Saturday, July 28, 2018

Super Saturday By-Elections Live And Postcount

MAYO: CALLED Sharkie (CA) retain
PERTH: CALLED Gorman (ALP) is new MP, retaining seat

Braddon Swings

Here is a graph showing the relationship between the vote for the independent Craig Garland and the 2PP swing in Braddon:

You can see maps of the Garland vote and the 2PP swing over at The Tally Room.  The Liberals did very well on 2PP swing on the West Coast, were smashed on King Island and the west end of the coast (Wynyard - Stanley area where fishing is important) and got small swings in Ulverstone and Devonport-Latrobe.

For every primary vote Craig Garland gained in a booth, the 2PP swing to Labor went up by a third of a vote. Obviously Garland's preferences flowed very strongly to Labor (about 75% was a Labor scrutineer estimate; exact figures will appear in due course.) Had the voters who voted Garland and then preferenced Labor been voters who intended to preference Labor anyway, then there would not be the relationship in the graph.  Rather, it's likely that about a third of the Garland vote consisted of voters who preferenced Whiteley over Keay in 2016 and switched to voting for Craig Garland and preferencing Keay over Whiteley in 2018.

If none of those voters had switched their 2PP vote, Brett Whiteley would probably have won the seat.  But this doesn't mean that Garland's candidacy or preference recommendations caused all those voters to switch their 2PP vote, as some of them were probably switching anyway.  Of voters intending to change their 2PP vote, some will vote directly for the opposite major party, and some will vote 1 for a non-major candidate. It would be surprising if Garland's how-to-vote card was followed strongly enough to account for most of the shifting of preferences through him that we saw here.  (Especially as it had a low hand-out rate.) An under-studied aspect of preferential voting is the extent to which some voters might be open to switching their preference only if they can find somebody else who they like (rather than the other major party) to vote 1 for.  Some voters have class-based or habit-based resistance to "voting Labor" or "voting Liberal" that may not extend to preferencing.

Brett Whiteley has made some comments connecting his defeat to compulsory preferencing.  With a primary vote lead of only 2% he would also have lost under optional preferencing, though probably by a smaller margin.

By the way, it may seem that Braddon and Longman are the norm in terms of the levels of protest voting for non-major candidates.  In fact in the most recent by-elections contested by both major parties there was a combined swing to the major parties in Bennelong, Canning and Griffith.  We have to go back to Aston 2001 for the last case where the combined major party vote was down significantly, and in that case there was a large increase in the number of candidates.

Late Night Wrap

I was busy with other things for a while but just putting up some late overnight comments.  In later counting Labor has come up again to a projected 0.5% swing in Braddon; when the individual booth details go up I will have a look at those.  In Longman the projected swing is running at 3.6% to Labor.  The Braddon outcome is not surprising (especially with so many uncertainties) but Longman was not expected by anyone much.

So in the two seats Labor has an average swing of just over 2%.  This exceeds the average for Opposition seat by-elections, but it would have been expected to based on the fact that the incumbents were recontesting and the fact that Labor is ahead in national polling.  So Labor's performance in Braddon at least is below what would be expected based on those factors, and here the debate can be endless about how much is down to Justine Keay's seat being not referred earlier, how much is down to specifically Tasmanian campaign factors, but what is very difficult here is to see any evidence of an anti-Shorten factor at work.  If we try to frame it as a bad night for the Liberals, on the other hand, we run into the objection that they punched above the weight of history in Braddon, while Longman might be written off as caused by the Ruthenberg medals scandal.

This is, then, a good night for Labor, but I don't think it tells us much about how the parties are travelling.  The narrative, though, will not be the Coalition's friend, precisely because they were so hyped up by the media for these contests, and we may see that harm them in the polls.

Some other winners and losers deserve mention.  Firstly it is a brilliant result for Craig Garland who I did not expect to break double figures after what he has had flung at him in this campaign.  Moreover, he has put the issues he stands for well and truly on the radar and it looks like his preferences have destroyed what chances the Liberals had, although my read of the booth swings even where Garland did well is that they would have lost anyway.  It should be kept in mind that the Liberals "leaked" (um yeah) internal polling claiming they had successfully bombed his vote down from 9% to below 5%, when he actually ended up finishing above the high point of their tracking.

Secondly the Longman result is another shocker for seat polling, and another curious case (like Bass and Macarthur, also Lindsay at the federal election) where multiple seat polls expected a close result and they were wrong. The seat polls in Longman for Labor ran 48, 50, 49, 49, 49, 51 (in chronological order) and they're currently on for in excess of 54.  (Newspoll can claim some credit for tipping the right winners in both its final polls, which at least shows that a large sample polled as late as possible might count for something.)

If it stays like that, questions need asking about why we are getting cases like this - are pollsters herding to previous results, are their models hedging towards 50-50 to reduce the risk of massive misses, or are commercial imperatives encouraging close-contest biases?  Or are both ReachTEL and Galaxy just making the same mistakes as each other in some kinds of seats - but not in lopsided seats such as Mayo?  It's also notable that the pollsters were all at sea in measuring the Others vote (excluding Greens and One Nation) in Longman, which is currently travelling at 9.6%, more than double the average in polling.  Labor and One Nation did roughly as expected on primaries, but the LNP much worse and Others much better.

There is not much to see in Mayo save that once crossbenchers are established and liked by their electorate any attempt to shift them is simply delusional (cf attempts to dislodge Andrew Wilkie in Denison/Clark) and the no-contests out west are only of interest in that the Greens were almost as lacklustre as the ReachTEL polls said they would be.


10:05 The Greens have dropped below 20 in Perth now, so that's not going to be interesting as they'll need three-quarters of preferences.

9:38 Liberals are not quite totally conceding Braddon but they should.  They might get closer from here, but might not. In fact the AEC projected swing is now in Labor's favour (0.02%).

9:34 Back to say there's nothing to see in Fremantle whatsoever as a sniff of a contest.  Massive win for Wilson.

8:55 Scaling back commentary for the next hour while I write an article.  Will get more serious again after 10 by which time the pattern in WA should be clearer.

8:50 We have something a bit interesting in Perth - an 18% swing to the Greens in a booth in the absence of the Liberal candidate.  In Fremantle so far no swing to speak of and the story there which could take a long time to figure out is who is second.

8:32 The swing to Lamb is blowing out, this is looking like a big failure for seat polls which all had the outcome very close and mostly had the LNP winning.  After some similar seats in 2016 questions may need to be asked about whether pollsters are hedging by producing strings of close seat results in seats that then are not close.  But let's see where it finishes up.

7:59 I've been very cautious on Braddon because of what we've seen through the night but we now have 63% in with no booth swings (and probably 15% or so that will never show up).  So we would need an enormous difference in prepolls and postals (which are strong for the Liberals anyway) and that won't be happening.  I expect the count to finish with a small swing to Whiteley but it might not,  In Longman the swing is rising again and the swing against the LNP on primaries is approaching 10%.  This is a terrible result for the LNP and the seat polling has performed poorly in this seat.

7:48 Although he'll drop on postals and PPVCs Garland is looking like he's on for double figures.  He'll do very nicely off the public funding here.  The Shooters will probably get their deposit back but the Greens have fallen below 4% and may well not.

7:43 In Braddon we only have a few non-PPVC booths remaining for the primary count, one of those being Wynyard Central which is going to be Garland Central based on the state election so more damage coming for the Liberals.

7:40 Unlike in Braddon where the swing by booth is going to be all over the show (once those numbers come out) the swing in Longman seems very constant.  I am not seeing any sign of improvement in Ruthenberg's position through the night at all.  We still have only 13% counted to 2PP so would like to see more in this very slow count but it does look like Lamb will retain with a swing to her.

7:35 Labor swing going up again in Braddon with now 33% counted in 2PP terms.  The lack of progress in the recent booths added is not a good sign for Whiteley.  Also we have nearly all polling places in on primaries with no sign of a coming recovery.  It's getting towards miracle territory to win it from here for the Liberals.

7:29 The swing in Braddon is hanging around zero at the moment in counted booths.  Braddon is not a seat where postals tend to spring a big surprise so Whiteley will need to close by at least a point, probably a point and a half, by the end of the night.  I think it is still too early for a definitive call on the seat though it is very strong for Labor.

7:25 We are seeing further movement in the Braddon swing which has now come down to a swing of 0.25 points to Whiteley based on booths actually counted to 2PP.  The ABC has more or less (but not quite) called it but I would like to wait for more.  Garland has come down to 11% and will probably still come down a bit more but he's done very well.

7:24 CALLED Mayo Sharkie (CA) retain.

7:20 17% counted to 2PP in Braddon and still a small swing to Keay (+0.7%).  Whiteley needs to get rid of that swing and then get another 2.2% on top of it.

7:18  A modest 2PP swing in Mayo to Sharkie so far.

7:14 As Antony is pointing out in a normal election you would have lost interest in all of these by now.

7:08 Braddon - ALP scrutineers recording 75% of Garland preferences according to ABC.  We now have 6% in the can two-party preferred and we are seeing the swing to the ALP stabilise, not decline.  In Longman the 2PP swing to Lamb is 6% off the first four booths.  If I was one of the pollsters responsible for those borderline seat polls I would be very nervous right now!

7:05 The first booth in on 2PP in Longman has a 2PP swing to Lamb of 2.2%.

7:01 Booths in for Longman so far are very representative.  We'll need to see the preference flows but at the moment it is looking like the LNP will need about two-thirds of preferences, which if so is about a Category 1 from the Tour de France.

6:59 Still a swing to Keay with 3% of the vote in the 2PP can, now at about 2% for her.

6:55 Garland is on for a vote much higher than the polls or anyone including him expected.  I've had booth figures through with him on 13% at Havenview, Burnie, where he got 3% at the state election!

6:51 Longman is starting now and in some early booths Lamb's primary is up, and the LNP is down substantially.  Labor's primary should be up a little because of the decline of competitive Others but this is a good start for Lamb as the LNP primary so far looks low, tracking for about 30 off the early booths.  We will have to see some preference figures though.

6:49 Preference count available in 8 booths now and the early lead for Keay has quickly almost disappeared and is now at 1.63%.  I would still ignore this at this stage.

6:46 It will take a while for the 2PP to settle down from the impact of the Moorleah booth so ignore it for a little while. The AEC projection is now running at 4% to Keay but that should be ignored for now until these early booths with high Garland votes get swamped by others.  (Who knows, they may have high Garland votes too!)

6:39 Votes are pouring in from small polling places.  The AEC's primary totals are booth matched and with eight booths counted Garland is on 17% (which will go down, but is strong), Keay is down 3%, and Whiteley is down 8%, but the Shooters are at 7% (they will go down too).  Keay has a 9% 2PP swing in the one booth counted so far (Moorleah, but that is unrepresentative!)

6:33 The early votes coming in are only primaries so the ABC's estimates are projections only.  Unfortunately I don't have a feed of results by booth just yet but hoping to find one - usually the ABC has this online but I'm not seeing it yet.

6:25 We have the first booth in and it's Moorleah with a massive vote for Craig Garland. However Moorleah was Garland's best booth in the state election and he is only up 3.7 points on his state vote.  The majors are both down about eight points on 2016 in this very small booth.

6:10 Worth bearing in mind that Longman is a seat where postal voting very heavily favours the LNP, which closed by 1.08 points on late counting (votes counted not on the night) in 2016.  Also by-elections don't have absent votes to any significant degree, so in theory it could be larger, but Labor have made more of an effort on postals this time.  If you see Labor well ahead but me reluctant to call, that is why.

5:35 A little note re betting odds and polls.  On the weight of 51-49 results in the final Newspoll, Labor retook the betting lead in Longman today (finishing at 1.58/2.30 on Sportsbet) and the Liberals in Braddon blew out to $3.50 (vs 1.28).  This makes both seats relatively rare cases where betting odds and the average of all non-party campaign seat polls disagree as to the winner (though for Braddon, the early 54-46 to Whiteley is skewing the outcome there).  In 2016 there were five such cases by my count with the betting odds beating the polls in Gilmore and Grey and the other way round in Macquarie, Cowper and Cowan.  There was also one seat with a 50-50 seat poll average (Hindmarsh) where the betting was wrong, one seat with a 50-50 average where it was right (Dawson - though the only seat poll there was from GetUp! which was a biased source in that case), and nine cases where both were wrong together.  Of over 100 seats with no seat polls, the betting odds were correct in every single one.


Super Saturday is here at last!

Welcome to my on-the-day and postcount coverage of the Super Saturday by-elections for the seats of Braddon, Longman, Mayo, Perth and Fremantle.  I will be posting live comments tonight starting from 6 pm though I don't expect to post anything significant before the first booths come through from 6:30 onwards.  As usual I intend my comments to be something of a companion to the ABC coverage, picking up on issues it may have missed.  All times for comments will be in AEST.  When I think there is no significant doubt about the fate of a seat it will appear as CALLED at the top of the page; until then there will be summaries.

I will be focusing especially heavily on Braddon but also on Longman and to a lesser degree the others (unless something exciting happens in them!)  Comments will scroll to the top; once the action gets going refresh every 5-10 minutes to see if anything has been added.  Postcounting coverage will continue here for any seats still in doubt or snippets of interest, save that if there are two or more seats still in serious doubt I will probably give them each a new thread.  I also did leadup articles for Longman and Braddon.

My rough setting of the goalposts for Labor for Braddon and Longman (I like to do this as an advance measure for the post-election party spin that always happens):

* substantial (2+ points) swing to Labor in both seats: excellent
* both retained and both clear on the night: good
* both retained but one or both inside 51-49: relief, but shows Coalition is more competitive than it had seemed, suggests work needed for federal election
* one loss but swing to Labor in the other: very concerning, but might be explained away using local factors provided Labor keeps national 2PP polling lead
* one loss and one swing against: very bad
* double loss: catastrophe

For the Coalition these by-elections are a free hit unless there is a substantial swing to Labor.  If there is a substantial swing to Labor then this opens up questions about the Coalition's competitiveness at a general election, tactics and candidate choice.

I will try to remember to clear comments as they come in but I will probably not have time to reply to them.  Scrutineering tips can also be sent to me by email, on Twitter or by text on my phone if you have my phone number.  With the exception of The Mercury, journalists are asked not to call me between 6 pm and 10 pm AEST.

At a certain point tonight, yet to be determined when, I will need to take a break to write a commissioned article; this may take me an hour or so.

A few things to watch for:

* The Longman prepoll and postal vote is enormous, meaning that we'll need to wait for the big prepoll voting centres to be counted later in the night before the vote settles to a high percentage counted.  One of the Longman prepolls (Caboolture) had over 13000 votes last time.  Could be a long night!

* In Braddon, at the state election independent Craig Garland polled a very regionally concentrated vote - below 1% in the Devonport booths but as much as 20% in some small booths in the far north-west.  This concentration will be less of a factor at this election but will still be a factor, so if you see Garland polling something off the wall in the early small booths, it probably won't last.  I will be booth-matching the Garland vote off the state election to see how he is tracking.

* Some postals are expected to be counted on the night but I am unsure how many.

* In both Longman and Braddon (especially Longman) we will need to see several 2PP counted booths to get an idea of how preference flows are going.  Projections off the early primaries may not be reliable.

Donations to support my efforts on this site (and Twitter) are very welcome but please only give money if you are sure you can afford to.  There is a Paypal box in the sidebar.


  1. 6000 postal votes will be counted in Longman tonight I'm thinking this is unusual.. Government wants quick results . Staff starting around 11 tomorrow for more counting

    1. I've also heard there will be some postal counting in Braddon but in the version I heard indirectly there was doubt over whether those results would be uploaded tonight or tomorrow.

    2. Well, they're online now. 7775 postals counted in Longman. Big Trev has reduced Lamb's lead on primaries by a whole 200 votes. I recall that about 10000 had been issued so not a lot more to come in over the next week.

    3. Oh, and he's gained 350 on the 2PP. A very small hill of beans...


  3. Your site is a little treasure with your insights and deciphering various polls, most of it goes through to the keeper when it comes to methodology. I donate what I can when I can along with Poll Bludger. Getting back to the results it's going to be interesting to see how the L/NP spin this one. Maybe they shot themselves in the foot yet again with such a long campaign.

    1. In the case of Braddon at least that may well have been so, because it allowed time for the impact of the state election to dissipate. Some voters may even have thought "what, two months of this and then if the Coalition win this one they'll call another election right away".

  4. This might be of use in looking at booth results: