Monday, July 4, 2016

2016 House of Reps Postcount: Vanilla Reps Seats

Expected final outcome Coalition 76 Labor 68 Others 5 Undecided 1
Herbert: Labor leading, much too close to call, recount very likely (moved to new thread)

Coalition assumed to win: Grey, Dunkley, Chisholm, Gilmore, Flynn, Forde, Capricornia

Labor assumed to win: Batman, Cowan, Hindmarsh, Melbourne Ports (see separate thread)

Note: You can now vote in the sidebar Not-A-Poll on what will be the closest seat!


This thread deals with the standard (hence the "vanilla" in title) two-party preferred seats at the 2016 federal election, as distinct from more exotic offerings like Grey and Melbourne Ports which have their own threads.  Usually the Coalition performs slightly better on post-counting than Labor and for this reason I have (for now) not included Robertson, Petrie, Dickson or La Trobe, in all of which the Coalition leads by more than 1000 votes.  Should any of these close up I may cover them. In 2010 and 2013 Labor never won any seat in which the Coalition had more than 50.53% at the end of the ordinary votes.  However with the different mix of postal votes (see below) perhaps Labor might make such a gain somewhere.

The seats included are those that on election night were projected to within 51:49 either way, or any other seat that becomes close on the live 2PP count (so far only Dunkley has been added.)  However I have for now excluded Lindsay, where the Coalition was projected to close to 49.3, but from a long way back, making it much more difficult to beat the projection by enough to win and Longman (though I am keeping an eye on these two).  The Coalition has performed extremely well on early postals in Longman but I cannot get the seat within 1000 votes on my projection even assuming that flow continues so I think that one is over.  The projection has now been switched off but over time I'll be doing my own in the more interesting seats, or bludging (pun intended) off those done by others.  Batman had a close on-the-night projection because the Greens came back strongly on absents and declaration prepolls in 2013 but a strong batch of postals looks to have sealed the deal for David Feeney.  (Michael McCarthy has a lot of goodies on Batman).

The big question for whether the Coalition scrapes into majority government concerns the accuracy of projections based on past postcounts.  These projections (as seen on the ABC and AEC sites before the Coalition suddenly fell behind in several seats when they were turned off) consider evidence from 2013 of performance in the postcount relative to the ordinary vote count.  The projections had the Coalition ahead in three seats where it trails: Flynn, Cowan and Capricornia.  However the Liberals went down following one of the prepoll booths in Cowan so probably fell behind there just before it was switched off.

As I start this thread, winning Grey (if that happens) gets the Coalition to 70 seats.  If they hold the lead in Dunkley, Gilmore, Chisholm they reach 73. If they retake the lead as projected in Capricornia and Flynn that's 75, and then any of Forde, Hindmarsh, Herbert or Melbourne Ports (of which Forde may be winnable on postals and the other three difficult) is a one-seat majority government.

On the other hand, Flynn is shaky because it relies on assumptions that extreme postcount patterns seen there will continue (they might not; then again they might grow stronger). Gilmore is very close and the comeback required to win Capricornia is no certainty.

The median outcome seems to be 75 (I'm treating Capricornia as likely and Forde and Flynn as tossups) assuming the postcount patterns from last election stay.  But if they don't then the Coalition's prospects of a scraped majority grow.  The Coalition wins a majority if it gets lucky or if postcount voting favours it by a fair deal more than it normally does.  On the other hand, if no leads change (aside from NXT's likely-to-be-spurious lead on the booth realignment in Grey), then we will be back where we were in 2010, but with McGowan and Sharkie replacing Oakeshott and Windsor - whatever that might mean.  

Seats will be covered in alphabetical order.  When a seat is called it will be moved to the bottom of the page.

Post-count votes and 2PP changes

As the postcount goes on I will be looking in more detail at things like evidence of party postal effort in various seats, and the 2016 breakdowns of non-ordinary votes and how they've changed.  So my assessments of some seats could change when I've factored in all these details.

Normally the Coalition tends to perform better on post-count votes than on booth votes, and so it tends to improve its position by about 0.3% two-party preferred from the end of the primary count to the final result.  However, different seats have their own characteristics that cause some to have favourable postcounts for Labor and some to have extremely favourable postcounts for the Coalition.  While the patterns will vary in a given seat from time to time, there is usually a correlation between how Coalition-leaning a seat's post-count is one election to the next.  Here (on request!) is a graph showing this from 2010 and 2013:

This graph only includes the seats that were classic 2PP seats in both years.  In most seats the Coalition's benefit from the post-count in 2013 was within 0.3 points of the 2010 benefit but there were a few outliers where the benefit changed by several tenths of a point.  The three seats where the Coalition gained massive benefits from postals in both years were Maranoa and Flynn (Queensland) and Melbourne Ports (Vic - large Jewish community).

There is a view that the Coalition's postal campaign was particularly strong and that this may lead to stronger postcount performances. I am seeing evidence in the AEC downloads that there were particularly strong increases in postal voting in Queensland, Victorian and WA seats, which include seven of the nine I list below.  As noted by Morgan in comments, these states are on school holidays.  The question then is whether the increased postal votes will show similar skews to last time, or whether they will be diluted by a broader range of voters than in 2013 using them.   Given that an increasing proportion of postal voters are therefore young parents, I suspected the answer was diluted, but early counts are not showing this. This may be a result of the LNP campaign on postals at the start of the campaign and late postals might moderate.

Conceding seats

There will be a lot of focus in these close contests on whether or not the trailing candidate concedes.  Whether or not a candidate who is clearly losing concedes is utterly irrelevant to the outcome.  It is just a nice thing to do to realise you have lost, congratulate your opponent and reduce uncertainty for voters.  A candidate might mistakenly concede then actually win, and their victory is not affected.  


Seats are not recounted automatically just because they are close.  A recount is automatic for a margin of 100 votes or less, otherwise a convincing case has to be stated and accepted as to why the seat should be recounted.  See AEC recount policies.  If any seat goes to a recount I will give the recount its own thread.

A note on the 2PP

The 2PP figure showing on the AEC website (currently 50.22 to Labor) is based only on the votes counted; it is not a projection.  Sixteen non-classic seats (Denison, Grayndler, Melbourne, Batman, Wills, Indi, New England, Cowper, Durack, Higgins, O'Connor, Barker, Mayo, Grey, Kennedy. Murray) are not included and these have an average lean to the Coalition that should increase the final 2PP by about 0.14 points (assuming the swing in those seats equals the national average, or more if it is below the national average).  Within the remaining seats the proportion of votes counted varies, and I get the average swing in the 134 presumably classic seats at 3.5%, so the 2PP would probably be about 50.14 to Coalition if the same number of votes had been counted in all seats.  If the Coalition's performance in postcounting is the same as last time the final 2PP will be about 50.4. It could yet be this changes in the Coalition's favour because the postcounts are larger, but based on 2013 I would not assume this.

A note about absent votes and out of electorate prepolls

Often there is excitement if a batch of absent votes does something at odds with past patterns.  However a given seat will often have a Liberal-voting end and a Labor-voting end (for instance) and the behaviour of the absent votes then varies based on what edge of the electorate they come from.  For this reason you can't project off absents until you have multiple batches of them (at least not reliably).


Herbert (Qld, LNP, 6.2%) - moved to new thread

Herbert is another of the group of Queensland seats that has either fallen or come under risk on a large margin while others on smaller margins have held.  Labor's Cathy O'Toole leads Liberal incumbent Ewen Jones by 1084 votes (50.74) with 75.3% counted.  In 2013 the ordinary vote was 80.3% of 93.3% total and Jones gained 0.32% in the post-count.  The postcount for Herbert is, however, substantially bigger than last time and if it turns out that postals are a larger part of that then the seat may close up more.

Tues 5 July 9 pm: Labor lead cut to 913 on late-counted ordinaries, a significant gain.

10 pm: The first batch of 1333 formal postals have broken 58.2% to Jones, which is about the same as last time, notwithstanding the large primary vote swing.  Labor lead cut to 694.  Postals will not necessarily remain as strong for Jones but today has already wiped off most of Labor's projected election-night winning margin.

Wednesday 10 pm: Jones continues performing strongly on postals though the overall flow has moderated to 57.3% and the lead is 620 votes.  Based on the greatly increased number of postals in this seat (it was 7000 last time but 11320 envelopes have been issued this time) it seems that either the strengthened postal flow (if it holds up) or the increased number of postals are capable of causing Jones to win. But we'll see if the remaining postals keep breaking so strongly.

Thursday 5 pm: There is no further change in the postal flow and the lead is just 449 votes.

Friday 2 pm: Crikey's "Tips and rumours" section today claimed that 177 votes initially classified as informal at a booth are missing in Herbert.  There is at least one booth in Herbert (Kelso) that had 177 informal votes counted on the night.  There is no official confirmation on the claim at this stage.

Monday 2pm:  There is no major change in the patterns, but counting of more postals and some absents have reduced the lead to 162 votes.  There is still no word on the missing votes claim above.

Tues 3pm: Some prepolls have broken 54:46 to Jones, about a point better than projected. He now trails by 51 but there are plenty of postals left to erase that.  He should win the count, the question being whether he can do so by enough for a safe margin.

Wed 12pm: The margin is now 66 to Labor but there are only at most 571 absents remaining, plus an unclear number of provisionals.  Absents have not helped Labor all that much here and the remaining postals and declaration prepolls should put Jones over the line with a non-trivial margin.

Wed 1 pm: Senator Ian Macdonald is reporting on Twitter that further postal counting is putting Jones ahead by 113 with "1500" postals to go.

Wed 4 pm:  Labor back in front by 35! When it's this close anything can happen but Jones should still retain the seat.  A few hundred postals were counted but I wonder if there may have been a correction to ordinary vote data.

Wed 9 pm: It does look like there was a correction to the ordinary vote total there.  Jones is now back in front by 34 votes and I am still projecting him to win by about 180 if the number of provisionals accepted is the same as last time.  But if many more provisionals are accepted we could be in for an even closer result.

Thurs 12 pm: Kieren Gilbert (Sky) reporting Jones' lead up to 148 after counting of 757 postals.  That will leave up to 257 absents, 856 prepolls and at least 305 postals (a few hundred more may make it) plus however many of the 945 provisionals are accepted.  However it is common for a small number of absents and prepolls to also be rejected (eg 100 absents rejected in one seat in 2013 and 232 in another).

Thurs 12:30: The VTR now shows the lead as 139 with 301 remaining known postals.

Thurs 2:30:  Another adjustment to the overall totals (presumably rechecking from a booth) has cut Jones' lead again, this time to only 31 votes.  The seat is still projecting to Jones, but by less than 100 votes meaning that a recount is quite likely.

Thurs 6:40: We are down to the provisionals and the dregs with Jones leading by 44.  There are 198 absents. 404 prepolls and 34 postals in theory to go, but a small number of extra postals will arrive while some of the others will be rejected.  Given that provisionals skew to Labor it is not easy for Jones to avoid an automatic recount here, assuming that he keeps the lead.

Note that before any recount is considered comes the distribution of preferences.  It is normal for mistakes to be found during the distribution.  In 2013 Clive Palmer provisionally won Fairfax by 36 votes, which became 7 votes after the distribution of preferences and then 53 votes after the recount.  The distribution of preferences rechecks all the non-major-party votes, but not the votes for Jones and O'Toole.  It is in theory possible that a margin could be within the automatic recount margin before the distribution, and outside it afterwards, in which case the loser would have to show cause for a recount.

Fri 6:00: 153 provisionals have been thrown breaking 61% to Labor and the margin is just 19 votes. No significant change in other categories and 505 provisionals still to be dealt with, though how many are accepted remains to be seen.

Fri 7:00: I can't see where the change has come from so it might have been rechecking.  The margin is 12 votes.  They're playing for sheep stations here given the thinness of the Coalition's majority, so if this goes to a recount after the preference distribution (which now seems very likely) it may very well also go to court.  If it does go to court the declared winner following the recount is seated until the case is resolved.

Sat 12:20: Remaining declaration votes are being checked against the electoral roll today and will be thrown on Monday, according to the AEC.

Monday 1:00: The lead is still 12 but several dozen provisionals have been thrown (doing approximately nothing) and there are now 130 absents, 316 provisionals, 249 prepolls and 15 postals listed as in theory available.  How many of these are actually usable remains to be seen but even if all were thrown and split 55:45 one way or the other that wouldn't get us out of the automatic recount zone.

Monday 4:00 Labor in front by eight votes.  Still in theory 573 remaining.

Monday 6:00 The 573 do not exist.  Too much media misreporting so this count has been moved to a new thread.


Seats no longer being followed unless something dramatic happens.

Capricornia (Qld, LNP 0.8%)

Capricornia was invariably expected to fall in the election leadup, but has been close.  Labor leads by 991 votes (50.68%) with 76.6% counted.  In 2013, the ordinary vote was 79.9% and total turnout was 94.8%.  Michelle Landry gained 0.88% on a two-party preferred basis in 2013, has the advantage of incumbency and the outstanding vote this time is slightly larger. It seems in theory that she has a solid chance of pulling in that margin.

Wednesday 10 pm: Some good news for Labor in a Queensland postcount at last with the first batch of 1994 postals splitting only 54.4% to Landry, which is three points below the projection. However this seat has a greatly increased postal count so projecting even the weak flow from this batch puts the seat on the wire.  Landry would want to improve on subsequent postals.

Thursday 3:30 pm: More postals and the flow to Landry has improved to 56.4% (closer to the projection) which gives her an edge at this early stage; absents have apparently started.  The margin is 472.

Friday 3:00 pm: More postals and the flow to Landry has improved to 57.6% (about as projected) which cuts the gap to 174 with at least 5000 postals to go, so this is looking like a Coalition retain.

Saturday 7 pm: Michelle Landry is now in the lead by 148 votes and there are still over 3000 postals to go.

Sunday 11 pm: The ABC computer has called this one and I am inclined to agree since Landry leads by 691 and even if all remaining absents and prepolls split 60:40 to Labor (which won't happen) Landry would still win it on postals.

Chisholm (Vic, ALP 1.0%)

Chisholm could be Labor's only casualty at this election (following the retirement of Anna Burke).  Julia Banks (Liberal) leads by just 66 votes (50.04%) with 68.1% counted.  In 2013 the ordinary vote was 70.6% (final turnout 93.4%). The Liberal candidate gained 0.61% in the postcount in 2013.

Tues 5 July 9 pm: A batch of almost 2000 postals broke 57.3% to Banks putting her 362 votes ahead.  That's about two points better than the break expected from 2013 but nothing to get excited about yet in terms of the possibility of super-strong postal flows for the Government elsewhere. It is looking rather likely Banks will win the seat.

Wednesday 6 July 1:30 pm: Banks' lead increases to 754 with postals continuing to break at around the projection (though they have dropped back a little now) and there seems little doubt the Coalition has won the seat.

Wednesday 10 pm: With Banks' lead out to 1394 this seat is done and dusted.

Cowan (WA, Lib 4.5%)

In Cowan, Anne Aly currently leads Luke Simpkins by 959 votes (50.73%) with 71.9% counted.  In 2013 the ordinary vote was 75.3% (turnout 93.1%) and Simpkins gained 0.33% in the postcount, but that was in an uncompetitive contest, and postcount patterns could also be affected by the redistribution.  This looks rather strong for Labor but still a very long way to go.

Tuesday 5 July: Indirect reports from Labor scrutineers that postals are breaking 55:45 to Simpkins, which if true is in line with the projection and fine for Labor.

Wednesday 6 July 11 am: The reports were accurate as the postals broke 54.6 to Simpkins.  That beats the projection for Labor, plus postals in Cowan are no more numerous than last time.  Cowan (margin currently 750) is strongly firming as a Labor gain.

Thursday 5 pm: A small report of goodies for Dr Aly - apparently 50 informals have been found in her opponent's pile! Also the postals break has weakened slightly so this is looking the best for Labor of the seriously undecided seats.

Friday 6:20 pm:  Apparently bad news for Labor with a parcel of absents breaking 291-214 to Simpkins.  The 57% flow is 8.5 points stronger for Simpkins than the ordinary votes, while in 2013 he was 2.1 points weaker on absents.  So if you plug that flow into a projection it will blow the seat completely.  But at this stage all this is is a lesson in why we should not project off early parcels of absents.  Cowan is a very politically diverse seat with areas leaning up to 70:30 one way or the other.   These absent votes have probably just come from a booth on a Liberal-leaning edge.  We will need to wait a while to see if absents display any consistent pattern that could be hazardous to Labor's chances here.

Friday 11 pm: Another batch of absents also broke to Simpkins, but more weakly (329-271) cutting Aly's lead to 427.

Saturday 7 pm: Aly leads by 399 but there don't seem to be many postals left (790 known and up to another 1361 that may or may not arrive).  On the other hand there are 5000 absents which should break in her favour, especially with those two presumably unrepresentative batches gone.

Sunday 8 pm: Aly leads by 487.

Monday 2:00 pm: A major development with reports of 200 Green votes being found in Simpkins' pile in the Tapping booth.  It is quite obvious that there is indeed a discrepancy in this booth with the Greens vote at only 1.36% on the night compared to 8.04% in 2013; adding 200 votes brings it to 7.5%.  Presumably these will mostly flow to Labor and once included the margin will blow out by at least 240, since 200 former Liberal 2PP votes will become at worst 120-80 to Labor.  If this is confirmed then barring similar mistakes the other way it could be just about game over!

Monday 11:00 pm: Labor have claimed victory and barring significant such mistakes in Simpkins' favour I agree with this claim and have moved the seat below the line, with Dr Aly leading by 946.

Dunkley (Vic, Lib, 5.0%)

Dunkley was widely thought to be at risk before the election based on the retirement of Bruce Billson, but the CFA saga dented Labor's winning chances in Victoria.  Chris Crewther leads by 438 (50.32) with 69.8% counted.  In 2013 the ordinary vote was 73.6% (turnout 93.3) and Billson gained 0.80% in the postcount.  That gain might not be repeated for a vacant and competitive seat but it will be surprising if the lead gets overturned.

Wednesday 10 pm: A 60-40 split on postals has beaten the projection, which had Crewther comfortable anyway, so Dunkley is a fizzer.

Flynn (Qld, LNP, 6.5%)

Flynn might seem a spurious inclusion given that Zac Beers' lead stands at 2058 votes (51.53) with 71.2% in the can.  However in 2013 Ken O'Dowd stretched his lead by an enormous 1.81% in the postcount, mainly because postals went 69:31.   That was certainly not a once-off; the gap closed by 1.9 points in 2010 and 1.97 points in 2007, so the kind of resurgence that would save the seat for O'Dowd is a standard feature of this seat.  It was generally thought that FIFO workers might be causing this pattern of postcount shift to the LNP, and William Bowe suggests that maybe the pattern will therefore weaken this time.  There will be many beers for Beers if he can manage to turn off this tap.  Whether he will or not, I wouldn't have the foggiest.  In 2013 the ordinary vote count was 76.2% (total 94.6%).

Tuesday 9:40pm: And it begins,  A batch of 766 postals breaks 65.27% to O'Dowd.  That's 16.8 points better than the ordinary vote but this is only slightly more than the 14.5 point difference last time.  The lead drops to 1824.

Wednesday 10 pm: More postals and the lead continues coming down, now to 1075.  O'Dowd has almost halved the lead already, and though the postals are slightly less strong overall now, he is still beating the projection.

Thursday 3 pm: More postals but the total break to O'Dowd has now moderated to 62.7% (14.2% above last time) and Beers' lead has increased to 1167.  O'Dowd is no longer beating the projection on postals but on the other hand there are thousands more of them than in 2013 so O'Dowd is still projecting to win unless the postals flow weakens a lot.

Friday 2:30 pm: Still more postals and the break to O'Dowd has again gone over 63%, cutting Labor's lead to seven votes with still five to seven thousand postals to be added.  It is now highly likely that O'Dowd has won the seat.

Saturday 7 pm: There is no letup for Labor against this endless wall of postals.  O'Dowd has now taken the lead by 266 votes and there are still 3000+ of the things to go.

Sunday 8:10 pm:  There has been a substantial correction in the Wondai booth which has stretched the margin by 125 votes.  As a result O'Dowd now leads by 391 votes.

Monday 2:00 pm:  A small number of absents and prepolls have been added.  The prepolls broke 52.4 to O'Dowd which is marginally weaker than projected, the absents broke 54.6 to Beers which is more or less in line with projection (especially without knowing where they came from).  O'Dowd leads by 411.

Monday 11 pm: With O'Dowd now 933 votes in front I have looked at the votes remaining and decided to stick a fork in this one.  Hopefully this result will cause people to be more careful about this seat in the future!

Forde (Qld, LNP, 4.6%)

Forde is the seat where former Queensland Premier Peter Beattie attempted a comeback and was defeated by local member Bert van Manen. Now van Manen's up against Des Hardman, who was pushed aside for Beattie.  Forde looks ridiculously close.  Hardman leads by 149 votes (50.11) with 70.14% counted.  In 2013 the ordinary vote count was 74.5% (turnout 92.6%) and van Manen gained by 0.14 points.  So Forde is tracking into recount territory if patterns from last time hold up, but the increased postcount and the increased postal postcount especially may be positive signs for van Manen.  Another point to bear in mind here is that if a seat has a low post-count flow to the Coalition one election then on average that flow will increase next election (regression to the mean.)

Wednesday 12 pm: The gap has dropped to 77 votes, but no postals have yet been counted.

Wednesday 1 pm: The gap has dropped to 53 votes.

1:20 pm: van Manen has taken the lead by 94 votes, presumably on late ordinaries or rechecking as we still have not seen any postals here.  Labor now has to win the postcount, which usually doesn't happen.

Wednesday 10 pm: 2594 postals split 56:44 to van Manen.  Again the Coalition slightly beats the projection so van Manen must be a lukewarm favourite to retain now.  His lead is now 265.

Thursday 3 pm: Van Manen continues to get a 56:44 split on postals and leads by 441.  His position looks fairly good.

Thursday 11 pm: More postals at much the same going rate and van Manen now 687 ahead.

Friday 6 pm: The first 1332 absents have been counted, breaking 58.2% to Hardman, which is 1.5 points better than projected.  The lead is now 470. Because we don't know where absents come from, we can't use a batch to project the total performance on absents, but even if we did it wouldn't quite overturn the projection yet.

Saturday 7 pm: van Manen leads by 915 votes.  Virtually all of the received postals are gone now but there are bound to be more to come, which should put paid to any hope of a successful Labor comeback on absents. Van Manen is also performing well on out-of-electorate prepolls thus far.

Monday 11 pm: Called.

Gilmore (NSW, Lib, 2.7%)

In Gilmore Ann Sudmalis (Liberal) has 50.23% for a lead of 405 with a high (for this election) 80.22% counted.  In 2013 the ordinary vote was 82.5% (of 94.3%) and Sudmalis gained 0.11% of the vote.  So on 2013 patterns Sudmalis has good chances to retain but it is too close to be certain.

Tues 5 July 9 pm: Lead cut to 353 on late-counted ordinaries.

Wed 6 July 12 pm: 1855 postals broke 59% to Sudmalis, beating the projection for her by a few points.  The lead is 688 and Gilmore is looking likely to hold.

Wed 10 pm: The AEC header page seems to have more up-to-date figures than the AEC seat page for Gilmore.  Apparently the lead is 991 now.

Thurs 3 pm: Sudmalis now leads by over 1000 and I don't see any real prospect of Labor getting this back since there are still thousands of postals left.

Hindmarsh (SA, Lib, 1.9%)

It was hard to know what was going on with the SA two-party-preferred vote in South Australia in polling before the election because of all the noise created by NXT.  The dust has mostly cleared and as expected only Hindmarsh is in play.  Labor's Steve Georganas, the former member, leads by 432 votes (50.27%) with 75% complete.  In 2013 the ordinary count was 76.1% (of 93%) and the net change in the 2PP during the postcount was nothing. A perhaps relevant difference is that Matt Williams is the incumbent this time.

Monday: As noted in comments, the addition of an outstanding booth has put Labor out to 639 votes (50.4)

Tues 5 July 9 pm: 4294 formal postals broke 54.39% to Williams which is a whole 17 votes better than he would have been expected to do on such a batch based on 2013 preferences, so unless the share of postals in the postcount has increased that's actually a really good start for Georganas.  The gap closed to 347 votes but Labor have good chances here.

Tues 6 July: More postals but they haven't changed the pattern.  The lead is down to 167 and Williams could well take the lead on the remaining postals by 150 or so if things go well, but even the 2013 patterns on absents and pre-polls would then put Georganas in front, and there should be a swing to him on those this time around.  The number of postals is up slightly but not enough to be likely to matter on my calculations.

Thurs 3 pm: Hindmarsh has gone to the wire with a margin of eight for now, but there is no change to speak of in the postals pattern.  There is however an increase in the potential number of postals in the seat, which might come to (say) 2000-2500 more formal postals than last time depending on how many are received.  This projects the seat to be very close - I still have Labor ahead by a few hundred though, but it is by no means done and dusted.

Thurs 11 pm: Georganas leads by 68 (I understand a minor error was detected).

Friday 11 pm:  Some absents have broken to Georganas 64-36 but don't get excited, they could be from anywhere (update: and are in fact from the ALP seat of Port Adelaide).  Georganas leads by 177.

Monday 11 pm: With Georganas now leading by 583 and at least as many absents as postals still to throw, I don't see any realistic room for doubt that Georganas will recover Hindmarsh, and I've moved it below the line.


  1. A bit of boasting here

    In 1996, I made a bit of a name for myself by publishing in the AFR a perfect prediction of the Howard majority, based on a method of inter-poll and time-series averaging, which I don't think anyone had used before. Everybody uses it now. I have continued to play with the technique in the subsequent 20 years, but have never come anywhere as close as I did in 1996.

    But perhaps this year ... ?

    This is what my predictor was telling me on Friday night:

    69 75 1 5

  2. Do you think if The Greens and Labor are within 100 votes of each other on second and third place in Melbourne Ports there will be a recount?

    1. I would think that would be accepted as a valid ground for at least a partial recount, yes.

  3. I noticed in Hindmarsh the Ascot Park booth hasn't had a TCP count announced yet, and Labor outpolled Lib by 87 votes there, so this should increase their lead after it is added.

  4. FWIW, Jones was the incumbent in Herbert in 2013 as well as 2016. So the inbuilt effects of a sitting members with postals should already exist.

  5. Hi Kevin,
    A question about postal votes, upon which so many close contests now depend: we hear repeatedly that (in general) the Labor Party are usually errant in organising their postal votes and this will hurt them. In fact I can remember hearing this same remark made in many past elections too about the ALP and postal votes. Just wondering why the hell the ALP hasn't sorted this out by now. Is it something too expensive for a non-incumbent to afford? Or is this just an urban myth?

    1. The perception is that Labor chooses to direct resources a bit more towards other ways of getting votes simply because they are more likely to be effective for them, per unit of cost/effort. I haven't looked at comparisons with 2013 yet in terms of postal vote effort.

    2. I thought it was also that older voters (pensioners) favour the coalition (slightly) and are more likely to vote postal.

    3. I seem to recall some post 2013 analysis which showed that the Labor Party was actually shooting itself in the foot by doing postals, i.e. recruiting a higher proportion of non-ALP voters than on polling day, pre-polls etc... although its not commented in the actual report:

      The argument probably runs that if the ALP is not organising postal votes, possibly more Coalition voters do not exercise their ballot as they would then need to take advantage of the Coalition postal effort, organise their own postal via the commission or pre-poll or turn up on the day. I believe that this has been a well noted trend for some time. Even if deliberately not running a postal campaign does not reduce the actual Coalition vote its quite arguable that its pushing a cost on to the Coalition and therefore freeing up valuable resources for other activities.

  6. Whether or not the end result is a very weak Coalition majority of a minority, do you think the Coalition are toast at the next federal election? Or do you think they can get a new leader with enough authority for them to get back into majority government?

    Howard in 2001 was a very special case where both Tampa and 9/11 attacks occurred.

  7. Howard in 2001 was on an upwards poll swing even before Tampa and 9/11. It's quite plausible, in my view likely even, that he would have retained office anyway.

    Looking at the current scenario it's extremely hard to see how it ends well for them down the track. If it's a minority government then I think its chances of even making the next election without falling on the floor of the House are weaker than they were for Gillard's government.

    1. The Coalition were polling atrociously in the first half of 2001. As the election neared there was always going to be natural movement back to the government they were on 40/60 tpp. The turning point was thought to be the Aston by-election but even then Labor got a very big swing and Coalition only held it barely. I believe that they would not have been re-elected if Tampa and the 9/11 attacks not BOTH occurred.

      A borderline minority government with 75 seats in the House cannot fall on the floor as there would never be enough votes for a no confidence motion to stand. However, it might occur with a defection in the duration of this term.

    2. This is very true - Tampa produced a blip, but it bounced back. We cynics said it was de-indexing the petrol price that did it.

    3. Unless one of the Liberals become the speaker of the house

    4. In a deadlocked parliament with the government providing the Speaker my understanding is that a no-confidence motion might pass if put to the vote, but the Opposition would not have the numbers to pass an SSO to bring it to the floor. However the government still needs by convention to maintain control of the House sufficient to be passing legislation and would presumably resign if defeated on the Budget or something else important. (The Gillard government lost some votes but didn't resign as they were not core government business.)

      I suspect that if the Coalition had 75 seats would secure the agreement of at least one crossbencher for support on confidence and supply and would survive. But with anything below 75 seats I think there could be problems.

  8. Final booth (Ascot Park) has dropped into 2PP in Hindmarsh, increasing lead to ALP by some 200 votes to 639 votes.

    As a question on your analysis, with large postal vote numbers 10-15k in many seats (and a strong LNP postal campaign - often 3-5k more than ALP) do you expect the postal count to be even greater than in prior years.

  9. I checked on the number of own division postal envelopes issued and there seems to be an average increase of 1600 per division, but with huge variation. Especially high on the increase list for whatever reason are Queensland and Victorian marginals including a number of key seats.

    1. Queensland and Victoria are in the middle of school holidays, so that's a probable cause.

    2. Thankyou! That is enormously helpful.

  10. Kevin, great analysis as usual. Just one question: a key part of your commentary is the assumption that the distribution of postcount votes in a seat is correlated with the postcount distribution in that seat in the previous election.

    Your Flynn example makes me think this correlation may be quite high, but to give us a better idea of how strong this correlation is, would you be able to provide a graph of the 2pp postcount differences in each seat from 2013 vs 2010?
    i.e. the data in the graph 2/3 of the way down this post, plotted against the corresponding data from 2013. (That is an interesting and relevant graph in itself btw)

    Of course this may not be particularly useful this time if there is a fundamental shift in the postcount vote as some commentators have suggested.

    1. That sounds like a fun exercise and I've worked out how to do it. Not sure if it will happen tonight or not!

    2. Thanks for the very quick response! It's interesting how strong the relationship is, although we will have to wait and see how well it holds up for 2016.

  11. Kevin, there's a lot of attention in the media currently about how the 'Mediscare' campaign had a massive effect on voters particularly in the final few days before the election.

    As far as I can remember, your poll tracker and others had Labor close to around 51%PP maybe 5 or 6 weeks from the election (there was a 52 Reachtel and a 51 Newspoll around this time I believe). Polls were even for a few weeks (as I recall there wasn't a whole lot of movement post-Brexit even). Labor started running Medicare heavily around 2 weeks prior to the election which to me seemed to coincide with their worst polling of the entire campaign. Coalition 2PP got out to high 50s or 51 and I believe you're predicting around 50.4% when all's said and done. So perhaps there was movement of a few tenths of a percent in the last few days.

    It seems to me that a lot of the media are trying to justify some fairly terrible bandwagoning for the Coalition by pretending there was a big swing late. Am I way off base here?

    1. The best Labor got on my aggregate as an end-of-week reading was 50.3.

      In terms of late swing, it is possible even that the aggregates just overestimated the Coalition 2PP slightly for whatever reasons and that there was no late swing to speak of. In my own, I suspect that herding by polls that had previously skewed to Labor in the last few weeks of the campaign might have caused most of the error. That said I would not be that surprised if Turnbull's silly bulk-billing rate response right at the end did some damage. Possibly the real issue is not whether there was a late swing but whether a late swing to the Coalition should have happened and did not.

      Projecting the final 2PP is a bit rubbery so maybe it will end up lower than I predict for some reason, in which case there might be more to talk about in the late swing department. But yes it was quite interesting to see the same papers that ran pieces on how Labor's campaign had fallen over before the election, now run pieces about how brilliant it was.

    2. Thanks for your insight, Kevin. I am following the count with great interest :)

      How quickly will the count proceed on Tuesday (today)? Should I be spamming refresh on the AEC website from 8AM?

    3. The AEC website has been remarkably keen to tell me it now auto-refreshes every five minutes!

  12. Hey Kevin,
    Great analysis, you really put a tremendous amount of effort into everything!
    I've noticed most projections generally have the Coalition performing better on pre-poll votes compared to Labor. Is this a trend usually seen over past elections, or simply in comparison to 2013?

    The 2013 campaign went atrociously for Labor. With the Rudd v.2 honeymoon bounce rapidly fading during the campaign and infighting/miscommunication between the PM and party HQ, not to mention the overall abysmal result that year.

    My point is, I understand why the Coalition will do better on postals as that's a well established trend, but is there much reason to expect pre-polls to behave the same way in 2016 as 2013?

  13. The trend of the Coalition performing better on the total of all non-ordinary votes is a standard one. These include postal, out-of-electorate pre-poll, provisional and absent votes.

    Within electorate pre-polls (the majority) are counted on the night.

    Regarding prepolls there was a decline in Labor support through the campaign and it's possible therefore that the Coalition's performance on out-of-electorate prepolls will be marginally better relative to its vote on the night than in 2013. However, firstly there will be very little in it and secondly my long experience is that when we go looking for these kinds of trails of campaign shifts in voting intention in the prepoll vote, they are lost in the noise of other factors.

    Projections of results for prepolls, postals and so on already factor in the assumption that the swing seen on the night will be repeated in the prepoll voting. So for instance if Labor scored 47% of ordinary votes but 40% of postal votes in a seat in 2013, and scored 51% of ordinary votes this year, then the projection will expect Labor to get 44% of postals this time.

  14. Hindmarsh
    Another strong flow to LNP with more postals counted.
    ALP Lead down to 167

    LNP has picked up 478 votes on 66.6% of postals.
    By trend thats another 239 on the remainder, which would put LNP into a slight lead

    Before the Absent, Provisional and Declaration - some 6500 votes.

    So this one is knife edge, but a very slight lead to LNP