Saturday, March 3, 2018

Election Day: Blue Skies With A Fair Chance Of A Poll Fail

Welcome to my on-the-day election coverage. (See my main guide page with links to electorate pages.) I may be adding comments now and then through the day on anything I feel like commenting on.  I expect this to continue up to the release of the Southern Cross exit poll around 6 pm.  I will be live blogging on the Mercury tonight with coverage expected to start around 6:30.  Once that goes live a link will be posted here at the top of the page. After I finish the coverage there will be comments posted here overnight - I am hoping this will include the rollout of postcount threads, but it may not. 

Note for media: I won't be available for any interviews other than the Mercury between 5:30 and 11; I may be available briefly after 11.  Also, tomorrow (because nobody paid me to stay at home) I will be on a field trip to Tooms Lake, will not be available for in-person interviews and may at times be out of mobile phone range.  

My advice to those still to vote is simple: number all of the boxes.  Even if you find you are getting into candidates you cannot stand or have never heard of, putting the lesser evils ahead of the greater evils will make your vote more powerful than if you stop.  Numbering all the boxes will never disadvantage the candidates you prefer or advantage those you most dislike, because your vote only flows on to your less preferred candidates once those you most prefer have all been elected or eliminated.

Normally at this stage I would be fine-tuning predictions based on the latest polling data, but this election I cannot do that, because there isn't any.  The reason for this being ...

No Newspoll

In a very sad day in Australian polling history, no Newspoll has appeared for the Tasmanian state election.  This is the first time this has ever happened.  I believe it's also the first time any state has not had a more or less pre-election Newspoll in the history of the Newspoll brand, with the possible exception

The ability of The Australian to publish a Newspoll today may have been affected by the electoral laws restricting print media coverage on the day of Tasmanian elections.  Last time, this was avoided by publishing results on Thursday and Friday.  It may also have been affected by the near-clash with South Australia (the Australian has released a Newspoll on SA today), or by difficulties applying the current Newspoll methods to Tasmania.  (Newspoll is now partly a phone poll and partly an online pollster).  If I find out more about why there was no Newspoll, I will note it here if I can.

This means that there have only been two independent polls published all year in Tasmania: the Mercury ReachTEL taken late last week, and the EMRS poll taken Saturday to Monday.  (The results of these polls were also consistent with a limited sample of commissioned polls that has been released.) If you've ever watched weather forecasts develop over the course of a week you'll have noticed that they are much more accurate as you get closer to the day than, say, five or six days out.  Polls are not quite as extreme as that, but a similar principle applies and has been proven time and time again in Australian polling: the best way to get the result right is to keep polling up until as close to election day as you can.  When you don't have late data, the chance of the polls - and any forecast that is based on polls - being wrong goes up.

It's still very unlikely to turn around to the extent that, say, Labor gets 43% of the vote and wins outright.  But the chance that some party does a fair bit better or a fair bit worse than the two polls we have seen - already substantial given the past history of both those polls - goes up with the lack of any fresher data.  If there is either a hung parliament or a decisive Liberal victory (14+ seats) then both those outcomes would be within the normal range of polling error in the circumstances.

It might be thought especially likely that there would be a late shift because of yesterday's gun policy revelations.  My own experience is that so-called last-minute game-changers usually aren't.  If late-campaign events caused a big shift in voting intention, then one would expect to find that the swing in prepoll and postal voting differs from the swing in on the day voting.  Yet this doesn't ever seem to happen - a striking case being the removal of Tony Abbott as Prime Minister failing to generate any such signal in the Canning by-election.  However, last-minute incidents can reveal something about the character of the campaign that was already known to voters.  So if the swing against the Liberals is greater than the public polls suggest, it is possible that this will have in fact been the case all along.

In general, last-day matters don't swing election outcomes and while there's a lot of media-bubble and social-media bubble stuff going on about the gun issue, I am not sure it is spilling through to punters generally.

I may add more comments later.


My reader Not-A-Poll has closed and 38.51% of 457 voters said the Liberals would win 12 seats, 26.48% for 13 seats, 16.41% 11 seats, 5.47% 14 seats, 4.81% 10 seats, 4.81% 9 or below, 3.5% 15 or more, for an expected outcome of 12.1 seats.  However as the campaign went on, the share for 13 seats rose from about 13% to its final level.

One consistent pattern with polling is that people asked to predict poll results tend more to predict good results for the side they support.  People voting on this site tend to be left-wing and in 2014 were predicting only 13 seats before a presumed Liberal supporter sabotaged the poll (the Liberals won 15 seats).  It's not very scientific but if the same skew holds up again the Liberals should win at least 13 seats, if not 14 seats.

How Soon Will We Know?

At the last few elections under the 25-seat system, there has usually been one seat, sometimes two seats, that was very close at party level and remained so for several days.  There may be a few others at mild levels of doubt.  There could also be unresolved within-party battles that do not affect the seat tally (last election there was one in Denison and one in Bass).


  1. Labor have signalled they won't support a no confidence motion against a Hodgman minority government. Would that scenario be the first time Labor has supported a Liberal government?

    1. I think that should there be a hung parliament we still need to see exactly what that means. They've said they won't support the Greens' motion of no confidence over disclosure issues, but they could well move their own motion over something else, or support a Greens motion on something else. They could also signal in other ways that the government had lost confidence, eg electing their own Speaker, amending the address in reply, persistently defeating government motions that control the business of the House and so on. It would become a joke and presumably the government would eventually give up.

      In the leadup to the 1996 election Labor indicated that in the event of a hung parliament it would provide supply and confidence for a limited period (interpreted as about a year) to prevent an immediate second election. This turned out to be irrelevant as the Greens agreed to provide supply and confidence to the Liberals anyway.

  2. Many thanks Kevin for all your work putting together the site that I turn to first for information about elections affecting Tasmania. However I must question your frequent advice to number _all_ the boxes under Hare-Clark. If the number of candidates = n, is there any harm in only numbering to n – 1?

    There’s usually one candidate readily identifiable as the most loathed. Wouldn’t putting any number in Most Loathed’s box run a chance of actually helping them over the line in a tight battle for the 5th seat? That chance might perhaps be infinitesimal, but possibly (?) still finite. Whereas leaving Most Loathed’s box as only one blank would mean that there’s no chance your vote could contribute to their election.

    1. Putting a number in Most Loathed's box cannot possibly assist Most Loathed as your vote could only reach them if every other candidate in the election had been elected or excluded. However, if that is the case then the election has finished and there is no contest left for Most Loathed to be assisted in.

      The reason I say one should number n boxes rather than n-1 is this. It sometimes happens in the cut-up that all bar six candidates have been excluded, but the candidates in fifth and sixth have not reached quota and any candidates who have reached quota no longer have any surplus votes to throw. At this point, the candidate who is running sixth has lost and the election is over as a contest. But the rules still require the preferences of the candidate who finished sixth to be thrown for the point of attempting to bring the candidate who finished fifth up to a quota for any potential future recount.

      Now in this situation, suppose Most Loathed finished fifth, and at the time your vote was still active and held by the candidate who finished sixth. If you leave the last box blank, your vote exhausts from the system during that final throw. If you have numbered all the way, your vote now reaches Most Loathed. This does Most Loathed no good at all, since they have already won. However should Most Loathed resign or die during their term, you now have the last laugh on them, because your vote gets included in the votes used for their recount.

      So, by putting Most Loathed last rather than leaving their box blank, you cannot possibly assist them in any way, but you may be able to increase the chance that one of the less obnoxious alternatives replaces them should they retire!

    2. Thanks Kevin, as always you're way ahead of us mug punters. I'll know better next time.