Primary vote
| # | % | Proj. | Swing | |
| Lou Pollard (Arts Party) | 1276 | 1.9% | ||
| William Bourke (Sustainable Population) | 2032 | 3.0% | ||
| Sam Kennard (Liberal Democrats) | 1395 | 2.1% | ||
| Kerry Bromson (Voluntary Euthanasia) | 663 | 1.0% | ||
| James Jansson (Future Party) | 474 | 0.7% | ||
| Arthur Chesterfield-Evans (Greens) | 10883 | 16.1% | 16.1% | +0.8% |
| Maryann Beregi (Independent) | 2416 | 3.6% | ||
| Silvana Nero (Christian Democratic) | 1754 | 2.6% | 2.6% | +1.6% |
| Robert James Marks (Palmer United) | 320 | 0.5% | 0.5% | -1.2% |
| Trent Zimmerman (Liberal) | 32107 | 47.5% | 47.4% | -13.6% |
| Stephen Ruff (Independent) | 12732 | 18.8% | ||
| Luke Freeman (Australian Cyclists) | 717 | 1.1% | ||
| Tim Bohm (Bullet Train for Australia) | 800 | 1.2% | ||
| Formal | 67569 | |||
| % of enrolled voters | 64.8% | |||
| Booths returned (of 43) | 39 |
1am. I’ve finally updated with final numbers for the night. The Liberals reportedly say they expect to win 58-42, which is at the high end of what I would have been expecting for them.
8.59pm. The AEC have stopped publishing the Liberal-versus-Greens results, in recognition that this is a Zimmerman-versus-Ruff contest.
8.30pm. Shellbell in comments points out the Palmer United candidate is coming last, with all of 199 votes.
8.16pm. For a point of comparison, Sophie Mirabella got 21% of preferences when she was defeated in Indi. If Zimmerman does about that well, his winning margin will be about 5%.
8.04pm. Now up to 26 booths on the primary vote. The situation there isn’t fundamentally changing, apart from a slight improvement in the Liberals’ position.
8.02pm. For what it’s worth, preferences are favouring the Greens over the Liberals by 59.5-40.5.
7.57pm. Sorry, wasn’t thinking in that last entry. The 2CP result in Liberal-versus-Greens, so it’s not surprising it’s heavily favouring the Liberals. What we need to know is the flow of preferences between Zimmerman and Ruff, which we won’t know this evening. Nonetheless, I’ve managed to fix up my table so the Liberal-versus-Greens 2PP is flowing through to it correctly.
7.51pm. There are, in fact, eight booths down as reporting on 2PP on the AEC site, and they suggest a clearer win for the Liberals than I’d thought – a 15.3% margin, although the early reporting booths were particularly strong ones from the Liberals.
7.43pm. Primary vote booth results continue to trickle in, bringing us up to 18 out of 43, but none of them are radically changing the outlook of a Liberal primary vote of around 46%-47%. In other words, a two-party booth result will be needed before there will be much of interest to add, unless some of the outstanding primary vote results throw up some surprises.
7.38pm. Now out to fourteen booths, and I’m still tracking the Liberal primary vote to come in at 45% or maybe a bit higher. Stephen Ruff’s lead over the third-placed Greens has narrowed a little, but I expect he’ll do better on preferences so still looks set to finish second. On current indications, he’d need 86% of preferences to win.
7.30pm. Eleven booths in now, and the primary vote swing against the Liberals is a more moderate but still imposing 14.7%. Zimmerman clearly headed for victory, but we’ll need some two-party results before we have a clear idea what the margin will look like.
7.26pm. A further two booths don’t change the situation on the primary vote. Still no two-party booth results yet, so no idea of preference flows.
7.17pm. Seven booths in now, and Zimmerman’s position has improved enough to suggest he should make it home. But with Ruff maintaining second place, he won’t have a hugely impressive margin after preferences.
7.10pm. Two booths in – St Leonards and Neutral Bay West – and they suggest a bigger drop in the Liberal primary vote than they’d feel comfortable with, particularly given that independent Stephen Ruff, who could presumably expect a pretty strong preference flow, has so far outpolled the Greens.
6pm. Voting has closed in the North Sydney by-election. Results should be a while coming, given that this is a city electorate with large booths, and there is a very large field of 13 candidates. This post will feature result projections and commentary as results start to roll in, which I’m guessing will be in a bit over an hour.
(by here of course I mean in this result, not on this site!)
Kevin of course 🙂
The ABC is now projecting 62.4% for Zimmerman but their projection has the Greens still in the count on minus one percent and is hence broken.
I’m very pleased to report the ABC computer projection no longer has the Greens on minus one and even more pleased to report that Labor are now on minus one instead!
No Kevin, it is called a residual. The swing against the Liberal Party after preferences is 3.5% on 7 booths. The Independent has polled 38.6% so +38.6%. So on those seven booths, the Liberal + Independent 2CP swing is +35.1%, so the Labor (Green previously) 2PP is down 35.1%, but as they only polled 34.1% in 2013, that gives a residual of -1.0%. This always happens if you show more than two candidates in the preference count. I could allocate the Labor 2PP from 2013 to the Independent and then your comment about the swing being broken would not be correct, but the Independent would be shown as +3.5%, which as they didn’t stand in 2013, does not necessarily make sense to say.
Or I could ignore the polling places all together in which case the swing against the Liberal part would be 4.5%.
At general elections the ABC computer hides all residuals by adding together the percentage of the candidate who polled zero votes 2CP last times and the candidate that polled zero votes 2CP at this election. This means only two candidates appear and the residual in the calculation disappears. It also explains why on some occasions the ABC listed swing changes from being ‘to’ a candidate to being ‘from’ a candidate. It’s all designed around the fact the when you show a swing dial, you cannot allow there to be more than two candidates in the count. In this case I thought I’d be honest and show the residual which is in fact always there but by fiddling the numbers you can make disappear if you wish.
If we are looking at a swing of 3.5% to 4.5% then that is a good result for the Libs from my perspective.
What the so-called “residual” really reflects is that the preference projections for the Liberal and Independent candidate effectively make two contradictory assumptions.
The method assumes that the Liberal 2CP will vary as more booths are added because you are using the 2013 Liberal 2CP (vs Labor) as a model for that, and in 2013 the Liberal 2CP (vs Labor) varied by booth.
It also assumes that the independent 2CP won’t vary as more booths are added because you do not have any 2013 data for that candidate to use as a model.
No, not preference projections. It’s a simple matching calculation. The residual represents the variability of the 2CP% from booth to booth. It is the same principle as the projection and the swing figure in William’s table at the top of the page. William only shows a % for candidates with a matching candidate last time. If you do show a % for a new candidate then you will be very lucky to get your percentages to add to 100%, in which case the residual is a plus or minus figure on a ghost candidate. The way round the residual is never show a percentage for a candidate with zero votes this time or zero votes last time, which is what William’s table does. Or in 2PP or 2CP, you just add the relevant candidates together to get two lines in your table, in which case your swings are equal in absolute terms and you get no residuals, which I could do with the North Sydney result but I chose to display the residual instead.
Antony GREEN@59
I should have said “projections of results after preferences”.
It actually represents the incompatability of the methods used for the two candidates, because the methods for one accommodate variability and the methods for the other don’t. If you were using the same method to model a case where there was not a new party in the 2CP then the projections for the two parties at any time would sum to 100 irrespective of variability in the 2CP.
There are many reasonable ways to avoid displaying something that looks silly on the screen in a projection. One approach for dealing with contradictory outputs is to split the difference – if one line says Zimmerman will go up by a point and the other line says Ruff will stay the same, then all else being equal the truth is more likely to fall between the two (ie 61.9:38.1).
There are more complex approaches for cases like these but I realise nonlinear assumptions are difficult to implement in a program. I used one of those last night to project after seven booths that the ABC and AEC projected primary vote swings would drop, which they did. That was relatively easy as this simply relied on the observation that the primary vote was swinging more in booths where it started higher (which is hardly any great surprise.)
The miracle is that I understand even a little of what you, Kevin and Anthony, are discussing 🙂
davidwh@61
I’m not helping by getting the odd word wrong along the way. For “nonlinear” above read “nonuniform”.
davidwh @61
Well, you’re doing better than me, then!
Asha I would just have a hidden cell where I’d put the balancing item and hope no one noticed.
Fascinating that in North Sydney – very Anglo and one of the best educated electorates in the country – the informal vote is over 6%. On postals you can see why people are confused when the NSW system is different but everybody at the booth would have been told to number every box.
2CP projections now a shade below 60:40 so the so-called “swing” looking around 6% now.
How is the last result?