Saturday, October 20, 2018

Wentworth Live: Majority On The Line Again (Plus Post-Count)

WENTWORTH (Lib vs ALP 17.8%)
Dave Sharma (Lib) vs Kerryn Phelps (IND) (16 candidates total)

GAIN by Phelps (IND) - margin will exceed 51:49

Government to lose majority and seat held almost continuously since Federation.

The swings involved, while among the largest, are not an all-time record of any kind. (Not even if you discount Wills 1992)


Live Comments

Saturday Ah Wentworth, is this still a thing?  The final trickle of postals boosted Phelps' lead to 1851 (51.22%) - this may change trivially during further processing. The result will be declared on Monday.   It will still be subject to any Section 44 challenge that might occur.

Monday Remaining postals not exactly flooding in with the AEC reporting there are still not enough to undertake a count today.  As far as possible the AEC avoids counting very small batches at a time because of the risks to the secret ballot (voters might be identifiable from their votes).

Thursday 5:30 Phelps wins another batch of postals and leads by 1783.  The share of postals to Sharma was over 20 points weaker than in the opening postal counting.  I have never seen a turnaround on this scale and it is not even the final week yet.

Thursday 3:15 Phelps wins declaration prepolls 189-112 and leads by 1720.

Wednesday 5:30 Provisionals broke 58.4% to Phelps (246-175) so her lead expands slightly again.  I don't expect the margin (currently at 1643) to change by more than a few hundred from here.

Wednesday 2:00 Another batch of just over 500 postals has been counted and Phelps (narrowly) won them, though I don't have the exact figures.  I have mentioned before that late postals tend to split more weakly for conservatives but here we are seeing a rate of change from early to late postals that I've never seen before.  It is unclear whether Sharma will make further gains from here at all.

Tuesday 2:00 A very weak split on those postals which I have at 434-360 with 40 informal, around 54.7% to Sharma.  Phelps now leads by 1552.  At the current rate of informality, even if every remaining postal arrived, Sharma would need a 66% split, compared to the 62.6% rate so far. If 1000 don't arrive he needs 70-30, if 2000 don't arrive he needs 77-23.

Tuesday 1:00 Antony Green tweets reports from scrutineers that the flow of these postals is weaker than previous ones and they might break as weakly as 54-46 to Sharma.  With the issues the Liberals had later in the campaign it's possible the split between earlier and later postals will be stronger than normal here in which case Phelps could still win by 1000 or more.   There has been a theory that late postals from Israel might be strong for Sharma so that might be a little optimistic but we'll see.

Tuesday: Today the AEC will count a further 845 postals.  Meanwhile, further showing up the laziness of many journalists in declaring "record swings" without checking the facts, Peter Brent has tweeted a list of the largest pseudo-swings at general elections or by-elections.  The current by-election doesn't even make the list.  The top few entries involve Lang Labor MPs returning to actual Labor.  The fourth is also artificial because the UAP ran the only two candidates in Wentworth 1931 thus achieving a 100% 2CP.  Excluding these, the highest swing against a sitting government might be Angas 1931 with a 31.5% pseudo-swing against Labor because they were beaten by their own incumbent who had split from the party.


Monday midnight: Have been busy with other things but it's worth noting every booth has now been through some level of rechecking (including Rose Bay PPVC), which increases the confidence about there not being any major surprises out there.

5:34 The Paddington PPVC change has now reverted so not sure what is going on there.  Might be good to check back tomorrow on this one as sometimes AEC processes result in the online display reverting to earlier figures.

5:00 Now 1542.  If Sharma can close by a few hundred more on rechecking things will get dicey again, but so far he has gained only 71 since yesterday's postals, and quite a lot of booths will be rechecked.

4:50 Routine rechecking through the day has resulted in small changes in a number of booths.  There was one significant change with Sharma gaining 50 and Phelps losing 51 as in the Paddington PPVC booth.  Looks like a bundle of 50 in the wrong pile.  The margin has come down to 1564  but Sharma needs bigger or more gifts than that.

11:15 The Vaucluse preference flow was also incorrect so Phelps' lead increases by 35 votes there. Phelps now leads by 1651.  I think she will now win by several hundred rather than a few hundred.

Monday 9:35 Reports that the AEC will not count further postals today and will be rechecking already counted booths, including Vaucluse.

Sunday Night: The Wills Objection 

As noted below the result is not a by-election record in terms of 2PP swing against a sitting government, or primary vote swing against a sitting government.  But I am aware of sneaky attempts by ABC operatives to reclaim the 2CP "pseudo-swing" record on the grounds that Wills 1992 was voided by the High Court sitting as the Court of Disputed Returns, raising the question of whether you can have a swing to a result that was overturned (and not even rerun).

That's a nice technical point but even in that case Wentworth 2018 still loses out to the 20.4% pseudo-swing in Wentworth 1956.  (Lib vs Ind to Lib vs a different Ind).  And that's only by-elections; I am sure there have been bigger pseudo-swings in general election seats, though I haven't checked. [Update: Reader Mark Nettle correctly notes that Boothby 1911 (20.9%) is an even larger pseudo-swing.]

10:25: No significant changes; Phelps leads by 1616.  It seems that the scare caused by the Rose Bay PPVC has fizzled since the 2CP numbers at the time of it arising were not actually as close as shown by the AEC.  At 2 am on Twitter I wrote: "When you say someone’s got no realistic chance, there’s still a realistic chance they will seem to have a realistic chance at some stage."  At the moment that looks like the story - provided there are no more errors!

Sunday 7:15 pm: A further 1212 postals have been counted breaking 688-460 with 64 informals.  That's only a 60-40 split; Phelps now leads by 1613.  Allowing for informals the mathematical maximum of remaining formal postals is about 5650, but it won't be that many in reality.  The two early "miscounts" were simply cases of misrecording of numbers, easily done.

Sunday 5:45: Antony Green tweets scrutineer report that discrepancy in Bondi Beach is larger than estimated. The claim is that Phelps' lead will increase by 689 which if true would make it enormously difficult for Sharma without further errors. [EDIT: Confirmed and Phelps is 1862 ahead.]

Sunday 4:34: Bellevue Hill was an error and is fixed. Phelps' margin increased by 287 as a result of that (plus another eight votes from somewhere), still no word on Bondi Beach.  Word however of Sharma getting c. 60% of primaries in the current postal count, very slightly higher than his score to date.  Phelps lead = 1177.

Sunday 2:12: The AEC has noted four further booths being rechecked.  These are Double Bay, Paddington Central, Darling Point and Bronte (two won by each candidate).  I am not aware of anything suspicious in these booths.

Sunday 1:52: Antony Green relates that the suspect booths, and also Vaucluse, are being rechecked today.  This is excellent work by the AEC to prioritise these booths.  (Vaucluse is also an outlier in preference flow terms but this could be caused by small sample size as it has only 84 preferences.)

Sunday 1:30: Laura Jayes has reported that Labor scrutineers believe Greens' preferences were mis-tallied in the two booths in my naughty corner.

Sunday 12:20: Two Special Hospital booths were added, breaking 267-56 to Sharma and cutting Phelps' AEC lead to 884 (which I think is more like 1700).  Some other small categories of votes I haven't mentioned are provisionals (which will probably favour Phelps, though there may be 200 or less of those) and out-of-electorate prepolls (214).  Even assuming the errors I've identified are real and corrected, Sharma would only need a slight lift in his postal performance or some of these small parcels to go his way to still win, though it's more likely he would fall just short.

Sunday 9:20: 5463 postals were counted, breaking 3356-1858 with 249 informal.  There are now a maximum of 7164 postals to go, but on past form it may be only 5000 or so arrive.  The AEC has Phelps' lead now at 1092 but if my calculations on apparent booth errors are correct it may actually be around 1992.  In one case Sharma would just need to keep getting what he is getting and he would just win (note that late postals are often softer than early ones), but if I'm right about the error he would be needing about a 70-30 split on the remainder.  However there is still another unknown, which is how many postals might come back.  If there were, say, 6200 formal postals left then Sharma would only need 66%, which is still a bit much but would only take small errors in his favour elsewhere to be achievable.

Note that although Phelps' primary in the postals was only 17%, she still got 72% of preferences in them - again consistent with my view that the counts in Bellevue Hill and Bondi Beach have errors.

Phelps' scrutineers apparently expect there to be around 11K postals total, which would mean about 5500 to go (not all formal).  They've also said the counting was a mess.

Once again I should underline the potential for more counting errors with an example - in Indi in 2013 a batch of 1000 uncounted votes for Cathy McGowan was found on the Wednesday after counting.

Sunday: I am trying to catch up on missed sleep so there won't be much here for a few hours.  A summary of the situation: Sharma has bounced back into what looks like a competitive (if difficult) position courtesy of a stellar performance at the Rose Bay PPVC booth.  Assuming the figures in this booth are all correct, the problem with modelling that led to the seat being called very early last night was an expectation that prepolls would behave relatively similarly to booth votes in terms of the swing.  There are two reasons this particular large PPVC might not have done so, namely the dynamics of the race and the geography of the prepolls, which is not the same as in 2016.  This large prepoll seems to have sucked Liberal votes out of the surrounding booths and hence made Phelps' position look stronger than it actually is during the booth count.

However Phelps' 2CP position is likely to be stronger than in the official count because of apparent uncorrected errors at two booths (these things happen).  If these are the only errors they will push Phelps to a lead that requires some heroic assumptions about postals for Sharma to overturn.  However, if there are errors in Sharma's favour, the gap might still be catchable.

So while it still looks most likely Phelps will win, we need to wait for further counting to be sure of it.


3:00 I've had a look at booth results vs 2CPs in search of any possible errors.  A couple of booths look really odd (see graph):

* Bellevue Hill (2290 votes)where Sharma polled 45%, Phelps polled 31.9% and 49.0% of preferences went to Phelps.

* Bondi Beach (3846 votes) where Sharma polled 28.9%, Phelps polled 39.0% and 57.4% of preferences went to Phelps.

There was nothing unusual about these booths except that the Greens did unusually well at Bondi Beach.  Yet they are among Phelps' worst preference shares (supposedly).  They are massive outliers.

The numbers would be on trend if Phelps was actually c.150 votes better in Bellevue Hill on the 2CP and c.300 votes better on Bondi Beach, which would increase her lead by c. 900 (52.58% 2CP not 51.92%), and leave Sharma needing c.69% of postals (which is very unrealistic).  Unless, of course, there are errors in the other direction somewhere else!

(The other two outliers below the line are a BLV booth with negligible votes, and Vaucluse with only 84 preferences, so ignore those.)

2:00 I have had a look at claims that the Liberals did much better on the swing on postals than on other votes in recent by-elections but I don't find that to be consistently the case.  Slightly better in Bennelong, worse in Longman, slightly worse in Braddon.  However Mayo is interesting because there the overall 2CP for Rebekha Sharkie increased by 2.6 points but the 2PP on postals for the Liberals increased by 5.1.  If that disparity was repeated, Sharma would just win.

12:31 Not all the outstanding postals will be received and accepted (some people apply for a postal and then prepoll vote or vote at a booth anyway) and in the Bennelong by-election case 77% of postals arrived.  If that repeats we'll be getting around 9700 here.  A few percent may be informal - say 3% as the informal rate so far has been low (but will probably increase on checking).  Say 9400 postals to go (plus a few hundred prepolls, provisionals etc that will probably not do a lot.) This would require Sharma to beat Phelps about 63.8-36.2 on postals to win.  If we compare the non-postals the pseudo-swing so far sits at 18.5%.  Sharma would need to keep that swing on postals to 12.8%.  That isn't unachievable.

11:50 We have a bit of a late flare-up here! Waverley PPVC closes the 2PP up to 51.92%.  Indies do perform badly on postals and Phelps got up a head of steam late in the campaign (plus the last week was a shocker for the Liberals).  There are also no absent votes in by-elections.  So, what we have in the can is a lead for Phelps of 2,590.  There are a maximum 12,549 postals to come.  If all were received and formal Sharma would need 60% to win.  I'm checking whether we've actually all called this thing too early! (That would be an industry-wide embarrassment.)

There is a fair amount of suspicion about the Rose Bay PPVC having recorded extremely strong results for the Liberals so that should be treated with caution.  However, it wasn't a booth last time, and assumptions that it would behave like other Rose Bay booths are not necessarily sound.

11:30 I've been a bit distracted waiting for the PPVCs to come in, but they're getting there (slowly).  All are in on the primary vote with Sharma now on 41.74, Phelps on 30.33.  A big hit on one of the prepolls knocked Phelps down to 52.08%.

10:15 Graeme Orr has just noted that in the Wills case there was a swing large enough back to Labor at the general election to easily account for Phelps' margin over Sharma this time - and with the same Labor candidate.

9:55 Paddington is the first PPVC in; unhelpfully there wasn't one by that name at the general election, though it is the same as one of the booths.  Phelps polled 29% there compared to almost 39% at Paddington booths while Sharma polled 36.7% compared with 31.3%.  Murray polled 16.5% in the prepoll compared to 11.8% at the booths.

9:20 A couple of booths still have 2PPs to go, apart from that the next act tonight will be the counting of the PPVC booths, which will give us a better idea of the final margin.

8:57 Some media (in the sin-bin so far Phillip Coorey and someone at Sky News) have been making false claims about this result being an all-time record.   So, let's spell these out:

* The biggest 2PP swing against a sitting government in a by-election was 20.1% (Wakefield, 1938).  To beat that, on current numbers, Tim Murray would need 83% of preferences.  This number will increase.  Not happening.
* There was a 24.1% swing against the sitting government on primaries in the same by-election.  Currently, there is a 23.2% swing against the government in this one.  That number is projecting to go down, so that record won't be broken either.
* Considering 2CP swings irrespective of changes in opponent, the mark set in Wills 1992 will only fall if Kerryn Phelps gets above 55.87% 2CP.  She is currently on 54.65% and will probably fall back on postals and prepolls.  So that one's probably not falling either.

8:45 Phelps has come down to 54.3% so I think the Wills mark is pretty safe.

8:27 Phelps is now up to 57.0% on counted booths, so it seems she is doing well in larger booths.  This should come down in the post-count.  Sharma is projected to come up 2% on the primary vote - the pseudo-swing could threaten the Wills 1992 mark after all!

8:18 Sharma is improving on the projection, he is probably going to break 40%.  But Phelps is much too close to him.

8:06 My 2PP tracker for Phelps is now at 55.3 but that ignores any impact on her vote (not Sharma's) of postals and prepolls, so I'm expecting it to come down in the postcount.  The margin won't be as huge as the more outlying polls and might get back to 54-46 or so.  Hardly anyone will care.

7:56 On current numbers Labor, who have been gutted by strategic voting, need 74% of preferences to win the 2PP.  I reckon that probably won't happen so we can put away calculations about the biggest 2PP swing in history. (Wakefield 1938 takes the cake for by-elections there, I am unsure about general elections).  However there is a live question of the "pseudo-swing" record.  To beat Wills 1992 (which I am unsure is the record there anyway) Phelps needs 55.87%.

7:47 I'm currently tracking for about 56:44 to Phelps; this would probably drop a bit after postals and prepolls.  This is based on a very linear relationship (99% variation explained!) of the primary gap between Sharma and Phelps and the booth 2PP, with 8 data points.  This matches the current tally on the AEC site off 8 booths.

7:33 Indies tend to do poorly on postals and prepolls so some of the size of Phelps' win is likely to drop back on late counting.  At the moment I don't have a 2CP by booth to project the size of her win, but I'm looking for one.

7:25 Phelps has hit the lead on the 2CP vote based on counted booths.  I haven't attempted to project her 2CP margin but it will be massive.  High 50s at least at this rate.  Also, there will be no mathematical uncertainty tonight (dammit) as Phelps and Sharma certainly are the top two on current figures.

7:17 I see no reason on earth not to call this.

7:16 Sharma has come up on the AEC tracking to 37.5% but it is nowhere near enough.  The seat is not just being lost on current numbers but being lost heavily.

7:15 Phelps tracking for primary 33 off five booths.

7:13 I have done a quick regression albeit based on only four comparable booths and have Phelps tracking for a primary of around 35% off those, which would put her little behind Sharma on primaries.

7:07 What we might call the pseudo-swing (the swing from Liberal vs Labor 2016 to Liberal vs Phelps 2018) was at around 27% in the first booth; Phelps only needs 17.8%.  This is looking like an early night on these numbers.

7:03 Phelps had over 80% of preferences at Double Bay East, only just losing the booth on 2CP.  Two more booths are added and Sharma's tracking is back into the mid-30s while Phelps' total after four booths is already at 35%.  At the current rate Phelps is winning massively.  Also Labor is down 4% so there is no Labor vs Phelps issue for third.

6:52 Another booth in, Vaucluse, where the Liberals polled 86% primary last time.  Here the damage has been limited and Phelps has dropped back, so we will need to keep an eye on how much the Phelps vote varies by booth.  This booth is very unrepresentative!  Sharma tracking for high 30s after two votes, let's see how that holds up later.  Labor tracking to about match their 17% from last time.

6:46  A TAI exit poll has been reported but I can't find the numbers yet, just stuff blaming the result on the stuff you'd expect TAI to blame it on.

6:40 First numbers in, Double Bay East where Sharma is down 27.3% according to AEC which is a bad start (overall primary for Liberals was 62.3 so that would put him in mid 30s) and Phelps off and running with a 29.5.  So Phelps has ticked the first box.  Swing for Labor is -2.5 (they got 17.3 last time).  Round 1 to Phelps.  However Double Bay East was not a booth in 2016, unless under a different name, so I am not sure what the AEC is projecting off.

6:00 Polls have closed, awaiting figures.  Haven't seen any exit poll reports yet.  The AEC will be counting the 2PP as Sharma vs Phelps, meaning we will not get exact 2PP swings tonight, only primary swings.  However we can still look at 2PP "swings" by booth to see if the target is being hit or not once booth figures start going up.

I'll be a little slow on comments til about 6:30.

3:00 Bevan Shields (Fairfax): " Several Liberal volunteers manning the booths say the situation is terrible and most voters aren’t interested in taking Liberal how-to-vote material "


Welcome to my very basic live comments thread for the Wentworth by-election in which the Liberal Party attempts to defend one of the jewels in its Federation crown, and appears likely (based on spotty polling but perhaps more tellingly a general impression of chaos piled on desperation) to fall short.  See my guide to the seat here (including accounts of the 11 polling results released.) History in the making, potentially, but we have not even been blessed with a Newspoll.  For those watching betting odds, Kerryn Phelps (IND) is now at $1.32, about the same odds John Alexander won Bennelong at, for what that's worth.

As usual my aim will be to fill in the gaps left by the mainstream coverage, but I suggest you also keep an eye on Poll Bludger which is offering much  greater snazziness than I am capable of.The Tally Room will also have live commentary.

Comments will start here from about 6 pm and will scroll with most recent at the top; once the count gets going refresh every 10 minutes for the latest.  At some point I will probably take a short dinner break.

If this seat falls, the Morrison government loses its majority (which is somewhat debatable anyway in view of the "government crossbencher" status of Kevin Hogan).  It would be unlikely to collapse for that reason alone as some crossbenchers will provide confidence and supply, and if things got really grim the Speaker could resign to give the government an extra vote.  There has been some thought that if defeated in Wentworth the government might go to an early election in, say, December, though it would be more likely to be smashed badly in such than if it waited.

Things to watch tonight and in following days:

* The AEC will do a two-candidate preferred count tonight, but we don't know yet if it will do Sharma vs Murray (2PP) or Sharma vs Phelps.  It is possible it will pick the wrong pair of candidates.  If it does this and it is clear on the night that it has the wrong pair, then the count will have to be re-aligned to the most likely pair.  This would take a few days, and during this process the two-candidate vote can sway around wildly based on the order of booths that are aligned.  If this happens I will be covering it clearly on this page in the post-count section, as past realignments (eg Denison 2010) have caused much confusion and misreporting.

* A strong possibility is that the final two candidates will not be clear tonight.  For instance if one of Murray and Phelps leads the other on primaries by anything less than double figures, it may not be known who the final two are for up to two weeks.  As the number of postals reduces the AEC might be able to do a 3CP count to establish which of the top three has finished third. but it could also be that the spread of votes for minor candidates means not even this is possible and we have to wait for the final distribution of preferences.  However, if there is a snowball of anti-Liberal votes to Phelps, it could be clear enough tonight who the final two are.

Enrolment in Wentworth is 103,810.  I expect a high turnout because of the circumstances and the range of parties contesting (some comparisons: 86% for Bennelong, 84.3% Longman, 90.4% Braddon which was the highest for a by-election since Aston 2001).  However with 16 candidates I expect a high informality rate, maybe around 7%, on account of the absurdly strict formality rules.

Around 12,549 postal votes were sent out (excluding those withdrawn) and there have been 18,713 prepolls.  The big prepoll centres are Waverley (7107), Rose Bay (6780), Paddington (3427) and Haymarket (1142).  The remainder are mostly out-of-electorate.  The AEC postal vote stats do not appear to provide the usual breakdown concerning which postals were "party postals".


  1. Hi Kevin,

    I have tried and mostly failed to explain the strategic implications here to a few people this week.

    If it turns out the AEC are right to select Sharma v Phelps 2PP will we still get a clear idea of what would have happened if Phelps dropped to 3rd and was cut up in a Sharma v Murray and if so when?

    Beers rest on it!

    1. If Sharma vs Phelps is correct we won't know the 2PP for quite a while.

  2. But someone (scrutineers?) still count where her preferences would have gone if she had have been eliminated?

    1. The AEC will do a 2PP count for information but I would expect that to take several weeks now. I expect Sharma will win it.

    2. very rough 2pp 60% Phelps to labour 20 plus alp 11 plus 80% of 13...11 plus 2% of remainder say 43 alp 57 lib

  3. Hi Kevin,

    "Pseudo-swing" record probably Lyne 2008 (32.4%). Other contenders are Groom 1988 (Lib v Nat, 25.2) and Echuca 1919 (VFU v Nationalist), which didn't go to preferences but if it did was probably around 33%.

    1. Yes, Peter Brent just posted a list with Lyne and Groom at the top, which were both swings against Oppositions (Groom being a rather weird example.)

  4. scrutineers are only needed for contested votes, where the ballot isn’t clear. All of the counting including preferences is done by AEC staff.

  5. Re Rose Bay PPCV ... pollbludger has a swing for that, must be based on something from the past. Rose Bay PPVC Lib FP swing is -2% (next best is somewhere near -16%) but all other swings in that prepoll center are in line with the rest of the electorate. I could be that a stack of bundles (1,000 FP votes) for Phelps got into the LIB register … if you take 1,000 votes out, the FP swing is -17% (back of envelope calc). That’s still “less bad” for Lib than all other PPVC. These things happen and you wouldnt notice it doing the 2PP count, but it will show up in the Sunday check count

    1. Perhaps it's based off the prepoll average. It isn't wildly off my trendline above so I'm still not sure it is an error.

    2. Well if the base-line is off (just for the Libs), then there's no point looking at the swing. Makes Rose Bay PPVC the top FP, and second TCP for Libs (excluding hospitals). Be interesting to see where those votes came from (and thus why the baseline was off) ... there's bound to be more new prevoting centres in the future and understanding how they draw voters from polling places might be interesting to see

    3. Yes, we're going to need more sophisticated ways of modelling new medium and large sized prepoll booths that haven't been used before. Assuming them to match the electorate prepoll average doesn't cut it.

  6. Any idea what happens if a major bookmaker pays out a winning bet but the result then changes?!

    1. They lose their money but get a lot of free publicity that sucks in more new punters.

  7. What did you make of Barrie Cassidy's largest swing defence on Insiders (i.e. that it didn't count if the candidate wasn't found to be eligible)?

    1. I'll think about that. Maybe see if it is still the largest discounting Wills when the dust settles to see if we should worry about it. :)

  8. Your early prediction of the 2PP allocation errors will certainly enhance your pseph credentials Kevin. Good call.

  9. Obviously a super low sample size but do you think the low preference percentages in vaucluse for Phelps was due to tactical voting by lib voters, voting labor first then preferencing libs (only booth to see a higher labor vote, coming from the lowest possible base though)

  10. I think the most likely explanation for the swing to Labor in Vaucluse is a bit of reversion from an extremely low base (and swing against) at the previous election.

  11. "I think she will now win by several hundred rather than a few hundred"

    Which is bigger? ;)

  12. Have they recounted Rose Bay prepoll ..looks too good for the libs

    1. What checking of names checked off/ballot papers issued vs ballot papers counted is done, and when are these results reported?
      I would assume that some preliminary looking has already occurred.

    2. That's an interesting question. I am not sure if markoff numbers are reported. There is often a slight discrepancy between markoff and votes cast based on some voters actually not putting their ballot in the box. There can also be clerical error in markoff (eg apparent double voting that is often a result of mistakes - I once got a please-explain over being marked off twice for this reason.) I'd expect there would be some kind of check to avoid severe mismatches though in the Indi case it took til around now for the missing 1000 votes to be cast, and in Fairfax 2013 there were discrepancies between House and Senate numbers as displayed for over a week.

  13. Wouldn’t Boothby 1911 count as a higher swing against a sitting government in a by-election. 20.9%,_1911

  14. Yes that's a slightly larger "pseudo-swing" than Wentworth 1956, well spotted.

  15. TCP vs Primary chart
    This is an interesting “new” tool that should prove useful in future. I wish I’d thought of that when wresting with the mysteries of Bellevue Hill on Saturday Night.
    The outliers have gone of course
    It was fascinating to do the same for Sharma – a P<0.05 straight line with an R-sq of 0.96.
    Not sure what to make of it – possibly that Sharma-preferencers are convinced, whereas the Phelps-preferencers are a bit more wobbly

    1. The issue would have been caught earlier in the night had I kept running the regression I was running earlier, but I stopped doing data entry for it when it settled down. It also had a very strong r-squared. I've used the same methods to spot outliers before but I think they were in counts that were much less high profile than this one.

  16. If Postals total 77% of issued (average of the 8 most-recent by-elections) and all are formal, then Sharma needs 78% of the remaining postal TCPs. Antony reckons 70%

    Slice and dice as you will, there is hardly likely to be a big difference - the final result now seems foreordained.

    Phelps people seem very happy

    Everybody has packed up at Alexandria and the scrutineers have gone home. Tomorrow they will continue with postals and also start on dec votes.

    I am told that they won't start until noon

    Phelps has a team of six scrutineers - two experienced, four learning the ropes

  17. How do you see provisional/declaration prepoll breaking. In 2016, both of these were not as strongly Turnbull as the average for the electorate.

    I am more after thought processes than probably inconsequential for this by-election numbers.

    My back of envelope thinks Phelps gets an advantage of 20 provisional (prov done on the day, mostly invalid, less strong in 2016 than average),

    Also an advantage of 10 declaration prepoll votes to Phelps? (based on difference between 2016 postal and dec prepoll - almost being 50% each way, 500 being valid?)

    That is, for an election where these might matter, what do we look for next time around?

  18. Provisionals I expect to break to Phelps, possibly as much as 60-40 given the booth voting, as they generally break to the left. Declaration prepolls might be seen as a tricky one because of the dynamic of pre-day and on-the-day votes breaking differently. However 80% of the dec prepolls were cast in the final week including 40% on the final day, with the main contributing electorates being Canberra, Melbourne and Brisbane. I don't think they will do a lot either way.

  19. What's your take on the Green vote Kevin? Their first preference vote fell off a cliff compared to 2016 (down around 48%). Many commentators were calling this by-election a referendum on climate change so you would expect the Green vote to be high but it simply tanked to a level even lower than 2004.

    1. The Green vote usually crashes when they are up against independents who make noises about any of their pet issues. Phelps was very vocal about climate change during the campaign, Heath ditto, and I believe a lot of people who normally vote Greens voted directly for Phelps or Heath instead. It is possible also that infighting over Greens preferencing contributed to their result. Overall I think it is very similar to what has happened to the federal Green vote in Denison following the rise of Andrew Wilkie. There are a lot of voters who prefer the Greens to the major parties but still have reservations about the Greens and would prefer an independent who ticks the same boxes.

    2. As of Thursday 5PM, Sharma needs 99% (+1) of the likely remaining postals to win

      On mark-off (checking names against electoral roll, etc); this won't happen for weeks and weeks.

      However, # of ballots ISSUED is checked against # found in the ballot boxes. Everything is counted again to see if the discrepancy continues. If it does, the numbers are submitted as is, with a caveat. Three of these recounts occurred at Bellevue Hill on Saturday night - with three different answers - they were all a bit stressed and went on to make other errors.

  20. why have the postal changed so much.....did the liberal desperation lose them the seat?

    1. Sure looks like it! People can argue otherwise but I have never seen stronger evidence in actual voting patterns for a party blowing it in the final stage of a campaign. The polls don't support this (beyond that the Phelps vote clearly rose) but the polls seem to have all been junk anyway.

  21. "I have never seen a turnaround on this scale and it is not even the final week yet."
    I think the message that the coalition is trying to read from this is not the message that was sent.
    While dumping a PM hurt (call it 1/2 the swing), to me it looks like the remaining swing was last week issues. (betting odds, differences in early votes vs later votes)

    1. I think dumping Turnbull was well over half the "pseudo-swing" but that events of the final week pushed it over the line whereas otherwise the Liberals would have just held on.

      I ignore betting odds in terms of their ability to predict this sort of outcome. They got it right in the end but there is a long history of betting odds being slow to smell the coffee in third-party contests and sometimes missing it entirely - at other times prominent indies have been overvalued.