity of the past week has been the lack of any progress in the count for MacKillop, which my system rates an almost but not quite certain One Nation win over the Liberals. Nothing has been added since the polling booths and early voting centre counts were completed, which was probably on Sunday or Monday.

My system has uncalled Light, where Labor’s win probability fell fractionally below the 99% threshold after postals 1016-939 in favour of One Nation. Labor leads by 538 votes though, and by this stage of the cuont my system is allowing for more outstanding votes than we are actually likely to get.

Thursday night

Disappointingly, there is still no sign of the Labor-versus-One Nation count that could settle the issue as to which of the two will win Hammond. There were around 3000 postals and “polling day declaration votes” (which turn out not to be what I thought they were, since otherwise there would not have been 1698 of them) and they leave the two almost dead level on first preferences. It can thus be inferred that whoever gets the most preferences will very likely win the seat. The only other regional seat with a Labor-versus-One Nation two-candidate count is Ngadjuri, and in that case One Nation is getting about 60% of them. The main difference between the two is that 10.0% of the vote in Hammond has gone to Airlie Keen, whose how-to-vote card had Liberal second, One Nation sixth and Labor second last. So it seems unlikely that the preference flow here will be radically different from Ngadjuri, which makes me confident in my system’s assessment that One Nation will very likely win the seat.

The Liberal lead over Labor on the two-candidate count in Heysen increased from 54 to 156 after the first batch of postals broke 911-576 to Liberal and the first batch of absents broke 928-704 to Labor. The Greens did poorly on postals, as always, and causing their deficit against Labor on first preferences to increase from 0.9% to 1.6%, which they will need to close on preferences, three quarters of which will be from One Nation, to make the final count against the Liberals. The most proximate useful example I can find from the federal election is the seat of Adelaide, where the Greens got 20.2% of preferences when One Nation was excluded (this included One Nation’s 4.02% primary vote and another 1.76% they picked up on preferences) and Labor got 17.64%. It is for this sort of reason that my three-candidate estimates are giving the Greens an even money chance of closing the gap, but things may be very different in the context of a greatly increased One Nation vote. Whether they would be better placed than Labor to defeat the Liberals at the final count is another open question. I would also reiterate my point about ECSA persisting with a Liberal-Labor two-candidate count and the likelihood that they have made an informed decision in doing so.

In Morphett, Labor candidate Toby Priest’s lead over Liberal incumbent Stephen Patterson increased today from 34 to 293 after a second batch of polling day absents broke 592-459 and a first batch of early voting absents broke 553-427. To close the gap, Patterson will need outstanding two-thirds of postals to split about 57-43 in his favour, as compared with the 53-47 he got from the first batch, and to break even on whatever else remains oustanding.

A second batch of postals in Narungga behaved similarly to the first in breaking 598-572 in favour of Liberal over One Nation, reducing the latter’s lead from 173 to 147. However, the Liberals now have less runway in terms of postals, each of the two batches having accounted for about a third of what’s likely to be the total, and federal precedent suggests absents are more likely to favour One Nation.

Continue reading “South Australian election: late counting”

Click here for full display of South Australian election results.

Some very general observations on the result before diving into specifics. As expected, Labor appears to have won every seat in Adelaide except Bragg, where Liberal member Jack Batty achieved a surprisingly normal-looking result. That involved unseating the Liberals in Colton, Hartley and Morialta, which Labor had held at various points during its previous time in government, together with Morphett and Unley, which it hadn’t. In a great many other seats in Adelaide, the Liberals have been reduced to minor party status, in many places running fourth and in one place even fifth. My system is calling four seats for the Liberals and has them leading in a fifth, making them likely but not certain to retain opposition status ahead of One Nation, who are being called in three and ahead in a fourth.

Labor may not be entirely spared the One Nation lash though, which needless to say has been the other big story of the election. My own preference estimates suggest a knife-edge result in Light, which the long-serving and highly popular incumbent Tony Piccolo abandoned for an unsuccessful run in Ngadjuri, presumably to help Labor win both seats but in fact putting them at the risk of winning neither. The ABC’s preference estimates give Labor a slightly more comfortable margin of 2.0%, but we really won’t know until we get a fresh two-candidate count, which hopefully ECSA will conduct over the coming days. Also in northern Adelaide, Elizabeth and Taylor have also gone from being safe Labor seats against Liberal to marginal ones against One Nation.

My system is calling three and very nearly four regional seats for One Nation, though none are being given away by the ABC. However, it’s difficult to see the Liberals closing their substantial primary vote deficits in MacKillop and Narungga, and in neither case is the independent incumbent competitive. I’m a little less confident about my system’s call of Ngadjuri for One Nation, but Liberal incumbent Penny Pratt will undoubtedly have a difficult time overcoming Labor to make the final count, as she’s effectively tied with them on the primary vote at present and the minor parties whose preferences will decide the issue are mostly of the left. If she can’t manage it, my system is undoubtedly correct in projecting that her preferences will give the seat to One Nation over Labor.

I am just about giving Hammond to One Nation, where Liberal member Adrian Pederick has fallen to a fairly distant third, with his preferences set to decide the result for One Nation over Labor, each of whom have about 27% on the primary vote. Neither my system nor the ABC’s are writing off One Nation’s chances of deposing independent Geoff Brock in Stuart, where we’re both estimating margins for Brock of 4.5% and applying very conservative error margins due to the complications involved in projecting an independent-held seat that has gained territory in a redistribution. Brock would have to be rated a pretty clear favourite though.

There are two other seats that my system rates as in doubt. Heysen is rated a three-way contest between Liberal, Labor and the Greens: the latter would have to close a small deficit against Labor to make the final count, but my system gives them a solid chance mostly by giving them more preferences than Labor from One Nation, based on observation of preference distributions from the federal election. It’s a close run thing though, and should they fall short, ECSA’s notional two-candidate count shows Labor leading Liberal in two-candidate terms by all of two votes, although late counting will presumably favour the Liberals.

Kavel looks fiendishly complicated, with no candidate polling higher than 23.9%. That candidate is Labor’s, who my system is rating the favourite, but its preference assumptions wouldn’t need to be too far wrong to overturn this. ECSA’s two-candidate count shows independent Matt Schultz will win if he faces off against Labor in the final count, but to do that he will have to get ahead of One Nation, who he slightly leads on the primary vote. However, my system assumes Liberal will drop out before all of the above, which I think very likely, and has more of their preferences going to One Nation than to Schultz, which is little more than a guess.

Those who were following my results late last night may notice some changes to the projections, which are due not to late counting but my reconfigurations of who the last three candidates are likely to be and changes to preference estimates. The system has been dealing with combinations of candidates that hadn’t previously been encountered, and in some cases was splitting preferences evenly where it had no defaults to fall back on. I haven’t been keeping score, but I believe my system had been calling Kavel to Matt Schultz on this basis, and I’m certain it wasn’t calling Finniss for independent Lou Nicholson, as it’s now doing. It’s more than possible that I’ll unearth another anomaly or two when I review the matter in the morning, or more likely afternoon.

I’m not exactly sure what can be expected today in the way of counting, but around half of the early voting centres were evidently unable to complete their primary vote counts by the end of last night, and only a few reported two-candidate numbers. There has been a fair bit of talk about the impact of South Australia’s unique savings provisions that cause ballots that have not been fully numbered to be deemed to conform with preference tickets pre-registered by the parties in question. This is of particular interest due to One Nation how-to-vote cards that show only a number one for the One Nation box and advise voters to fill out the rest in order of preference, which some may have misinterpreted as meaning a single number constituted a valid vote for One Nation.

However, the effect of this in late counting will be minor. One Nation lodged multiple registered tickets with the effect that its saved preferences will divide evenly between Labor and Liberal. Only in the four seats where it was included in the notional two-candidate count (Chaffey, Giles, Schubert and Wright) are its own soon-to-be-saved votes currently withheld from a published count, and none of these are in doubt. Furthermore, the differences in the number of votes reported on the primary vote and two-candidate counts is not all that dramatic, so the number of votes should in any case not be over-estimated.

Before I finally turn in for the evening, my first look at the Legislative Council result. Only 219,769 first preference votes have been counted, which is barely a third of the lower house count total, so this must be considered highly preliminary. It currently looks very much like Labor four, One Nation three, Liberal two and Greens one, with one more seat in doubt. With Labor on 4.4 quotas and the Greens on 1.4, one would think their collective surplus of 0.8 quotas would mean a further seat for one or the other. That would add up to nine or ten seats for Labor in the new parliament along with two or three for the Greens, for a combined total of 12 out of 22. The Liberals would have six, and One Nation’s new cohort of three would be supplemented by former party member Sarah Game, whose Fair Go for Australians party was one of a number of right-of-centre concerns that accomplished very little at this election.

Click here for full display of South Australian election results.

I’m otherwise engaged this evening, so won’t be offering anything in the way of live commentary, but hopefully the Poll Bludger live results system will tell you everything you need to know. As well as the main landing platform linked to above there is also a map display that will colour in as results are reported to reflect who the system deems to be ahead or to have won, respectively indicated with a lighter or darker shade.

Each individual seat page comes with projections (which start from assumptions about who the three leading candidates will be, which in some cases I may have to change on the fly if I find time) and results at swings at booth level, displayed in both tabular and mapped form (for the latter, click the “active” button at the bottom of the page). As always, I’m crossing my fingers hoping this will all work okay. I have made life harder for myself by implementing a new method for determining win probabilities, which would ideally have been rolled out at a less complicated election. Hopefully its errors will be on the side of caution – a may implement some rather crude fudges on the fly to make it so if that doesn’t seem to be what’s happening.

This will be a landmark election night so far as the Electoral Commission of South Australia’s reporting of the results is concerned, as early voting centres will be counted on the night for the first time (though experience suggest some of them will handle too many votes for them to finish the job before the close of business). What were formerly reported in one line as “declaration votes” will now be reported at an appropriate level of complexity – early voting centres no longer count as declaration votes, and will be reported on a location by location basis, and the rest will for the first time be reported separately as postals, absents and the best. This complicated matters for those of us who need to produce datasets of previous election results to match against those that will be reported tonight for swing calculation purposes, but is undoubtedly a welcome development in the long run.

An added layer of challenge lies in the fact that the ECSA offers no indication as to which candidates it has selected for the notional two-candidate preferred counts, which are usually provided to the media on a confidential basis. A couple of seat pages are in fact not working as they should be because one of my tables says one thing about who the candidates will be another one says another, but I believe I’ve geared things so it will correct itself when the appropriate candidates can be identified. If however individual seat results pages continue to misfire, the likeliest explanation (though my no means the only one possible) will have something to do with this issue. In an ideal world I would be free to patch things together, but as it stands I will be otherwise engaged, so it’s a question of fingers crossed.

Naturally all of this involved a vast amount of unpaid work, so if you believe that situation should be rectified, your donations are gratefully received through the “become a supporter” buttons at the top of the site and the bottom of each post.

Two late results from South Australia make it four out of four polls from the last week of the campaign with One Nation ahead of the Liberals:

• Courtesy of The Australian, Newspoll has the Liberals on 16% of the primary vote, their consolation being that this is two points higher than the Newspoll result from the start of the campaign. Labor is down four to 40%, One Nation down two to 22% and the Greens steady on 12%, and others up four to 10%. Peter Malinauskas is down two on approval to 65% and up four on disapproval to 31%, while Ashton Hurn is respectively up four to 43% and steady on 35%, for all the good the improvement seems to have done for the Liberals. Malinauskas leads 64-22 on preferred premier, in from 67-19. Newspoll’s policy appears to be not to provide two-party numbers when One Nation is running second. The poll was conducted last Thursday to this Wednesday from a sample of 1048.

• A DemosAU/Ace Strategies poll for InDaily has Labor at 37%, significantly moderating the 43% recorded in its last poll four weeks ago, with One Nation up four to 23%, the Liberals down one to 17% and the Greens down one to 11%. Peter Malinauskas records personal ratings of 49% positive, 31% neutral and 20% negative, while Ashton Hurn is respectively at 21%, 58% and 21%, and Cory Bernardi is at 20%, 44% and 36%. Malinauskas leads Hurn 56-21 on preferred premier. Like Newspoll, the poll was conducted last Thursday to this Wednesday, the sample in this case being 1242.

Three of this week’s four polls have broken down their results by inner metropolitan, outer metropolitan and regional, as detailed on the table below, and the story they tell is consistent enough that no more need be added to my earlier analysis of the Fox & Hedgehog result, which concluded that Labor could potentially sweep the board in Adelaide, Bragg being the most likely Liberal hold-out. The situation in the regions is a good deal harder to read, with One Nation needing to overcome Labor preferences and a number of strong independents to make anything out of its commanding primary vote. The report by David Penberthy in The Australian accompanying the Newspoll result rates its strongest chances as Narungga, Hammond, Flinders and Ngadjuri.

UPDATE (Resolve Strategic): One last poll comes from Resolve Strategic, which it describes as “experimental” on the grounds that respondents were met with an interactive AI voice. Their caution is no doubt informed by the fact that the poll has an even higher One Nation vote than everyone else, at 28%, with Labor coming in well below par on 32%, the Liberals about evens at 18% and the Greens on 10%. The poll was conducted on Monday from a sample of 1112.